]> granicus.if.org Git - graphviz/commitdiff
adding lib/rxspencer - unmodified source copy
authorJohn Ellson <ellson@research.att.com>
Tue, 12 Apr 2016 19:21:34 +0000 (15:21 -0400)
committerJohn Ellson <ellson@research.att.com>
Tue, 12 Apr 2016 19:21:34 +0000 (15:21 -0400)
35 files changed:
configure.ac
lib/Makefile.am
lib/rxspencer.README [new file with mode: 0644]
lib/rxspencer/COPYRIGHT [new file with mode: 0644]
lib/rxspencer/Makefile.am [new file with mode: 0644]
lib/rxspencer/Makefile.orig [new file with mode: 0644]
lib/rxspencer/README [new file with mode: 0644]
lib/rxspencer/WHATSNEW [new file with mode: 0644]
lib/rxspencer/autogen.sh [new file with mode: 0755]
lib/rxspencer/cclass.h [new file with mode: 0644]
lib/rxspencer/cname.h [new file with mode: 0644]
lib/rxspencer/compile [new file with mode: 0755]
lib/rxspencer/configure.ac [new file with mode: 0644]
lib/rxspencer/debug.c [new file with mode: 0644]
lib/rxspencer/engine.c [new file with mode: 0644]
lib/rxspencer/install-sh [new file with mode: 0755]
lib/rxspencer/ltmain.sh [new file with mode: 0644]
lib/rxspencer/main.c [new file with mode: 0644]
lib/rxspencer/missing [new file with mode: 0755]
lib/rxspencer/mkh [new file with mode: 0644]
lib/rxspencer/mkinstalldirs [new file with mode: 0755]
lib/rxspencer/regcomp.c [new file with mode: 0644]
lib/rxspencer/regerror.c [new file with mode: 0644]
lib/rxspencer/regex2.h [new file with mode: 0644]
lib/rxspencer/regexec.c [new file with mode: 0644]
lib/rxspencer/regfree.c [new file with mode: 0644]
lib/rxspencer/rxspencer.3 [new file with mode: 0644]
lib/rxspencer/rxspencer.7 [new file with mode: 0644]
lib/rxspencer/split.c [new file with mode: 0644]
lib/rxspencer/test-driver [new file with mode: 0755]
lib/rxspencer/test1.sh [new file with mode: 0755]
lib/rxspencer/test2.sh [new file with mode: 0755]
lib/rxspencer/test3.sh [new file with mode: 0755]
lib/rxspencer/tests [new file with mode: 0644]
lib/rxspencer/utils.h [new file with mode: 0644]

index 8a2a8c7a302cb64a24c8b6fc22ba9108fe623873..7afaa0950d63b13aed57ee413a2c4cfa2f1e0728 100644 (file)
@@ -3151,6 +3151,7 @@ AC_CONFIG_FILES(Makefile
        lib/twopigen/Makefile
        lib/patchwork/Makefile
        lib/pack/Makefile
+       lib/rxspencer/Makefile
        lib/ortho/Makefile
        lib/expr/Makefile
        lib/expr/libexpr.pc
index 7e0788bf21b13955631201cb7fd347ffcb35fac9..e20f02f8eb0dfdaee36e4464e3ffd22ecd7e813e 100644 (file)
@@ -1,7 +1,7 @@
 # $Id$ $Revision$
 ## Process this file with automake to produce Makefile.in
 
-SUBDIRS = cdt cgraph pathplan sfio vmalloc ast \
+SUBDIRS = cdt cgraph pathplan sfio vmalloc ast rxspencer \
        vpsc rbtree ortho sparse patchwork expr common \
        pack xdot label gvc ingraphs topfish glcomp mingle edgepaint \
        circogen dotgen fdpgen neatogen twopigen sfdpgen osage gvpr
diff --git a/lib/rxspencer.README b/lib/rxspencer.README
new file mode 100644 (file)
index 0000000..4aa7a72
--- /dev/null
@@ -0,0 +1,6 @@
+The rxspencer directory contains an unmodified snapshot copy
+of Henry Spencer's BSD regex library, as maintained by Gary Houston.
+
+       https://github.com/garyhouston/rxspencer
+
+The snapshot was grabbed by John Ellson (john.ellson@comcast.net) on April 12, 1016
diff --git a/lib/rxspencer/COPYRIGHT b/lib/rxspencer/COPYRIGHT
new file mode 100644 (file)
index 0000000..30c1f7a
--- /dev/null
@@ -0,0 +1,20 @@
+Copyright 1992, 1993, 1994, 1997 Henry Spencer.  All rights reserved.
+This software is not subject to any license of the American Telephone
+and Telegraph Company or of the Regents of the University of California.
+
+Permission is granted to anyone to use this software for any purpose on
+any computer system, and to alter it and redistribute it, subject
+to the following restrictions:
+
+1. The author is not responsible for the consequences of use of this
+   software, no matter how awful, even if they arise from flaws in it.
+
+2. The origin of this software must not be misrepresented, either by
+   explicit claim or by omission.  Since few users ever read sources,
+   credits must appear in the documentation.
+
+3. Altered versions must be plainly marked as such, and must not be
+   misrepresented as being the original software.  Since few users
+   ever read sources, credits must appear in the documentation.
+
+4. This notice may not be removed or altered.
diff --git a/lib/rxspencer/Makefile.am b/lib/rxspencer/Makefile.am
new file mode 100644 (file)
index 0000000..7cd4d55
--- /dev/null
@@ -0,0 +1,36 @@
+AUTOMAKE_OPTIONS = foreign no-dependencies
+lib_LTLIBRARIES = librxspencer.la
+librxspencer_la_LDFLAGS = -version-info 0:2:0 -no-undefined -module
+AM_CPPFLAGS = -DPOSIX_MISTAKE
+librxspencer_la_SOURCES = regcomp.c regerror.c regexec.c regfree.c
+noinst_HEADERS = cclass.h cname.h engine.c regex2.h utils.h
+man_MANS = rxspencer.3 rxspencer.7
+BUILT_SOURCES = main.ih debug.ih regerror.ih engine.ih regcomp.ih regex.h
+test_sources = split.c debug.c main.c 
+re_oo_objs = split.oo debug.oo main.oo regcomp.oo regerror.oo regexec.oo regfree.oo
+re: $(BUILT_SOURCES) $(re_oo_objs)
+       $(LINK) -o re $(re_oo_objs)
+SUFFIXES = .oo .ih
+MKHFLAGS =
+.c.oo:
+       $(COMPILE) -c -DREDEBUG $< -o $@
+.c.ih:
+       sh $(srcdir)/mkh $(MKHFLAGS) -p $< > $@
+REGEXHSRC = $(srcdir)/regex2.h $(srcdir)/regcomp.c $(srcdir)/regerror.c\
+       $(srcdir)/regexec.c $(srcdir)/regfree.c
+regex.h: $(REGEXHSRC) mkh
+       sh $(srcdir)/mkh $(MKHFLAGS) -i _REGEX_H_ $(REGEXHSRC) > regex.h
+install-data-local:
+       $(mkinstalldirs) $(DESTDIR)$(pkgincludedir)
+       $(INSTALL_DATA) regex.h $(DESTDIR)$(pkgincludedir)
+CLEANFILES = re *.oo *.ih regex.h
+TESTS = test1.sh test2.sh test3.sh
+test1.sh: re
+EXTRA_DIST = COPYRIGHT WHATSNEW tests autogen.sh compile mkh\
+       Makefile.orig $(TESTS) $(test_sources) $(man_MANS)
+debug.lo: debug.c debug.ih regex.h utils.h regex2.h
+main.lo: main.c main.ih regex.h 
+regcomp.lo: regcomp.c regcomp.ih cclass.h cname.h regex.h regex2.h utils.h 
+regerror.lo: regerror.c regerror.ih regex.h utils.h
+regexec.lo: regexec.c engine.c engine.ih regex.h regex2.h utils.h
+split.lo: split.c
diff --git a/lib/rxspencer/Makefile.orig b/lib/rxspencer/Makefile.orig
new file mode 100644 (file)
index 0000000..3882b37
--- /dev/null
@@ -0,0 +1,130 @@
+# You probably want to take -DREDEBUG out of CFLAGS, and put something like
+# -O in, *after* testing (-DREDEBUG strengthens testing by enabling a lot of
+# internal assertion checking and some debugging facilities).
+# Put -Dconst= in for a pre-ANSI compiler.
+# Do not take -DPOSIX_MISTAKE out.
+# REGCFLAGS isn't important to you (it's for my use in some special contexts).
+CFLAGS=-I. -DPOSIX_MISTAKE -DREDEBUG $(REGCFLAGS)
+
+# If you have a pre-ANSI compiler, put -o into MKHFLAGS.  If you want
+# the Berkeley __P macro, put -b in.
+MKHFLAGS=
+
+# Flags for linking but not compiling, if any.
+LDFLAGS=
+
+# Extra libraries for linking, if any.
+LIBS=
+
+# Internal stuff, should not need changing.
+OBJPRODN=regcomp.o regexec.o regerror.o regfree.o
+OBJS=$(OBJPRODN) split.o debug.o main.o
+H=cclass.h cname.h regex2.h utils.h
+REGSRC=regcomp.c regerror.c regexec.c regfree.c
+ALLSRC=$(REGSRC) engine.c debug.c main.c split.c
+
+# Stuff that matters only if you're trying to lint the package.
+LINTFLAGS=-I. -Dstatic= -Dconst= -DREDEBUG
+LINTC=regcomp.c regexec.c regerror.c regfree.c debug.c main.c
+JUNKLINT=possible pointer alignment|null effect
+
+# arrangements to build forward-reference header files
+.SUFFIXES:     .ih .h
+.c.ih:
+       sh ./mkh $(MKHFLAGS) -p $< >$@
+
+default:       r
+
+lib:   purge $(OBJPRODN)
+       rm -f libregex.a
+       ar crv libregex.a $(OBJPRODN)
+
+purge:
+       rm -f *.o
+
+# stuff to build regex.h
+REGEXH=regex.h
+REGEXHSRC=regex2.h $(REGSRC)
+$(REGEXH):     $(REGEXHSRC) mkh
+       sh ./mkh $(MKHFLAGS) -i _REGEX_H_ $(REGEXHSRC) >regex.tmp
+       cmp -s regex.tmp regex.h 2>/dev/null || cp regex.tmp regex.h
+       rm -f regex.tmp
+
+# dependencies
+$(OBJPRODN) debug.o:   utils.h regex.h regex2.h
+regcomp.o:     cclass.h cname.h regcomp.ih
+regexec.o:     engine.c engine.ih
+regerror.o:    regerror.ih
+debug.o:       debug.ih
+main.o:        main.ih
+
+# tester
+re:    $(OBJS)
+       $(CC) $(CFLAGS) $(LDFLAGS) $(OBJS) $(LIBS) -o $@
+
+# regression test
+r:     re tests
+       ./re <tests
+       ./re -el <tests
+       ./re -er <tests
+
+# 57 variants, and other stuff, for development use -- not useful to you
+ra:    ./re tests
+       -./re <tests
+       -./re -el <tests
+       -./re -er <tests
+
+rx:    ./re tests
+       ./re -x <tests
+       ./re -x -el <tests
+       ./re -x -er <tests
+
+t:     ./re tests
+       -time ./re <tests
+       -time ./re -cs <tests
+       -time ./re -el <tests
+       -time ./re -cs -el <tests
+
+l:     $(LINTC)
+       lint $(LINTFLAGS) -h $(LINTC) 2>&1 | egrep -v '$(JUNKLINT)' | tee lint
+
+fullprint:
+       ti README WHATSNEW notes todo | list
+       ti *.h | list
+       list *.c
+       list regex.3 regex.7
+
+print:
+       ti README WHATSNEW notes todo | list
+       ti *.h | list
+       list reg*.c engine.c
+
+
+mf.tmp:        Makefile
+       sed '/^REGEXH=/s/=.*/=regex.h/' Makefile | sed '/#DEL$$/d' >$@
+
+DTRH=cclass.h cname.h regex2.h utils.h
+PRE=COPYRIGHT README WHATSNEW
+POST=mkh regex.3 regex.7 tests $(DTRH) $(ALLSRC) fake/*.[ch]
+FILES=$(PRE) Makefile $(POST)
+DTR=$(PRE) Makefile=mf.tmp $(POST)
+dtr:   $(FILES) mf.tmp
+       makedtr $(DTR) >$@
+       rm mf.tmp
+
+cio:   $(FILES)
+       cio $(FILES)
+
+rdf:   $(FILES)
+       rcsdiff -c $(FILES) 2>&1 | p
+
+# various forms of cleanup
+tidy:
+       rm -f junk* core core.* *.core dtr *.tmp lint
+
+clean: tidy
+       rm -f *.o *.s *.ih re libregex.a
+
+# don't do this one unless you know what you're doing
+spotless:      clean
+       rm -f mkh regex.h
diff --git a/lib/rxspencer/README b/lib/rxspencer/README
new file mode 100644 (file)
index 0000000..b3faeda
--- /dev/null
@@ -0,0 +1,69 @@
+This is a copy of Henry Spencer's regular expression library, with
+modifications to the build system:
+
+autoconf/automake/libtool support has been added to allow building a
+shared library on many platforms (in theory).  In other cases, a
+static library will be built.
+
+The names of installed files have been changed to avoid conflict with
+other regex libraries.
+
+Related URLs:
+
+http://www.arglist.com/regex/
+ftp://ftp.zoo.toronto.edu/pub/ for the original regex.shar.
+
+Installation
+============
+
+Ignore the notes in the original README below and try something like:
+
+./configure
+make
+make check
+make install
+
+See the INSTALL file for more details.
+
+Usage
+=====
+
+Library:  -lrxspencer
+Header: #include <rxspencer/regex.h>
+Man Pages: rxspencer(3), rxspencer(7)
+
+Gary Houston,  ghouston@arglist.com
+
+---- original README follows --------
+alpha3.8 release.
+Tue Aug 10 15:51:48 EDT 1999
+henry@spsystems.net  (formerly henry@zoo.toronto.edu)
+
+See WHATSNEW for change listing.
+
+installation notes:
+--------
+Read the comments at the beginning of Makefile before running.
+
+Utils.h contains some things that just might have to be modified on
+some systems, as well as a nested include (ugh) of <assert.h>.
+
+The "fake" directory contains quick-and-dirty fakes for some header
+files and routines that old systems may not have.  Note also that
+-DUSEBCOPY will make utils.h substitute bcopy() for memmove().
+
+After that, "make r" will build regcomp.o, regexec.o, regfree.o,
+and regerror.o (the actual routines), bundle them together into a test
+program, and run regression tests on them.  No output is good output.
+
+"make lib" builds just the .o files for the actual routines (when
+you're happy with testing and have adjusted CFLAGS for production),
+and puts them together into libregex.a.  You can pick up either the
+library or *.o ("make lib" makes sure there are no other .o files left
+around to confuse things).
+
+Main.c, debug.c, split.c are used for regression testing but are not part
+of the RE routines themselves.
+
+Regex.h goes in /usr/include.  All other .h files are internal only.
+--------
diff --git a/lib/rxspencer/WHATSNEW b/lib/rxspencer/WHATSNEW
new file mode 100644 (file)
index 0000000..b7eb2cc
--- /dev/null
@@ -0,0 +1,133 @@
+New in alpha3.8.g7 (Gary Houston): Configuration scripts updated. Make sure
+symlinks aren't created by libtool. 26 Nov 2015.
+
+New in alpha3.8.g6 (Gary Houston): Bug fix for integer overflow in regcomp
+for excessively long pattern strings. CERT Vulnerability Note VU#695940.
+Found by Guido Vranken. Configuration scripts updated. 25 Feb 2015.
+
+New in alpha3.8.g5 (Gary Houston): Configuration scripts updated. Fixed
+"make check". Deleted unused "fake" C library headers.
+
+New in alpha3.8.g4 (Gary Houston): Makefile.am: account for $(DESTDIR)
+in install-data-local, thanks to  Aleksey Cheusov. Configuration scripts
+updated.
+
+New in alpha3.8.g3 (Gary Houston): configuration scripts updated.
+
+New in alpha3.8.g2 (Gary Houston): configure.in, Makefile.am: updated
+for newer versions of autoconf/automake/libtool, with miscellaneous
+changes.  mkh: don't set PATH, in case utilities are in non-standard
+locations (i.e., cygwin).
+
+New in alpha3.8.g1 (Gary Houston): Use autoconf/automake/libtool to
+build, test and install the library.  Renamed the installed files to
+avoid conflicts with other regex libraries.
+
+New in alpha3.8:  Bug fix for signed/unsigned mixup, found and fixed
+by the FreeBSD folks.
+
+New in alpha3.7:  A bit of cleanup aimed at maximizing portability,
+possibly at slight cost in efficiency.  "ul" suffixes and "unsigned long"
+no longer appear, in particular.
+
+New in alpha3.6:  A couple more portability glitches fixed.
+
+New in alpha3.5:  Active development of this code has been stopped --
+I'm working on a complete reimplementation -- but folks have found some
+minor portability glitches and the like, hence this release to fix them.
+One penalty:  slightly reduced compatibility with old compilers, because
+the ANSI C `unsigned long' type and `ul' constant suffix are used in a
+few places (I could avoid this but it would be considerably more work).
+
+New in alpha3.4:  The complex bug alluded to below has been fixed (in a
+slightly kludgey temporary way that may hurt efficiency a bit; this is
+another "get it out the door for 4.4" release).  The tests at the end of
+the tests file have accordingly been uncommented.  The primary sign of
+the bug was that something like a?b matching ab matched b rather than ab.
+(The bug was essentially specific to this exact situation, else it would
+have shown up earlier.)
+
+New in alpha3.3:  The definition of word boundaries has been altered
+slightly, to more closely match the usual programming notion that "_"
+is an alphabetic.  Stuff used for pre-ANSI systems is now in a subdir,
+and the makefile no longer alludes to it in mysterious ways.  The
+makefile has generally been cleaned up some.  Fixes have been made
+(again!) so that the regression test will run without -DREDEBUG, at
+the cost of weaker checking.  A workaround for a bug in some folks'
+<assert.h> has been added.  And some more things have been added to
+tests, including a couple right at the end which are commented out
+because the code currently flunks them (complex bug; fix coming).
+Plus the usual minor cleanup.
+
+New in alpha3.2:  Assorted bits of cleanup and portability improvement
+(the development base is now a BSDI system using GCC instead of an ancient
+Sun system, and the newer compiler exposed some glitches).  Fix for a
+serious bug that affected REs using many [] (including REG_ICASE REs
+because of the way they are implemented), *sometimes*, depending on
+memory-allocation patterns.  The header-file prototypes no longer name
+the parameters, avoiding possible name conflicts.  The possibility that
+some clot has defined CHAR_MIN as (say) `-128' instead of `(-128)' is
+now handled gracefully.  "uchar" is no longer used as an internal type
+name (too many people have the same idea).  Still the same old lousy
+performance, alas.
+
+New in alpha3.1:  Basically nothing, this release is just a bookkeeping
+convenience.  Stay tuned.
+
+New in alpha3.0:  Performance is no better, alas, but some fixes have been
+made and some functionality has been added.  (This is basically the "get
+it out the door in time for 4.4" release.)  One bug fix:  regfree() didn't
+free the main internal structure (how embarrassing).  It is now possible
+to put NULs in either the RE or the target string, using (resp.) a new
+REG_PEND flag and the old REG_STARTEND flag.  The REG_NOSPEC flag to
+regcomp() makes all characters ordinary, so you can match a literal
+string easily (this will become more useful when performance improves!).
+There are now primitives to match beginnings and ends of words, although
+the syntax is disgusting and so is the implementation.  The REG_ATOI
+debugging interface has changed a bit.  And there has been considerable
+internal cleanup of various kinds.
+
+New in alpha2.3:  Split change list out of README, and moved flags notes
+into Makefile.  Macro-ized the name of regex(7) in regex(3), since it has
+to change for 4.4BSD.  Cleanup work in engine.c, and some new regression
+tests to catch tricky cases thereof.
+
+New in alpha2.2:  Out-of-date manpages updated.  Regerror() acquires two
+small extensions -- REG_ITOA and REG_ATOI -- which avoid debugging kludges
+in my own test program and might be useful to others for similar purposes.
+The regression test will now compile (and run) without REDEBUG.  The
+BRE \$ bug is fixed.  Most uses of "uchar" are gone; it's all chars now.
+Char/uchar parameters are now written int/unsigned, to avoid possible
+portability problems with unpromoted parameters.  Some unsigned casts have
+been introduced to minimize portability problems with shifting into sign
+bits.
+
+New in alpha2.1:  Lots of little stuff, cleanup and fixes.  The one big
+thing is that regex.h is now generated, using mkh, rather than being
+supplied in the distribution; due to circularities in dependencies,
+you have to build regex.h explicitly by "make h".  The two known bugs
+have been fixed (and the regression test now checks for them), as has a
+problem with assertions not being suppressed in the absence of REDEBUG.
+No performance work yet.
+
+New in alpha2:  Backslash-anything is an ordinary character, not an
+error (except, of course, for the handful of backslashed metacharacters
+in BREs), which should reduce script breakage.  The regression test
+checks *where* null strings are supposed to match, and has generally
+been tightened up somewhat.  Small bug fixes in parameter passing (not
+harmful, but technically errors) and some other areas.  Debugging
+invoked by defining REDEBUG rather than not defining NDEBUG.
+
+New in alpha+3:  full prototyping for internal routines, using a little
+helper program, mkh, which extracts prototypes given in stylized comments.
+More minor cleanup.  Buglet fix:  it's CHAR_BIT, not CHAR_BITS.  Simple
+pre-screening of input when a literal string is known to be part of the
+RE; this does wonders for performance.
+
+New in alpha+2:  minor bits of cleanup.  Notably, the number "32" for the
+word width isn't hardwired into regexec.c any more, the public header
+file prototypes the functions if __STDC__ is defined, and some small typos
+in the manpages have been fixed.
+
+New in alpha+1:  improvements to the manual pages, and an important
+extension, the REG_STARTEND option to regexec().
diff --git a/lib/rxspencer/autogen.sh b/lib/rxspencer/autogen.sh
new file mode 100755 (executable)
index 0000000..81835c1
--- /dev/null
@@ -0,0 +1,5 @@
+#!/bin/sh
+libtoolize -c --automake --force
+aclocal
+autoconf -Wall --force
+automake -Wall --add-missing --force-missing --copy
diff --git a/lib/rxspencer/cclass.h b/lib/rxspencer/cclass.h
new file mode 100644 (file)
index 0000000..0c29302
--- /dev/null
@@ -0,0 +1,31 @@
+/* character-class table */
+static struct cclass {
+       char *name;
+       char *chars;
+       char *multis;
+} cclasses[] = {
+       "alnum",        "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz\
+0123456789",                           "",
+       "alpha",        "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz",
+                                       "",
+       "blank",        " \t",          "",
+       "cntrl",        "\007\b\t\n\v\f\r\1\2\3\4\5\6\16\17\20\21\22\23\24\
+\25\26\27\30\31\32\33\34\35\36\37\177",        "",
+       "digit",        "0123456789",   "",
+       "graph",        "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz\
+0123456789!\"#$%&'()*+,-./:;<=>?@[\\]^_`{|}~",
+                                       "",
+       "lower",        "abcdefghijklmnopqrstuvwxyz",
+                                       "",
+       "print",        "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz\
+0123456789!\"#$%&'()*+,-./:;<=>?@[\\]^_`{|}~ ",
+                                       "",
+       "punct",        "!\"#$%&'()*+,-./:;<=>?@[\\]^_`{|}~",
+                                       "",
+       "space",        "\t\n\v\f\r ",  "",
+       "upper",        "ABCDEFGHIJKLMNOPQRSTUVWXYZ",
+                                       "",
+       "xdigit",       "0123456789ABCDEFabcdef",
+                                       "",
+       NULL,           0,              ""
+};
diff --git a/lib/rxspencer/cname.h b/lib/rxspencer/cname.h
new file mode 100644 (file)
index 0000000..02e86e9
--- /dev/null
@@ -0,0 +1,102 @@
+/* character-name table */
+static struct cname {
+       char *name;
+       char code;
+} cnames[] = {
+       "NUL",  '\0',
+       "SOH",  '\001',
+       "STX",  '\002',
+       "ETX",  '\003',
+       "EOT",  '\004',
+       "ENQ",  '\005',
+       "ACK",  '\006',
+       "BEL",  '\007',
+       "alert",        '\007',
+       "BS",           '\010',
+       "backspace",    '\b',
+       "HT",           '\011',
+       "tab",          '\t',
+       "LF",           '\012',
+       "newline",      '\n',
+       "VT",           '\013',
+       "vertical-tab", '\v',
+       "FF",           '\014',
+       "form-feed",    '\f',
+       "CR",           '\015',
+       "carriage-return",      '\r',
+       "SO",   '\016',
+       "SI",   '\017',
+       "DLE",  '\020',
+       "DC1",  '\021',
+       "DC2",  '\022',
+       "DC3",  '\023',
+       "DC4",  '\024',
+       "NAK",  '\025',
+       "SYN",  '\026',
+       "ETB",  '\027',
+       "CAN",  '\030',
+       "EM",   '\031',
+       "SUB",  '\032',
+       "ESC",  '\033',
+       "IS4",  '\034',
+       "FS",   '\034',
+       "IS3",  '\035',
+       "GS",   '\035',
+       "IS2",  '\036',
+       "RS",   '\036',
+       "IS1",  '\037',
+       "US",   '\037',
+       "space",                ' ',
+       "exclamation-mark",     '!',
+       "quotation-mark",       '"',
+       "number-sign",          '#',
+       "dollar-sign",          '$',
+       "percent-sign",         '%',
+       "ampersand",            '&',
+       "apostrophe",           '\'',
+       "left-parenthesis",     '(',
+       "right-parenthesis",    ')',
+       "asterisk",     '*',
+       "plus-sign",    '+',
+       "comma",        ',',
+       "hyphen",       '-',
+       "hyphen-minus", '-',
+       "period",       '.',
+       "full-stop",    '.',
+       "slash",        '/',
+       "solidus",      '/',
+       "zero",         '0',
+       "one",          '1',
+       "two",          '2',
+       "three",        '3',
+       "four",         '4',
+       "five",         '5',
+       "six",          '6',
+       "seven",        '7',
+       "eight",        '8',
+       "nine",         '9',
+       "colon",        ':',
+       "semicolon",    ';',
+       "less-than-sign",       '<',
+       "equals-sign",          '=',
+       "greater-than-sign",    '>',
+       "question-mark",        '?',
+       "commercial-at",        '@',
+       "left-square-bracket",  '[',
+       "backslash",            '\\',
+       "reverse-solidus",      '\\',
+       "right-square-bracket", ']',
+       "circumflex",           '^',
+       "circumflex-accent",    '^',
+       "underscore",           '_',
+       "low-line",             '_',
+       "grave-accent",         '`',
+       "left-brace",           '{',
+       "left-curly-bracket",   '{',
+       "vertical-line",        '|',
+       "right-brace",          '}',
+       "right-curly-bracket",  '}',
+       "tilde",                '~',
+       "DEL",  '\177',
+       NULL,   0,
+};
diff --git a/lib/rxspencer/compile b/lib/rxspencer/compile
new file mode 100755 (executable)
index 0000000..a85b723
--- /dev/null
@@ -0,0 +1,347 @@
+#! /bin/sh
+# Wrapper for compilers which do not understand '-c -o'.
+
+scriptversion=2012-10-14.11; # UTC
+
+# Copyright (C) 1999-2014 Free Software Foundation, Inc.
+# Written by Tom Tromey <tromey@cygnus.com>.
+#
+# This program is free software; you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation; either version 2, or (at your option)
+# any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+# As a special exception to the GNU General Public License, if you
+# distribute this file as part of a program that contains a
+# configuration script generated by Autoconf, you may include it under
+# the same distribution terms that you use for the rest of that program.
+
+# This file is maintained in Automake, please report
+# bugs to <bug-automake@gnu.org> or send patches to
+# <automake-patches@gnu.org>.
+
+nl='
+'
+
+# We need space, tab and new line, in precisely that order.  Quoting is
+# there to prevent tools from complaining about whitespace usage.
+IFS=" ""       $nl"
+
+file_conv=
+
+# func_file_conv build_file lazy
+# Convert a $build file to $host form and store it in $file
+# Currently only supports Windows hosts. If the determined conversion
+# type is listed in (the comma separated) LAZY, no conversion will
+# take place.
+func_file_conv ()
+{
+  file=$1
+  case $file in
+    / | /[!/]*) # absolute file, and not a UNC file
+      if test -z "$file_conv"; then
+       # lazily determine how to convert abs files
+       case `uname -s` in
+         MINGW*)
+           file_conv=mingw
+           ;;
+         CYGWIN*)
+           file_conv=cygwin
+           ;;
+         *)
+           file_conv=wine
+           ;;
+       esac
+      fi
+      case $file_conv/,$2, in
+       *,$file_conv,*)
+         ;;
+       mingw/*)
+         file=`cmd //C echo "$file " | sed -e 's/"\(.*\) " *$/\1/'`
+         ;;
+       cygwin/*)
+         file=`cygpath -m "$file" || echo "$file"`
+         ;;
+       wine/*)
+         file=`winepath -w "$file" || echo "$file"`
+         ;;
+      esac
+      ;;
+  esac
+}
+
+# func_cl_dashL linkdir
+# Make cl look for libraries in LINKDIR
+func_cl_dashL ()
+{
+  func_file_conv "$1"
+  if test -z "$lib_path"; then
+    lib_path=$file
+  else
+    lib_path="$lib_path;$file"
+  fi
+  linker_opts="$linker_opts -LIBPATH:$file"
+}
+
+# func_cl_dashl library
+# Do a library search-path lookup for cl
+func_cl_dashl ()
+{
+  lib=$1
+  found=no
+  save_IFS=$IFS
+  IFS=';'
+  for dir in $lib_path $LIB
+  do
+    IFS=$save_IFS
+    if $shared && test -f "$dir/$lib.dll.lib"; then
+      found=yes
+      lib=$dir/$lib.dll.lib
+      break
+    fi
+    if test -f "$dir/$lib.lib"; then
+      found=yes
+      lib=$dir/$lib.lib
+      break
+    fi
+    if test -f "$dir/lib$lib.a"; then
+      found=yes
+      lib=$dir/lib$lib.a
+      break
+    fi
+  done
+  IFS=$save_IFS
+
+  if test "$found" != yes; then
+    lib=$lib.lib
+  fi
+}
+
+# func_cl_wrapper cl arg...
+# Adjust compile command to suit cl
+func_cl_wrapper ()
+{
+  # Assume a capable shell
+  lib_path=
+  shared=:
+  linker_opts=
+  for arg
+  do
+    if test -n "$eat"; then
+      eat=
+    else
+      case $1 in
+       -o)
+         # configure might choose to run compile as 'compile cc -o foo foo.c'.
+         eat=1
+         case $2 in
+           *.o | *.[oO][bB][jJ])
+             func_file_conv "$2"
+             set x "$@" -Fo"$file"
+             shift
+             ;;
+           *)
+             func_file_conv "$2"
+             set x "$@" -Fe"$file"
+             shift
+             ;;
+         esac
+         ;;
+       -I)
+         eat=1
+         func_file_conv "$2" mingw
+         set x "$@" -I"$file"
+         shift
+         ;;
+       -I*)
+         func_file_conv "${1#-I}" mingw
+         set x "$@" -I"$file"
+         shift
+         ;;
+       -l)
+         eat=1
+         func_cl_dashl "$2"
+         set x "$@" "$lib"
+         shift
+         ;;
+       -l*)
+         func_cl_dashl "${1#-l}"
+         set x "$@" "$lib"
+         shift
+         ;;
+       -L)
+         eat=1
+         func_cl_dashL "$2"
+         ;;
+       -L*)
+         func_cl_dashL "${1#-L}"
+         ;;
+       -static)
+         shared=false
+         ;;
+       -Wl,*)
+         arg=${1#-Wl,}
+         save_ifs="$IFS"; IFS=','
+         for flag in $arg; do
+           IFS="$save_ifs"
+           linker_opts="$linker_opts $flag"
+         done
+         IFS="$save_ifs"
+         ;;
+       -Xlinker)
+         eat=1
+         linker_opts="$linker_opts $2"
+         ;;
+       -*)
+         set x "$@" "$1"
+         shift
+         ;;
+       *.cc | *.CC | *.cxx | *.CXX | *.[cC]++)
+         func_file_conv "$1"
+         set x "$@" -Tp"$file"
+         shift
+         ;;
+       *.c | *.cpp | *.CPP | *.lib | *.LIB | *.Lib | *.OBJ | *.obj | *.[oO])
+         func_file_conv "$1" mingw
+         set x "$@" "$file"
+         shift
+         ;;
+       *)
+         set x "$@" "$1"
+         shift
+         ;;
+      esac
+    fi
+    shift
+  done
+  if test -n "$linker_opts"; then
+    linker_opts="-link$linker_opts"
+  fi
+  exec "$@" $linker_opts
+  exit 1
+}
+
+eat=
+
+case $1 in
+  '')
+     echo "$0: No command.  Try '$0 --help' for more information." 1>&2
+     exit 1;
+     ;;
+  -h | --h*)
+    cat <<\EOF
+Usage: compile [--help] [--version] PROGRAM [ARGS]
+
+Wrapper for compilers which do not understand '-c -o'.
+Remove '-o dest.o' from ARGS, run PROGRAM with the remaining
+arguments, and rename the output as expected.
+
+If you are trying to build a whole package this is not the
+right script to run: please start by reading the file 'INSTALL'.
+
+Report bugs to <bug-automake@gnu.org>.
+EOF
+    exit $?
+    ;;
+  -v | --v*)
+    echo "compile $scriptversion"
+    exit $?
+    ;;
+  cl | *[/\\]cl | cl.exe | *[/\\]cl.exe )
+    func_cl_wrapper "$@"      # Doesn't return...
+    ;;
+esac
+
+ofile=
+cfile=
+
+for arg
+do
+  if test -n "$eat"; then
+    eat=
+  else
+    case $1 in
+      -o)
+       # configure might choose to run compile as 'compile cc -o foo foo.c'.
+       # So we strip '-o arg' only if arg is an object.
+       eat=1
+       case $2 in
+         *.o | *.obj)
+           ofile=$2
+           ;;
+         *)
+           set x "$@" -o "$2"
+           shift
+           ;;
+       esac
+       ;;
+      *.c)
+       cfile=$1
+       set x "$@" "$1"
+       shift
+       ;;
+      *)
+       set x "$@" "$1"
+       shift
+       ;;
+    esac
+  fi
+  shift
+done
+
+if test -z "$ofile" || test -z "$cfile"; then
+  # If no '-o' option was seen then we might have been invoked from a
+  # pattern rule where we don't need one.  That is ok -- this is a
+  # normal compilation that the losing compiler can handle.  If no
+  # '.c' file was seen then we are probably linking.  That is also
+  # ok.
+  exec "$@"
+fi
+
+# Name of file we expect compiler to create.
+cofile=`echo "$cfile" | sed 's|^.*[\\/]||; s|^[a-zA-Z]:||; s/\.c$/.o/'`
+
+# Create the lock directory.
+# Note: use '[/\\:.-]' here to ensure that we don't use the same name
+# that we are using for the .o file.  Also, base the name on the expected
+# object file name, since that is what matters with a parallel build.
+lockdir=`echo "$cofile" | sed -e 's|[/\\:.-]|_|g'`.d
+while true; do
+  if mkdir "$lockdir" >/dev/null 2>&1; then
+    break
+  fi
+  sleep 1
+done
+# FIXME: race condition here if user kills between mkdir and trap.
+trap "rmdir '$lockdir'; exit 1" 1 2 15
+
+# Run the compile.
+"$@"
+ret=$?
+
+if test -f "$cofile"; then
+  test "$cofile" = "$ofile" || mv "$cofile" "$ofile"
+elif test -f "${cofile}bj"; then
+  test "${cofile}bj" = "$ofile" || mv "${cofile}bj" "$ofile"
+fi
+
+rmdir "$lockdir"
+exit $ret
+
+# Local Variables:
+# mode: shell-script
+# sh-indentation: 2
+# eval: (add-hook 'write-file-hooks 'time-stamp)
+# time-stamp-start: "scriptversion="
+# time-stamp-format: "%:y-%02m-%02d.%02H"
+# time-stamp-time-zone: "UTC"
+# time-stamp-end: "; # UTC"
+# End:
diff --git a/lib/rxspencer/configure.ac b/lib/rxspencer/configure.ac
new file mode 100644 (file)
index 0000000..b55008e
--- /dev/null
@@ -0,0 +1,6 @@
+AC_INIT([rxspencer],[alpha3.8.g5])
+AC_CONFIG_SRCDIR(regcomp.c)
+AM_INIT_AUTOMAKE
+LT_INIT
+AC_CONFIG_FILES(Makefile)
+AC_OUTPUT
diff --git a/lib/rxspencer/debug.c b/lib/rxspencer/debug.c
new file mode 100644 (file)
index 0000000..99ce7da
--- /dev/null
@@ -0,0 +1,242 @@
+#include <stdio.h>
+#include <string.h>
+#include <ctype.h>
+#include <limits.h>
+#include <stdlib.h>
+#include <sys/types.h>
+#include <regex.h>
+
+#include "utils.h"
+#include "regex2.h"
+#include "debug.ih"
+
+/*
+ - regprint - print a regexp for debugging
+ == void regprint(regex_t *r, FILE *d);
+ */
+void
+regprint(r, d)
+regex_t *r;
+FILE *d;
+{
+       register struct re_guts *g = r->re_g;
+       register int i;
+       register int c;
+       register int last;
+       int nincat[NC];
+
+       fprintf(d, "%ld states, %d categories", (long)g->nstates,
+                                                       g->ncategories);
+       fprintf(d, ", first %ld last %ld", (long)g->firststate,
+                                               (long)g->laststate);
+       if (g->iflags&USEBOL)
+               fprintf(d, ", USEBOL");
+       if (g->iflags&USEEOL)
+               fprintf(d, ", USEEOL");
+       if (g->iflags&BAD)
+               fprintf(d, ", BAD");
+       if (g->nsub > 0)
+               fprintf(d, ", nsub=%ld", (long)g->nsub);
+       if (g->must != NULL)
+               fprintf(d, ", must(%ld) `%*s'", (long)g->mlen, (int)g->mlen,
+                                                               g->must);
+       if (g->backrefs)
+               fprintf(d, ", backrefs");
+       if (g->nplus > 0)
+               fprintf(d, ", nplus %ld", (long)g->nplus);
+       fprintf(d, "\n");
+       s_print(g, d);
+       for (i = 0; i < g->ncategories; i++) {
+               nincat[i] = 0;
+               for (c = CHAR_MIN; c <= CHAR_MAX; c++)
+                       if (g->categories[c] == i)
+                               nincat[i]++;
+       }
+       fprintf(d, "cc0#%d", nincat[0]);
+       for (i = 1; i < g->ncategories; i++)
+               if (nincat[i] == 1) {
+                       for (c = CHAR_MIN; c <= CHAR_MAX; c++)
+                               if (g->categories[c] == i)
+                                       break;
+                       fprintf(d, ", %d=%s", i, regchar(c));
+               }
+       fprintf(d, "\n");
+       for (i = 1; i < g->ncategories; i++)
+               if (nincat[i] != 1) {
+                       fprintf(d, "cc%d\t", i);
+                       last = -1;
+                       for (c = CHAR_MIN; c <= CHAR_MAX+1; c++)        /* +1 does flush */
+                               if (c <= CHAR_MAX && g->categories[c] == i) {
+                                       if (last < 0) {
+                                               fprintf(d, "%s", regchar(c));
+                                               last = c;
+                                       }
+                               } else {
+                                       if (last >= 0) {
+                                               if (last != c-1)
+                                                       fprintf(d, "-%s",
+                                                               regchar(c-1));
+                                               last = -1;
+                                       }
+                               }
+                       fprintf(d, "\n");
+               }
+}
+
+/*
+ - s_print - print the strip for debugging
+ == static void s_print(register struct re_guts *g, FILE *d);
+ */
+static void
+s_print(g, d)
+register struct re_guts *g;
+FILE *d;
+{
+       register sop *s;
+       register cset *cs;
+       register int i;
+       register int done = 0;
+       register sop opnd;
+       register int col = 0;
+       register int last;
+       register sopno offset = 2;
+#      define  GAP()   {       if (offset % 5 == 0) { \
+                                       if (col > 40) { \
+                                               fprintf(d, "\n\t"); \
+                                               col = 0; \
+                                       } else { \
+                                               fprintf(d, " "); \
+                                               col++; \
+                                       } \
+                               } else \
+                                       col++; \
+                               offset++; \
+                       }
+
+       if (OP(g->strip[0]) != OEND)
+               fprintf(d, "missing initial OEND!\n");
+       for (s = &g->strip[1]; !done; s++) {
+               opnd = OPND(*s);
+               switch (OP(*s)) {
+               case OEND:
+                       fprintf(d, "\n");
+                       done = 1;
+                       break;
+               case OCHAR:
+                       if (strchr("\\|()^$.[+*?{}!<> ", (char)opnd) != NULL)
+                               fprintf(d, "\\%c", (char)opnd);
+                       else
+                               fprintf(d, "%s", regchar((char)opnd));
+                       break;
+               case OBOL:
+                       fprintf(d, "^");
+                       break;
+               case OEOL:
+                       fprintf(d, "$");
+                       break;
+               case OBOW:
+                       fprintf(d, "\\{");
+                       break;
+               case OEOW:
+                       fprintf(d, "\\}");
+                       break;
+               case OANY:
+                       fprintf(d, ".");
+                       break;
+               case OANYOF:
+                       fprintf(d, "[(%ld)", (long)opnd);
+                       cs = &g->sets[opnd];
+                       last = -1;
+                       for (i = 0; i < g->csetsize+1; i++)     /* +1 flushes */
+                               if (CHIN(cs, i) && i < g->csetsize) {
+                                       if (last < 0) {
+                                               fprintf(d, "%s", regchar(i));
+                                               last = i;
+                                       }
+                               } else {
+                                       if (last >= 0) {
+                                               if (last != i-1)
+                                                       fprintf(d, "-%s",
+                                                               regchar(i-1));
+                                               last = -1;
+                                       }
+                               }
+                       fprintf(d, "]");
+                       break;
+               case OBACK_:
+                       fprintf(d, "(\\<%ld>", (long)opnd);
+                       break;
+               case O_BACK:
+                       fprintf(d, "<%ld>\\)", (long)opnd);
+                       break;
+               case OPLUS_:
+                       fprintf(d, "(+");
+                       if (OP(*(s+opnd)) != O_PLUS)
+                               fprintf(d, "<%ld>", (long)opnd);
+                       break;
+               case O_PLUS:
+                       if (OP(*(s-opnd)) != OPLUS_)
+                               fprintf(d, "<%ld>", (long)opnd);
+                       fprintf(d, "+)");
+                       break;
+               case OQUEST_:
+                       fprintf(d, "(?");
+                       if (OP(*(s+opnd)) != O_QUEST)
+                               fprintf(d, "<%ld>", (long)opnd);
+                       break;
+               case O_QUEST:
+                       if (OP(*(s-opnd)) != OQUEST_)
+                               fprintf(d, "<%ld>", (long)opnd);
+                       fprintf(d, "?)");
+                       break;
+               case OLPAREN:
+                       fprintf(d, "((<%ld>", (long)opnd);
+                       break;
+               case ORPAREN:
+                       fprintf(d, "<%ld>))", (long)opnd);
+                       break;
+               case OCH_:
+                       fprintf(d, "<");
+                       if (OP(*(s+opnd)) != OOR2)
+                               fprintf(d, "<%ld>", (long)opnd);
+                       break;
+               case OOR1:
+                       if (OP(*(s-opnd)) != OOR1 && OP(*(s-opnd)) != OCH_)
+                               fprintf(d, "<%ld>", (long)opnd);
+                       fprintf(d, "|");
+                       break;
+               case OOR2:
+                       fprintf(d, "|");
+                       if (OP(*(s+opnd)) != OOR2 && OP(*(s+opnd)) != O_CH)
+                               fprintf(d, "<%ld>", (long)opnd);
+                       break;
+               case O_CH:
+                       if (OP(*(s-opnd)) != OOR1)
+                               fprintf(d, "<%ld>", (long)opnd);
+                       fprintf(d, ">");
+                       break;
+               default:
+                       fprintf(d, "!%d(%d)!", OP(*s), opnd);
+                       break;
+               }
+               if (!done)
+                       GAP();
+       }
+}
+
+/*
+ - regchar - make a character printable
+ == static char *regchar(int ch);
+ */
+static char *                  /* -> representation */
+regchar(ch)
+int ch;
+{
+       static char buf[10];
+
+       if (isprint(ch) || ch == ' ')
+               sprintf(buf, "%c", ch);
+       else
+               sprintf(buf, "\\%o", ch);
+       return(buf);
+}
diff --git a/lib/rxspencer/engine.c b/lib/rxspencer/engine.c
new file mode 100644 (file)
index 0000000..919fe3f
--- /dev/null
@@ -0,0 +1,1019 @@
+/*
+ * The matching engine and friends.  This file is #included by regexec.c
+ * after suitable #defines of a variety of macros used herein, so that
+ * different state representations can be used without duplicating masses
+ * of code.
+ */
+
+#ifdef SNAMES
+#define        matcher smatcher
+#define        fast    sfast
+#define        slow    sslow
+#define        dissect sdissect
+#define        backref sbackref
+#define        step    sstep
+#define        print   sprint
+#define        at      sat
+#define        match   smat
+#endif
+#ifdef LNAMES
+#define        matcher lmatcher
+#define        fast    lfast
+#define        slow    lslow
+#define        dissect ldissect
+#define        backref lbackref
+#define        step    lstep
+#define        print   lprint
+#define        at      lat
+#define        match   lmat
+#endif
+
+/* another structure passed up and down to avoid zillions of parameters */
+struct match {
+       struct re_guts *g;
+       int eflags;
+       regmatch_t *pmatch;     /* [nsub+1] (0 element unused) */
+       char *offp;             /* offsets work from here */
+       char *beginp;           /* start of string -- virtual NUL precedes */
+       char *endp;             /* end of string -- virtual NUL here */
+       char *coldp;            /* can be no match starting before here */
+       char **lastpos;         /* [nplus+1] */
+       STATEVARS;
+       states st;              /* current states */
+       states fresh;           /* states for a fresh start */
+       states tmp;             /* temporary */
+       states empty;           /* empty set of states */
+};
+
+#include "engine.ih"
+
+#ifdef REDEBUG
+#define        SP(t, s, c)     print(m, t, s, c, stdout)
+#define        AT(t, p1, p2, s1, s2)   at(m, t, p1, p2, s1, s2)
+#define        NOTE(str)       { if (m->eflags&REG_TRACE) printf("=%s\n", (str)); }
+#else
+#define        SP(t, s, c)     /* nothing */
+#define        AT(t, p1, p2, s1, s2)   /* nothing */
+#define        NOTE(s) /* nothing */
+#endif
+
+/*
+ - matcher - the actual matching engine
+ == static int matcher(register struct re_guts *g, char *string, \
+ ==    size_t nmatch, regmatch_t pmatch[], int eflags);
+ */
+static int                     /* 0 success, REG_NOMATCH failure */
+matcher(g, string, nmatch, pmatch, eflags)
+register struct re_guts *g;
+char *string;
+size_t nmatch;
+regmatch_t pmatch[];
+int eflags;
+{
+       register char *endp;
+       register int i;
+       struct match mv;
+       register struct match *m = &mv;
+       register char *dp;
+       const register sopno gf = g->firststate+1;      /* +1 for OEND */
+       const register sopno gl = g->laststate;
+       char *start;
+       char *stop;
+
+       /* simplify the situation where possible */
+       if (g->cflags&REG_NOSUB)
+               nmatch = 0;
+       if (eflags&REG_STARTEND) {
+               start = string + pmatch[0].rm_so;
+               stop = string + pmatch[0].rm_eo;
+       } else {
+               start = string;
+               stop = start + strlen(start);
+       }
+       if (stop < start)
+               return(REG_INVARG);
+
+       /* prescreening; this does wonders for this rather slow code */
+       if (g->must != NULL) {
+               for (dp = start; dp < stop; dp++)
+                       if (*dp == g->must[0] && stop - dp >= g->mlen &&
+                               memcmp(dp, g->must, (size_t)g->mlen) == 0)
+                               break;
+               if (dp == stop)         /* we didn't find g->must */
+                       return(REG_NOMATCH);
+       }
+
+       /* match struct setup */
+       m->g = g;
+       m->eflags = eflags;
+       m->pmatch = NULL;
+       m->lastpos = NULL;
+       m->offp = string;
+       m->beginp = start;
+       m->endp = stop;
+       STATESETUP(m, 4);
+       SETUP(m->st);
+       SETUP(m->fresh);
+       SETUP(m->tmp);
+       SETUP(m->empty);
+       CLEAR(m->empty);
+
+       /* this loop does only one repetition except for backrefs */
+       for (;;) {
+               endp = fast(m, start, stop, gf, gl);
+               if (endp == NULL) {             /* a miss */
+                       STATETEARDOWN(m);
+                       return(REG_NOMATCH);
+               }
+               if (nmatch == 0 && !g->backrefs)
+                       break;          /* no further info needed */
+
+               /* where? */
+               assert(m->coldp != NULL);
+               for (;;) {
+                       NOTE("finding start");
+                       endp = slow(m, m->coldp, stop, gf, gl);
+                       if (endp != NULL)
+                               break;
+                       assert(m->coldp < m->endp);
+                       m->coldp++;
+               }
+               if (nmatch == 1 && !g->backrefs)
+                       break;          /* no further info needed */
+
+               /* oh my, he wants the subexpressions... */
+               if (m->pmatch == NULL)
+                       m->pmatch = (regmatch_t *)malloc((m->g->nsub + 1) *
+                                                       sizeof(regmatch_t));
+               if (m->pmatch == NULL) {
+                       STATETEARDOWN(m);
+                       return(REG_ESPACE);
+               }
+               for (i = 1; i <= m->g->nsub; i++)
+                       m->pmatch[i].rm_so = m->pmatch[i].rm_eo = -1;
+               if (!g->backrefs && !(m->eflags&REG_BACKR)) {
+                       NOTE("dissecting");
+                       dp = dissect(m, m->coldp, endp, gf, gl);
+               } else {
+                       if (g->nplus > 0 && m->lastpos == NULL)
+                               m->lastpos = (char **)malloc((g->nplus+1) *
+                                                       sizeof(char *));
+                       if (g->nplus > 0 && m->lastpos == NULL) {
+                               free(m->pmatch);
+                               STATETEARDOWN(m);
+                               return(REG_ESPACE);
+                       }
+                       NOTE("backref dissect");
+                       dp = backref(m, m->coldp, endp, gf, gl, (sopno)0);
+               }
+               if (dp != NULL)
+                       break;
+
+               /* uh-oh... we couldn't find a subexpression-level match */
+               assert(g->backrefs);    /* must be back references doing it */
+               assert(g->nplus == 0 || m->lastpos != NULL);
+               for (;;) {
+                       if (dp != NULL || endp <= m->coldp)
+                               break;          /* defeat */
+                       NOTE("backoff");
+                       endp = slow(m, m->coldp, endp-1, gf, gl);
+                       if (endp == NULL)
+                               break;          /* defeat */
+                       /* try it on a shorter possibility */
+#ifndef NDEBUG
+                       for (i = 1; i <= m->g->nsub; i++) {
+                               assert(m->pmatch[i].rm_so == -1);
+                               assert(m->pmatch[i].rm_eo == -1);
+                       }
+#endif
+                       NOTE("backoff dissect");
+                       dp = backref(m, m->coldp, endp, gf, gl, (sopno)0);
+               }
+               assert(dp == NULL || dp == endp);
+               if (dp != NULL)         /* found a shorter one */
+                       break;
+
+               /* despite initial appearances, there is no match here */
+               NOTE("false alarm");
+               start = m->coldp + 1;   /* recycle starting later */
+               assert(start <= stop);
+       }
+
+       /* fill in the details if requested */
+       if (nmatch > 0) {
+               pmatch[0].rm_so = m->coldp - m->offp;
+               pmatch[0].rm_eo = endp - m->offp;
+       }
+       if (nmatch > 1) {
+               assert(m->pmatch != NULL);
+               for (i = 1; i < nmatch; i++)
+                       if (i <= m->g->nsub)
+                               pmatch[i] = m->pmatch[i];
+                       else {
+                               pmatch[i].rm_so = -1;
+                               pmatch[i].rm_eo = -1;
+                       }
+       }
+
+       if (m->pmatch != NULL)
+               free((char *)m->pmatch);
+       if (m->lastpos != NULL)
+               free((char *)m->lastpos);
+       STATETEARDOWN(m);
+       return(0);
+}
+
+/*
+ - dissect - figure out what matched what, no back references
+ == static char *dissect(register struct match *m, char *start, \
+ ==    char *stop, sopno startst, sopno stopst);
+ */
+static char *                  /* == stop (success) always */
+dissect(m, start, stop, startst, stopst)
+register struct match *m;
+char *start;
+char *stop;
+sopno startst;
+sopno stopst;
+{
+       register int i;
+       register sopno ss;      /* start sop of current subRE */
+       register sopno es;      /* end sop of current subRE */
+       register char *sp;      /* start of string matched by it */
+       register char *stp;     /* string matched by it cannot pass here */
+       register char *rest;    /* start of rest of string */
+       register char *tail;    /* string unmatched by rest of RE */
+       register sopno ssub;    /* start sop of subsubRE */
+       register sopno esub;    /* end sop of subsubRE */
+       register char *ssp;     /* start of string matched by subsubRE */
+       register char *sep;     /* end of string matched by subsubRE */
+       register char *oldssp;  /* previous ssp */
+       register char *dp;
+
+       AT("diss", start, stop, startst, stopst);
+       sp = start;
+       for (ss = startst; ss < stopst; ss = es) {
+               /* identify end of subRE */
+               es = ss;
+               switch (OP(m->g->strip[es])) {
+               case OPLUS_:
+               case OQUEST_:
+                       es += OPND(m->g->strip[es]);
+                       break;
+               case OCH_:
+                       while (OP(m->g->strip[es]) != O_CH)
+                               es += OPND(m->g->strip[es]);
+                       break;
+               }
+               es++;
+
+               /* figure out what it matched */
+               switch (OP(m->g->strip[ss])) {
+               case OEND:
+                       assert(nope);
+                       break;
+               case OCHAR:
+                       sp++;
+                       break;
+               case OBOL:
+               case OEOL:
+               case OBOW:
+               case OEOW:
+                       break;
+               case OANY:
+               case OANYOF:
+                       sp++;
+                       break;
+               case OBACK_:
+               case O_BACK:
+                       assert(nope);
+                       break;
+               /* cases where length of match is hard to find */
+               case OQUEST_:
+                       stp = stop;
+                       for (;;) {
+                               /* how long could this one be? */
+                               rest = slow(m, sp, stp, ss, es);
+                               assert(rest != NULL);   /* it did match */
+                               /* could the rest match the rest? */
+                               tail = slow(m, rest, stop, es, stopst);
+                               if (tail == stop)
+                                       break;          /* yes! */
+                               /* no -- try a shorter match for this one */
+                               stp = rest - 1;
+                               assert(stp >= sp);      /* it did work */
+                       }
+                       ssub = ss + 1;
+                       esub = es - 1;
+                       /* did innards match? */
+                       if (slow(m, sp, rest, ssub, esub) != NULL) {
+                               dp = dissect(m, sp, rest, ssub, esub);
+                               assert(dp == rest);
+                       } else          /* no */
+                               assert(sp == rest);
+                       sp = rest;
+                       break;
+               case OPLUS_:
+                       stp = stop;
+                       for (;;) {
+                               /* how long could this one be? */
+                               rest = slow(m, sp, stp, ss, es);
+                               assert(rest != NULL);   /* it did match */
+                               /* could the rest match the rest? */
+                               tail = slow(m, rest, stop, es, stopst);
+                               if (tail == stop)
+                                       break;          /* yes! */
+                               /* no -- try a shorter match for this one */
+                               stp = rest - 1;
+                               assert(stp >= sp);      /* it did work */
+                       }
+                       ssub = ss + 1;
+                       esub = es - 1;
+                       ssp = sp;
+                       oldssp = ssp;
+                       for (;;) {      /* find last match of innards */
+                               sep = slow(m, ssp, rest, ssub, esub);
+                               if (sep == NULL || sep == ssp)
+                                       break;  /* failed or matched null */
+                               oldssp = ssp;   /* on to next try */
+                               ssp = sep;
+                       }
+                       if (sep == NULL) {
+                               /* last successful match */
+                               sep = ssp;
+                               ssp = oldssp;
+                       }
+                       assert(sep == rest);    /* must exhaust substring */
+                       assert(slow(m, ssp, sep, ssub, esub) == rest);
+                       dp = dissect(m, ssp, sep, ssub, esub);
+                       assert(dp == sep);
+                       sp = rest;
+                       break;
+               case OCH_:
+                       stp = stop;
+                       for (;;) {
+                               /* how long could this one be? */
+                               rest = slow(m, sp, stp, ss, es);
+                               assert(rest != NULL);   /* it did match */
+                               /* could the rest match the rest? */
+                               tail = slow(m, rest, stop, es, stopst);
+                               if (tail == stop)
+                                       break;          /* yes! */
+                               /* no -- try a shorter match for this one */
+                               stp = rest - 1;
+                               assert(stp >= sp);      /* it did work */
+                       }
+                       ssub = ss + 1;
+                       esub = ss + OPND(m->g->strip[ss]) - 1;
+                       assert(OP(m->g->strip[esub]) == OOR1);
+                       for (;;) {      /* find first matching branch */
+                               if (slow(m, sp, rest, ssub, esub) == rest)
+                                       break;  /* it matched all of it */
+                               /* that one missed, try next one */
+                               assert(OP(m->g->strip[esub]) == OOR1);
+                               esub++;
+                               assert(OP(m->g->strip[esub]) == OOR2);
+                               ssub = esub + 1;
+                               esub += OPND(m->g->strip[esub]);
+                               if (OP(m->g->strip[esub]) == OOR2)
+                                       esub--;
+                               else
+                                       assert(OP(m->g->strip[esub]) == O_CH);
+                       }
+                       dp = dissect(m, sp, rest, ssub, esub);
+                       assert(dp == rest);
+                       sp = rest;
+                       break;
+               case O_PLUS:
+               case O_QUEST:
+               case OOR1:
+               case OOR2:
+               case O_CH:
+                       assert(nope);
+                       break;
+               case OLPAREN:
+                       i = OPND(m->g->strip[ss]);
+                       assert(0 < i && i <= m->g->nsub);
+                       m->pmatch[i].rm_so = sp - m->offp;
+                       break;
+               case ORPAREN:
+                       i = OPND(m->g->strip[ss]);
+                       assert(0 < i && i <= m->g->nsub);
+                       m->pmatch[i].rm_eo = sp - m->offp;
+                       break;
+               default:                /* uh oh */
+                       assert(nope);
+                       break;
+               }
+       }
+
+       assert(sp == stop);
+       return(sp);
+}
+
+/*
+ - backref - figure out what matched what, figuring in back references
+ == static char *backref(register struct match *m, char *start, \
+ ==    char *stop, sopno startst, sopno stopst, sopno lev);
+ */
+static char *                  /* == stop (success) or NULL (failure) */
+backref(m, start, stop, startst, stopst, lev)
+register struct match *m;
+char *start;
+char *stop;
+sopno startst;
+sopno stopst;
+sopno lev;                     /* PLUS nesting level */
+{
+       register int i;
+       register sopno ss;      /* start sop of current subRE */
+       register char *sp;      /* start of string matched by it */
+       register sopno ssub;    /* start sop of subsubRE */
+       register sopno esub;    /* end sop of subsubRE */
+       register char *ssp;     /* start of string matched by subsubRE */
+       register char *dp;
+       register size_t len;
+       register int hard;
+       register sop s;
+       register regoff_t offsave;
+       register cset *cs;
+
+       AT("back", start, stop, startst, stopst);
+       sp = start;
+
+       /* get as far as we can with easy stuff */
+       hard = 0;
+       for (ss = startst; !hard && ss < stopst; ss++)
+               switch (OP(s = m->g->strip[ss])) {
+               case OCHAR:
+                       if (sp == stop || *sp++ != (char)OPND(s))
+                               return(NULL);
+                       break;
+               case OANY:
+                       if (sp == stop)
+                               return(NULL);
+                       sp++;
+                       break;
+               case OANYOF:
+                       cs = &m->g->sets[OPND(s)];
+                       if (sp == stop || !CHIN(cs, *sp++))
+                               return(NULL);
+                       break;
+               case OBOL:
+                       if ( (sp == m->beginp && !(m->eflags&REG_NOTBOL)) ||
+                                       (sp < m->endp && *(sp-1) == '\n' &&
+                                               (m->g->cflags&REG_NEWLINE)) )
+                               { /* yes */ }
+                       else
+                               return(NULL);
+                       break;
+               case OEOL:
+                       if ( (sp == m->endp && !(m->eflags&REG_NOTEOL)) ||
+                                       (sp < m->endp && *sp == '\n' &&
+                                               (m->g->cflags&REG_NEWLINE)) )
+                               { /* yes */ }
+                       else
+                               return(NULL);
+                       break;
+               case OBOW:
+                       if (( (sp == m->beginp && !(m->eflags&REG_NOTBOL)) ||
+                                       (sp < m->endp && *(sp-1) == '\n' &&
+                                               (m->g->cflags&REG_NEWLINE)) ||
+                                       (sp > m->beginp &&
+                                                       !ISWORD(*(sp-1))) ) &&
+                                       (sp < m->endp && ISWORD(*sp)) )
+                               { /* yes */ }
+                       else
+                               return(NULL);
+                       break;
+               case OEOW:
+                       if (( (sp == m->endp && !(m->eflags&REG_NOTEOL)) ||
+                                       (sp < m->endp && *sp == '\n' &&
+                                               (m->g->cflags&REG_NEWLINE)) ||
+                                       (sp < m->endp && !ISWORD(*sp)) ) &&
+                                       (sp > m->beginp && ISWORD(*(sp-1))) )
+                               { /* yes */ }
+                       else
+                               return(NULL);
+                       break;
+               case O_QUEST:
+                       break;
+               case OOR1:      /* matches null but needs to skip */
+                       ss++;
+                       s = m->g->strip[ss];
+                       do {
+                               assert(OP(s) == OOR2);
+                               ss += OPND(s);
+                       } while (OP(s = m->g->strip[ss]) != O_CH);
+                       /* note that the ss++ gets us past the O_CH */
+                       break;
+               default:        /* have to make a choice */
+                       hard = 1;
+                       break;
+               }
+       if (!hard) {            /* that was it! */
+               if (sp != stop)
+                       return(NULL);
+               return(sp);
+       }
+       ss--;                   /* adjust for the for's final increment */
+
+       /* the hard stuff */
+       AT("hard", sp, stop, ss, stopst);
+       s = m->g->strip[ss];
+       switch (OP(s)) {
+       case OBACK_:            /* the vilest depths */
+               i = OPND(s);
+               assert(0 < i && i <= m->g->nsub);
+               if (m->pmatch[i].rm_eo == -1)
+                       return(NULL);
+               assert(m->pmatch[i].rm_so != -1);
+               len = m->pmatch[i].rm_eo - m->pmatch[i].rm_so;
+               assert(stop - m->beginp >= len);
+               if (sp > stop - len)
+                       return(NULL);   /* not enough left to match */
+               ssp = m->offp + m->pmatch[i].rm_so;
+               if (memcmp(sp, ssp, len) != 0)
+                       return(NULL);
+               while (m->g->strip[ss] != SOP(O_BACK, i))
+                       ss++;
+               return(backref(m, sp+len, stop, ss+1, stopst, lev));
+               break;
+       case OQUEST_:           /* to null or not */
+               dp = backref(m, sp, stop, ss+1, stopst, lev);
+               if (dp != NULL)
+                       return(dp);     /* not */
+               return(backref(m, sp, stop, ss+OPND(s)+1, stopst, lev));
+               break;
+       case OPLUS_:
+               assert(m->lastpos != NULL);
+               assert(lev+1 <= m->g->nplus);
+               m->lastpos[lev+1] = sp;
+               return(backref(m, sp, stop, ss+1, stopst, lev+1));
+               break;
+       case O_PLUS:
+               if (sp == m->lastpos[lev])      /* last pass matched null */
+                       return(backref(m, sp, stop, ss+1, stopst, lev-1));
+               /* try another pass */
+               m->lastpos[lev] = sp;
+               dp = backref(m, sp, stop, ss-OPND(s)+1, stopst, lev);
+               if (dp == NULL)
+                       return(backref(m, sp, stop, ss+1, stopst, lev-1));
+               else
+                       return(dp);
+               break;
+       case OCH_:              /* find the right one, if any */
+               ssub = ss + 1;
+               esub = ss + OPND(s) - 1;
+               assert(OP(m->g->strip[esub]) == OOR1);
+               for (;;) {      /* find first matching branch */
+                       dp = backref(m, sp, stop, ssub, esub, lev);
+                       if (dp != NULL)
+                               return(dp);
+                       /* that one missed, try next one */
+                       if (OP(m->g->strip[esub]) == O_CH)
+                               return(NULL);   /* there is none */
+                       esub++;
+                       assert(OP(m->g->strip[esub]) == OOR2);
+                       ssub = esub + 1;
+                       esub += OPND(m->g->strip[esub]);
+                       if (OP(m->g->strip[esub]) == OOR2)
+                               esub--;
+                       else
+                               assert(OP(m->g->strip[esub]) == O_CH);
+               }
+               break;
+       case OLPAREN:           /* must undo assignment if rest fails */
+               i = OPND(s);
+               assert(0 < i && i <= m->g->nsub);
+               offsave = m->pmatch[i].rm_so;
+               m->pmatch[i].rm_so = sp - m->offp;
+               dp = backref(m, sp, stop, ss+1, stopst, lev);
+               if (dp != NULL)
+                       return(dp);
+               m->pmatch[i].rm_so = offsave;
+               return(NULL);
+               break;
+       case ORPAREN:           /* must undo assignment if rest fails */
+               i = OPND(s);
+               assert(0 < i && i <= m->g->nsub);
+               offsave = m->pmatch[i].rm_eo;
+               m->pmatch[i].rm_eo = sp - m->offp;
+               dp = backref(m, sp, stop, ss+1, stopst, lev);
+               if (dp != NULL)
+                       return(dp);
+               m->pmatch[i].rm_eo = offsave;
+               return(NULL);
+               break;
+       default:                /* uh oh */
+               assert(nope);
+               break;
+       }
+
+       /* "can't happen" */
+       assert(nope);
+       /* NOTREACHED */
+       return((char *)NULL);   /* dummy */
+}
+
+/*
+ - fast - step through the string at top speed
+ == static char *fast(register struct match *m, char *start, \
+ ==    char *stop, sopno startst, sopno stopst);
+ */
+static char *                  /* where tentative match ended, or NULL */
+fast(m, start, stop, startst, stopst)
+register struct match *m;
+char *start;
+char *stop;
+sopno startst;
+sopno stopst;
+{
+       register states st = m->st;
+       register states fresh = m->fresh;
+       register states tmp = m->tmp;
+       register char *p = start;
+       register int c = (start == m->beginp) ? OUT : *(start-1);
+       register int lastc;     /* previous c */
+       register int flagch;
+       register int i;
+       register char *coldp;   /* last p after which no match was underway */
+
+       CLEAR(st);
+       SET1(st, startst);
+       st = step(m->g, startst, stopst, st, NOTHING, st);
+       ASSIGN(fresh, st);
+       SP("start", st, *p);
+       coldp = NULL;
+       for (;;) {
+               /* next character */
+               lastc = c;
+               c = (p == m->endp) ? OUT : *p;
+               if (EQ(st, fresh))
+                       coldp = p;
+
+               /* is there an EOL and/or BOL between lastc and c? */
+               flagch = '\0';
+               i = 0;
+               if ( (lastc == '\n' && m->g->cflags&REG_NEWLINE) ||
+                               (lastc == OUT && !(m->eflags&REG_NOTBOL)) ) {
+                       flagch = BOL;
+                       i = m->g->nbol;
+               }
+               if ( (c == '\n' && m->g->cflags&REG_NEWLINE) ||
+                               (c == OUT && !(m->eflags&REG_NOTEOL)) ) {
+                       flagch = (flagch == BOL) ? BOLEOL : EOL;
+                       i += m->g->neol;
+               }
+               if (i != 0) {
+                       for (; i > 0; i--)
+                               st = step(m->g, startst, stopst, st, flagch, st);
+                       SP("boleol", st, c);
+               }
+
+               /* how about a word boundary? */
+               if ( (flagch == BOL || (lastc != OUT && !ISWORD(lastc))) &&
+                                       (c != OUT && ISWORD(c)) ) {
+                       flagch = BOW;
+               }
+               if ( (lastc != OUT && ISWORD(lastc)) &&
+                               (flagch == EOL || (c != OUT && !ISWORD(c))) ) {
+                       flagch = EOW;
+               }
+               if (flagch == BOW || flagch == EOW) {
+                       st = step(m->g, startst, stopst, st, flagch, st);
+                       SP("boweow", st, c);
+               }
+
+               /* are we done? */
+               if (ISSET(st, stopst) || p == stop)
+                       break;          /* NOTE BREAK OUT */
+
+               /* no, we must deal with this character */
+               ASSIGN(tmp, st);
+               ASSIGN(st, fresh);
+               assert(c != OUT);
+               st = step(m->g, startst, stopst, tmp, c, st);
+               SP("aft", st, c);
+               assert(EQ(step(m->g, startst, stopst, st, NOTHING, st), st));
+               p++;
+       }
+
+       assert(coldp != NULL);
+       m->coldp = coldp;
+       if (ISSET(st, stopst))
+               return(p+1);
+       else
+               return(NULL);
+}
+
+/*
+ - slow - step through the string more deliberately
+ == static char *slow(register struct match *m, char *start, \
+ ==    char *stop, sopno startst, sopno stopst);
+ */
+static char *                  /* where it ended */
+slow(m, start, stop, startst, stopst)
+register struct match *m;
+char *start;
+char *stop;
+sopno startst;
+sopno stopst;
+{
+       register states st = m->st;
+       register states empty = m->empty;
+       register states tmp = m->tmp;
+       register char *p = start;
+       register int c = (start == m->beginp) ? OUT : *(start-1);
+       register int lastc;     /* previous c */
+       register int flagch;
+       register int i;
+       register char *matchp;  /* last p at which a match ended */
+
+       AT("slow", start, stop, startst, stopst);
+       CLEAR(st);
+       SET1(st, startst);
+       SP("sstart", st, *p);
+       st = step(m->g, startst, stopst, st, NOTHING, st);
+       matchp = NULL;
+       for (;;) {
+               /* next character */
+               lastc = c;
+               c = (p == m->endp) ? OUT : *p;
+
+               /* is there an EOL and/or BOL between lastc and c? */
+               flagch = '\0';
+               i = 0;
+               if ( (lastc == '\n' && m->g->cflags&REG_NEWLINE) ||
+                               (lastc == OUT && !(m->eflags&REG_NOTBOL)) ) {
+                       flagch = BOL;
+                       i = m->g->nbol;
+               }
+               if ( (c == '\n' && m->g->cflags&REG_NEWLINE) ||
+                               (c == OUT && !(m->eflags&REG_NOTEOL)) ) {
+                       flagch = (flagch == BOL) ? BOLEOL : EOL;
+                       i += m->g->neol;
+               }
+               if (i != 0) {
+                       for (; i > 0; i--)
+                               st = step(m->g, startst, stopst, st, flagch, st);
+                       SP("sboleol", st, c);
+               }
+
+               /* how about a word boundary? */
+               if ( (flagch == BOL || (lastc != OUT && !ISWORD(lastc))) &&
+                                       (c != OUT && ISWORD(c)) ) {
+                       flagch = BOW;
+               }
+               if ( (lastc != OUT && ISWORD(lastc)) &&
+                               (flagch == EOL || (c != OUT && !ISWORD(c))) ) {
+                       flagch = EOW;
+               }
+               if (flagch == BOW || flagch == EOW) {
+                       st = step(m->g, startst, stopst, st, flagch, st);
+                       SP("sboweow", st, c);
+               }
+
+               /* are we done? */
+               if (ISSET(st, stopst))
+                       matchp = p;
+               if (EQ(st, empty) || p == stop)
+                       break;          /* NOTE BREAK OUT */
+
+               /* no, we must deal with this character */
+               ASSIGN(tmp, st);
+               ASSIGN(st, empty);
+               assert(c != OUT);
+               st = step(m->g, startst, stopst, tmp, c, st);
+               SP("saft", st, c);
+               assert(EQ(step(m->g, startst, stopst, st, NOTHING, st), st));
+               p++;
+       }
+
+       return(matchp);
+}
+
+
+/*
+ - step - map set of states reachable before char to set reachable after
+ == static states step(register struct re_guts *g, sopno start, sopno stop, \
+ ==    register states bef, int ch, register states aft);
+ == #define    BOL     (OUT+1)
+ == #define    EOL     (BOL+1)
+ == #define    BOLEOL  (BOL+2)
+ == #define    NOTHING (BOL+3)
+ == #define    BOW     (BOL+4)
+ == #define    EOW     (BOL+5)
+ == #define    CODEMAX (BOL+5)         // highest code used
+ == #define    NONCHAR(c)      ((c) > CHAR_MAX)
+ == #define    NNONCHAR        (CODEMAX-CHAR_MAX)
+ */
+static states
+step(g, start, stop, bef, ch, aft)
+register struct re_guts *g;
+sopno start;                   /* start state within strip */
+sopno stop;                    /* state after stop state within strip */
+register states bef;           /* states reachable before */
+int ch;                                /* character or NONCHAR code */
+register states aft;           /* states already known reachable after */
+{
+       register cset *cs;
+       register sop s;
+       register sopno pc;
+       register onestate here;         /* note, macros know this name */
+       register sopno look;
+       register long i;
+
+       for (pc = start, INIT(here, pc); pc != stop; pc++, INC(here)) {
+               s = g->strip[pc];
+               switch (OP(s)) {
+               case OEND:
+                       assert(pc == stop-1);
+                       break;
+               case OCHAR:
+                       /* only characters can match */
+                       assert(!NONCHAR(ch) || ch != (char)OPND(s));
+                       if (ch == (char)OPND(s))
+                               FWD(aft, bef, 1);
+                       break;
+               case OBOL:
+                       if (ch == BOL || ch == BOLEOL)
+                               FWD(aft, bef, 1);
+                       break;
+               case OEOL:
+                       if (ch == EOL || ch == BOLEOL)
+                               FWD(aft, bef, 1);
+                       break;
+               case OBOW:
+                       if (ch == BOW)
+                               FWD(aft, bef, 1);
+                       break;
+               case OEOW:
+                       if (ch == EOW)
+                               FWD(aft, bef, 1);
+                       break;
+               case OANY:
+                       if (!NONCHAR(ch))
+                               FWD(aft, bef, 1);
+                       break;
+               case OANYOF:
+                       cs = &g->sets[OPND(s)];
+                       if (!NONCHAR(ch) && CHIN(cs, ch))
+                               FWD(aft, bef, 1);
+                       break;
+               case OBACK_:            /* ignored here */
+               case O_BACK:
+                       FWD(aft, aft, 1);
+                       break;
+               case OPLUS_:            /* forward, this is just an empty */
+                       FWD(aft, aft, 1);
+                       break;
+               case O_PLUS:            /* both forward and back */
+                       FWD(aft, aft, 1);
+                       i = ISSETBACK(aft, OPND(s));
+                       BACK(aft, aft, OPND(s));
+                       if (!i && ISSETBACK(aft, OPND(s))) {
+                               /* oho, must reconsider loop body */
+                               pc -= OPND(s) + 1;
+                               INIT(here, pc);
+                       }
+                       break;
+               case OQUEST_:           /* two branches, both forward */
+                       FWD(aft, aft, 1);
+                       FWD(aft, aft, OPND(s));
+                       break;
+               case O_QUEST:           /* just an empty */
+                       FWD(aft, aft, 1);
+                       break;
+               case OLPAREN:           /* not significant here */
+               case ORPAREN:
+                       FWD(aft, aft, 1);
+                       break;
+               case OCH_:              /* mark the first two branches */
+                       FWD(aft, aft, 1);
+                       assert(OP(g->strip[pc+OPND(s)]) == OOR2);
+                       FWD(aft, aft, OPND(s));
+                       break;
+               case OOR1:              /* done a branch, find the O_CH */
+                       if (ISSTATEIN(aft, here)) {
+                               for (look = 1;
+                                               OP(s = g->strip[pc+look]) != O_CH;
+                                               look += OPND(s))
+                                       assert(OP(s) == OOR2);
+                               FWD(aft, aft, look);
+                       }
+                       break;
+               case OOR2:              /* propagate OCH_'s marking */
+                       FWD(aft, aft, 1);
+                       if (OP(g->strip[pc+OPND(s)]) != O_CH) {
+                               assert(OP(g->strip[pc+OPND(s)]) == OOR2);
+                               FWD(aft, aft, OPND(s));
+                       }
+                       break;
+               case O_CH:              /* just empty */
+                       FWD(aft, aft, 1);
+                       break;
+               default:                /* ooooops... */
+                       assert(nope);
+                       break;
+               }
+       }
+
+       return(aft);
+}
+
+#ifdef REDEBUG
+/*
+ - print - print a set of states
+ == #ifdef REDEBUG
+ == static void print(struct match *m, char *caption, states st, \
+ ==    int ch, FILE *d);
+ == #endif
+ */
+static void
+print(m, caption, st, ch, d)
+struct match *m;
+char *caption;
+states st;
+int ch;
+FILE *d;
+{
+       register struct re_guts *g = m->g;
+       register int i;
+       register int first = 1;
+
+       if (!(m->eflags&REG_TRACE))
+               return;
+
+       fprintf(d, "%s", caption);
+       if (ch != '\0')
+               fprintf(d, " %s", pchar(ch));
+       for (i = 0; i < g->nstates; i++)
+               if (ISSET(st, i)) {
+                       fprintf(d, "%s%d", (first) ? "\t" : ", ", i);
+                       first = 0;
+               }
+       fprintf(d, "\n");
+}
+
+/* 
+ - at - print current situation
+ == #ifdef REDEBUG
+ == static void at(struct match *m, char *title, char *start, char *stop, \
+ ==                                            sopno startst, sopno stopst);
+ == #endif
+ */
+static void
+at(m, title, start, stop, startst, stopst)
+struct match *m;
+char *title;
+char *start;
+char *stop;
+sopno startst;
+sopno stopst;
+{
+       if (!(m->eflags&REG_TRACE))
+               return;
+
+       printf("%s %s-", title, pchar(*start));
+       printf("%s ", pchar(*stop));
+       printf("%ld-%ld\n", (long)startst, (long)stopst);
+}
+
+#ifndef PCHARDONE
+#define        PCHARDONE       /* never again */
+/*
+ - pchar - make a character printable
+ == #ifdef REDEBUG
+ == static char *pchar(int ch);
+ == #endif
+ *
+ * Is this identical to regchar() over in debug.c?  Well, yes.  But a
+ * duplicate here avoids having a debugging-capable regexec.o tied to
+ * a matching debug.o, and this is convenient.  It all disappears in
+ * the non-debug compilation anyway, so it doesn't matter much.
+ */
+static char *                  /* -> representation */
+pchar(ch)
+int ch;
+{
+       static char pbuf[10];
+
+       if (isprint(ch) || ch == ' ')
+               sprintf(pbuf, "%c", ch);
+       else
+               sprintf(pbuf, "\\%o", ch);
+       return(pbuf);
+}
+#endif
+#endif
+
+#undef matcher
+#undef fast
+#undef slow
+#undef dissect
+#undef backref
+#undef step
+#undef print
+#undef at
+#undef match
diff --git a/lib/rxspencer/install-sh b/lib/rxspencer/install-sh
new file mode 100755 (executable)
index 0000000..0b0fdcb
--- /dev/null
@@ -0,0 +1,501 @@
+#!/bin/sh
+# install - install a program, script, or datafile
+
+scriptversion=2013-12-25.23; # UTC
+
+# This originates from X11R5 (mit/util/scripts/install.sh), which was
+# later released in X11R6 (xc/config/util/install.sh) with the
+# following copyright and license.
+#
+# Copyright (C) 1994 X Consortium
+#
+# Permission is hereby granted, free of charge, to any person obtaining a copy
+# of this software and associated documentation files (the "Software"), to
+# deal in the Software without restriction, including without limitation the
+# rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+# sell copies of the Software, and to permit persons to whom the Software is
+# furnished to do so, subject to the following conditions:
+#
+# The above copyright notice and this permission notice shall be included in
+# all copies or substantial portions of the Software.
+#
+# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL THE
+# X CONSORTIUM BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+# AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNEC-
+# TION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+#
+# Except as contained in this notice, the name of the X Consortium shall not
+# be used in advertising or otherwise to promote the sale, use or other deal-
+# ings in this Software without prior written authorization from the X Consor-
+# tium.
+#
+#
+# FSF changes to this file are in the public domain.
+#
+# Calling this script install-sh is preferred over install.sh, to prevent
+# 'make' implicit rules from creating a file called install from it
+# when there is no Makefile.
+#
+# This script is compatible with the BSD install script, but was written
+# from scratch.
+
+tab='  '
+nl='
+'
+IFS=" $tab$nl"
+
+# Set DOITPROG to "echo" to test this script.
+
+doit=${DOITPROG-}
+doit_exec=${doit:-exec}
+
+# Put in absolute file names if you don't have them in your path;
+# or use environment vars.
+
+chgrpprog=${CHGRPPROG-chgrp}
+chmodprog=${CHMODPROG-chmod}
+chownprog=${CHOWNPROG-chown}
+cmpprog=${CMPPROG-cmp}
+cpprog=${CPPROG-cp}
+mkdirprog=${MKDIRPROG-mkdir}
+mvprog=${MVPROG-mv}
+rmprog=${RMPROG-rm}
+stripprog=${STRIPPROG-strip}
+
+posix_mkdir=
+
+# Desired mode of installed file.
+mode=0755
+
+chgrpcmd=
+chmodcmd=$chmodprog
+chowncmd=
+mvcmd=$mvprog
+rmcmd="$rmprog -f"
+stripcmd=
+
+src=
+dst=
+dir_arg=
+dst_arg=
+
+copy_on_change=false
+is_target_a_directory=possibly
+
+usage="\
+Usage: $0 [OPTION]... [-T] SRCFILE DSTFILE
+   or: $0 [OPTION]... SRCFILES... DIRECTORY
+   or: $0 [OPTION]... -t DIRECTORY SRCFILES...
+   or: $0 [OPTION]... -d DIRECTORIES...
+
+In the 1st form, copy SRCFILE to DSTFILE.
+In the 2nd and 3rd, copy all SRCFILES to DIRECTORY.
+In the 4th, create DIRECTORIES.
+
+Options:
+     --help     display this help and exit.
+     --version  display version info and exit.
+
+  -c            (ignored)
+  -C            install only if different (preserve the last data modification time)
+  -d            create directories instead of installing files.
+  -g GROUP      $chgrpprog installed files to GROUP.
+  -m MODE       $chmodprog installed files to MODE.
+  -o USER       $chownprog installed files to USER.
+  -s            $stripprog installed files.
+  -t DIRECTORY  install into DIRECTORY.
+  -T            report an error if DSTFILE is a directory.
+
+Environment variables override the default commands:
+  CHGRPPROG CHMODPROG CHOWNPROG CMPPROG CPPROG MKDIRPROG MVPROG
+  RMPROG STRIPPROG
+"
+
+while test $# -ne 0; do
+  case $1 in
+    -c) ;;
+
+    -C) copy_on_change=true;;
+
+    -d) dir_arg=true;;
+
+    -g) chgrpcmd="$chgrpprog $2"
+        shift;;
+
+    --help) echo "$usage"; exit $?;;
+
+    -m) mode=$2
+        case $mode in
+          *' '* | *"$tab"* | *"$nl"* | *'*'* | *'?'* | *'['*)
+            echo "$0: invalid mode: $mode" >&2
+            exit 1;;
+        esac
+        shift;;
+
+    -o) chowncmd="$chownprog $2"
+        shift;;
+
+    -s) stripcmd=$stripprog;;
+
+    -t)
+        is_target_a_directory=always
+        dst_arg=$2
+        # Protect names problematic for 'test' and other utilities.
+        case $dst_arg in
+          -* | [=\(\)!]) dst_arg=./$dst_arg;;
+        esac
+        shift;;
+
+    -T) is_target_a_directory=never;;
+
+    --version) echo "$0 $scriptversion"; exit $?;;
+
+    --) shift
+        break;;
+
+    -*) echo "$0: invalid option: $1" >&2
+        exit 1;;
+
+    *)  break;;
+  esac
+  shift
+done
+
+# We allow the use of options -d and -T together, by making -d
+# take the precedence; this is for compatibility with GNU install.
+
+if test -n "$dir_arg"; then
+  if test -n "$dst_arg"; then
+    echo "$0: target directory not allowed when installing a directory." >&2
+    exit 1
+  fi
+fi
+
+if test $# -ne 0 && test -z "$dir_arg$dst_arg"; then
+  # When -d is used, all remaining arguments are directories to create.
+  # When -t is used, the destination is already specified.
+  # Otherwise, the last argument is the destination.  Remove it from $@.
+  for arg
+  do
+    if test -n "$dst_arg"; then
+      # $@ is not empty: it contains at least $arg.
+      set fnord "$@" "$dst_arg"
+      shift # fnord
+    fi
+    shift # arg
+    dst_arg=$arg
+    # Protect names problematic for 'test' and other utilities.
+    case $dst_arg in
+      -* | [=\(\)!]) dst_arg=./$dst_arg;;
+    esac
+  done
+fi
+
+if test $# -eq 0; then
+  if test -z "$dir_arg"; then
+    echo "$0: no input file specified." >&2
+    exit 1
+  fi
+  # It's OK to call 'install-sh -d' without argument.
+  # This can happen when creating conditional directories.
+  exit 0
+fi
+
+if test -z "$dir_arg"; then
+  if test $# -gt 1 || test "$is_target_a_directory" = always; then
+    if test ! -d "$dst_arg"; then
+      echo "$0: $dst_arg: Is not a directory." >&2
+      exit 1
+    fi
+  fi
+fi
+
+if test -z "$dir_arg"; then
+  do_exit='(exit $ret); exit $ret'
+  trap "ret=129; $do_exit" 1
+  trap "ret=130; $do_exit" 2
+  trap "ret=141; $do_exit" 13
+  trap "ret=143; $do_exit" 15
+
+  # Set umask so as not to create temps with too-generous modes.
+  # However, 'strip' requires both read and write access to temps.
+  case $mode in
+    # Optimize common cases.
+    *644) cp_umask=133;;
+    *755) cp_umask=22;;
+
+    *[0-7])
+      if test -z "$stripcmd"; then
+        u_plus_rw=
+      else
+        u_plus_rw='% 200'
+      fi
+      cp_umask=`expr '(' 777 - $mode % 1000 ')' $u_plus_rw`;;
+    *)
+      if test -z "$stripcmd"; then
+        u_plus_rw=
+      else
+        u_plus_rw=,u+rw
+      fi
+      cp_umask=$mode$u_plus_rw;;
+  esac
+fi
+
+for src
+do
+  # Protect names problematic for 'test' and other utilities.
+  case $src in
+    -* | [=\(\)!]) src=./$src;;
+  esac
+
+  if test -n "$dir_arg"; then
+    dst=$src
+    dstdir=$dst
+    test -d "$dstdir"
+    dstdir_status=$?
+  else
+
+    # Waiting for this to be detected by the "$cpprog $src $dsttmp" command
+    # might cause directories to be created, which would be especially bad
+    # if $src (and thus $dsttmp) contains '*'.
+    if test ! -f "$src" && test ! -d "$src"; then
+      echo "$0: $src does not exist." >&2
+      exit 1
+    fi
+
+    if test -z "$dst_arg"; then
+      echo "$0: no destination specified." >&2
+      exit 1
+    fi
+    dst=$dst_arg
+
+    # If destination is a directory, append the input filename; won't work
+    # if double slashes aren't ignored.
+    if test -d "$dst"; then
+      if test "$is_target_a_directory" = never; then
+        echo "$0: $dst_arg: Is a directory" >&2
+        exit 1
+      fi
+      dstdir=$dst
+      dst=$dstdir/`basename "$src"`
+      dstdir_status=0
+    else
+      dstdir=`dirname "$dst"`
+      test -d "$dstdir"
+      dstdir_status=$?
+    fi
+  fi
+
+  obsolete_mkdir_used=false
+
+  if test $dstdir_status != 0; then
+    case $posix_mkdir in
+      '')
+        # Create intermediate dirs using mode 755 as modified by the umask.
+        # This is like FreeBSD 'install' as of 1997-10-28.
+        umask=`umask`
+        case $stripcmd.$umask in
+          # Optimize common cases.
+          *[2367][2367]) mkdir_umask=$umask;;
+          .*0[02][02] | .[02][02] | .[02]) mkdir_umask=22;;
+
+          *[0-7])
+            mkdir_umask=`expr $umask + 22 \
+              - $umask % 100 % 40 + $umask % 20 \
+              - $umask % 10 % 4 + $umask % 2
+            `;;
+          *) mkdir_umask=$umask,go-w;;
+        esac
+
+        # With -d, create the new directory with the user-specified mode.
+        # Otherwise, rely on $mkdir_umask.
+        if test -n "$dir_arg"; then
+          mkdir_mode=-m$mode
+        else
+          mkdir_mode=
+        fi
+
+        posix_mkdir=false
+        case $umask in
+          *[123567][0-7][0-7])
+            # POSIX mkdir -p sets u+wx bits regardless of umask, which
+            # is incompatible with FreeBSD 'install' when (umask & 300) != 0.
+            ;;
+          *)
+            tmpdir=${TMPDIR-/tmp}/ins$RANDOM-$$
+            trap 'ret=$?; rmdir "$tmpdir/d" "$tmpdir" 2>/dev/null; exit $ret' 0
+
+            if (umask $mkdir_umask &&
+                exec $mkdirprog $mkdir_mode -p -- "$tmpdir/d") >/dev/null 2>&1
+            then
+              if test -z "$dir_arg" || {
+                   # Check for POSIX incompatibilities with -m.
+                   # HP-UX 11.23 and IRIX 6.5 mkdir -m -p sets group- or
+                   # other-writable bit of parent directory when it shouldn't.
+                   # FreeBSD 6.1 mkdir -m -p sets mode of existing directory.
+                   ls_ld_tmpdir=`ls -ld "$tmpdir"`
+                   case $ls_ld_tmpdir in
+                     d????-?r-*) different_mode=700;;
+                     d????-?--*) different_mode=755;;
+                     *) false;;
+                   esac &&
+                   $mkdirprog -m$different_mode -p -- "$tmpdir" && {
+                     ls_ld_tmpdir_1=`ls -ld "$tmpdir"`
+                     test "$ls_ld_tmpdir" = "$ls_ld_tmpdir_1"
+                   }
+                 }
+              then posix_mkdir=:
+              fi
+              rmdir "$tmpdir/d" "$tmpdir"
+            else
+              # Remove any dirs left behind by ancient mkdir implementations.
+              rmdir ./$mkdir_mode ./-p ./-- 2>/dev/null
+            fi
+            trap '' 0;;
+        esac;;
+    esac
+
+    if
+      $posix_mkdir && (
+        umask $mkdir_umask &&
+        $doit_exec $mkdirprog $mkdir_mode -p -- "$dstdir"
+      )
+    then :
+    else
+
+      # The umask is ridiculous, or mkdir does not conform to POSIX,
+      # or it failed possibly due to a race condition.  Create the
+      # directory the slow way, step by step, checking for races as we go.
+
+      case $dstdir in
+        /*) prefix='/';;
+        [-=\(\)!]*) prefix='./';;
+        *)  prefix='';;
+      esac
+
+      oIFS=$IFS
+      IFS=/
+      set -f
+      set fnord $dstdir
+      shift
+      set +f
+      IFS=$oIFS
+
+      prefixes=
+
+      for d
+      do
+        test X"$d" = X && continue
+
+        prefix=$prefix$d
+        if test -d "$prefix"; then
+          prefixes=
+        else
+          if $posix_mkdir; then
+            (umask=$mkdir_umask &&
+             $doit_exec $mkdirprog $mkdir_mode -p -- "$dstdir") && break
+            # Don't fail if two instances are running concurrently.
+            test -d "$prefix" || exit 1
+          else
+            case $prefix in
+              *\'*) qprefix=`echo "$prefix" | sed "s/'/'\\\\\\\\''/g"`;;
+              *) qprefix=$prefix;;
+            esac
+            prefixes="$prefixes '$qprefix'"
+          fi
+        fi
+        prefix=$prefix/
+      done
+
+      if test -n "$prefixes"; then
+        # Don't fail if two instances are running concurrently.
+        (umask $mkdir_umask &&
+         eval "\$doit_exec \$mkdirprog $prefixes") ||
+          test -d "$dstdir" || exit 1
+        obsolete_mkdir_used=true
+      fi
+    fi
+  fi
+
+  if test -n "$dir_arg"; then
+    { test -z "$chowncmd" || $doit $chowncmd "$dst"; } &&
+    { test -z "$chgrpcmd" || $doit $chgrpcmd "$dst"; } &&
+    { test "$obsolete_mkdir_used$chowncmd$chgrpcmd" = false ||
+      test -z "$chmodcmd" || $doit $chmodcmd $mode "$dst"; } || exit 1
+  else
+
+    # Make a couple of temp file names in the proper directory.
+    dsttmp=$dstdir/_inst.$$_
+    rmtmp=$dstdir/_rm.$$_
+
+    # Trap to clean up those temp files at exit.
+    trap 'ret=$?; rm -f "$dsttmp" "$rmtmp" && exit $ret' 0
+
+    # Copy the file name to the temp name.
+    (umask $cp_umask && $doit_exec $cpprog "$src" "$dsttmp") &&
+
+    # and set any options; do chmod last to preserve setuid bits.
+    #
+    # If any of these fail, we abort the whole thing.  If we want to
+    # ignore errors from any of these, just make sure not to ignore
+    # errors from the above "$doit $cpprog $src $dsttmp" command.
+    #
+    { test -z "$chowncmd" || $doit $chowncmd "$dsttmp"; } &&
+    { test -z "$chgrpcmd" || $doit $chgrpcmd "$dsttmp"; } &&
+    { test -z "$stripcmd" || $doit $stripcmd "$dsttmp"; } &&
+    { test -z "$chmodcmd" || $doit $chmodcmd $mode "$dsttmp"; } &&
+
+    # If -C, don't bother to copy if it wouldn't change the file.
+    if $copy_on_change &&
+       old=`LC_ALL=C ls -dlL "$dst"     2>/dev/null` &&
+       new=`LC_ALL=C ls -dlL "$dsttmp"  2>/dev/null` &&
+       set -f &&
+       set X $old && old=:$2:$4:$5:$6 &&
+       set X $new && new=:$2:$4:$5:$6 &&
+       set +f &&
+       test "$old" = "$new" &&
+       $cmpprog "$dst" "$dsttmp" >/dev/null 2>&1
+    then
+      rm -f "$dsttmp"
+    else
+      # Rename the file to the real destination.
+      $doit $mvcmd -f "$dsttmp" "$dst" 2>/dev/null ||
+
+      # The rename failed, perhaps because mv can't rename something else
+      # to itself, or perhaps because mv is so ancient that it does not
+      # support -f.
+      {
+        # Now remove or move aside any old file at destination location.
+        # We try this two ways since rm can't unlink itself on some
+        # systems and the destination file might be busy for other
+        # reasons.  In this case, the final cleanup might fail but the new
+        # file should still install successfully.
+        {
+          test ! -f "$dst" ||
+          $doit $rmcmd -f "$dst" 2>/dev/null ||
+          { $doit $mvcmd -f "$dst" "$rmtmp" 2>/dev/null &&
+            { $doit $rmcmd -f "$rmtmp" 2>/dev/null; :; }
+          } ||
+          { echo "$0: cannot unlink or rename $dst" >&2
+            (exit 1); exit 1
+          }
+        } &&
+
+        # Now rename the file to the real destination.
+        $doit $mvcmd "$dsttmp" "$dst"
+      }
+    fi || exit 1
+
+    trap '' 0
+  fi
+done
+
+# Local variables:
+# eval: (add-hook 'write-file-hooks 'time-stamp)
+# time-stamp-start: "scriptversion="
+# time-stamp-format: "%:y-%02m-%02d.%02H"
+# time-stamp-time-zone: "UTC"
+# time-stamp-end: "; # UTC"
+# End:
diff --git a/lib/rxspencer/ltmain.sh b/lib/rxspencer/ltmain.sh
new file mode 100644 (file)
index 0000000..bffda54
--- /dev/null
@@ -0,0 +1,9661 @@
+
+# libtool (GNU libtool) 2.4.2
+# Written by Gordon Matzigkeit <gord@gnu.ai.mit.edu>, 1996
+
+# Copyright (C) 1996, 1997, 1998, 1999, 2000, 2001, 2003, 2004, 2005, 2006,
+# 2007, 2008, 2009, 2010, 2011 Free Software Foundation, Inc.
+# This is free software; see the source for copying conditions.  There is NO
+# warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+
+# GNU Libtool is free software; you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation; either version 2 of the License, or
+# (at your option) any later version.
+#
+# As a special exception to the GNU General Public License,
+# if you distribute this file as part of a program or library that
+# is built using GNU Libtool, you may include this file under the
+# same distribution terms that you use for the rest of that program.
+#
+# GNU Libtool is distributed in the hope that it will be useful, but
+# WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+# General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with GNU Libtool; see the file COPYING.  If not, a copy
+# can be downloaded from http://www.gnu.org/licenses/gpl.html,
+# or obtained by writing to the Free Software Foundation, Inc.,
+# 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
+
+# Usage: $progname [OPTION]... [MODE-ARG]...
+#
+# Provide generalized library-building support services.
+#
+#       --config             show all configuration variables
+#       --debug              enable verbose shell tracing
+#   -n, --dry-run            display commands without modifying any files
+#       --features           display basic configuration information and exit
+#       --mode=MODE          use operation mode MODE
+#       --preserve-dup-deps  don't remove duplicate dependency libraries
+#       --quiet, --silent    don't print informational messages
+#       --no-quiet, --no-silent
+#                            print informational messages (default)
+#       --no-warn            don't display warning messages
+#       --tag=TAG            use configuration variables from tag TAG
+#   -v, --verbose            print more informational messages than default
+#       --no-verbose         don't print the extra informational messages
+#       --version            print version information
+#   -h, --help, --help-all   print short, long, or detailed help message
+#
+# MODE must be one of the following:
+#
+#         clean              remove files from the build directory
+#         compile            compile a source file into a libtool object
+#         execute            automatically set library path, then run a program
+#         finish             complete the installation of libtool libraries
+#         install            install libraries or executables
+#         link               create a library or an executable
+#         uninstall          remove libraries from an installed directory
+#
+# MODE-ARGS vary depending on the MODE.  When passed as first option,
+# `--mode=MODE' may be abbreviated as `MODE' or a unique abbreviation of that.
+# Try `$progname --help --mode=MODE' for a more detailed description of MODE.
+#
+# When reporting a bug, please describe a test case to reproduce it and
+# include the following information:
+#
+#         host-triplet:        $host
+#         shell:               $SHELL
+#         compiler:            $LTCC
+#         compiler flags:              $LTCFLAGS
+#         linker:              $LD (gnu? $with_gnu_ld)
+#         $progname:   (GNU libtool) 2.4.2 Debian-2.4.2-1.11
+#         automake:    $automake_version
+#         autoconf:    $autoconf_version
+#
+# Report bugs to <bug-libtool@gnu.org>.
+# GNU libtool home page: <http://www.gnu.org/software/libtool/>.
+# General help using GNU software: <http://www.gnu.org/gethelp/>.
+
+PROGRAM=libtool
+PACKAGE=libtool
+VERSION="2.4.2 Debian-2.4.2-1.11"
+TIMESTAMP=""
+package_revision=1.3337
+
+# Be Bourne compatible
+if test -n "${ZSH_VERSION+set}" && (emulate sh) >/dev/null 2>&1; then
+  emulate sh
+  NULLCMD=:
+  # Zsh 3.x and 4.x performs word splitting on ${1+"$@"}, which
+  # is contrary to our usage.  Disable this feature.
+  alias -g '${1+"$@"}'='"$@"'
+  setopt NO_GLOB_SUBST
+else
+  case `(set -o) 2>/dev/null` in *posix*) set -o posix;; esac
+fi
+BIN_SH=xpg4; export BIN_SH # for Tru64
+DUALCASE=1; export DUALCASE # for MKS sh
+
+# A function that is used when there is no print builtin or printf.
+func_fallback_echo ()
+{
+  eval 'cat <<_LTECHO_EOF
+$1
+_LTECHO_EOF'
+}
+
+# NLS nuisances: We save the old values to restore during execute mode.
+lt_user_locale=
+lt_safe_locale=
+for lt_var in LANG LANGUAGE LC_ALL LC_CTYPE LC_COLLATE LC_MESSAGES
+do
+  eval "if test \"\${$lt_var+set}\" = set; then
+          save_$lt_var=\$$lt_var
+          $lt_var=C
+         export $lt_var
+         lt_user_locale=\"$lt_var=\\\$save_\$lt_var; \$lt_user_locale\"
+         lt_safe_locale=\"$lt_var=C; \$lt_safe_locale\"
+       fi"
+done
+LC_ALL=C
+LANGUAGE=C
+export LANGUAGE LC_ALL
+
+$lt_unset CDPATH
+
+
+# Work around backward compatibility issue on IRIX 6.5. On IRIX 6.4+, sh
+# is ksh but when the shell is invoked as "sh" and the current value of
+# the _XPG environment variable is not equal to 1 (one), the special
+# positional parameter $0, within a function call, is the name of the
+# function.
+progpath="$0"
+
+
+
+: ${CP="cp -f"}
+test "${ECHO+set}" = set || ECHO=${as_echo-'printf %s\n'}
+: ${MAKE="make"}
+: ${MKDIR="mkdir"}
+: ${MV="mv -f"}
+: ${RM="rm -f"}
+: ${SHELL="${CONFIG_SHELL-/bin/sh}"}
+: ${Xsed="$SED -e 1s/^X//"}
+
+# Global variables:
+EXIT_SUCCESS=0
+EXIT_FAILURE=1
+EXIT_MISMATCH=63  # $? = 63 is used to indicate version mismatch to missing.
+EXIT_SKIP=77     # $? = 77 is used to indicate a skipped test to automake.
+
+exit_status=$EXIT_SUCCESS
+
+# Make sure IFS has a sensible default
+lt_nl='
+'
+IFS="  $lt_nl"
+
+dirname="s,/[^/]*$,,"
+basename="s,^.*/,,"
+
+# func_dirname file append nondir_replacement
+# Compute the dirname of FILE.  If nonempty, add APPEND to the result,
+# otherwise set result to NONDIR_REPLACEMENT.
+func_dirname ()
+{
+    func_dirname_result=`$ECHO "${1}" | $SED "$dirname"`
+    if test "X$func_dirname_result" = "X${1}"; then
+      func_dirname_result="${3}"
+    else
+      func_dirname_result="$func_dirname_result${2}"
+    fi
+} # func_dirname may be replaced by extended shell implementation
+
+
+# func_basename file
+func_basename ()
+{
+    func_basename_result=`$ECHO "${1}" | $SED "$basename"`
+} # func_basename may be replaced by extended shell implementation
+
+
+# func_dirname_and_basename file append nondir_replacement
+# perform func_basename and func_dirname in a single function
+# call:
+#   dirname:  Compute the dirname of FILE.  If nonempty,
+#             add APPEND to the result, otherwise set result
+#             to NONDIR_REPLACEMENT.
+#             value returned in "$func_dirname_result"
+#   basename: Compute filename of FILE.
+#             value retuned in "$func_basename_result"
+# Implementation must be kept synchronized with func_dirname
+# and func_basename. For efficiency, we do not delegate to
+# those functions but instead duplicate the functionality here.
+func_dirname_and_basename ()
+{
+    # Extract subdirectory from the argument.
+    func_dirname_result=`$ECHO "${1}" | $SED -e "$dirname"`
+    if test "X$func_dirname_result" = "X${1}"; then
+      func_dirname_result="${3}"
+    else
+      func_dirname_result="$func_dirname_result${2}"
+    fi
+    func_basename_result=`$ECHO "${1}" | $SED -e "$basename"`
+} # func_dirname_and_basename may be replaced by extended shell implementation
+
+
+# func_stripname prefix suffix name
+# strip PREFIX and SUFFIX off of NAME.
+# PREFIX and SUFFIX must not contain globbing or regex special
+# characters, hashes, percent signs, but SUFFIX may contain a leading
+# dot (in which case that matches only a dot).
+# func_strip_suffix prefix name
+func_stripname ()
+{
+    case ${2} in
+      .*) func_stripname_result=`$ECHO "${3}" | $SED "s%^${1}%%; s%\\\\${2}\$%%"`;;
+      *)  func_stripname_result=`$ECHO "${3}" | $SED "s%^${1}%%; s%${2}\$%%"`;;
+    esac
+} # func_stripname may be replaced by extended shell implementation
+
+
+# These SED scripts presuppose an absolute path with a trailing slash.
+pathcar='s,^/\([^/]*\).*$,\1,'
+pathcdr='s,^/[^/]*,,'
+removedotparts=':dotsl
+               s@/\./@/@g
+               t dotsl
+               s,/\.$,/,'
+collapseslashes='s@/\{1,\}@/@g'
+finalslash='s,/*$,/,'
+
+# func_normal_abspath PATH
+# Remove doubled-up and trailing slashes, "." path components,
+# and cancel out any ".." path components in PATH after making
+# it an absolute path.
+#             value returned in "$func_normal_abspath_result"
+func_normal_abspath ()
+{
+  # Start from root dir and reassemble the path.
+  func_normal_abspath_result=
+  func_normal_abspath_tpath=$1
+  func_normal_abspath_altnamespace=
+  case $func_normal_abspath_tpath in
+    "")
+      # Empty path, that just means $cwd.
+      func_stripname '' '/' "`pwd`"
+      func_normal_abspath_result=$func_stripname_result
+      return
+    ;;
+    # The next three entries are used to spot a run of precisely
+    # two leading slashes without using negated character classes;
+    # we take advantage of case's first-match behaviour.
+    ///*)
+      # Unusual form of absolute path, do nothing.
+    ;;
+    //*)
+      # Not necessarily an ordinary path; POSIX reserves leading '//'
+      # and for example Cygwin uses it to access remote file shares
+      # over CIFS/SMB, so we conserve a leading double slash if found.
+      func_normal_abspath_altnamespace=/
+    ;;
+    /*)
+      # Absolute path, do nothing.
+    ;;
+    *)
+      # Relative path, prepend $cwd.
+      func_normal_abspath_tpath=`pwd`/$func_normal_abspath_tpath
+    ;;
+  esac
+  # Cancel out all the simple stuff to save iterations.  We also want
+  # the path to end with a slash for ease of parsing, so make sure
+  # there is one (and only one) here.
+  func_normal_abspath_tpath=`$ECHO "$func_normal_abspath_tpath" | $SED \
+        -e "$removedotparts" -e "$collapseslashes" -e "$finalslash"`
+  while :; do
+    # Processed it all yet?
+    if test "$func_normal_abspath_tpath" = / ; then
+      # If we ascended to the root using ".." the result may be empty now.
+      if test -z "$func_normal_abspath_result" ; then
+        func_normal_abspath_result=/
+      fi
+      break
+    fi
+    func_normal_abspath_tcomponent=`$ECHO "$func_normal_abspath_tpath" | $SED \
+        -e "$pathcar"`
+    func_normal_abspath_tpath=`$ECHO "$func_normal_abspath_tpath" | $SED \
+        -e "$pathcdr"`
+    # Figure out what to do with it
+    case $func_normal_abspath_tcomponent in
+      "")
+        # Trailing empty path component, ignore it.
+      ;;
+      ..)
+        # Parent dir; strip last assembled component from result.
+        func_dirname "$func_normal_abspath_result"
+        func_normal_abspath_result=$func_dirname_result
+      ;;
+      *)
+        # Actual path component, append it.
+        func_normal_abspath_result=$func_normal_abspath_result/$func_normal_abspath_tcomponent
+      ;;
+    esac
+  done
+  # Restore leading double-slash if one was found on entry.
+  func_normal_abspath_result=$func_normal_abspath_altnamespace$func_normal_abspath_result
+}
+
+# func_relative_path SRCDIR DSTDIR
+# generates a relative path from SRCDIR to DSTDIR, with a trailing
+# slash if non-empty, suitable for immediately appending a filename
+# without needing to append a separator.
+#             value returned in "$func_relative_path_result"
+func_relative_path ()
+{
+  func_relative_path_result=
+  func_normal_abspath "$1"
+  func_relative_path_tlibdir=$func_normal_abspath_result
+  func_normal_abspath "$2"
+  func_relative_path_tbindir=$func_normal_abspath_result
+
+  # Ascend the tree starting from libdir
+  while :; do
+    # check if we have found a prefix of bindir
+    case $func_relative_path_tbindir in
+      $func_relative_path_tlibdir)
+        # found an exact match
+        func_relative_path_tcancelled=
+        break
+        ;;
+      $func_relative_path_tlibdir*)
+        # found a matching prefix
+        func_stripname "$func_relative_path_tlibdir" '' "$func_relative_path_tbindir"
+        func_relative_path_tcancelled=$func_stripname_result
+        if test -z "$func_relative_path_result"; then
+          func_relative_path_result=.
+        fi
+        break
+        ;;
+      *)
+        func_dirname $func_relative_path_tlibdir
+        func_relative_path_tlibdir=${func_dirname_result}
+        if test "x$func_relative_path_tlibdir" = x ; then
+          # Have to descend all the way to the root!
+          func_relative_path_result=../$func_relative_path_result
+          func_relative_path_tcancelled=$func_relative_path_tbindir
+          break
+        fi
+        func_relative_path_result=../$func_relative_path_result
+        ;;
+    esac
+  done
+
+  # Now calculate path; take care to avoid doubling-up slashes.
+  func_stripname '' '/' "$func_relative_path_result"
+  func_relative_path_result=$func_stripname_result
+  func_stripname '/' '/' "$func_relative_path_tcancelled"
+  if test "x$func_stripname_result" != x ; then
+    func_relative_path_result=${func_relative_path_result}/${func_stripname_result}
+  fi
+
+  # Normalisation. If bindir is libdir, return empty string,
+  # else relative path ending with a slash; either way, target
+  # file name can be directly appended.
+  if test ! -z "$func_relative_path_result"; then
+    func_stripname './' '' "$func_relative_path_result/"
+    func_relative_path_result=$func_stripname_result
+  fi
+}
+
+# The name of this program:
+func_dirname_and_basename "$progpath"
+progname=$func_basename_result
+
+# Make sure we have an absolute path for reexecution:
+case $progpath in
+  [\\/]*|[A-Za-z]:\\*) ;;
+  *[\\/]*)
+     progdir=$func_dirname_result
+     progdir=`cd "$progdir" && pwd`
+     progpath="$progdir/$progname"
+     ;;
+  *)
+     save_IFS="$IFS"
+     IFS=${PATH_SEPARATOR-:}
+     for progdir in $PATH; do
+       IFS="$save_IFS"
+       test -x "$progdir/$progname" && break
+     done
+     IFS="$save_IFS"
+     test -n "$progdir" || progdir=`pwd`
+     progpath="$progdir/$progname"
+     ;;
+esac
+
+# Sed substitution that helps us do robust quoting.  It backslashifies
+# metacharacters that are still active within double-quoted strings.
+Xsed="${SED}"' -e 1s/^X//'
+sed_quote_subst='s/\([`"$\\]\)/\\\1/g'
+
+# Same as above, but do not quote variable references.
+double_quote_subst='s/\(["`\\]\)/\\\1/g'
+
+# Sed substitution that turns a string into a regex matching for the
+# string literally.
+sed_make_literal_regex='s,[].[^$\\*\/],\\&,g'
+
+# Sed substitution that converts a w32 file name or path
+# which contains forward slashes, into one that contains
+# (escaped) backslashes.  A very naive implementation.
+lt_sed_naive_backslashify='s|\\\\*|\\|g;s|/|\\|g;s|\\|\\\\|g'
+
+# Re-`\' parameter expansions in output of double_quote_subst that were
+# `\'-ed in input to the same.  If an odd number of `\' preceded a '$'
+# in input to double_quote_subst, that '$' was protected from expansion.
+# Since each input `\' is now two `\'s, look for any number of runs of
+# four `\'s followed by two `\'s and then a '$'.  `\' that '$'.
+bs='\\'
+bs2='\\\\'
+bs4='\\\\\\\\'
+dollar='\$'
+sed_double_backslash="\
+  s/$bs4/&\\
+/g
+  s/^$bs2$dollar/$bs&/
+  s/\\([^$bs]\\)$bs2$dollar/\\1$bs2$bs$dollar/g
+  s/\n//g"
+
+# Standard options:
+opt_dry_run=false
+opt_help=false
+opt_quiet=false
+opt_verbose=false
+opt_warning=:
+
+# func_echo arg...
+# Echo program name prefixed message, along with the current mode
+# name if it has been set yet.
+func_echo ()
+{
+    $ECHO "$progname: ${opt_mode+$opt_mode: }$*"
+}
+
+# func_verbose arg...
+# Echo program name prefixed message in verbose mode only.
+func_verbose ()
+{
+    $opt_verbose && func_echo ${1+"$@"}
+
+    # A bug in bash halts the script if the last line of a function
+    # fails when set -e is in force, so we need another command to
+    # work around that:
+    :
+}
+
+# func_echo_all arg...
+# Invoke $ECHO with all args, space-separated.
+func_echo_all ()
+{
+    $ECHO "$*"
+}
+
+# func_error arg...
+# Echo program name prefixed message to standard error.
+func_error ()
+{
+    $ECHO "$progname: ${opt_mode+$opt_mode: }"${1+"$@"} 1>&2
+}
+
+# func_warning arg...
+# Echo program name prefixed warning message to standard error.
+func_warning ()
+{
+    $opt_warning && $ECHO "$progname: ${opt_mode+$opt_mode: }warning: "${1+"$@"} 1>&2
+
+    # bash bug again:
+    :
+}
+
+# func_fatal_error arg...
+# Echo program name prefixed message to standard error, and exit.
+func_fatal_error ()
+{
+    func_error ${1+"$@"}
+    exit $EXIT_FAILURE
+}
+
+# func_fatal_help arg...
+# Echo program name prefixed message to standard error, followed by
+# a help hint, and exit.
+func_fatal_help ()
+{
+    func_error ${1+"$@"}
+    func_fatal_error "$help"
+}
+help="Try \`$progname --help' for more information."  ## default
+
+
+# func_grep expression filename
+# Check whether EXPRESSION matches any line of FILENAME, without output.
+func_grep ()
+{
+    $GREP "$1" "$2" >/dev/null 2>&1
+}
+
+
+# func_mkdir_p directory-path
+# Make sure the entire path to DIRECTORY-PATH is available.
+func_mkdir_p ()
+{
+    my_directory_path="$1"
+    my_dir_list=
+
+    if test -n "$my_directory_path" && test "$opt_dry_run" != ":"; then
+
+      # Protect directory names starting with `-'
+      case $my_directory_path in
+        -*) my_directory_path="./$my_directory_path" ;;
+      esac
+
+      # While some portion of DIR does not yet exist...
+      while test ! -d "$my_directory_path"; do
+        # ...make a list in topmost first order.  Use a colon delimited
+       # list incase some portion of path contains whitespace.
+        my_dir_list="$my_directory_path:$my_dir_list"
+
+        # If the last portion added has no slash in it, the list is done
+        case $my_directory_path in */*) ;; *) break ;; esac
+
+        # ...otherwise throw away the child directory and loop
+        my_directory_path=`$ECHO "$my_directory_path" | $SED -e "$dirname"`
+      done
+      my_dir_list=`$ECHO "$my_dir_list" | $SED 's,:*$,,'`
+
+      save_mkdir_p_IFS="$IFS"; IFS=':'
+      for my_dir in $my_dir_list; do
+       IFS="$save_mkdir_p_IFS"
+        # mkdir can fail with a `File exist' error if two processes
+        # try to create one of the directories concurrently.  Don't
+        # stop in that case!
+        $MKDIR "$my_dir" 2>/dev/null || :
+      done
+      IFS="$save_mkdir_p_IFS"
+
+      # Bail out if we (or some other process) failed to create a directory.
+      test -d "$my_directory_path" || \
+        func_fatal_error "Failed to create \`$1'"
+    fi
+}
+
+
+# func_mktempdir [string]
+# Make a temporary directory that won't clash with other running
+# libtool processes, and avoids race conditions if possible.  If
+# given, STRING is the basename for that directory.
+func_mktempdir ()
+{
+    my_template="${TMPDIR-/tmp}/${1-$progname}"
+
+    if test "$opt_dry_run" = ":"; then
+      # Return a directory name, but don't create it in dry-run mode
+      my_tmpdir="${my_template}-$$"
+    else
+
+      # If mktemp works, use that first and foremost
+      my_tmpdir=`mktemp -d "${my_template}-XXXXXXXX" 2>/dev/null`
+
+      if test ! -d "$my_tmpdir"; then
+        # Failing that, at least try and use $RANDOM to avoid a race
+        my_tmpdir="${my_template}-${RANDOM-0}$$"
+
+        save_mktempdir_umask=`umask`
+        umask 0077
+        $MKDIR "$my_tmpdir"
+        umask $save_mktempdir_umask
+      fi
+
+      # If we're not in dry-run mode, bomb out on failure
+      test -d "$my_tmpdir" || \
+        func_fatal_error "cannot create temporary directory \`$my_tmpdir'"
+    fi
+
+    $ECHO "$my_tmpdir"
+}
+
+
+# func_quote_for_eval arg
+# Aesthetically quote ARG to be evaled later.
+# This function returns two values: FUNC_QUOTE_FOR_EVAL_RESULT
+# is double-quoted, suitable for a subsequent eval, whereas
+# FUNC_QUOTE_FOR_EVAL_UNQUOTED_RESULT has merely all characters
+# which are still active within double quotes backslashified.
+func_quote_for_eval ()
+{
+    case $1 in
+      *[\\\`\"\$]*)
+       func_quote_for_eval_unquoted_result=`$ECHO "$1" | $SED "$sed_quote_subst"` ;;
+      *)
+        func_quote_for_eval_unquoted_result="$1" ;;
+    esac
+
+    case $func_quote_for_eval_unquoted_result in
+      # Double-quote args containing shell metacharacters to delay
+      # word splitting, command substitution and and variable
+      # expansion for a subsequent eval.
+      # Many Bourne shells cannot handle close brackets correctly
+      # in scan sets, so we specify it separately.
+      *[\[\~\#\^\&\*\(\)\{\}\|\;\<\>\?\'\ \    ]*|*]*|"")
+        func_quote_for_eval_result="\"$func_quote_for_eval_unquoted_result\""
+        ;;
+      *)
+        func_quote_for_eval_result="$func_quote_for_eval_unquoted_result"
+    esac
+}
+
+
+# func_quote_for_expand arg
+# Aesthetically quote ARG to be evaled later; same as above,
+# but do not quote variable references.
+func_quote_for_expand ()
+{
+    case $1 in
+      *[\\\`\"]*)
+       my_arg=`$ECHO "$1" | $SED \
+           -e "$double_quote_subst" -e "$sed_double_backslash"` ;;
+      *)
+        my_arg="$1" ;;
+    esac
+
+    case $my_arg in
+      # Double-quote args containing shell metacharacters to delay
+      # word splitting and command substitution for a subsequent eval.
+      # Many Bourne shells cannot handle close brackets correctly
+      # in scan sets, so we specify it separately.
+      *[\[\~\#\^\&\*\(\)\{\}\|\;\<\>\?\'\ \    ]*|*]*|"")
+        my_arg="\"$my_arg\""
+        ;;
+    esac
+
+    func_quote_for_expand_result="$my_arg"
+}
+
+
+# func_show_eval cmd [fail_exp]
+# Unless opt_silent is true, then output CMD.  Then, if opt_dryrun is
+# not true, evaluate CMD.  If the evaluation of CMD fails, and FAIL_EXP
+# is given, then evaluate it.
+func_show_eval ()
+{
+    my_cmd="$1"
+    my_fail_exp="${2-:}"
+
+    ${opt_silent-false} || {
+      func_quote_for_expand "$my_cmd"
+      eval "func_echo $func_quote_for_expand_result"
+    }
+
+    if ${opt_dry_run-false}; then :; else
+      eval "$my_cmd"
+      my_status=$?
+      if test "$my_status" -eq 0; then :; else
+       eval "(exit $my_status); $my_fail_exp"
+      fi
+    fi
+}
+
+
+# func_show_eval_locale cmd [fail_exp]
+# Unless opt_silent is true, then output CMD.  Then, if opt_dryrun is
+# not true, evaluate CMD.  If the evaluation of CMD fails, and FAIL_EXP
+# is given, then evaluate it.  Use the saved locale for evaluation.
+func_show_eval_locale ()
+{
+    my_cmd="$1"
+    my_fail_exp="${2-:}"
+
+    ${opt_silent-false} || {
+      func_quote_for_expand "$my_cmd"
+      eval "func_echo $func_quote_for_expand_result"
+    }
+
+    if ${opt_dry_run-false}; then :; else
+      eval "$lt_user_locale
+           $my_cmd"
+      my_status=$?
+      eval "$lt_safe_locale"
+      if test "$my_status" -eq 0; then :; else
+       eval "(exit $my_status); $my_fail_exp"
+      fi
+    fi
+}
+
+# func_tr_sh
+# Turn $1 into a string suitable for a shell variable name.
+# Result is stored in $func_tr_sh_result.  All characters
+# not in the set a-zA-Z0-9_ are replaced with '_'. Further,
+# if $1 begins with a digit, a '_' is prepended as well.
+func_tr_sh ()
+{
+  case $1 in
+  [0-9]* | *[!a-zA-Z0-9_]*)
+    func_tr_sh_result=`$ECHO "$1" | $SED 's/^\([0-9]\)/_\1/; s/[^a-zA-Z0-9_]/_/g'`
+    ;;
+  * )
+    func_tr_sh_result=$1
+    ;;
+  esac
+}
+
+
+# func_version
+# Echo version message to standard output and exit.
+func_version ()
+{
+    $opt_debug
+
+    $SED -n '/(C)/!b go
+       :more
+       /\./!{
+         N
+         s/\n# / /
+         b more
+       }
+       :go
+       /^# '$PROGRAM' (GNU /,/# warranty; / {
+        s/^# //
+       s/^# *$//
+        s/\((C)\)[ 0-9,-]*\( [1-9][0-9]*\)/\1\2/
+        p
+     }' < "$progpath"
+     exit $?
+}
+
+# func_usage
+# Echo short help message to standard output and exit.
+func_usage ()
+{
+    $opt_debug
+
+    $SED -n '/^# Usage:/,/^#  *.*--help/ {
+        s/^# //
+       s/^# *$//
+       s/\$progname/'$progname'/
+       p
+    }' < "$progpath"
+    echo
+    $ECHO "run \`$progname --help | more' for full usage"
+    exit $?
+}
+
+# func_help [NOEXIT]
+# Echo long help message to standard output and exit,
+# unless 'noexit' is passed as argument.
+func_help ()
+{
+    $opt_debug
+
+    $SED -n '/^# Usage:/,/# Report bugs to/ {
+       :print
+        s/^# //
+       s/^# *$//
+       s*\$progname*'$progname'*
+       s*\$host*'"$host"'*
+       s*\$SHELL*'"$SHELL"'*
+       s*\$LTCC*'"$LTCC"'*
+       s*\$LTCFLAGS*'"$LTCFLAGS"'*
+       s*\$LD*'"$LD"'*
+       s/\$with_gnu_ld/'"$with_gnu_ld"'/
+       s/\$automake_version/'"`(${AUTOMAKE-automake} --version) 2>/dev/null |$SED 1q`"'/
+       s/\$autoconf_version/'"`(${AUTOCONF-autoconf} --version) 2>/dev/null |$SED 1q`"'/
+       p
+       d
+     }
+     /^# .* home page:/b print
+     /^# General help using/b print
+     ' < "$progpath"
+    ret=$?
+    if test -z "$1"; then
+      exit $ret
+    fi
+}
+
+# func_missing_arg argname
+# Echo program name prefixed message to standard error and set global
+# exit_cmd.
+func_missing_arg ()
+{
+    $opt_debug
+
+    func_error "missing argument for $1."
+    exit_cmd=exit
+}
+
+
+# func_split_short_opt shortopt
+# Set func_split_short_opt_name and func_split_short_opt_arg shell
+# variables after splitting SHORTOPT after the 2nd character.
+func_split_short_opt ()
+{
+    my_sed_short_opt='1s/^\(..\).*$/\1/;q'
+    my_sed_short_rest='1s/^..\(.*\)$/\1/;q'
+
+    func_split_short_opt_name=`$ECHO "$1" | $SED "$my_sed_short_opt"`
+    func_split_short_opt_arg=`$ECHO "$1" | $SED "$my_sed_short_rest"`
+} # func_split_short_opt may be replaced by extended shell implementation
+
+
+# func_split_long_opt longopt
+# Set func_split_long_opt_name and func_split_long_opt_arg shell
+# variables after splitting LONGOPT at the `=' sign.
+func_split_long_opt ()
+{
+    my_sed_long_opt='1s/^\(--[^=]*\)=.*/\1/;q'
+    my_sed_long_arg='1s/^--[^=]*=//'
+
+    func_split_long_opt_name=`$ECHO "$1" | $SED "$my_sed_long_opt"`
+    func_split_long_opt_arg=`$ECHO "$1" | $SED "$my_sed_long_arg"`
+} # func_split_long_opt may be replaced by extended shell implementation
+
+exit_cmd=:
+
+
+
+
+
+magic="%%%MAGIC variable%%%"
+magic_exe="%%%MAGIC EXE variable%%%"
+
+# Global variables.
+nonopt=
+preserve_args=
+lo2o="s/\\.lo\$/.${objext}/"
+o2lo="s/\\.${objext}\$/.lo/"
+extracted_archives=
+extracted_serial=0
+
+# If this variable is set in any of the actions, the command in it
+# will be execed at the end.  This prevents here-documents from being
+# left over by shells.
+exec_cmd=
+
+# func_append var value
+# Append VALUE to the end of shell variable VAR.
+func_append ()
+{
+    eval "${1}=\$${1}\${2}"
+} # func_append may be replaced by extended shell implementation
+
+# func_append_quoted var value
+# Quote VALUE and append to the end of shell variable VAR, separated
+# by a space.
+func_append_quoted ()
+{
+    func_quote_for_eval "${2}"
+    eval "${1}=\$${1}\\ \$func_quote_for_eval_result"
+} # func_append_quoted may be replaced by extended shell implementation
+
+
+# func_arith arithmetic-term...
+func_arith ()
+{
+    func_arith_result=`expr "${@}"`
+} # func_arith may be replaced by extended shell implementation
+
+
+# func_len string
+# STRING may not start with a hyphen.
+func_len ()
+{
+    func_len_result=`expr "${1}" : ".*" 2>/dev/null || echo $max_cmd_len`
+} # func_len may be replaced by extended shell implementation
+
+
+# func_lo2o object
+func_lo2o ()
+{
+    func_lo2o_result=`$ECHO "${1}" | $SED "$lo2o"`
+} # func_lo2o may be replaced by extended shell implementation
+
+
+# func_xform libobj-or-source
+func_xform ()
+{
+    func_xform_result=`$ECHO "${1}" | $SED 's/\.[^.]*$/.lo/'`
+} # func_xform may be replaced by extended shell implementation
+
+
+# func_fatal_configuration arg...
+# Echo program name prefixed message to standard error, followed by
+# a configuration failure hint, and exit.
+func_fatal_configuration ()
+{
+    func_error ${1+"$@"}
+    func_error "See the $PACKAGE documentation for more information."
+    func_fatal_error "Fatal configuration error."
+}
+
+
+# func_config
+# Display the configuration for all the tags in this script.
+func_config ()
+{
+    re_begincf='^# ### BEGIN LIBTOOL'
+    re_endcf='^# ### END LIBTOOL'
+
+    # Default configuration.
+    $SED "1,/$re_begincf CONFIG/d;/$re_endcf CONFIG/,\$d" < "$progpath"
+
+    # Now print the configurations for the tags.
+    for tagname in $taglist; do
+      $SED -n "/$re_begincf TAG CONFIG: $tagname\$/,/$re_endcf TAG CONFIG: $tagname\$/p" < "$progpath"
+    done
+
+    exit $?
+}
+
+# func_features
+# Display the features supported by this script.
+func_features ()
+{
+    echo "host: $host"
+    if test "$build_libtool_libs" = yes; then
+      echo "enable shared libraries"
+    else
+      echo "disable shared libraries"
+    fi
+    if test "$build_old_libs" = yes; then
+      echo "enable static libraries"
+    else
+      echo "disable static libraries"
+    fi
+
+    exit $?
+}
+
+# func_enable_tag tagname
+# Verify that TAGNAME is valid, and either flag an error and exit, or
+# enable the TAGNAME tag.  We also add TAGNAME to the global $taglist
+# variable here.
+func_enable_tag ()
+{
+  # Global variable:
+  tagname="$1"
+
+  re_begincf="^# ### BEGIN LIBTOOL TAG CONFIG: $tagname\$"
+  re_endcf="^# ### END LIBTOOL TAG CONFIG: $tagname\$"
+  sed_extractcf="/$re_begincf/,/$re_endcf/p"
+
+  # Validate tagname.
+  case $tagname in
+    *[!-_A-Za-z0-9,/]*)
+      func_fatal_error "invalid tag name: $tagname"
+      ;;
+  esac
+
+  # Don't test for the "default" C tag, as we know it's
+  # there but not specially marked.
+  case $tagname in
+    CC) ;;
+    *)
+      if $GREP "$re_begincf" "$progpath" >/dev/null 2>&1; then
+       taglist="$taglist $tagname"
+
+       # Evaluate the configuration.  Be careful to quote the path
+       # and the sed script, to avoid splitting on whitespace, but
+       # also don't use non-portable quotes within backquotes within
+       # quotes we have to do it in 2 steps:
+       extractedcf=`$SED -n -e "$sed_extractcf" < "$progpath"`
+       eval "$extractedcf"
+      else
+       func_error "ignoring unknown tag $tagname"
+      fi
+      ;;
+  esac
+}
+
+# func_check_version_match
+# Ensure that we are using m4 macros, and libtool script from the same
+# release of libtool.
+func_check_version_match ()
+{
+  if test "$package_revision" != "$macro_revision"; then
+    if test "$VERSION" != "$macro_version"; then
+      if test -z "$macro_version"; then
+        cat >&2 <<_LT_EOF
+$progname: Version mismatch error.  This is $PACKAGE $VERSION, but the
+$progname: definition of this LT_INIT comes from an older release.
+$progname: You should recreate aclocal.m4 with macros from $PACKAGE $VERSION
+$progname: and run autoconf again.
+_LT_EOF
+      else
+        cat >&2 <<_LT_EOF
+$progname: Version mismatch error.  This is $PACKAGE $VERSION, but the
+$progname: definition of this LT_INIT comes from $PACKAGE $macro_version.
+$progname: You should recreate aclocal.m4 with macros from $PACKAGE $VERSION
+$progname: and run autoconf again.
+_LT_EOF
+      fi
+    else
+      cat >&2 <<_LT_EOF
+$progname: Version mismatch error.  This is $PACKAGE $VERSION, revision $package_revision,
+$progname: but the definition of this LT_INIT comes from revision $macro_revision.
+$progname: You should recreate aclocal.m4 with macros from revision $package_revision
+$progname: of $PACKAGE $VERSION and run autoconf again.
+_LT_EOF
+    fi
+
+    exit $EXIT_MISMATCH
+  fi
+}
+
+
+# Shorthand for --mode=foo, only valid as the first argument
+case $1 in
+clean|clea|cle|cl)
+  shift; set dummy --mode clean ${1+"$@"}; shift
+  ;;
+compile|compil|compi|comp|com|co|c)
+  shift; set dummy --mode compile ${1+"$@"}; shift
+  ;;
+execute|execut|execu|exec|exe|ex|e)
+  shift; set dummy --mode execute ${1+"$@"}; shift
+  ;;
+finish|finis|fini|fin|fi|f)
+  shift; set dummy --mode finish ${1+"$@"}; shift
+  ;;
+install|instal|insta|inst|ins|in|i)
+  shift; set dummy --mode install ${1+"$@"}; shift
+  ;;
+link|lin|li|l)
+  shift; set dummy --mode link ${1+"$@"}; shift
+  ;;
+uninstall|uninstal|uninsta|uninst|unins|unin|uni|un|u)
+  shift; set dummy --mode uninstall ${1+"$@"}; shift
+  ;;
+esac
+
+
+
+# Option defaults:
+opt_debug=:
+opt_dry_run=false
+opt_config=false
+opt_preserve_dup_deps=false
+opt_features=false
+opt_finish=false
+opt_help=false
+opt_help_all=false
+opt_silent=:
+opt_warning=:
+opt_verbose=:
+opt_silent=false
+opt_verbose=false
+
+
+# Parse options once, thoroughly.  This comes as soon as possible in the
+# script to make things like `--version' happen as quickly as we can.
+{
+  # this just eases exit handling
+  while test $# -gt 0; do
+    opt="$1"
+    shift
+    case $opt in
+      --debug|-x)      opt_debug='set -x'
+                       func_echo "enabling shell trace mode"
+                       $opt_debug
+                       ;;
+      --dry-run|--dryrun|-n)
+                       opt_dry_run=:
+                       ;;
+      --config)
+                       opt_config=:
+func_config
+                       ;;
+      --dlopen|-dlopen)
+                       optarg="$1"
+                       opt_dlopen="${opt_dlopen+$opt_dlopen
+}$optarg"
+                       shift
+                       ;;
+      --preserve-dup-deps)
+                       opt_preserve_dup_deps=:
+                       ;;
+      --features)
+                       opt_features=:
+func_features
+                       ;;
+      --finish)
+                       opt_finish=:
+set dummy --mode finish ${1+"$@"}; shift
+                       ;;
+      --help)
+                       opt_help=:
+                       ;;
+      --help-all)
+                       opt_help_all=:
+opt_help=': help-all'
+                       ;;
+      --mode)
+                       test $# = 0 && func_missing_arg $opt && break
+                       optarg="$1"
+                       opt_mode="$optarg"
+case $optarg in
+  # Valid mode arguments:
+  clean|compile|execute|finish|install|link|relink|uninstall) ;;
+
+  # Catch anything else as an error
+  *) func_error "invalid argument for $opt"
+     exit_cmd=exit
+     break
+     ;;
+esac
+                       shift
+                       ;;
+      --no-silent|--no-quiet)
+                       opt_silent=false
+func_append preserve_args " $opt"
+                       ;;
+      --no-warning|--no-warn)
+                       opt_warning=false
+func_append preserve_args " $opt"
+                       ;;
+      --no-verbose)
+                       opt_verbose=false
+func_append preserve_args " $opt"
+                       ;;
+      --silent|--quiet)
+                       opt_silent=:
+func_append preserve_args " $opt"
+        opt_verbose=false
+                       ;;
+      --verbose|-v)
+                       opt_verbose=:
+func_append preserve_args " $opt"
+opt_silent=false
+                       ;;
+      --tag)
+                       test $# = 0 && func_missing_arg $opt && break
+                       optarg="$1"
+                       opt_tag="$optarg"
+func_append preserve_args " $opt $optarg"
+func_enable_tag "$optarg"
+                       shift
+                       ;;
+
+      -\?|-h)          func_usage                              ;;
+      --help)          func_help                               ;;
+      --version)       func_version                            ;;
+
+      # Separate optargs to long options:
+      --*=*)
+                       func_split_long_opt "$opt"
+                       set dummy "$func_split_long_opt_name" "$func_split_long_opt_arg" ${1+"$@"}
+                       shift
+                       ;;
+
+      # Separate non-argument short options:
+      -\?*|-h*|-n*|-v*)
+                       func_split_short_opt "$opt"
+                       set dummy "$func_split_short_opt_name" "-$func_split_short_opt_arg" ${1+"$@"}
+                       shift
+                       ;;
+
+      --)              break                                   ;;
+      -*)              func_fatal_help "unrecognized option \`$opt'" ;;
+      *)               set dummy "$opt" ${1+"$@"};     shift; break  ;;
+    esac
+  done
+
+  # Validate options:
+
+  # save first non-option argument
+  if test "$#" -gt 0; then
+    nonopt="$opt"
+    shift
+  fi
+
+  # preserve --debug
+  test "$opt_debug" = : || func_append preserve_args " --debug"
+
+  case $host in
+    *cygwin* | *mingw* | *pw32* | *cegcc*)
+      # don't eliminate duplications in $postdeps and $predeps
+      opt_duplicate_compiler_generated_deps=:
+      ;;
+    *)
+      opt_duplicate_compiler_generated_deps=$opt_preserve_dup_deps
+      ;;
+  esac
+
+  $opt_help || {
+    # Sanity checks first:
+    func_check_version_match
+
+    if test "$build_libtool_libs" != yes && test "$build_old_libs" != yes; then
+      func_fatal_configuration "not configured to build any kind of library"
+    fi
+
+    # Darwin sucks
+    eval std_shrext=\"$shrext_cmds\"
+
+    # Only execute mode is allowed to have -dlopen flags.
+    if test -n "$opt_dlopen" && test "$opt_mode" != execute; then
+      func_error "unrecognized option \`-dlopen'"
+      $ECHO "$help" 1>&2
+      exit $EXIT_FAILURE
+    fi
+
+    # Change the help message to a mode-specific one.
+    generic_help="$help"
+    help="Try \`$progname --help --mode=$opt_mode' for more information."
+  }
+
+
+  # Bail if the options were screwed
+  $exit_cmd $EXIT_FAILURE
+}
+
+
+
+
+## ----------- ##
+##    Main.    ##
+## ----------- ##
+
+# func_lalib_p file
+# True iff FILE is a libtool `.la' library or `.lo' object file.
+# This function is only a basic sanity check; it will hardly flush out
+# determined imposters.
+func_lalib_p ()
+{
+    test -f "$1" &&
+      $SED -e 4q "$1" 2>/dev/null \
+        | $GREP "^# Generated by .*$PACKAGE" > /dev/null 2>&1
+}
+
+# func_lalib_unsafe_p file
+# True iff FILE is a libtool `.la' library or `.lo' object file.
+# This function implements the same check as func_lalib_p without
+# resorting to external programs.  To this end, it redirects stdin and
+# closes it afterwards, without saving the original file descriptor.
+# As a safety measure, use it only where a negative result would be
+# fatal anyway.  Works if `file' does not exist.
+func_lalib_unsafe_p ()
+{
+    lalib_p=no
+    if test -f "$1" && test -r "$1" && exec 5<&0 <"$1"; then
+       for lalib_p_l in 1 2 3 4
+       do
+           read lalib_p_line
+           case "$lalib_p_line" in
+               \#\ Generated\ by\ *$PACKAGE* ) lalib_p=yes; break;;
+           esac
+       done
+       exec 0<&5 5<&-
+    fi
+    test "$lalib_p" = yes
+}
+
+# func_ltwrapper_script_p file
+# True iff FILE is a libtool wrapper script
+# This function is only a basic sanity check; it will hardly flush out
+# determined imposters.
+func_ltwrapper_script_p ()
+{
+    func_lalib_p "$1"
+}
+
+# func_ltwrapper_executable_p file
+# True iff FILE is a libtool wrapper executable
+# This function is only a basic sanity check; it will hardly flush out
+# determined imposters.
+func_ltwrapper_executable_p ()
+{
+    func_ltwrapper_exec_suffix=
+    case $1 in
+    *.exe) ;;
+    *) func_ltwrapper_exec_suffix=.exe ;;
+    esac
+    $GREP "$magic_exe" "$1$func_ltwrapper_exec_suffix" >/dev/null 2>&1
+}
+
+# func_ltwrapper_scriptname file
+# Assumes file is an ltwrapper_executable
+# uses $file to determine the appropriate filename for a
+# temporary ltwrapper_script.
+func_ltwrapper_scriptname ()
+{
+    func_dirname_and_basename "$1" "" "."
+    func_stripname '' '.exe' "$func_basename_result"
+    func_ltwrapper_scriptname_result="$func_dirname_result/$objdir/${func_stripname_result}_ltshwrapper"
+}
+
+# func_ltwrapper_p file
+# True iff FILE is a libtool wrapper script or wrapper executable
+# This function is only a basic sanity check; it will hardly flush out
+# determined imposters.
+func_ltwrapper_p ()
+{
+    func_ltwrapper_script_p "$1" || func_ltwrapper_executable_p "$1"
+}
+
+
+# func_execute_cmds commands fail_cmd
+# Execute tilde-delimited COMMANDS.
+# If FAIL_CMD is given, eval that upon failure.
+# FAIL_CMD may read-access the current command in variable CMD!
+func_execute_cmds ()
+{
+    $opt_debug
+    save_ifs=$IFS; IFS='~'
+    for cmd in $1; do
+      IFS=$save_ifs
+      eval cmd=\"$cmd\"
+      func_show_eval "$cmd" "${2-:}"
+    done
+    IFS=$save_ifs
+}
+
+
+# func_source file
+# Source FILE, adding directory component if necessary.
+# Note that it is not necessary on cygwin/mingw to append a dot to
+# FILE even if both FILE and FILE.exe exist: automatic-append-.exe
+# behavior happens only for exec(3), not for open(2)!  Also, sourcing
+# `FILE.' does not work on cygwin managed mounts.
+func_source ()
+{
+    $opt_debug
+    case $1 in
+    */* | *\\*)        . "$1" ;;
+    *)         . "./$1" ;;
+    esac
+}
+
+
+# func_resolve_sysroot PATH
+# Replace a leading = in PATH with a sysroot.  Store the result into
+# func_resolve_sysroot_result
+func_resolve_sysroot ()
+{
+  func_resolve_sysroot_result=$1
+  case $func_resolve_sysroot_result in
+  =*)
+    func_stripname '=' '' "$func_resolve_sysroot_result"
+    func_resolve_sysroot_result=$lt_sysroot$func_stripname_result
+    ;;
+  esac
+}
+
+# func_replace_sysroot PATH
+# If PATH begins with the sysroot, replace it with = and
+# store the result into func_replace_sysroot_result.
+func_replace_sysroot ()
+{
+  case "$lt_sysroot:$1" in
+  ?*:"$lt_sysroot"*)
+    func_stripname "$lt_sysroot" '' "$1"
+    func_replace_sysroot_result="=$func_stripname_result"
+    ;;
+  *)
+    # Including no sysroot.
+    func_replace_sysroot_result=$1
+    ;;
+  esac
+}
+
+# func_infer_tag arg
+# Infer tagged configuration to use if any are available and
+# if one wasn't chosen via the "--tag" command line option.
+# Only attempt this if the compiler in the base compile
+# command doesn't match the default compiler.
+# arg is usually of the form 'gcc ...'
+func_infer_tag ()
+{
+    $opt_debug
+    if test -n "$available_tags" && test -z "$tagname"; then
+      CC_quoted=
+      for arg in $CC; do
+       func_append_quoted CC_quoted "$arg"
+      done
+      CC_expanded=`func_echo_all $CC`
+      CC_quoted_expanded=`func_echo_all $CC_quoted`
+      case $@ in
+      # Blanks in the command may have been stripped by the calling shell,
+      # but not from the CC environment variable when configure was run.
+      " $CC "* | "$CC "* | " $CC_expanded "* | "$CC_expanded "* | \
+      " $CC_quoted"* | "$CC_quoted "* | " $CC_quoted_expanded "* | "$CC_quoted_expanded "*) ;;
+      # Blanks at the start of $base_compile will cause this to fail
+      # if we don't check for them as well.
+      *)
+       for z in $available_tags; do
+         if $GREP "^# ### BEGIN LIBTOOL TAG CONFIG: $z$" < "$progpath" > /dev/null; then
+           # Evaluate the configuration.
+           eval "`${SED} -n -e '/^# ### BEGIN LIBTOOL TAG CONFIG: '$z'$/,/^# ### END LIBTOOL TAG CONFIG: '$z'$/p' < $progpath`"
+           CC_quoted=
+           for arg in $CC; do
+             # Double-quote args containing other shell metacharacters.
+             func_append_quoted CC_quoted "$arg"
+           done
+           CC_expanded=`func_echo_all $CC`
+           CC_quoted_expanded=`func_echo_all $CC_quoted`
+           case "$@ " in
+           " $CC "* | "$CC "* | " $CC_expanded "* | "$CC_expanded "* | \
+           " $CC_quoted"* | "$CC_quoted "* | " $CC_quoted_expanded "* | "$CC_quoted_expanded "*)
+             # The compiler in the base compile command matches
+             # the one in the tagged configuration.
+             # Assume this is the tagged configuration we want.
+             tagname=$z
+             break
+             ;;
+           esac
+         fi
+       done
+       # If $tagname still isn't set, then no tagged configuration
+       # was found and let the user know that the "--tag" command
+       # line option must be used.
+       if test -z "$tagname"; then
+         func_echo "unable to infer tagged configuration"
+         func_fatal_error "specify a tag with \`--tag'"
+#      else
+#        func_verbose "using $tagname tagged configuration"
+       fi
+       ;;
+      esac
+    fi
+}
+
+
+
+# func_write_libtool_object output_name pic_name nonpic_name
+# Create a libtool object file (analogous to a ".la" file),
+# but don't create it if we're doing a dry run.
+func_write_libtool_object ()
+{
+    write_libobj=${1}
+    if test "$build_libtool_libs" = yes; then
+      write_lobj=\'${2}\'
+    else
+      write_lobj=none
+    fi
+
+    if test "$build_old_libs" = yes; then
+      write_oldobj=\'${3}\'
+    else
+      write_oldobj=none
+    fi
+
+    $opt_dry_run || {
+      cat >${write_libobj}T <<EOF
+# $write_libobj - a libtool object file
+# Generated by $PROGRAM (GNU $PACKAGE$TIMESTAMP) $VERSION
+#
+# Please DO NOT delete this file!
+# It is necessary for linking the library.
+
+# Name of the PIC object.
+pic_object=$write_lobj
+
+# Name of the non-PIC object
+non_pic_object=$write_oldobj
+
+EOF
+      $MV "${write_libobj}T" "${write_libobj}"
+    }
+}
+
+
+##################################################
+# FILE NAME AND PATH CONVERSION HELPER FUNCTIONS #
+##################################################
+
+# func_convert_core_file_wine_to_w32 ARG
+# Helper function used by file name conversion functions when $build is *nix,
+# and $host is mingw, cygwin, or some other w32 environment. Relies on a
+# correctly configured wine environment available, with the winepath program
+# in $build's $PATH.
+#
+# ARG is the $build file name to be converted to w32 format.
+# Result is available in $func_convert_core_file_wine_to_w32_result, and will
+# be empty on error (or when ARG is empty)
+func_convert_core_file_wine_to_w32 ()
+{
+  $opt_debug
+  func_convert_core_file_wine_to_w32_result="$1"
+  if test -n "$1"; then
+    # Unfortunately, winepath does not exit with a non-zero error code, so we
+    # are forced to check the contents of stdout. On the other hand, if the
+    # command is not found, the shell will set an exit code of 127 and print
+    # *an error message* to stdout. So we must check for both error code of
+    # zero AND non-empty stdout, which explains the odd construction:
+    func_convert_core_file_wine_to_w32_tmp=`winepath -w "$1" 2>/dev/null`
+    if test "$?" -eq 0 && test -n "${func_convert_core_file_wine_to_w32_tmp}"; then
+      func_convert_core_file_wine_to_w32_result=`$ECHO "$func_convert_core_file_wine_to_w32_tmp" |
+        $SED -e "$lt_sed_naive_backslashify"`
+    else
+      func_convert_core_file_wine_to_w32_result=
+    fi
+  fi
+}
+# end: func_convert_core_file_wine_to_w32
+
+
+# func_convert_core_path_wine_to_w32 ARG
+# Helper function used by path conversion functions when $build is *nix, and
+# $host is mingw, cygwin, or some other w32 environment. Relies on a correctly
+# configured wine environment available, with the winepath program in $build's
+# $PATH. Assumes ARG has no leading or trailing path separator characters.
+#
+# ARG is path to be converted from $build format to win32.
+# Result is available in $func_convert_core_path_wine_to_w32_result.
+# Unconvertible file (directory) names in ARG are skipped; if no directory names
+# are convertible, then the result may be empty.
+func_convert_core_path_wine_to_w32 ()
+{
+  $opt_debug
+  # unfortunately, winepath doesn't convert paths, only file names
+  func_convert_core_path_wine_to_w32_result=""
+  if test -n "$1"; then
+    oldIFS=$IFS
+    IFS=:
+    for func_convert_core_path_wine_to_w32_f in $1; do
+      IFS=$oldIFS
+      func_convert_core_file_wine_to_w32 "$func_convert_core_path_wine_to_w32_f"
+      if test -n "$func_convert_core_file_wine_to_w32_result" ; then
+        if test -z "$func_convert_core_path_wine_to_w32_result"; then
+          func_convert_core_path_wine_to_w32_result="$func_convert_core_file_wine_to_w32_result"
+        else
+          func_append func_convert_core_path_wine_to_w32_result ";$func_convert_core_file_wine_to_w32_result"
+        fi
+      fi
+    done
+    IFS=$oldIFS
+  fi
+}
+# end: func_convert_core_path_wine_to_w32
+
+
+# func_cygpath ARGS...
+# Wrapper around calling the cygpath program via LT_CYGPATH. This is used when
+# when (1) $build is *nix and Cygwin is hosted via a wine environment; or (2)
+# $build is MSYS and $host is Cygwin, or (3) $build is Cygwin. In case (1) or
+# (2), returns the Cygwin file name or path in func_cygpath_result (input
+# file name or path is assumed to be in w32 format, as previously converted
+# from $build's *nix or MSYS format). In case (3), returns the w32 file name
+# or path in func_cygpath_result (input file name or path is assumed to be in
+# Cygwin format). Returns an empty string on error.
+#
+# ARGS are passed to cygpath, with the last one being the file name or path to
+# be converted.
+#
+# Specify the absolute *nix (or w32) name to cygpath in the LT_CYGPATH
+# environment variable; do not put it in $PATH.
+func_cygpath ()
+{
+  $opt_debug
+  if test -n "$LT_CYGPATH" && test -f "$LT_CYGPATH"; then
+    func_cygpath_result=`$LT_CYGPATH "$@" 2>/dev/null`
+    if test "$?" -ne 0; then
+      # on failure, ensure result is empty
+      func_cygpath_result=
+    fi
+  else
+    func_cygpath_result=
+    func_error "LT_CYGPATH is empty or specifies non-existent file: \`$LT_CYGPATH'"
+  fi
+}
+#end: func_cygpath
+
+
+# func_convert_core_msys_to_w32 ARG
+# Convert file name or path ARG from MSYS format to w32 format.  Return
+# result in func_convert_core_msys_to_w32_result.
+func_convert_core_msys_to_w32 ()
+{
+  $opt_debug
+  # awkward: cmd appends spaces to result
+  func_convert_core_msys_to_w32_result=`( cmd //c echo "$1" ) 2>/dev/null |
+    $SED -e 's/[ ]*$//' -e "$lt_sed_naive_backslashify"`
+}
+#end: func_convert_core_msys_to_w32
+
+
+# func_convert_file_check ARG1 ARG2
+# Verify that ARG1 (a file name in $build format) was converted to $host
+# format in ARG2. Otherwise, emit an error message, but continue (resetting
+# func_to_host_file_result to ARG1).
+func_convert_file_check ()
+{
+  $opt_debug
+  if test -z "$2" && test -n "$1" ; then
+    func_error "Could not determine host file name corresponding to"
+    func_error "  \`$1'"
+    func_error "Continuing, but uninstalled executables may not work."
+    # Fallback:
+    func_to_host_file_result="$1"
+  fi
+}
+# end func_convert_file_check
+
+
+# func_convert_path_check FROM_PATHSEP TO_PATHSEP FROM_PATH TO_PATH
+# Verify that FROM_PATH (a path in $build format) was converted to $host
+# format in TO_PATH. Otherwise, emit an error message, but continue, resetting
+# func_to_host_file_result to a simplistic fallback value (see below).
+func_convert_path_check ()
+{
+  $opt_debug
+  if test -z "$4" && test -n "$3"; then
+    func_error "Could not determine the host path corresponding to"
+    func_error "  \`$3'"
+    func_error "Continuing, but uninstalled executables may not work."
+    # Fallback.  This is a deliberately simplistic "conversion" and
+    # should not be "improved".  See libtool.info.
+    if test "x$1" != "x$2"; then
+      lt_replace_pathsep_chars="s|$1|$2|g"
+      func_to_host_path_result=`echo "$3" |
+        $SED -e "$lt_replace_pathsep_chars"`
+    else
+      func_to_host_path_result="$3"
+    fi
+  fi
+}
+# end func_convert_path_check
+
+
+# func_convert_path_front_back_pathsep FRONTPAT BACKPAT REPL ORIG
+# Modifies func_to_host_path_result by prepending REPL if ORIG matches FRONTPAT
+# and appending REPL if ORIG matches BACKPAT.
+func_convert_path_front_back_pathsep ()
+{
+  $opt_debug
+  case $4 in
+  $1 ) func_to_host_path_result="$3$func_to_host_path_result"
+    ;;
+  esac
+  case $4 in
+  $2 ) func_append func_to_host_path_result "$3"
+    ;;
+  esac
+}
+# end func_convert_path_front_back_pathsep
+
+
+##################################################
+# $build to $host FILE NAME CONVERSION FUNCTIONS #
+##################################################
+# invoked via `$to_host_file_cmd ARG'
+#
+# In each case, ARG is the path to be converted from $build to $host format.
+# Result will be available in $func_to_host_file_result.
+
+
+# func_to_host_file ARG
+# Converts the file name ARG from $build format to $host format. Return result
+# in func_to_host_file_result.
+func_to_host_file ()
+{
+  $opt_debug
+  $to_host_file_cmd "$1"
+}
+# end func_to_host_file
+
+
+# func_to_tool_file ARG LAZY
+# converts the file name ARG from $build format to toolchain format. Return
+# result in func_to_tool_file_result.  If the conversion in use is listed
+# in (the comma separated) LAZY, no conversion takes place.
+func_to_tool_file ()
+{
+  $opt_debug
+  case ,$2, in
+    *,"$to_tool_file_cmd",*)
+      func_to_tool_file_result=$1
+      ;;
+    *)
+      $to_tool_file_cmd "$1"
+      func_to_tool_file_result=$func_to_host_file_result
+      ;;
+  esac
+}
+# end func_to_tool_file
+
+
+# func_convert_file_noop ARG
+# Copy ARG to func_to_host_file_result.
+func_convert_file_noop ()
+{
+  func_to_host_file_result="$1"
+}
+# end func_convert_file_noop
+
+
+# func_convert_file_msys_to_w32 ARG
+# Convert file name ARG from (mingw) MSYS to (mingw) w32 format; automatic
+# conversion to w32 is not available inside the cwrapper.  Returns result in
+# func_to_host_file_result.
+func_convert_file_msys_to_w32 ()
+{
+  $opt_debug
+  func_to_host_file_result="$1"
+  if test -n "$1"; then
+    func_convert_core_msys_to_w32 "$1"
+    func_to_host_file_result="$func_convert_core_msys_to_w32_result"
+  fi
+  func_convert_file_check "$1" "$func_to_host_file_result"
+}
+# end func_convert_file_msys_to_w32
+
+
+# func_convert_file_cygwin_to_w32 ARG
+# Convert file name ARG from Cygwin to w32 format.  Returns result in
+# func_to_host_file_result.
+func_convert_file_cygwin_to_w32 ()
+{
+  $opt_debug
+  func_to_host_file_result="$1"
+  if test -n "$1"; then
+    # because $build is cygwin, we call "the" cygpath in $PATH; no need to use
+    # LT_CYGPATH in this case.
+    func_to_host_file_result=`cygpath -m "$1"`
+  fi
+  func_convert_file_check "$1" "$func_to_host_file_result"
+}
+# end func_convert_file_cygwin_to_w32
+
+
+# func_convert_file_nix_to_w32 ARG
+# Convert file name ARG from *nix to w32 format.  Requires a wine environment
+# and a working winepath. Returns result in func_to_host_file_result.
+func_convert_file_nix_to_w32 ()
+{
+  $opt_debug
+  func_to_host_file_result="$1"
+  if test -n "$1"; then
+    func_convert_core_file_wine_to_w32 "$1"
+    func_to_host_file_result="$func_convert_core_file_wine_to_w32_result"
+  fi
+  func_convert_file_check "$1" "$func_to_host_file_result"
+}
+# end func_convert_file_nix_to_w32
+
+
+# func_convert_file_msys_to_cygwin ARG
+# Convert file name ARG from MSYS to Cygwin format.  Requires LT_CYGPATH set.
+# Returns result in func_to_host_file_result.
+func_convert_file_msys_to_cygwin ()
+{
+  $opt_debug
+  func_to_host_file_result="$1"
+  if test -n "$1"; then
+    func_convert_core_msys_to_w32 "$1"
+    func_cygpath -u "$func_convert_core_msys_to_w32_result"
+    func_to_host_file_result="$func_cygpath_result"
+  fi
+  func_convert_file_check "$1" "$func_to_host_file_result"
+}
+# end func_convert_file_msys_to_cygwin
+
+
+# func_convert_file_nix_to_cygwin ARG
+# Convert file name ARG from *nix to Cygwin format.  Requires Cygwin installed
+# in a wine environment, working winepath, and LT_CYGPATH set.  Returns result
+# in func_to_host_file_result.
+func_convert_file_nix_to_cygwin ()
+{
+  $opt_debug
+  func_to_host_file_result="$1"
+  if test -n "$1"; then
+    # convert from *nix to w32, then use cygpath to convert from w32 to cygwin.
+    func_convert_core_file_wine_to_w32 "$1"
+    func_cygpath -u "$func_convert_core_file_wine_to_w32_result"
+    func_to_host_file_result="$func_cygpath_result"
+  fi
+  func_convert_file_check "$1" "$func_to_host_file_result"
+}
+# end func_convert_file_nix_to_cygwin
+
+
+#############################################
+# $build to $host PATH CONVERSION FUNCTIONS #
+#############################################
+# invoked via `$to_host_path_cmd ARG'
+#
+# In each case, ARG is the path to be converted from $build to $host format.
+# The result will be available in $func_to_host_path_result.
+#
+# Path separators are also converted from $build format to $host format.  If
+# ARG begins or ends with a path separator character, it is preserved (but
+# converted to $host format) on output.
+#
+# All path conversion functions are named using the following convention:
+#   file name conversion function    : func_convert_file_X_to_Y ()
+#   path conversion function         : func_convert_path_X_to_Y ()
+# where, for any given $build/$host combination the 'X_to_Y' value is the
+# same.  If conversion functions are added for new $build/$host combinations,
+# the two new functions must follow this pattern, or func_init_to_host_path_cmd
+# will break.
+
+
+# func_init_to_host_path_cmd
+# Ensures that function "pointer" variable $to_host_path_cmd is set to the
+# appropriate value, based on the value of $to_host_file_cmd.
+to_host_path_cmd=
+func_init_to_host_path_cmd ()
+{
+  $opt_debug
+  if test -z "$to_host_path_cmd"; then
+    func_stripname 'func_convert_file_' '' "$to_host_file_cmd"
+    to_host_path_cmd="func_convert_path_${func_stripname_result}"
+  fi
+}
+
+
+# func_to_host_path ARG
+# Converts the path ARG from $build format to $host format. Return result
+# in func_to_host_path_result.
+func_to_host_path ()
+{
+  $opt_debug
+  func_init_to_host_path_cmd
+  $to_host_path_cmd "$1"
+}
+# end func_to_host_path
+
+
+# func_convert_path_noop ARG
+# Copy ARG to func_to_host_path_result.
+func_convert_path_noop ()
+{
+  func_to_host_path_result="$1"
+}
+# end func_convert_path_noop
+
+
+# func_convert_path_msys_to_w32 ARG
+# Convert path ARG from (mingw) MSYS to (mingw) w32 format; automatic
+# conversion to w32 is not available inside the cwrapper.  Returns result in
+# func_to_host_path_result.
+func_convert_path_msys_to_w32 ()
+{
+  $opt_debug
+  func_to_host_path_result="$1"
+  if test -n "$1"; then
+    # Remove leading and trailing path separator characters from ARG.  MSYS
+    # behavior is inconsistent here; cygpath turns them into '.;' and ';.';
+    # and winepath ignores them completely.
+    func_stripname : : "$1"
+    func_to_host_path_tmp1=$func_stripname_result
+    func_convert_core_msys_to_w32 "$func_to_host_path_tmp1"
+    func_to_host_path_result="$func_convert_core_msys_to_w32_result"
+    func_convert_path_check : ";" \
+      "$func_to_host_path_tmp1" "$func_to_host_path_result"
+    func_convert_path_front_back_pathsep ":*" "*:" ";" "$1"
+  fi
+}
+# end func_convert_path_msys_to_w32
+
+
+# func_convert_path_cygwin_to_w32 ARG
+# Convert path ARG from Cygwin to w32 format.  Returns result in
+# func_to_host_file_result.
+func_convert_path_cygwin_to_w32 ()
+{
+  $opt_debug
+  func_to_host_path_result="$1"
+  if test -n "$1"; then
+    # See func_convert_path_msys_to_w32:
+    func_stripname : : "$1"
+    func_to_host_path_tmp1=$func_stripname_result
+    func_to_host_path_result=`cygpath -m -p "$func_to_host_path_tmp1"`
+    func_convert_path_check : ";" \
+      "$func_to_host_path_tmp1" "$func_to_host_path_result"
+    func_convert_path_front_back_pathsep ":*" "*:" ";" "$1"
+  fi
+}
+# end func_convert_path_cygwin_to_w32
+
+
+# func_convert_path_nix_to_w32 ARG
+# Convert path ARG from *nix to w32 format.  Requires a wine environment and
+# a working winepath.  Returns result in func_to_host_file_result.
+func_convert_path_nix_to_w32 ()
+{
+  $opt_debug
+  func_to_host_path_result="$1"
+  if test -n "$1"; then
+    # See func_convert_path_msys_to_w32:
+    func_stripname : : "$1"
+    func_to_host_path_tmp1=$func_stripname_result
+    func_convert_core_path_wine_to_w32 "$func_to_host_path_tmp1"
+    func_to_host_path_result="$func_convert_core_path_wine_to_w32_result"
+    func_convert_path_check : ";" \
+      "$func_to_host_path_tmp1" "$func_to_host_path_result"
+    func_convert_path_front_back_pathsep ":*" "*:" ";" "$1"
+  fi
+}
+# end func_convert_path_nix_to_w32
+
+
+# func_convert_path_msys_to_cygwin ARG
+# Convert path ARG from MSYS to Cygwin format.  Requires LT_CYGPATH set.
+# Returns result in func_to_host_file_result.
+func_convert_path_msys_to_cygwin ()
+{
+  $opt_debug
+  func_to_host_path_result="$1"
+  if test -n "$1"; then
+    # See func_convert_path_msys_to_w32:
+    func_stripname : : "$1"
+    func_to_host_path_tmp1=$func_stripname_result
+    func_convert_core_msys_to_w32 "$func_to_host_path_tmp1"
+    func_cygpath -u -p "$func_convert_core_msys_to_w32_result"
+    func_to_host_path_result="$func_cygpath_result"
+    func_convert_path_check : : \
+      "$func_to_host_path_tmp1" "$func_to_host_path_result"
+    func_convert_path_front_back_pathsep ":*" "*:" : "$1"
+  fi
+}
+# end func_convert_path_msys_to_cygwin
+
+
+# func_convert_path_nix_to_cygwin ARG
+# Convert path ARG from *nix to Cygwin format.  Requires Cygwin installed in a
+# a wine environment, working winepath, and LT_CYGPATH set.  Returns result in
+# func_to_host_file_result.
+func_convert_path_nix_to_cygwin ()
+{
+  $opt_debug
+  func_to_host_path_result="$1"
+  if test -n "$1"; then
+    # Remove leading and trailing path separator characters from
+    # ARG. msys behavior is inconsistent here, cygpath turns them
+    # into '.;' and ';.', and winepath ignores them completely.
+    func_stripname : : "$1"
+    func_to_host_path_tmp1=$func_stripname_result
+    func_convert_core_path_wine_to_w32 "$func_to_host_path_tmp1"
+    func_cygpath -u -p "$func_convert_core_path_wine_to_w32_result"
+    func_to_host_path_result="$func_cygpath_result"
+    func_convert_path_check : : \
+      "$func_to_host_path_tmp1" "$func_to_host_path_result"
+    func_convert_path_front_back_pathsep ":*" "*:" : "$1"
+  fi
+}
+# end func_convert_path_nix_to_cygwin
+
+
+# func_mode_compile arg...
+func_mode_compile ()
+{
+    $opt_debug
+    # Get the compilation command and the source file.
+    base_compile=
+    srcfile="$nonopt"  #  always keep a non-empty value in "srcfile"
+    suppress_opt=yes
+    suppress_output=
+    arg_mode=normal
+    libobj=
+    later=
+    pie_flag=
+
+    for arg
+    do
+      case $arg_mode in
+      arg  )
+       # do not "continue".  Instead, add this to base_compile
+       lastarg="$arg"
+       arg_mode=normal
+       ;;
+
+      target )
+       libobj="$arg"
+       arg_mode=normal
+       continue
+       ;;
+
+      normal )
+       # Accept any command-line options.
+       case $arg in
+       -o)
+         test -n "$libobj" && \
+           func_fatal_error "you cannot specify \`-o' more than once"
+         arg_mode=target
+         continue
+         ;;
+
+       -pie | -fpie | -fPIE)
+          func_append pie_flag " $arg"
+         continue
+         ;;
+
+       -shared | -static | -prefer-pic | -prefer-non-pic)
+         func_append later " $arg"
+         continue
+         ;;
+
+       -no-suppress)
+         suppress_opt=no
+         continue
+         ;;
+
+       -Xcompiler)
+         arg_mode=arg  #  the next one goes into the "base_compile" arg list
+         continue      #  The current "srcfile" will either be retained or
+         ;;            #  replaced later.  I would guess that would be a bug.
+
+       -Wc,*)
+         func_stripname '-Wc,' '' "$arg"
+         args=$func_stripname_result
+         lastarg=
+         save_ifs="$IFS"; IFS=','
+         for arg in $args; do
+           IFS="$save_ifs"
+           func_append_quoted lastarg "$arg"
+         done
+         IFS="$save_ifs"
+         func_stripname ' ' '' "$lastarg"
+         lastarg=$func_stripname_result
+
+         # Add the arguments to base_compile.
+         func_append base_compile " $lastarg"
+         continue
+         ;;
+
+       *)
+         # Accept the current argument as the source file.
+         # The previous "srcfile" becomes the current argument.
+         #
+         lastarg="$srcfile"
+         srcfile="$arg"
+         ;;
+       esac  #  case $arg
+       ;;
+      esac    #  case $arg_mode
+
+      # Aesthetically quote the previous argument.
+      func_append_quoted base_compile "$lastarg"
+    done # for arg
+
+    case $arg_mode in
+    arg)
+      func_fatal_error "you must specify an argument for -Xcompile"
+      ;;
+    target)
+      func_fatal_error "you must specify a target with \`-o'"
+      ;;
+    *)
+      # Get the name of the library object.
+      test -z "$libobj" && {
+       func_basename "$srcfile"
+       libobj="$func_basename_result"
+      }
+      ;;
+    esac
+
+    # Recognize several different file suffixes.
+    # If the user specifies -o file.o, it is replaced with file.lo
+    case $libobj in
+    *.[cCFSifmso] | \
+    *.ada | *.adb | *.ads | *.asm | \
+    *.c++ | *.cc | *.ii | *.class | *.cpp | *.cxx | \
+    *.[fF][09]? | *.for | *.java | *.go | *.obj | *.sx | *.cu | *.cup)
+      func_xform "$libobj"
+      libobj=$func_xform_result
+      ;;
+    esac
+
+    case $libobj in
+    *.lo) func_lo2o "$libobj"; obj=$func_lo2o_result ;;
+    *)
+      func_fatal_error "cannot determine name of library object from \`$libobj'"
+      ;;
+    esac
+
+    func_infer_tag $base_compile
+
+    for arg in $later; do
+      case $arg in
+      -shared)
+       test "$build_libtool_libs" != yes && \
+         func_fatal_configuration "can not build a shared library"
+       build_old_libs=no
+       continue
+       ;;
+
+      -static)
+       build_libtool_libs=no
+       build_old_libs=yes
+       continue
+       ;;
+
+      -prefer-pic)
+       pic_mode=yes
+       continue
+       ;;
+
+      -prefer-non-pic)
+       pic_mode=no
+       continue
+       ;;
+      esac
+    done
+
+    func_quote_for_eval "$libobj"
+    test "X$libobj" != "X$func_quote_for_eval_result" \
+      && $ECHO "X$libobj" | $GREP '[]~#^*{};<>?"'"'"'   &()|`$[]' \
+      && func_warning "libobj name \`$libobj' may not contain shell special characters."
+    func_dirname_and_basename "$obj" "/" ""
+    objname="$func_basename_result"
+    xdir="$func_dirname_result"
+    lobj=${xdir}$objdir/$objname
+
+    test -z "$base_compile" && \
+      func_fatal_help "you must specify a compilation command"
+
+    # Delete any leftover library objects.
+    if test "$build_old_libs" = yes; then
+      removelist="$obj $lobj $libobj ${libobj}T"
+    else
+      removelist="$lobj $libobj ${libobj}T"
+    fi
+
+    # On Cygwin there's no "real" PIC flag so we must build both object types
+    case $host_os in
+    cygwin* | mingw* | pw32* | os2* | cegcc*)
+      pic_mode=default
+      ;;
+    esac
+    if test "$pic_mode" = no && test "$deplibs_check_method" != pass_all; then
+      # non-PIC code in shared libraries is not supported
+      pic_mode=default
+    fi
+
+    # Calculate the filename of the output object if compiler does
+    # not support -o with -c
+    if test "$compiler_c_o" = no; then
+      output_obj=`$ECHO "$srcfile" | $SED 's%^.*/%%; s%\.[^.]*$%%'`.${objext}
+      lockfile="$output_obj.lock"
+    else
+      output_obj=
+      need_locks=no
+      lockfile=
+    fi
+
+    # Lock this critical section if it is needed
+    # We use this script file to make the link, it avoids creating a new file
+    if test "$need_locks" = yes; then
+      until $opt_dry_run || ln "$progpath" "$lockfile" 2>/dev/null; do
+       func_echo "Waiting for $lockfile to be removed"
+       sleep 2
+      done
+    elif test "$need_locks" = warn; then
+      if test -f "$lockfile"; then
+       $ECHO "\
+*** ERROR, $lockfile exists and contains:
+`cat $lockfile 2>/dev/null`
+
+This indicates that another process is trying to use the same
+temporary object file, and libtool could not work around it because
+your compiler does not support \`-c' and \`-o' together.  If you
+repeat this compilation, it may succeed, by chance, but you had better
+avoid parallel builds (make -j) in this platform, or get a better
+compiler."
+
+       $opt_dry_run || $RM $removelist
+       exit $EXIT_FAILURE
+      fi
+      func_append removelist " $output_obj"
+      $ECHO "$srcfile" > "$lockfile"
+    fi
+
+    $opt_dry_run || $RM $removelist
+    func_append removelist " $lockfile"
+    trap '$opt_dry_run || $RM $removelist; exit $EXIT_FAILURE' 1 2 15
+
+    func_to_tool_file "$srcfile" func_convert_file_msys_to_w32
+    srcfile=$func_to_tool_file_result
+    func_quote_for_eval "$srcfile"
+    qsrcfile=$func_quote_for_eval_result
+
+    # Only build a PIC object if we are building libtool libraries.
+    if test "$build_libtool_libs" = yes; then
+      # Without this assignment, base_compile gets emptied.
+      fbsd_hideous_sh_bug=$base_compile
+
+      if test "$pic_mode" != no; then
+       command="$base_compile $qsrcfile $pic_flag"
+      else
+       # Don't build PIC code
+       command="$base_compile $qsrcfile"
+      fi
+
+      func_mkdir_p "$xdir$objdir"
+
+      if test -z "$output_obj"; then
+       # Place PIC objects in $objdir
+       func_append command " -o $lobj"
+      fi
+
+      func_show_eval_locale "$command" \
+          'test -n "$output_obj" && $RM $removelist; exit $EXIT_FAILURE'
+
+      if test "$need_locks" = warn &&
+        test "X`cat $lockfile 2>/dev/null`" != "X$srcfile"; then
+       $ECHO "\
+*** ERROR, $lockfile contains:
+`cat $lockfile 2>/dev/null`
+
+but it should contain:
+$srcfile
+
+This indicates that another process is trying to use the same
+temporary object file, and libtool could not work around it because
+your compiler does not support \`-c' and \`-o' together.  If you
+repeat this compilation, it may succeed, by chance, but you had better
+avoid parallel builds (make -j) in this platform, or get a better
+compiler."
+
+       $opt_dry_run || $RM $removelist
+       exit $EXIT_FAILURE
+      fi
+
+      # Just move the object if needed, then go on to compile the next one
+      if test -n "$output_obj" && test "X$output_obj" != "X$lobj"; then
+       func_show_eval '$MV "$output_obj" "$lobj"' \
+         'error=$?; $opt_dry_run || $RM $removelist; exit $error'
+      fi
+
+      # Allow error messages only from the first compilation.
+      if test "$suppress_opt" = yes; then
+       suppress_output=' >/dev/null 2>&1'
+      fi
+    fi
+
+    # Only build a position-dependent object if we build old libraries.
+    if test "$build_old_libs" = yes; then
+      if test "$pic_mode" != yes; then
+       # Don't build PIC code
+       command="$base_compile $qsrcfile$pie_flag"
+      else
+       command="$base_compile $qsrcfile $pic_flag"
+      fi
+      if test "$compiler_c_o" = yes; then
+       func_append command " -o $obj"
+      fi
+
+      # Suppress compiler output if we already did a PIC compilation.
+      func_append command "$suppress_output"
+      func_show_eval_locale "$command" \
+        '$opt_dry_run || $RM $removelist; exit $EXIT_FAILURE'
+
+      if test "$need_locks" = warn &&
+        test "X`cat $lockfile 2>/dev/null`" != "X$srcfile"; then
+       $ECHO "\
+*** ERROR, $lockfile contains:
+`cat $lockfile 2>/dev/null`
+
+but it should contain:
+$srcfile
+
+This indicates that another process is trying to use the same
+temporary object file, and libtool could not work around it because
+your compiler does not support \`-c' and \`-o' together.  If you
+repeat this compilation, it may succeed, by chance, but you had better
+avoid parallel builds (make -j) in this platform, or get a better
+compiler."
+
+       $opt_dry_run || $RM $removelist
+       exit $EXIT_FAILURE
+      fi
+
+      # Just move the object if needed
+      if test -n "$output_obj" && test "X$output_obj" != "X$obj"; then
+       func_show_eval '$MV "$output_obj" "$obj"' \
+         'error=$?; $opt_dry_run || $RM $removelist; exit $error'
+      fi
+    fi
+
+    $opt_dry_run || {
+      func_write_libtool_object "$libobj" "$objdir/$objname" "$objname"
+
+      # Unlock the critical section if it was locked
+      if test "$need_locks" != no; then
+       removelist=$lockfile
+        $RM "$lockfile"
+      fi
+    }
+
+    exit $EXIT_SUCCESS
+}
+
+$opt_help || {
+  test "$opt_mode" = compile && func_mode_compile ${1+"$@"}
+}
+
+func_mode_help ()
+{
+    # We need to display help for each of the modes.
+    case $opt_mode in
+      "")
+        # Generic help is extracted from the usage comments
+        # at the start of this file.
+        func_help
+        ;;
+
+      clean)
+        $ECHO \
+"Usage: $progname [OPTION]... --mode=clean RM [RM-OPTION]... FILE...
+
+Remove files from the build directory.
+
+RM is the name of the program to use to delete files associated with each FILE
+(typically \`/bin/rm').  RM-OPTIONS are options (such as \`-f') to be passed
+to RM.
+
+If FILE is a libtool library, object or program, all the files associated
+with it are deleted. Otherwise, only FILE itself is deleted using RM."
+        ;;
+
+      compile)
+      $ECHO \
+"Usage: $progname [OPTION]... --mode=compile COMPILE-COMMAND... SOURCEFILE
+
+Compile a source file into a libtool library object.
+
+This mode accepts the following additional options:
+
+  -o OUTPUT-FILE    set the output file name to OUTPUT-FILE
+  -no-suppress      do not suppress compiler output for multiple passes
+  -prefer-pic       try to build PIC objects only
+  -prefer-non-pic   try to build non-PIC objects only
+  -shared           do not build a \`.o' file suitable for static linking
+  -static           only build a \`.o' file suitable for static linking
+  -Wc,FLAG          pass FLAG directly to the compiler
+
+COMPILE-COMMAND is a command to be used in creating a \`standard' object file
+from the given SOURCEFILE.
+
+The output file name is determined by removing the directory component from
+SOURCEFILE, then substituting the C source code suffix \`.c' with the
+library object suffix, \`.lo'."
+        ;;
+
+      execute)
+        $ECHO \
+"Usage: $progname [OPTION]... --mode=execute COMMAND [ARGS]...
+
+Automatically set library path, then run a program.
+
+This mode accepts the following additional options:
+
+  -dlopen FILE      add the directory containing FILE to the library path
+
+This mode sets the library path environment variable according to \`-dlopen'
+flags.
+
+If any of the ARGS are libtool executable wrappers, then they are translated
+into their corresponding uninstalled binary, and any of their required library
+directories are added to the library path.
+
+Then, COMMAND is executed, with ARGS as arguments."
+        ;;
+
+      finish)
+        $ECHO \
+"Usage: $progname [OPTION]... --mode=finish [LIBDIR]...
+
+Complete the installation of libtool libraries.
+
+Each LIBDIR is a directory that contains libtool libraries.
+
+The commands that this mode executes may require superuser privileges.  Use
+the \`--dry-run' option if you just want to see what would be executed."
+        ;;
+
+      install)
+        $ECHO \
+"Usage: $progname [OPTION]... --mode=install INSTALL-COMMAND...
+
+Install executables or libraries.
+
+INSTALL-COMMAND is the installation command.  The first component should be
+either the \`install' or \`cp' program.
+
+The following components of INSTALL-COMMAND are treated specially:
+
+  -inst-prefix-dir PREFIX-DIR  Use PREFIX-DIR as a staging area for installation
+
+The rest of the components are interpreted as arguments to that command (only
+BSD-compatible install options are recognized)."
+        ;;
+
+      link)
+        $ECHO \
+"Usage: $progname [OPTION]... --mode=link LINK-COMMAND...
+
+Link object files or libraries together to form another library, or to
+create an executable program.
+
+LINK-COMMAND is a command using the C compiler that you would use to create
+a program from several object files.
+
+The following components of LINK-COMMAND are treated specially:
+
+  -all-static       do not do any dynamic linking at all
+  -avoid-version    do not add a version suffix if possible
+  -bindir BINDIR    specify path to binaries directory (for systems where
+                    libraries must be found in the PATH setting at runtime)
+  -dlopen FILE      \`-dlpreopen' FILE if it cannot be dlopened at runtime
+  -dlpreopen FILE   link in FILE and add its symbols to lt_preloaded_symbols
+  -export-dynamic   allow symbols from OUTPUT-FILE to be resolved with dlsym(3)
+  -export-symbols SYMFILE
+                    try to export only the symbols listed in SYMFILE
+  -export-symbols-regex REGEX
+                    try to export only the symbols matching REGEX
+  -LLIBDIR          search LIBDIR for required installed libraries
+  -lNAME            OUTPUT-FILE requires the installed library libNAME
+  -module           build a library that can dlopened
+  -no-fast-install  disable the fast-install mode
+  -no-install       link a not-installable executable
+  -no-undefined     declare that a library does not refer to external symbols
+  -o OUTPUT-FILE    create OUTPUT-FILE from the specified objects
+  -objectlist FILE  Use a list of object files found in FILE to specify objects
+  -precious-files-regex REGEX
+                    don't remove output files matching REGEX
+  -release RELEASE  specify package release information
+  -rpath LIBDIR     the created library will eventually be installed in LIBDIR
+  -R[ ]LIBDIR       add LIBDIR to the runtime path of programs and libraries
+  -shared           only do dynamic linking of libtool libraries
+  -shrext SUFFIX    override the standard shared library file extension
+  -static           do not do any dynamic linking of uninstalled libtool libraries
+  -static-libtool-libs
+                    do not do any dynamic linking of libtool libraries
+  -version-info CURRENT[:REVISION[:AGE]]
+                    specify library version info [each variable defaults to 0]
+  -weak LIBNAME     declare that the target provides the LIBNAME interface
+  -Wc,FLAG
+  -Xcompiler FLAG   pass linker-specific FLAG directly to the compiler
+  -Wl,FLAG
+  -Xlinker FLAG     pass linker-specific FLAG directly to the linker
+  -XCClinker FLAG   pass link-specific FLAG to the compiler driver (CC)
+
+All other options (arguments beginning with \`-') are ignored.
+
+Every other argument is treated as a filename.  Files ending in \`.la' are
+treated as uninstalled libtool libraries, other files are standard or library
+object files.
+
+If the OUTPUT-FILE ends in \`.la', then a libtool library is created,
+only library objects (\`.lo' files) may be specified, and \`-rpath' is
+required, except when creating a convenience library.
+
+If OUTPUT-FILE ends in \`.a' or \`.lib', then a standard library is created
+using \`ar' and \`ranlib', or on Windows using \`lib'.
+
+If OUTPUT-FILE ends in \`.lo' or \`.${objext}', then a reloadable object file
+is created, otherwise an executable program is created."
+        ;;
+
+      uninstall)
+        $ECHO \
+"Usage: $progname [OPTION]... --mode=uninstall RM [RM-OPTION]... FILE...
+
+Remove libraries from an installation directory.
+
+RM is the name of the program to use to delete files associated with each FILE
+(typically \`/bin/rm').  RM-OPTIONS are options (such as \`-f') to be passed
+to RM.
+
+If FILE is a libtool library, all the files associated with it are deleted.
+Otherwise, only FILE itself is deleted using RM."
+        ;;
+
+      *)
+        func_fatal_help "invalid operation mode \`$opt_mode'"
+        ;;
+    esac
+
+    echo
+    $ECHO "Try \`$progname --help' for more information about other modes."
+}
+
+# Now that we've collected a possible --mode arg, show help if necessary
+if $opt_help; then
+  if test "$opt_help" = :; then
+    func_mode_help
+  else
+    {
+      func_help noexit
+      for opt_mode in compile link execute install finish uninstall clean; do
+       func_mode_help
+      done
+    } | sed -n '1p; 2,$s/^Usage:/  or: /p'
+    {
+      func_help noexit
+      for opt_mode in compile link execute install finish uninstall clean; do
+       echo
+       func_mode_help
+      done
+    } |
+    sed '1d
+      /^When reporting/,/^Report/{
+       H
+       d
+      }
+      $x
+      /information about other modes/d
+      /more detailed .*MODE/d
+      s/^Usage:.*--mode=\([^ ]*\) .*/Description of \1 mode:/'
+  fi
+  exit $?
+fi
+
+
+# func_mode_execute arg...
+func_mode_execute ()
+{
+    $opt_debug
+    # The first argument is the command name.
+    cmd="$nonopt"
+    test -z "$cmd" && \
+      func_fatal_help "you must specify a COMMAND"
+
+    # Handle -dlopen flags immediately.
+    for file in $opt_dlopen; do
+      test -f "$file" \
+       || func_fatal_help "\`$file' is not a file"
+
+      dir=
+      case $file in
+      *.la)
+       func_resolve_sysroot "$file"
+       file=$func_resolve_sysroot_result
+
+       # Check to see that this really is a libtool archive.
+       func_lalib_unsafe_p "$file" \
+         || func_fatal_help "\`$lib' is not a valid libtool archive"
+
+       # Read the libtool library.
+       dlname=
+       library_names=
+       func_source "$file"
+
+       # Skip this library if it cannot be dlopened.
+       if test -z "$dlname"; then
+         # Warn if it was a shared library.
+         test -n "$library_names" && \
+           func_warning "\`$file' was not linked with \`-export-dynamic'"
+         continue
+       fi
+
+       func_dirname "$file" "" "."
+       dir="$func_dirname_result"
+
+       if test -f "$dir/$objdir/$dlname"; then
+         func_append dir "/$objdir"
+       else
+         if test ! -f "$dir/$dlname"; then
+           func_fatal_error "cannot find \`$dlname' in \`$dir' or \`$dir/$objdir'"
+         fi
+       fi
+       ;;
+
+      *.lo)
+       # Just add the directory containing the .lo file.
+       func_dirname "$file" "" "."
+       dir="$func_dirname_result"
+       ;;
+
+      *)
+       func_warning "\`-dlopen' is ignored for non-libtool libraries and objects"
+       continue
+       ;;
+      esac
+
+      # Get the absolute pathname.
+      absdir=`cd "$dir" && pwd`
+      test -n "$absdir" && dir="$absdir"
+
+      # Now add the directory to shlibpath_var.
+      if eval "test -z \"\$$shlibpath_var\""; then
+       eval "$shlibpath_var=\"\$dir\""
+      else
+       eval "$shlibpath_var=\"\$dir:\$$shlibpath_var\""
+      fi
+    done
+
+    # This variable tells wrapper scripts just to set shlibpath_var
+    # rather than running their programs.
+    libtool_execute_magic="$magic"
+
+    # Check if any of the arguments is a wrapper script.
+    args=
+    for file
+    do
+      case $file in
+      -* | *.la | *.lo ) ;;
+      *)
+       # Do a test to see if this is really a libtool program.
+       if func_ltwrapper_script_p "$file"; then
+         func_source "$file"
+         # Transform arg to wrapped name.
+         file="$progdir/$program"
+       elif func_ltwrapper_executable_p "$file"; then
+         func_ltwrapper_scriptname "$file"
+         func_source "$func_ltwrapper_scriptname_result"
+         # Transform arg to wrapped name.
+         file="$progdir/$program"
+       fi
+       ;;
+      esac
+      # Quote arguments (to preserve shell metacharacters).
+      func_append_quoted args "$file"
+    done
+
+    if test "X$opt_dry_run" = Xfalse; then
+      if test -n "$shlibpath_var"; then
+       # Export the shlibpath_var.
+       eval "export $shlibpath_var"
+      fi
+
+      # Restore saved environment variables
+      for lt_var in LANG LANGUAGE LC_ALL LC_CTYPE LC_COLLATE LC_MESSAGES
+      do
+       eval "if test \"\${save_$lt_var+set}\" = set; then
+                $lt_var=\$save_$lt_var; export $lt_var
+             else
+               $lt_unset $lt_var
+             fi"
+      done
+
+      # Now prepare to actually exec the command.
+      exec_cmd="\$cmd$args"
+    else
+      # Display what would be done.
+      if test -n "$shlibpath_var"; then
+       eval "\$ECHO \"\$shlibpath_var=\$$shlibpath_var\""
+       echo "export $shlibpath_var"
+      fi
+      $ECHO "$cmd$args"
+      exit $EXIT_SUCCESS
+    fi
+}
+
+test "$opt_mode" = execute && func_mode_execute ${1+"$@"}
+
+
+# func_mode_finish arg...
+func_mode_finish ()
+{
+    $opt_debug
+    libs=
+    libdirs=
+    admincmds=
+
+    for opt in "$nonopt" ${1+"$@"}
+    do
+      if test -d "$opt"; then
+       func_append libdirs " $opt"
+
+      elif test -f "$opt"; then
+       if func_lalib_unsafe_p "$opt"; then
+         func_append libs " $opt"
+       else
+         func_warning "\`$opt' is not a valid libtool archive"
+       fi
+
+      else
+       func_fatal_error "invalid argument \`$opt'"
+      fi
+    done
+
+    if test -n "$libs"; then
+      if test -n "$lt_sysroot"; then
+        sysroot_regex=`$ECHO "$lt_sysroot" | $SED "$sed_make_literal_regex"`
+        sysroot_cmd="s/\([ ']\)$sysroot_regex/\1/g;"
+      else
+        sysroot_cmd=
+      fi
+
+      # Remove sysroot references
+      if $opt_dry_run; then
+        for lib in $libs; do
+          echo "removing references to $lt_sysroot and \`=' prefixes from $lib"
+        done
+      else
+        tmpdir=`func_mktempdir`
+        for lib in $libs; do
+         sed -e "${sysroot_cmd} s/\([ ']-[LR]\)=/\1/g; s/\([ ']\)=/\1/g" $lib \
+           > $tmpdir/tmp-la
+         mv -f $tmpdir/tmp-la $lib
+       done
+        ${RM}r "$tmpdir"
+      fi
+    fi
+
+    if test -n "$finish_cmds$finish_eval" && test -n "$libdirs"; then
+      for libdir in $libdirs; do
+       if test -n "$finish_cmds"; then
+         # Do each command in the finish commands.
+         func_execute_cmds "$finish_cmds" 'admincmds="$admincmds
+'"$cmd"'"'
+       fi
+       if test -n "$finish_eval"; then
+         # Do the single finish_eval.
+         eval cmds=\"$finish_eval\"
+         $opt_dry_run || eval "$cmds" || func_append admincmds "
+       $cmds"
+       fi
+      done
+    fi
+
+    # Exit here if they wanted silent mode.
+    $opt_silent && exit $EXIT_SUCCESS
+
+    if test -n "$finish_cmds$finish_eval" && test -n "$libdirs"; then
+      echo "----------------------------------------------------------------------"
+      echo "Libraries have been installed in:"
+      for libdir in $libdirs; do
+       $ECHO "   $libdir"
+      done
+      echo
+      echo "If you ever happen to want to link against installed libraries"
+      echo "in a given directory, LIBDIR, you must either use libtool, and"
+      echo "specify the full pathname of the library, or use the \`-LLIBDIR'"
+      echo "flag during linking and do at least one of the following:"
+      if test -n "$shlibpath_var"; then
+       echo "   - add LIBDIR to the \`$shlibpath_var' environment variable"
+       echo "     during execution"
+      fi
+      if test -n "$runpath_var"; then
+       echo "   - add LIBDIR to the \`$runpath_var' environment variable"
+       echo "     during linking"
+      fi
+      if test -n "$hardcode_libdir_flag_spec"; then
+       libdir=LIBDIR
+       eval flag=\"$hardcode_libdir_flag_spec\"
+
+       $ECHO "   - use the \`$flag' linker flag"
+      fi
+      if test -n "$admincmds"; then
+       $ECHO "   - have your system administrator run these commands:$admincmds"
+      fi
+      if test -f /etc/ld.so.conf; then
+       echo "   - have your system administrator add LIBDIR to \`/etc/ld.so.conf'"
+      fi
+      echo
+
+      echo "See any operating system documentation about shared libraries for"
+      case $host in
+       solaris2.[6789]|solaris2.1[0-9])
+         echo "more information, such as the ld(1), crle(1) and ld.so(8) manual"
+         echo "pages."
+         ;;
+       *)
+         echo "more information, such as the ld(1) and ld.so(8) manual pages."
+         ;;
+      esac
+      echo "----------------------------------------------------------------------"
+    fi
+    exit $EXIT_SUCCESS
+}
+
+test "$opt_mode" = finish && func_mode_finish ${1+"$@"}
+
+
+# func_mode_install arg...
+func_mode_install ()
+{
+    $opt_debug
+    # There may be an optional sh(1) argument at the beginning of
+    # install_prog (especially on Windows NT).
+    if test "$nonopt" = "$SHELL" || test "$nonopt" = /bin/sh ||
+       # Allow the use of GNU shtool's install command.
+       case $nonopt in *shtool*) :;; *) false;; esac; then
+      # Aesthetically quote it.
+      func_quote_for_eval "$nonopt"
+      install_prog="$func_quote_for_eval_result "
+      arg=$1
+      shift
+    else
+      install_prog=
+      arg=$nonopt
+    fi
+
+    # The real first argument should be the name of the installation program.
+    # Aesthetically quote it.
+    func_quote_for_eval "$arg"
+    func_append install_prog "$func_quote_for_eval_result"
+    install_shared_prog=$install_prog
+    case " $install_prog " in
+      *[\\\ /]cp\ *) install_cp=: ;;
+      *) install_cp=false ;;
+    esac
+
+    # We need to accept at least all the BSD install flags.
+    dest=
+    files=
+    opts=
+    prev=
+    install_type=
+    isdir=no
+    stripme=
+    no_mode=:
+    for arg
+    do
+      arg2=
+      if test -n "$dest"; then
+       func_append files " $dest"
+       dest=$arg
+       continue
+      fi
+
+      case $arg in
+      -d) isdir=yes ;;
+      -f)
+       if $install_cp; then :; else
+         prev=$arg
+       fi
+       ;;
+      -g | -m | -o)
+       prev=$arg
+       ;;
+      -s)
+       stripme=" -s"
+       continue
+       ;;
+      -*)
+       ;;
+      *)
+       # If the previous option needed an argument, then skip it.
+       if test -n "$prev"; then
+         if test "x$prev" = x-m && test -n "$install_override_mode"; then
+           arg2=$install_override_mode
+           no_mode=false
+         fi
+         prev=
+       else
+         dest=$arg
+         continue
+       fi
+       ;;
+      esac
+
+      # Aesthetically quote the argument.
+      func_quote_for_eval "$arg"
+      func_append install_prog " $func_quote_for_eval_result"
+      if test -n "$arg2"; then
+       func_quote_for_eval "$arg2"
+      fi
+      func_append install_shared_prog " $func_quote_for_eval_result"
+    done
+
+    test -z "$install_prog" && \
+      func_fatal_help "you must specify an install program"
+
+    test -n "$prev" && \
+      func_fatal_help "the \`$prev' option requires an argument"
+
+    if test -n "$install_override_mode" && $no_mode; then
+      if $install_cp; then :; else
+       func_quote_for_eval "$install_override_mode"
+       func_append install_shared_prog " -m $func_quote_for_eval_result"
+      fi
+    fi
+
+    if test -z "$files"; then
+      if test -z "$dest"; then
+       func_fatal_help "no file or destination specified"
+      else
+       func_fatal_help "you must specify a destination"
+      fi
+    fi
+
+    # Strip any trailing slash from the destination.
+    func_stripname '' '/' "$dest"
+    dest=$func_stripname_result
+
+    # Check to see that the destination is a directory.
+    test -d "$dest" && isdir=yes
+    if test "$isdir" = yes; then
+      destdir="$dest"
+      destname=
+    else
+      func_dirname_and_basename "$dest" "" "."
+      destdir="$func_dirname_result"
+      destname="$func_basename_result"
+
+      # Not a directory, so check to see that there is only one file specified.
+      set dummy $files; shift
+      test "$#" -gt 1 && \
+       func_fatal_help "\`$dest' is not a directory"
+    fi
+    case $destdir in
+    [\\/]* | [A-Za-z]:[\\/]*) ;;
+    *)
+      for file in $files; do
+       case $file in
+       *.lo) ;;
+       *)
+         func_fatal_help "\`$destdir' must be an absolute directory name"
+         ;;
+       esac
+      done
+      ;;
+    esac
+
+    # This variable tells wrapper scripts just to set variables rather
+    # than running their programs.
+    libtool_install_magic="$magic"
+
+    staticlibs=
+    future_libdirs=
+    current_libdirs=
+    for file in $files; do
+
+      # Do each installation.
+      case $file in
+      *.$libext)
+       # Do the static libraries later.
+       func_append staticlibs " $file"
+       ;;
+
+      *.la)
+       func_resolve_sysroot "$file"
+       file=$func_resolve_sysroot_result
+
+       # Check to see that this really is a libtool archive.
+       func_lalib_unsafe_p "$file" \
+         || func_fatal_help "\`$file' is not a valid libtool archive"
+
+       library_names=
+       old_library=
+       relink_command=
+       func_source "$file"
+
+       # Add the libdir to current_libdirs if it is the destination.
+       if test "X$destdir" = "X$libdir"; then
+         case "$current_libdirs " in
+         *" $libdir "*) ;;
+         *) func_append current_libdirs " $libdir" ;;
+         esac
+       else
+         # Note the libdir as a future libdir.
+         case "$future_libdirs " in
+         *" $libdir "*) ;;
+         *) func_append future_libdirs " $libdir" ;;
+         esac
+       fi
+
+       func_dirname "$file" "/" ""
+       dir="$func_dirname_result"
+       func_append dir "$objdir"
+
+       if test -n "$relink_command"; then
+         # Determine the prefix the user has applied to our future dir.
+         inst_prefix_dir=`$ECHO "$destdir" | $SED -e "s%$libdir\$%%"`
+
+         # Don't allow the user to place us outside of our expected
+         # location b/c this prevents finding dependent libraries that
+         # are installed to the same prefix.
+         # At present, this check doesn't affect windows .dll's that
+         # are installed into $libdir/../bin (currently, that works fine)
+         # but it's something to keep an eye on.
+         test "$inst_prefix_dir" = "$destdir" && \
+           func_fatal_error "error: cannot install \`$file' to a directory not ending in $libdir"
+
+         if test -n "$inst_prefix_dir"; then
+           # Stick the inst_prefix_dir data into the link command.
+           relink_command=`$ECHO "$relink_command" | $SED "s%@inst_prefix_dir@%-inst-prefix-dir $inst_prefix_dir%"`
+         else
+           relink_command=`$ECHO "$relink_command" | $SED "s%@inst_prefix_dir@%%"`
+         fi
+
+         func_warning "relinking \`$file'"
+         func_show_eval "$relink_command" \
+           'func_fatal_error "error: relink \`$file'\'' with the above command before installing it"'
+       fi
+
+       # See the names of the shared library.
+       set dummy $library_names; shift
+       if test -n "$1"; then
+         realname="$1"
+         shift
+
+         srcname="$realname"
+         test -n "$relink_command" && srcname="$realname"T
+
+         # Install the shared library and build the symlinks.
+         func_show_eval "$install_shared_prog $dir/$srcname $destdir/$realname" \
+             'exit $?'
+         tstripme="$stripme"
+         case $host_os in
+         cygwin* | mingw* | pw32* | cegcc*)
+           case $realname in
+           *.dll.a)
+             tstripme=""
+             ;;
+           esac
+           ;;
+         esac
+         if test -n "$tstripme" && test -n "$striplib"; then
+           func_show_eval "$striplib $destdir/$realname" 'exit $?'
+         fi
+
+         if test "$#" -gt 0; then
+           # Delete the old symlinks, and create new ones.
+           # Try `ln -sf' first, because the `ln' binary might depend on
+           # the symlink we replace!  Solaris /bin/ln does not understand -f,
+           # so we also need to try rm && ln -s.
+           for linkname
+           do
+             test "$linkname" != "$realname" \
+               && func_show_eval "(cd $destdir && { $LN_S -f $realname $linkname || { $RM $linkname && $LN_S $realname $linkname; }; })"
+           done
+         fi
+
+         # Do each command in the postinstall commands.
+         lib="$destdir/$realname"
+         func_execute_cmds "$postinstall_cmds" 'exit $?'
+       fi
+
+       # Install the pseudo-library for information purposes.
+       func_basename "$file"
+       name="$func_basename_result"
+       instname="$dir/$name"i
+       func_show_eval "$install_prog $instname $destdir/$name" 'exit $?'
+
+       # Maybe install the static library, too.
+       test -n "$old_library" && func_append staticlibs " $dir/$old_library"
+       ;;
+
+      *.lo)
+       # Install (i.e. copy) a libtool object.
+
+       # Figure out destination file name, if it wasn't already specified.
+       if test -n "$destname"; then
+         destfile="$destdir/$destname"
+       else
+         func_basename "$file"
+         destfile="$func_basename_result"
+         destfile="$destdir/$destfile"
+       fi
+
+       # Deduce the name of the destination old-style object file.
+       case $destfile in
+       *.lo)
+         func_lo2o "$destfile"
+         staticdest=$func_lo2o_result
+         ;;
+       *.$objext)
+         staticdest="$destfile"
+         destfile=
+         ;;
+       *)
+         func_fatal_help "cannot copy a libtool object to \`$destfile'"
+         ;;
+       esac
+
+       # Install the libtool object if requested.
+       test -n "$destfile" && \
+         func_show_eval "$install_prog $file $destfile" 'exit $?'
+
+       # Install the old object if enabled.
+       if test "$build_old_libs" = yes; then
+         # Deduce the name of the old-style object file.
+         func_lo2o "$file"
+         staticobj=$func_lo2o_result
+         func_show_eval "$install_prog \$staticobj \$staticdest" 'exit $?'
+       fi
+       exit $EXIT_SUCCESS
+       ;;
+
+      *)
+       # Figure out destination file name, if it wasn't already specified.
+       if test -n "$destname"; then
+         destfile="$destdir/$destname"
+       else
+         func_basename "$file"
+         destfile="$func_basename_result"
+         destfile="$destdir/$destfile"
+       fi
+
+       # If the file is missing, and there is a .exe on the end, strip it
+       # because it is most likely a libtool script we actually want to
+       # install
+       stripped_ext=""
+       case $file in
+         *.exe)
+           if test ! -f "$file"; then
+             func_stripname '' '.exe' "$file"
+             file=$func_stripname_result
+             stripped_ext=".exe"
+           fi
+           ;;
+       esac
+
+       # Do a test to see if this is really a libtool program.
+       case $host in
+       *cygwin* | *mingw*)
+           if func_ltwrapper_executable_p "$file"; then
+             func_ltwrapper_scriptname "$file"
+             wrapper=$func_ltwrapper_scriptname_result
+           else
+             func_stripname '' '.exe' "$file"
+             wrapper=$func_stripname_result
+           fi
+           ;;
+       *)
+           wrapper=$file
+           ;;
+       esac
+       if func_ltwrapper_script_p "$wrapper"; then
+         notinst_deplibs=
+         relink_command=
+
+         func_source "$wrapper"
+
+         # Check the variables that should have been set.
+         test -z "$generated_by_libtool_version" && \
+           func_fatal_error "invalid libtool wrapper script \`$wrapper'"
+
+         finalize=yes
+         for lib in $notinst_deplibs; do
+           # Check to see that each library is installed.
+           libdir=
+           if test -f "$lib"; then
+             func_source "$lib"
+           fi
+           libfile="$libdir/"`$ECHO "$lib" | $SED 's%^.*/%%g'` ### testsuite: skip nested quoting test
+           if test -n "$libdir" && test ! -f "$libfile"; then
+             func_warning "\`$lib' has not been installed in \`$libdir'"
+             finalize=no
+           fi
+         done
+
+         relink_command=
+         func_source "$wrapper"
+
+         outputname=
+         if test "$fast_install" = no && test -n "$relink_command"; then
+           $opt_dry_run || {
+             if test "$finalize" = yes; then
+               tmpdir=`func_mktempdir`
+               func_basename "$file$stripped_ext"
+               file="$func_basename_result"
+               outputname="$tmpdir/$file"
+               # Replace the output file specification.
+               relink_command=`$ECHO "$relink_command" | $SED 's%@OUTPUT@%'"$outputname"'%g'`
+
+               $opt_silent || {
+                 func_quote_for_expand "$relink_command"
+                 eval "func_echo $func_quote_for_expand_result"
+               }
+               if eval "$relink_command"; then :
+                 else
+                 func_error "error: relink \`$file' with the above command before installing it"
+                 $opt_dry_run || ${RM}r "$tmpdir"
+                 continue
+               fi
+               file="$outputname"
+             else
+               func_warning "cannot relink \`$file'"
+             fi
+           }
+         else
+           # Install the binary that we compiled earlier.
+           file=`$ECHO "$file$stripped_ext" | $SED "s%\([^/]*\)$%$objdir/\1%"`
+         fi
+       fi
+
+       # remove .exe since cygwin /usr/bin/install will append another
+       # one anyway
+       case $install_prog,$host in
+       */usr/bin/install*,*cygwin*)
+         case $file:$destfile in
+         *.exe:*.exe)
+           # this is ok
+           ;;
+         *.exe:*)
+           destfile=$destfile.exe
+           ;;
+         *:*.exe)
+           func_stripname '' '.exe' "$destfile"
+           destfile=$func_stripname_result
+           ;;
+         esac
+         ;;
+       esac
+       func_show_eval "$install_prog\$stripme \$file \$destfile" 'exit $?'
+       $opt_dry_run || if test -n "$outputname"; then
+         ${RM}r "$tmpdir"
+       fi
+       ;;
+      esac
+    done
+
+    for file in $staticlibs; do
+      func_basename "$file"
+      name="$func_basename_result"
+
+      # Set up the ranlib parameters.
+      oldlib="$destdir/$name"
+      func_to_tool_file "$oldlib" func_convert_file_msys_to_w32
+      tool_oldlib=$func_to_tool_file_result
+
+      func_show_eval "$install_prog \$file \$oldlib" 'exit $?'
+
+      if test -n "$stripme" && test -n "$old_striplib"; then
+       func_show_eval "$old_striplib $tool_oldlib" 'exit $?'
+      fi
+
+      # Do each command in the postinstall commands.
+      func_execute_cmds "$old_postinstall_cmds" 'exit $?'
+    done
+
+    test -n "$future_libdirs" && \
+      func_warning "remember to run \`$progname --finish$future_libdirs'"
+
+    if test -n "$current_libdirs"; then
+      # Maybe just do a dry run.
+      $opt_dry_run && current_libdirs=" -n$current_libdirs"
+      exec_cmd='$SHELL $progpath $preserve_args --finish$current_libdirs'
+    else
+      exit $EXIT_SUCCESS
+    fi
+}
+
+test "$opt_mode" = install && func_mode_install ${1+"$@"}
+
+
+# func_generate_dlsyms outputname originator pic_p
+# Extract symbols from dlprefiles and create ${outputname}S.o with
+# a dlpreopen symbol table.
+func_generate_dlsyms ()
+{
+    $opt_debug
+    my_outputname="$1"
+    my_originator="$2"
+    my_pic_p="${3-no}"
+    my_prefix=`$ECHO "$my_originator" | sed 's%[^a-zA-Z0-9]%_%g'`
+    my_dlsyms=
+
+    if test -n "$dlfiles$dlprefiles" || test "$dlself" != no; then
+      if test -n "$NM" && test -n "$global_symbol_pipe"; then
+       my_dlsyms="${my_outputname}S.c"
+      else
+       func_error "not configured to extract global symbols from dlpreopened files"
+      fi
+    fi
+
+    if test -n "$my_dlsyms"; then
+      case $my_dlsyms in
+      "") ;;
+      *.c)
+       # Discover the nlist of each of the dlfiles.
+       nlist="$output_objdir/${my_outputname}.nm"
+
+       func_show_eval "$RM $nlist ${nlist}S ${nlist}T"
+
+       # Parse the name list into a source file.
+       func_verbose "creating $output_objdir/$my_dlsyms"
+
+       $opt_dry_run || $ECHO > "$output_objdir/$my_dlsyms" "\
+/* $my_dlsyms - symbol resolution table for \`$my_outputname' dlsym emulation. */
+/* Generated by $PROGRAM (GNU $PACKAGE$TIMESTAMP) $VERSION */
+
+#ifdef __cplusplus
+extern \"C\" {
+#endif
+
+#if defined(__GNUC__) && (((__GNUC__ == 4) && (__GNUC_MINOR__ >= 4)) || (__GNUC__ > 4))
+#pragma GCC diagnostic ignored \"-Wstrict-prototypes\"
+#endif
+
+/* Keep this code in sync between libtool.m4, ltmain, lt_system.h, and tests.  */
+#if defined(_WIN32) || defined(__CYGWIN__) || defined(_WIN32_WCE)
+/* DATA imports from DLLs on WIN32 con't be const, because runtime
+   relocations are performed -- see ld's documentation on pseudo-relocs.  */
+# define LT_DLSYM_CONST
+#elif defined(__osf__)
+/* This system does not cope well with relocations in const data.  */
+# define LT_DLSYM_CONST
+#else
+# define LT_DLSYM_CONST const
+#endif
+
+/* External symbol declarations for the compiler. */\
+"
+
+       if test "$dlself" = yes; then
+         func_verbose "generating symbol list for \`$output'"
+
+         $opt_dry_run || echo ': @PROGRAM@ ' > "$nlist"
+
+         # Add our own program objects to the symbol list.
+         progfiles=`$ECHO "$objs$old_deplibs" | $SP2NL | $SED "$lo2o" | $NL2SP`
+         for progfile in $progfiles; do
+           func_to_tool_file "$progfile" func_convert_file_msys_to_w32
+           func_verbose "extracting global C symbols from \`$func_to_tool_file_result'"
+           $opt_dry_run || eval "$NM $func_to_tool_file_result | $global_symbol_pipe >> '$nlist'"
+         done
+
+         if test -n "$exclude_expsyms"; then
+           $opt_dry_run || {
+             eval '$EGREP -v " ($exclude_expsyms)$" "$nlist" > "$nlist"T'
+             eval '$MV "$nlist"T "$nlist"'
+           }
+         fi
+
+         if test -n "$export_symbols_regex"; then
+           $opt_dry_run || {
+             eval '$EGREP -e "$export_symbols_regex" "$nlist" > "$nlist"T'
+             eval '$MV "$nlist"T "$nlist"'
+           }
+         fi
+
+         # Prepare the list of exported symbols
+         if test -z "$export_symbols"; then
+           export_symbols="$output_objdir/$outputname.exp"
+           $opt_dry_run || {
+             $RM $export_symbols
+             eval "${SED} -n -e '/^: @PROGRAM@ $/d' -e 's/^.* \(.*\)$/\1/p' "'< "$nlist" > "$export_symbols"'
+             case $host in
+             *cygwin* | *mingw* | *cegcc* )
+                eval "echo EXPORTS "'> "$output_objdir/$outputname.def"'
+                eval 'cat "$export_symbols" >> "$output_objdir/$outputname.def"'
+               ;;
+             esac
+           }
+         else
+           $opt_dry_run || {
+             eval "${SED} -e 's/\([].[*^$]\)/\\\\\1/g' -e 's/^/ /' -e 's/$/$/'"' < "$export_symbols" > "$output_objdir/$outputname.exp"'
+             eval '$GREP -f "$output_objdir/$outputname.exp" < "$nlist" > "$nlist"T'
+             eval '$MV "$nlist"T "$nlist"'
+             case $host in
+               *cygwin* | *mingw* | *cegcc* )
+                 eval "echo EXPORTS "'> "$output_objdir/$outputname.def"'
+                 eval 'cat "$nlist" >> "$output_objdir/$outputname.def"'
+                 ;;
+             esac
+           }
+         fi
+       fi
+
+       for dlprefile in $dlprefiles; do
+         func_verbose "extracting global C symbols from \`$dlprefile'"
+         func_basename "$dlprefile"
+         name="$func_basename_result"
+          case $host in
+           *cygwin* | *mingw* | *cegcc* )
+             # if an import library, we need to obtain dlname
+             if func_win32_import_lib_p "$dlprefile"; then
+               func_tr_sh "$dlprefile"
+               eval "curr_lafile=\$libfile_$func_tr_sh_result"
+               dlprefile_dlbasename=""
+               if test -n "$curr_lafile" && func_lalib_p "$curr_lafile"; then
+                 # Use subshell, to avoid clobbering current variable values
+                 dlprefile_dlname=`source "$curr_lafile" && echo "$dlname"`
+                 if test -n "$dlprefile_dlname" ; then
+                   func_basename "$dlprefile_dlname"
+                   dlprefile_dlbasename="$func_basename_result"
+                 else
+                   # no lafile. user explicitly requested -dlpreopen <import library>.
+                   $sharedlib_from_linklib_cmd "$dlprefile"
+                   dlprefile_dlbasename=$sharedlib_from_linklib_result
+                 fi
+               fi
+               $opt_dry_run || {
+                 if test -n "$dlprefile_dlbasename" ; then
+                   eval '$ECHO ": $dlprefile_dlbasename" >> "$nlist"'
+                 else
+                   func_warning "Could not compute DLL name from $name"
+                   eval '$ECHO ": $name " >> "$nlist"'
+                 fi
+                 func_to_tool_file "$dlprefile" func_convert_file_msys_to_w32
+                 eval "$NM \"$func_to_tool_file_result\" 2>/dev/null | $global_symbol_pipe |
+                   $SED -e '/I __imp/d' -e 's/I __nm_/D /;s/_nm__//' >> '$nlist'"
+               }
+             else # not an import lib
+               $opt_dry_run || {
+                 eval '$ECHO ": $name " >> "$nlist"'
+                 func_to_tool_file "$dlprefile" func_convert_file_msys_to_w32
+                 eval "$NM \"$func_to_tool_file_result\" 2>/dev/null | $global_symbol_pipe >> '$nlist'"
+               }
+             fi
+           ;;
+           *)
+             $opt_dry_run || {
+               eval '$ECHO ": $name " >> "$nlist"'
+               func_to_tool_file "$dlprefile" func_convert_file_msys_to_w32
+               eval "$NM \"$func_to_tool_file_result\" 2>/dev/null | $global_symbol_pipe >> '$nlist'"
+             }
+           ;;
+          esac
+       done
+
+       $opt_dry_run || {
+         # Make sure we have at least an empty file.
+         test -f "$nlist" || : > "$nlist"
+
+         if test -n "$exclude_expsyms"; then
+           $EGREP -v " ($exclude_expsyms)$" "$nlist" > "$nlist"T
+           $MV "$nlist"T "$nlist"
+         fi
+
+         # Try sorting and uniquifying the output.
+         if $GREP -v "^: " < "$nlist" |
+             if sort -k 3 </dev/null >/dev/null 2>&1; then
+               sort -k 3
+             else
+               sort +2
+             fi |
+             uniq > "$nlist"S; then
+           :
+         else
+           $GREP -v "^: " < "$nlist" > "$nlist"S
+         fi
+
+         if test -f "$nlist"S; then
+           eval "$global_symbol_to_cdecl"' < "$nlist"S >> "$output_objdir/$my_dlsyms"'
+         else
+           echo '/* NONE */' >> "$output_objdir/$my_dlsyms"
+         fi
+
+         echo >> "$output_objdir/$my_dlsyms" "\
+
+/* The mapping between symbol names and symbols.  */
+typedef struct {
+  const char *name;
+  void *address;
+} lt_dlsymlist;
+extern LT_DLSYM_CONST lt_dlsymlist
+lt_${my_prefix}_LTX_preloaded_symbols[];
+LT_DLSYM_CONST lt_dlsymlist
+lt_${my_prefix}_LTX_preloaded_symbols[] =
+{\
+  { \"$my_originator\", (void *) 0 },"
+
+         case $need_lib_prefix in
+         no)
+           eval "$global_symbol_to_c_name_address" < "$nlist" >> "$output_objdir/$my_dlsyms"
+           ;;
+         *)
+           eval "$global_symbol_to_c_name_address_lib_prefix" < "$nlist" >> "$output_objdir/$my_dlsyms"
+           ;;
+         esac
+         echo >> "$output_objdir/$my_dlsyms" "\
+  {0, (void *) 0}
+};
+
+/* This works around a problem in FreeBSD linker */
+#ifdef FREEBSD_WORKAROUND
+static const void *lt_preloaded_setup() {
+  return lt_${my_prefix}_LTX_preloaded_symbols;
+}
+#endif
+
+#ifdef __cplusplus
+}
+#endif\
+"
+       } # !$opt_dry_run
+
+       pic_flag_for_symtable=
+       case "$compile_command " in
+       *" -static "*) ;;
+       *)
+         case $host in
+         # compiling the symbol table file with pic_flag works around
+         # a FreeBSD bug that causes programs to crash when -lm is
+         # linked before any other PIC object.  But we must not use
+         # pic_flag when linking with -static.  The problem exists in
+         # FreeBSD 2.2.6 and is fixed in FreeBSD 3.1.
+         *-*-freebsd2.*|*-*-freebsd3.0*|*-*-freebsdelf3.0*)
+           pic_flag_for_symtable=" $pic_flag -DFREEBSD_WORKAROUND" ;;
+         *-*-hpux*)
+           pic_flag_for_symtable=" $pic_flag"  ;;
+         *)
+           if test "X$my_pic_p" != Xno; then
+             pic_flag_for_symtable=" $pic_flag"
+           fi
+           ;;
+         esac
+         ;;
+       esac
+       symtab_cflags=
+       for arg in $LTCFLAGS; do
+         case $arg in
+         -pie | -fpie | -fPIE) ;;
+         *) func_append symtab_cflags " $arg" ;;
+         esac
+       done
+
+       # Now compile the dynamic symbol file.
+       func_show_eval '(cd $output_objdir && $LTCC$symtab_cflags -c$no_builtin_flag$pic_flag_for_symtable "$my_dlsyms")' 'exit $?'
+
+       # Clean up the generated files.
+       func_show_eval '$RM "$output_objdir/$my_dlsyms" "$nlist" "${nlist}S" "${nlist}T"'
+
+       # Transform the symbol file into the correct name.
+       symfileobj="$output_objdir/${my_outputname}S.$objext"
+       case $host in
+       *cygwin* | *mingw* | *cegcc* )
+         if test -f "$output_objdir/$my_outputname.def"; then
+           compile_command=`$ECHO "$compile_command" | $SED "s%@SYMFILE@%$output_objdir/$my_outputname.def $symfileobj%"`
+           finalize_command=`$ECHO "$finalize_command" | $SED "s%@SYMFILE@%$output_objdir/$my_outputname.def $symfileobj%"`
+         else
+           compile_command=`$ECHO "$compile_command" | $SED "s%@SYMFILE@%$symfileobj%"`
+           finalize_command=`$ECHO "$finalize_command" | $SED "s%@SYMFILE@%$symfileobj%"`
+         fi
+         ;;
+       *)
+         compile_command=`$ECHO "$compile_command" | $SED "s%@SYMFILE@%$symfileobj%"`
+         finalize_command=`$ECHO "$finalize_command" | $SED "s%@SYMFILE@%$symfileobj%"`
+         ;;
+       esac
+       ;;
+      *)
+       func_fatal_error "unknown suffix for \`$my_dlsyms'"
+       ;;
+      esac
+    else
+      # We keep going just in case the user didn't refer to
+      # lt_preloaded_symbols.  The linker will fail if global_symbol_pipe
+      # really was required.
+
+      # Nullify the symbol file.
+      compile_command=`$ECHO "$compile_command" | $SED "s% @SYMFILE@%%"`
+      finalize_command=`$ECHO "$finalize_command" | $SED "s% @SYMFILE@%%"`
+    fi
+}
+
+# func_win32_libid arg
+# return the library type of file 'arg'
+#
+# Need a lot of goo to handle *both* DLLs and import libs
+# Has to be a shell function in order to 'eat' the argument
+# that is supplied when $file_magic_command is called.
+# Despite the name, also deal with 64 bit binaries.
+func_win32_libid ()
+{
+  $opt_debug
+  win32_libid_type="unknown"
+  win32_fileres=`file -L $1 2>/dev/null`
+  case $win32_fileres in
+  *ar\ archive\ import\ library*) # definitely import
+    win32_libid_type="x86 archive import"
+    ;;
+  *ar\ archive*) # could be an import, or static
+    # Keep the egrep pattern in sync with the one in _LT_CHECK_MAGIC_METHOD.
+    if eval $OBJDUMP -f $1 | $SED -e '10q' 2>/dev/null |
+       $EGREP 'file format (pei*-i386(.*architecture: i386)?|pe-arm-wince|pe-x86-64)' >/dev/null; then
+      func_to_tool_file "$1" func_convert_file_msys_to_w32
+      win32_nmres=`eval $NM -f posix -A \"$func_to_tool_file_result\" |
+       $SED -n -e '
+           1,100{
+               / I /{
+                   s,.*,import,
+                   p
+                   q
+               }
+           }'`
+      case $win32_nmres in
+      import*)  win32_libid_type="x86 archive import";;
+      *)        win32_libid_type="x86 archive static";;
+      esac
+    fi
+    ;;
+  *DLL*)
+    win32_libid_type="x86 DLL"
+    ;;
+  *executable*) # but shell scripts are "executable" too...
+    case $win32_fileres in
+    *MS\ Windows\ PE\ Intel*)
+      win32_libid_type="x86 DLL"
+      ;;
+    esac
+    ;;
+  esac
+  $ECHO "$win32_libid_type"
+}
+
+# func_cygming_dll_for_implib ARG
+#
+# Platform-specific function to extract the
+# name of the DLL associated with the specified
+# import library ARG.
+# Invoked by eval'ing the libtool variable
+#    $sharedlib_from_linklib_cmd
+# Result is available in the variable
+#    $sharedlib_from_linklib_result
+func_cygming_dll_for_implib ()
+{
+  $opt_debug
+  sharedlib_from_linklib_result=`$DLLTOOL --identify-strict --identify "$1"`
+}
+
+# func_cygming_dll_for_implib_fallback_core SECTION_NAME LIBNAMEs
+#
+# The is the core of a fallback implementation of a
+# platform-specific function to extract the name of the
+# DLL associated with the specified import library LIBNAME.
+#
+# SECTION_NAME is either .idata$6 or .idata$7, depending
+# on the platform and compiler that created the implib.
+#
+# Echos the name of the DLL associated with the
+# specified import library.
+func_cygming_dll_for_implib_fallback_core ()
+{
+  $opt_debug
+  match_literal=`$ECHO "$1" | $SED "$sed_make_literal_regex"`
+  $OBJDUMP -s --section "$1" "$2" 2>/dev/null |
+    $SED '/^Contents of section '"$match_literal"':/{
+      # Place marker at beginning of archive member dllname section
+      s/.*/====MARK====/
+      p
+      d
+    }
+    # These lines can sometimes be longer than 43 characters, but
+    # are always uninteresting
+    /:[         ]*file format pe[i]\{,1\}-/d
+    /^In archive [^:]*:/d
+    # Ensure marker is printed
+    /^====MARK====/p
+    # Remove all lines with less than 43 characters
+    /^.\{43\}/!d
+    # From remaining lines, remove first 43 characters
+    s/^.\{43\}//' |
+    $SED -n '
+      # Join marker and all lines until next marker into a single line
+      /^====MARK====/ b para
+      H
+      $ b para
+      b
+      :para
+      x
+      s/\n//g
+      # Remove the marker
+      s/^====MARK====//
+      # Remove trailing dots and whitespace
+      s/[\. \t]*$//
+      # Print
+      /./p' |
+    # we now have a list, one entry per line, of the stringified
+    # contents of the appropriate section of all members of the
+    # archive which possess that section. Heuristic: eliminate
+    # all those which have a first or second character that is
+    # a '.' (that is, objdump's representation of an unprintable
+    # character.) This should work for all archives with less than
+    # 0x302f exports -- but will fail for DLLs whose name actually
+    # begins with a literal '.' or a single character followed by
+    # a '.'.
+    #
+    # Of those that remain, print the first one.
+    $SED -e '/^\./d;/^.\./d;q'
+}
+
+# func_cygming_gnu_implib_p ARG
+# This predicate returns with zero status (TRUE) if
+# ARG is a GNU/binutils-style import library. Returns
+# with nonzero status (FALSE) otherwise.
+func_cygming_gnu_implib_p ()
+{
+  $opt_debug
+  func_to_tool_file "$1" func_convert_file_msys_to_w32
+  func_cygming_gnu_implib_tmp=`$NM "$func_to_tool_file_result" | eval "$global_symbol_pipe" | $EGREP ' (_head_[A-Za-z0-9_]+_[ad]l*|[A-Za-z0-9_]+_[ad]l*_iname)$'`
+  test -n "$func_cygming_gnu_implib_tmp"
+}
+
+# func_cygming_ms_implib_p ARG
+# This predicate returns with zero status (TRUE) if
+# ARG is an MS-style import library. Returns
+# with nonzero status (FALSE) otherwise.
+func_cygming_ms_implib_p ()
+{
+  $opt_debug
+  func_to_tool_file "$1" func_convert_file_msys_to_w32
+  func_cygming_ms_implib_tmp=`$NM "$func_to_tool_file_result" | eval "$global_symbol_pipe" | $GREP '_NULL_IMPORT_DESCRIPTOR'`
+  test -n "$func_cygming_ms_implib_tmp"
+}
+
+# func_cygming_dll_for_implib_fallback ARG
+# Platform-specific function to extract the
+# name of the DLL associated with the specified
+# import library ARG.
+#
+# This fallback implementation is for use when $DLLTOOL
+# does not support the --identify-strict option.
+# Invoked by eval'ing the libtool variable
+#    $sharedlib_from_linklib_cmd
+# Result is available in the variable
+#    $sharedlib_from_linklib_result
+func_cygming_dll_for_implib_fallback ()
+{
+  $opt_debug
+  if func_cygming_gnu_implib_p "$1" ; then
+    # binutils import library
+    sharedlib_from_linklib_result=`func_cygming_dll_for_implib_fallback_core '.idata$7' "$1"`
+  elif func_cygming_ms_implib_p "$1" ; then
+    # ms-generated import library
+    sharedlib_from_linklib_result=`func_cygming_dll_for_implib_fallback_core '.idata$6' "$1"`
+  else
+    # unknown
+    sharedlib_from_linklib_result=""
+  fi
+}
+
+
+# func_extract_an_archive dir oldlib
+func_extract_an_archive ()
+{
+    $opt_debug
+    f_ex_an_ar_dir="$1"; shift
+    f_ex_an_ar_oldlib="$1"
+    if test "$lock_old_archive_extraction" = yes; then
+      lockfile=$f_ex_an_ar_oldlib.lock
+      until $opt_dry_run || ln "$progpath" "$lockfile" 2>/dev/null; do
+       func_echo "Waiting for $lockfile to be removed"
+       sleep 2
+      done
+    fi
+    func_show_eval "(cd \$f_ex_an_ar_dir && $AR x \"\$f_ex_an_ar_oldlib\")" \
+                  'stat=$?; rm -f "$lockfile"; exit $stat'
+    if test "$lock_old_archive_extraction" = yes; then
+      $opt_dry_run || rm -f "$lockfile"
+    fi
+    if ($AR t "$f_ex_an_ar_oldlib" | sort | sort -uc >/dev/null 2>&1); then
+     :
+    else
+      func_fatal_error "object name conflicts in archive: $f_ex_an_ar_dir/$f_ex_an_ar_oldlib"
+    fi
+}
+
+
+# func_extract_archives gentop oldlib ...
+func_extract_archives ()
+{
+    $opt_debug
+    my_gentop="$1"; shift
+    my_oldlibs=${1+"$@"}
+    my_oldobjs=""
+    my_xlib=""
+    my_xabs=""
+    my_xdir=""
+
+    for my_xlib in $my_oldlibs; do
+      # Extract the objects.
+      case $my_xlib in
+       [\\/]* | [A-Za-z]:[\\/]*) my_xabs="$my_xlib" ;;
+       *) my_xabs=`pwd`"/$my_xlib" ;;
+      esac
+      func_basename "$my_xlib"
+      my_xlib="$func_basename_result"
+      my_xlib_u=$my_xlib
+      while :; do
+        case " $extracted_archives " in
+       *" $my_xlib_u "*)
+         func_arith $extracted_serial + 1
+         extracted_serial=$func_arith_result
+         my_xlib_u=lt$extracted_serial-$my_xlib ;;
+       *) break ;;
+       esac
+      done
+      extracted_archives="$extracted_archives $my_xlib_u"
+      my_xdir="$my_gentop/$my_xlib_u"
+
+      func_mkdir_p "$my_xdir"
+
+      case $host in
+      *-darwin*)
+       func_verbose "Extracting $my_xabs"
+       # Do not bother doing anything if just a dry run
+       $opt_dry_run || {
+         darwin_orig_dir=`pwd`
+         cd $my_xdir || exit $?
+         darwin_archive=$my_xabs
+         darwin_curdir=`pwd`
+         darwin_base_archive=`basename "$darwin_archive"`
+         darwin_arches=`$LIPO -info "$darwin_archive" 2>/dev/null | $GREP Architectures 2>/dev/null || true`
+         if test -n "$darwin_arches"; then
+           darwin_arches=`$ECHO "$darwin_arches" | $SED -e 's/.*are://'`
+           darwin_arch=
+           func_verbose "$darwin_base_archive has multiple architectures $darwin_arches"
+           for darwin_arch in  $darwin_arches ; do
+             func_mkdir_p "unfat-$$/${darwin_base_archive}-${darwin_arch}"
+             $LIPO -thin $darwin_arch -output "unfat-$$/${darwin_base_archive}-${darwin_arch}/${darwin_base_archive}" "${darwin_archive}"
+             cd "unfat-$$/${darwin_base_archive}-${darwin_arch}"
+             func_extract_an_archive "`pwd`" "${darwin_base_archive}"
+             cd "$darwin_curdir"
+             $RM "unfat-$$/${darwin_base_archive}-${darwin_arch}/${darwin_base_archive}"
+           done # $darwin_arches
+            ## Okay now we've a bunch of thin objects, gotta fatten them up :)
+           darwin_filelist=`find unfat-$$ -type f -name \*.o -print -o -name \*.lo -print | $SED -e "$basename" | sort -u`
+           darwin_file=
+           darwin_files=
+           for darwin_file in $darwin_filelist; do
+             darwin_files=`find unfat-$$ -name $darwin_file -print | sort | $NL2SP`
+             $LIPO -create -output "$darwin_file" $darwin_files
+           done # $darwin_filelist
+           $RM -rf unfat-$$
+           cd "$darwin_orig_dir"
+         else
+           cd $darwin_orig_dir
+           func_extract_an_archive "$my_xdir" "$my_xabs"
+         fi # $darwin_arches
+       } # !$opt_dry_run
+       ;;
+      *)
+        func_extract_an_archive "$my_xdir" "$my_xabs"
+       ;;
+      esac
+      my_oldobjs="$my_oldobjs "`find $my_xdir -name \*.$objext -print -o -name \*.lo -print | sort | $NL2SP`
+    done
+
+    func_extract_archives_result="$my_oldobjs"
+}
+
+
+# func_emit_wrapper [arg=no]
+#
+# Emit a libtool wrapper script on stdout.
+# Don't directly open a file because we may want to
+# incorporate the script contents within a cygwin/mingw
+# wrapper executable.  Must ONLY be called from within
+# func_mode_link because it depends on a number of variables
+# set therein.
+#
+# ARG is the value that the WRAPPER_SCRIPT_BELONGS_IN_OBJDIR
+# variable will take.  If 'yes', then the emitted script
+# will assume that the directory in which it is stored is
+# the $objdir directory.  This is a cygwin/mingw-specific
+# behavior.
+func_emit_wrapper ()
+{
+       func_emit_wrapper_arg1=${1-no}
+
+       $ECHO "\
+#! $SHELL
+
+# $output - temporary wrapper script for $objdir/$outputname
+# Generated by $PROGRAM (GNU $PACKAGE$TIMESTAMP) $VERSION
+#
+# The $output program cannot be directly executed until all the libtool
+# libraries that it depends on are installed.
+#
+# This wrapper script should never be moved out of the build directory.
+# If it is, it will not operate correctly.
+
+# Sed substitution that helps us do robust quoting.  It backslashifies
+# metacharacters that are still active within double-quoted strings.
+sed_quote_subst='$sed_quote_subst'
+
+# Be Bourne compatible
+if test -n \"\${ZSH_VERSION+set}\" && (emulate sh) >/dev/null 2>&1; then
+  emulate sh
+  NULLCMD=:
+  # Zsh 3.x and 4.x performs word splitting on \${1+\"\$@\"}, which
+  # is contrary to our usage.  Disable this feature.
+  alias -g '\${1+\"\$@\"}'='\"\$@\"'
+  setopt NO_GLOB_SUBST
+else
+  case \`(set -o) 2>/dev/null\` in *posix*) set -o posix;; esac
+fi
+BIN_SH=xpg4; export BIN_SH # for Tru64
+DUALCASE=1; export DUALCASE # for MKS sh
+
+# The HP-UX ksh and POSIX shell print the target directory to stdout
+# if CDPATH is set.
+(unset CDPATH) >/dev/null 2>&1 && unset CDPATH
+
+relink_command=\"$relink_command\"
+
+# This environment variable determines our operation mode.
+if test \"\$libtool_install_magic\" = \"$magic\"; then
+  # install mode needs the following variables:
+  generated_by_libtool_version='$macro_version'
+  notinst_deplibs='$notinst_deplibs'
+else
+  # When we are sourced in execute mode, \$file and \$ECHO are already set.
+  if test \"\$libtool_execute_magic\" != \"$magic\"; then
+    file=\"\$0\""
+
+    qECHO=`$ECHO "$ECHO" | $SED "$sed_quote_subst"`
+    $ECHO "\
+
+# A function that is used when there is no print builtin or printf.
+func_fallback_echo ()
+{
+  eval 'cat <<_LTECHO_EOF
+\$1
+_LTECHO_EOF'
+}
+    ECHO=\"$qECHO\"
+  fi
+
+# Very basic option parsing. These options are (a) specific to
+# the libtool wrapper, (b) are identical between the wrapper
+# /script/ and the wrapper /executable/ which is used only on
+# windows platforms, and (c) all begin with the string "--lt-"
+# (application programs are unlikely to have options which match
+# this pattern).
+#
+# There are only two supported options: --lt-debug and
+# --lt-dump-script. There is, deliberately, no --lt-help.
+#
+# The first argument to this parsing function should be the
+# script's $0 value, followed by "$@".
+lt_option_debug=
+func_parse_lt_options ()
+{
+  lt_script_arg0=\$0
+  shift
+  for lt_opt
+  do
+    case \"\$lt_opt\" in
+    --lt-debug) lt_option_debug=1 ;;
+    --lt-dump-script)
+        lt_dump_D=\`\$ECHO \"X\$lt_script_arg0\" | $SED -e 's/^X//' -e 's%/[^/]*$%%'\`
+        test \"X\$lt_dump_D\" = \"X\$lt_script_arg0\" && lt_dump_D=.
+        lt_dump_F=\`\$ECHO \"X\$lt_script_arg0\" | $SED -e 's/^X//' -e 's%^.*/%%'\`
+        cat \"\$lt_dump_D/\$lt_dump_F\"
+        exit 0
+      ;;
+    --lt-*)
+        \$ECHO \"Unrecognized --lt- option: '\$lt_opt'\" 1>&2
+        exit 1
+      ;;
+    esac
+  done
+
+  # Print the debug banner immediately:
+  if test -n \"\$lt_option_debug\"; then
+    echo \"${outputname}:${output}:\${LINENO}: libtool wrapper (GNU $PACKAGE$TIMESTAMP) $VERSION\" 1>&2
+  fi
+}
+
+# Used when --lt-debug. Prints its arguments to stdout
+# (redirection is the responsibility of the caller)
+func_lt_dump_args ()
+{
+  lt_dump_args_N=1;
+  for lt_arg
+  do
+    \$ECHO \"${outputname}:${output}:\${LINENO}: newargv[\$lt_dump_args_N]: \$lt_arg\"
+    lt_dump_args_N=\`expr \$lt_dump_args_N + 1\`
+  done
+}
+
+# Core function for launching the target application
+func_exec_program_core ()
+{
+"
+  case $host in
+  # Backslashes separate directories on plain windows
+  *-*-mingw | *-*-os2* | *-cegcc*)
+    $ECHO "\
+      if test -n \"\$lt_option_debug\"; then
+        \$ECHO \"${outputname}:${output}:\${LINENO}: newargv[0]: \$progdir\\\\\$program\" 1>&2
+        func_lt_dump_args \${1+\"\$@\"} 1>&2
+      fi
+      exec \"\$progdir\\\\\$program\" \${1+\"\$@\"}
+"
+    ;;
+
+  *)
+    $ECHO "\
+      if test -n \"\$lt_option_debug\"; then
+        \$ECHO \"${outputname}:${output}:\${LINENO}: newargv[0]: \$progdir/\$program\" 1>&2
+        func_lt_dump_args \${1+\"\$@\"} 1>&2
+      fi
+      exec \"\$progdir/\$program\" \${1+\"\$@\"}
+"
+    ;;
+  esac
+  $ECHO "\
+      \$ECHO \"\$0: cannot exec \$program \$*\" 1>&2
+      exit 1
+}
+
+# A function to encapsulate launching the target application
+# Strips options in the --lt-* namespace from \$@ and
+# launches target application with the remaining arguments.
+func_exec_program ()
+{
+  case \" \$* \" in
+  *\\ --lt-*)
+    for lt_wr_arg
+    do
+      case \$lt_wr_arg in
+      --lt-*) ;;
+      *) set x \"\$@\" \"\$lt_wr_arg\"; shift;;
+      esac
+      shift
+    done ;;
+  esac
+  func_exec_program_core \${1+\"\$@\"}
+}
+
+  # Parse options
+  func_parse_lt_options \"\$0\" \${1+\"\$@\"}
+
+  # Find the directory that this script lives in.
+  thisdir=\`\$ECHO \"\$file\" | $SED 's%/[^/]*$%%'\`
+  test \"x\$thisdir\" = \"x\$file\" && thisdir=.
+
+  # Follow symbolic links until we get to the real thisdir.
+  file=\`ls -ld \"\$file\" | $SED -n 's/.*-> //p'\`
+  while test -n \"\$file\"; do
+    destdir=\`\$ECHO \"\$file\" | $SED 's%/[^/]*\$%%'\`
+
+    # If there was a directory component, then change thisdir.
+    if test \"x\$destdir\" != \"x\$file\"; then
+      case \"\$destdir\" in
+      [\\\\/]* | [A-Za-z]:[\\\\/]*) thisdir=\"\$destdir\" ;;
+      *) thisdir=\"\$thisdir/\$destdir\" ;;
+      esac
+    fi
+
+    file=\`\$ECHO \"\$file\" | $SED 's%^.*/%%'\`
+    file=\`ls -ld \"\$thisdir/\$file\" | $SED -n 's/.*-> //p'\`
+  done
+
+  # Usually 'no', except on cygwin/mingw when embedded into
+  # the cwrapper.
+  WRAPPER_SCRIPT_BELONGS_IN_OBJDIR=$func_emit_wrapper_arg1
+  if test \"\$WRAPPER_SCRIPT_BELONGS_IN_OBJDIR\" = \"yes\"; then
+    # special case for '.'
+    if test \"\$thisdir\" = \".\"; then
+      thisdir=\`pwd\`
+    fi
+    # remove .libs from thisdir
+    case \"\$thisdir\" in
+    *[\\\\/]$objdir ) thisdir=\`\$ECHO \"\$thisdir\" | $SED 's%[\\\\/][^\\\\/]*$%%'\` ;;
+    $objdir )   thisdir=. ;;
+    esac
+  fi
+
+  # Try to get the absolute directory name.
+  absdir=\`cd \"\$thisdir\" && pwd\`
+  test -n \"\$absdir\" && thisdir=\"\$absdir\"
+"
+
+       if test "$fast_install" = yes; then
+         $ECHO "\
+  program=lt-'$outputname'$exeext
+  progdir=\"\$thisdir/$objdir\"
+
+  if test ! -f \"\$progdir/\$program\" ||
+     { file=\`ls -1dt \"\$progdir/\$program\" \"\$progdir/../\$program\" 2>/dev/null | ${SED} 1q\`; \\
+       test \"X\$file\" != \"X\$progdir/\$program\"; }; then
+
+    file=\"\$\$-\$program\"
+
+    if test ! -d \"\$progdir\"; then
+      $MKDIR \"\$progdir\"
+    else
+      $RM \"\$progdir/\$file\"
+    fi"
+
+         $ECHO "\
+
+    # relink executable if necessary
+    if test -n \"\$relink_command\"; then
+      if relink_command_output=\`eval \$relink_command 2>&1\`; then :
+      else
+       $ECHO \"\$relink_command_output\" >&2
+       $RM \"\$progdir/\$file\"
+       exit 1
+      fi
+    fi
+
+    $MV \"\$progdir/\$file\" \"\$progdir/\$program\" 2>/dev/null ||
+    { $RM \"\$progdir/\$program\";
+      $MV \"\$progdir/\$file\" \"\$progdir/\$program\"; }
+    $RM \"\$progdir/\$file\"
+  fi"
+       else
+         $ECHO "\
+  program='$outputname'
+  progdir=\"\$thisdir/$objdir\"
+"
+       fi
+
+       $ECHO "\
+
+  if test -f \"\$progdir/\$program\"; then"
+
+       # fixup the dll searchpath if we need to.
+       #
+       # Fix the DLL searchpath if we need to.  Do this before prepending
+       # to shlibpath, because on Windows, both are PATH and uninstalled
+       # libraries must come first.
+       if test -n "$dllsearchpath"; then
+         $ECHO "\
+    # Add the dll search path components to the executable PATH
+    PATH=$dllsearchpath:\$PATH
+"
+       fi
+
+       # Export our shlibpath_var if we have one.
+       if test "$shlibpath_overrides_runpath" = yes && test -n "$shlibpath_var" && test -n "$temp_rpath"; then
+         $ECHO "\
+    # Add our own library path to $shlibpath_var
+    $shlibpath_var=\"$temp_rpath\$$shlibpath_var\"
+
+    # Some systems cannot cope with colon-terminated $shlibpath_var
+    # The second colon is a workaround for a bug in BeOS R4 sed
+    $shlibpath_var=\`\$ECHO \"\$$shlibpath_var\" | $SED 's/::*\$//'\`
+
+    export $shlibpath_var
+"
+       fi
+
+       $ECHO "\
+    if test \"\$libtool_execute_magic\" != \"$magic\"; then
+      # Run the actual program with our arguments.
+      func_exec_program \${1+\"\$@\"}
+    fi
+  else
+    # The program doesn't exist.
+    \$ECHO \"\$0: error: \\\`\$progdir/\$program' does not exist\" 1>&2
+    \$ECHO \"This script is just a wrapper for \$program.\" 1>&2
+    \$ECHO \"See the $PACKAGE documentation for more information.\" 1>&2
+    exit 1
+  fi
+fi\
+"
+}
+
+
+# func_emit_cwrapperexe_src
+# emit the source code for a wrapper executable on stdout
+# Must ONLY be called from within func_mode_link because
+# it depends on a number of variable set therein.
+func_emit_cwrapperexe_src ()
+{
+       cat <<EOF
+
+/* $cwrappersource - temporary wrapper executable for $objdir/$outputname
+   Generated by $PROGRAM (GNU $PACKAGE$TIMESTAMP) $VERSION
+
+   The $output program cannot be directly executed until all the libtool
+   libraries that it depends on are installed.
+
+   This wrapper executable should never be moved out of the build directory.
+   If it is, it will not operate correctly.
+*/
+EOF
+           cat <<"EOF"
+#ifdef _MSC_VER
+# define _CRT_SECURE_NO_DEPRECATE 1
+#endif
+#include <stdio.h>
+#include <stdlib.h>
+#ifdef _MSC_VER
+# include <direct.h>
+# include <process.h>
+# include <io.h>
+#else
+# include <unistd.h>
+# include <stdint.h>
+# ifdef __CYGWIN__
+#  include <io.h>
+# endif
+#endif
+#include <malloc.h>
+#include <stdarg.h>
+#include <assert.h>
+#include <string.h>
+#include <ctype.h>
+#include <errno.h>
+#include <fcntl.h>
+#include <sys/stat.h>
+
+/* declarations of non-ANSI functions */
+#if defined(__MINGW32__)
+# ifdef __STRICT_ANSI__
+int _putenv (const char *);
+# endif
+#elif defined(__CYGWIN__)
+# ifdef __STRICT_ANSI__
+char *realpath (const char *, char *);
+int putenv (char *);
+int setenv (const char *, const char *, int);
+# endif
+/* #elif defined (other platforms) ... */
+#endif
+
+/* portability defines, excluding path handling macros */
+#if defined(_MSC_VER)
+# define setmode _setmode
+# define stat    _stat
+# define chmod   _chmod
+# define getcwd  _getcwd
+# define putenv  _putenv
+# define S_IXUSR _S_IEXEC
+# ifndef _INTPTR_T_DEFINED
+#  define _INTPTR_T_DEFINED
+#  define intptr_t int
+# endif
+#elif defined(__MINGW32__)
+# define setmode _setmode
+# define stat    _stat
+# define chmod   _chmod
+# define getcwd  _getcwd
+# define putenv  _putenv
+#elif defined(__CYGWIN__)
+# define HAVE_SETENV
+# define FOPEN_WB "wb"
+/* #elif defined (other platforms) ... */
+#endif
+
+#if defined(PATH_MAX)
+# define LT_PATHMAX PATH_MAX
+#elif defined(MAXPATHLEN)
+# define LT_PATHMAX MAXPATHLEN
+#else
+# define LT_PATHMAX 1024
+#endif
+
+#ifndef S_IXOTH
+# define S_IXOTH 0
+#endif
+#ifndef S_IXGRP
+# define S_IXGRP 0
+#endif
+
+/* path handling portability macros */
+#ifndef DIR_SEPARATOR
+# define DIR_SEPARATOR '/'
+# define PATH_SEPARATOR ':'
+#endif
+
+#if defined (_WIN32) || defined (__MSDOS__) || defined (__DJGPP__) || \
+  defined (__OS2__)
+# define HAVE_DOS_BASED_FILE_SYSTEM
+# define FOPEN_WB "wb"
+# ifndef DIR_SEPARATOR_2
+#  define DIR_SEPARATOR_2 '\\'
+# endif
+# ifndef PATH_SEPARATOR_2
+#  define PATH_SEPARATOR_2 ';'
+# endif
+#endif
+
+#ifndef DIR_SEPARATOR_2
+# define IS_DIR_SEPARATOR(ch) ((ch) == DIR_SEPARATOR)
+#else /* DIR_SEPARATOR_2 */
+# define IS_DIR_SEPARATOR(ch) \
+       (((ch) == DIR_SEPARATOR) || ((ch) == DIR_SEPARATOR_2))
+#endif /* DIR_SEPARATOR_2 */
+
+#ifndef PATH_SEPARATOR_2
+# define IS_PATH_SEPARATOR(ch) ((ch) == PATH_SEPARATOR)
+#else /* PATH_SEPARATOR_2 */
+# define IS_PATH_SEPARATOR(ch) ((ch) == PATH_SEPARATOR_2)
+#endif /* PATH_SEPARATOR_2 */
+
+#ifndef FOPEN_WB
+# define FOPEN_WB "w"
+#endif
+#ifndef _O_BINARY
+# define _O_BINARY 0
+#endif
+
+#define XMALLOC(type, num)      ((type *) xmalloc ((num) * sizeof(type)))
+#define XFREE(stale) do { \
+  if (stale) { free ((void *) stale); stale = 0; } \
+} while (0)
+
+#if defined(LT_DEBUGWRAPPER)
+static int lt_debug = 1;
+#else
+static int lt_debug = 0;
+#endif
+
+const char *program_name = "libtool-wrapper"; /* in case xstrdup fails */
+
+void *xmalloc (size_t num);
+char *xstrdup (const char *string);
+const char *base_name (const char *name);
+char *find_executable (const char *wrapper);
+char *chase_symlinks (const char *pathspec);
+int make_executable (const char *path);
+int check_executable (const char *path);
+char *strendzap (char *str, const char *pat);
+void lt_debugprintf (const char *file, int line, const char *fmt, ...);
+void lt_fatal (const char *file, int line, const char *message, ...);
+static const char *nonnull (const char *s);
+static const char *nonempty (const char *s);
+void lt_setenv (const char *name, const char *value);
+char *lt_extend_str (const char *orig_value, const char *add, int to_end);
+void lt_update_exe_path (const char *name, const char *value);
+void lt_update_lib_path (const char *name, const char *value);
+char **prepare_spawn (char **argv);
+void lt_dump_script (FILE *f);
+EOF
+
+           cat <<EOF
+volatile const char * MAGIC_EXE = "$magic_exe";
+const char * LIB_PATH_VARNAME = "$shlibpath_var";
+EOF
+
+           if test "$shlibpath_overrides_runpath" = yes && test -n "$shlibpath_var" && test -n "$temp_rpath"; then
+              func_to_host_path "$temp_rpath"
+             cat <<EOF
+const char * LIB_PATH_VALUE   = "$func_to_host_path_result";
+EOF
+           else
+             cat <<"EOF"
+const char * LIB_PATH_VALUE   = "";
+EOF
+           fi
+
+           if test -n "$dllsearchpath"; then
+              func_to_host_path "$dllsearchpath:"
+             cat <<EOF
+const char * EXE_PATH_VARNAME = "PATH";
+const char * EXE_PATH_VALUE   = "$func_to_host_path_result";
+EOF
+           else
+             cat <<"EOF"
+const char * EXE_PATH_VARNAME = "";
+const char * EXE_PATH_VALUE   = "";
+EOF
+           fi
+
+           if test "$fast_install" = yes; then
+             cat <<EOF
+const char * TARGET_PROGRAM_NAME = "lt-$outputname"; /* hopefully, no .exe */
+EOF
+           else
+             cat <<EOF
+const char * TARGET_PROGRAM_NAME = "$outputname"; /* hopefully, no .exe */
+EOF
+           fi
+
+
+           cat <<"EOF"
+
+#define LTWRAPPER_OPTION_PREFIX         "--lt-"
+
+static const char *ltwrapper_option_prefix = LTWRAPPER_OPTION_PREFIX;
+static const char *dumpscript_opt       = LTWRAPPER_OPTION_PREFIX "dump-script";
+static const char *debug_opt            = LTWRAPPER_OPTION_PREFIX "debug";
+
+int
+main (int argc, char *argv[])
+{
+  char **newargz;
+  int  newargc;
+  char *tmp_pathspec;
+  char *actual_cwrapper_path;
+  char *actual_cwrapper_name;
+  char *target_name;
+  char *lt_argv_zero;
+  intptr_t rval = 127;
+
+  int i;
+
+  program_name = (char *) xstrdup (base_name (argv[0]));
+  newargz = XMALLOC (char *, argc + 1);
+
+  /* very simple arg parsing; don't want to rely on getopt
+   * also, copy all non cwrapper options to newargz, except
+   * argz[0], which is handled differently
+   */
+  newargc=0;
+  for (i = 1; i < argc; i++)
+    {
+      if (strcmp (argv[i], dumpscript_opt) == 0)
+       {
+EOF
+           case "$host" in
+             *mingw* | *cygwin* )
+               # make stdout use "unix" line endings
+               echo "          setmode(1,_O_BINARY);"
+               ;;
+             esac
+
+           cat <<"EOF"
+         lt_dump_script (stdout);
+         return 0;
+       }
+      if (strcmp (argv[i], debug_opt) == 0)
+       {
+          lt_debug = 1;
+          continue;
+       }
+      if (strcmp (argv[i], ltwrapper_option_prefix) == 0)
+        {
+          /* however, if there is an option in the LTWRAPPER_OPTION_PREFIX
+             namespace, but it is not one of the ones we know about and
+             have already dealt with, above (inluding dump-script), then
+             report an error. Otherwise, targets might begin to believe
+             they are allowed to use options in the LTWRAPPER_OPTION_PREFIX
+             namespace. The first time any user complains about this, we'll
+             need to make LTWRAPPER_OPTION_PREFIX a configure-time option
+             or a configure.ac-settable value.
+           */
+          lt_fatal (__FILE__, __LINE__,
+                   "unrecognized %s option: '%s'",
+                    ltwrapper_option_prefix, argv[i]);
+        }
+      /* otherwise ... */
+      newargz[++newargc] = xstrdup (argv[i]);
+    }
+  newargz[++newargc] = NULL;
+
+EOF
+           cat <<EOF
+  /* The GNU banner must be the first non-error debug message */
+  lt_debugprintf (__FILE__, __LINE__, "libtool wrapper (GNU $PACKAGE$TIMESTAMP) $VERSION\n");
+EOF
+           cat <<"EOF"
+  lt_debugprintf (__FILE__, __LINE__, "(main) argv[0]: %s\n", argv[0]);
+  lt_debugprintf (__FILE__, __LINE__, "(main) program_name: %s\n", program_name);
+
+  tmp_pathspec = find_executable (argv[0]);
+  if (tmp_pathspec == NULL)
+    lt_fatal (__FILE__, __LINE__, "couldn't find %s", argv[0]);
+  lt_debugprintf (__FILE__, __LINE__,
+                  "(main) found exe (before symlink chase) at: %s\n",
+                 tmp_pathspec);
+
+  actual_cwrapper_path = chase_symlinks (tmp_pathspec);
+  lt_debugprintf (__FILE__, __LINE__,
+                  "(main) found exe (after symlink chase) at: %s\n",
+                 actual_cwrapper_path);
+  XFREE (tmp_pathspec);
+
+  actual_cwrapper_name = xstrdup (base_name (actual_cwrapper_path));
+  strendzap (actual_cwrapper_path, actual_cwrapper_name);
+
+  /* wrapper name transforms */
+  strendzap (actual_cwrapper_name, ".exe");
+  tmp_pathspec = lt_extend_str (actual_cwrapper_name, ".exe", 1);
+  XFREE (actual_cwrapper_name);
+  actual_cwrapper_name = tmp_pathspec;
+  tmp_pathspec = 0;
+
+  /* target_name transforms -- use actual target program name; might have lt- prefix */
+  target_name = xstrdup (base_name (TARGET_PROGRAM_NAME));
+  strendzap (target_name, ".exe");
+  tmp_pathspec = lt_extend_str (target_name, ".exe", 1);
+  XFREE (target_name);
+  target_name = tmp_pathspec;
+  tmp_pathspec = 0;
+
+  lt_debugprintf (__FILE__, __LINE__,
+                 "(main) libtool target name: %s\n",
+                 target_name);
+EOF
+
+           cat <<EOF
+  newargz[0] =
+    XMALLOC (char, (strlen (actual_cwrapper_path) +
+                   strlen ("$objdir") + 1 + strlen (actual_cwrapper_name) + 1));
+  strcpy (newargz[0], actual_cwrapper_path);
+  strcat (newargz[0], "$objdir");
+  strcat (newargz[0], "/");
+EOF
+
+           cat <<"EOF"
+  /* stop here, and copy so we don't have to do this twice */
+  tmp_pathspec = xstrdup (newargz[0]);
+
+  /* do NOT want the lt- prefix here, so use actual_cwrapper_name */
+  strcat (newargz[0], actual_cwrapper_name);
+
+  /* DO want the lt- prefix here if it exists, so use target_name */
+  lt_argv_zero = lt_extend_str (tmp_pathspec, target_name, 1);
+  XFREE (tmp_pathspec);
+  tmp_pathspec = NULL;
+EOF
+
+           case $host_os in
+             mingw*)
+           cat <<"EOF"
+  {
+    char* p;
+    while ((p = strchr (newargz[0], '\\')) != NULL)
+      {
+       *p = '/';
+      }
+    while ((p = strchr (lt_argv_zero, '\\')) != NULL)
+      {
+       *p = '/';
+      }
+  }
+EOF
+           ;;
+           esac
+
+           cat <<"EOF"
+  XFREE (target_name);
+  XFREE (actual_cwrapper_path);
+  XFREE (actual_cwrapper_name);
+
+  lt_setenv ("BIN_SH", "xpg4"); /* for Tru64 */
+  lt_setenv ("DUALCASE", "1");  /* for MSK sh */
+  /* Update the DLL searchpath.  EXE_PATH_VALUE ($dllsearchpath) must
+     be prepended before (that is, appear after) LIB_PATH_VALUE ($temp_rpath)
+     because on Windows, both *_VARNAMEs are PATH but uninstalled
+     libraries must come first. */
+  lt_update_exe_path (EXE_PATH_VARNAME, EXE_PATH_VALUE);
+  lt_update_lib_path (LIB_PATH_VARNAME, LIB_PATH_VALUE);
+
+  lt_debugprintf (__FILE__, __LINE__, "(main) lt_argv_zero: %s\n",
+                 nonnull (lt_argv_zero));
+  for (i = 0; i < newargc; i++)
+    {
+      lt_debugprintf (__FILE__, __LINE__, "(main) newargz[%d]: %s\n",
+                     i, nonnull (newargz[i]));
+    }
+
+EOF
+
+           case $host_os in
+             mingw*)
+               cat <<"EOF"
+  /* execv doesn't actually work on mingw as expected on unix */
+  newargz = prepare_spawn (newargz);
+  rval = _spawnv (_P_WAIT, lt_argv_zero, (const char * const *) newargz);
+  if (rval == -1)
+    {
+      /* failed to start process */
+      lt_debugprintf (__FILE__, __LINE__,
+                     "(main) failed to launch target \"%s\": %s\n",
+                     lt_argv_zero, nonnull (strerror (errno)));
+      return 127;
+    }
+  return rval;
+EOF
+               ;;
+             *)
+               cat <<"EOF"
+  execv (lt_argv_zero, newargz);
+  return rval; /* =127, but avoids unused variable warning */
+EOF
+               ;;
+           esac
+
+           cat <<"EOF"
+}
+
+void *
+xmalloc (size_t num)
+{
+  void *p = (void *) malloc (num);
+  if (!p)
+    lt_fatal (__FILE__, __LINE__, "memory exhausted");
+
+  return p;
+}
+
+char *
+xstrdup (const char *string)
+{
+  return string ? strcpy ((char *) xmalloc (strlen (string) + 1),
+                         string) : NULL;
+}
+
+const char *
+base_name (const char *name)
+{
+  const char *base;
+
+#if defined (HAVE_DOS_BASED_FILE_SYSTEM)
+  /* Skip over the disk name in MSDOS pathnames. */
+  if (isalpha ((unsigned char) name[0]) && name[1] == ':')
+    name += 2;
+#endif
+
+  for (base = name; *name; name++)
+    if (IS_DIR_SEPARATOR (*name))
+      base = name + 1;
+  return base;
+}
+
+int
+check_executable (const char *path)
+{
+  struct stat st;
+
+  lt_debugprintf (__FILE__, __LINE__, "(check_executable): %s\n",
+                  nonempty (path));
+  if ((!path) || (!*path))
+    return 0;
+
+  if ((stat (path, &st) >= 0)
+      && (st.st_mode & (S_IXUSR | S_IXGRP | S_IXOTH)))
+    return 1;
+  else
+    return 0;
+}
+
+int
+make_executable (const char *path)
+{
+  int rval = 0;
+  struct stat st;
+
+  lt_debugprintf (__FILE__, __LINE__, "(make_executable): %s\n",
+                  nonempty (path));
+  if ((!path) || (!*path))
+    return 0;
+
+  if (stat (path, &st) >= 0)
+    {
+      rval = chmod (path, st.st_mode | S_IXOTH | S_IXGRP | S_IXUSR);
+    }
+  return rval;
+}
+
+/* Searches for the full path of the wrapper.  Returns
+   newly allocated full path name if found, NULL otherwise
+   Does not chase symlinks, even on platforms that support them.
+*/
+char *
+find_executable (const char *wrapper)
+{
+  int has_slash = 0;
+  const char *p;
+  const char *p_next;
+  /* static buffer for getcwd */
+  char tmp[LT_PATHMAX + 1];
+  int tmp_len;
+  char *concat_name;
+
+  lt_debugprintf (__FILE__, __LINE__, "(find_executable): %s\n",
+                  nonempty (wrapper));
+
+  if ((wrapper == NULL) || (*wrapper == '\0'))
+    return NULL;
+
+  /* Absolute path? */
+#if defined (HAVE_DOS_BASED_FILE_SYSTEM)
+  if (isalpha ((unsigned char) wrapper[0]) && wrapper[1] == ':')
+    {
+      concat_name = xstrdup (wrapper);
+      if (check_executable (concat_name))
+       return concat_name;
+      XFREE (concat_name);
+    }
+  else
+    {
+#endif
+      if (IS_DIR_SEPARATOR (wrapper[0]))
+       {
+         concat_name = xstrdup (wrapper);
+         if (check_executable (concat_name))
+           return concat_name;
+         XFREE (concat_name);
+       }
+#if defined (HAVE_DOS_BASED_FILE_SYSTEM)
+    }
+#endif
+
+  for (p = wrapper; *p; p++)
+    if (*p == '/')
+      {
+       has_slash = 1;
+       break;
+      }
+  if (!has_slash)
+    {
+      /* no slashes; search PATH */
+      const char *path = getenv ("PATH");
+      if (path != NULL)
+       {
+         for (p = path; *p; p = p_next)
+           {
+             const char *q;
+             size_t p_len;
+             for (q = p; *q; q++)
+               if (IS_PATH_SEPARATOR (*q))
+                 break;
+             p_len = q - p;
+             p_next = (*q == '\0' ? q : q + 1);
+             if (p_len == 0)
+               {
+                 /* empty path: current directory */
+                 if (getcwd (tmp, LT_PATHMAX) == NULL)
+                   lt_fatal (__FILE__, __LINE__, "getcwd failed: %s",
+                              nonnull (strerror (errno)));
+                 tmp_len = strlen (tmp);
+                 concat_name =
+                   XMALLOC (char, tmp_len + 1 + strlen (wrapper) + 1);
+                 memcpy (concat_name, tmp, tmp_len);
+                 concat_name[tmp_len] = '/';
+                 strcpy (concat_name + tmp_len + 1, wrapper);
+               }
+             else
+               {
+                 concat_name =
+                   XMALLOC (char, p_len + 1 + strlen (wrapper) + 1);
+                 memcpy (concat_name, p, p_len);
+                 concat_name[p_len] = '/';
+                 strcpy (concat_name + p_len + 1, wrapper);
+               }
+             if (check_executable (concat_name))
+               return concat_name;
+             XFREE (concat_name);
+           }
+       }
+      /* not found in PATH; assume curdir */
+    }
+  /* Relative path | not found in path: prepend cwd */
+  if (getcwd (tmp, LT_PATHMAX) == NULL)
+    lt_fatal (__FILE__, __LINE__, "getcwd failed: %s",
+              nonnull (strerror (errno)));
+  tmp_len = strlen (tmp);
+  concat_name = XMALLOC (char, tmp_len + 1 + strlen (wrapper) + 1);
+  memcpy (concat_name, tmp, tmp_len);
+  concat_name[tmp_len] = '/';
+  strcpy (concat_name + tmp_len + 1, wrapper);
+
+  if (check_executable (concat_name))
+    return concat_name;
+  XFREE (concat_name);
+  return NULL;
+}
+
+char *
+chase_symlinks (const char *pathspec)
+{
+#ifndef S_ISLNK
+  return xstrdup (pathspec);
+#else
+  char buf[LT_PATHMAX];
+  struct stat s;
+  char *tmp_pathspec = xstrdup (pathspec);
+  char *p;
+  int has_symlinks = 0;
+  while (strlen (tmp_pathspec) && !has_symlinks)
+    {
+      lt_debugprintf (__FILE__, __LINE__,
+                     "checking path component for symlinks: %s\n",
+                     tmp_pathspec);
+      if (lstat (tmp_pathspec, &s) == 0)
+       {
+         if (S_ISLNK (s.st_mode) != 0)
+           {
+             has_symlinks = 1;
+             break;
+           }
+
+         /* search backwards for last DIR_SEPARATOR */
+         p = tmp_pathspec + strlen (tmp_pathspec) - 1;
+         while ((p > tmp_pathspec) && (!IS_DIR_SEPARATOR (*p)))
+           p--;
+         if ((p == tmp_pathspec) && (!IS_DIR_SEPARATOR (*p)))
+           {
+             /* no more DIR_SEPARATORS left */
+             break;
+           }
+         *p = '\0';
+       }
+      else
+       {
+         lt_fatal (__FILE__, __LINE__,
+                   "error accessing file \"%s\": %s",
+                   tmp_pathspec, nonnull (strerror (errno)));
+       }
+    }
+  XFREE (tmp_pathspec);
+
+  if (!has_symlinks)
+    {
+      return xstrdup (pathspec);
+    }
+
+  tmp_pathspec = realpath (pathspec, buf);
+  if (tmp_pathspec == 0)
+    {
+      lt_fatal (__FILE__, __LINE__,
+               "could not follow symlinks for %s", pathspec);
+    }
+  return xstrdup (tmp_pathspec);
+#endif
+}
+
+char *
+strendzap (char *str, const char *pat)
+{
+  size_t len, patlen;
+
+  assert (str != NULL);
+  assert (pat != NULL);
+
+  len = strlen (str);
+  patlen = strlen (pat);
+
+  if (patlen <= len)
+    {
+      str += len - patlen;
+      if (strcmp (str, pat) == 0)
+       *str = '\0';
+    }
+  return str;
+}
+
+void
+lt_debugprintf (const char *file, int line, const char *fmt, ...)
+{
+  va_list args;
+  if (lt_debug)
+    {
+      (void) fprintf (stderr, "%s:%s:%d: ", program_name, file, line);
+      va_start (args, fmt);
+      (void) vfprintf (stderr, fmt, args);
+      va_end (args);
+    }
+}
+
+static void
+lt_error_core (int exit_status, const char *file,
+              int line, const char *mode,
+              const char *message, va_list ap)
+{
+  fprintf (stderr, "%s:%s:%d: %s: ", program_name, file, line, mode);
+  vfprintf (stderr, message, ap);
+  fprintf (stderr, ".\n");
+
+  if (exit_status >= 0)
+    exit (exit_status);
+}
+
+void
+lt_fatal (const char *file, int line, const char *message, ...)
+{
+  va_list ap;
+  va_start (ap, message);
+  lt_error_core (EXIT_FAILURE, file, line, "FATAL", message, ap);
+  va_end (ap);
+}
+
+static const char *
+nonnull (const char *s)
+{
+  return s ? s : "(null)";
+}
+
+static const char *
+nonempty (const char *s)
+{
+  return (s && !*s) ? "(empty)" : nonnull (s);
+}
+
+void
+lt_setenv (const char *name, const char *value)
+{
+  lt_debugprintf (__FILE__, __LINE__,
+                 "(lt_setenv) setting '%s' to '%s'\n",
+                  nonnull (name), nonnull (value));
+  {
+#ifdef HAVE_SETENV
+    /* always make a copy, for consistency with !HAVE_SETENV */
+    char *str = xstrdup (value);
+    setenv (name, str, 1);
+#else
+    int len = strlen (name) + 1 + strlen (value) + 1;
+    char *str = XMALLOC (char, len);
+    sprintf (str, "%s=%s", name, value);
+    if (putenv (str) != EXIT_SUCCESS)
+      {
+        XFREE (str);
+      }
+#endif
+  }
+}
+
+char *
+lt_extend_str (const char *orig_value, const char *add, int to_end)
+{
+  char *new_value;
+  if (orig_value && *orig_value)
+    {
+      int orig_value_len = strlen (orig_value);
+      int add_len = strlen (add);
+      new_value = XMALLOC (char, add_len + orig_value_len + 1);
+      if (to_end)
+        {
+          strcpy (new_value, orig_value);
+          strcpy (new_value + orig_value_len, add);
+        }
+      else
+        {
+          strcpy (new_value, add);
+          strcpy (new_value + add_len, orig_value);
+        }
+    }
+  else
+    {
+      new_value = xstrdup (add);
+    }
+  return new_value;
+}
+
+void
+lt_update_exe_path (const char *name, const char *value)
+{
+  lt_debugprintf (__FILE__, __LINE__,
+                 "(lt_update_exe_path) modifying '%s' by prepending '%s'\n",
+                  nonnull (name), nonnull (value));
+
+  if (name && *name && value && *value)
+    {
+      char *new_value = lt_extend_str (getenv (name), value, 0);
+      /* some systems can't cope with a ':'-terminated path #' */
+      int len = strlen (new_value);
+      while (((len = strlen (new_value)) > 0) && IS_PATH_SEPARATOR (new_value[len-1]))
+        {
+          new_value[len-1] = '\0';
+        }
+      lt_setenv (name, new_value);
+      XFREE (new_value);
+    }
+}
+
+void
+lt_update_lib_path (const char *name, const char *value)
+{
+  lt_debugprintf (__FILE__, __LINE__,
+                 "(lt_update_lib_path) modifying '%s' by prepending '%s'\n",
+                  nonnull (name), nonnull (value));
+
+  if (name && *name && value && *value)
+    {
+      char *new_value = lt_extend_str (getenv (name), value, 0);
+      lt_setenv (name, new_value);
+      XFREE (new_value);
+    }
+}
+
+EOF
+           case $host_os in
+             mingw*)
+               cat <<"EOF"
+
+/* Prepares an argument vector before calling spawn().
+   Note that spawn() does not by itself call the command interpreter
+     (getenv ("COMSPEC") != NULL ? getenv ("COMSPEC") :
+      ({ OSVERSIONINFO v; v.dwOSVersionInfoSize = sizeof(OSVERSIONINFO);
+         GetVersionEx(&v);
+         v.dwPlatformId == VER_PLATFORM_WIN32_NT;
+      }) ? "cmd.exe" : "command.com").
+   Instead it simply concatenates the arguments, separated by ' ', and calls
+   CreateProcess().  We must quote the arguments since Win32 CreateProcess()
+   interprets characters like ' ', '\t', '\\', '"' (but not '<' and '>') in a
+   special way:
+   - Space and tab are interpreted as delimiters. They are not treated as
+     delimiters if they are surrounded by double quotes: "...".
+   - Unescaped double quotes are removed from the input. Their only effect is
+     that within double quotes, space and tab are treated like normal
+     characters.
+   - Backslashes not followed by double quotes are not special.
+   - But 2*n+1 backslashes followed by a double quote become
+     n backslashes followed by a double quote (n >= 0):
+       \" -> "
+       \\\" -> \"
+       \\\\\" -> \\"
+ */
+#define SHELL_SPECIAL_CHARS "\"\\ \001\002\003\004\005\006\007\010\011\012\013\014\015\016\017\020\021\022\023\024\025\026\027\030\031\032\033\034\035\036\037"
+#define SHELL_SPACE_CHARS " \001\002\003\004\005\006\007\010\011\012\013\014\015\016\017\020\021\022\023\024\025\026\027\030\031\032\033\034\035\036\037"
+char **
+prepare_spawn (char **argv)
+{
+  size_t argc;
+  char **new_argv;
+  size_t i;
+
+  /* Count number of arguments.  */
+  for (argc = 0; argv[argc] != NULL; argc++)
+    ;
+
+  /* Allocate new argument vector.  */
+  new_argv = XMALLOC (char *, argc + 1);
+
+  /* Put quoted arguments into the new argument vector.  */
+  for (i = 0; i < argc; i++)
+    {
+      const char *string = argv[i];
+
+      if (string[0] == '\0')
+       new_argv[i] = xstrdup ("\"\"");
+      else if (strpbrk (string, SHELL_SPECIAL_CHARS) != NULL)
+       {
+         int quote_around = (strpbrk (string, SHELL_SPACE_CHARS) != NULL);
+         size_t length;
+         unsigned int backslashes;
+         const char *s;
+         char *quoted_string;
+         char *p;
+
+         length = 0;
+         backslashes = 0;
+         if (quote_around)
+           length++;
+         for (s = string; *s != '\0'; s++)
+           {
+             char c = *s;
+             if (c == '"')
+               length += backslashes + 1;
+             length++;
+             if (c == '\\')
+               backslashes++;
+             else
+               backslashes = 0;
+           }
+         if (quote_around)
+           length += backslashes + 1;
+
+         quoted_string = XMALLOC (char, length + 1);
+
+         p = quoted_string;
+         backslashes = 0;
+         if (quote_around)
+           *p++ = '"';
+         for (s = string; *s != '\0'; s++)
+           {
+             char c = *s;
+             if (c == '"')
+               {
+                 unsigned int j;
+                 for (j = backslashes + 1; j > 0; j--)
+                   *p++ = '\\';
+               }
+             *p++ = c;
+             if (c == '\\')
+               backslashes++;
+             else
+               backslashes = 0;
+           }
+         if (quote_around)
+           {
+             unsigned int j;
+             for (j = backslashes; j > 0; j--)
+               *p++ = '\\';
+             *p++ = '"';
+           }
+         *p = '\0';
+
+         new_argv[i] = quoted_string;
+       }
+      else
+       new_argv[i] = (char *) string;
+    }
+  new_argv[argc] = NULL;
+
+  return new_argv;
+}
+EOF
+               ;;
+           esac
+
+            cat <<"EOF"
+void lt_dump_script (FILE* f)
+{
+EOF
+           func_emit_wrapper yes |
+             $SED -n -e '
+s/^\(.\{79\}\)\(..*\)/\1\
+\2/
+h
+s/\([\\"]\)/\\\1/g
+s/$/\\n/
+s/\([^\n]*\).*/  fputs ("\1", f);/p
+g
+D'
+            cat <<"EOF"
+}
+EOF
+}
+# end: func_emit_cwrapperexe_src
+
+# func_win32_import_lib_p ARG
+# True if ARG is an import lib, as indicated by $file_magic_cmd
+func_win32_import_lib_p ()
+{
+    $opt_debug
+    case `eval $file_magic_cmd \"\$1\" 2>/dev/null | $SED -e 10q` in
+    *import*) : ;;
+    *) false ;;
+    esac
+}
+
+# func_mode_link arg...
+func_mode_link ()
+{
+    $opt_debug
+    case $host in
+    *-*-cygwin* | *-*-mingw* | *-*-pw32* | *-*-os2* | *-cegcc*)
+      # It is impossible to link a dll without this setting, and
+      # we shouldn't force the makefile maintainer to figure out
+      # which system we are compiling for in order to pass an extra
+      # flag for every libtool invocation.
+      # allow_undefined=no
+
+      # FIXME: Unfortunately, there are problems with the above when trying
+      # to make a dll which has undefined symbols, in which case not
+      # even a static library is built.  For now, we need to specify
+      # -no-undefined on the libtool link line when we can be certain
+      # that all symbols are satisfied, otherwise we get a static library.
+      allow_undefined=yes
+      ;;
+    *)
+      allow_undefined=yes
+      ;;
+    esac
+    libtool_args=$nonopt
+    base_compile="$nonopt $@"
+    compile_command=$nonopt
+    finalize_command=$nonopt
+
+    compile_rpath=
+    finalize_rpath=
+    compile_shlibpath=
+    finalize_shlibpath=
+    convenience=
+    old_convenience=
+    deplibs=
+    old_deplibs=
+    compiler_flags=
+    linker_flags=
+    dllsearchpath=
+    lib_search_path=`pwd`
+    inst_prefix_dir=
+    new_inherited_linker_flags=
+
+    avoid_version=no
+    bindir=
+    dlfiles=
+    dlprefiles=
+    dlself=no
+    export_dynamic=no
+    export_symbols=
+    export_symbols_regex=
+    generated=
+    libobjs=
+    ltlibs=
+    module=no
+    no_install=no
+    objs=
+    non_pic_objects=
+    precious_files_regex=
+    prefer_static_libs=no
+    preload=no
+    prev=
+    prevarg=
+    release=
+    rpath=
+    xrpath=
+    perm_rpath=
+    temp_rpath=
+    thread_safe=no
+    vinfo=
+    vinfo_number=no
+    weak_libs=
+    single_module="${wl}-single_module"
+    func_infer_tag $base_compile
+
+    # We need to know -static, to get the right output filenames.
+    for arg
+    do
+      case $arg in
+      -shared)
+       test "$build_libtool_libs" != yes && \
+         func_fatal_configuration "can not build a shared library"
+       build_old_libs=no
+       break
+       ;;
+      -all-static | -static | -static-libtool-libs)
+       case $arg in
+       -all-static)
+         if test "$build_libtool_libs" = yes && test -z "$link_static_flag"; then
+           func_warning "complete static linking is impossible in this configuration"
+         fi
+         if test -n "$link_static_flag"; then
+           dlopen_self=$dlopen_self_static
+         fi
+         prefer_static_libs=yes
+         ;;
+       -static)
+         if test -z "$pic_flag" && test -n "$link_static_flag"; then
+           dlopen_self=$dlopen_self_static
+         fi
+         prefer_static_libs=built
+         ;;
+       -static-libtool-libs)
+         if test -z "$pic_flag" && test -n "$link_static_flag"; then
+           dlopen_self=$dlopen_self_static
+         fi
+         prefer_static_libs=yes
+         ;;
+       esac
+       build_libtool_libs=no
+       build_old_libs=yes
+       break
+       ;;
+      esac
+    done
+
+    # See if our shared archives depend on static archives.
+    test -n "$old_archive_from_new_cmds" && build_old_libs=yes
+
+    # Go through the arguments, transforming them on the way.
+    while test "$#" -gt 0; do
+      arg="$1"
+      shift
+      func_quote_for_eval "$arg"
+      qarg=$func_quote_for_eval_unquoted_result
+      func_append libtool_args " $func_quote_for_eval_result"
+
+      # If the previous option needs an argument, assign it.
+      if test -n "$prev"; then
+       case $prev in
+       output)
+         func_append compile_command " @OUTPUT@"
+         func_append finalize_command " @OUTPUT@"
+         ;;
+       esac
+
+       case $prev in
+       bindir)
+         bindir="$arg"
+         prev=
+         continue
+         ;;
+       dlfiles|dlprefiles)
+         if test "$preload" = no; then
+           # Add the symbol object into the linking commands.
+           func_append compile_command " @SYMFILE@"
+           func_append finalize_command " @SYMFILE@"
+           preload=yes
+         fi
+         case $arg in
+         *.la | *.lo) ;;  # We handle these cases below.
+         force)
+           if test "$dlself" = no; then
+             dlself=needless
+             export_dynamic=yes
+           fi
+           prev=
+           continue
+           ;;
+         self)
+           if test "$prev" = dlprefiles; then
+             dlself=yes
+           elif test "$prev" = dlfiles && test "$dlopen_self" != yes; then
+             dlself=yes
+           else
+             dlself=needless
+             export_dynamic=yes
+           fi
+           prev=
+           continue
+           ;;
+         *)
+           if test "$prev" = dlfiles; then
+             func_append dlfiles " $arg"
+           else
+             func_append dlprefiles " $arg"
+           fi
+           prev=
+           continue
+           ;;
+         esac
+         ;;
+       expsyms)
+         export_symbols="$arg"
+         test -f "$arg" \
+           || func_fatal_error "symbol file \`$arg' does not exist"
+         prev=
+         continue
+         ;;
+       expsyms_regex)
+         export_symbols_regex="$arg"
+         prev=
+         continue
+         ;;
+       framework)
+         case $host in
+           *-*-darwin*)
+             case "$deplibs " in
+               *" $qarg.ltframework "*) ;;
+               *) func_append deplibs " $qarg.ltframework" # this is fixed later
+                  ;;
+             esac
+             ;;
+         esac
+         prev=
+         continue
+         ;;
+       inst_prefix)
+         inst_prefix_dir="$arg"
+         prev=
+         continue
+         ;;
+       objectlist)
+         if test -f "$arg"; then
+           save_arg=$arg
+           moreargs=
+           for fil in `cat "$save_arg"`
+           do
+#            func_append moreargs " $fil"
+             arg=$fil
+             # A libtool-controlled object.
+
+             # Check to see that this really is a libtool object.
+             if func_lalib_unsafe_p "$arg"; then
+               pic_object=
+               non_pic_object=
+
+               # Read the .lo file
+               func_source "$arg"
+
+               if test -z "$pic_object" ||
+                  test -z "$non_pic_object" ||
+                  test "$pic_object" = none &&
+                  test "$non_pic_object" = none; then
+                 func_fatal_error "cannot find name of object for \`$arg'"
+               fi
+
+               # Extract subdirectory from the argument.
+               func_dirname "$arg" "/" ""
+               xdir="$func_dirname_result"
+
+               if test "$pic_object" != none; then
+                 # Prepend the subdirectory the object is found in.
+                 pic_object="$xdir$pic_object"
+
+                 if test "$prev" = dlfiles; then
+                   if test "$build_libtool_libs" = yes && test "$dlopen_support" = yes; then
+                     func_append dlfiles " $pic_object"
+                     prev=
+                     continue
+                   else
+                     # If libtool objects are unsupported, then we need to preload.
+                     prev=dlprefiles
+                   fi
+                 fi
+
+                 # CHECK ME:  I think I busted this.  -Ossama
+                 if test "$prev" = dlprefiles; then
+                   # Preload the old-style object.
+                   func_append dlprefiles " $pic_object"
+                   prev=
+                 fi
+
+                 # A PIC object.
+                 func_append libobjs " $pic_object"
+                 arg="$pic_object"
+               fi
+
+               # Non-PIC object.
+               if test "$non_pic_object" != none; then
+                 # Prepend the subdirectory the object is found in.
+                 non_pic_object="$xdir$non_pic_object"
+
+                 # A standard non-PIC object
+                 func_append non_pic_objects " $non_pic_object"
+                 if test -z "$pic_object" || test "$pic_object" = none ; then
+                   arg="$non_pic_object"
+                 fi
+               else
+                 # If the PIC object exists, use it instead.
+                 # $xdir was prepended to $pic_object above.
+                 non_pic_object="$pic_object"
+                 func_append non_pic_objects " $non_pic_object"
+               fi
+             else
+               # Only an error if not doing a dry-run.
+               if $opt_dry_run; then
+                 # Extract subdirectory from the argument.
+                 func_dirname "$arg" "/" ""
+                 xdir="$func_dirname_result"
+
+                 func_lo2o "$arg"
+                 pic_object=$xdir$objdir/$func_lo2o_result
+                 non_pic_object=$xdir$func_lo2o_result
+                 func_append libobjs " $pic_object"
+                 func_append non_pic_objects " $non_pic_object"
+               else
+                 func_fatal_error "\`$arg' is not a valid libtool object"
+               fi
+             fi
+           done
+         else
+           func_fatal_error "link input file \`$arg' does not exist"
+         fi
+         arg=$save_arg
+         prev=
+         continue
+         ;;
+       precious_regex)
+         precious_files_regex="$arg"
+         prev=
+         continue
+         ;;
+       release)
+         release="-$arg"
+         prev=
+         continue
+         ;;
+       rpath | xrpath)
+         # We need an absolute path.
+         case $arg in
+         [\\/]* | [A-Za-z]:[\\/]*) ;;
+         *)
+           func_fatal_error "only absolute run-paths are allowed"
+           ;;
+         esac
+         if test "$prev" = rpath; then
+           case "$rpath " in
+           *" $arg "*) ;;
+           *) func_append rpath " $arg" ;;
+           esac
+         else
+           case "$xrpath " in
+           *" $arg "*) ;;
+           *) func_append xrpath " $arg" ;;
+           esac
+         fi
+         prev=
+         continue
+         ;;
+       shrext)
+         shrext_cmds="$arg"
+         prev=
+         continue
+         ;;
+       weak)
+         func_append weak_libs " $arg"
+         prev=
+         continue
+         ;;
+       xcclinker)
+         func_append linker_flags " $qarg"
+         func_append compiler_flags " $qarg"
+         prev=
+         func_append compile_command " $qarg"
+         func_append finalize_command " $qarg"
+         continue
+         ;;
+       xcompiler)
+         func_append compiler_flags " $qarg"
+         prev=
+         func_append compile_command " $qarg"
+         func_append finalize_command " $qarg"
+         continue
+         ;;
+       xlinker)
+         func_append linker_flags " $qarg"
+         func_append compiler_flags " $wl$qarg"
+         prev=
+         func_append compile_command " $wl$qarg"
+         func_append finalize_command " $wl$qarg"
+         continue
+         ;;
+       *)
+         eval "$prev=\"\$arg\""
+         prev=
+         continue
+         ;;
+       esac
+      fi # test -n "$prev"
+
+      prevarg="$arg"
+
+      case $arg in
+      -all-static)
+       if test -n "$link_static_flag"; then
+         # See comment for -static flag below, for more details.
+         func_append compile_command " $link_static_flag"
+         func_append finalize_command " $link_static_flag"
+       fi
+       continue
+       ;;
+
+      -allow-undefined)
+       # FIXME: remove this flag sometime in the future.
+       func_fatal_error "\`-allow-undefined' must not be used because it is the default"
+       ;;
+
+      -avoid-version)
+       avoid_version=yes
+       continue
+       ;;
+
+      -bindir)
+       prev=bindir
+       continue
+       ;;
+
+      -dlopen)
+       prev=dlfiles
+       continue
+       ;;
+
+      -dlpreopen)
+       prev=dlprefiles
+       continue
+       ;;
+
+      -export-dynamic)
+       export_dynamic=yes
+       continue
+       ;;
+
+      -export-symbols | -export-symbols-regex)
+       if test -n "$export_symbols" || test -n "$export_symbols_regex"; then
+         func_fatal_error "more than one -exported-symbols argument is not allowed"
+       fi
+       if test "X$arg" = "X-export-symbols"; then
+         prev=expsyms
+       else
+         prev=expsyms_regex
+       fi
+       continue
+       ;;
+
+      -framework)
+       prev=framework
+       continue
+       ;;
+
+      -inst-prefix-dir)
+       prev=inst_prefix
+       continue
+       ;;
+
+      # The native IRIX linker understands -LANG:*, -LIST:* and -LNO:*
+      # so, if we see these flags be careful not to treat them like -L
+      -L[A-Z][A-Z]*:*)
+       case $with_gcc/$host in
+       no/*-*-irix* | /*-*-irix*)
+         func_append compile_command " $arg"
+         func_append finalize_command " $arg"
+         ;;
+       esac
+       continue
+       ;;
+
+      -L*)
+       func_stripname "-L" '' "$arg"
+       if test -z "$func_stripname_result"; then
+         if test "$#" -gt 0; then
+           func_fatal_error "require no space between \`-L' and \`$1'"
+         else
+           func_fatal_error "need path for \`-L' option"
+         fi
+       fi
+       func_resolve_sysroot "$func_stripname_result"
+       dir=$func_resolve_sysroot_result
+       # We need an absolute path.
+       case $dir in
+       [\\/]* | [A-Za-z]:[\\/]*) ;;
+       *)
+         absdir=`cd "$dir" && pwd`
+         test -z "$absdir" && \
+           func_fatal_error "cannot determine absolute directory name of \`$dir'"
+         dir="$absdir"
+         ;;
+       esac
+       case "$deplibs " in
+       *" -L$dir "* | *" $arg "*)
+         # Will only happen for absolute or sysroot arguments
+         ;;
+       *)
+         # Preserve sysroot, but never include relative directories
+         case $dir in
+           [\\/]* | [A-Za-z]:[\\/]* | =*) func_append deplibs " $arg" ;;
+           *) func_append deplibs " -L$dir" ;;
+         esac
+         func_append lib_search_path " $dir"
+         ;;
+       esac
+       case $host in
+       *-*-cygwin* | *-*-mingw* | *-*-pw32* | *-*-os2* | *-cegcc*)
+         testbindir=`$ECHO "$dir" | $SED 's*/lib$*/bin*'`
+         case :$dllsearchpath: in
+         *":$dir:"*) ;;
+         ::) dllsearchpath=$dir;;
+         *) func_append dllsearchpath ":$dir";;
+         esac
+         case :$dllsearchpath: in
+         *":$testbindir:"*) ;;
+         ::) dllsearchpath=$testbindir;;
+         *) func_append dllsearchpath ":$testbindir";;
+         esac
+         ;;
+       esac
+       continue
+       ;;
+
+      -l*)
+       if test "X$arg" = "X-lc" || test "X$arg" = "X-lm"; then
+         case $host in
+         *-*-cygwin* | *-*-mingw* | *-*-pw32* | *-*-beos* | *-cegcc* | *-*-haiku*)
+           # These systems don't actually have a C or math library (as such)
+           continue
+           ;;
+         *-*-os2*)
+           # These systems don't actually have a C library (as such)
+           test "X$arg" = "X-lc" && continue
+           ;;
+         *-*-openbsd* | *-*-freebsd* | *-*-dragonfly*)
+           # Do not include libc due to us having libc/libc_r.
+           test "X$arg" = "X-lc" && continue
+           ;;
+         *-*-rhapsody* | *-*-darwin1.[012])
+           # Rhapsody C and math libraries are in the System framework
+           func_append deplibs " System.ltframework"
+           continue
+           ;;
+         *-*-sco3.2v5* | *-*-sco5v6*)
+           # Causes problems with __ctype
+           test "X$arg" = "X-lc" && continue
+           ;;
+         *-*-sysv4.2uw2* | *-*-sysv5* | *-*-unixware* | *-*-OpenUNIX*)
+           # Compiler inserts libc in the correct place for threads to work
+           test "X$arg" = "X-lc" && continue
+           ;;
+         esac
+       elif test "X$arg" = "X-lc_r"; then
+        case $host in
+        *-*-openbsd* | *-*-freebsd* | *-*-dragonfly*)
+          # Do not include libc_r directly, use -pthread flag.
+          continue
+          ;;
+        esac
+       fi
+       func_append deplibs " $arg"
+       continue
+       ;;
+
+      -module)
+       module=yes
+       continue
+       ;;
+
+      # Tru64 UNIX uses -model [arg] to determine the layout of C++
+      # classes, name mangling, and exception handling.
+      # Darwin uses the -arch flag to determine output architecture.
+      -model|-arch|-isysroot|--sysroot)
+       func_append compiler_flags " $arg"
+       func_append compile_command " $arg"
+       func_append finalize_command " $arg"
+       prev=xcompiler
+       continue
+       ;;
+
+      -mt|-mthreads|-kthread|-Kthread|-pthread|-pthreads|--thread-safe \
+      |-threads|-fopenmp|-openmp|-mp|-xopenmp|-omp|-qsmp=*)
+       func_append compiler_flags " $arg"
+       func_append compile_command " $arg"
+       func_append finalize_command " $arg"
+       case "$new_inherited_linker_flags " in
+           *" $arg "*) ;;
+           * ) func_append new_inherited_linker_flags " $arg" ;;
+       esac
+       continue
+       ;;
+
+      -multi_module)
+       single_module="${wl}-multi_module"
+       continue
+       ;;
+
+      -no-fast-install)
+       fast_install=no
+       continue
+       ;;
+
+      -no-install)
+       case $host in
+       *-*-cygwin* | *-*-mingw* | *-*-pw32* | *-*-os2* | *-*-darwin* | *-cegcc*)
+         # The PATH hackery in wrapper scripts is required on Windows
+         # and Darwin in order for the loader to find any dlls it needs.
+         func_warning "\`-no-install' is ignored for $host"
+         func_warning "assuming \`-no-fast-install' instead"
+         fast_install=no
+         ;;
+       *) no_install=yes ;;
+       esac
+       continue
+       ;;
+
+      -no-undefined)
+       allow_undefined=no
+       continue
+       ;;
+
+      -objectlist)
+       prev=objectlist
+       continue
+       ;;
+
+      -o) prev=output ;;
+
+      -precious-files-regex)
+       prev=precious_regex
+       continue
+       ;;
+
+      -release)
+       prev=release
+       continue
+       ;;
+
+      -rpath)
+       prev=rpath
+       continue
+       ;;
+
+      -R)
+       prev=xrpath
+       continue
+       ;;
+
+      -R*)
+       func_stripname '-R' '' "$arg"
+       dir=$func_stripname_result
+       # We need an absolute path.
+       case $dir in
+       [\\/]* | [A-Za-z]:[\\/]*) ;;
+       =*)
+         func_stripname '=' '' "$dir"
+         dir=$lt_sysroot$func_stripname_result
+         ;;
+       *)
+         func_fatal_error "only absolute run-paths are allowed"
+         ;;
+       esac
+       case "$xrpath " in
+       *" $dir "*) ;;
+       *) func_append xrpath " $dir" ;;
+       esac
+       continue
+       ;;
+
+      -shared)
+       # The effects of -shared are defined in a previous loop.
+       continue
+       ;;
+
+      -shrext)
+       prev=shrext
+       continue
+       ;;
+
+      -static | -static-libtool-libs)
+       # The effects of -static are defined in a previous loop.
+       # We used to do the same as -all-static on platforms that
+       # didn't have a PIC flag, but the assumption that the effects
+       # would be equivalent was wrong.  It would break on at least
+       # Digital Unix and AIX.
+       continue
+       ;;
+
+      -thread-safe)
+       thread_safe=yes
+       continue
+       ;;
+
+      -version-info)
+       prev=vinfo
+       continue
+       ;;
+
+      -version-number)
+       prev=vinfo
+       vinfo_number=yes
+       continue
+       ;;
+
+      -weak)
+        prev=weak
+       continue
+       ;;
+
+      -Wc,*)
+       func_stripname '-Wc,' '' "$arg"
+       args=$func_stripname_result
+       arg=
+       save_ifs="$IFS"; IFS=','
+       for flag in $args; do
+         IFS="$save_ifs"
+          func_quote_for_eval "$flag"
+         func_append arg " $func_quote_for_eval_result"
+         func_append compiler_flags " $func_quote_for_eval_result"
+       done
+       IFS="$save_ifs"
+       func_stripname ' ' '' "$arg"
+       arg=$func_stripname_result
+       ;;
+
+      -Wl,*)
+       func_stripname '-Wl,' '' "$arg"
+       args=$func_stripname_result
+       arg=
+       save_ifs="$IFS"; IFS=','
+       for flag in $args; do
+         IFS="$save_ifs"
+          func_quote_for_eval "$flag"
+         func_append arg " $wl$func_quote_for_eval_result"
+         func_append compiler_flags " $wl$func_quote_for_eval_result"
+         func_append linker_flags " $func_quote_for_eval_result"
+       done
+       IFS="$save_ifs"
+       func_stripname ' ' '' "$arg"
+       arg=$func_stripname_result
+       ;;
+
+      -Xcompiler)
+       prev=xcompiler
+       continue
+       ;;
+
+      -Xlinker)
+       prev=xlinker
+       continue
+       ;;
+
+      -XCClinker)
+       prev=xcclinker
+       continue
+       ;;
+
+      # -msg_* for osf cc
+      -msg_*)
+       func_quote_for_eval "$arg"
+       arg="$func_quote_for_eval_result"
+       ;;
+
+      # Flags to be passed through unchanged, with rationale:
+      # -64, -mips[0-9]      enable 64-bit mode for the SGI compiler
+      # -r[0-9][0-9]*        specify processor for the SGI compiler
+      # -xarch=*, -xtarget=* enable 64-bit mode for the Sun compiler
+      # +DA*, +DD*           enable 64-bit mode for the HP compiler
+      # -q*                  compiler args for the IBM compiler
+      # -m*, -t[45]*, -txscale* architecture-specific flags for GCC
+      # -F/path              path to uninstalled frameworks, gcc on darwin
+      # -p, -pg, --coverage, -fprofile-*  profiling flags for GCC
+      # @file                GCC response files
+      # -tp=*                Portland pgcc target processor selection
+      # --sysroot=*          for sysroot support
+      # -O*, -flto*, -fwhopr*, -fuse-linker-plugin GCC link-time optimization
+      -64|-mips[0-9]|-r[0-9][0-9]*|-xarch=*|-xtarget=*|+DA*|+DD*|-q*|-m*| \
+      -t[45]*|-txscale*|-p|-pg|--coverage|-fprofile-*|-F*|@*|-tp=*|--sysroot=*| \
+      -O*|-flto*|-fwhopr*|-fuse-linker-plugin)
+        func_quote_for_eval "$arg"
+       arg="$func_quote_for_eval_result"
+        func_append compile_command " $arg"
+        func_append finalize_command " $arg"
+        func_append compiler_flags " $arg"
+        continue
+        ;;
+
+      # Some other compiler flag.
+      -* | +*)
+        func_quote_for_eval "$arg"
+       arg="$func_quote_for_eval_result"
+       ;;
+
+      *.$objext)
+       # A standard object.
+       func_append objs " $arg"
+       ;;
+
+      *.lo)
+       # A libtool-controlled object.
+
+       # Check to see that this really is a libtool object.
+       if func_lalib_unsafe_p "$arg"; then
+         pic_object=
+         non_pic_object=
+
+         # Read the .lo file
+         func_source "$arg"
+
+         if test -z "$pic_object" ||
+            test -z "$non_pic_object" ||
+            test "$pic_object" = none &&
+            test "$non_pic_object" = none; then
+           func_fatal_error "cannot find name of object for \`$arg'"
+         fi
+
+         # Extract subdirectory from the argument.
+         func_dirname "$arg" "/" ""
+         xdir="$func_dirname_result"
+
+         if test "$pic_object" != none; then
+           # Prepend the subdirectory the object is found in.
+           pic_object="$xdir$pic_object"
+
+           if test "$prev" = dlfiles; then
+             if test "$build_libtool_libs" = yes && test "$dlopen_support" = yes; then
+               func_append dlfiles " $pic_object"
+               prev=
+               continue
+             else
+               # If libtool objects are unsupported, then we need to preload.
+               prev=dlprefiles
+             fi
+           fi
+
+           # CHECK ME:  I think I busted this.  -Ossama
+           if test "$prev" = dlprefiles; then
+             # Preload the old-style object.
+             func_append dlprefiles " $pic_object"
+             prev=
+           fi
+
+           # A PIC object.
+           func_append libobjs " $pic_object"
+           arg="$pic_object"
+         fi
+
+         # Non-PIC object.
+         if test "$non_pic_object" != none; then
+           # Prepend the subdirectory the object is found in.
+           non_pic_object="$xdir$non_pic_object"
+
+           # A standard non-PIC object
+           func_append non_pic_objects " $non_pic_object"
+           if test -z "$pic_object" || test "$pic_object" = none ; then
+             arg="$non_pic_object"
+           fi
+         else
+           # If the PIC object exists, use it instead.
+           # $xdir was prepended to $pic_object above.
+           non_pic_object="$pic_object"
+           func_append non_pic_objects " $non_pic_object"
+         fi
+       else
+         # Only an error if not doing a dry-run.
+         if $opt_dry_run; then
+           # Extract subdirectory from the argument.
+           func_dirname "$arg" "/" ""
+           xdir="$func_dirname_result"
+
+           func_lo2o "$arg"
+           pic_object=$xdir$objdir/$func_lo2o_result
+           non_pic_object=$xdir$func_lo2o_result
+           func_append libobjs " $pic_object"
+           func_append non_pic_objects " $non_pic_object"
+         else
+           func_fatal_error "\`$arg' is not a valid libtool object"
+         fi
+       fi
+       ;;
+
+      *.$libext)
+       # An archive.
+       func_append deplibs " $arg"
+       func_append old_deplibs " $arg"
+       continue
+       ;;
+
+      *.la)
+       # A libtool-controlled library.
+
+       func_resolve_sysroot "$arg"
+       if test "$prev" = dlfiles; then
+         # This library was specified with -dlopen.
+         func_append dlfiles " $func_resolve_sysroot_result"
+         prev=
+       elif test "$prev" = dlprefiles; then
+         # The library was specified with -dlpreopen.
+         func_append dlprefiles " $func_resolve_sysroot_result"
+         prev=
+       else
+         func_append deplibs " $func_resolve_sysroot_result"
+       fi
+       continue
+       ;;
+
+      # Some other compiler argument.
+      *)
+       # Unknown arguments in both finalize_command and compile_command need
+       # to be aesthetically quoted because they are evaled later.
+       func_quote_for_eval "$arg"
+       arg="$func_quote_for_eval_result"
+       ;;
+      esac # arg
+
+      # Now actually substitute the argument into the commands.
+      if test -n "$arg"; then
+       func_append compile_command " $arg"
+       func_append finalize_command " $arg"
+      fi
+    done # argument parsing loop
+
+    test -n "$prev" && \
+      func_fatal_help "the \`$prevarg' option requires an argument"
+
+    if test "$export_dynamic" = yes && test -n "$export_dynamic_flag_spec"; then
+      eval arg=\"$export_dynamic_flag_spec\"
+      func_append compile_command " $arg"
+      func_append finalize_command " $arg"
+    fi
+
+    oldlibs=
+    # calculate the name of the file, without its directory
+    func_basename "$output"
+    outputname="$func_basename_result"
+    libobjs_save="$libobjs"
+
+    if test -n "$shlibpath_var"; then
+      # get the directories listed in $shlibpath_var
+      eval shlib_search_path=\`\$ECHO \"\${$shlibpath_var}\" \| \$SED \'s/:/ /g\'\`
+    else
+      shlib_search_path=
+    fi
+    eval sys_lib_search_path=\"$sys_lib_search_path_spec\"
+    eval sys_lib_dlsearch_path=\"$sys_lib_dlsearch_path_spec\"
+
+    func_dirname "$output" "/" ""
+    output_objdir="$func_dirname_result$objdir"
+    func_to_tool_file "$output_objdir/"
+    tool_output_objdir=$func_to_tool_file_result
+    # Create the object directory.
+    func_mkdir_p "$output_objdir"
+
+    # Determine the type of output
+    case $output in
+    "")
+      func_fatal_help "you must specify an output file"
+      ;;
+    *.$libext) linkmode=oldlib ;;
+    *.lo | *.$objext) linkmode=obj ;;
+    *.la) linkmode=lib ;;
+    *) linkmode=prog ;; # Anything else should be a program.
+    esac
+
+    specialdeplibs=
+
+    libs=
+    # Find all interdependent deplibs by searching for libraries
+    # that are linked more than once (e.g. -la -lb -la)
+    for deplib in $deplibs; do
+      if $opt_preserve_dup_deps ; then
+       case "$libs " in
+       *" $deplib "*) func_append specialdeplibs " $deplib" ;;
+       esac
+      fi
+      func_append libs " $deplib"
+    done
+
+    if test "$linkmode" = lib; then
+      libs="$predeps $libs $compiler_lib_search_path $postdeps"
+
+      # Compute libraries that are listed more than once in $predeps
+      # $postdeps and mark them as special (i.e., whose duplicates are
+      # not to be eliminated).
+      pre_post_deps=
+      if $opt_duplicate_compiler_generated_deps; then
+       for pre_post_dep in $predeps $postdeps; do
+         case "$pre_post_deps " in
+         *" $pre_post_dep "*) func_append specialdeplibs " $pre_post_deps" ;;
+         esac
+         func_append pre_post_deps " $pre_post_dep"
+       done
+      fi
+      pre_post_deps=
+    fi
+
+    deplibs=
+    newdependency_libs=
+    newlib_search_path=
+    need_relink=no # whether we're linking any uninstalled libtool libraries
+    notinst_deplibs= # not-installed libtool libraries
+    notinst_path= # paths that contain not-installed libtool libraries
+
+    case $linkmode in
+    lib)
+       passes="conv dlpreopen link"
+       for file in $dlfiles $dlprefiles; do
+         case $file in
+         *.la) ;;
+         *)
+           func_fatal_help "libraries can \`-dlopen' only libtool libraries: $file"
+           ;;
+         esac
+       done
+       ;;
+    prog)
+       compile_deplibs=
+       finalize_deplibs=
+       alldeplibs=no
+       newdlfiles=
+       newdlprefiles=
+       passes="conv scan dlopen dlpreopen link"
+       ;;
+    *)  passes="conv"
+       ;;
+    esac
+
+    for pass in $passes; do
+      # The preopen pass in lib mode reverses $deplibs; put it back here
+      # so that -L comes before libs that need it for instance...
+      if test "$linkmode,$pass" = "lib,link"; then
+       ## FIXME: Find the place where the list is rebuilt in the wrong
+       ##        order, and fix it there properly
+        tmp_deplibs=
+       for deplib in $deplibs; do
+         tmp_deplibs="$deplib $tmp_deplibs"
+       done
+       deplibs="$tmp_deplibs"
+      fi
+
+      if test "$linkmode,$pass" = "lib,link" ||
+        test "$linkmode,$pass" = "prog,scan"; then
+       libs="$deplibs"
+       deplibs=
+      fi
+      if test "$linkmode" = prog; then
+       case $pass in
+       dlopen) libs="$dlfiles" ;;
+       dlpreopen) libs="$dlprefiles" ;;
+       link)
+         libs="$deplibs %DEPLIBS%"
+         test "X$link_all_deplibs" != Xno && libs="$libs $dependency_libs"
+         ;;
+       esac
+      fi
+      if test "$linkmode,$pass" = "lib,dlpreopen"; then
+       # Collect and forward deplibs of preopened libtool libs
+       for lib in $dlprefiles; do
+         # Ignore non-libtool-libs
+         dependency_libs=
+         func_resolve_sysroot "$lib"
+         case $lib in
+         *.la) func_source "$func_resolve_sysroot_result" ;;
+         esac
+
+         # Collect preopened libtool deplibs, except any this library
+         # has declared as weak libs
+         for deplib in $dependency_libs; do
+           func_basename "$deplib"
+            deplib_base=$func_basename_result
+           case " $weak_libs " in
+           *" $deplib_base "*) ;;
+           *) func_append deplibs " $deplib" ;;
+           esac
+         done
+       done
+       libs="$dlprefiles"
+      fi
+      if test "$pass" = dlopen; then
+       # Collect dlpreopened libraries
+       save_deplibs="$deplibs"
+       deplibs=
+      fi
+
+      for deplib in $libs; do
+       lib=
+       found=no
+       case $deplib in
+       -mt|-mthreads|-kthread|-Kthread|-pthread|-pthreads|--thread-safe \
+        |-threads|-fopenmp|-openmp|-mp|-xopenmp|-omp|-qsmp=*)
+         if test "$linkmode,$pass" = "prog,link"; then
+           compile_deplibs="$deplib $compile_deplibs"
+           finalize_deplibs="$deplib $finalize_deplibs"
+         else
+           func_append compiler_flags " $deplib"
+           if test "$linkmode" = lib ; then
+               case "$new_inherited_linker_flags " in
+                   *" $deplib "*) ;;
+                   * ) func_append new_inherited_linker_flags " $deplib" ;;
+               esac
+           fi
+         fi
+         continue
+         ;;
+       -l*)
+         if test "$linkmode" != lib && test "$linkmode" != prog; then
+           func_warning "\`-l' is ignored for archives/objects"
+           continue
+         fi
+         func_stripname '-l' '' "$deplib"
+         name=$func_stripname_result
+         if test "$linkmode" = lib; then
+           searchdirs="$newlib_search_path $lib_search_path $compiler_lib_search_dirs $sys_lib_search_path $shlib_search_path"
+         else
+           searchdirs="$newlib_search_path $lib_search_path $sys_lib_search_path $shlib_search_path"
+         fi
+         for searchdir in $searchdirs; do
+           for search_ext in .la $std_shrext .so .a; do
+             # Search the libtool library
+             lib="$searchdir/lib${name}${search_ext}"
+             if test -f "$lib"; then
+               if test "$search_ext" = ".la"; then
+                 found=yes
+               else
+                 found=no
+               fi
+               break 2
+             fi
+           done
+         done
+         if test "$found" != yes; then
+           # deplib doesn't seem to be a libtool library
+           if test "$linkmode,$pass" = "prog,link"; then
+             compile_deplibs="$deplib $compile_deplibs"
+             finalize_deplibs="$deplib $finalize_deplibs"
+           else
+             deplibs="$deplib $deplibs"
+             test "$linkmode" = lib && newdependency_libs="$deplib $newdependency_libs"
+           fi
+           continue
+         else # deplib is a libtool library
+           # If $allow_libtool_libs_with_static_runtimes && $deplib is a stdlib,
+           # We need to do some special things here, and not later.
+           if test "X$allow_libtool_libs_with_static_runtimes" = "Xyes" ; then
+             case " $predeps $postdeps " in
+             *" $deplib "*)
+               if func_lalib_p "$lib"; then
+                 library_names=
+                 old_library=
+                 func_source "$lib"
+                 for l in $old_library $library_names; do
+                   ll="$l"
+                 done
+                 if test "X$ll" = "X$old_library" ; then # only static version available
+                   found=no
+                   func_dirname "$lib" "" "."
+                   ladir="$func_dirname_result"
+                   lib=$ladir/$old_library
+                   if test "$linkmode,$pass" = "prog,link"; then
+                     compile_deplibs="$deplib $compile_deplibs"
+                     finalize_deplibs="$deplib $finalize_deplibs"
+                   else
+                     deplibs="$deplib $deplibs"
+                     test "$linkmode" = lib && newdependency_libs="$deplib $newdependency_libs"
+                   fi
+                   continue
+                 fi
+               fi
+               ;;
+             *) ;;
+             esac
+           fi
+         fi
+         ;; # -l
+       *.ltframework)
+         if test "$linkmode,$pass" = "prog,link"; then
+           compile_deplibs="$deplib $compile_deplibs"
+           finalize_deplibs="$deplib $finalize_deplibs"
+         else
+           deplibs="$deplib $deplibs"
+           if test "$linkmode" = lib ; then
+               case "$new_inherited_linker_flags " in
+                   *" $deplib "*) ;;
+                   * ) func_append new_inherited_linker_flags " $deplib" ;;
+               esac
+           fi
+         fi
+         continue
+         ;;
+       -L*)
+         case $linkmode in
+         lib)
+           deplibs="$deplib $deplibs"
+           test "$pass" = conv && continue
+           newdependency_libs="$deplib $newdependency_libs"
+           func_stripname '-L' '' "$deplib"
+           func_resolve_sysroot "$func_stripname_result"
+           func_append newlib_search_path " $func_resolve_sysroot_result"
+           ;;
+         prog)
+           if test "$pass" = conv; then
+             deplibs="$deplib $deplibs"
+             continue
+           fi
+           if test "$pass" = scan; then
+             deplibs="$deplib $deplibs"
+           else
+             compile_deplibs="$deplib $compile_deplibs"
+             finalize_deplibs="$deplib $finalize_deplibs"
+           fi
+           func_stripname '-L' '' "$deplib"
+           func_resolve_sysroot "$func_stripname_result"
+           func_append newlib_search_path " $func_resolve_sysroot_result"
+           ;;
+         *)
+           func_warning "\`-L' is ignored for archives/objects"
+           ;;
+         esac # linkmode
+         continue
+         ;; # -L
+       -R*)
+         if test "$pass" = link; then
+           func_stripname '-R' '' "$deplib"
+           func_resolve_sysroot "$func_stripname_result"
+           dir=$func_resolve_sysroot_result
+           # Make sure the xrpath contains only unique directories.
+           case "$xrpath " in
+           *" $dir "*) ;;
+           *) func_append xrpath " $dir" ;;
+           esac
+         fi
+         deplibs="$deplib $deplibs"
+         continue
+         ;;
+       *.la)
+         func_resolve_sysroot "$deplib"
+         lib=$func_resolve_sysroot_result
+         ;;
+       *.$libext)
+         if test "$pass" = conv; then
+           deplibs="$deplib $deplibs"
+           continue
+         fi
+         case $linkmode in
+         lib)
+           # Linking convenience modules into shared libraries is allowed,
+           # but linking other static libraries is non-portable.
+           case " $dlpreconveniencelibs " in
+           *" $deplib "*) ;;
+           *)
+             valid_a_lib=no
+             case $deplibs_check_method in
+               match_pattern*)
+                 set dummy $deplibs_check_method; shift
+                 match_pattern_regex=`expr "$deplibs_check_method" : "$1 \(.*\)"`
+                 if eval "\$ECHO \"$deplib\"" 2>/dev/null | $SED 10q \
+                   | $EGREP "$match_pattern_regex" > /dev/null; then
+                   valid_a_lib=yes
+                 fi
+               ;;
+               pass_all)
+                 valid_a_lib=yes
+               ;;
+             esac
+             if test "$valid_a_lib" != yes; then
+               echo
+               $ECHO "*** Warning: Trying to link with static lib archive $deplib."
+               echo "*** I have the capability to make that library automatically link in when"
+               echo "*** you link to this library.  But I can only do this if you have a"
+               echo "*** shared version of the library, which you do not appear to have"
+               echo "*** because the file extensions .$libext of this argument makes me believe"
+               echo "*** that it is just a static archive that I should not use here."
+             else
+               echo
+               $ECHO "*** Warning: Linking the shared library $output against the"
+               $ECHO "*** static library $deplib is not portable!"
+               deplibs="$deplib $deplibs"
+             fi
+             ;;
+           esac
+           continue
+           ;;
+         prog)
+           if test "$pass" != link; then
+             deplibs="$deplib $deplibs"
+           else
+             compile_deplibs="$deplib $compile_deplibs"
+             finalize_deplibs="$deplib $finalize_deplibs"
+           fi
+           continue
+           ;;
+         esac # linkmode
+         ;; # *.$libext
+       *.lo | *.$objext)
+         if test "$pass" = conv; then
+           deplibs="$deplib $deplibs"
+         elif test "$linkmode" = prog; then
+           if test "$pass" = dlpreopen || test "$dlopen_support" != yes || test "$build_libtool_libs" = no; then
+             # If there is no dlopen support or we're linking statically,
+             # we need to preload.
+             func_append newdlprefiles " $deplib"
+             compile_deplibs="$deplib $compile_deplibs"
+             finalize_deplibs="$deplib $finalize_deplibs"
+           else
+             func_append newdlfiles " $deplib"
+           fi
+         fi
+         continue
+         ;;
+       %DEPLIBS%)
+         alldeplibs=yes
+         continue
+         ;;
+       esac # case $deplib
+
+       if test "$found" = yes || test -f "$lib"; then :
+       else
+         func_fatal_error "cannot find the library \`$lib' or unhandled argument \`$deplib'"
+       fi
+
+       # Check to see that this really is a libtool archive.
+       func_lalib_unsafe_p "$lib" \
+         || func_fatal_error "\`$lib' is not a valid libtool archive"
+
+       func_dirname "$lib" "" "."
+       ladir="$func_dirname_result"
+
+       dlname=
+       dlopen=
+       dlpreopen=
+       libdir=
+       library_names=
+       old_library=
+       inherited_linker_flags=
+       # If the library was installed with an old release of libtool,
+       # it will not redefine variables installed, or shouldnotlink
+       installed=yes
+       shouldnotlink=no
+       avoidtemprpath=
+
+
+       # Read the .la file
+       func_source "$lib"
+
+       # Convert "-framework foo" to "foo.ltframework"
+       if test -n "$inherited_linker_flags"; then
+         tmp_inherited_linker_flags=`$ECHO "$inherited_linker_flags" | $SED 's/-framework \([^ $]*\)/\1.ltframework/g'`
+         for tmp_inherited_linker_flag in $tmp_inherited_linker_flags; do
+           case " $new_inherited_linker_flags " in
+             *" $tmp_inherited_linker_flag "*) ;;
+             *) func_append new_inherited_linker_flags " $tmp_inherited_linker_flag";;
+           esac
+         done
+       fi
+       dependency_libs=`$ECHO " $dependency_libs" | $SED 's% \([^ $]*\).ltframework% -framework \1%g'`
+       if test "$linkmode,$pass" = "lib,link" ||
+          test "$linkmode,$pass" = "prog,scan" ||
+          { test "$linkmode" != prog && test "$linkmode" != lib; }; then
+         test -n "$dlopen" && func_append dlfiles " $dlopen"
+         test -n "$dlpreopen" && func_append dlprefiles " $dlpreopen"
+       fi
+
+       if test "$pass" = conv; then
+         # Only check for convenience libraries
+         deplibs="$lib $deplibs"
+         if test -z "$libdir"; then
+           if test -z "$old_library"; then
+             func_fatal_error "cannot find name of link library for \`$lib'"
+           fi
+           # It is a libtool convenience library, so add in its objects.
+           func_append convenience " $ladir/$objdir/$old_library"
+           func_append old_convenience " $ladir/$objdir/$old_library"
+           tmp_libs=
+           for deplib in $dependency_libs; do
+             deplibs="$deplib $deplibs"
+             if $opt_preserve_dup_deps ; then
+               case "$tmp_libs " in
+               *" $deplib "*) func_append specialdeplibs " $deplib" ;;
+               esac
+             fi
+             func_append tmp_libs " $deplib"
+           done
+         elif test "$linkmode" != prog && test "$linkmode" != lib; then
+           func_fatal_error "\`$lib' is not a convenience library"
+         fi
+         continue
+       fi # $pass = conv
+
+
+       # Get the name of the library we link against.
+       linklib=
+       if test -n "$old_library" &&
+          { test "$prefer_static_libs" = yes ||
+            test "$prefer_static_libs,$installed" = "built,no"; }; then
+         linklib=$old_library
+       else
+         for l in $old_library $library_names; do
+           linklib="$l"
+         done
+       fi
+       if test -z "$linklib"; then
+         func_fatal_error "cannot find name of link library for \`$lib'"
+       fi
+
+       # This library was specified with -dlopen.
+       if test "$pass" = dlopen; then
+         if test -z "$libdir"; then
+           func_fatal_error "cannot -dlopen a convenience library: \`$lib'"
+         fi
+         if test -z "$dlname" ||
+            test "$dlopen_support" != yes ||
+            test "$build_libtool_libs" = no; then
+           # If there is no dlname, no dlopen support or we're linking
+           # statically, we need to preload.  We also need to preload any
+           # dependent libraries so libltdl's deplib preloader doesn't
+           # bomb out in the load deplibs phase.
+           func_append dlprefiles " $lib $dependency_libs"
+         else
+           func_append newdlfiles " $lib"
+         fi
+         continue
+       fi # $pass = dlopen
+
+       # We need an absolute path.
+       case $ladir in
+       [\\/]* | [A-Za-z]:[\\/]*) abs_ladir="$ladir" ;;
+       *)
+         abs_ladir=`cd "$ladir" && pwd`
+         if test -z "$abs_ladir"; then
+           func_warning "cannot determine absolute directory name of \`$ladir'"
+           func_warning "passing it literally to the linker, although it might fail"
+           abs_ladir="$ladir"
+         fi
+         ;;
+       esac
+       func_basename "$lib"
+       laname="$func_basename_result"
+
+       # Find the relevant object directory and library name.
+       if test "X$installed" = Xyes; then
+         if test ! -f "$lt_sysroot$libdir/$linklib" && test -f "$abs_ladir/$linklib"; then
+           func_warning "library \`$lib' was moved."
+           dir="$ladir"
+           absdir="$abs_ladir"
+           libdir="$abs_ladir"
+         else
+           dir="$lt_sysroot$libdir"
+           absdir="$lt_sysroot$libdir"
+         fi
+         test "X$hardcode_automatic" = Xyes && avoidtemprpath=yes
+       else
+         if test ! -f "$ladir/$objdir/$linklib" && test -f "$abs_ladir/$linklib"; then
+           dir="$ladir"
+           absdir="$abs_ladir"
+           # Remove this search path later
+           func_append notinst_path " $abs_ladir"
+         else
+           dir="$ladir/$objdir"
+           absdir="$abs_ladir/$objdir"
+           # Remove this search path later
+           func_append notinst_path " $abs_ladir"
+         fi
+       fi # $installed = yes
+       func_stripname 'lib' '.la' "$laname"
+       name=$func_stripname_result
+
+       # This library was specified with -dlpreopen.
+       if test "$pass" = dlpreopen; then
+         if test -z "$libdir" && test "$linkmode" = prog; then
+           func_fatal_error "only libraries may -dlpreopen a convenience library: \`$lib'"
+         fi
+         case "$host" in
+           # special handling for platforms with PE-DLLs.
+           *cygwin* | *mingw* | *cegcc* )
+             # Linker will automatically link against shared library if both
+             # static and shared are present.  Therefore, ensure we extract
+             # symbols from the import library if a shared library is present
+             # (otherwise, the dlopen module name will be incorrect).  We do
+             # this by putting the import library name into $newdlprefiles.
+             # We recover the dlopen module name by 'saving' the la file
+             # name in a special purpose variable, and (later) extracting the
+             # dlname from the la file.
+             if test -n "$dlname"; then
+               func_tr_sh "$dir/$linklib"
+               eval "libfile_$func_tr_sh_result=\$abs_ladir/\$laname"
+               func_append newdlprefiles " $dir/$linklib"
+             else
+               func_append newdlprefiles " $dir/$old_library"
+               # Keep a list of preopened convenience libraries to check
+               # that they are being used correctly in the link pass.
+               test -z "$libdir" && \
+                 func_append dlpreconveniencelibs " $dir/$old_library"
+             fi
+           ;;
+           * )
+             # Prefer using a static library (so that no silly _DYNAMIC symbols
+             # are required to link).
+             if test -n "$old_library"; then
+               func_append newdlprefiles " $dir/$old_library"
+               # Keep a list of preopened convenience libraries to check
+               # that they are being used correctly in the link pass.
+               test -z "$libdir" && \
+                 func_append dlpreconveniencelibs " $dir/$old_library"
+             # Otherwise, use the dlname, so that lt_dlopen finds it.
+             elif test -n "$dlname"; then
+               func_append newdlprefiles " $dir/$dlname"
+             else
+               func_append newdlprefiles " $dir/$linklib"
+             fi
+           ;;
+         esac
+       fi # $pass = dlpreopen
+
+       if test -z "$libdir"; then
+         # Link the convenience library
+         if test "$linkmode" = lib; then
+           deplibs="$dir/$old_library $deplibs"
+         elif test "$linkmode,$pass" = "prog,link"; then
+           compile_deplibs="$dir/$old_library $compile_deplibs"
+           finalize_deplibs="$dir/$old_library $finalize_deplibs"
+         else
+           deplibs="$lib $deplibs" # used for prog,scan pass
+         fi
+         continue
+       fi
+
+
+       if test "$linkmode" = prog && test "$pass" != link; then
+         func_append newlib_search_path " $ladir"
+         deplibs="$lib $deplibs"
+
+         linkalldeplibs=no
+         if test "$link_all_deplibs" != no || test -z "$library_names" ||
+            test "$build_libtool_libs" = no; then
+           linkalldeplibs=yes
+         fi
+
+         tmp_libs=
+         for deplib in $dependency_libs; do
+           case $deplib in
+           -L*) func_stripname '-L' '' "$deplib"
+                func_resolve_sysroot "$func_stripname_result"
+                func_append newlib_search_path " $func_resolve_sysroot_result"
+                ;;
+           esac
+           # Need to link against all dependency_libs?
+           if test "$linkalldeplibs" = yes; then
+             deplibs="$deplib $deplibs"
+           else
+             # Need to hardcode shared library paths
+             # or/and link against static libraries
+             newdependency_libs="$deplib $newdependency_libs"
+           fi
+           if $opt_preserve_dup_deps ; then
+             case "$tmp_libs " in
+             *" $deplib "*) func_append specialdeplibs " $deplib" ;;
+             esac
+           fi
+           func_append tmp_libs " $deplib"
+         done # for deplib
+         continue
+       fi # $linkmode = prog...
+
+       if test "$linkmode,$pass" = "prog,link"; then
+         if test -n "$library_names" &&
+            { { test "$prefer_static_libs" = no ||
+                test "$prefer_static_libs,$installed" = "built,yes"; } ||
+              test -z "$old_library"; }; then
+           # We need to hardcode the library path
+           if test -n "$shlibpath_var" && test -z "$avoidtemprpath" ; then
+             # Make sure the rpath contains only unique directories.
+             case "$temp_rpath:" in
+             *"$absdir:"*) ;;
+             *) func_append temp_rpath "$absdir:" ;;
+             esac
+           fi
+
+           # Hardcode the library path.
+           # Skip directories that are in the system default run-time
+           # search path.
+           case " $sys_lib_dlsearch_path " in
+           *" $absdir "*) ;;
+           *)
+             case "$compile_rpath " in
+             *" $absdir "*) ;;
+             *) func_append compile_rpath " $absdir" ;;
+             esac
+             ;;
+           esac
+           case " $sys_lib_dlsearch_path " in
+           *" $libdir "*) ;;
+           *)
+             case "$finalize_rpath " in
+             *" $libdir "*) ;;
+             *) func_append finalize_rpath " $libdir" ;;
+             esac
+             ;;
+           esac
+         fi # $linkmode,$pass = prog,link...
+
+         if test "$alldeplibs" = yes &&
+            { test "$deplibs_check_method" = pass_all ||
+              { test "$build_libtool_libs" = yes &&
+                test -n "$library_names"; }; }; then
+           # We only need to search for static libraries
+           continue
+         fi
+       fi
+
+       link_static=no # Whether the deplib will be linked statically
+       use_static_libs=$prefer_static_libs
+       if test "$use_static_libs" = built && test "$installed" = yes; then
+         use_static_libs=no
+       fi
+       if test -n "$library_names" &&
+          { test "$use_static_libs" = no || test -z "$old_library"; }; then
+         case $host in
+         *cygwin* | *mingw* | *cegcc*)
+             # No point in relinking DLLs because paths are not encoded
+             func_append notinst_deplibs " $lib"
+             need_relink=no
+           ;;
+         *)
+           if test "$installed" = no; then
+             func_append notinst_deplibs " $lib"
+             need_relink=yes
+           fi
+           ;;
+         esac
+         # This is a shared library
+
+         # Warn about portability, can't link against -module's on some
+         # systems (darwin).  Don't bleat about dlopened modules though!
+         dlopenmodule=""
+         for dlpremoduletest in $dlprefiles; do
+           if test "X$dlpremoduletest" = "X$lib"; then
+             dlopenmodule="$dlpremoduletest"
+             break
+           fi
+         done
+         if test -z "$dlopenmodule" && test "$shouldnotlink" = yes && test "$pass" = link; then
+           echo
+           if test "$linkmode" = prog; then
+             $ECHO "*** Warning: Linking the executable $output against the loadable module"
+           else
+             $ECHO "*** Warning: Linking the shared library $output against the loadable module"
+           fi
+           $ECHO "*** $linklib is not portable!"
+         fi
+         if test "$linkmode" = lib &&
+            test "$hardcode_into_libs" = yes; then
+           # Hardcode the library path.
+           # Skip directories that are in the system default run-time
+           # search path.
+           case " $sys_lib_dlsearch_path " in
+           *" $absdir "*) ;;
+           *)
+             case "$compile_rpath " in
+             *" $absdir "*) ;;
+             *) func_append compile_rpath " $absdir" ;;
+             esac
+             ;;
+           esac
+           case " $sys_lib_dlsearch_path " in
+           *" $libdir "*) ;;
+           *)
+             case "$finalize_rpath " in
+             *" $libdir "*) ;;
+             *) func_append finalize_rpath " $libdir" ;;
+             esac
+             ;;
+           esac
+         fi
+
+         if test -n "$old_archive_from_expsyms_cmds"; then
+           # figure out the soname
+           set dummy $library_names
+           shift
+           realname="$1"
+           shift
+           libname=`eval "\\$ECHO \"$libname_spec\""`
+           # use dlname if we got it. it's perfectly good, no?
+           if test -n "$dlname"; then
+             soname="$dlname"
+           elif test -n "$soname_spec"; then
+             # bleh windows
+             case $host in
+             *cygwin* | mingw* | *cegcc*)
+               func_arith $current - $age
+               major=$func_arith_result
+               versuffix="-$major"
+               ;;
+             esac
+             eval soname=\"$soname_spec\"
+           else
+             soname="$realname"
+           fi
+
+           # Make a new name for the extract_expsyms_cmds to use
+           soroot="$soname"
+           func_basename "$soroot"
+           soname="$func_basename_result"
+           func_stripname 'lib' '.dll' "$soname"
+           newlib=libimp-$func_stripname_result.a
+
+           # If the library has no export list, then create one now
+           if test -f "$output_objdir/$soname-def"; then :
+           else
+             func_verbose "extracting exported symbol list from \`$soname'"
+             func_execute_cmds "$extract_expsyms_cmds" 'exit $?'
+           fi
+
+           # Create $newlib
+           if test -f "$output_objdir/$newlib"; then :; else
+             func_verbose "generating import library for \`$soname'"
+             func_execute_cmds "$old_archive_from_expsyms_cmds" 'exit $?'
+           fi
+           # make sure the library variables are pointing to the new library
+           dir=$output_objdir
+           linklib=$newlib
+         fi # test -n "$old_archive_from_expsyms_cmds"
+
+         if test "$linkmode" = prog || test "$opt_mode" != relink; then
+           add_shlibpath=
+           add_dir=
+           add=
+           lib_linked=yes
+           case $hardcode_action in
+           immediate | unsupported)
+             if test "$hardcode_direct" = no; then
+               add="$dir/$linklib"
+               case $host in
+                 *-*-sco3.2v5.0.[024]*) add_dir="-L$dir" ;;
+                 *-*-sysv4*uw2*) add_dir="-L$dir" ;;
+                 *-*-sysv5OpenUNIX* | *-*-sysv5UnixWare7.[01].[10]* | \
+                   *-*-unixware7*) add_dir="-L$dir" ;;
+                 *-*-darwin* )
+                   # if the lib is a (non-dlopened) module then we can not
+                   # link against it, someone is ignoring the earlier warnings
+                   if /usr/bin/file -L $add 2> /dev/null |
+                        $GREP ": [^:]* bundle" >/dev/null ; then
+                     if test "X$dlopenmodule" != "X$lib"; then
+                       $ECHO "*** Warning: lib $linklib is a module, not a shared library"
+                       if test -z "$old_library" ; then
+                         echo
+                         echo "*** And there doesn't seem to be a static archive available"
+                         echo "*** The link will probably fail, sorry"
+                       else
+                         add="$dir/$old_library"
+                       fi
+                     elif test -n "$old_library"; then
+                       add="$dir/$old_library"
+                     fi
+                   fi
+               esac
+             elif test "$hardcode_minus_L" = no; then
+               case $host in
+               *-*-sunos*) add_shlibpath="$dir" ;;
+               esac
+               add_dir="-L$dir"
+               add="-l$name"
+             elif test "$hardcode_shlibpath_var" = no; then
+               add_shlibpath="$dir"
+               add="-l$name"
+             else
+               lib_linked=no
+             fi
+             ;;
+           relink)
+             if test "$hardcode_direct" = yes &&
+                test "$hardcode_direct_absolute" = no; then
+               add="$dir/$linklib"
+             elif test "$hardcode_minus_L" = yes; then
+               add_dir="-L$absdir"
+               # Try looking first in the location we're being installed to.
+               if test -n "$inst_prefix_dir"; then
+                 case $libdir in
+                   [\\/]*)
+                     func_append add_dir " -L$inst_prefix_dir$libdir"
+                     ;;
+                 esac
+               fi
+               add="-l$name"
+             elif test "$hardcode_shlibpath_var" = yes; then
+               add_shlibpath="$dir"
+               add="-l$name"
+             else
+               lib_linked=no
+             fi
+             ;;
+           *) lib_linked=no ;;
+           esac
+
+           if test "$lib_linked" != yes; then
+             func_fatal_configuration "unsupported hardcode properties"
+           fi
+
+           if test -n "$add_shlibpath"; then
+             case :$compile_shlibpath: in
+             *":$add_shlibpath:"*) ;;
+             *) func_append compile_shlibpath "$add_shlibpath:" ;;
+             esac
+           fi
+           if test "$linkmode" = prog; then
+             test -n "$add_dir" && compile_deplibs="$add_dir $compile_deplibs"
+             test -n "$add" && compile_deplibs="$add $compile_deplibs"
+           else
+             test -n "$add_dir" && deplibs="$add_dir $deplibs"
+             test -n "$add" && deplibs="$add $deplibs"
+             if test "$hardcode_direct" != yes &&
+                test "$hardcode_minus_L" != yes &&
+                test "$hardcode_shlibpath_var" = yes; then
+               case :$finalize_shlibpath: in
+               *":$libdir:"*) ;;
+               *) func_append finalize_shlibpath "$libdir:" ;;
+               esac
+             fi
+           fi
+         fi
+
+         if test "$linkmode" = prog || test "$opt_mode" = relink; then
+           add_shlibpath=
+           add_dir=
+           add=
+           # Finalize command for both is simple: just hardcode it.
+           if test "$hardcode_direct" = yes &&
+              test "$hardcode_direct_absolute" = no; then
+             add="$libdir/$linklib"
+           elif test "$hardcode_minus_L" = yes; then
+             add_dir="-L$libdir"
+             add="-l$name"
+           elif test "$hardcode_shlibpath_var" = yes; then
+             case :$finalize_shlibpath: in
+             *":$libdir:"*) ;;
+             *) func_append finalize_shlibpath "$libdir:" ;;
+             esac
+             add="-l$name"
+           elif test "$hardcode_automatic" = yes; then
+             if test -n "$inst_prefix_dir" &&
+                test -f "$inst_prefix_dir$libdir/$linklib" ; then
+               add="$inst_prefix_dir$libdir/$linklib"
+             else
+               add="$libdir/$linklib"
+             fi
+           else
+             # We cannot seem to hardcode it, guess we'll fake it.
+             add_dir="-L$libdir"
+             # Try looking first in the location we're being installed to.
+             if test -n "$inst_prefix_dir"; then
+               case $libdir in
+                 [\\/]*)
+                   func_append add_dir " -L$inst_prefix_dir$libdir"
+                   ;;
+               esac
+             fi
+             add="-l$name"
+           fi
+
+           if test "$linkmode" = prog; then
+             test -n "$add_dir" && finalize_deplibs="$add_dir $finalize_deplibs"
+             test -n "$add" && finalize_deplibs="$add $finalize_deplibs"
+           else
+             test -n "$add_dir" && deplibs="$add_dir $deplibs"
+             test -n "$add" && deplibs="$add $deplibs"
+           fi
+         fi
+       elif test "$linkmode" = prog; then
+         # Here we assume that one of hardcode_direct or hardcode_minus_L
+         # is not unsupported.  This is valid on all known static and
+         # shared platforms.
+         if test "$hardcode_direct" != unsupported; then
+           test -n "$old_library" && linklib="$old_library"
+           compile_deplibs="$dir/$linklib $compile_deplibs"
+           finalize_deplibs="$dir/$linklib $finalize_deplibs"
+         else
+           compile_deplibs="-l$name -L$dir $compile_deplibs"
+           finalize_deplibs="-l$name -L$dir $finalize_deplibs"
+         fi
+       elif test "$build_libtool_libs" = yes; then
+         # Not a shared library
+         if test "$deplibs_check_method" != pass_all; then
+           # We're trying link a shared library against a static one
+           # but the system doesn't support it.
+
+           # Just print a warning and add the library to dependency_libs so
+           # that the program can be linked against the static library.
+           echo
+           $ECHO "*** Warning: This system can not link to static lib archive $lib."
+           echo "*** I have the capability to make that library automatically link in when"
+           echo "*** you link to this library.  But I can only do this if you have a"
+           echo "*** shared version of the library, which you do not appear to have."
+           if test "$module" = yes; then
+             echo "*** But as you try to build a module library, libtool will still create "
+             echo "*** a static module, that should work as long as the dlopening application"
+             echo "*** is linked with the -dlopen flag to resolve symbols at runtime."
+             if test -z "$global_symbol_pipe"; then
+               echo
+               echo "*** However, this would only work if libtool was able to extract symbol"
+               echo "*** lists from a program, using \`nm' or equivalent, but libtool could"
+               echo "*** not find such a program.  So, this module is probably useless."
+               echo "*** \`nm' from GNU binutils and a full rebuild may help."
+             fi
+             if test "$build_old_libs" = no; then
+               build_libtool_libs=module
+               build_old_libs=yes
+             else
+               build_libtool_libs=no
+             fi
+           fi
+         else
+           deplibs="$dir/$old_library $deplibs"
+           link_static=yes
+         fi
+       fi # link shared/static library?
+
+       if test "$linkmode" = lib; then
+         if test -n "$dependency_libs" &&
+            { test "$hardcode_into_libs" != yes ||
+              test "$build_old_libs" = yes ||
+              test "$link_static" = yes; }; then
+           # Extract -R from dependency_libs
+           temp_deplibs=
+           for libdir in $dependency_libs; do
+             case $libdir in
+             -R*) func_stripname '-R' '' "$libdir"
+                  temp_xrpath=$func_stripname_result
+                  case " $xrpath " in
+                  *" $temp_xrpath "*) ;;
+                  *) func_append xrpath " $temp_xrpath";;
+                  esac;;
+             *) func_append temp_deplibs " $libdir";;
+             esac
+           done
+           dependency_libs="$temp_deplibs"
+         fi
+
+         func_append newlib_search_path " $absdir"
+         # Link against this library
+         test "$link_static" = no && newdependency_libs="$abs_ladir/$laname $newdependency_libs"
+         # ... and its dependency_libs
+         tmp_libs=
+         for deplib in $dependency_libs; do
+           newdependency_libs="$deplib $newdependency_libs"
+           case $deplib in
+              -L*) func_stripname '-L' '' "$deplib"
+                   func_resolve_sysroot "$func_stripname_result";;
+              *) func_resolve_sysroot "$deplib" ;;
+            esac
+           if $opt_preserve_dup_deps ; then
+             case "$tmp_libs " in
+             *" $func_resolve_sysroot_result "*)
+                func_append specialdeplibs " $func_resolve_sysroot_result" ;;
+             esac
+           fi
+           func_append tmp_libs " $func_resolve_sysroot_result"
+         done
+
+         if test "$link_all_deplibs" != no; then
+           # Add the search paths of all dependency libraries
+           for deplib in $dependency_libs; do
+             path=
+             case $deplib in
+             -L*) path="$deplib" ;;
+             *.la)
+               func_resolve_sysroot "$deplib"
+               deplib=$func_resolve_sysroot_result
+               func_dirname "$deplib" "" "."
+               dir=$func_dirname_result
+               # We need an absolute path.
+               case $dir in
+               [\\/]* | [A-Za-z]:[\\/]*) absdir="$dir" ;;
+               *)
+                 absdir=`cd "$dir" && pwd`
+                 if test -z "$absdir"; then
+                   func_warning "cannot determine absolute directory name of \`$dir'"
+                   absdir="$dir"
+                 fi
+                 ;;
+               esac
+               if $GREP "^installed=no" $deplib > /dev/null; then
+               case $host in
+               *-*-darwin*)
+                 depdepl=
+                 eval deplibrary_names=`${SED} -n -e 's/^library_names=\(.*\)$/\1/p' $deplib`
+                 if test -n "$deplibrary_names" ; then
+                   for tmp in $deplibrary_names ; do
+                     depdepl=$tmp
+                   done
+                   if test -f "$absdir/$objdir/$depdepl" ; then
+                     depdepl="$absdir/$objdir/$depdepl"
+                     darwin_install_name=`${OTOOL} -L $depdepl | awk '{if (NR == 2) {print $1;exit}}'`
+                      if test -z "$darwin_install_name"; then
+                          darwin_install_name=`${OTOOL64} -L $depdepl  | awk '{if (NR == 2) {print $1;exit}}'`
+                      fi
+                     func_append compiler_flags " ${wl}-dylib_file ${wl}${darwin_install_name}:${depdepl}"
+                     func_append linker_flags " -dylib_file ${darwin_install_name}:${depdepl}"
+                     path=
+                   fi
+                 fi
+                 ;;
+               *)
+                 path="-L$absdir/$objdir"
+                 ;;
+               esac
+               else
+                 eval libdir=`${SED} -n -e 's/^libdir=\(.*\)$/\1/p' $deplib`
+                 test -z "$libdir" && \
+                   func_fatal_error "\`$deplib' is not a valid libtool archive"
+                 test "$absdir" != "$libdir" && \
+                   func_warning "\`$deplib' seems to be moved"
+
+                 path="-L$absdir"
+               fi
+               ;;
+             esac
+             case " $deplibs " in
+             *" $path "*) ;;
+             *) deplibs="$path $deplibs" ;;
+             esac
+           done
+         fi # link_all_deplibs != no
+       fi # linkmode = lib
+      done # for deplib in $libs
+      if test "$pass" = link; then
+       if test "$linkmode" = "prog"; then
+         compile_deplibs="$new_inherited_linker_flags $compile_deplibs"
+         finalize_deplibs="$new_inherited_linker_flags $finalize_deplibs"
+       else
+         compiler_flags="$compiler_flags "`$ECHO " $new_inherited_linker_flags" | $SED 's% \([^ $]*\).ltframework% -framework \1%g'`
+       fi
+      fi
+      dependency_libs="$newdependency_libs"
+      if test "$pass" = dlpreopen; then
+       # Link the dlpreopened libraries before other libraries
+       for deplib in $save_deplibs; do
+         deplibs="$deplib $deplibs"
+       done
+      fi
+      if test "$pass" != dlopen; then
+       if test "$pass" != conv; then
+         # Make sure lib_search_path contains only unique directories.
+         lib_search_path=
+         for dir in $newlib_search_path; do
+           case "$lib_search_path " in
+           *" $dir "*) ;;
+           *) func_append lib_search_path " $dir" ;;
+           esac
+         done
+         newlib_search_path=
+       fi
+
+       if test "$linkmode,$pass" != "prog,link"; then
+         vars="deplibs"
+       else
+         vars="compile_deplibs finalize_deplibs"
+       fi
+       for var in $vars dependency_libs; do
+         # Add libraries to $var in reverse order
+         eval tmp_libs=\"\$$var\"
+         new_libs=
+         for deplib in $tmp_libs; do
+           # FIXME: Pedantically, this is the right thing to do, so
+           #        that some nasty dependency loop isn't accidentally
+           #        broken:
+           #new_libs="$deplib $new_libs"
+           # Pragmatically, this seems to cause very few problems in
+           # practice:
+           case $deplib in
+           -L*) new_libs="$deplib $new_libs" ;;
+           -R*) ;;
+           *)
+             # And here is the reason: when a library appears more
+             # than once as an explicit dependence of a library, or
+             # is implicitly linked in more than once by the
+             # compiler, it is considered special, and multiple
+             # occurrences thereof are not removed.  Compare this
+             # with having the same library being listed as a
+             # dependency of multiple other libraries: in this case,
+             # we know (pedantically, we assume) the library does not
+             # need to be listed more than once, so we keep only the
+             # last copy.  This is not always right, but it is rare
+             # enough that we require users that really mean to play
+             # such unportable linking tricks to link the library
+             # using -Wl,-lname, so that libtool does not consider it
+             # for duplicate removal.
+             case " $specialdeplibs " in
+             *" $deplib "*) new_libs="$deplib $new_libs" ;;
+             *)
+               case " $new_libs " in
+               *" $deplib "*) ;;
+               *) new_libs="$deplib $new_libs" ;;
+               esac
+               ;;
+             esac
+             ;;
+           esac
+         done
+         tmp_libs=
+         for deplib in $new_libs; do
+           case $deplib in
+           -L*)
+             case " $tmp_libs " in
+             *" $deplib "*) ;;
+             *) func_append tmp_libs " $deplib" ;;
+             esac
+             ;;
+           *) func_append tmp_libs " $deplib" ;;
+           esac
+         done
+         eval $var=\"$tmp_libs\"
+       done # for var
+      fi
+      # Last step: remove runtime libs from dependency_libs
+      # (they stay in deplibs)
+      tmp_libs=
+      for i in $dependency_libs ; do
+       case " $predeps $postdeps $compiler_lib_search_path " in
+       *" $i "*)
+         i=""
+         ;;
+       esac
+       if test -n "$i" ; then
+         func_append tmp_libs " $i"
+       fi
+      done
+      dependency_libs=$tmp_libs
+    done # for pass
+    if test "$linkmode" = prog; then
+      dlfiles="$newdlfiles"
+    fi
+    if test "$linkmode" = prog || test "$linkmode" = lib; then
+      dlprefiles="$newdlprefiles"
+    fi
+
+    case $linkmode in
+    oldlib)
+      if test -n "$dlfiles$dlprefiles" || test "$dlself" != no; then
+       func_warning "\`-dlopen' is ignored for archives"
+      fi
+
+      case " $deplibs" in
+      *\ -l* | *\ -L*)
+       func_warning "\`-l' and \`-L' are ignored for archives" ;;
+      esac
+
+      test -n "$rpath" && \
+       func_warning "\`-rpath' is ignored for archives"
+
+      test -n "$xrpath" && \
+       func_warning "\`-R' is ignored for archives"
+
+      test -n "$vinfo" && \
+       func_warning "\`-version-info/-version-number' is ignored for archives"
+
+      test -n "$release" && \
+       func_warning "\`-release' is ignored for archives"
+
+      test -n "$export_symbols$export_symbols_regex" && \
+       func_warning "\`-export-symbols' is ignored for archives"
+
+      # Now set the variables for building old libraries.
+      build_libtool_libs=no
+      oldlibs="$output"
+      func_append objs "$old_deplibs"
+      ;;
+
+    lib)
+      # Make sure we only generate libraries of the form `libNAME.la'.
+      case $outputname in
+      lib*)
+       func_stripname 'lib' '.la' "$outputname"
+       name=$func_stripname_result
+       eval shared_ext=\"$shrext_cmds\"
+       eval libname=\"$libname_spec\"
+       ;;
+      *)
+       test "$module" = no && \
+         func_fatal_help "libtool library \`$output' must begin with \`lib'"
+
+       if test "$need_lib_prefix" != no; then
+         # Add the "lib" prefix for modules if required
+         func_stripname '' '.la' "$outputname"
+         name=$func_stripname_result
+         eval shared_ext=\"$shrext_cmds\"
+         eval libname=\"$libname_spec\"
+       else
+         func_stripname '' '.la' "$outputname"
+         libname=$func_stripname_result
+       fi
+       ;;
+      esac
+
+      if test -n "$objs"; then
+       if test "$deplibs_check_method" != pass_all; then
+         func_fatal_error "cannot build libtool library \`$output' from non-libtool objects on this host:$objs"
+       else
+         echo
+         $ECHO "*** Warning: Linking the shared library $output against the non-libtool"
+         $ECHO "*** objects $objs is not portable!"
+         func_append libobjs " $objs"
+       fi
+      fi
+
+      test "$dlself" != no && \
+       func_warning "\`-dlopen self' is ignored for libtool libraries"
+
+      set dummy $rpath
+      shift
+      test "$#" -gt 1 && \
+       func_warning "ignoring multiple \`-rpath's for a libtool library"
+
+      install_libdir="$1"
+
+      oldlibs=
+      if test -z "$rpath"; then
+       if test "$build_libtool_libs" = yes; then
+         # Building a libtool convenience library.
+         # Some compilers have problems with a `.al' extension so
+         # convenience libraries should have the same extension an
+         # archive normally would.
+         oldlibs="$output_objdir/$libname.$libext $oldlibs"
+         build_libtool_libs=convenience
+         build_old_libs=yes
+       fi
+
+       test -n "$vinfo" && \
+         func_warning "\`-version-info/-version-number' is ignored for convenience libraries"
+
+       test -n "$release" && \
+         func_warning "\`-release' is ignored for convenience libraries"
+      else
+
+       # Parse the version information argument.
+       save_ifs="$IFS"; IFS=':'
+       set dummy $vinfo 0 0 0
+       shift
+       IFS="$save_ifs"
+
+       test -n "$7" && \
+         func_fatal_help "too many parameters to \`-version-info'"
+
+       # convert absolute version numbers to libtool ages
+       # this retains compatibility with .la files and attempts
+       # to make the code below a bit more comprehensible
+
+       case $vinfo_number in
+       yes)
+         number_major="$1"
+         number_minor="$2"
+         number_revision="$3"
+         #
+         # There are really only two kinds -- those that
+         # use the current revision as the major version
+         # and those that subtract age and use age as
+         # a minor version.  But, then there is irix
+         # which has an extra 1 added just for fun
+         #
+         case $version_type in
+         # correct linux to gnu/linux during the next big refactor
+         darwin|linux|osf|windows|none)
+           func_arith $number_major + $number_minor
+           current=$func_arith_result
+           age="$number_minor"
+           revision="$number_revision"
+           ;;
+         freebsd-aout|freebsd-elf|qnx|sunos)
+           current="$number_major"
+           revision="$number_minor"
+           age="0"
+           ;;
+         irix|nonstopux)
+           func_arith $number_major + $number_minor
+           current=$func_arith_result
+           age="$number_minor"
+           revision="$number_minor"
+           lt_irix_increment=no
+           ;;
+         *)
+           func_fatal_configuration "$modename: unknown library version type \`$version_type'"
+           ;;
+         esac
+         ;;
+       no)
+         current="$1"
+         revision="$2"
+         age="$3"
+         ;;
+       esac
+
+       # Check that each of the things are valid numbers.
+       case $current in
+       0|[1-9]|[1-9][0-9]|[1-9][0-9][0-9]|[1-9][0-9][0-9][0-9]|[1-9][0-9][0-9][0-9][0-9]) ;;
+       *)
+         func_error "CURRENT \`$current' must be a nonnegative integer"
+         func_fatal_error "\`$vinfo' is not valid version information"
+         ;;
+       esac
+
+       case $revision in
+       0|[1-9]|[1-9][0-9]|[1-9][0-9][0-9]|[1-9][0-9][0-9][0-9]|[1-9][0-9][0-9][0-9][0-9]) ;;
+       *)
+         func_error "REVISION \`$revision' must be a nonnegative integer"
+         func_fatal_error "\`$vinfo' is not valid version information"
+         ;;
+       esac
+
+       case $age in
+       0|[1-9]|[1-9][0-9]|[1-9][0-9][0-9]|[1-9][0-9][0-9][0-9]|[1-9][0-9][0-9][0-9][0-9]) ;;
+       *)
+         func_error "AGE \`$age' must be a nonnegative integer"
+         func_fatal_error "\`$vinfo' is not valid version information"
+         ;;
+       esac
+
+       if test "$age" -gt "$current"; then
+         func_error "AGE \`$age' is greater than the current interface number \`$current'"
+         func_fatal_error "\`$vinfo' is not valid version information"
+       fi
+
+       # Calculate the version variables.
+       major=
+       versuffix=
+       verstring=
+       case $version_type in
+       none) ;;
+
+       darwin)
+         # Like Linux, but with the current version available in
+         # verstring for coding it into the library header
+         func_arith $current - $age
+         major=.$func_arith_result
+         versuffix="$major.$age.$revision"
+         # Darwin ld doesn't like 0 for these options...
+         func_arith $current + 1
+         minor_current=$func_arith_result
+         xlcverstring="${wl}-compatibility_version ${wl}$minor_current ${wl}-current_version ${wl}$minor_current.$revision"
+         verstring="-compatibility_version $minor_current -current_version $minor_current.$revision"
+         ;;
+
+       freebsd-aout)
+         major=".$current"
+         versuffix=".$current.$revision";
+         ;;
+
+       freebsd-elf)
+         major=".$current"
+         versuffix=".$current"
+         ;;
+
+       irix | nonstopux)
+         if test "X$lt_irix_increment" = "Xno"; then
+           func_arith $current - $age
+         else
+           func_arith $current - $age + 1
+         fi
+         major=$func_arith_result
+
+         case $version_type in
+           nonstopux) verstring_prefix=nonstopux ;;
+           *)         verstring_prefix=sgi ;;
+         esac
+         verstring="$verstring_prefix$major.$revision"
+
+         # Add in all the interfaces that we are compatible with.
+         loop=$revision
+         while test "$loop" -ne 0; do
+           func_arith $revision - $loop
+           iface=$func_arith_result
+           func_arith $loop - 1
+           loop=$func_arith_result
+           verstring="$verstring_prefix$major.$iface:$verstring"
+         done
+
+         # Before this point, $major must not contain `.'.
+         major=.$major
+         versuffix="$major.$revision"
+         ;;
+
+       linux) # correct to gnu/linux during the next big refactor
+         func_arith $current - $age
+         major=.$func_arith_result
+         versuffix="$major.$age.$revision"
+         ;;
+
+       osf)
+         func_arith $current - $age
+         major=.$func_arith_result
+         versuffix=".$current.$age.$revision"
+         verstring="$current.$age.$revision"
+
+         # Add in all the interfaces that we are compatible with.
+         loop=$age
+         while test "$loop" -ne 0; do
+           func_arith $current - $loop
+           iface=$func_arith_result
+           func_arith $loop - 1
+           loop=$func_arith_result
+           verstring="$verstring:${iface}.0"
+         done
+
+         # Make executables depend on our current version.
+         func_append verstring ":${current}.0"
+         ;;
+
+       qnx)
+         major=".$current"
+         versuffix=".$current"
+         ;;
+
+       sunos)
+         major=".$current"
+         versuffix=".$current.$revision"
+         ;;
+
+       windows)
+         # Use '-' rather than '.', since we only want one
+         # extension on DOS 8.3 filesystems.
+         func_arith $current - $age
+         major=$func_arith_result
+         versuffix="-$major"
+         ;;
+
+       *)
+         func_fatal_configuration "unknown library version type \`$version_type'"
+         ;;
+       esac
+
+       # Clear the version info if we defaulted, and they specified a release.
+       if test -z "$vinfo" && test -n "$release"; then
+         major=
+         case $version_type in
+         darwin)
+           # we can't check for "0.0" in archive_cmds due to quoting
+           # problems, so we reset it completely
+           verstring=
+           ;;
+         *)
+           verstring="0.0"
+           ;;
+         esac
+         if test "$need_version" = no; then
+           versuffix=
+         else
+           versuffix=".0.0"
+         fi
+       fi
+
+       # Remove version info from name if versioning should be avoided
+       if test "$avoid_version" = yes && test "$need_version" = no; then
+         major=
+         versuffix=
+         verstring=""
+       fi
+
+       # Check to see if the archive will have undefined symbols.
+       if test "$allow_undefined" = yes; then
+         if test "$allow_undefined_flag" = unsupported; then
+           func_warning "undefined symbols not allowed in $host shared libraries"
+           build_libtool_libs=no
+           build_old_libs=yes
+         fi
+       else
+         # Don't allow undefined symbols.
+         allow_undefined_flag="$no_undefined_flag"
+       fi
+
+      fi
+
+      func_generate_dlsyms "$libname" "$libname" "yes"
+      func_append libobjs " $symfileobj"
+      test "X$libobjs" = "X " && libobjs=
+
+      if test "$opt_mode" != relink; then
+       # Remove our outputs, but don't remove object files since they
+       # may have been created when compiling PIC objects.
+       removelist=
+       tempremovelist=`$ECHO "$output_objdir/*"`
+       for p in $tempremovelist; do
+         case $p in
+           *.$objext | *.gcno)
+              ;;
+           $output_objdir/$outputname | $output_objdir/$libname.* | $output_objdir/${libname}${release}.*)
+              if test "X$precious_files_regex" != "X"; then
+                if $ECHO "$p" | $EGREP -e "$precious_files_regex" >/dev/null 2>&1
+                then
+                  continue
+                fi
+              fi
+              func_append removelist " $p"
+              ;;
+           *) ;;
+         esac
+       done
+       test -n "$removelist" && \
+         func_show_eval "${RM}r \$removelist"
+      fi
+
+      # Now set the variables for building old libraries.
+      if test "$build_old_libs" = yes && test "$build_libtool_libs" != convenience ; then
+       func_append oldlibs " $output_objdir/$libname.$libext"
+
+       # Transform .lo files to .o files.
+       oldobjs="$objs "`$ECHO "$libobjs" | $SP2NL | $SED "/\.${libext}$/d; $lo2o" | $NL2SP`
+      fi
+
+      # Eliminate all temporary directories.
+      #for path in $notinst_path; do
+      #        lib_search_path=`$ECHO "$lib_search_path " | $SED "s% $path % %g"`
+      #        deplibs=`$ECHO "$deplibs " | $SED "s% -L$path % %g"`
+      #        dependency_libs=`$ECHO "$dependency_libs " | $SED "s% -L$path % %g"`
+      #done
+
+      if test -n "$xrpath"; then
+       # If the user specified any rpath flags, then add them.
+       temp_xrpath=
+       for libdir in $xrpath; do
+         func_replace_sysroot "$libdir"
+         func_append temp_xrpath " -R$func_replace_sysroot_result"
+         case "$finalize_rpath " in
+         *" $libdir "*) ;;
+         *) func_append finalize_rpath " $libdir" ;;
+         esac
+       done
+       if test "$hardcode_into_libs" != yes || test "$build_old_libs" = yes; then
+         dependency_libs="$temp_xrpath $dependency_libs"
+       fi
+      fi
+
+      # Make sure dlfiles contains only unique files that won't be dlpreopened
+      old_dlfiles="$dlfiles"
+      dlfiles=
+      for lib in $old_dlfiles; do
+       case " $dlprefiles $dlfiles " in
+       *" $lib "*) ;;
+       *) func_append dlfiles " $lib" ;;
+       esac
+      done
+
+      # Make sure dlprefiles contains only unique files
+      old_dlprefiles="$dlprefiles"
+      dlprefiles=
+      for lib in $old_dlprefiles; do
+       case "$dlprefiles " in
+       *" $lib "*) ;;
+       *) func_append dlprefiles " $lib" ;;
+       esac
+      done
+
+      if test "$build_libtool_libs" = yes; then
+       if test -n "$rpath"; then
+         case $host in
+         *-*-cygwin* | *-*-mingw* | *-*-pw32* | *-*-os2* | *-*-beos* | *-cegcc* | *-*-haiku*)
+           # these systems don't actually have a c library (as such)!
+           ;;
+         *-*-rhapsody* | *-*-darwin1.[012])
+           # Rhapsody C library is in the System framework
+           func_append deplibs " System.ltframework"
+           ;;
+         *-*-netbsd*)
+           # Don't link with libc until the a.out ld.so is fixed.
+           ;;
+         *-*-openbsd* | *-*-freebsd* | *-*-dragonfly*)
+           # Do not include libc due to us having libc/libc_r.
+           ;;
+         *-*-sco3.2v5* | *-*-sco5v6*)
+           # Causes problems with __ctype
+           ;;
+         *-*-sysv4.2uw2* | *-*-sysv5* | *-*-unixware* | *-*-OpenUNIX*)
+           # Compiler inserts libc in the correct place for threads to work
+           ;;
+         *)
+           # Add libc to deplibs on all other systems if necessary.
+           if test "$build_libtool_need_lc" = "yes"; then
+             func_append deplibs " -lc"
+           fi
+           ;;
+         esac
+       fi
+
+       # Transform deplibs into only deplibs that can be linked in shared.
+       name_save=$name
+       libname_save=$libname
+       release_save=$release
+       versuffix_save=$versuffix
+       major_save=$major
+       # I'm not sure if I'm treating the release correctly.  I think
+       # release should show up in the -l (ie -lgmp5) so we don't want to
+       # add it in twice.  Is that correct?
+       release=""
+       versuffix=""
+       major=""
+       newdeplibs=
+       droppeddeps=no
+       case $deplibs_check_method in
+       pass_all)
+         # Don't check for shared/static.  Everything works.
+         # This might be a little naive.  We might want to check
+         # whether the library exists or not.  But this is on
+         # osf3 & osf4 and I'm not really sure... Just
+         # implementing what was already the behavior.
+         newdeplibs=$deplibs
+         ;;
+       test_compile)
+         # This code stresses the "libraries are programs" paradigm to its
+         # limits. Maybe even breaks it.  We compile a program, linking it
+         # against the deplibs as a proxy for the library.  Then we can check
+         # whether they linked in statically or dynamically with ldd.
+         $opt_dry_run || $RM conftest.c
+         cat > conftest.c <<EOF
+         int main() { return 0; }
+EOF
+         $opt_dry_run || $RM conftest
+         if $LTCC $LTCFLAGS -o conftest conftest.c $deplibs; then
+           ldd_output=`ldd conftest`
+           for i in $deplibs; do
+             case $i in
+             -l*)
+               func_stripname -l '' "$i"
+               name=$func_stripname_result
+               if test "X$allow_libtool_libs_with_static_runtimes" = "Xyes" ; then
+                 case " $predeps $postdeps " in
+                 *" $i "*)
+                   func_append newdeplibs " $i"
+                   i=""
+                   ;;
+                 esac
+               fi
+               if test -n "$i" ; then
+                 libname=`eval "\\$ECHO \"$libname_spec\""`
+                 deplib_matches=`eval "\\$ECHO \"$library_names_spec\""`
+                 set dummy $deplib_matches; shift
+                 deplib_match=$1
+                 if test `expr "$ldd_output" : ".*$deplib_match"` -ne 0 ; then
+                   func_append newdeplibs " $i"
+                 else
+                   droppeddeps=yes
+                   echo
+                   $ECHO "*** Warning: dynamic linker does not accept needed library $i."
+                   echo "*** I have the capability to make that library automatically link in when"
+                   echo "*** you link to this library.  But I can only do this if you have a"
+                   echo "*** shared version of the library, which I believe you do not have"
+                   echo "*** because a test_compile did reveal that the linker did not use it for"
+                   echo "*** its dynamic dependency list that programs get resolved with at runtime."
+                 fi
+               fi
+               ;;
+             *)
+               func_append newdeplibs " $i"
+               ;;
+             esac
+           done
+         else
+           # Error occurred in the first compile.  Let's try to salvage
+           # the situation: Compile a separate program for each library.
+           for i in $deplibs; do
+             case $i in
+             -l*)
+               func_stripname -l '' "$i"
+               name=$func_stripname_result
+               $opt_dry_run || $RM conftest
+               if $LTCC $LTCFLAGS -o conftest conftest.c $i; then
+                 ldd_output=`ldd conftest`
+                 if test "X$allow_libtool_libs_with_static_runtimes" = "Xyes" ; then
+                   case " $predeps $postdeps " in
+                   *" $i "*)
+                     func_append newdeplibs " $i"
+                     i=""
+                     ;;
+                   esac
+                 fi
+                 if test -n "$i" ; then
+                   libname=`eval "\\$ECHO \"$libname_spec\""`
+                   deplib_matches=`eval "\\$ECHO \"$library_names_spec\""`
+                   set dummy $deplib_matches; shift
+                   deplib_match=$1
+                   if test `expr "$ldd_output" : ".*$deplib_match"` -ne 0 ; then
+                     func_append newdeplibs " $i"
+                   else
+                     droppeddeps=yes
+                     echo
+                     $ECHO "*** Warning: dynamic linker does not accept needed library $i."
+                     echo "*** I have the capability to make that library automatically link in when"
+                     echo "*** you link to this library.  But I can only do this if you have a"
+                     echo "*** shared version of the library, which you do not appear to have"
+                     echo "*** because a test_compile did reveal that the linker did not use this one"
+                     echo "*** as a dynamic dependency that programs can get resolved with at runtime."
+                   fi
+                 fi
+               else
+                 droppeddeps=yes
+                 echo
+                 $ECHO "*** Warning!  Library $i is needed by this library but I was not able to"
+                 echo "*** make it link in!  You will probably need to install it or some"
+                 echo "*** library that it depends on before this library will be fully"
+                 echo "*** functional.  Installing it before continuing would be even better."
+               fi
+               ;;
+             *)
+               func_append newdeplibs " $i"
+               ;;
+             esac
+           done
+         fi
+         ;;
+       file_magic*)
+         set dummy $deplibs_check_method; shift
+         file_magic_regex=`expr "$deplibs_check_method" : "$1 \(.*\)"`
+         for a_deplib in $deplibs; do
+           case $a_deplib in
+           -l*)
+             func_stripname -l '' "$a_deplib"
+             name=$func_stripname_result
+             if test "X$allow_libtool_libs_with_static_runtimes" = "Xyes" ; then
+               case " $predeps $postdeps " in
+               *" $a_deplib "*)
+                 func_append newdeplibs " $a_deplib"
+                 a_deplib=""
+                 ;;
+               esac
+             fi
+             if test -n "$a_deplib" ; then
+               libname=`eval "\\$ECHO \"$libname_spec\""`
+               if test -n "$file_magic_glob"; then
+                 libnameglob=`func_echo_all "$libname" | $SED -e $file_magic_glob`
+               else
+                 libnameglob=$libname
+               fi
+               test "$want_nocaseglob" = yes && nocaseglob=`shopt -p nocaseglob`
+               for i in $lib_search_path $sys_lib_search_path $shlib_search_path; do
+                 if test "$want_nocaseglob" = yes; then
+                   shopt -s nocaseglob
+                   potential_libs=`ls $i/$libnameglob[.-]* 2>/dev/null`
+                   $nocaseglob
+                 else
+                   potential_libs=`ls $i/$libnameglob[.-]* 2>/dev/null`
+                 fi
+                 for potent_lib in $potential_libs; do
+                     # Follow soft links.
+                     if ls -lLd "$potent_lib" 2>/dev/null |
+                        $GREP " -> " >/dev/null; then
+                       continue
+                     fi
+                     # The statement above tries to avoid entering an
+                     # endless loop below, in case of cyclic links.
+                     # We might still enter an endless loop, since a link
+                     # loop can be closed while we follow links,
+                     # but so what?
+                     potlib="$potent_lib"
+                     while test -h "$potlib" 2>/dev/null; do
+                       potliblink=`ls -ld $potlib | ${SED} 's/.* -> //'`
+                       case $potliblink in
+                       [\\/]* | [A-Za-z]:[\\/]*) potlib="$potliblink";;
+                       *) potlib=`$ECHO "$potlib" | $SED 's,[^/]*$,,'`"$potliblink";;
+                       esac
+                     done
+                     if eval $file_magic_cmd \"\$potlib\" 2>/dev/null |
+                        $SED -e 10q |
+                        $EGREP "$file_magic_regex" > /dev/null; then
+                       func_append newdeplibs " $a_deplib"
+                       a_deplib=""
+                       break 2
+                     fi
+                 done
+               done
+             fi
+             if test -n "$a_deplib" ; then
+               droppeddeps=yes
+               echo
+               $ECHO "*** Warning: linker path does not have real file for library $a_deplib."
+               echo "*** I have the capability to make that library automatically link in when"
+               echo "*** you link to this library.  But I can only do this if you have a"
+               echo "*** shared version of the library, which you do not appear to have"
+               echo "*** because I did check the linker path looking for a file starting"
+               if test -z "$potlib" ; then
+                 $ECHO "*** with $libname but no candidates were found. (...for file magic test)"
+               else
+                 $ECHO "*** with $libname and none of the candidates passed a file format test"
+                 $ECHO "*** using a file magic. Last file checked: $potlib"
+               fi
+             fi
+             ;;
+           *)
+             # Add a -L argument.
+             func_append newdeplibs " $a_deplib"
+             ;;
+           esac
+         done # Gone through all deplibs.
+         ;;
+       match_pattern*)
+         set dummy $deplibs_check_method; shift
+         match_pattern_regex=`expr "$deplibs_check_method" : "$1 \(.*\)"`
+         for a_deplib in $deplibs; do
+           case $a_deplib in
+           -l*)
+             func_stripname -l '' "$a_deplib"
+             name=$func_stripname_result
+             if test "X$allow_libtool_libs_with_static_runtimes" = "Xyes" ; then
+               case " $predeps $postdeps " in
+               *" $a_deplib "*)
+                 func_append newdeplibs " $a_deplib"
+                 a_deplib=""
+                 ;;
+               esac
+             fi
+             if test -n "$a_deplib" ; then
+               libname=`eval "\\$ECHO \"$libname_spec\""`
+               for i in $lib_search_path $sys_lib_search_path $shlib_search_path; do
+                 potential_libs=`ls $i/$libname[.-]* 2>/dev/null`
+                 for potent_lib in $potential_libs; do
+                   potlib="$potent_lib" # see symlink-check above in file_magic test
+                   if eval "\$ECHO \"$potent_lib\"" 2>/dev/null | $SED 10q | \
+                      $EGREP "$match_pattern_regex" > /dev/null; then
+                     func_append newdeplibs " $a_deplib"
+                     a_deplib=""
+                     break 2
+                   fi
+                 done
+               done
+             fi
+             if test -n "$a_deplib" ; then
+               droppeddeps=yes
+               echo
+               $ECHO "*** Warning: linker path does not have real file for library $a_deplib."
+               echo "*** I have the capability to make that library automatically link in when"
+               echo "*** you link to this library.  But I can only do this if you have a"
+               echo "*** shared version of the library, which you do not appear to have"
+               echo "*** because I did check the linker path looking for a file starting"
+               if test -z "$potlib" ; then
+                 $ECHO "*** with $libname but no candidates were found. (...for regex pattern test)"
+               else
+                 $ECHO "*** with $libname and none of the candidates passed a file format test"
+                 $ECHO "*** using a regex pattern. Last file checked: $potlib"
+               fi
+             fi
+             ;;
+           *)
+             # Add a -L argument.
+             func_append newdeplibs " $a_deplib"
+             ;;
+           esac
+         done # Gone through all deplibs.
+         ;;
+       none | unknown | *)
+         newdeplibs=""
+         tmp_deplibs=`$ECHO " $deplibs" | $SED 's/ -lc$//; s/ -[LR][^ ]*//g'`
+         if test "X$allow_libtool_libs_with_static_runtimes" = "Xyes" ; then
+           for i in $predeps $postdeps ; do
+             # can't use Xsed below, because $i might contain '/'
+             tmp_deplibs=`$ECHO " $tmp_deplibs" | $SED "s,$i,,"`
+           done
+         fi
+         case $tmp_deplibs in
+         *[!\  \ ]*)
+           echo
+           if test "X$deplibs_check_method" = "Xnone"; then
+             echo "*** Warning: inter-library dependencies are not supported in this platform."
+           else
+             echo "*** Warning: inter-library dependencies are not known to be supported."
+           fi
+           echo "*** All declared inter-library dependencies are being dropped."
+           droppeddeps=yes
+           ;;
+         esac
+         ;;
+       esac
+       versuffix=$versuffix_save
+       major=$major_save
+       release=$release_save
+       libname=$libname_save
+       name=$name_save
+
+       case $host in
+       *-*-rhapsody* | *-*-darwin1.[012])
+         # On Rhapsody replace the C library with the System framework
+         newdeplibs=`$ECHO " $newdeplibs" | $SED 's/ -lc / System.ltframework /'`
+         ;;
+       esac
+
+       if test "$droppeddeps" = yes; then
+         if test "$module" = yes; then
+           echo
+           echo "*** Warning: libtool could not satisfy all declared inter-library"
+           $ECHO "*** dependencies of module $libname.  Therefore, libtool will create"
+           echo "*** a static module, that should work as long as the dlopening"
+           echo "*** application is linked with the -dlopen flag."
+           if test -z "$global_symbol_pipe"; then
+             echo
+             echo "*** However, this would only work if libtool was able to extract symbol"
+             echo "*** lists from a program, using \`nm' or equivalent, but libtool could"
+             echo "*** not find such a program.  So, this module is probably useless."
+             echo "*** \`nm' from GNU binutils and a full rebuild may help."
+           fi
+           if test "$build_old_libs" = no; then
+             oldlibs="$output_objdir/$libname.$libext"
+             build_libtool_libs=module
+             build_old_libs=yes
+           else
+             build_libtool_libs=no
+           fi
+         else
+           echo "*** The inter-library dependencies that have been dropped here will be"
+           echo "*** automatically added whenever a program is linked with this library"
+           echo "*** or is declared to -dlopen it."
+
+           if test "$allow_undefined" = no; then
+             echo
+             echo "*** Since this library must not contain undefined symbols,"
+             echo "*** because either the platform does not support them or"
+             echo "*** it was explicitly requested with -no-undefined,"
+             echo "*** libtool will only create a static version of it."
+             if test "$build_old_libs" = no; then
+               oldlibs="$output_objdir/$libname.$libext"
+               build_libtool_libs=module
+               build_old_libs=yes
+             else
+               build_libtool_libs=no
+             fi
+           fi
+         fi
+       fi
+       # Done checking deplibs!
+       deplibs=$newdeplibs
+      fi
+      # Time to change all our "foo.ltframework" stuff back to "-framework foo"
+      case $host in
+       *-*-darwin*)
+         newdeplibs=`$ECHO " $newdeplibs" | $SED 's% \([^ $]*\).ltframework% -framework \1%g'`
+         new_inherited_linker_flags=`$ECHO " $new_inherited_linker_flags" | $SED 's% \([^ $]*\).ltframework% -framework \1%g'`
+         deplibs=`$ECHO " $deplibs" | $SED 's% \([^ $]*\).ltframework% -framework \1%g'`
+         ;;
+      esac
+
+      # move library search paths that coincide with paths to not yet
+      # installed libraries to the beginning of the library search list
+      new_libs=
+      for path in $notinst_path; do
+       case " $new_libs " in
+       *" -L$path/$objdir "*) ;;
+       *)
+         case " $deplibs " in
+         *" -L$path/$objdir "*)
+           func_append new_libs " -L$path/$objdir" ;;
+         esac
+         ;;
+       esac
+      done
+      for deplib in $deplibs; do
+       case $deplib in
+       -L*)
+         case " $new_libs " in
+         *" $deplib "*) ;;
+         *) func_append new_libs " $deplib" ;;
+         esac
+         ;;
+       *) func_append new_libs " $deplib" ;;
+       esac
+      done
+      deplibs="$new_libs"
+
+      # All the library-specific variables (install_libdir is set above).
+      library_names=
+      old_library=
+      dlname=
+
+      # Test again, we may have decided not to build it any more
+      if test "$build_libtool_libs" = yes; then
+       # Remove ${wl} instances when linking with ld.
+       # FIXME: should test the right _cmds variable.
+       case $archive_cmds in
+         *\$LD\ *) wl= ;;
+        esac
+       if test "$hardcode_into_libs" = yes; then
+         # Hardcode the library paths
+         hardcode_libdirs=
+         dep_rpath=
+         rpath="$finalize_rpath"
+         test "$opt_mode" != relink && rpath="$compile_rpath$rpath"
+         for libdir in $rpath; do
+           if test -n "$hardcode_libdir_flag_spec"; then
+             if test -n "$hardcode_libdir_separator"; then
+               func_replace_sysroot "$libdir"
+               libdir=$func_replace_sysroot_result
+               if test -z "$hardcode_libdirs"; then
+                 hardcode_libdirs="$libdir"
+               else
+                 # Just accumulate the unique libdirs.
+                 case $hardcode_libdir_separator$hardcode_libdirs$hardcode_libdir_separator in
+                 *"$hardcode_libdir_separator$libdir$hardcode_libdir_separator"*)
+                   ;;
+                 *)
+                   func_append hardcode_libdirs "$hardcode_libdir_separator$libdir"
+                   ;;
+                 esac
+               fi
+             else
+               eval flag=\"$hardcode_libdir_flag_spec\"
+               func_append dep_rpath " $flag"
+             fi
+           elif test -n "$runpath_var"; then
+             case "$perm_rpath " in
+             *" $libdir "*) ;;
+             *) func_append perm_rpath " $libdir" ;;
+             esac
+           fi
+         done
+         # Substitute the hardcoded libdirs into the rpath.
+         if test -n "$hardcode_libdir_separator" &&
+            test -n "$hardcode_libdirs"; then
+           libdir="$hardcode_libdirs"
+           eval "dep_rpath=\"$hardcode_libdir_flag_spec\""
+         fi
+         if test -n "$runpath_var" && test -n "$perm_rpath"; then
+           # We should set the runpath_var.
+           rpath=
+           for dir in $perm_rpath; do
+             func_append rpath "$dir:"
+           done
+           eval "$runpath_var='$rpath\$$runpath_var'; export $runpath_var"
+         fi
+         test -n "$dep_rpath" && deplibs="$dep_rpath $deplibs"
+       fi
+
+       shlibpath="$finalize_shlibpath"
+       test "$opt_mode" != relink && shlibpath="$compile_shlibpath$shlibpath"
+       if test -n "$shlibpath"; then
+         eval "$shlibpath_var='$shlibpath\$$shlibpath_var'; export $shlibpath_var"
+       fi
+
+       # Get the real and link names of the library.
+       eval shared_ext=\"$shrext_cmds\"
+       eval library_names=\"$library_names_spec\"
+       set dummy $library_names
+       shift
+       realname="$1"
+       shift
+
+       if test -n "$soname_spec"; then
+         eval soname=\"$soname_spec\"
+       else
+         soname="$realname"
+       fi
+       if test -z "$dlname"; then
+         dlname=$soname
+       fi
+
+       lib="$output_objdir/$realname"
+       linknames=
+       for link
+       do
+         func_append linknames " $link"
+       done
+
+       # Use standard objects if they are pic
+       test -z "$pic_flag" && libobjs=`$ECHO "$libobjs" | $SP2NL | $SED "$lo2o" | $NL2SP`
+       test "X$libobjs" = "X " && libobjs=
+
+       delfiles=
+       if test -n "$export_symbols" && test -n "$include_expsyms"; then
+         $opt_dry_run || cp "$export_symbols" "$output_objdir/$libname.uexp"
+         export_symbols="$output_objdir/$libname.uexp"
+         func_append delfiles " $export_symbols"
+       fi
+
+       orig_export_symbols=
+       case $host_os in
+       cygwin* | mingw* | cegcc*)
+         if test -n "$export_symbols" && test -z "$export_symbols_regex"; then
+           # exporting using user supplied symfile
+           if test "x`$SED 1q $export_symbols`" != xEXPORTS; then
+             # and it's NOT already a .def file. Must figure out
+             # which of the given symbols are data symbols and tag
+             # them as such. So, trigger use of export_symbols_cmds.
+             # export_symbols gets reassigned inside the "prepare
+             # the list of exported symbols" if statement, so the
+             # include_expsyms logic still works.
+             orig_export_symbols="$export_symbols"
+             export_symbols=
+             always_export_symbols=yes
+           fi
+         fi
+         ;;
+       esac
+
+       # Prepare the list of exported symbols
+       if test -z "$export_symbols"; then
+         if test "$always_export_symbols" = yes || test -n "$export_symbols_regex"; then
+           func_verbose "generating symbol list for \`$libname.la'"
+           export_symbols="$output_objdir/$libname.exp"
+           $opt_dry_run || $RM $export_symbols
+           cmds=$export_symbols_cmds
+           save_ifs="$IFS"; IFS='~'
+           for cmd1 in $cmds; do
+             IFS="$save_ifs"
+             # Take the normal branch if the nm_file_list_spec branch
+             # doesn't work or if tool conversion is not needed.
+             case $nm_file_list_spec~$to_tool_file_cmd in
+               *~func_convert_file_noop | *~func_convert_file_msys_to_w32 | ~*)
+                 try_normal_branch=yes
+                 eval cmd=\"$cmd1\"
+                 func_len " $cmd"
+                 len=$func_len_result
+                 ;;
+               *)
+                 try_normal_branch=no
+                 ;;
+             esac
+             if test "$try_normal_branch" = yes \
+                && { test "$len" -lt "$max_cmd_len" \
+                     || test "$max_cmd_len" -le -1; }
+             then
+               func_show_eval "$cmd" 'exit $?'
+               skipped_export=false
+             elif test -n "$nm_file_list_spec"; then
+               func_basename "$output"
+               output_la=$func_basename_result
+               save_libobjs=$libobjs
+               save_output=$output
+               output=${output_objdir}/${output_la}.nm
+               func_to_tool_file "$output"
+               libobjs=$nm_file_list_spec$func_to_tool_file_result
+               func_append delfiles " $output"
+               func_verbose "creating $NM input file list: $output"
+               for obj in $save_libobjs; do
+                 func_to_tool_file "$obj"
+                 $ECHO "$func_to_tool_file_result"
+               done > "$output"
+               eval cmd=\"$cmd1\"
+               func_show_eval "$cmd" 'exit $?'
+               output=$save_output
+               libobjs=$save_libobjs
+               skipped_export=false
+             else
+               # The command line is too long to execute in one step.
+               func_verbose "using reloadable object file for export list..."
+               skipped_export=:
+               # Break out early, otherwise skipped_export may be
+               # set to false by a later but shorter cmd.
+               break
+             fi
+           done
+           IFS="$save_ifs"
+           if test -n "$export_symbols_regex" && test "X$skipped_export" != "X:"; then
+             func_show_eval '$EGREP -e "$export_symbols_regex" "$export_symbols" > "${export_symbols}T"'
+             func_show_eval '$MV "${export_symbols}T" "$export_symbols"'
+           fi
+         fi
+       fi
+
+       if test -n "$export_symbols" && test -n "$include_expsyms"; then
+         tmp_export_symbols="$export_symbols"
+         test -n "$orig_export_symbols" && tmp_export_symbols="$orig_export_symbols"
+         $opt_dry_run || eval '$ECHO "$include_expsyms" | $SP2NL >> "$tmp_export_symbols"'
+       fi
+
+       if test "X$skipped_export" != "X:" && test -n "$orig_export_symbols"; then
+         # The given exports_symbols file has to be filtered, so filter it.
+         func_verbose "filter symbol list for \`$libname.la' to tag DATA exports"
+         # FIXME: $output_objdir/$libname.filter potentially contains lots of
+         # 's' commands which not all seds can handle. GNU sed should be fine
+         # though. Also, the filter scales superlinearly with the number of
+         # global variables. join(1) would be nice here, but unfortunately
+         # isn't a blessed tool.
+         $opt_dry_run || $SED -e '/[ ,]DATA/!d;s,\(.*\)\([ \,].*\),s|^\1$|\1\2|,' < $export_symbols > $output_objdir/$libname.filter
+         func_append delfiles " $export_symbols $output_objdir/$libname.filter"
+         export_symbols=$output_objdir/$libname.def
+         $opt_dry_run || $SED -f $output_objdir/$libname.filter < $orig_export_symbols > $export_symbols
+       fi
+
+       tmp_deplibs=
+       for test_deplib in $deplibs; do
+         case " $convenience " in
+         *" $test_deplib "*) ;;
+         *)
+           func_append tmp_deplibs " $test_deplib"
+           ;;
+         esac
+       done
+       deplibs="$tmp_deplibs"
+
+       if test -n "$convenience"; then
+         if test -n "$whole_archive_flag_spec" &&
+           test "$compiler_needs_object" = yes &&
+           test -z "$libobjs"; then
+           # extract the archives, so we have objects to list.
+           # TODO: could optimize this to just extract one archive.
+           whole_archive_flag_spec=
+         fi
+         if test -n "$whole_archive_flag_spec"; then
+           save_libobjs=$libobjs
+           eval libobjs=\"\$libobjs $whole_archive_flag_spec\"
+           test "X$libobjs" = "X " && libobjs=
+         else
+           gentop="$output_objdir/${outputname}x"
+           func_append generated " $gentop"
+
+           func_extract_archives $gentop $convenience
+           func_append libobjs " $func_extract_archives_result"
+           test "X$libobjs" = "X " && libobjs=
+         fi
+       fi
+
+       if test "$thread_safe" = yes && test -n "$thread_safe_flag_spec"; then
+         eval flag=\"$thread_safe_flag_spec\"
+         func_append linker_flags " $flag"
+       fi
+
+       # Make a backup of the uninstalled library when relinking
+       if test "$opt_mode" = relink; then
+         $opt_dry_run || eval '(cd $output_objdir && $RM ${realname}U && $MV $realname ${realname}U)' || exit $?
+       fi
+
+       # Do each of the archive commands.
+       if test "$module" = yes && test -n "$module_cmds" ; then
+         if test -n "$export_symbols" && test -n "$module_expsym_cmds"; then
+           eval test_cmds=\"$module_expsym_cmds\"
+           cmds=$module_expsym_cmds
+         else
+           eval test_cmds=\"$module_cmds\"
+           cmds=$module_cmds
+         fi
+       else
+         if test -n "$export_symbols" && test -n "$archive_expsym_cmds"; then
+           eval test_cmds=\"$archive_expsym_cmds\"
+           cmds=$archive_expsym_cmds
+         else
+           eval test_cmds=\"$archive_cmds\"
+           cmds=$archive_cmds
+         fi
+       fi
+
+       if test "X$skipped_export" != "X:" &&
+          func_len " $test_cmds" &&
+          len=$func_len_result &&
+          test "$len" -lt "$max_cmd_len" || test "$max_cmd_len" -le -1; then
+         :
+       else
+         # The command line is too long to link in one step, link piecewise
+         # or, if using GNU ld and skipped_export is not :, use a linker
+         # script.
+
+         # Save the value of $output and $libobjs because we want to
+         # use them later.  If we have whole_archive_flag_spec, we
+         # want to use save_libobjs as it was before
+         # whole_archive_flag_spec was expanded, because we can't
+         # assume the linker understands whole_archive_flag_spec.
+         # This may have to be revisited, in case too many
+         # convenience libraries get linked in and end up exceeding
+         # the spec.
+         if test -z "$convenience" || test -z "$whole_archive_flag_spec"; then
+           save_libobjs=$libobjs
+         fi
+         save_output=$output
+         func_basename "$output"
+         output_la=$func_basename_result
+
+         # Clear the reloadable object creation command queue and
+         # initialize k to one.
+         test_cmds=
+         concat_cmds=
+         objlist=
+         last_robj=
+         k=1
+
+         if test -n "$save_libobjs" && test "X$skipped_export" != "X:" && test "$with_gnu_ld" = yes; then
+           output=${output_objdir}/${output_la}.lnkscript
+           func_verbose "creating GNU ld script: $output"
+           echo 'INPUT (' > $output
+           for obj in $save_libobjs
+           do
+             func_to_tool_file "$obj"
+             $ECHO "$func_to_tool_file_result" >> $output
+           done
+           echo ')' >> $output
+           func_append delfiles " $output"
+           func_to_tool_file "$output"
+           output=$func_to_tool_file_result
+         elif test -n "$save_libobjs" && test "X$skipped_export" != "X:" && test "X$file_list_spec" != X; then
+           output=${output_objdir}/${output_la}.lnk
+           func_verbose "creating linker input file list: $output"
+           : > $output
+           set x $save_libobjs
+           shift
+           firstobj=
+           if test "$compiler_needs_object" = yes; then
+             firstobj="$1 "
+             shift
+           fi
+           for obj
+           do
+             func_to_tool_file "$obj"
+             $ECHO "$func_to_tool_file_result" >> $output
+           done
+           func_append delfiles " $output"
+           func_to_tool_file "$output"
+           output=$firstobj\"$file_list_spec$func_to_tool_file_result\"
+         else
+           if test -n "$save_libobjs"; then
+             func_verbose "creating reloadable object files..."
+             output=$output_objdir/$output_la-${k}.$objext
+             eval test_cmds=\"$reload_cmds\"
+             func_len " $test_cmds"
+             len0=$func_len_result
+             len=$len0
+
+             # Loop over the list of objects to be linked.
+             for obj in $save_libobjs
+             do
+               func_len " $obj"
+               func_arith $len + $func_len_result
+               len=$func_arith_result
+               if test "X$objlist" = X ||
+                  test "$len" -lt "$max_cmd_len"; then
+                 func_append objlist " $obj"
+               else
+                 # The command $test_cmds is almost too long, add a
+                 # command to the queue.
+                 if test "$k" -eq 1 ; then
+                   # The first file doesn't have a previous command to add.
+                   reload_objs=$objlist
+                   eval concat_cmds=\"$reload_cmds\"
+                 else
+                   # All subsequent reloadable object files will link in
+                   # the last one created.
+                   reload_objs="$objlist $last_robj"
+                   eval concat_cmds=\"\$concat_cmds~$reload_cmds~\$RM $last_robj\"
+                 fi
+                 last_robj=$output_objdir/$output_la-${k}.$objext
+                 func_arith $k + 1
+                 k=$func_arith_result
+                 output=$output_objdir/$output_la-${k}.$objext
+                 objlist=" $obj"
+                 func_len " $last_robj"
+                 func_arith $len0 + $func_len_result
+                 len=$func_arith_result
+               fi
+             done
+             # Handle the remaining objects by creating one last
+             # reloadable object file.  All subsequent reloadable object
+             # files will link in the last one created.
+             test -z "$concat_cmds" || concat_cmds=$concat_cmds~
+             reload_objs="$objlist $last_robj"
+             eval concat_cmds=\"\${concat_cmds}$reload_cmds\"
+             if test -n "$last_robj"; then
+               eval concat_cmds=\"\${concat_cmds}~\$RM $last_robj\"
+             fi
+             func_append delfiles " $output"
+
+           else
+             output=
+           fi
+
+           if ${skipped_export-false}; then
+             func_verbose "generating symbol list for \`$libname.la'"
+             export_symbols="$output_objdir/$libname.exp"
+             $opt_dry_run || $RM $export_symbols
+             libobjs=$output
+             # Append the command to create the export file.
+             test -z "$concat_cmds" || concat_cmds=$concat_cmds~
+             eval concat_cmds=\"\$concat_cmds$export_symbols_cmds\"
+             if test -n "$last_robj"; then
+               eval concat_cmds=\"\$concat_cmds~\$RM $last_robj\"
+             fi
+           fi
+
+           test -n "$save_libobjs" &&
+             func_verbose "creating a temporary reloadable object file: $output"
+
+           # Loop through the commands generated above and execute them.
+           save_ifs="$IFS"; IFS='~'
+           for cmd in $concat_cmds; do
+             IFS="$save_ifs"
+             $opt_silent || {
+                 func_quote_for_expand "$cmd"
+                 eval "func_echo $func_quote_for_expand_result"
+             }
+             $opt_dry_run || eval "$cmd" || {
+               lt_exit=$?
+
+               # Restore the uninstalled library and exit
+               if test "$opt_mode" = relink; then
+                 ( cd "$output_objdir" && \
+                   $RM "${realname}T" && \
+                   $MV "${realname}U" "$realname" )
+               fi
+
+               exit $lt_exit
+             }
+           done
+           IFS="$save_ifs"
+
+           if test -n "$export_symbols_regex" && ${skipped_export-false}; then
+             func_show_eval '$EGREP -e "$export_symbols_regex" "$export_symbols" > "${export_symbols}T"'
+             func_show_eval '$MV "${export_symbols}T" "$export_symbols"'
+           fi
+         fi
+
+          if ${skipped_export-false}; then
+           if test -n "$export_symbols" && test -n "$include_expsyms"; then
+             tmp_export_symbols="$export_symbols"
+             test -n "$orig_export_symbols" && tmp_export_symbols="$orig_export_symbols"
+             $opt_dry_run || eval '$ECHO "$include_expsyms" | $SP2NL >> "$tmp_export_symbols"'
+           fi
+
+           if test -n "$orig_export_symbols"; then
+             # The given exports_symbols file has to be filtered, so filter it.
+             func_verbose "filter symbol list for \`$libname.la' to tag DATA exports"
+             # FIXME: $output_objdir/$libname.filter potentially contains lots of
+             # 's' commands which not all seds can handle. GNU sed should be fine
+             # though. Also, the filter scales superlinearly with the number of
+             # global variables. join(1) would be nice here, but unfortunately
+             # isn't a blessed tool.
+             $opt_dry_run || $SED -e '/[ ,]DATA/!d;s,\(.*\)\([ \,].*\),s|^\1$|\1\2|,' < $export_symbols > $output_objdir/$libname.filter
+             func_append delfiles " $export_symbols $output_objdir/$libname.filter"
+             export_symbols=$output_objdir/$libname.def
+             $opt_dry_run || $SED -f $output_objdir/$libname.filter < $orig_export_symbols > $export_symbols
+           fi
+         fi
+
+         libobjs=$output
+         # Restore the value of output.
+         output=$save_output
+
+         if test -n "$convenience" && test -n "$whole_archive_flag_spec"; then
+           eval libobjs=\"\$libobjs $whole_archive_flag_spec\"
+           test "X$libobjs" = "X " && libobjs=
+         fi
+         # Expand the library linking commands again to reset the
+         # value of $libobjs for piecewise linking.
+
+         # Do each of the archive commands.
+         if test "$module" = yes && test -n "$module_cmds" ; then
+           if test -n "$export_symbols" && test -n "$module_expsym_cmds"; then
+             cmds=$module_expsym_cmds
+           else
+             cmds=$module_cmds
+           fi
+         else
+           if test -n "$export_symbols" && test -n "$archive_expsym_cmds"; then
+             cmds=$archive_expsym_cmds
+           else
+             cmds=$archive_cmds
+           fi
+         fi
+       fi
+
+       if test -n "$delfiles"; then
+         # Append the command to remove temporary files to $cmds.
+         eval cmds=\"\$cmds~\$RM $delfiles\"
+       fi
+
+       # Add any objects from preloaded convenience libraries
+       if test -n "$dlprefiles"; then
+         gentop="$output_objdir/${outputname}x"
+         func_append generated " $gentop"
+
+         func_extract_archives $gentop $dlprefiles
+         func_append libobjs " $func_extract_archives_result"
+         test "X$libobjs" = "X " && libobjs=
+       fi
+
+       save_ifs="$IFS"; IFS='~'
+       for cmd in $cmds; do
+         IFS="$save_ifs"
+         eval cmd=\"$cmd\"
+         $opt_silent || {
+           func_quote_for_expand "$cmd"
+           eval "func_echo $func_quote_for_expand_result"
+         }
+         $opt_dry_run || eval "$cmd" || {
+           lt_exit=$?
+
+           # Restore the uninstalled library and exit
+           if test "$opt_mode" = relink; then
+             ( cd "$output_objdir" && \
+               $RM "${realname}T" && \
+               $MV "${realname}U" "$realname" )
+           fi
+
+           exit $lt_exit
+         }
+       done
+       IFS="$save_ifs"
+
+       # Restore the uninstalled library and exit
+       if test "$opt_mode" = relink; then
+         $opt_dry_run || eval '(cd $output_objdir && $RM ${realname}T && $MV $realname ${realname}T && $MV ${realname}U $realname)' || exit $?
+
+         if test -n "$convenience"; then
+           if test -z "$whole_archive_flag_spec"; then
+             func_show_eval '${RM}r "$gentop"'
+           fi
+         fi
+
+         exit $EXIT_SUCCESS
+       fi
+
+       # Create links to the real library.
+       for linkname in $linknames; do
+         if test "$realname" != "$linkname"; then
+           func_show_eval '(cd "$output_objdir" && $RM "$linkname" && $LN_S "$realname" "$linkname")' 'exit $?'
+         fi
+       done
+
+       # If -module or -export-dynamic was specified, set the dlname.
+       if test "$module" = yes || test "$export_dynamic" = yes; then
+         # On all known operating systems, these are identical.
+         dlname="$soname"
+       fi
+      fi
+      ;;
+
+    obj)
+      if test -n "$dlfiles$dlprefiles" || test "$dlself" != no; then
+       func_warning "\`-dlopen' is ignored for objects"
+      fi
+
+      case " $deplibs" in
+      *\ -l* | *\ -L*)
+       func_warning "\`-l' and \`-L' are ignored for objects" ;;
+      esac
+
+      test -n "$rpath" && \
+       func_warning "\`-rpath' is ignored for objects"
+
+      test -n "$xrpath" && \
+       func_warning "\`-R' is ignored for objects"
+
+      test -n "$vinfo" && \
+       func_warning "\`-version-info' is ignored for objects"
+
+      test -n "$release" && \
+       func_warning "\`-release' is ignored for objects"
+
+      case $output in
+      *.lo)
+       test -n "$objs$old_deplibs" && \
+         func_fatal_error "cannot build library object \`$output' from non-libtool objects"
+
+       libobj=$output
+       func_lo2o "$libobj"
+       obj=$func_lo2o_result
+       ;;
+      *)
+       libobj=
+       obj="$output"
+       ;;
+      esac
+
+      # Delete the old objects.
+      $opt_dry_run || $RM $obj $libobj
+
+      # Objects from convenience libraries.  This assumes
+      # single-version convenience libraries.  Whenever we create
+      # different ones for PIC/non-PIC, this we'll have to duplicate
+      # the extraction.
+      reload_conv_objs=
+      gentop=
+      # reload_cmds runs $LD directly, so let us get rid of
+      # -Wl from whole_archive_flag_spec and hope we can get by with
+      # turning comma into space..
+      wl=
+
+      if test -n "$convenience"; then
+       if test -n "$whole_archive_flag_spec"; then
+         eval tmp_whole_archive_flags=\"$whole_archive_flag_spec\"
+         reload_conv_objs=$reload_objs\ `$ECHO "$tmp_whole_archive_flags" | $SED 's|,| |g'`
+       else
+         gentop="$output_objdir/${obj}x"
+         func_append generated " $gentop"
+
+         func_extract_archives $gentop $convenience
+         reload_conv_objs="$reload_objs $func_extract_archives_result"
+       fi
+      fi
+
+      # If we're not building shared, we need to use non_pic_objs
+      test "$build_libtool_libs" != yes && libobjs="$non_pic_objects"
+
+      # Create the old-style object.
+      reload_objs="$objs$old_deplibs "`$ECHO "$libobjs" | $SP2NL | $SED "/\.${libext}$/d; /\.lib$/d; $lo2o" | $NL2SP`" $reload_conv_objs" ### testsuite: skip nested quoting test
+
+      output="$obj"
+      func_execute_cmds "$reload_cmds" 'exit $?'
+
+      # Exit if we aren't doing a library object file.
+      if test -z "$libobj"; then
+       if test -n "$gentop"; then
+         func_show_eval '${RM}r "$gentop"'
+       fi
+
+       exit $EXIT_SUCCESS
+      fi
+
+      if test "$build_libtool_libs" != yes; then
+       if test -n "$gentop"; then
+         func_show_eval '${RM}r "$gentop"'
+       fi
+
+       # Create an invalid libtool object if no PIC, so that we don't
+       # accidentally link it into a program.
+       # $show "echo timestamp > $libobj"
+       # $opt_dry_run || eval "echo timestamp > $libobj" || exit $?
+       exit $EXIT_SUCCESS
+      fi
+
+      if test -n "$pic_flag" || test "$pic_mode" != default; then
+       # Only do commands if we really have different PIC objects.
+       reload_objs="$libobjs $reload_conv_objs"
+       output="$libobj"
+       func_execute_cmds "$reload_cmds" 'exit $?'
+      fi
+
+      if test -n "$gentop"; then
+       func_show_eval '${RM}r "$gentop"'
+      fi
+
+      exit $EXIT_SUCCESS
+      ;;
+
+    prog)
+      case $host in
+       *cygwin*) func_stripname '' '.exe' "$output"
+                 output=$func_stripname_result.exe;;
+      esac
+      test -n "$vinfo" && \
+       func_warning "\`-version-info' is ignored for programs"
+
+      test -n "$release" && \
+       func_warning "\`-release' is ignored for programs"
+
+      test "$preload" = yes \
+        && test "$dlopen_support" = unknown \
+       && test "$dlopen_self" = unknown \
+       && test "$dlopen_self_static" = unknown && \
+         func_warning "\`LT_INIT([dlopen])' not used. Assuming no dlopen support."
+
+      case $host in
+      *-*-rhapsody* | *-*-darwin1.[012])
+       # On Rhapsody replace the C library is the System framework
+       compile_deplibs=`$ECHO " $compile_deplibs" | $SED 's/ -lc / System.ltframework /'`
+       finalize_deplibs=`$ECHO " $finalize_deplibs" | $SED 's/ -lc / System.ltframework /'`
+       ;;
+      esac
+
+      case $host in
+      *-*-darwin*)
+       # Don't allow lazy linking, it breaks C++ global constructors
+       # But is supposedly fixed on 10.4 or later (yay!).
+       if test "$tagname" = CXX ; then
+         case ${MACOSX_DEPLOYMENT_TARGET-10.0} in
+           10.[0123])
+             func_append compile_command " ${wl}-bind_at_load"
+             func_append finalize_command " ${wl}-bind_at_load"
+           ;;
+         esac
+       fi
+       # Time to change all our "foo.ltframework" stuff back to "-framework foo"
+       compile_deplibs=`$ECHO " $compile_deplibs" | $SED 's% \([^ $]*\).ltframework% -framework \1%g'`
+       finalize_deplibs=`$ECHO " $finalize_deplibs" | $SED 's% \([^ $]*\).ltframework% -framework \1%g'`
+       ;;
+      esac
+
+
+      # move library search paths that coincide with paths to not yet
+      # installed libraries to the beginning of the library search list
+      new_libs=
+      for path in $notinst_path; do
+       case " $new_libs " in
+       *" -L$path/$objdir "*) ;;
+       *)
+         case " $compile_deplibs " in
+         *" -L$path/$objdir "*)
+           func_append new_libs " -L$path/$objdir" ;;
+         esac
+         ;;
+       esac
+      done
+      for deplib in $compile_deplibs; do
+       case $deplib in
+       -L*)
+         case " $new_libs " in
+         *" $deplib "*) ;;
+         *) func_append new_libs " $deplib" ;;
+         esac
+         ;;
+       *) func_append new_libs " $deplib" ;;
+       esac
+      done
+      compile_deplibs="$new_libs"
+
+
+      func_append compile_command " $compile_deplibs"
+      func_append finalize_command " $finalize_deplibs"
+
+      if test -n "$rpath$xrpath"; then
+       # If the user specified any rpath flags, then add them.
+       for libdir in $rpath $xrpath; do
+         # This is the magic to use -rpath.
+         case "$finalize_rpath " in
+         *" $libdir "*) ;;
+         *) func_append finalize_rpath " $libdir" ;;
+         esac
+       done
+      fi
+
+      # Now hardcode the library paths
+      rpath=
+      hardcode_libdirs=
+      for libdir in $compile_rpath $finalize_rpath; do
+       if test -n "$hardcode_libdir_flag_spec"; then
+         if test -n "$hardcode_libdir_separator"; then
+           if test -z "$hardcode_libdirs"; then
+             hardcode_libdirs="$libdir"
+           else
+             # Just accumulate the unique libdirs.
+             case $hardcode_libdir_separator$hardcode_libdirs$hardcode_libdir_separator in
+             *"$hardcode_libdir_separator$libdir$hardcode_libdir_separator"*)
+               ;;
+             *)
+               func_append hardcode_libdirs "$hardcode_libdir_separator$libdir"
+               ;;
+             esac
+           fi
+         else
+           eval flag=\"$hardcode_libdir_flag_spec\"
+           func_append rpath " $flag"
+         fi
+       elif test -n "$runpath_var"; then
+         case "$perm_rpath " in
+         *" $libdir "*) ;;
+         *) func_append perm_rpath " $libdir" ;;
+         esac
+       fi
+       case $host in
+       *-*-cygwin* | *-*-mingw* | *-*-pw32* | *-*-os2* | *-cegcc*)
+         testbindir=`${ECHO} "$libdir" | ${SED} -e 's*/lib$*/bin*'`
+         case :$dllsearchpath: in
+         *":$libdir:"*) ;;
+         ::) dllsearchpath=$libdir;;
+         *) func_append dllsearchpath ":$libdir";;
+         esac
+         case :$dllsearchpath: in
+         *":$testbindir:"*) ;;
+         ::) dllsearchpath=$testbindir;;
+         *) func_append dllsearchpath ":$testbindir";;
+         esac
+         ;;
+       esac
+      done
+      # Substitute the hardcoded libdirs into the rpath.
+      if test -n "$hardcode_libdir_separator" &&
+        test -n "$hardcode_libdirs"; then
+       libdir="$hardcode_libdirs"
+       eval rpath=\" $hardcode_libdir_flag_spec\"
+      fi
+      compile_rpath="$rpath"
+
+      rpath=
+      hardcode_libdirs=
+      for libdir in $finalize_rpath; do
+       if test -n "$hardcode_libdir_flag_spec"; then
+         if test -n "$hardcode_libdir_separator"; then
+           if test -z "$hardcode_libdirs"; then
+             hardcode_libdirs="$libdir"
+           else
+             # Just accumulate the unique libdirs.
+             case $hardcode_libdir_separator$hardcode_libdirs$hardcode_libdir_separator in
+             *"$hardcode_libdir_separator$libdir$hardcode_libdir_separator"*)
+               ;;
+             *)
+               func_append hardcode_libdirs "$hardcode_libdir_separator$libdir"
+               ;;
+             esac
+           fi
+         else
+           eval flag=\"$hardcode_libdir_flag_spec\"
+           func_append rpath " $flag"
+         fi
+       elif test -n "$runpath_var"; then
+         case "$finalize_perm_rpath " in
+         *" $libdir "*) ;;
+         *) func_append finalize_perm_rpath " $libdir" ;;
+         esac
+       fi
+      done
+      # Substitute the hardcoded libdirs into the rpath.
+      if test -n "$hardcode_libdir_separator" &&
+        test -n "$hardcode_libdirs"; then
+       libdir="$hardcode_libdirs"
+       eval rpath=\" $hardcode_libdir_flag_spec\"
+      fi
+      finalize_rpath="$rpath"
+
+      if test -n "$libobjs" && test "$build_old_libs" = yes; then
+       # Transform all the library objects into standard objects.
+       compile_command=`$ECHO "$compile_command" | $SP2NL | $SED "$lo2o" | $NL2SP`
+       finalize_command=`$ECHO "$finalize_command" | $SP2NL | $SED "$lo2o" | $NL2SP`
+      fi
+
+      func_generate_dlsyms "$outputname" "@PROGRAM@" "no"
+
+      # template prelinking step
+      if test -n "$prelink_cmds"; then
+       func_execute_cmds "$prelink_cmds" 'exit $?'
+      fi
+
+      wrappers_required=yes
+      case $host in
+      *cegcc* | *mingw32ce*)
+        # Disable wrappers for cegcc and mingw32ce hosts, we are cross compiling anyway.
+        wrappers_required=no
+        ;;
+      *cygwin* | *mingw* )
+        if test "$build_libtool_libs" != yes; then
+          wrappers_required=no
+        fi
+        ;;
+      *)
+        if test "$need_relink" = no || test "$build_libtool_libs" != yes; then
+          wrappers_required=no
+        fi
+        ;;
+      esac
+      if test "$wrappers_required" = no; then
+       # Replace the output file specification.
+       compile_command=`$ECHO "$compile_command" | $SED 's%@OUTPUT@%'"$output"'%g'`
+       link_command="$compile_command$compile_rpath"
+
+       # We have no uninstalled library dependencies, so finalize right now.
+       exit_status=0
+       func_show_eval "$link_command" 'exit_status=$?'
+
+       if test -n "$postlink_cmds"; then
+         func_to_tool_file "$output"
+         postlink_cmds=`func_echo_all "$postlink_cmds" | $SED -e 's%@OUTPUT@%'"$output"'%g' -e 's%@TOOL_OUTPUT@%'"$func_to_tool_file_result"'%g'`
+         func_execute_cmds "$postlink_cmds" 'exit $?'
+       fi
+
+       # Delete the generated files.
+       if test -f "$output_objdir/${outputname}S.${objext}"; then
+         func_show_eval '$RM "$output_objdir/${outputname}S.${objext}"'
+       fi
+
+       exit $exit_status
+      fi
+
+      if test -n "$compile_shlibpath$finalize_shlibpath"; then
+       compile_command="$shlibpath_var=\"$compile_shlibpath$finalize_shlibpath\$$shlibpath_var\" $compile_command"
+      fi
+      if test -n "$finalize_shlibpath"; then
+       finalize_command="$shlibpath_var=\"$finalize_shlibpath\$$shlibpath_var\" $finalize_command"
+      fi
+
+      compile_var=
+      finalize_var=
+      if test -n "$runpath_var"; then
+       if test -n "$perm_rpath"; then
+         # We should set the runpath_var.
+         rpath=
+         for dir in $perm_rpath; do
+           func_append rpath "$dir:"
+         done
+         compile_var="$runpath_var=\"$rpath\$$runpath_var\" "
+       fi
+       if test -n "$finalize_perm_rpath"; then
+         # We should set the runpath_var.
+         rpath=
+         for dir in $finalize_perm_rpath; do
+           func_append rpath "$dir:"
+         done
+         finalize_var="$runpath_var=\"$rpath\$$runpath_var\" "
+       fi
+      fi
+
+      if test "$no_install" = yes; then
+       # We don't need to create a wrapper script.
+       link_command="$compile_var$compile_command$compile_rpath"
+       # Replace the output file specification.
+       link_command=`$ECHO "$link_command" | $SED 's%@OUTPUT@%'"$output"'%g'`
+       # Delete the old output file.
+       $opt_dry_run || $RM $output
+       # Link the executable and exit
+       func_show_eval "$link_command" 'exit $?'
+
+       if test -n "$postlink_cmds"; then
+         func_to_tool_file "$output"
+         postlink_cmds=`func_echo_all "$postlink_cmds" | $SED -e 's%@OUTPUT@%'"$output"'%g' -e 's%@TOOL_OUTPUT@%'"$func_to_tool_file_result"'%g'`
+         func_execute_cmds "$postlink_cmds" 'exit $?'
+       fi
+
+       exit $EXIT_SUCCESS
+      fi
+
+      if test "$hardcode_action" = relink; then
+       # Fast installation is not supported
+       link_command="$compile_var$compile_command$compile_rpath"
+       relink_command="$finalize_var$finalize_command$finalize_rpath"
+
+       func_warning "this platform does not like uninstalled shared libraries"
+       func_warning "\`$output' will be relinked during installation"
+      else
+       if test "$fast_install" != no; then
+         link_command="$finalize_var$compile_command$finalize_rpath"
+         if test "$fast_install" = yes; then
+           relink_command=`$ECHO "$compile_var$compile_command$compile_rpath" | $SED 's%@OUTPUT@%\$progdir/\$file%g'`
+         else
+           # fast_install is set to needless
+           relink_command=
+         fi
+       else
+         link_command="$compile_var$compile_command$compile_rpath"
+         relink_command="$finalize_var$finalize_command$finalize_rpath"
+       fi
+      fi
+
+      # Replace the output file specification.
+      link_command=`$ECHO "$link_command" | $SED 's%@OUTPUT@%'"$output_objdir/$outputname"'%g'`
+
+      # Delete the old output files.
+      $opt_dry_run || $RM $output $output_objdir/$outputname $output_objdir/lt-$outputname
+
+      func_show_eval "$link_command" 'exit $?'
+
+      if test -n "$postlink_cmds"; then
+       func_to_tool_file "$output_objdir/$outputname"
+       postlink_cmds=`func_echo_all "$postlink_cmds" | $SED -e 's%@OUTPUT@%'"$output_objdir/$outputname"'%g' -e 's%@TOOL_OUTPUT@%'"$func_to_tool_file_result"'%g'`
+       func_execute_cmds "$postlink_cmds" 'exit $?'
+      fi
+
+      # Now create the wrapper script.
+      func_verbose "creating $output"
+
+      # Quote the relink command for shipping.
+      if test -n "$relink_command"; then
+       # Preserve any variables that may affect compiler behavior
+       for var in $variables_saved_for_relink; do
+         if eval test -z \"\${$var+set}\"; then
+           relink_command="{ test -z \"\${$var+set}\" || $lt_unset $var || { $var=; export $var; }; }; $relink_command"
+         elif eval var_value=\$$var; test -z "$var_value"; then
+           relink_command="$var=; export $var; $relink_command"
+         else
+           func_quote_for_eval "$var_value"
+           relink_command="$var=$func_quote_for_eval_result; export $var; $relink_command"
+         fi
+       done
+       relink_command="(cd `pwd`; $relink_command)"
+       relink_command=`$ECHO "$relink_command" | $SED "$sed_quote_subst"`
+      fi
+
+      # Only actually do things if not in dry run mode.
+      $opt_dry_run || {
+       # win32 will think the script is a binary if it has
+       # a .exe suffix, so we strip it off here.
+       case $output in
+         *.exe) func_stripname '' '.exe' "$output"
+                output=$func_stripname_result ;;
+       esac
+       # test for cygwin because mv fails w/o .exe extensions
+       case $host in
+         *cygwin*)
+           exeext=.exe
+           func_stripname '' '.exe' "$outputname"
+           outputname=$func_stripname_result ;;
+         *) exeext= ;;
+       esac
+       case $host in
+         *cygwin* | *mingw* )
+           func_dirname_and_basename "$output" "" "."
+           output_name=$func_basename_result
+           output_path=$func_dirname_result
+           cwrappersource="$output_path/$objdir/lt-$output_name.c"
+           cwrapper="$output_path/$output_name.exe"
+           $RM $cwrappersource $cwrapper
+           trap "$RM $cwrappersource $cwrapper; exit $EXIT_FAILURE" 1 2 15
+
+           func_emit_cwrapperexe_src > $cwrappersource
+
+           # The wrapper executable is built using the $host compiler,
+           # because it contains $host paths and files. If cross-
+           # compiling, it, like the target executable, must be
+           # executed on the $host or under an emulation environment.
+           $opt_dry_run || {
+             $LTCC $LTCFLAGS -o $cwrapper $cwrappersource
+             $STRIP $cwrapper
+           }
+
+           # Now, create the wrapper script for func_source use:
+           func_ltwrapper_scriptname $cwrapper
+           $RM $func_ltwrapper_scriptname_result
+           trap "$RM $func_ltwrapper_scriptname_result; exit $EXIT_FAILURE" 1 2 15
+           $opt_dry_run || {
+             # note: this script will not be executed, so do not chmod.
+             if test "x$build" = "x$host" ; then
+               $cwrapper --lt-dump-script > $func_ltwrapper_scriptname_result
+             else
+               func_emit_wrapper no > $func_ltwrapper_scriptname_result
+             fi
+           }
+         ;;
+         * )
+           $RM $output
+           trap "$RM $output; exit $EXIT_FAILURE" 1 2 15
+
+           func_emit_wrapper no > $output
+           chmod +x $output
+         ;;
+       esac
+      }
+      exit $EXIT_SUCCESS
+      ;;
+    esac
+
+    # See if we need to build an old-fashioned archive.
+    for oldlib in $oldlibs; do
+
+      if test "$build_libtool_libs" = convenience; then
+       oldobjs="$libobjs_save $symfileobj"
+       addlibs="$convenience"
+       build_libtool_libs=no
+      else
+       if test "$build_libtool_libs" = module; then
+         oldobjs="$libobjs_save"
+         build_libtool_libs=no
+       else
+         oldobjs="$old_deplibs $non_pic_objects"
+         if test "$preload" = yes && test -f "$symfileobj"; then
+           func_append oldobjs " $symfileobj"
+         fi
+       fi
+       addlibs="$old_convenience"
+      fi
+
+      if test -n "$addlibs"; then
+       gentop="$output_objdir/${outputname}x"
+       func_append generated " $gentop"
+
+       func_extract_archives $gentop $addlibs
+       func_append oldobjs " $func_extract_archives_result"
+      fi
+
+      # Do each command in the archive commands.
+      if test -n "$old_archive_from_new_cmds" && test "$build_libtool_libs" = yes; then
+       cmds=$old_archive_from_new_cmds
+      else
+
+       # Add any objects from preloaded convenience libraries
+       if test -n "$dlprefiles"; then
+         gentop="$output_objdir/${outputname}x"
+         func_append generated " $gentop"
+
+         func_extract_archives $gentop $dlprefiles
+         func_append oldobjs " $func_extract_archives_result"
+       fi
+
+       # POSIX demands no paths to be encoded in archives.  We have
+       # to avoid creating archives with duplicate basenames if we
+       # might have to extract them afterwards, e.g., when creating a
+       # static archive out of a convenience library, or when linking
+       # the entirety of a libtool archive into another (currently
+       # not supported by libtool).
+       if (for obj in $oldobjs
+           do
+             func_basename "$obj"
+             $ECHO "$func_basename_result"
+           done | sort | sort -uc >/dev/null 2>&1); then
+         :
+       else
+         echo "copying selected object files to avoid basename conflicts..."
+         gentop="$output_objdir/${outputname}x"
+         func_append generated " $gentop"
+         func_mkdir_p "$gentop"
+         save_oldobjs=$oldobjs
+         oldobjs=
+         counter=1
+         for obj in $save_oldobjs
+         do
+           func_basename "$obj"
+           objbase="$func_basename_result"
+           case " $oldobjs " in
+           " ") oldobjs=$obj ;;
+           *[\ /]"$objbase "*)
+             while :; do
+               # Make sure we don't pick an alternate name that also
+               # overlaps.
+               newobj=lt$counter-$objbase
+               func_arith $counter + 1
+               counter=$func_arith_result
+               case " $oldobjs " in
+               *[\ /]"$newobj "*) ;;
+               *) if test ! -f "$gentop/$newobj"; then break; fi ;;
+               esac
+             done
+             func_show_eval "ln $obj $gentop/$newobj || cp $obj $gentop/$newobj"
+             func_append oldobjs " $gentop/$newobj"
+             ;;
+           *) func_append oldobjs " $obj" ;;
+           esac
+         done
+       fi
+       func_to_tool_file "$oldlib" func_convert_file_msys_to_w32
+       tool_oldlib=$func_to_tool_file_result
+       eval cmds=\"$old_archive_cmds\"
+
+       func_len " $cmds"
+       len=$func_len_result
+       if test "$len" -lt "$max_cmd_len" || test "$max_cmd_len" -le -1; then
+         cmds=$old_archive_cmds
+       elif test -n "$archiver_list_spec"; then
+         func_verbose "using command file archive linking..."
+         for obj in $oldobjs
+         do
+           func_to_tool_file "$obj"
+           $ECHO "$func_to_tool_file_result"
+         done > $output_objdir/$libname.libcmd
+         func_to_tool_file "$output_objdir/$libname.libcmd"
+         oldobjs=" $archiver_list_spec$func_to_tool_file_result"
+         cmds=$old_archive_cmds
+       else
+         # the command line is too long to link in one step, link in parts
+         func_verbose "using piecewise archive linking..."
+         save_RANLIB=$RANLIB
+         RANLIB=:
+         objlist=
+         concat_cmds=
+         save_oldobjs=$oldobjs
+         oldobjs=
+         # Is there a better way of finding the last object in the list?
+         for obj in $save_oldobjs
+         do
+           last_oldobj=$obj
+         done
+         eval test_cmds=\"$old_archive_cmds\"
+         func_len " $test_cmds"
+         len0=$func_len_result
+         len=$len0
+         for obj in $save_oldobjs
+         do
+           func_len " $obj"
+           func_arith $len + $func_len_result
+           len=$func_arith_result
+           func_append objlist " $obj"
+           if test "$len" -lt "$max_cmd_len"; then
+             :
+           else
+             # the above command should be used before it gets too long
+             oldobjs=$objlist
+             if test "$obj" = "$last_oldobj" ; then
+               RANLIB=$save_RANLIB
+             fi
+             test -z "$concat_cmds" || concat_cmds=$concat_cmds~
+             eval concat_cmds=\"\${concat_cmds}$old_archive_cmds\"
+             objlist=
+             len=$len0
+           fi
+         done
+         RANLIB=$save_RANLIB
+         oldobjs=$objlist
+         if test "X$oldobjs" = "X" ; then
+           eval cmds=\"\$concat_cmds\"
+         else
+           eval cmds=\"\$concat_cmds~\$old_archive_cmds\"
+         fi
+       fi
+      fi
+      func_execute_cmds "$cmds" 'exit $?'
+    done
+
+    test -n "$generated" && \
+      func_show_eval "${RM}r$generated"
+
+    # Now create the libtool archive.
+    case $output in
+    *.la)
+      old_library=
+      test "$build_old_libs" = yes && old_library="$libname.$libext"
+      func_verbose "creating $output"
+
+      # Preserve any variables that may affect compiler behavior
+      for var in $variables_saved_for_relink; do
+       if eval test -z \"\${$var+set}\"; then
+         relink_command="{ test -z \"\${$var+set}\" || $lt_unset $var || { $var=; export $var; }; }; $relink_command"
+       elif eval var_value=\$$var; test -z "$var_value"; then
+         relink_command="$var=; export $var; $relink_command"
+       else
+         func_quote_for_eval "$var_value"
+         relink_command="$var=$func_quote_for_eval_result; export $var; $relink_command"
+       fi
+      done
+      # Quote the link command for shipping.
+      relink_command="(cd `pwd`; $SHELL $progpath $preserve_args --mode=relink $libtool_args @inst_prefix_dir@)"
+      relink_command=`$ECHO "$relink_command" | $SED "$sed_quote_subst"`
+      if test "$hardcode_automatic" = yes ; then
+       relink_command=
+      fi
+
+      # Only create the output if not a dry run.
+      $opt_dry_run || {
+       for installed in no yes; do
+         if test "$installed" = yes; then
+           if test -z "$install_libdir"; then
+             break
+           fi
+           output="$output_objdir/$outputname"i
+           # Replace all uninstalled libtool libraries with the installed ones
+           newdependency_libs=
+           for deplib in $dependency_libs; do
+             case $deplib in
+             *.la)
+               func_basename "$deplib"
+               name="$func_basename_result"
+               func_resolve_sysroot "$deplib"
+               eval libdir=`${SED} -n -e 's/^libdir=\(.*\)$/\1/p' $func_resolve_sysroot_result`
+               test -z "$libdir" && \
+                 func_fatal_error "\`$deplib' is not a valid libtool archive"
+               func_append newdependency_libs " ${lt_sysroot:+=}$libdir/$name"
+               ;;
+             -L*)
+               func_stripname -L '' "$deplib"
+               func_replace_sysroot "$func_stripname_result"
+               func_append newdependency_libs " -L$func_replace_sysroot_result"
+               ;;
+             -R*)
+               func_stripname -R '' "$deplib"
+               func_replace_sysroot "$func_stripname_result"
+               func_append newdependency_libs " -R$func_replace_sysroot_result"
+               ;;
+             *) func_append newdependency_libs " $deplib" ;;
+             esac
+           done
+           dependency_libs="$newdependency_libs"
+           newdlfiles=
+
+           for lib in $dlfiles; do
+             case $lib in
+             *.la)
+               func_basename "$lib"
+               name="$func_basename_result"
+               eval libdir=`${SED} -n -e 's/^libdir=\(.*\)$/\1/p' $lib`
+               test -z "$libdir" && \
+                 func_fatal_error "\`$lib' is not a valid libtool archive"
+               func_append newdlfiles " ${lt_sysroot:+=}$libdir/$name"
+               ;;
+             *) func_append newdlfiles " $lib" ;;
+             esac
+           done
+           dlfiles="$newdlfiles"
+           newdlprefiles=
+           for lib in $dlprefiles; do
+             case $lib in
+             *.la)
+               # Only pass preopened files to the pseudo-archive (for
+               # eventual linking with the app. that links it) if we
+               # didn't already link the preopened objects directly into
+               # the library:
+               func_basename "$lib"
+               name="$func_basename_result"
+               eval libdir=`${SED} -n -e 's/^libdir=\(.*\)$/\1/p' $lib`
+               test -z "$libdir" && \
+                 func_fatal_error "\`$lib' is not a valid libtool archive"
+               func_append newdlprefiles " ${lt_sysroot:+=}$libdir/$name"
+               ;;
+             esac
+           done
+           dlprefiles="$newdlprefiles"
+         else
+           newdlfiles=
+           for lib in $dlfiles; do
+             case $lib in
+               [\\/]* | [A-Za-z]:[\\/]*) abs="$lib" ;;
+               *) abs=`pwd`"/$lib" ;;
+             esac
+             func_append newdlfiles " $abs"
+           done
+           dlfiles="$newdlfiles"
+           newdlprefiles=
+           for lib in $dlprefiles; do
+             case $lib in
+               [\\/]* | [A-Za-z]:[\\/]*) abs="$lib" ;;
+               *) abs=`pwd`"/$lib" ;;
+             esac
+             func_append newdlprefiles " $abs"
+           done
+           dlprefiles="$newdlprefiles"
+         fi
+         $RM $output
+         # place dlname in correct position for cygwin
+         # In fact, it would be nice if we could use this code for all target
+         # systems that can't hard-code library paths into their executables
+         # and that have no shared library path variable independent of PATH,
+         # but it turns out we can't easily determine that from inspecting
+         # libtool variables, so we have to hard-code the OSs to which it
+         # applies here; at the moment, that means platforms that use the PE
+         # object format with DLL files.  See the long comment at the top of
+         # tests/bindir.at for full details.
+         tdlname=$dlname
+         case $host,$output,$installed,$module,$dlname in
+           *cygwin*,*lai,yes,no,*.dll | *mingw*,*lai,yes,no,*.dll | *cegcc*,*lai,yes,no,*.dll)
+             # If a -bindir argument was supplied, place the dll there.
+             if test "x$bindir" != x ;
+             then
+               func_relative_path "$install_libdir" "$bindir"
+               tdlname=$func_relative_path_result$dlname
+             else
+               # Otherwise fall back on heuristic.
+               tdlname=../bin/$dlname
+             fi
+             ;;
+         esac
+         $ECHO > $output "\
+# $outputname - a libtool library file
+# Generated by $PROGRAM (GNU $PACKAGE$TIMESTAMP) $VERSION
+#
+# Please DO NOT delete this file!
+# It is necessary for linking the library.
+
+# The name that we can dlopen(3).
+dlname='$tdlname'
+
+# Names of this library.
+library_names='$library_names'
+
+# The name of the static archive.
+old_library='$old_library'
+
+# Linker flags that can not go in dependency_libs.
+inherited_linker_flags='$new_inherited_linker_flags'
+
+# Libraries that this one depends upon.
+dependency_libs='$dependency_libs'
+
+# Names of additional weak libraries provided by this library
+weak_library_names='$weak_libs'
+
+# Version information for $libname.
+current=$current
+age=$age
+revision=$revision
+
+# Is this an already installed library?
+installed=$installed
+
+# Should we warn about portability when linking against -modules?
+shouldnotlink=$module
+
+# Files to dlopen/dlpreopen
+dlopen='$dlfiles'
+dlpreopen='$dlprefiles'
+
+# Directory that this library needs to be installed in:
+libdir='$install_libdir'"
+         if test "$installed" = no && test "$need_relink" = yes; then
+           $ECHO >> $output "\
+relink_command=\"$relink_command\""
+         fi
+       done
+      }
+
+      # Do a symbolic link so that the libtool archive can be found in
+      # LD_LIBRARY_PATH before the program is installed.
+      func_show_eval '( cd "$output_objdir" && $RM "$outputname" && $LN_S "../$outputname" "$outputname" )' 'exit $?'
+      ;;
+    esac
+    exit $EXIT_SUCCESS
+}
+
+{ test "$opt_mode" = link || test "$opt_mode" = relink; } &&
+    func_mode_link ${1+"$@"}
+
+
+# func_mode_uninstall arg...
+func_mode_uninstall ()
+{
+    $opt_debug
+    RM="$nonopt"
+    files=
+    rmforce=
+    exit_status=0
+
+    # This variable tells wrapper scripts just to set variables rather
+    # than running their programs.
+    libtool_install_magic="$magic"
+
+    for arg
+    do
+      case $arg in
+      -f) func_append RM " $arg"; rmforce=yes ;;
+      -*) func_append RM " $arg" ;;
+      *) func_append files " $arg" ;;
+      esac
+    done
+
+    test -z "$RM" && \
+      func_fatal_help "you must specify an RM program"
+
+    rmdirs=
+
+    for file in $files; do
+      func_dirname "$file" "" "."
+      dir="$func_dirname_result"
+      if test "X$dir" = X.; then
+       odir="$objdir"
+      else
+       odir="$dir/$objdir"
+      fi
+      func_basename "$file"
+      name="$func_basename_result"
+      test "$opt_mode" = uninstall && odir="$dir"
+
+      # Remember odir for removal later, being careful to avoid duplicates
+      if test "$opt_mode" = clean; then
+       case " $rmdirs " in
+         *" $odir "*) ;;
+         *) func_append rmdirs " $odir" ;;
+       esac
+      fi
+
+      # Don't error if the file doesn't exist and rm -f was used.
+      if { test -L "$file"; } >/dev/null 2>&1 ||
+        { test -h "$file"; } >/dev/null 2>&1 ||
+        test -f "$file"; then
+       :
+      elif test -d "$file"; then
+       exit_status=1
+       continue
+      elif test "$rmforce" = yes; then
+       continue
+      fi
+
+      rmfiles="$file"
+
+      case $name in
+      *.la)
+       # Possibly a libtool archive, so verify it.
+       if func_lalib_p "$file"; then
+         func_source $dir/$name
+
+         # Delete the libtool libraries and symlinks.
+         for n in $library_names; do
+           func_append rmfiles " $odir/$n"
+         done
+         test -n "$old_library" && func_append rmfiles " $odir/$old_library"
+
+         case "$opt_mode" in
+         clean)
+           case " $library_names " in
+           *" $dlname "*) ;;
+           *) test -n "$dlname" && func_append rmfiles " $odir/$dlname" ;;
+           esac
+           test -n "$libdir" && func_append rmfiles " $odir/$name $odir/${name}i"
+           ;;
+         uninstall)
+           if test -n "$library_names"; then
+             # Do each command in the postuninstall commands.
+             func_execute_cmds "$postuninstall_cmds" 'test "$rmforce" = yes || exit_status=1'
+           fi
+
+           if test -n "$old_library"; then
+             # Do each command in the old_postuninstall commands.
+             func_execute_cmds "$old_postuninstall_cmds" 'test "$rmforce" = yes || exit_status=1'
+           fi
+           # FIXME: should reinstall the best remaining shared library.
+           ;;
+         esac
+       fi
+       ;;
+
+      *.lo)
+       # Possibly a libtool object, so verify it.
+       if func_lalib_p "$file"; then
+
+         # Read the .lo file
+         func_source $dir/$name
+
+         # Add PIC object to the list of files to remove.
+         if test -n "$pic_object" &&
+            test "$pic_object" != none; then
+           func_append rmfiles " $dir/$pic_object"
+         fi
+
+         # Add non-PIC object to the list of files to remove.
+         if test -n "$non_pic_object" &&
+            test "$non_pic_object" != none; then
+           func_append rmfiles " $dir/$non_pic_object"
+         fi
+       fi
+       ;;
+
+      *)
+       if test "$opt_mode" = clean ; then
+         noexename=$name
+         case $file in
+         *.exe)
+           func_stripname '' '.exe' "$file"
+           file=$func_stripname_result
+           func_stripname '' '.exe' "$name"
+           noexename=$func_stripname_result
+           # $file with .exe has already been added to rmfiles,
+           # add $file without .exe
+           func_append rmfiles " $file"
+           ;;
+         esac
+         # Do a test to see if this is a libtool program.
+         if func_ltwrapper_p "$file"; then
+           if func_ltwrapper_executable_p "$file"; then
+             func_ltwrapper_scriptname "$file"
+             relink_command=
+             func_source $func_ltwrapper_scriptname_result
+             func_append rmfiles " $func_ltwrapper_scriptname_result"
+           else
+             relink_command=
+             func_source $dir/$noexename
+           fi
+
+           # note $name still contains .exe if it was in $file originally
+           # as does the version of $file that was added into $rmfiles
+           func_append rmfiles " $odir/$name $odir/${name}S.${objext}"
+           if test "$fast_install" = yes && test -n "$relink_command"; then
+             func_append rmfiles " $odir/lt-$name"
+           fi
+           if test "X$noexename" != "X$name" ; then
+             func_append rmfiles " $odir/lt-${noexename}.c"
+           fi
+         fi
+       fi
+       ;;
+      esac
+      func_show_eval "$RM $rmfiles" 'exit_status=1'
+    done
+
+    # Try to remove the ${objdir}s in the directories where we deleted files
+    for dir in $rmdirs; do
+      if test -d "$dir"; then
+       func_show_eval "rmdir $dir >/dev/null 2>&1"
+      fi
+    done
+
+    exit $exit_status
+}
+
+{ test "$opt_mode" = uninstall || test "$opt_mode" = clean; } &&
+    func_mode_uninstall ${1+"$@"}
+
+test -z "$opt_mode" && {
+  help="$generic_help"
+  func_fatal_help "you must specify a MODE"
+}
+
+test -z "$exec_cmd" && \
+  func_fatal_help "invalid operation mode \`$opt_mode'"
+
+if test -n "$exec_cmd"; then
+  eval exec "$exec_cmd"
+  exit $EXIT_FAILURE
+fi
+
+exit $exit_status
+
+
+# The TAGs below are defined such that we never get into a situation
+# in which we disable both kinds of libraries.  Given conflicting
+# choices, we go for a static library, that is the most portable,
+# since we can't tell whether shared libraries were disabled because
+# the user asked for that or because the platform doesn't support
+# them.  This is particularly important on AIX, because we don't
+# support having both static and shared libraries enabled at the same
+# time on that platform, so we default to a shared-only configuration.
+# If a disable-shared tag is given, we'll fallback to a static-only
+# configuration.  But we'll never go from static-only to shared-only.
+
+# ### BEGIN LIBTOOL TAG CONFIG: disable-shared
+build_libtool_libs=no
+build_old_libs=yes
+# ### END LIBTOOL TAG CONFIG: disable-shared
+
+# ### BEGIN LIBTOOL TAG CONFIG: disable-static
+build_old_libs=`case $build_libtool_libs in yes) echo no;; *) echo yes;; esac`
+# ### END LIBTOOL TAG CONFIG: disable-static
+
+# Local Variables:
+# mode:shell-script
+# sh-indentation:2
+# End:
+# vi:sw=2
+
diff --git a/lib/rxspencer/main.c b/lib/rxspencer/main.c
new file mode 100644 (file)
index 0000000..0221e77
--- /dev/null
@@ -0,0 +1,510 @@
+#include <stdio.h>
+#include <string.h>
+#include <sys/types.h>
+#include <regex.h>
+#include <assert.h>
+
+#include "main.ih"
+
+char *progname;
+int debug = 0;
+int line = 0;
+int status = 0;
+
+int copts = REG_EXTENDED;
+int eopts = 0;
+regoff_t startoff = 0;
+regoff_t endoff = 0;
+
+
+extern int split();
+extern void regprint();
+
+/*
+ - main - do the simple case, hand off to regress() for regression
+ */
+main(argc, argv)
+int argc;
+char *argv[];
+{
+       regex_t re;
+#      define  NS      10
+       regmatch_t subs[NS];
+       char erbuf[100];
+       int err;
+       size_t len;
+       int c;
+       int errflg = 0;
+       register int i;
+       extern int optind;
+       extern char *optarg;
+
+       progname = argv[0];
+
+       while ((c = getopt(argc, argv, "c:e:S:E:x")) != EOF)
+               switch (c) {
+               case 'c':       /* compile options */
+                       copts = options('c', optarg);
+                       break;
+               case 'e':       /* execute options */
+                       eopts = options('e', optarg);
+                       break;
+               case 'S':       /* start offset */
+                       startoff = (regoff_t)atoi(optarg);
+                       break;
+               case 'E':       /* end offset */
+                       endoff = (regoff_t)atoi(optarg);
+                       break;
+               case 'x':       /* Debugging. */
+                       debug++;
+                       break;
+               case '?':
+               default:
+                       errflg++;
+                       break;
+               }
+       if (errflg) {
+               fprintf(stderr, "usage: %s ", progname);
+               fprintf(stderr, "[-c copt][-C][-d] [re]\n");
+               exit(2);
+       }
+
+       if (optind >= argc) {
+               regress(stdin);
+               exit(status);
+       }
+
+       err = regcomp(&re, argv[optind++], copts);
+       if (err) {
+               len = regerror(err, &re, erbuf, sizeof(erbuf));
+               fprintf(stderr, "error %s, %d/%d `%s'\n",
+                       eprint(err), len, sizeof(erbuf), erbuf);
+               exit(status);
+       }
+       regprint(&re, stdout);  
+
+       if (optind >= argc) {
+               regfree(&re);
+               exit(status);
+       }
+
+       if (eopts&REG_STARTEND) {
+               subs[0].rm_so = startoff;
+               subs[0].rm_eo = strlen(argv[optind]) - endoff;
+       }
+       err = regexec(&re, argv[optind], (size_t)NS, subs, eopts);
+       if (err) {
+               len = regerror(err, &re, erbuf, sizeof(erbuf));
+               fprintf(stderr, "error %s, %d/%d `%s'\n",
+                       eprint(err), len, sizeof(erbuf), erbuf);
+               exit(status);
+       }
+       if (!(copts&REG_NOSUB)) {
+               len = (int)(subs[0].rm_eo - subs[0].rm_so);
+               if (subs[0].rm_so != -1) {
+                       if (len != 0)
+                               printf("match `%.*s'\n", len,
+                                       argv[optind] + subs[0].rm_so);
+                       else
+                               printf("match `'@%.1s\n",
+                                       argv[optind] + subs[0].rm_so);
+               }
+               for (i = 1; i < NS; i++)
+                       if (subs[i].rm_so != -1)
+                               printf("(%d) `%.*s'\n", i,
+                                       (int)(subs[i].rm_eo - subs[i].rm_so),
+                                       argv[optind] + subs[i].rm_so);
+       }
+       exit(status);
+}
+
+/*
+ - regress - main loop of regression test
+ == void regress(FILE *in);
+ */
+void
+regress(in)
+FILE *in;
+{
+       char inbuf[1000];
+#      define  MAXF    10
+       char *f[MAXF];
+       int nf;
+       int i;
+       char erbuf[100];
+       size_t ne;
+       char *badpat = "invalid regular expression";
+#      define  SHORT   10
+       char *bpname = "REG_BADPAT";
+       regex_t re;
+
+       while (fgets(inbuf, sizeof(inbuf), in) != NULL) {
+               line++;
+               if (inbuf[0] == '#' || inbuf[0] == '\n')
+                       continue;                       /* NOTE CONTINUE */
+               inbuf[strlen(inbuf)-1] = '\0';  /* get rid of stupid \n */
+               if (debug)
+                       fprintf(stdout, "%d:\n", line);
+               nf = split(inbuf, f, MAXF, "\t\t");
+               if (nf < 3) {
+                       fprintf(stderr, "bad input, line %d\n", line);
+                       exit(1);
+               }
+               for (i = 0; i < nf; i++)
+                       if (strcmp(f[i], "\"\"") == 0)
+                               f[i] = "";
+               if (nf <= 3)
+                       f[3] = NULL;
+               if (nf <= 4)
+                       f[4] = NULL;
+               try(f[0], f[1], f[2], f[3], f[4], options('c', f[1]));
+               if (opt('&', f[1]))     /* try with either type of RE */
+                       try(f[0], f[1], f[2], f[3], f[4],
+                                       options('c', f[1]) &~ REG_EXTENDED);
+       }
+
+       ne = regerror(REG_BADPAT, (regex_t *)NULL, erbuf, sizeof(erbuf));
+       if (strcmp(erbuf, badpat) != 0 || ne != strlen(badpat)+1) {
+               fprintf(stderr, "end: regerror() test gave `%s' not `%s'\n",
+                                                       erbuf, badpat);
+               status = 1;
+       }
+       ne = regerror(REG_BADPAT, (regex_t *)NULL, erbuf, (size_t)SHORT);
+       if (strncmp(erbuf, badpat, SHORT-1) != 0 || erbuf[SHORT-1] != '\0' ||
+                                               ne != strlen(badpat)+1) {
+               fprintf(stderr, "end: regerror() short test gave `%s' not `%.*s'\n",
+                                               erbuf, SHORT-1, badpat);
+               status = 1;
+       }
+       ne = regerror(REG_ITOA|REG_BADPAT, (regex_t *)NULL, erbuf, sizeof(erbuf));
+       if (strcmp(erbuf, bpname) != 0 || ne != strlen(bpname)+1) {
+               fprintf(stderr, "end: regerror() ITOA test gave `%s' not `%s'\n",
+                                               erbuf, bpname);
+               status = 1;
+       }
+       re.re_endp = bpname;
+       ne = regerror(REG_ATOI, &re, erbuf, sizeof(erbuf));
+       if (atoi(erbuf) != (int)REG_BADPAT) {
+               fprintf(stderr, "end: regerror() ATOI test gave `%s' not `%ld'\n",
+                                               erbuf, (long)REG_BADPAT);
+               status = 1;
+       } else if (ne != strlen(erbuf)+1) {
+               fprintf(stderr, "end: regerror() ATOI test len(`%s') = %ld\n",
+                                               erbuf, (long)REG_BADPAT);
+               status = 1;
+       }
+}
+
+/*
+ - try - try it, and report on problems
+ == void try(char *f0, char *f1, char *f2, char *f3, char *f4, int opts);
+ */
+void
+try(f0, f1, f2, f3, f4, opts)
+char *f0;
+char *f1;
+char *f2;
+char *f3;
+char *f4;
+int opts;                      /* may not match f1 */
+{
+       regex_t re;
+#      define  NSUBS   10
+       regmatch_t subs[NSUBS];
+#      define  NSHOULD 15
+       char *should[NSHOULD];
+       int nshould;
+       char erbuf[100];
+       int err;
+       int len;
+       char *type = (opts & REG_EXTENDED) ? "ERE" : "BRE";
+       register int i;
+       char *grump;
+       char f0copy[1000];
+       char f2copy[1000];
+
+       strcpy(f0copy, f0);
+       re.re_endp = (opts&REG_PEND) ? f0copy + strlen(f0copy) : NULL;
+       fixstr(f0copy);
+       err = regcomp(&re, f0copy, opts);
+       if (err != 0 && (!opt('C', f1) || err != efind(f2))) {
+               /* unexpected error or wrong error */
+               len = regerror(err, &re, erbuf, sizeof(erbuf));
+               fprintf(stderr, "%d: %s error %s, %d/%d `%s'\n",
+                                       line, type, eprint(err), len,
+                                       sizeof(erbuf), erbuf);
+               status = 1;
+       } else if (err == 0 && opt('C', f1)) {
+               /* unexpected success */
+               fprintf(stderr, "%d: %s should have given REG_%s\n",
+                                               line, type, f2);
+               status = 1;
+               err = 1;        /* so we won't try regexec */
+       }
+
+       if (err != 0) {
+               regfree(&re);
+               return;
+       }
+
+       strcpy(f2copy, f2);
+       fixstr(f2copy);
+
+       if (options('e', f1)&REG_STARTEND) {
+               if (strchr(f2, '(') == NULL || strchr(f2, ')') == NULL)
+                       fprintf(stderr, "%d: bad STARTEND syntax\n", line);
+               subs[0].rm_so = strchr(f2, '(') - f2 + 1;
+               subs[0].rm_eo = strchr(f2, ')') - f2;
+       }
+       err = regexec(&re, f2copy, NSUBS, subs, options('e', f1));
+
+       if (err != 0 && (f3 != NULL || err != REG_NOMATCH)) {
+               /* unexpected error or wrong error */
+               len = regerror(err, &re, erbuf, sizeof(erbuf));
+               fprintf(stderr, "%d: %s exec error %s, %d/%d `%s'\n",
+                                       line, type, eprint(err), len,
+                                       sizeof(erbuf), erbuf);
+               status = 1;
+       } else if (err != 0) {
+               /* nothing more to check */
+       } else if (f3 == NULL) {
+               /* unexpected success */
+               fprintf(stderr, "%d: %s exec should have failed\n",
+                                               line, type);
+               status = 1;
+               err = 1;                /* just on principle */
+       } else if (opts&REG_NOSUB) {
+               /* nothing more to check */
+       } else if ((grump = check(f2, subs[0], f3)) != NULL) {
+               fprintf(stderr, "%d: %s %s\n", line, type, grump);
+               status = 1;
+               err = 1;
+       }
+
+       if (err != 0 || f4 == NULL) {
+               regfree(&re);
+               return;
+       }
+
+       for (i = 1; i < NSHOULD; i++)
+               should[i] = NULL;
+       nshould = split(f4, should+1, NSHOULD-1, ",");
+       if (nshould == 0) {
+               nshould = 1;
+               should[1] = "";
+       }
+       for (i = 1; i < NSUBS; i++) {
+               grump = check(f2, subs[i], should[i]);
+               if (grump != NULL) {
+                       fprintf(stderr, "%d: %s $%d %s\n", line,
+                                                       type, i, grump);
+                       status = 1;
+                       err = 1;
+               }
+       }
+
+       regfree(&re);
+}
+
+/*
+ - options - pick options out of a regression-test string
+ == int options(int type, char *s);
+ */
+int
+options(type, s)
+int type;                      /* 'c' compile, 'e' exec */
+char *s;
+{
+       register char *p;
+       register int o = (type == 'c') ? copts : eopts;
+       register char *legal = (type == 'c') ? "bisnmp" : "^$#tl";
+
+       for (p = s; *p != '\0'; p++)
+               if (strchr(legal, *p) != NULL)
+                       switch (*p) {
+                       case 'b':
+                               o &= ~REG_EXTENDED;
+                               break;
+                       case 'i':
+                               o |= REG_ICASE;
+                               break;
+                       case 's':
+                               o |= REG_NOSUB;
+                               break;
+                       case 'n':
+                               o |= REG_NEWLINE;
+                               break;
+                       case 'm':
+                               o &= ~REG_EXTENDED;
+                               o |= REG_NOSPEC;
+                               break;
+                       case 'p':
+                               o |= REG_PEND;
+                               break;
+                       case '^':
+                               o |= REG_NOTBOL;
+                               break;
+                       case '$':
+                               o |= REG_NOTEOL;
+                               break;
+                       case '#':
+                               o |= REG_STARTEND;
+                               break;
+                       case 't':       /* trace */
+                               o |= REG_TRACE;
+                               break;
+                       case 'l':       /* force long representation */
+                               o |= REG_LARGE;
+                               break;
+                       case 'r':       /* force backref use */
+                               o |= REG_BACKR;
+                               break;
+                       }
+       return(o);
+}
+
+/*
+ - opt - is a particular option in a regression string?
+ == int opt(int c, char *s);
+ */
+int                            /* predicate */
+opt(c, s)
+int c;
+char *s;
+{
+       return(strchr(s, c) != NULL);
+}
+
+/*
+ - fixstr - transform magic characters in strings
+ == void fixstr(register char *p);
+ */
+void
+fixstr(p)
+register char *p;
+{
+       if (p == NULL)
+               return;
+
+       for (; *p != '\0'; p++)
+               if (*p == 'N')
+                       *p = '\n';
+               else if (*p == 'T')
+                       *p = '\t';
+               else if (*p == 'S')
+                       *p = ' ';
+               else if (*p == 'Z')
+                       *p = '\0';
+}
+
+/*
+ - check - check a substring match
+ == char *check(char *str, regmatch_t sub, char *should);
+ */
+char *                         /* NULL or complaint */
+check(str, sub, should)
+char *str;
+regmatch_t sub;
+char *should;
+{
+       register int len;
+       register int shlen;
+       register char *p;
+       static char grump[500];
+       register char *at = NULL;
+
+       if (should != NULL && strcmp(should, "-") == 0)
+               should = NULL;
+       if (should != NULL && should[0] == '@') {
+               at = should + 1;
+               should = "";
+       }
+
+       /* check rm_so and rm_eo for consistency */
+       if (sub.rm_so > sub.rm_eo || (sub.rm_so == -1 && sub.rm_eo != -1) ||
+                               (sub.rm_so != -1 && sub.rm_eo == -1) ||
+                               (sub.rm_so != -1 && sub.rm_so < 0) ||
+                               (sub.rm_eo != -1 && sub.rm_eo < 0) ) {
+               sprintf(grump, "start %ld end %ld", (long)sub.rm_so,
+                                                       (long)sub.rm_eo);
+               return(grump);
+       }
+
+       /* check for no match */
+       if (sub.rm_so == -1 && should == NULL)
+               return(NULL);
+       if (sub.rm_so == -1)
+               return("did not match");
+
+       /* check for in range */
+       if (sub.rm_eo > strlen(str)) {
+               sprintf(grump, "start %ld end %ld, past end of string",
+                                       (long)sub.rm_so, (long)sub.rm_eo);
+               return(grump);
+       }
+
+       len = (int)(sub.rm_eo - sub.rm_so);
+       shlen = (int)strlen(should);
+       p = str + sub.rm_so;
+
+       /* check for not supposed to match */
+       if (should == NULL) {
+               sprintf(grump, "matched `%.*s'", len, p);
+               return(grump);
+       }
+
+       /* check for wrong match */
+       if (len != shlen || strncmp(p, should, (size_t)shlen) != 0) {
+               sprintf(grump, "matched `%.*s' instead", len, p);
+               return(grump);
+       }
+       if (shlen > 0)
+               return(NULL);
+
+       /* check null match in right place */
+       if (at == NULL)
+               return(NULL);
+       shlen = strlen(at);
+       if (shlen == 0)
+               shlen = 1;      /* force check for end-of-string */
+       if (strncmp(p, at, shlen) != 0) {
+               sprintf(grump, "matched null at `%.20s'", p);
+               return(grump);
+       }
+       return(NULL);
+}
+
+/*
+ - eprint - convert error number to name
+ == static char *eprint(int err);
+ */
+static char *
+eprint(err)
+int err;
+{
+       static char epbuf[100];
+       size_t len;
+
+       len = regerror(REG_ITOA|err, (regex_t *)NULL, epbuf, sizeof(epbuf));
+       assert(len <= sizeof(epbuf));
+       return(epbuf);
+}
+
+/*
+ - efind - convert error name to number
+ == static int efind(char *name);
+ */
+static int
+efind(name)
+char *name;
+{
+       static char efbuf[100];
+       size_t n;
+       regex_t re;
+
+       sprintf(efbuf, "REG_%s", name);
+       assert(strlen(efbuf) < sizeof(efbuf));
+       re.re_endp = efbuf;
+       (void) regerror(REG_ATOI, &re, efbuf, sizeof(efbuf));
+       return(atoi(efbuf));
+}
diff --git a/lib/rxspencer/missing b/lib/rxspencer/missing
new file mode 100755 (executable)
index 0000000..f62bbae
--- /dev/null
@@ -0,0 +1,215 @@
+#! /bin/sh
+# Common wrapper for a few potentially missing GNU programs.
+
+scriptversion=2013-10-28.13; # UTC
+
+# Copyright (C) 1996-2014 Free Software Foundation, Inc.
+# Originally written by Fran,cois Pinard <pinard@iro.umontreal.ca>, 1996.
+
+# This program is free software; you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation; either version 2, or (at your option)
+# any later version.
+
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+
+# You should have received a copy of the GNU General Public License
+# along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+# As a special exception to the GNU General Public License, if you
+# distribute this file as part of a program that contains a
+# configuration script generated by Autoconf, you may include it under
+# the same distribution terms that you use for the rest of that program.
+
+if test $# -eq 0; then
+  echo 1>&2 "Try '$0 --help' for more information"
+  exit 1
+fi
+
+case $1 in
+
+  --is-lightweight)
+    # Used by our autoconf macros to check whether the available missing
+    # script is modern enough.
+    exit 0
+    ;;
+
+  --run)
+    # Back-compat with the calling convention used by older automake.
+    shift
+    ;;
+
+  -h|--h|--he|--hel|--help)
+    echo "\
+$0 [OPTION]... PROGRAM [ARGUMENT]...
+
+Run 'PROGRAM [ARGUMENT]...', returning a proper advice when this fails due
+to PROGRAM being missing or too old.
+
+Options:
+  -h, --help      display this help and exit
+  -v, --version   output version information and exit
+
+Supported PROGRAM values:
+  aclocal   autoconf  autoheader   autom4te  automake  makeinfo
+  bison     yacc      flex         lex       help2man
+
+Version suffixes to PROGRAM as well as the prefixes 'gnu-', 'gnu', and
+'g' are ignored when checking the name.
+
+Send bug reports to <bug-automake@gnu.org>."
+    exit $?
+    ;;
+
+  -v|--v|--ve|--ver|--vers|--versi|--versio|--version)
+    echo "missing $scriptversion (GNU Automake)"
+    exit $?
+    ;;
+
+  -*)
+    echo 1>&2 "$0: unknown '$1' option"
+    echo 1>&2 "Try '$0 --help' for more information"
+    exit 1
+    ;;
+
+esac
+
+# Run the given program, remember its exit status.
+"$@"; st=$?
+
+# If it succeeded, we are done.
+test $st -eq 0 && exit 0
+
+# Also exit now if we it failed (or wasn't found), and '--version' was
+# passed; such an option is passed most likely to detect whether the
+# program is present and works.
+case $2 in --version|--help) exit $st;; esac
+
+# Exit code 63 means version mismatch.  This often happens when the user
+# tries to use an ancient version of a tool on a file that requires a
+# minimum version.
+if test $st -eq 63; then
+  msg="probably too old"
+elif test $st -eq 127; then
+  # Program was missing.
+  msg="missing on your system"
+else
+  # Program was found and executed, but failed.  Give up.
+  exit $st
+fi
+
+perl_URL=http://www.perl.org/
+flex_URL=http://flex.sourceforge.net/
+gnu_software_URL=http://www.gnu.org/software
+
+program_details ()
+{
+  case $1 in
+    aclocal|automake)
+      echo "The '$1' program is part of the GNU Automake package:"
+      echo "<$gnu_software_URL/automake>"
+      echo "It also requires GNU Autoconf, GNU m4 and Perl in order to run:"
+      echo "<$gnu_software_URL/autoconf>"
+      echo "<$gnu_software_URL/m4/>"
+      echo "<$perl_URL>"
+      ;;
+    autoconf|autom4te|autoheader)
+      echo "The '$1' program is part of the GNU Autoconf package:"
+      echo "<$gnu_software_URL/autoconf/>"
+      echo "It also requires GNU m4 and Perl in order to run:"
+      echo "<$gnu_software_URL/m4/>"
+      echo "<$perl_URL>"
+      ;;
+  esac
+}
+
+give_advice ()
+{
+  # Normalize program name to check for.
+  normalized_program=`echo "$1" | sed '
+    s/^gnu-//; t
+    s/^gnu//; t
+    s/^g//; t'`
+
+  printf '%s\n' "'$1' is $msg."
+
+  configure_deps="'configure.ac' or m4 files included by 'configure.ac'"
+  case $normalized_program in
+    autoconf*)
+      echo "You should only need it if you modified 'configure.ac',"
+      echo "or m4 files included by it."
+      program_details 'autoconf'
+      ;;
+    autoheader*)
+      echo "You should only need it if you modified 'acconfig.h' or"
+      echo "$configure_deps."
+      program_details 'autoheader'
+      ;;
+    automake*)
+      echo "You should only need it if you modified 'Makefile.am' or"
+      echo "$configure_deps."
+      program_details 'automake'
+      ;;
+    aclocal*)
+      echo "You should only need it if you modified 'acinclude.m4' or"
+      echo "$configure_deps."
+      program_details 'aclocal'
+      ;;
+   autom4te*)
+      echo "You might have modified some maintainer files that require"
+      echo "the 'autom4te' program to be rebuilt."
+      program_details 'autom4te'
+      ;;
+    bison*|yacc*)
+      echo "You should only need it if you modified a '.y' file."
+      echo "You may want to install the GNU Bison package:"
+      echo "<$gnu_software_URL/bison/>"
+      ;;
+    lex*|flex*)
+      echo "You should only need it if you modified a '.l' file."
+      echo "You may want to install the Fast Lexical Analyzer package:"
+      echo "<$flex_URL>"
+      ;;
+    help2man*)
+      echo "You should only need it if you modified a dependency" \
+           "of a man page."
+      echo "You may want to install the GNU Help2man package:"
+      echo "<$gnu_software_URL/help2man/>"
+    ;;
+    makeinfo*)
+      echo "You should only need it if you modified a '.texi' file, or"
+      echo "any other file indirectly affecting the aspect of the manual."
+      echo "You might want to install the Texinfo package:"
+      echo "<$gnu_software_URL/texinfo/>"
+      echo "The spurious makeinfo call might also be the consequence of"
+      echo "using a buggy 'make' (AIX, DU, IRIX), in which case you might"
+      echo "want to install GNU make:"
+      echo "<$gnu_software_URL/make/>"
+      ;;
+    *)
+      echo "You might have modified some files without having the proper"
+      echo "tools for further handling them.  Check the 'README' file, it"
+      echo "often tells you about the needed prerequisites for installing"
+      echo "this package.  You may also peek at any GNU archive site, in"
+      echo "case some other package contains this missing '$1' program."
+      ;;
+  esac
+}
+
+give_advice "$1" | sed -e '1s/^/WARNING: /' \
+                       -e '2,$s/^/         /' >&2
+
+# Propagate the correct exit status (expected to be 127 for a program
+# not found, 63 for a program that failed due to version mismatch).
+exit $st
+
+# Local variables:
+# eval: (add-hook 'write-file-hooks 'time-stamp)
+# time-stamp-start: "scriptversion="
+# time-stamp-format: "%:y-%02m-%02d.%02H"
+# time-stamp-time-zone: "UTC"
+# time-stamp-end: "; # UTC"
+# End:
diff --git a/lib/rxspencer/mkh b/lib/rxspencer/mkh
new file mode 100644 (file)
index 0000000..5939d34
--- /dev/null
@@ -0,0 +1,75 @@
+#! /bin/sh
+# mkh - pull headers out of C source
+
+# egrep pattern to pick out marked lines
+egrep='^ =([   ]|$)'
+
+# Sed program to process marked lines into lines for the header file.
+# The markers have already been removed.  Two things are done here:  removal
+# of backslashed newlines, and some fudging of comments.  The first is done
+# because -o needs to have prototypes on one line to strip them down.
+# Getting comments into the output is tricky; we turn C++-style // comments
+# into /* */ comments, after altering any existing */'s to avoid trouble.
+peel=' /\\$/N
+       /\\\n[  ]*/s///g
+       /\/\//s;\*/;* /;g
+       /\/\//s;//\(.*\);/*\1 */;'
+
+for a
+do
+       case "$a" in
+       -o)     # old (pre-function-prototype) compiler
+               # add code to comment out argument lists
+               peel="$peel
+                       "'/^\([^#\/][^\/]*[a-zA-Z0-9_)]\)(\(.*\))/s;;\1(/*\2*/);'
+               shift
+               ;;
+       -b)     # funny Berkeley __P macro
+               peel="$peel
+                       "'/^\([^#\/][^\/]*[a-zA-Z0-9_)]\)(\(.*\))/s;;\1 __P((\2));'
+               shift
+               ;;
+       -s)     # compiler doesn't like `static foo();'
+               # add code to get rid of the `static'
+               peel="$peel
+                       "'/^static[     ][^\/]*[a-zA-Z0-9_)](.*)/s;static.;;'
+               shift
+               ;;
+       -p)     # private declarations
+               egrep='^ ==([   ]|$)'
+               shift
+               ;;
+       -i)     # wrap in #ifndef, argument is name
+               ifndef="$2"
+               shift ; shift
+               ;;
+       *)      break
+               ;;
+       esac
+done
+
+if test " $ifndef" != " "
+then
+       echo "#ifndef $ifndef"
+       echo "#define   $ifndef /* never again */"
+fi
+echo "/* ========= begin header generated by $0 ========= */"
+echo '#ifdef __cplusplus'
+echo 'extern "C" {'
+echo '#endif'
+for f
+do
+       echo
+       echo "/* === $f === */"
+       egrep "$egrep" $f | sed 's/^ ==*[       ]//;s/^ ==*$//' | sed "$peel"
+       echo
+done
+echo '#ifdef __cplusplus'
+echo '}'
+echo '#endif'
+echo "/* ========= end header generated by $0 ========= */"
+if test " $ifndef" != " "
+then
+       echo "#endif"
+fi
+exit 0
diff --git a/lib/rxspencer/mkinstalldirs b/lib/rxspencer/mkinstalldirs
new file mode 100755 (executable)
index 0000000..55d537f
--- /dev/null
@@ -0,0 +1,162 @@
+#! /bin/sh
+# mkinstalldirs --- make directory hierarchy
+
+scriptversion=2009-04-28.21; # UTC
+
+# Original author: Noah Friedman <friedman@prep.ai.mit.edu>
+# Created: 1993-05-16
+# Public domain.
+#
+# This file is maintained in Automake, please report
+# bugs to <bug-automake@gnu.org> or send patches to
+# <automake-patches@gnu.org>.
+
+nl='
+'
+IFS=" ""       $nl"
+errstatus=0
+dirmode=
+
+usage="\
+Usage: mkinstalldirs [-h] [--help] [--version] [-m MODE] DIR ...
+
+Create each directory DIR (with mode MODE, if specified), including all
+leading file name components.
+
+Report bugs to <bug-automake@gnu.org>."
+
+# process command line arguments
+while test $# -gt 0 ; do
+  case $1 in
+    -h | --help | --h*)         # -h for help
+      echo "$usage"
+      exit $?
+      ;;
+    -m)                         # -m PERM arg
+      shift
+      test $# -eq 0 && { echo "$usage" 1>&2; exit 1; }
+      dirmode=$1
+      shift
+      ;;
+    --version)
+      echo "$0 $scriptversion"
+      exit $?
+      ;;
+    --)                         # stop option processing
+      shift
+      break
+      ;;
+    -*)                         # unknown option
+      echo "$usage" 1>&2
+      exit 1
+      ;;
+    *)                          # first non-opt arg
+      break
+      ;;
+  esac
+done
+
+for file
+do
+  if test -d "$file"; then
+    shift
+  else
+    break
+  fi
+done
+
+case $# in
+  0) exit 0 ;;
+esac
+
+# Solaris 8's mkdir -p isn't thread-safe.  If you mkdir -p a/b and
+# mkdir -p a/c at the same time, both will detect that a is missing,
+# one will create a, then the other will try to create a and die with
+# a "File exists" error.  This is a problem when calling mkinstalldirs
+# from a parallel make.  We use --version in the probe to restrict
+# ourselves to GNU mkdir, which is thread-safe.
+case $dirmode in
+  '')
+    if mkdir -p --version . >/dev/null 2>&1 && test ! -d ./--version; then
+      echo "mkdir -p -- $*"
+      exec mkdir -p -- "$@"
+    else
+      # On NextStep and OpenStep, the 'mkdir' command does not
+      # recognize any option.  It will interpret all options as
+      # directories to create, and then abort because '.' already
+      # exists.
+      test -d ./-p && rmdir ./-p
+      test -d ./--version && rmdir ./--version
+    fi
+    ;;
+  *)
+    if mkdir -m "$dirmode" -p --version . >/dev/null 2>&1 &&
+       test ! -d ./--version; then
+      echo "mkdir -m $dirmode -p -- $*"
+      exec mkdir -m "$dirmode" -p -- "$@"
+    else
+      # Clean up after NextStep and OpenStep mkdir.
+      for d in ./-m ./-p ./--version "./$dirmode";
+      do
+        test -d $d && rmdir $d
+      done
+    fi
+    ;;
+esac
+
+for file
+do
+  case $file in
+    /*) pathcomp=/ ;;
+    *)  pathcomp= ;;
+  esac
+  oIFS=$IFS
+  IFS=/
+  set fnord $file
+  shift
+  IFS=$oIFS
+
+  for d
+  do
+    test "x$d" = x && continue
+
+    pathcomp=$pathcomp$d
+    case $pathcomp in
+      -*) pathcomp=./$pathcomp ;;
+    esac
+
+    if test ! -d "$pathcomp"; then
+      echo "mkdir $pathcomp"
+
+      mkdir "$pathcomp" || lasterr=$?
+
+      if test ! -d "$pathcomp"; then
+       errstatus=$lasterr
+      else
+       if test ! -z "$dirmode"; then
+         echo "chmod $dirmode $pathcomp"
+         lasterr=
+         chmod "$dirmode" "$pathcomp" || lasterr=$?
+
+         if test ! -z "$lasterr"; then
+           errstatus=$lasterr
+         fi
+       fi
+      fi
+    fi
+
+    pathcomp=$pathcomp/
+  done
+done
+
+exit $errstatus
+
+# Local Variables:
+# mode: shell-script
+# sh-indentation: 2
+# eval: (add-hook 'write-file-hooks 'time-stamp)
+# time-stamp-start: "scriptversion="
+# time-stamp-format: "%:y-%02m-%02d.%02H"
+# time-stamp-time-zone: "UTC"
+# time-stamp-end: "; # UTC"
+# End:
diff --git a/lib/rxspencer/regcomp.c b/lib/rxspencer/regcomp.c
new file mode 100644 (file)
index 0000000..12da9b1
--- /dev/null
@@ -0,0 +1,1611 @@
+#include <sys/types.h>
+#include <stdio.h>
+#include <string.h>
+#include <ctype.h>
+#include <limits.h>
+#include <stdlib.h>
+#include <regex.h>
+
+#include "utils.h"
+#include "regex2.h"
+
+#include "cclass.h"
+#include "cname.h"
+
+/*
+ * parse structure, passed up and down to avoid global variables and
+ * other clumsinesses
+ */
+struct parse {
+       char *next;             /* next character in RE */
+       char *end;              /* end of string (-> NUL normally) */
+       int error;              /* has an error been seen? */
+       sop *strip;             /* malloced strip */
+       sopno ssize;            /* malloced strip size (allocated) */
+       sopno slen;             /* malloced strip length (used) */
+       int ncsalloc;           /* number of csets allocated */
+       struct re_guts *g;
+#      define  NPAREN  10      /* we need to remember () 1-9 for back refs */
+       sopno pbegin[NPAREN];   /* -> ( ([0] unused) */
+       sopno pend[NPAREN];     /* -> ) ([0] unused) */
+};
+
+#include "regcomp.ih"
+
+static char nuls[10];          /* place to point scanner in event of error */
+
+/*
+ * macros for use with parse structure
+ * BEWARE:  these know that the parse structure is named `p' !!!
+ */
+#define        PEEK()  (*p->next)
+#define        PEEK2() (*(p->next+1))
+#define        MORE()  (p->next < p->end)
+#define        MORE2() (p->next+1 < p->end)
+#define        SEE(c)  (MORE() && PEEK() == (c))
+#define        SEETWO(a, b)    (MORE() && MORE2() && PEEK() == (a) && PEEK2() == (b))
+#define        EAT(c)  ((SEE(c)) ? (NEXT(), 1) : 0)
+#define        EATTWO(a, b)    ((SEETWO(a, b)) ? (NEXT2(), 1) : 0)
+#define        NEXT()  (p->next++)
+#define        NEXT2() (p->next += 2)
+#define        NEXTn(n)        (p->next += (n))
+#define        GETNEXT()       (*p->next++)
+#define        SETERROR(e)     seterr(p, (e))
+#define        REQUIRE(co, e)  ((co) || SETERROR(e))
+#define        MUSTSEE(c, e)   (REQUIRE(MORE() && PEEK() == (c), e))
+#define        MUSTEAT(c, e)   (REQUIRE(MORE() && GETNEXT() == (c), e))
+#define        MUSTNOTSEE(c, e)        (REQUIRE(!MORE() || PEEK() != (c), e))
+#define        EMIT(op, sopnd) doemit(p, (sop)(op), (size_t)(sopnd))
+#define        INSERT(op, pos) doinsert(p, (sop)(op), HERE()-(pos)+1, pos)
+#define        AHEAD(pos)              dofwd(p, pos, HERE()-(pos))
+#define        ASTERN(sop, pos)        EMIT(sop, HERE()-pos)
+#define        HERE()          (p->slen)
+#define        THERE()         (p->slen - 1)
+#define        THERETHERE()    (p->slen - 2)
+#define        DROP(n) (p->slen -= (n))
+
+#ifndef NDEBUG
+static int never = 0;          /* for use in asserts; shuts lint up */
+#else
+#define        never   0               /* some <assert.h>s have bugs too */
+#endif
+
+/*
+ - regcomp - interface for parser and compilation
+ = extern int regcomp(regex_t *, const char *, int);
+ = #define     REG_BASIC       0000
+ = #define     REG_EXTENDED    0001
+ = #define     REG_ICASE       0002
+ = #define     REG_NOSUB       0004
+ = #define     REG_NEWLINE     0010
+ = #define     REG_NOSPEC      0020
+ = #define     REG_PEND        0040
+ = #define     REG_DUMP        0200
+ */
+int                            /* 0 success, otherwise REG_something */
+regcomp(preg, pattern, cflags)
+regex_t *preg;
+const char *pattern;
+int cflags;
+{
+       struct parse pa;
+       register struct re_guts *g;
+       register struct parse *p = &pa;
+       register int i;
+       register size_t len;
+#ifdef REDEBUG
+#      define  GOODFLAGS(f)    (f)
+#else
+#      define  GOODFLAGS(f)    ((f)&~REG_DUMP)
+#endif
+
+       cflags = GOODFLAGS(cflags);
+       if ((cflags&REG_EXTENDED) && (cflags&REG_NOSPEC))
+               return(REG_INVARG);
+
+       if (cflags&REG_PEND) {
+               if (preg->re_endp < pattern)
+                       return(REG_INVARG);
+               len = preg->re_endp - pattern;
+       } else
+               len = strlen((char *)pattern);
+
+       /* do the mallocs early so failure handling is easy */
+       g = (struct re_guts *)malloc(sizeof(struct re_guts) +
+                                                       (NC-1)*sizeof(cat_t));
+       if (g == NULL)
+               return(REG_ESPACE);
+       {
+         /* Patched for CERT Vulnerability Note VU#695940, Feb 2015. */
+         size_t new_ssize = len/(size_t)2*(size_t)3 + (size_t)1; /* ugh */
+         if (new_ssize < len || new_ssize > LONG_MAX / sizeof(sop)) {
+           free((char *) g);
+           return REG_INVARG;
+         }
+         p->ssize = new_ssize;
+       }
+       p->strip = (sop *)malloc(p->ssize * sizeof(sop));
+       p->slen = 0;
+       if (p->strip == NULL) {
+               free((char *)g);
+               return(REG_ESPACE);
+       }
+
+       /* set things up */
+       p->g = g;
+       p->next = (char *)pattern;      /* convenience; we do not modify it */
+       p->end = p->next + len;
+       p->error = 0;
+       p->ncsalloc = 0;
+       for (i = 0; i < NPAREN; i++) {
+               p->pbegin[i] = 0;
+               p->pend[i] = 0;
+       }
+       g->csetsize = NC;
+       g->sets = NULL;
+       g->setbits = NULL;
+       g->ncsets = 0;
+       g->cflags = cflags;
+       g->iflags = 0;
+       g->nbol = 0;
+       g->neol = 0;
+       g->must = NULL;
+       g->mlen = 0;
+       g->nsub = 0;
+       g->ncategories = 1;     /* category 0 is "everything else" */
+       g->categories = &g->catspace[-(CHAR_MIN)];
+       (void) memset((char *)g->catspace, 0, NC*sizeof(cat_t));
+       g->backrefs = 0;
+
+       /* do it */
+       EMIT(OEND, 0);
+       g->firststate = THERE();
+       if (cflags&REG_EXTENDED)
+               p_ere(p, OUT);
+       else if (cflags&REG_NOSPEC)
+               p_str(p);
+       else
+               p_bre(p, OUT, OUT);
+       EMIT(OEND, 0);
+       g->laststate = THERE();
+
+       /* tidy up loose ends and fill things in */
+       categorize(p, g);
+       stripsnug(p, g);
+       findmust(p, g);
+       g->nplus = pluscount(p, g);
+       g->magic = MAGIC2;
+       preg->re_nsub = g->nsub;
+       preg->re_g = g;
+       preg->re_magic = MAGIC1;
+#ifndef REDEBUG
+       /* not debugging, so can't rely on the assert() in regexec() */
+       if (g->iflags&BAD)
+               SETERROR(REG_ASSERT);
+#endif
+
+       /* win or lose, we're done */
+       if (p->error != 0)      /* lose */
+               regfree(preg);
+       return(p->error);
+}
+
+/*
+ - p_ere - ERE parser top level, concatenation and alternation
+ == static void p_ere(register struct parse *p, int stop);
+ */
+static void
+p_ere(p, stop)
+register struct parse *p;
+int stop;                      /* character this ERE should end at */
+{
+       register char c;
+       register sopno prevback;
+       register sopno prevfwd;
+       register sopno conc;
+       register int first = 1;         /* is this the first alternative? */
+
+       for (;;) {
+               /* do a bunch of concatenated expressions */
+               conc = HERE();
+               while (MORE() && (c = PEEK()) != '|' && c != stop)
+                       p_ere_exp(p);
+               REQUIRE(HERE() != conc, REG_EMPTY);     /* require nonempty */
+
+               if (!EAT('|'))
+                       break;          /* NOTE BREAK OUT */
+
+               if (first) {
+                       INSERT(OCH_, conc);     /* offset is wrong */
+                       prevfwd = conc;
+                       prevback = conc;
+                       first = 0;
+               }
+               ASTERN(OOR1, prevback);
+               prevback = THERE();
+               AHEAD(prevfwd);                 /* fix previous offset */
+               prevfwd = HERE();
+               EMIT(OOR2, 0);                  /* offset is very wrong */
+       }
+
+       if (!first) {           /* tail-end fixups */
+               AHEAD(prevfwd);
+               ASTERN(O_CH, prevback);
+       }
+
+       assert(!MORE() || SEE(stop));
+}
+
+/*
+ - p_ere_exp - parse one subERE, an atom possibly followed by a repetition op
+ == static void p_ere_exp(register struct parse *p);
+ */
+static void
+p_ere_exp(p)
+register struct parse *p;
+{
+       register char c;
+       register sopno pos;
+       register int count;
+       register int count2;
+       register sopno subno;
+       int wascaret = 0;
+
+       assert(MORE());         /* caller should have ensured this */
+       c = GETNEXT();
+
+       pos = HERE();
+       switch (c) {
+       case '(':
+               REQUIRE(MORE(), REG_EPAREN);
+               p->g->nsub++;
+               subno = p->g->nsub;
+               if (subno < NPAREN)
+                       p->pbegin[subno] = HERE();
+               EMIT(OLPAREN, subno);
+               if (!SEE(')'))
+                       p_ere(p, ')');
+               if (subno < NPAREN) {
+                       p->pend[subno] = HERE();
+                       assert(p->pend[subno] != 0);
+               }
+               EMIT(ORPAREN, subno);
+               MUSTEAT(')', REG_EPAREN);
+               break;
+#ifndef POSIX_MISTAKE
+       case ')':               /* happens only if no current unmatched ( */
+               /*
+                * You may ask, why the ifndef?  Because I didn't notice
+                * this until slightly too late for 1003.2, and none of the
+                * other 1003.2 regular-expression reviewers noticed it at
+                * all.  So an unmatched ) is legal POSIX, at least until
+                * we can get it fixed.
+                */
+               SETERROR(REG_EPAREN);
+               break;
+#endif
+       case '^':
+               EMIT(OBOL, 0);
+               p->g->iflags |= USEBOL;
+               p->g->nbol++;
+               wascaret = 1;
+               break;
+       case '$':
+               EMIT(OEOL, 0);
+               p->g->iflags |= USEEOL;
+               p->g->neol++;
+               break;
+       case '|':
+               SETERROR(REG_EMPTY);
+               break;
+       case '*':
+       case '+':
+       case '?':
+               SETERROR(REG_BADRPT);
+               break;
+       case '.':
+               if (p->g->cflags&REG_NEWLINE)
+                       nonnewline(p);
+               else
+                       EMIT(OANY, 0);
+               break;
+       case '[':
+               p_bracket(p);
+               break;
+       case '\\':
+               REQUIRE(MORE(), REG_EESCAPE);
+               c = GETNEXT();
+               ordinary(p, c);
+               break;
+       case '{':               /* okay as ordinary except if digit follows */
+               REQUIRE(!MORE() || !isdigit(PEEK()), REG_BADRPT);
+               /* FALLTHROUGH */
+       default:
+               ordinary(p, c);
+               break;
+       }
+
+       if (!MORE())
+               return;
+       c = PEEK();
+       /* we call { a repetition if followed by a digit */
+       if (!( c == '*' || c == '+' || c == '?' ||
+                               (c == '{' && MORE2() && isdigit(PEEK2())) ))
+               return;         /* no repetition, we're done */
+       NEXT();
+
+       REQUIRE(!wascaret, REG_BADRPT);
+       switch (c) {
+       case '*':       /* implemented as +? */
+               /* this case does not require the (y|) trick, noKLUDGE */
+               INSERT(OPLUS_, pos);
+               ASTERN(O_PLUS, pos);
+               INSERT(OQUEST_, pos);
+               ASTERN(O_QUEST, pos);
+               break;
+       case '+':
+               INSERT(OPLUS_, pos);
+               ASTERN(O_PLUS, pos);
+               break;
+       case '?':
+               /* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
+               INSERT(OCH_, pos);              /* offset slightly wrong */
+               ASTERN(OOR1, pos);              /* this one's right */
+               AHEAD(pos);                     /* fix the OCH_ */
+               EMIT(OOR2, 0);                  /* offset very wrong... */
+               AHEAD(THERE());                 /* ...so fix it */
+               ASTERN(O_CH, THERETHERE());
+               break;
+       case '{':
+               count = p_count(p);
+               if (EAT(',')) {
+                       if (isdigit(PEEK())) {
+                               count2 = p_count(p);
+                               REQUIRE(count <= count2, REG_BADBR);
+                       } else          /* single number with comma */
+                               count2 = INFINITY;
+               } else          /* just a single number */
+                       count2 = count;
+               repeat(p, pos, count, count2);
+               if (!EAT('}')) {        /* error heuristics */
+                       while (MORE() && PEEK() != '}')
+                               NEXT();
+                       REQUIRE(MORE(), REG_EBRACE);
+                       SETERROR(REG_BADBR);
+               }
+               break;
+       }
+
+       if (!MORE())
+               return;
+       c = PEEK();
+       if (!( c == '*' || c == '+' || c == '?' ||
+                               (c == '{' && MORE2() && isdigit(PEEK2())) ) )
+               return;
+       SETERROR(REG_BADRPT);
+}
+
+/*
+ - p_str - string (no metacharacters) "parser"
+ == static void p_str(register struct parse *p);
+ */
+static void
+p_str(p)
+register struct parse *p;
+{
+       REQUIRE(MORE(), REG_EMPTY);
+       while (MORE())
+               ordinary(p, GETNEXT());
+}
+
+/*
+ - p_bre - BRE parser top level, anchoring and concatenation
+ == static void p_bre(register struct parse *p, register int end1, \
+ ==    register int end2);
+ * Giving end1 as OUT essentially eliminates the end1/end2 check.
+ *
+ * This implementation is a bit of a kludge, in that a trailing $ is first
+ * taken as an ordinary character and then revised to be an anchor.  The
+ * only undesirable side effect is that '$' gets included as a character
+ * category in such cases.  This is fairly harmless; not worth fixing.
+ * The amount of lookahead needed to avoid this kludge is excessive.
+ */
+static void
+p_bre(p, end1, end2)
+register struct parse *p;
+register int end1;             /* first terminating character */
+register int end2;             /* second terminating character */
+{
+       register sopno start = HERE();
+       register int first = 1;                 /* first subexpression? */
+       register int wasdollar = 0;
+
+       if (EAT('^')) {
+               EMIT(OBOL, 0);
+               p->g->iflags |= USEBOL;
+               p->g->nbol++;
+       }
+       while (MORE() && !SEETWO(end1, end2)) {
+               wasdollar = p_simp_re(p, first);
+               first = 0;
+       }
+       if (wasdollar) {        /* oops, that was a trailing anchor */
+               DROP(1);
+               EMIT(OEOL, 0);
+               p->g->iflags |= USEEOL;
+               p->g->neol++;
+       }
+
+       REQUIRE(HERE() != start, REG_EMPTY);    /* require nonempty */
+}
+
+/*
+ - p_simp_re - parse a simple RE, an atom possibly followed by a repetition
+ == static int p_simp_re(register struct parse *p, int starordinary);
+ */
+static int                     /* was the simple RE an unbackslashed $? */
+p_simp_re(p, starordinary)
+register struct parse *p;
+int starordinary;              /* is a leading * an ordinary character? */
+{
+       register int c;
+       register int count;
+       register int count2;
+       register sopno pos;
+       register int i;
+       register sopno subno;
+#      define  BACKSL  (1<<CHAR_BIT)
+
+       pos = HERE();           /* repetion op, if any, covers from here */
+
+       assert(MORE());         /* caller should have ensured this */
+       c = GETNEXT();
+       if (c == '\\') {
+               REQUIRE(MORE(), REG_EESCAPE);
+               c = BACKSL | (unsigned char)GETNEXT();
+       }
+       switch (c) {
+       case '.':
+               if (p->g->cflags&REG_NEWLINE)
+                       nonnewline(p);
+               else
+                       EMIT(OANY, 0);
+               break;
+       case '[':
+               p_bracket(p);
+               break;
+       case BACKSL|'{':
+               SETERROR(REG_BADRPT);
+               break;
+       case BACKSL|'(':
+               p->g->nsub++;
+               subno = p->g->nsub;
+               if (subno < NPAREN)
+                       p->pbegin[subno] = HERE();
+               EMIT(OLPAREN, subno);
+               /* the MORE here is an error heuristic */
+               if (MORE() && !SEETWO('\\', ')'))
+                       p_bre(p, '\\', ')');
+               if (subno < NPAREN) {
+                       p->pend[subno] = HERE();
+                       assert(p->pend[subno] != 0);
+               }
+               EMIT(ORPAREN, subno);
+               REQUIRE(EATTWO('\\', ')'), REG_EPAREN);
+               break;
+       case BACKSL|')':        /* should not get here -- must be user */
+       case BACKSL|'}':
+               SETERROR(REG_EPAREN);
+               break;
+       case BACKSL|'1':
+       case BACKSL|'2':
+       case BACKSL|'3':
+       case BACKSL|'4':
+       case BACKSL|'5':
+       case BACKSL|'6':
+       case BACKSL|'7':
+       case BACKSL|'8':
+       case BACKSL|'9':
+               i = (c&~BACKSL) - '0';
+               assert(i < NPAREN);
+               if (p->pend[i] != 0) {
+                       assert(i <= p->g->nsub);
+                       EMIT(OBACK_, i);
+                       assert(p->pbegin[i] != 0);
+                       assert(OP(p->strip[p->pbegin[i]]) == OLPAREN);
+                       assert(OP(p->strip[p->pend[i]]) == ORPAREN);
+                       (void) dupl(p, p->pbegin[i]+1, p->pend[i]);
+                       EMIT(O_BACK, i);
+               } else
+                       SETERROR(REG_ESUBREG);
+               p->g->backrefs = 1;
+               break;
+       case '*':
+               REQUIRE(starordinary, REG_BADRPT);
+               /* FALLTHROUGH */
+       default:
+               ordinary(p, (char)c);   /* takes off BACKSL, if any */
+               break;
+       }
+
+       if (EAT('*')) {         /* implemented as +? */
+               /* this case does not require the (y|) trick, noKLUDGE */
+               INSERT(OPLUS_, pos);
+               ASTERN(O_PLUS, pos);
+               INSERT(OQUEST_, pos);
+               ASTERN(O_QUEST, pos);
+       } else if (EATTWO('\\', '{')) {
+               count = p_count(p);
+               if (EAT(',')) {
+                       if (MORE() && isdigit(PEEK())) {
+                               count2 = p_count(p);
+                               REQUIRE(count <= count2, REG_BADBR);
+                       } else          /* single number with comma */
+                               count2 = INFINITY;
+               } else          /* just a single number */
+                       count2 = count;
+               repeat(p, pos, count, count2);
+               if (!EATTWO('\\', '}')) {       /* error heuristics */
+                       while (MORE() && !SEETWO('\\', '}'))
+                               NEXT();
+                       REQUIRE(MORE(), REG_EBRACE);
+                       SETERROR(REG_BADBR);
+               }
+       } else if (c == (unsigned char)'$')     /* $ (but not \$) ends it */
+               return(1);
+
+       return(0);
+}
+
+/*
+ - p_count - parse a repetition count
+ == static int p_count(register struct parse *p);
+ */
+static int                     /* the value */
+p_count(p)
+register struct parse *p;
+{
+       register int count = 0;
+       register int ndigits = 0;
+
+       while (MORE() && isdigit(PEEK()) && count <= DUPMAX) {
+               count = count*10 + (GETNEXT() - '0');
+               ndigits++;
+       }
+
+       REQUIRE(ndigits > 0 && count <= DUPMAX, REG_BADBR);
+       return(count);
+}
+
+/*
+ - p_bracket - parse a bracketed character list
+ == static void p_bracket(register struct parse *p);
+ *
+ * Note a significant property of this code:  if the allocset() did SETERROR,
+ * no set operations are done.
+ */
+static void
+p_bracket(p)
+register struct parse *p;
+{
+       register cset *cs = allocset(p);
+       register int invert = 0;
+
+       /* Dept of Truly Sickening Special-Case Kludges */
+       if (p->next + 5 < p->end && strncmp(p->next, "[:<:]]", 6) == 0) {
+               EMIT(OBOW, 0);
+               NEXTn(6);
+               return;
+       }
+       if (p->next + 5 < p->end && strncmp(p->next, "[:>:]]", 6) == 0) {
+               EMIT(OEOW, 0);
+               NEXTn(6);
+               return;
+       }
+
+       if (EAT('^'))
+               invert++;       /* make note to invert set at end */
+       if (EAT(']'))
+               CHadd(cs, ']');
+       else if (EAT('-'))
+               CHadd(cs, '-');
+       while (MORE() && PEEK() != ']' && !SEETWO('-', ']'))
+               p_b_term(p, cs);
+       if (EAT('-'))
+               CHadd(cs, '-');
+       MUSTEAT(']', REG_EBRACK);
+
+       if (p->error != 0)      /* don't mess things up further */
+               return;
+
+       if (p->g->cflags&REG_ICASE) {
+               register int i;
+               register int ci;
+
+               for (i = p->g->csetsize - 1; i >= 0; i--)
+                       if (CHIN(cs, i) && isalpha(i)) {
+                               ci = othercase(i);
+                               if (ci != i)
+                                       CHadd(cs, ci);
+                       }
+               if (cs->multis != NULL)
+                       mccase(p, cs);
+       }
+       if (invert) {
+               register int i;
+
+               for (i = p->g->csetsize - 1; i >= 0; i--)
+                       if (CHIN(cs, i))
+                               CHsub(cs, i);
+                       else
+                               CHadd(cs, i);
+               if (p->g->cflags&REG_NEWLINE)
+                       CHsub(cs, '\n');
+               if (cs->multis != NULL)
+                       mcinvert(p, cs);
+       }
+
+       assert(cs->multis == NULL);             /* xxx */
+
+       if (nch(p, cs) == 1) {          /* optimize singleton sets */
+               ordinary(p, firstch(p, cs));
+               freeset(p, cs);
+       } else
+               EMIT(OANYOF, freezeset(p, cs));
+}
+
+/*
+ - p_b_term - parse one term of a bracketed character list
+ == static void p_b_term(register struct parse *p, register cset *cs);
+ */
+static void
+p_b_term(p, cs)
+register struct parse *p;
+register cset *cs;
+{
+       register char c;
+       register char start, finish;
+       register int i;
+
+       /* classify what we've got */
+       switch ((MORE()) ? PEEK() : '\0') {
+       case '[':
+               c = (MORE2()) ? PEEK2() : '\0';
+               break;
+       case '-':
+               SETERROR(REG_ERANGE);
+               return;                 /* NOTE RETURN */
+               break;
+       default:
+               c = '\0';
+               break;
+       }
+
+       switch (c) {
+       case ':':               /* character class */
+               NEXT2();
+               REQUIRE(MORE(), REG_EBRACK);
+               c = PEEK();
+               REQUIRE(c != '-' && c != ']', REG_ECTYPE);
+               p_b_cclass(p, cs);
+               REQUIRE(MORE(), REG_EBRACK);
+               REQUIRE(EATTWO(':', ']'), REG_ECTYPE);
+               break;
+       case '=':               /* equivalence class */
+               NEXT2();
+               REQUIRE(MORE(), REG_EBRACK);
+               c = PEEK();
+               REQUIRE(c != '-' && c != ']', REG_ECOLLATE);
+               p_b_eclass(p, cs);
+               REQUIRE(MORE(), REG_EBRACK);
+               REQUIRE(EATTWO('=', ']'), REG_ECOLLATE);
+               break;
+       default:                /* symbol, ordinary character, or range */
+/* xxx revision needed for multichar stuff */
+               start = p_b_symbol(p);
+               if (SEE('-') && MORE2() && PEEK2() != ']') {
+                       /* range */
+                       NEXT();
+                       if (EAT('-'))
+                               finish = '-';
+                       else
+                               finish = p_b_symbol(p);
+               } else
+                       finish = start;
+/* xxx what about signed chars here... */
+               REQUIRE(start <= finish, REG_ERANGE);
+               for (i = start; i <= finish; i++)
+                       CHadd(cs, i);
+               break;
+       }
+}
+
+/*
+ - p_b_cclass - parse a character-class name and deal with it
+ == static void p_b_cclass(register struct parse *p, register cset *cs);
+ */
+static void
+p_b_cclass(p, cs)
+register struct parse *p;
+register cset *cs;
+{
+       register char *sp = p->next;
+       register struct cclass *cp;
+       register size_t len;
+       register char *u;
+       register char c;
+
+       while (MORE() && isalpha(PEEK()))
+               NEXT();
+       len = p->next - sp;
+       for (cp = cclasses; cp->name != NULL; cp++)
+               if (strncmp(cp->name, sp, len) == 0 && cp->name[len] == '\0')
+                       break;
+       if (cp->name == NULL) {
+               /* oops, didn't find it */
+               SETERROR(REG_ECTYPE);
+               return;
+       }
+
+       u = cp->chars;
+       while ((c = *u++) != '\0')
+               CHadd(cs, c);
+       for (u = cp->multis; *u != '\0'; u += strlen(u) + 1)
+               MCadd(p, cs, u);
+}
+
+/*
+ - p_b_eclass - parse an equivalence-class name and deal with it
+ == static void p_b_eclass(register struct parse *p, register cset *cs);
+ *
+ * This implementation is incomplete. xxx
+ */
+static void
+p_b_eclass(p, cs)
+register struct parse *p;
+register cset *cs;
+{
+       register char c;
+
+       c = p_b_coll_elem(p, '=');
+       CHadd(cs, c);
+}
+
+/*
+ - p_b_symbol - parse a character or [..]ed multicharacter collating symbol
+ == static char p_b_symbol(register struct parse *p);
+ */
+static char                    /* value of symbol */
+p_b_symbol(p)
+register struct parse *p;
+{
+       register char value;
+
+       REQUIRE(MORE(), REG_EBRACK);
+       if (!EATTWO('[', '.'))
+               return(GETNEXT());
+
+       /* collating symbol */
+       value = p_b_coll_elem(p, '.');
+       REQUIRE(EATTWO('.', ']'), REG_ECOLLATE);
+       return(value);
+}
+
+/*
+ - p_b_coll_elem - parse a collating-element name and look it up
+ == static char p_b_coll_elem(register struct parse *p, int endc);
+ */
+static char                    /* value of collating element */
+p_b_coll_elem(p, endc)
+register struct parse *p;
+int endc;                      /* name ended by endc,']' */
+{
+       register char *sp = p->next;
+       register struct cname *cp;
+       register int len;
+
+       while (MORE() && !SEETWO(endc, ']'))
+               NEXT();
+       if (!MORE()) {
+               SETERROR(REG_EBRACK);
+               return(0);
+       }
+       len = p->next - sp;
+       for (cp = cnames; cp->name != NULL; cp++)
+               if (strncmp(cp->name, sp, len) == 0 && cp->name[len] == '\0')
+                       return(cp->code);       /* known name */
+       if (len == 1)
+               return(*sp);    /* single character */
+       SETERROR(REG_ECOLLATE);                 /* neither */
+       return(0);
+}
+
+/*
+ - othercase - return the case counterpart of an alphabetic
+ == static char othercase(int ch);
+ */
+static char                    /* if no counterpart, return ch */
+othercase(ch)
+int ch;
+{
+       assert(isalpha(ch));
+       if (isupper(ch))
+               return(tolower(ch));
+       else if (islower(ch))
+               return(toupper(ch));
+       else                    /* peculiar, but could happen */
+               return(ch);
+}
+
+/*
+ - bothcases - emit a dualcase version of a two-case character
+ == static void bothcases(register struct parse *p, int ch);
+ *
+ * Boy, is this implementation ever a kludge...
+ */
+static void
+bothcases(p, ch)
+register struct parse *p;
+int ch;
+{
+       register char *oldnext = p->next;
+       register char *oldend = p->end;
+       char bracket[3];
+
+       assert(othercase(ch) != ch);    /* p_bracket() would recurse */
+       p->next = bracket;
+       p->end = bracket+2;
+       bracket[0] = ch;
+       bracket[1] = ']';
+       bracket[2] = '\0';
+       p_bracket(p);
+       assert(p->next == bracket+2);
+       p->next = oldnext;
+       p->end = oldend;
+}
+
+/*
+ - ordinary - emit an ordinary character
+ == static void ordinary(register struct parse *p, register int ch);
+ */
+static void
+ordinary(p, ch)
+register struct parse *p;
+register int ch;
+{
+       register cat_t *cap = p->g->categories;
+
+       if ((p->g->cflags&REG_ICASE) && isalpha(ch) && othercase(ch) != ch)
+               bothcases(p, ch);
+       else {
+               EMIT(OCHAR, (unsigned char)ch);
+               if (cap[ch] == 0)
+                       cap[ch] = p->g->ncategories++;
+       }
+}
+
+/*
+ - nonnewline - emit REG_NEWLINE version of OANY
+ == static void nonnewline(register struct parse *p);
+ *
+ * Boy, is this implementation ever a kludge...
+ */
+static void
+nonnewline(p)
+register struct parse *p;
+{
+       register char *oldnext = p->next;
+       register char *oldend = p->end;
+       char bracket[4];
+
+       p->next = bracket;
+       p->end = bracket+3;
+       bracket[0] = '^';
+       bracket[1] = '\n';
+       bracket[2] = ']';
+       bracket[3] = '\0';
+       p_bracket(p);
+       assert(p->next == bracket+3);
+       p->next = oldnext;
+       p->end = oldend;
+}
+
+/*
+ - repeat - generate code for a bounded repetition, recursively if needed
+ == static void repeat(register struct parse *p, sopno start, int from, int to);
+ */
+static void
+repeat(p, start, from, to)
+register struct parse *p;
+sopno start;                   /* operand from here to end of strip */
+int from;                      /* repeated from this number */
+int to;                                /* to this number of times (maybe INFINITY) */
+{
+       register sopno finish = HERE();
+#      define  N       2
+#      define  INF     3
+#      define  REP(f, t)       ((f)*8 + (t))
+#      define  MAP(n)  (((n) <= 1) ? (n) : ((n) == INFINITY) ? INF : N)
+       register sopno copy;
+
+       if (p->error != 0)      /* head off possible runaway recursion */
+               return;
+
+       assert(from <= to);
+
+       switch (REP(MAP(from), MAP(to))) {
+       case REP(0, 0):                 /* must be user doing this */
+               DROP(finish-start);     /* drop the operand */
+               break;
+       case REP(0, 1):                 /* as x{1,1}? */
+       case REP(0, N):                 /* as x{1,n}? */
+       case REP(0, INF):               /* as x{1,}? */
+               /* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
+               INSERT(OCH_, start);            /* offset is wrong... */
+               repeat(p, start+1, 1, to);
+               ASTERN(OOR1, start);
+               AHEAD(start);                   /* ... fix it */
+               EMIT(OOR2, 0);
+               AHEAD(THERE());
+               ASTERN(O_CH, THERETHERE());
+               break;
+       case REP(1, 1):                 /* trivial case */
+               /* done */
+               break;
+       case REP(1, N):                 /* as x?x{1,n-1} */
+               /* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
+               INSERT(OCH_, start);
+               ASTERN(OOR1, start);
+               AHEAD(start);
+               EMIT(OOR2, 0);                  /* offset very wrong... */
+               AHEAD(THERE());                 /* ...so fix it */
+               ASTERN(O_CH, THERETHERE());
+               copy = dupl(p, start+1, finish+1);
+               assert(copy == finish+4);
+               repeat(p, copy, 1, to-1);
+               break;
+       case REP(1, INF):               /* as x+ */
+               INSERT(OPLUS_, start);
+               ASTERN(O_PLUS, start);
+               break;
+       case REP(N, N):                 /* as xx{m-1,n-1} */
+               copy = dupl(p, start, finish);
+               repeat(p, copy, from-1, to-1);
+               break;
+       case REP(N, INF):               /* as xx{n-1,INF} */
+               copy = dupl(p, start, finish);
+               repeat(p, copy, from-1, to);
+               break;
+       default:                        /* "can't happen" */
+               SETERROR(REG_ASSERT);   /* just in case */
+               break;
+       }
+}
+
+/*
+ - seterr - set an error condition
+ == static int seterr(register struct parse *p, int e);
+ */
+static int                     /* useless but makes type checking happy */
+seterr(p, e)
+register struct parse *p;
+int e;
+{
+       if (p->error == 0)      /* keep earliest error condition */
+               p->error = e;
+       p->next = nuls;         /* try to bring things to a halt */
+       p->end = nuls;
+       return(0);              /* make the return value well-defined */
+}
+
+/*
+ - allocset - allocate a set of characters for []
+ == static cset *allocset(register struct parse *p);
+ */
+static cset *
+allocset(p)
+register struct parse *p;
+{
+       register int no = p->g->ncsets++;
+       register size_t nc;
+       register size_t nbytes;
+       register cset *cs;
+       register size_t css = (size_t)p->g->csetsize;
+       register int i;
+
+       if (no >= p->ncsalloc) {        /* need another column of space */
+               p->ncsalloc += CHAR_BIT;
+               nc = p->ncsalloc;
+               assert(nc % CHAR_BIT == 0);
+               nbytes = nc / CHAR_BIT * css;
+               if (p->g->sets == NULL)
+                       p->g->sets = (cset *)malloc(nc * sizeof(cset));
+               else
+                       p->g->sets = (cset *)realloc((char *)p->g->sets,
+                                                       nc * sizeof(cset));
+               if (p->g->setbits == NULL)
+                       p->g->setbits = (uch *)malloc(nbytes);
+               else {
+                       p->g->setbits = (uch *)realloc((char *)p->g->setbits,
+                                                               nbytes);
+                       /* xxx this isn't right if setbits is now NULL */
+                       for (i = 0; i < no; i++)
+                               p->g->sets[i].ptr = p->g->setbits + css*(i/CHAR_BIT);
+               }
+               if (p->g->sets != NULL && p->g->setbits != NULL)
+                       (void) memset((char *)p->g->setbits + (nbytes - css),
+                                                               0, css);
+               else {
+                       no = 0;
+                       SETERROR(REG_ESPACE);
+                       /* caller's responsibility not to do set ops */
+               }
+       }
+
+       assert(p->g->sets != NULL);     /* xxx */
+       cs = &p->g->sets[no];
+       cs->ptr = p->g->setbits + css*((no)/CHAR_BIT);
+       cs->mask = 1 << ((no) % CHAR_BIT);
+       cs->hash = 0;
+       cs->smultis = 0;
+       cs->multis = NULL;
+
+       return(cs);
+}
+
+/*
+ - freeset - free a now-unused set
+ == static void freeset(register struct parse *p, register cset *cs);
+ */
+static void
+freeset(p, cs)
+register struct parse *p;
+register cset *cs;
+{
+       register int i;
+       register cset *top = &p->g->sets[p->g->ncsets];
+       register size_t css = (size_t)p->g->csetsize;
+
+       for (i = 0; i < css; i++)
+               CHsub(cs, i);
+       if (cs == top-1)        /* recover only the easy case */
+               p->g->ncsets--;
+}
+
+/*
+ - freezeset - final processing on a set of characters
+ == static int freezeset(register struct parse *p, register cset *cs);
+ *
+ * The main task here is merging identical sets.  This is usually a waste
+ * of time (although the hash code minimizes the overhead), but can win
+ * big if REG_ICASE is being used.  REG_ICASE, by the way, is why the hash
+ * is done using addition rather than xor -- all ASCII [aA] sets xor to
+ * the same value!
+ */
+static int                     /* set number */
+freezeset(p, cs)
+register struct parse *p;
+register cset *cs;
+{
+       register uch h = cs->hash;
+       register int i;
+       register cset *top = &p->g->sets[p->g->ncsets];
+       register cset *cs2;
+       register size_t css = (size_t)p->g->csetsize;
+
+       /* look for an earlier one which is the same */
+       for (cs2 = &p->g->sets[0]; cs2 < top; cs2++)
+               if (cs2->hash == h && cs2 != cs) {
+                       /* maybe */
+                       for (i = 0; i < css; i++)
+                               if (!!CHIN(cs2, i) != !!CHIN(cs, i))
+                                       break;          /* no */
+                       if (i == css)
+                               break;                  /* yes */
+               }
+
+       if (cs2 < top) {        /* found one */
+               freeset(p, cs);
+               cs = cs2;
+       }
+
+       return((int)(cs - p->g->sets));
+}
+
+/*
+ - firstch - return first character in a set (which must have at least one)
+ == static int firstch(register struct parse *p, register cset *cs);
+ */
+static int                     /* character; there is no "none" value */
+firstch(p, cs)
+register struct parse *p;
+register cset *cs;
+{
+       register int i;
+       register size_t css = (size_t)p->g->csetsize;
+
+       for (i = 0; i < css; i++)
+               if (CHIN(cs, i))
+                       return((char)i);
+       assert(never);
+       return(0);              /* arbitrary */
+}
+
+/*
+ - nch - number of characters in a set
+ == static int nch(register struct parse *p, register cset *cs);
+ */
+static int
+nch(p, cs)
+register struct parse *p;
+register cset *cs;
+{
+       register int i;
+       register size_t css = (size_t)p->g->csetsize;
+       register int n = 0;
+
+       for (i = 0; i < css; i++)
+               if (CHIN(cs, i))
+                       n++;
+       return(n);
+}
+
+/*
+ - mcadd - add a collating element to a cset
+ == static void mcadd(register struct parse *p, register cset *cs, \
+ ==    register char *cp);
+ */
+static void
+mcadd(p, cs, cp)
+register struct parse *p;
+register cset *cs;
+register char *cp;
+{
+       register size_t oldend = cs->smultis;
+
+       cs->smultis += strlen(cp) + 1;
+       if (cs->multis == NULL)
+               cs->multis = malloc(cs->smultis);
+       else
+               cs->multis = realloc(cs->multis, cs->smultis);
+       if (cs->multis == NULL) {
+               SETERROR(REG_ESPACE);
+               return;
+       }
+
+       (void) strcpy(cs->multis + oldend - 1, cp);
+       cs->multis[cs->smultis - 1] = '\0';
+}
+
+/*
+ - mcsub - subtract a collating element from a cset
+ == static void mcsub(register cset *cs, register char *cp);
+ */
+static void
+mcsub(cs, cp)
+register cset *cs;
+register char *cp;
+{
+       register char *fp = mcfind(cs, cp);
+       register size_t len = strlen(fp);
+
+       assert(fp != NULL);
+       (void) memmove(fp, fp + len + 1,
+                               cs->smultis - (fp + len + 1 - cs->multis));
+       cs->smultis -= len;
+
+       if (cs->smultis == 0) {
+               free(cs->multis);
+               cs->multis = NULL;
+               return;
+       }
+
+       cs->multis = realloc(cs->multis, cs->smultis);
+       assert(cs->multis != NULL);
+}
+
+/*
+ - mcin - is a collating element in a cset?
+ == static int mcin(register cset *cs, register char *cp);
+ */
+static int
+mcin(cs, cp)
+register cset *cs;
+register char *cp;
+{
+       return(mcfind(cs, cp) != NULL);
+}
+
+/*
+ - mcfind - find a collating element in a cset
+ == static char *mcfind(register cset *cs, register char *cp);
+ */
+static char *
+mcfind(cs, cp)
+register cset *cs;
+register char *cp;
+{
+       register char *p;
+
+       if (cs->multis == NULL)
+               return(NULL);
+       for (p = cs->multis; *p != '\0'; p += strlen(p) + 1)
+               if (strcmp(cp, p) == 0)
+                       return(p);
+       return(NULL);
+}
+
+/*
+ - mcinvert - invert the list of collating elements in a cset
+ == static void mcinvert(register struct parse *p, register cset *cs);
+ *
+ * This would have to know the set of possibilities.  Implementation
+ * is deferred.
+ */
+static void
+mcinvert(p, cs)
+register struct parse *p;
+register cset *cs;
+{
+       assert(cs->multis == NULL);     /* xxx */
+}
+
+/*
+ - mccase - add case counterparts of the list of collating elements in a cset
+ == static void mccase(register struct parse *p, register cset *cs);
+ *
+ * This would have to know the set of possibilities.  Implementation
+ * is deferred.
+ */
+static void
+mccase(p, cs)
+register struct parse *p;
+register cset *cs;
+{
+       assert(cs->multis == NULL);     /* xxx */
+}
+
+/*
+ - isinsets - is this character in any sets?
+ == static int isinsets(register struct re_guts *g, int c);
+ */
+static int                     /* predicate */
+isinsets(g, c)
+register struct re_guts *g;
+int c;
+{
+       register uch *col;
+       register int i;
+       register int ncols = (g->ncsets+(CHAR_BIT-1)) / CHAR_BIT;
+       register unsigned uc = (unsigned char)c;
+
+       for (i = 0, col = g->setbits; i < ncols; i++, col += g->csetsize)
+               if (col[uc] != 0)
+                       return(1);
+       return(0);
+}
+
+/*
+ - samesets - are these two characters in exactly the same sets?
+ == static int samesets(register struct re_guts *g, int c1, int c2);
+ */
+static int                     /* predicate */
+samesets(g, c1, c2)
+register struct re_guts *g;
+int c1;
+int c2;
+{
+       register uch *col;
+       register int i;
+       register int ncols = (g->ncsets+(CHAR_BIT-1)) / CHAR_BIT;
+       register unsigned uc1 = (unsigned char)c1;
+       register unsigned uc2 = (unsigned char)c2;
+
+       for (i = 0, col = g->setbits; i < ncols; i++, col += g->csetsize)
+               if (col[uc1] != col[uc2])
+                       return(0);
+       return(1);
+}
+
+/*
+ - categorize - sort out character categories
+ == static void categorize(struct parse *p, register struct re_guts *g);
+ */
+static void
+categorize(p, g)
+struct parse *p;
+register struct re_guts *g;
+{
+       register cat_t *cats = g->categories;
+       register int c;
+       register int c2;
+       register cat_t cat;
+
+       /* avoid making error situations worse */
+       if (p->error != 0)
+               return;
+
+       for (c = CHAR_MIN; c <= CHAR_MAX; c++)
+               if (cats[c] == 0 && isinsets(g, c)) {
+                       cat = g->ncategories++;
+                       cats[c] = cat;
+                       for (c2 = c+1; c2 <= CHAR_MAX; c2++)
+                               if (cats[c2] == 0 && samesets(g, c, c2))
+                                       cats[c2] = cat;
+               }
+}
+
+/*
+ - dupl - emit a duplicate of a bunch of sops
+ == static sopno dupl(register struct parse *p, sopno start, sopno finish);
+ */
+static sopno                   /* start of duplicate */
+dupl(p, start, finish)
+register struct parse *p;
+sopno start;                   /* from here */
+sopno finish;                  /* to this less one */
+{
+       register sopno ret = HERE();
+       register sopno len = finish - start;
+
+       assert(finish >= start);
+       if (len == 0)
+               return(ret);
+       enlarge(p, p->ssize + len);     /* this many unexpected additions */
+       assert(p->ssize >= p->slen + len);
+       (void) memcpy((char *)(p->strip + p->slen),
+               (char *)(p->strip + start), (size_t)len*sizeof(sop));
+       p->slen += len;
+       return(ret);
+}
+
+/*
+ - doemit - emit a strip operator
+ == static void doemit(register struct parse *p, sop op, size_t opnd);
+ *
+ * It might seem better to implement this as a macro with a function as
+ * hard-case backup, but it's just too big and messy unless there are
+ * some changes to the data structures.  Maybe later.
+ */
+static void
+doemit(p, op, opnd)
+register struct parse *p;
+sop op;
+size_t opnd;
+{
+       /* avoid making error situations worse */
+       if (p->error != 0)
+               return;
+
+       /* deal with oversize operands ("can't happen", more or less) */
+       assert(opnd < 1<<OPSHIFT);
+
+       /* deal with undersized strip */
+       if (p->slen >= p->ssize)
+               enlarge(p, (p->ssize+1) / 2 * 3);       /* +50% */
+       assert(p->slen < p->ssize);
+
+       /* finally, it's all reduced to the easy case */
+       p->strip[p->slen++] = SOP(op, opnd);
+}
+
+/*
+ - doinsert - insert a sop into the strip
+ == static void doinsert(register struct parse *p, sop op, size_t opnd, sopno pos);
+ */
+static void
+doinsert(p, op, opnd, pos)
+register struct parse *p;
+sop op;
+size_t opnd;
+sopno pos;
+{
+       register sopno sn;
+       register sop s;
+       register int i;
+
+       /* avoid making error situations worse */
+       if (p->error != 0)
+               return;
+
+       sn = HERE();
+       EMIT(op, opnd);         /* do checks, ensure space */
+       assert(HERE() == sn+1);
+       s = p->strip[sn];
+
+       /* adjust paren pointers */
+       assert(pos > 0);
+       for (i = 1; i < NPAREN; i++) {
+               if (p->pbegin[i] >= pos) {
+                       p->pbegin[i]++;
+               }
+               if (p->pend[i] >= pos) {
+                       p->pend[i]++;
+               }
+       }
+
+       memmove((char *)&p->strip[pos+1], (char *)&p->strip[pos],
+                                               (HERE()-pos-1)*sizeof(sop));
+       p->strip[pos] = s;
+}
+
+/*
+ - dofwd - complete a forward reference
+ == static void dofwd(register struct parse *p, sopno pos, sop value);
+ */
+static void
+dofwd(p, pos, value)
+register struct parse *p;
+register sopno pos;
+sop value;
+{
+       /* avoid making error situations worse */
+       if (p->error != 0)
+               return;
+
+       assert(value < 1<<OPSHIFT);
+       p->strip[pos] = OP(p->strip[pos]) | value;
+}
+
+/*
+ - enlarge - enlarge the strip
+ == static void enlarge(register struct parse *p, sopno size);
+ */
+static void
+enlarge(p, size)
+register struct parse *p;
+register sopno size;
+{
+       register sop *sp;
+
+       if (p->ssize >= size)
+               return;
+
+       sp = (sop *)realloc(p->strip, size*sizeof(sop));
+       if (sp == NULL) {
+               SETERROR(REG_ESPACE);
+               return;
+       }
+       p->strip = sp;
+       p->ssize = size;
+}
+
+/*
+ - stripsnug - compact the strip
+ == static void stripsnug(register struct parse *p, register struct re_guts *g);
+ */
+static void
+stripsnug(p, g)
+register struct parse *p;
+register struct re_guts *g;
+{
+       g->nstates = p->slen;
+       g->strip = (sop *)realloc((char *)p->strip, p->slen * sizeof(sop));
+       if (g->strip == NULL) {
+               SETERROR(REG_ESPACE);
+               g->strip = p->strip;
+       }
+}
+
+/*
+ - findmust - fill in must and mlen with longest mandatory literal string
+ == static void findmust(register struct parse *p, register struct re_guts *g);
+ *
+ * This algorithm could do fancy things like analyzing the operands of |
+ * for common subsequences.  Someday.  This code is simple and finds most
+ * of the interesting cases.
+ *
+ * Note that must and mlen got initialized during setup.
+ */
+static void
+findmust(p, g)
+struct parse *p;
+register struct re_guts *g;
+{
+       register sop *scan;
+       sop *start;
+       register sop *newstart;
+       register sopno newlen;
+       register sop s;
+       register char *cp;
+       register sopno i;
+
+       /* avoid making error situations worse */
+       if (p->error != 0)
+               return;
+
+       /* find the longest OCHAR sequence in strip */
+       newlen = 0;
+       scan = g->strip + 1;
+       do {
+               s = *scan++;
+               switch (OP(s)) {
+               case OCHAR:             /* sequence member */
+                       if (newlen == 0)                /* new sequence */
+                               newstart = scan - 1;
+                       newlen++;
+                       break;
+               case OPLUS_:            /* things that don't break one */
+               case OLPAREN:
+               case ORPAREN:
+                       break;
+               case OQUEST_:           /* things that must be skipped */
+               case OCH_:
+                       scan--;
+                       do {
+                               scan += OPND(s);
+                               s = *scan;
+                               /* assert() interferes w debug printouts */
+                               if (OP(s) != O_QUEST && OP(s) != O_CH &&
+                                                       OP(s) != OOR2) {
+                                       g->iflags |= BAD;
+                                       return;
+                               }
+                       } while (OP(s) != O_QUEST && OP(s) != O_CH);
+                       /* fallthrough */
+               default:                /* things that break a sequence */
+                       if (newlen > g->mlen) {         /* ends one */
+                               start = newstart;
+                               g->mlen = newlen;
+                       }
+                       newlen = 0;
+                       break;
+               }
+       } while (OP(s) != OEND);
+
+       if (g->mlen == 0)               /* there isn't one */
+               return;
+
+       /* turn it into a character string */
+       g->must = malloc((size_t)g->mlen + 1);
+       if (g->must == NULL) {          /* argh; just forget it */
+               g->mlen = 0;
+               return;
+       }
+       cp = g->must;
+       scan = start;
+       for (i = g->mlen; i > 0; i--) {
+               while (OP(s = *scan++) != OCHAR)
+                       continue;
+               assert(cp < g->must + g->mlen);
+               *cp++ = (char)OPND(s);
+       }
+       assert(cp == g->must + g->mlen);
+       *cp++ = '\0';           /* just on general principles */
+}
+
+/*
+ - pluscount - count + nesting
+ == static sopno pluscount(register struct parse *p, register struct re_guts *g);
+ */
+static sopno                   /* nesting depth */
+pluscount(p, g)
+struct parse *p;
+register struct re_guts *g;
+{
+       register sop *scan;
+       register sop s;
+       register sopno plusnest = 0;
+       register sopno maxnest = 0;
+
+       if (p->error != 0)
+               return(0);      /* there may not be an OEND */
+
+       scan = g->strip + 1;
+       do {
+               s = *scan++;
+               switch (OP(s)) {
+               case OPLUS_:
+                       plusnest++;
+                       break;
+               case O_PLUS:
+                       if (plusnest > maxnest)
+                               maxnest = plusnest;
+                       plusnest--;
+                       break;
+               }
+       } while (OP(s) != OEND);
+       if (plusnest != 0)
+               g->iflags |= BAD;
+       return(maxnest);
+}
diff --git a/lib/rxspencer/regerror.c b/lib/rxspencer/regerror.c
new file mode 100644 (file)
index 0000000..e53dafc
--- /dev/null
@@ -0,0 +1,126 @@
+#include <sys/types.h>
+#include <stdio.h>
+#include <string.h>
+#include <ctype.h>
+#include <limits.h>
+#include <stdlib.h>
+#include <regex.h>
+
+#include "utils.h"
+#include "regerror.ih"
+
+/*
+ = #define     REG_OKAY         0
+ = #define     REG_NOMATCH      1
+ = #define     REG_BADPAT       2
+ = #define     REG_ECOLLATE     3
+ = #define     REG_ECTYPE       4
+ = #define     REG_EESCAPE      5
+ = #define     REG_ESUBREG      6
+ = #define     REG_EBRACK       7
+ = #define     REG_EPAREN       8
+ = #define     REG_EBRACE       9
+ = #define     REG_BADBR       10
+ = #define     REG_ERANGE      11
+ = #define     REG_ESPACE      12
+ = #define     REG_BADRPT      13
+ = #define     REG_EMPTY       14
+ = #define     REG_ASSERT      15
+ = #define     REG_INVARG      16
+ = #define     REG_ATOI        255     // convert name to number (!)
+ = #define     REG_ITOA        0400    // convert number to name (!)
+ */
+static struct rerr {
+       int code;
+       char *name;
+       char *explain;
+} rerrs[] = {
+       REG_OKAY,       "REG_OKAY",     "no errors detected",
+       REG_NOMATCH,    "REG_NOMATCH",  "regexec() failed to match",
+       REG_BADPAT,     "REG_BADPAT",   "invalid regular expression",
+       REG_ECOLLATE,   "REG_ECOLLATE", "invalid collating element",
+       REG_ECTYPE,     "REG_ECTYPE",   "invalid character class",
+       REG_EESCAPE,    "REG_EESCAPE",  "trailing backslash (\\)",
+       REG_ESUBREG,    "REG_ESUBREG",  "invalid backreference number",
+       REG_EBRACK,     "REG_EBRACK",   "brackets ([ ]) not balanced",
+       REG_EPAREN,     "REG_EPAREN",   "parentheses not balanced",
+       REG_EBRACE,     "REG_EBRACE",   "braces not balanced",
+       REG_BADBR,      "REG_BADBR",    "invalid repetition count(s)",
+       REG_ERANGE,     "REG_ERANGE",   "invalid character range",
+       REG_ESPACE,     "REG_ESPACE",   "out of memory",
+       REG_BADRPT,     "REG_BADRPT",   "repetition-operator operand invalid",
+       REG_EMPTY,      "REG_EMPTY",    "empty (sub)expression",
+       REG_ASSERT,     "REG_ASSERT",   "\"can't happen\" -- you found a bug",
+       REG_INVARG,     "REG_INVARG",   "invalid argument to regex routine",
+       -1,             "",             "*** unknown regexp error code ***",
+};
+
+/*
+ - regerror - the interface to error numbers
+ = extern size_t regerror(int, const regex_t *, char *, size_t);
+ */
+/* ARGSUSED */
+size_t
+regerror(errcode, preg, errbuf, errbuf_size)
+int errcode;
+const regex_t *preg;
+char *errbuf;
+size_t errbuf_size;
+{
+       register struct rerr *r;
+       register size_t len;
+       register int target = errcode &~ REG_ITOA;
+       register char *s;
+       char convbuf[50];
+
+       if (errcode == REG_ATOI)
+               s = regatoi(preg, convbuf);
+       else {
+               for (r = rerrs; r->code >= 0; r++)
+                       if (r->code == target)
+                               break;
+       
+               if (errcode&REG_ITOA) {
+                       if (r->code >= 0)
+                               (void) strcpy(convbuf, r->name);
+                       else
+                               sprintf(convbuf, "REG_0x%x", target);
+                       assert(strlen(convbuf) < sizeof(convbuf));
+                       s = convbuf;
+               } else
+                       s = r->explain;
+       }
+
+       len = strlen(s) + 1;
+       if (errbuf_size > 0) {
+               if (errbuf_size > len)
+                       (void) strcpy(errbuf, s);
+               else {
+                       (void) strncpy(errbuf, s, errbuf_size-1);
+                       errbuf[errbuf_size-1] = '\0';
+               }
+       }
+
+       return(len);
+}
+
+/*
+ - regatoi - internal routine to implement REG_ATOI
+ == static char *regatoi(const regex_t *preg, char *localbuf);
+ */
+static char *
+regatoi(preg, localbuf)
+const regex_t *preg;
+char *localbuf;
+{
+       register struct rerr *r;
+
+       for (r = rerrs; r->code >= 0; r++)
+               if (strcmp(r->name, preg->re_endp) == 0)
+                       break;
+       if (r->code < 0)
+               return("0");
+
+       sprintf(localbuf, "%d", r->code);
+       return(localbuf);
+}
diff --git a/lib/rxspencer/regex2.h b/lib/rxspencer/regex2.h
new file mode 100644 (file)
index 0000000..58fd8d8
--- /dev/null
@@ -0,0 +1,134 @@
+/*
+ * First, the stuff that ends up in the outside-world include file
+ = typedef off_t regoff_t;
+ = typedef struct {
+ =     int re_magic;
+ =     size_t re_nsub;         // number of parenthesized subexpressions
+ =     const char *re_endp;    // end pointer for REG_PEND
+ =     struct re_guts *re_g;   // none of your business :-)
+ = } regex_t;
+ = typedef struct {
+ =     regoff_t rm_so;         // start of match
+ =     regoff_t rm_eo;         // end of match
+ = } regmatch_t;
+ */
+/*
+ * internals of regex_t
+ */
+#define        MAGIC1  ((('r'^0200)<<8) | 'e')
+
+/*
+ * The internal representation is a *strip*, a sequence of
+ * operators ending with an endmarker.  (Some terminology etc. is a
+ * historical relic of earlier versions which used multiple strips.)
+ * Certain oddities in the representation are there to permit running
+ * the machinery backwards; in particular, any deviation from sequential
+ * flow must be marked at both its source and its destination.  Some
+ * fine points:
+ *
+ * - OPLUS_ and O_PLUS are *inside* the loop they create.
+ * - OQUEST_ and O_QUEST are *outside* the bypass they create.
+ * - OCH_ and O_CH are *outside* the multi-way branch they create, while
+ *   OOR1 and OOR2 are respectively the end and the beginning of one of
+ *   the branches.  Note that there is an implicit OOR2 following OCH_
+ *   and an implicit OOR1 preceding O_CH.
+ *
+ * In state representations, an operator's bit is on to signify a state
+ * immediately *preceding* "execution" of that operator.
+ */
+typedef long sop;              /* strip operator */
+typedef long sopno;
+#define        OPRMASK 0x7c000000
+#define        OPDMASK 0x03ffffff
+#define        OPSHIFT (26)
+#define        OP(n)   ((n)&OPRMASK)
+#define        OPND(n) ((n)&OPDMASK)
+#define        SOP(op, opnd)   ((op)|(opnd))
+/* operators                      meaning      operand                 */
+/*                                             (back, fwd are offsets) */
+#define        OEND    (1<<OPSHIFT)    /* endmarker    -                       */
+#define        OCHAR   (2<<OPSHIFT)    /* character    unsigned char           */
+#define        OBOL    (3<<OPSHIFT)    /* left anchor  -                       */
+#define        OEOL    (4<<OPSHIFT)    /* right anchor -                       */
+#define        OANY    (5<<OPSHIFT)    /* .            -                       */
+#define        OANYOF  (6<<OPSHIFT)    /* [...]        set number              */
+#define        OBACK_  (7<<OPSHIFT)    /* begin \d     paren number            */
+#define        O_BACK  (8<<OPSHIFT)    /* end \d       paren number            */
+#define        OPLUS_  (9<<OPSHIFT)    /* + prefix     fwd to suffix           */
+#define        O_PLUS  (10<<OPSHIFT)   /* + suffix     back to prefix          */
+#define        OQUEST_ (11<<OPSHIFT)   /* ? prefix     fwd to suffix           */
+#define        O_QUEST (12<<OPSHIFT)   /* ? suffix     back to prefix          */
+#define        OLPAREN (13<<OPSHIFT)   /* (            fwd to )                */
+#define        ORPAREN (14<<OPSHIFT)   /* )            back to (               */
+#define        OCH_    (15<<OPSHIFT)   /* begin choice fwd to OOR2             */
+#define        OOR1    (16<<OPSHIFT)   /* | pt. 1      back to OOR1 or OCH_    */
+#define        OOR2    (17<<OPSHIFT)   /* | pt. 2      fwd to OOR2 or O_CH     */
+#define        O_CH    (18<<OPSHIFT)   /* end choice   back to OOR1            */
+#define        OBOW    (19<<OPSHIFT)   /* begin word   -                       */
+#define        OEOW    (20<<OPSHIFT)   /* end word     -                       */
+
+/*
+ * Structure for [] character-set representation.  Character sets are
+ * done as bit vectors, grouped 8 to a byte vector for compactness.
+ * The individual set therefore has both a pointer to the byte vector
+ * and a mask to pick out the relevant bit of each byte.  A hash code
+ * simplifies testing whether two sets could be identical.
+ *
+ * This will get trickier for multicharacter collating elements.  As
+ * preliminary hooks for dealing with such things, we also carry along
+ * a string of multi-character elements, and decide the size of the
+ * vectors at run time.
+ */
+typedef struct {
+       uch *ptr;               /* -> uch [csetsize] */
+       uch mask;               /* bit within array */
+       uch hash;               /* hash code */
+       size_t smultis;
+       char *multis;           /* -> char[smulti]  ab\0cd\0ef\0\0 */
+} cset;
+/* note that CHadd and CHsub are unsafe, and CHIN doesn't yield 0/1 */
+#define        CHadd(cs, c)    ((cs)->ptr[(uch)(c)] |= (cs)->mask, (cs)->hash += (c))
+#define        CHsub(cs, c)    ((cs)->ptr[(uch)(c)] &= ~(cs)->mask, (cs)->hash -= (c))
+#define        CHIN(cs, c)     ((cs)->ptr[(uch)(c)] & (cs)->mask)
+#define        MCadd(p, cs, cp)        mcadd(p, cs, cp)        /* regcomp() internal fns */
+#define        MCsub(p, cs, cp)        mcsub(p, cs, cp)
+#define        MCin(p, cs, cp) mcin(p, cs, cp)
+
+/* stuff for character categories */
+typedef unsigned char cat_t;
+
+/*
+ * main compiled-expression structure
+ */
+struct re_guts {
+       int magic;
+#              define  MAGIC2  ((('R'^0200)<<8)|'E')
+       sop *strip;             /* malloced area for strip */
+       int csetsize;           /* number of bits in a cset vector */
+       int ncsets;             /* number of csets in use */
+       cset *sets;             /* -> cset [ncsets] */
+       uch *setbits;           /* -> uch[csetsize][ncsets/CHAR_BIT] */
+       int cflags;             /* copy of regcomp() cflags argument */
+       sopno nstates;          /* = number of sops */
+       sopno firststate;       /* the initial OEND (normally 0) */
+       sopno laststate;        /* the final OEND */
+       int iflags;             /* internal flags */
+#              define  USEBOL  01      /* used ^ */
+#              define  USEEOL  02      /* used $ */
+#              define  BAD     04      /* something wrong */
+       int nbol;               /* number of ^ used */
+       int neol;               /* number of $ used */
+       int ncategories;        /* how many character categories */
+       cat_t *categories;      /* ->catspace[-CHAR_MIN] */
+       char *must;             /* match must contain this string */
+       int mlen;               /* length of must */
+       size_t nsub;            /* copy of re_nsub */
+       int backrefs;           /* does it use back references? */
+       sopno nplus;            /* how deep does it nest +s? */
+       /* catspace must be last */
+       cat_t catspace[1];      /* actually [NC] */
+};
+
+/* misc utilities */
+#define        OUT     (CHAR_MAX+1)    /* a non-character value */
+#define        ISWORD(c)       (isalnum(c) || (c) == '_')
diff --git a/lib/rxspencer/regexec.c b/lib/rxspencer/regexec.c
new file mode 100644 (file)
index 0000000..dcb11b2
--- /dev/null
@@ -0,0 +1,138 @@
+/*
+ * the outer shell of regexec()
+ *
+ * This file includes engine.c *twice*, after muchos fiddling with the
+ * macros that code uses.  This lets the same code operate on two different
+ * representations for state sets.
+ */
+#include <sys/types.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <limits.h>
+#include <ctype.h>
+#include <regex.h>
+
+#include "utils.h"
+#include "regex2.h"
+
+static int nope = 0;           /* for use in asserts; shuts lint up */
+
+/* macros for manipulating states, small version */
+#define        states  unsigned
+#define        states1 unsigned        /* for later use in regexec() decision */
+#define        CLEAR(v)        ((v) = 0)
+#define        SET0(v, n)      ((v) &= ~((unsigned)1 << (n)))
+#define        SET1(v, n)      ((v) |= (unsigned)1 << (n))
+#define        ISSET(v, n)     ((v) & ((unsigned)1 << (n)))
+#define        ASSIGN(d, s)    ((d) = (s))
+#define        EQ(a, b)        ((a) == (b))
+#define        STATEVARS       int dummy       /* dummy version */
+#define        STATESETUP(m, n)        /* nothing */
+#define        STATETEARDOWN(m)        /* nothing */
+#define        SETUP(v)        ((v) = 0)
+#define        onestate        unsigned
+#define        INIT(o, n)      ((o) = (unsigned)1 << (n))
+#define        INC(o)  ((o) <<= 1)
+#define        ISSTATEIN(v, o) ((v) & (o))
+/* some abbreviations; note that some of these know variable names! */
+/* do "if I'm here, I can also be there" etc without branches */
+#define        FWD(dst, src, n)        ((dst) |= ((unsigned)(src)&(here)) << (n))
+#define        BACK(dst, src, n)       ((dst) |= ((unsigned)(src)&(here)) >> (n))
+#define        ISSETBACK(v, n) ((v) & ((unsigned)here >> (n)))
+/* function names */
+#define SNAMES                 /* engine.c looks after details */
+
+#include "engine.c"
+
+/* now undo things */
+#undef states
+#undef CLEAR
+#undef SET0
+#undef SET1
+#undef ISSET
+#undef ASSIGN
+#undef EQ
+#undef STATEVARS
+#undef STATESETUP
+#undef STATETEARDOWN
+#undef SETUP
+#undef onestate
+#undef INIT
+#undef INC
+#undef ISSTATEIN
+#undef FWD
+#undef BACK
+#undef ISSETBACK
+#undef SNAMES
+
+/* macros for manipulating states, large version */
+#define        states  char *
+#define        CLEAR(v)        memset(v, 0, m->g->nstates)
+#define        SET0(v, n)      ((v)[n] = 0)
+#define        SET1(v, n)      ((v)[n] = 1)
+#define        ISSET(v, n)     ((v)[n])
+#define        ASSIGN(d, s)    memcpy(d, s, m->g->nstates)
+#define        EQ(a, b)        (memcmp(a, b, m->g->nstates) == 0)
+#define        STATEVARS       int vn; char *space
+#define        STATESETUP(m, nv)       { (m)->space = malloc((nv)*(m)->g->nstates); \
+                               if ((m)->space == NULL) return(REG_ESPACE); \
+                               (m)->vn = 0; }
+#define        STATETEARDOWN(m)        { free((m)->space); }
+#define        SETUP(v)        ((v) = &m->space[m->vn++ * m->g->nstates])
+#define        onestate        int
+#define        INIT(o, n)      ((o) = (n))
+#define        INC(o)  ((o)++)
+#define        ISSTATEIN(v, o) ((v)[o])
+/* some abbreviations; note that some of these know variable names! */
+/* do "if I'm here, I can also be there" etc without branches */
+#define        FWD(dst, src, n)        ((dst)[here+(n)] |= (src)[here])
+#define        BACK(dst, src, n)       ((dst)[here-(n)] |= (src)[here])
+#define        ISSETBACK(v, n) ((v)[here - (n)])
+/* function names */
+#define        LNAMES                  /* flag */
+
+#include "engine.c"
+
+/*
+ - regexec - interface for matching
+ = extern int regexec(const regex_t *, const char *, size_t, \
+ =                                     regmatch_t [], int);
+ = #define     REG_NOTBOL      00001
+ = #define     REG_NOTEOL      00002
+ = #define     REG_STARTEND    00004
+ = #define     REG_TRACE       00400   // tracing of execution
+ = #define     REG_LARGE       01000   // force large representation
+ = #define     REG_BACKR       02000   // force use of backref code
+ *
+ * We put this here so we can exploit knowledge of the state representation
+ * when choosing which matcher to call.  Also, by this point the matchers
+ * have been prototyped.
+ */
+int                            /* 0 success, REG_NOMATCH failure */
+regexec(preg, string, nmatch, pmatch, eflags)
+const regex_t *preg;
+const char *string;
+size_t nmatch;
+regmatch_t pmatch[];
+int eflags;
+{
+       register struct re_guts *g = preg->re_g;
+#ifdef REDEBUG
+#      define  GOODFLAGS(f)    (f)
+#else
+#      define  GOODFLAGS(f)    ((f)&(REG_NOTBOL|REG_NOTEOL|REG_STARTEND))
+#endif
+
+       if (preg->re_magic != MAGIC1 || g->magic != MAGIC2)
+               return(REG_BADPAT);
+       assert(!(g->iflags&BAD));
+       if (g->iflags&BAD)              /* backstop for no-debug case */
+               return(REG_BADPAT);
+       eflags = GOODFLAGS(eflags);
+
+       if (g->nstates <= CHAR_BIT*sizeof(states1) && !(eflags&REG_LARGE))
+               return(smatcher(g, (char *)string, nmatch, pmatch, eflags));
+       else
+               return(lmatcher(g, (char *)string, nmatch, pmatch, eflags));
+}
diff --git a/lib/rxspencer/regfree.c b/lib/rxspencer/regfree.c
new file mode 100644 (file)
index 0000000..9a6acf1
--- /dev/null
@@ -0,0 +1,37 @@
+#include <sys/types.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <regex.h>
+
+#include "utils.h"
+#include "regex2.h"
+
+/*
+ - regfree - free everything
+ = extern void regfree(regex_t *);
+ */
+void
+regfree(preg)
+regex_t *preg;
+{
+       register struct re_guts *g;
+
+       if (preg->re_magic != MAGIC1)   /* oops */
+               return;                 /* nice to complain, but hard */
+
+       g = preg->re_g;
+       if (g == NULL || g->magic != MAGIC2)    /* oops again */
+               return;
+       preg->re_magic = 0;             /* mark it invalid */
+       g->magic = 0;                   /* mark it invalid */
+
+       if (g->strip != NULL)
+               free((char *)g->strip);
+       if (g->sets != NULL)
+               free((char *)g->sets);
+       if (g->setbits != NULL)
+               free((char *)g->setbits);
+       if (g->must != NULL)
+               free(g->must);
+       free((char *)g);
+}
diff --git a/lib/rxspencer/rxspencer.3 b/lib/rxspencer/rxspencer.3
new file mode 100644 (file)
index 0000000..3016432
--- /dev/null
@@ -0,0 +1,509 @@
+.TH REGEX 3 "25 Sept 1997"
+.BY "Henry Spencer"
+.de ZR
+.\" one other place knows this name:  the SEE ALSO section
+.IR rxspencer (7) \\$1
+..
+.SH NAME
+regcomp, regexec, regerror, regfree \- regular-expression library
+.SH SYNOPSIS
+.ft B
+.\".na
+#include <sys/types.h>
+.br
+#include <regex.h>
+.HP 10
+int regcomp(regex_t\ *preg, const\ char\ *pattern, int\ cflags);
+.HP
+int\ regexec(const\ regex_t\ *preg, const\ char\ *string,
+size_t\ nmatch, regmatch_t\ pmatch[], int\ eflags);
+.HP
+size_t\ regerror(int\ errcode, const\ regex_t\ *preg,
+char\ *errbuf, size_t\ errbuf_size);
+.HP
+void\ regfree(regex_t\ *preg);
+.\".ad
+.ft
+.SH DESCRIPTION
+These routines implement POSIX 1003.2 regular expressions (``RE''s);
+see
+.ZR .
+.I Regcomp
+compiles an RE written as a string into an internal form,
+.I regexec
+matches that internal form against a string and reports results,
+.I regerror
+transforms error codes from either into human-readable messages,
+and
+.I regfree
+frees any dynamically-allocated storage used by the internal form
+of an RE.
+.PP
+The header
+.I <regex.h>
+declares two structure types,
+.I regex_t
+and
+.IR regmatch_t ,
+the former for compiled internal forms and the latter for match reporting.
+It also declares the four functions,
+a type
+.IR regoff_t ,
+and a number of constants with names starting with ``REG_''.
+.PP
+.I Regcomp
+compiles the regular expression contained in the
+.I pattern
+string,
+subject to the flags in
+.IR cflags ,
+and places the results in the
+.I regex_t
+structure pointed to by
+.IR preg .
+.I Cflags
+is the bitwise OR of zero or more of the following flags:
+.IP REG_EXTENDED \w'REG_EXTENDED'u+2n
+Compile modern (``extended'') REs,
+rather than the obsolete (``basic'') REs that
+are the default.
+.IP REG_BASIC
+This is a synonym for 0,
+provided as a counterpart to REG_EXTENDED to improve readability.
+This is an extension,
+compatible with but not specified by POSIX 1003.2,
+and should be used with
+caution in software intended to be portable to other systems.
+.IP REG_NOSPEC
+Compile with recognition of all special characters turned off.
+All characters are thus considered ordinary,
+so the ``RE'' is a literal string.
+This is an extension,
+compatible with but not specified by POSIX 1003.2,
+and should be used with
+caution in software intended to be portable to other systems.
+REG_EXTENDED and REG_NOSPEC may not be used
+in the same call to
+.IR regcomp .
+.IP REG_ICASE
+Compile for matching that ignores upper/lower case distinctions.
+See
+.ZR .
+.IP REG_NOSUB
+Compile for matching that need only report success or failure,
+not what was matched.
+.IP REG_NEWLINE
+Compile for newline-sensitive matching.
+By default, newline is a completely ordinary character with no special
+meaning in either REs or strings.
+With this flag,
+`[^' bracket expressions and `.' never match newline,
+a `^' anchor matches the null string after any newline in the string
+in addition to its normal function,
+and the `$' anchor matches the null string before any newline in the
+string in addition to its normal function.
+.IP REG_PEND
+The regular expression ends,
+not at the first NUL,
+but just before the character pointed to by the
+.I re_endp
+member of the structure pointed to by
+.IR preg .
+The
+.I re_endp
+member is of type
+.IR const\ char\ * .
+This flag permits inclusion of NULs in the RE;
+they are considered ordinary characters.
+This is an extension,
+compatible with but not specified by POSIX 1003.2,
+and should be used with
+caution in software intended to be portable to other systems.
+.PP
+When successful,
+.I regcomp
+returns 0 and fills in the structure pointed to by
+.IR preg .
+One member of that structure
+(other than
+.IR re_endp )
+is publicized:
+.IR re_nsub ,
+of type
+.IR size_t ,
+contains the number of parenthesized subexpressions within the RE
+(except that the value of this member is undefined if the
+REG_NOSUB flag was used).
+If
+.I regcomp
+fails, it returns a non-zero error code;
+see DIAGNOSTICS.
+.PP
+.I Regexec
+matches the compiled RE pointed to by
+.I preg
+against the
+.IR string ,
+subject to the flags in
+.IR eflags ,
+and reports results using
+.IR nmatch ,
+.IR pmatch ,
+and the returned value.
+The RE must have been compiled by a previous invocation of
+.IR regcomp .
+The compiled form is not altered during execution of
+.IR regexec ,
+so a single compiled RE can be used simultaneously by multiple threads.
+.PP
+By default,
+the NUL-terminated string pointed to by
+.I string
+is considered to be the text of an entire line,
+with the NUL indicating the end of the line.
+(That is,
+any other end-of-line marker is considered to have been removed
+and replaced by the NUL.)
+The
+.I eflags
+argument is the bitwise OR of zero or more of the following flags:
+.IP REG_NOTBOL \w'REG_STARTEND'u+2n
+The first character of
+the string
+is not the beginning of a line, so the `^' anchor should not match before it.
+This does not affect the behavior of newlines under REG_NEWLINE.
+.IP REG_NOTEOL
+The NUL terminating
+the string
+does not end a line, so the `$' anchor should not match before it.
+This does not affect the behavior of newlines under REG_NEWLINE.
+.IP REG_STARTEND
+The string is considered to start at
+\fIstring\fR\ + \fIpmatch\fR[0].\fIrm_so\fR
+and to have a terminating NUL located at
+\fIstring\fR\ + \fIpmatch\fR[0].\fIrm_eo\fR
+(there need not actually be a NUL at that location),
+regardless of the value of
+.IR nmatch .
+See below for the definition of
+.IR pmatch
+and
+.IR nmatch .
+This is an extension,
+compatible with but not specified by POSIX 1003.2,
+and should be used with
+caution in software intended to be portable to other systems.
+Note that a non-zero \fIrm_so\fR does not imply REG_NOTBOL;
+REG_STARTEND affects only the location of the string,
+not how it is matched.
+.PP
+See
+.ZR
+for a discussion of what is matched in situations where an RE or a
+portion thereof could match any of several substrings of
+.IR string .
+.PP
+Normally,
+.I regexec
+returns 0 for success and the non-zero code REG_NOMATCH for failure.
+Other non-zero error codes may be returned in exceptional situations;
+see DIAGNOSTICS.
+.PP
+If REG_NOSUB was specified in the compilation of the RE,
+or if
+.I nmatch
+is 0,
+.I regexec
+ignores the
+.I pmatch
+argument (but see below for the case where REG_STARTEND is specified).
+Otherwise,
+.I pmatch
+points to an array of
+.I nmatch
+structures of type
+.IR regmatch_t .
+Such a structure has at least the members
+.I rm_so
+and
+.IR rm_eo ,
+both of type
+.I regoff_t
+(a signed arithmetic type at least as large as an
+.I off_t
+and a
+.IR ssize_t ),
+containing respectively the offset of the first character of a substring
+and the offset of the first character after the end of the substring.
+Offsets are measured from the beginning of the
+.I string
+argument given to
+.IR regexec .
+An empty substring is denoted by equal offsets,
+both indicating the character following the empty substring.
+.PP
+The 0th member of the
+.I pmatch
+array is filled in to indicate what substring of
+.I string
+was matched by the entire RE.
+Remaining members report what substring was matched by parenthesized
+subexpressions within the RE;
+member
+.I i
+reports subexpression
+.IR i ,
+with subexpressions counted (starting at 1) by the order of their opening
+parentheses in the RE, left to right.
+Unused entries in the array\(emcorresponding either to subexpressions that
+did not participate in the match at all, or to subexpressions that do not
+exist in the RE (that is, \fIi\fR\ > \fIpreg\fR\->\fIre_nsub\fR)\(emhave both
+.I rm_so
+and
+.I rm_eo
+set to \-1.
+If a subexpression participated in the match several times,
+the reported substring is the last one it matched.
+(Note, as an example in particular, that when the RE `(b*)+' matches `bbb',
+the parenthesized subexpression matches the three `b's and then
+an infinite number of empty strings following the last `b',
+so the reported substring is one of the empties.)
+.PP
+If REG_STARTEND is specified,
+.I pmatch
+must point to at least one
+.I regmatch_t
+(even if
+.I nmatch
+is 0 or REG_NOSUB was specified),
+to hold the input offsets for REG_STARTEND.
+Use for output is still entirely controlled by
+.IR nmatch ;
+if
+.I nmatch
+is 0 or REG_NOSUB was specified,
+the value of
+.IR pmatch [0]
+will not be changed by a successful
+.IR regexec .
+.PP
+.I Regerror
+maps a non-zero
+.I errcode
+from either
+.I regcomp
+or
+.I regexec
+to a human-readable, printable message.
+If
+.I preg
+is non-NULL,
+the error code should have arisen from use of
+the
+.I regex_t
+pointed to by
+.IR preg ,
+and if the error code came from
+.IR regcomp ,
+it should have been the result from the most recent
+.I regcomp
+using that
+.IR regex_t .
+.RI ( Regerror
+may be able to supply a more detailed message using information
+from the
+.IR regex_t .)
+.I Regerror
+places the NUL-terminated message into the buffer pointed to by
+.IR errbuf ,
+limiting the length (including the NUL) to at most
+.I errbuf_size
+bytes.
+If the whole message won't fit,
+as much of it as will fit before the terminating NUL is supplied.
+In any case,
+the returned value is the size of buffer needed to hold the whole
+message (including terminating NUL).
+If
+.I errbuf_size
+is 0,
+.I errbuf
+is ignored but the return value is still correct.
+.PP
+If the
+.I errcode
+given to
+.I regerror
+is first ORed with REG_ITOA,
+the ``message'' that results is the printable name of the error code,
+e.g. ``REG_NOMATCH'',
+rather than an explanation thereof.
+If
+.I errcode
+is REG_ATOI,
+then
+.I preg
+shall be non-NULL and the
+.I re_endp
+member of the structure it points to
+must point to the printable name of an error code;
+in this case, the result in
+.I errbuf
+is the decimal digits of
+the numeric value of the error code
+(0 if the name is not recognized).
+REG_ITOA and REG_ATOI are intended primarily as debugging facilities;
+they are extensions,
+compatible with but not specified by POSIX 1003.2,
+and should be used with
+caution in software intended to be portable to other systems.
+Be warned also that they are considered experimental and changes are possible.
+.PP
+.I Regfree
+frees any dynamically-allocated storage associated with the compiled RE
+pointed to by
+.IR preg .
+The remaining
+.I regex_t
+is no longer a valid compiled RE
+and the effect of supplying it to
+.I regexec
+or
+.I regerror
+is undefined.
+.PP
+None of these functions references global variables except for tables
+of constants;
+all are safe for use from multiple threads if the arguments are safe.
+.SH IMPLEMENTATION CHOICES
+There are a number of decisions that 1003.2 leaves up to the implementor,
+either by explicitly saying ``undefined'' or by virtue of them being
+forbidden by the RE grammar.
+This implementation treats them as follows.
+.PP
+See
+.ZR
+for a discussion of the definition of case-independent matching.
+.PP
+There is no particular limit on the length of REs,
+except insofar as memory is limited.
+Memory usage is approximately linear in RE size, and largely insensitive
+to RE complexity, except for bounded repetitions.
+See BUGS for one short RE using them
+that will run almost any system out of memory.
+.PP
+A backslashed character other than one specifically given a magic meaning
+by 1003.2 (such magic meanings occur only in obsolete [``basic''] REs)
+is taken as an ordinary character.
+.PP
+Any unmatched [ is a REG_EBRACK error.
+.PP
+Equivalence classes cannot begin or end bracket-expression ranges.
+The endpoint of one range cannot begin another.
+.PP
+RE_DUP_MAX, the limit on repetition counts in bounded repetitions, is 255.
+.PP
+A repetition operator (?, *, +, or bounds) cannot follow another
+repetition operator.
+A repetition operator cannot begin an expression or subexpression
+or follow `^' or `|'.
+.PP
+`|' cannot appear first or last in a (sub)expression or after another `|',
+i.e. an operand of `|' cannot be an empty subexpression.
+An empty parenthesized subexpression, `()', is legal and matches an
+empty (sub)string.
+An empty string is not a legal RE.
+.PP
+A `{' followed by a digit is considered the beginning of bounds for a
+bounded repetition, which must then follow the syntax for bounds.
+A `{' \fInot\fR followed by a digit is considered an ordinary character.
+.PP
+`^' and `$' beginning and ending subexpressions in obsolete (``basic'')
+REs are anchors, not ordinary characters.
+.SH SEE ALSO
+grep(1), rxspencer(7)
+.PP
+POSIX 1003.2, sections 2.8 (Regular Expression Notation)
+and
+B.5 (C Binding for Regular Expression Matching).
+.SH DIAGNOSTICS
+Non-zero error codes from
+.I regcomp
+and
+.I regexec
+include the following:
+.PP
+.nf
+.ta \w'REG_ECOLLATE'u+3n
+REG_NOMATCH    regexec() failed to match
+REG_BADPAT     invalid regular expression
+REG_ECOLLATE   invalid collating element
+REG_ECTYPE     invalid character class
+REG_EESCAPE    \e applied to unescapable character
+REG_ESUBREG    invalid backreference number
+REG_EBRACK     brackets [ ] not balanced
+REG_EPAREN     parentheses ( ) not balanced
+REG_EBRACE     braces { } not balanced
+REG_BADBR      invalid repetition count(s) in { }
+REG_ERANGE     invalid character range in [ ]
+REG_ESPACE     ran out of memory
+REG_BADRPT     ?, *, or + operand invalid
+REG_EMPTY      empty (sub)expression
+REG_ASSERT     ``can't happen''\(emyou found a bug
+REG_INVARG     invalid argument, e.g. negative-length string
+.fi
+.SH HISTORY
+Written by Henry Spencer,
+henry@zoo.toronto.edu.
+.SH BUGS
+This is an alpha release with known defects.
+Please report problems.
+.PP
+There is one known functionality bug.
+The implementation of internationalization is incomplete:
+the locale is always assumed to be the default one of 1003.2,
+and only the collating elements etc. of that locale are available.
+.PP
+The back-reference code is subtle and doubts linger about its correctness
+in complex cases.
+.PP
+.I Regexec
+performance is poor.
+This will improve with later releases.
+.I Nmatch
+exceeding 0 is expensive;
+.I nmatch
+exceeding 1 is worse.
+.I Regexec
+is largely insensitive to RE complexity \fIexcept\fR that back
+references are massively expensive.
+RE length does matter; in particular, there is a strong speed bonus
+for keeping RE length under about 30 characters,
+with most special characters counting roughly double.
+.PP
+.I Regcomp
+implements bounded repetitions by macro expansion,
+which is costly in time and space if counts are large
+or bounded repetitions are nested.
+An RE like, say,
+`((((a{1,100}){1,100}){1,100}){1,100}){1,100}'
+will (eventually) run almost any existing machine out of swap space.
+.PP
+There are suspected problems with response to obscure error conditions.
+Notably,
+certain kinds of internal overflow,
+produced only by truly enormous REs or by multiply nested bounded repetitions,
+are probably not handled well.
+.PP
+Due to a mistake in 1003.2, things like `a)b' are legal REs because `)' is
+a special character only in the presence of a previous unmatched `('.
+This can't be fixed until the spec is fixed.
+.PP
+The standard's definition of back references is vague.
+For example, does
+`a\e(\e(b\e)*\e2\e)*d' match `abbbd'?
+Until the standard is clarified,
+behavior in such cases should not be relied on.
+.PP
+The implementation of word-boundary matching is a bit of a kludge,
+and bugs may lurk in combinations of word-boundary matching and anchoring.
diff --git a/lib/rxspencer/rxspencer.7 b/lib/rxspencer/rxspencer.7
new file mode 100644 (file)
index 0000000..387790a
--- /dev/null
@@ -0,0 +1,235 @@
+.TH REGEX 7 "25 Oct 1995"
+.BY "Henry Spencer"
+.SH NAME
+regex \- POSIX 1003.2 regular expressions
+.SH DESCRIPTION
+Regular expressions (``RE''s),
+as defined in POSIX 1003.2, come in two forms:
+modern REs (roughly those of
+.IR egrep ;
+1003.2 calls these ``extended'' REs)
+and obsolete REs (roughly those of
+.IR ed ;
+1003.2 ``basic'' REs).
+Obsolete REs mostly exist for backward compatibility in some old programs;
+they will be discussed at the end.
+1003.2 leaves some aspects of RE syntax and semantics open;
+`\(dg' marks decisions on these aspects that
+may not be fully portable to other 1003.2 implementations.
+.PP
+A (modern) RE is one\(dg or more non-empty\(dg \fIbranches\fR,
+separated by `|'.
+It matches anything that matches one of the branches.
+.PP
+A branch is one\(dg or more \fIpieces\fR, concatenated.
+It matches a match for the first, followed by a match for the second, etc.
+.PP
+A piece is an \fIatom\fR possibly followed
+by a single\(dg `*', `+', `?', or \fIbound\fR.
+An atom followed by `*' matches a sequence of 0 or more matches of the atom.
+An atom followed by `+' matches a sequence of 1 or more matches of the atom.
+An atom followed by `?' matches a sequence of 0 or 1 matches of the atom.
+.PP
+A \fIbound\fR is `{' followed by an unsigned decimal integer,
+possibly followed by `,'
+possibly followed by another unsigned decimal integer,
+always followed by `}'.
+The integers must lie between 0 and RE_DUP_MAX (255\(dg) inclusive,
+and if there are two of them, the first may not exceed the second.
+An atom followed by a bound containing one integer \fIi\fR
+and no comma matches
+a sequence of exactly \fIi\fR matches of the atom.
+An atom followed by a bound
+containing one integer \fIi\fR and a comma matches
+a sequence of \fIi\fR or more matches of the atom.
+An atom followed by a bound
+containing two integers \fIi\fR and \fIj\fR matches
+a sequence of \fIi\fR through \fIj\fR (inclusive) matches of the atom.
+.PP
+An atom is a regular expression enclosed in `()' (matching a match for the
+regular expression),
+an empty set of `()' (matching the null string)\(dg,
+a \fIbracket expression\fR (see below), `.'
+(matching any single character), `^' (matching the null string at the
+beginning of a line), `$' (matching the null string at the
+end of a line), a `\e' followed by one of the characters
+`^.[$()|*+?{\e'
+(matching that character taken as an ordinary character),
+a `\e' followed by any other character\(dg
+(matching that character taken as an ordinary character,
+as if the `\e' had not been present\(dg),
+or a single character with no other significance (matching that character).
+A `{' followed by a character other than a digit is an ordinary
+character, not the beginning of a bound\(dg.
+It is illegal to end an RE with `\e'.
+.PP
+A \fIbracket expression\fR is a list of characters enclosed in `[]'.
+It normally matches any single character from the list (but see below).
+If the list begins with `^',
+it matches any single character
+(but see below) \fInot\fR from the rest of the list.
+If two characters in the list are separated by `\-', this is shorthand
+for the full \fIrange\fR of characters between those two (inclusive) in the
+collating sequence,
+e.g. `[0\-9]' in ASCII matches any decimal digit.
+It is illegal\(dg for two ranges to share an
+endpoint, e.g. `a\-c\-e'.
+Ranges are very collating-sequence-dependent,
+and portable programs should avoid relying on them.
+.PP
+To include a literal `]' in the list, make it the first character
+(following a possible `^').
+To include a literal `\-', make it the first or last character,
+or the second endpoint of a range.
+To use a literal `\-' as the first endpoint of a range,
+enclose it in `[.' and `.]' to make it a collating element (see below).
+With the exception of these and some combinations using `[' (see next
+paragraphs), all other special characters, including `\e', lose their
+special significance within a bracket expression.
+.PP
+Within a bracket expression, a collating element (a character,
+a multi-character sequence that collates as if it were a single character,
+or a collating-sequence name for either)
+enclosed in `[.' and `.]' stands for the
+sequence of characters of that collating element.
+The sequence is a single element of the bracket expression's list.
+A bracket expression containing a multi-character collating element 
+can thus match more than one character,
+e.g. if the collating sequence includes a `ch' collating element,
+then the RE `[[.ch.]]*c' matches the first five characters
+of `chchcc'.
+.PP
+Within a bracket expression, a collating element enclosed in `[=' and
+`=]' is an equivalence class, standing for the sequences of characters
+of all collating elements equivalent to that one, including itself.
+(If there are no other equivalent collating elements,
+the treatment is as if the enclosing delimiters were `[.' and `.]'.)
+For example, if o and \o'o^' are the members of an equivalence class,
+then `[[=o=]]', `[[=\o'o^'=]]', and `[o\o'o^']' are all synonymous.
+An equivalence class may not\(dg be an endpoint
+of a range.
+.PP
+Within a bracket expression, the name of a \fIcharacter class\fR enclosed
+in `[:' and `:]' stands for the list of all characters belonging to that
+class.
+Standard character class names are:
+.PP
+.RS
+.nf
+.ta 3c 6c 9c
+alnum  digit   punct
+alpha  graph   space
+blank  lower   upper
+cntrl  print   xdigit
+.fi
+.RE
+.PP
+These stand for the character classes defined in
+.IR ctype (3).
+A locale may provide others.
+A character class may not be used as an endpoint of a range.
+.PP
+There are two special cases\(dg of bracket expressions:
+the bracket expressions `[[:<:]]' and `[[:>:]]' match the null string at
+the beginning and end of a word respectively.
+A word is defined as a sequence of
+word characters
+which is neither preceded nor followed by
+word characters.
+A word character is an
+.I alnum
+character (as defined by
+.IR ctype (3))
+or an underscore.
+This is an extension,
+compatible with but not specified by POSIX 1003.2,
+and should be used with
+caution in software intended to be portable to other systems.
+.PP
+In the event that an RE could match more than one substring of a given
+string,
+the RE matches the one starting earliest in the string.
+If the RE could match more than one substring starting at that point,
+it matches the longest.
+Subexpressions also match the longest possible substrings, subject to
+the constraint that the whole match be as long as possible,
+with subexpressions starting earlier in the RE taking priority over
+ones starting later.
+Note that higher-level subexpressions thus take priority over
+their lower-level component subexpressions.
+.PP
+Match lengths are measured in characters, not collating elements.
+A null string is considered longer than no match at all.
+For example,
+`bb*' matches the three middle characters of `abbbc',
+`(wee|week)(knights|nights)' matches all ten characters of `weeknights',
+when `(.*).*' is matched against `abc' the parenthesized subexpression
+matches all three characters, and
+when `(a*)*' is matched against `bc' both the whole RE and the parenthesized
+subexpression match the null string.
+.PP
+If case-independent matching is specified,
+the effect is much as if all case distinctions had vanished from the
+alphabet.
+When an alphabetic that exists in multiple cases appears as an
+ordinary character outside a bracket expression, it is effectively
+transformed into a bracket expression containing both cases,
+e.g. `x' becomes `[xX]'.
+When it appears inside a bracket expression, all case counterparts
+of it are added to the bracket expression, so that (e.g.) `[x]'
+becomes `[xX]' and `[^x]' becomes `[^xX]'.
+.PP
+No particular limit is imposed on the length of REs\(dg.
+Programs intended to be portable should not employ REs longer
+than 256 bytes,
+as an implementation can refuse to accept such REs and remain
+POSIX-compliant.
+.PP
+Obsolete (``basic'') regular expressions differ in several respects.
+`|', `+', and `?' are ordinary characters and there is no equivalent
+for their functionality.
+The delimiters for bounds are `\e{' and `\e}',
+with `{' and `}' by themselves ordinary characters.
+The parentheses for nested subexpressions are `\e(' and `\e)',
+with `(' and `)' by themselves ordinary characters.
+`^' is an ordinary character except at the beginning of the
+RE or\(dg the beginning of a parenthesized subexpression,
+`$' is an ordinary character except at the end of the
+RE or\(dg the end of a parenthesized subexpression,
+and `*' is an ordinary character if it appears at the beginning of the
+RE or the beginning of a parenthesized subexpression
+(after a possible leading `^').
+Finally, there is one new type of atom, a \fIback reference\fR:
+`\e' followed by a non-zero decimal digit \fId\fR
+matches the same sequence of characters
+matched by the \fId\fRth parenthesized subexpression
+(numbering subexpressions by the positions of their opening parentheses,
+left to right),
+so that (e.g.) `\e([bc]\e)\e1' matches `bb' or `cc' but not `bc'.
+.SH SEE ALSO
+rxspencer(3)
+.PP
+POSIX 1003.2, section 2.8 (Regular Expression Notation).
+.SH HISTORY
+Written by Henry Spencer, based on the 1003.2 spec.
+.SH BUGS
+Having two kinds of REs is a botch.
+.PP
+The current 1003.2 spec says that `)' is an ordinary character in
+the absence of an unmatched `(';
+this was an unintentional result of a wording error,
+and change is likely.
+Avoid relying on it.
+.PP
+Back references are a dreadful botch,
+posing major problems for efficient implementations.
+They are also somewhat vaguely defined
+(does
+`a\e(\e(b\e)*\e2\e)*d' match `abbbd'?).
+Avoid using them.
+.PP
+1003.2's specification of case-independent matching is vague.
+The ``one case implies all cases'' definition given above
+is current consensus among implementors as to the right interpretation.
+.PP
+The syntax for word boundaries is incredibly ugly.
diff --git a/lib/rxspencer/split.c b/lib/rxspencer/split.c
new file mode 100644 (file)
index 0000000..188bdb7
--- /dev/null
@@ -0,0 +1,316 @@
+#include <stdio.h>
+#include <string.h>
+
+/*
+ - split - divide a string into fields, like awk split()
+ = int split(char *string, char *fields[], int nfields, char *sep);
+ */
+int                            /* number of fields, including overflow */
+split(string, fields, nfields, sep)
+char *string;
+char *fields[];                        /* list is not NULL-terminated */
+int nfields;                   /* number of entries available in fields[] */
+char *sep;                     /* "" white, "c" single char, "ab" [ab]+ */
+{
+       register char *p = string;
+       register char c;                        /* latest character */
+       register char sepc = sep[0];
+       register char sepc2;
+       register int fn;
+       register char **fp = fields;
+       register char *sepp;
+       register int trimtrail;
+
+       /* white space */
+       if (sepc == '\0') {
+               while ((c = *p++) == ' ' || c == '\t')
+                       continue;
+               p--;
+               trimtrail = 1;
+               sep = " \t";    /* note, code below knows this is 2 long */
+               sepc = ' ';
+       } else
+               trimtrail = 0;
+       sepc2 = sep[1];         /* now we can safely pick this up */
+
+       /* catch empties */
+       if (*p == '\0')
+               return(0);
+
+       /* single separator */
+       if (sepc2 == '\0') {
+               fn = nfields;
+               for (;;) {
+                       *fp++ = p;
+                       fn--;
+                       if (fn == 0)
+                               break;
+                       while ((c = *p++) != sepc)
+                               if (c == '\0')
+                                       return(nfields - fn);
+                       *(p-1) = '\0';
+               }
+               /* we have overflowed the fields vector -- just count them */
+               fn = nfields;
+               for (;;) {
+                       while ((c = *p++) != sepc)
+                               if (c == '\0')
+                                       return(fn);
+                       fn++;
+               }
+               /* not reached */
+       }
+
+       /* two separators */
+       if (sep[2] == '\0') {
+               fn = nfields;
+               for (;;) {
+                       *fp++ = p;
+                       fn--;
+                       while ((c = *p++) != sepc && c != sepc2)
+                               if (c == '\0') {
+                                       if (trimtrail && **(fp-1) == '\0')
+                                               fn++;
+                                       return(nfields - fn);
+                               }
+                       if (fn == 0)
+                               break;
+                       *(p-1) = '\0';
+                       while ((c = *p++) == sepc || c == sepc2)
+                               continue;
+                       p--;
+               }
+               /* we have overflowed the fields vector -- just count them */
+               fn = nfields;
+               while (c != '\0') {
+                       while ((c = *p++) == sepc || c == sepc2)
+                               continue;
+                       p--;
+                       fn++;
+                       while ((c = *p++) != '\0' && c != sepc && c != sepc2)
+                               continue;
+               }
+               /* might have to trim trailing white space */
+               if (trimtrail) {
+                       p--;
+                       while ((c = *--p) == sepc || c == sepc2)
+                               continue;
+                       p++;
+                       if (*p != '\0') {
+                               if (fn == nfields+1)
+                                       *p = '\0';
+                               fn--;
+                       }
+               }
+               return(fn);
+       }
+
+       /* n separators */
+       fn = 0;
+       for (;;) {
+               if (fn < nfields)
+                       *fp++ = p;
+               fn++;
+               for (;;) {
+                       c = *p++;
+                       if (c == '\0')
+                               return(fn);
+                       sepp = sep;
+                       while ((sepc = *sepp++) != '\0' && sepc != c)
+                               continue;
+                       if (sepc != '\0')       /* it was a separator */
+                               break;
+               }
+               if (fn < nfields)
+                       *(p-1) = '\0';
+               for (;;) {
+                       c = *p++;
+                       sepp = sep;
+                       while ((sepc = *sepp++) != '\0' && sepc != c)
+                               continue;
+                       if (sepc == '\0')       /* it wasn't a separator */
+                               break;
+               }
+               p--;
+       }
+
+       /* not reached */
+}
+
+#ifdef TEST_SPLIT
+
+
+/*
+ * test program
+ * pgm         runs regression
+ * pgm sep     splits stdin lines by sep
+ * pgm str sep splits str by sep
+ * pgm str sep n       splits str by sep n times
+ */
+int
+main(argc, argv)
+int argc;
+char *argv[];
+{
+       char buf[512];
+       register int n;
+#      define  MNF     10
+       char *fields[MNF];
+
+       if (argc > 4)
+               for (n = atoi(argv[3]); n > 0; n--) {
+                       (void) strcpy(buf, argv[1]);
+               }
+       else if (argc > 3)
+               for (n = atoi(argv[3]); n > 0; n--) {
+                       (void) strcpy(buf, argv[1]);
+                       (void) split(buf, fields, MNF, argv[2]);
+               }
+       else if (argc > 2)
+               dosplit(argv[1], argv[2]);
+       else if (argc > 1)
+               while (fgets(buf, sizeof(buf), stdin) != NULL) {
+                       buf[strlen(buf)-1] = '\0';      /* stomp newline */
+                       dosplit(buf, argv[1]);
+               }
+       else
+               regress();
+
+       exit(0);
+}
+
+dosplit(string, seps)
+char *string;
+char *seps;
+{
+#      define  NF      5
+       char *fields[NF];
+       register int nf;
+
+       nf = split(string, fields, NF, seps);
+       print(nf, NF, fields);
+}
+
+print(nf, nfp, fields)
+int nf;
+int nfp;
+char *fields[];
+{
+       register int fn;
+       register int bound;
+
+       bound = (nf > nfp) ? nfp : nf;
+       printf("%d:\t", nf);
+       for (fn = 0; fn < bound; fn++)
+               printf("\"%s\"%s", fields[fn], (fn+1 < nf) ? ", " : "\n");
+}
+
+#define        RNF     5               /* some table entries know this */
+struct {
+       char *str;
+       char *seps;
+       int nf;
+       char *fi[RNF];
+} tests[] = {
+       "",             " ",    0,      { "" },
+       " ",            " ",    2,      { "", "" },
+       "x",            " ",    1,      { "x" },
+       "xy",           " ",    1,      { "xy" },
+       "x y",          " ",    2,      { "x", "y" },
+       "abc def  g ",  " ",    5,      { "abc", "def", "", "g", "" },
+       "  a bcd",      " ",    4,      { "", "", "a", "bcd" },
+       "a b c d e f",  " ",    6,      { "a", "b", "c", "d", "e f" },
+       " a b c d ",    " ",    6,      { "", "a", "b", "c", "d " },
+
+       "",             " _",   0,      { "" },
+       " ",            " _",   2,      { "", "" },
+       "x",            " _",   1,      { "x" },
+       "x y",          " _",   2,      { "x", "y" },
+       "ab _ cd",      " _",   2,      { "ab", "cd" },
+       " a_b  c ",     " _",   5,      { "", "a", "b", "c", "" },
+       "a b c_d e f",  " _",   6,      { "a", "b", "c", "d", "e f" },
+       " a b c d ",    " _",   6,      { "", "a", "b", "c", "d " },
+
+       "",             " _~",  0,      { "" },
+       " ",            " _~",  2,      { "", "" },
+       "x",            " _~",  1,      { "x" },
+       "x y",          " _~",  2,      { "x", "y" },
+       "ab _~ cd",     " _~",  2,      { "ab", "cd" },
+       " a_b  c~",     " _~",  5,      { "", "a", "b", "c", "" },
+       "a b_c d~e f",  " _~",  6,      { "a", "b", "c", "d", "e f" },
+       "~a b c d ",    " _~",  6,      { "", "a", "b", "c", "d " },
+
+       "",             " _~-", 0,      { "" },
+       " ",            " _~-", 2,      { "", "" },
+       "x",            " _~-", 1,      { "x" },
+       "x y",          " _~-", 2,      { "x", "y" },
+       "ab _~- cd",    " _~-", 2,      { "ab", "cd" },
+       " a_b  c~",     " _~-", 5,      { "", "a", "b", "c", "" },
+       "a b_c-d~e f",  " _~-", 6,      { "a", "b", "c", "d", "e f" },
+       "~a-b c d ",    " _~-", 6,      { "", "a", "b", "c", "d " },
+
+       "",             "  ",   0,      { "" },
+       " ",            "  ",   2,      { "", "" },
+       "x",            "  ",   1,      { "x" },
+       "xy",           "  ",   1,      { "xy" },
+       "x y",          "  ",   2,      { "x", "y" },
+       "abc def  g ",  "  ",   4,      { "abc", "def", "g", "" },
+       "  a bcd",      "  ",   3,      { "", "a", "bcd" },
+       "a b c d e f",  "  ",   6,      { "a", "b", "c", "d", "e f" },
+       " a b c d ",    "  ",   6,      { "", "a", "b", "c", "d " },
+
+       "",             "",     0,      { "" },
+       " ",            "",     0,      { "" },
+       "x",            "",     1,      { "x" },
+       "xy",           "",     1,      { "xy" },
+       "x y",          "",     2,      { "x", "y" },
+       "abc def  g ",  "",     3,      { "abc", "def", "g" },
+       "\t a bcd",     "",     2,      { "a", "bcd" },
+       "  a \tb\t c ", "",     3,      { "a", "b", "c" },
+       "a b c d e ",   "",     5,      { "a", "b", "c", "d", "e" },
+       "a b\tc d e f", "",     6,      { "a", "b", "c", "d", "e f" },
+       " a b c d e f ",        "",     6,      { "a", "b", "c", "d", "e f " },
+
+       NULL,           NULL,   0,      { NULL },
+};
+
+regress()
+{
+       char buf[512];
+       register int n;
+       char *fields[RNF+1];
+       register int nf;
+       register int i;
+       register int printit;
+       register char *f;
+
+       for (n = 0; tests[n].str != NULL; n++) {
+               (void) strcpy(buf, tests[n].str);
+               fields[RNF] = NULL;
+               nf = split(buf, fields, RNF, tests[n].seps);
+               printit = 0;
+               if (nf != tests[n].nf) {
+                       printf("split `%s' by `%s' gave %d fields, not %d\n",
+                               tests[n].str, tests[n].seps, nf, tests[n].nf);
+                       printit = 1;
+               } else if (fields[RNF] != NULL) {
+                       printf("split() went beyond array end\n");
+                       printit = 1;
+               } else {
+                       for (i = 0; i < nf && i < RNF; i++) {
+                               f = fields[i];
+                               if (f == NULL)
+                                       f = "(NULL)";
+                               if (strcmp(f, tests[n].fi[i]) != 0) {
+                                       printf("split `%s' by `%s', field %d is `%s', not `%s'\n",
+                                               tests[n].str, tests[n].seps,
+                                               i, fields[i], tests[n].fi[i]);
+                                       printit = 1;
+                               }
+                       }
+               }
+               if (printit)
+                       print(nf, RNF, fields);
+       }
+}
+#endif
diff --git a/lib/rxspencer/test-driver b/lib/rxspencer/test-driver
new file mode 100755 (executable)
index 0000000..8e575b0
--- /dev/null
@@ -0,0 +1,148 @@
+#! /bin/sh
+# test-driver - basic testsuite driver script.
+
+scriptversion=2013-07-13.22; # UTC
+
+# Copyright (C) 2011-2014 Free Software Foundation, Inc.
+#
+# This program is free software; you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation; either version 2, or (at your option)
+# any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+# As a special exception to the GNU General Public License, if you
+# distribute this file as part of a program that contains a
+# configuration script generated by Autoconf, you may include it under
+# the same distribution terms that you use for the rest of that program.
+
+# This file is maintained in Automake, please report
+# bugs to <bug-automake@gnu.org> or send patches to
+# <automake-patches@gnu.org>.
+
+# Make unconditional expansion of undefined variables an error.  This
+# helps a lot in preventing typo-related bugs.
+set -u
+
+usage_error ()
+{
+  echo "$0: $*" >&2
+  print_usage >&2
+  exit 2
+}
+
+print_usage ()
+{
+  cat <<END
+Usage:
+  test-driver --test-name=NAME --log-file=PATH --trs-file=PATH
+              [--expect-failure={yes|no}] [--color-tests={yes|no}]
+              [--enable-hard-errors={yes|no}] [--]
+              TEST-SCRIPT [TEST-SCRIPT-ARGUMENTS]
+The '--test-name', '--log-file' and '--trs-file' options are mandatory.
+END
+}
+
+test_name= # Used for reporting.
+log_file=  # Where to save the output of the test script.
+trs_file=  # Where to save the metadata of the test run.
+expect_failure=no
+color_tests=no
+enable_hard_errors=yes
+while test $# -gt 0; do
+  case $1 in
+  --help) print_usage; exit $?;;
+  --version) echo "test-driver $scriptversion"; exit $?;;
+  --test-name) test_name=$2; shift;;
+  --log-file) log_file=$2; shift;;
+  --trs-file) trs_file=$2; shift;;
+  --color-tests) color_tests=$2; shift;;
+  --expect-failure) expect_failure=$2; shift;;
+  --enable-hard-errors) enable_hard_errors=$2; shift;;
+  --) shift; break;;
+  -*) usage_error "invalid option: '$1'";;
+   *) break;;
+  esac
+  shift
+done
+
+missing_opts=
+test x"$test_name" = x && missing_opts="$missing_opts --test-name"
+test x"$log_file"  = x && missing_opts="$missing_opts --log-file"
+test x"$trs_file"  = x && missing_opts="$missing_opts --trs-file"
+if test x"$missing_opts" != x; then
+  usage_error "the following mandatory options are missing:$missing_opts"
+fi
+
+if test $# -eq 0; then
+  usage_error "missing argument"
+fi
+
+if test $color_tests = yes; then
+  # Keep this in sync with 'lib/am/check.am:$(am__tty_colors)'.
+  red='\e[0;31m' # Red.
+  grn='\e[0;32m' # Green.
+  lgn='\e[1;32m' # Light green.
+  blu='\e[1;34m' # Blue.
+  mgn='\e[0;35m' # Magenta.
+  std='\e[m'     # No color.
+else
+  red= grn= lgn= blu= mgn= std=
+fi
+
+do_exit='rm -f $log_file $trs_file; (exit $st); exit $st'
+trap "st=129; $do_exit" 1
+trap "st=130; $do_exit" 2
+trap "st=141; $do_exit" 13
+trap "st=143; $do_exit" 15
+
+# Test script is run here.
+"$@" >$log_file 2>&1
+estatus=$?
+
+if test $enable_hard_errors = no && test $estatus -eq 99; then
+  tweaked_estatus=1
+else
+  tweaked_estatus=$estatus
+fi
+
+case $tweaked_estatus:$expect_failure in
+  0:yes) col=$red res=XPASS recheck=yes gcopy=yes;;
+  0:*)   col=$grn res=PASS  recheck=no  gcopy=no;;
+  77:*)  col=$blu res=SKIP  recheck=no  gcopy=yes;;
+  99:*)  col=$mgn res=ERROR recheck=yes gcopy=yes;;
+  *:yes) col=$lgn res=XFAIL recheck=no  gcopy=yes;;
+  *:*)   col=$red res=FAIL  recheck=yes gcopy=yes;;
+esac
+
+# Report the test outcome and exit status in the logs, so that one can
+# know whether the test passed or failed simply by looking at the '.log'
+# file, without the need of also peaking into the corresponding '.trs'
+# file (automake bug#11814).
+echo "$res $test_name (exit status: $estatus)" >>$log_file
+
+# Report outcome to console.
+echo "${col}${res}${std}: $test_name"
+
+# Register the test result, and other relevant metadata.
+echo ":test-result: $res" > $trs_file
+echo ":global-test-result: $res" >> $trs_file
+echo ":recheck: $recheck" >> $trs_file
+echo ":copy-in-global-log: $gcopy" >> $trs_file
+
+# Local Variables:
+# mode: shell-script
+# sh-indentation: 2
+# eval: (add-hook 'write-file-hooks 'time-stamp)
+# time-stamp-start: "scriptversion="
+# time-stamp-format: "%:y-%02m-%02d.%02H"
+# time-stamp-time-zone: "UTC"
+# time-stamp-end: "; # UTC"
+# End:
diff --git a/lib/rxspencer/test1.sh b/lib/rxspencer/test1.sh
new file mode 100755 (executable)
index 0000000..f15aadf
--- /dev/null
@@ -0,0 +1,2 @@
+#!/bin/sh
+./re < $srcdir/tests
diff --git a/lib/rxspencer/test2.sh b/lib/rxspencer/test2.sh
new file mode 100755 (executable)
index 0000000..035df56
--- /dev/null
@@ -0,0 +1,2 @@
+#!/bin/sh
+./re -el < $srcdir/tests
diff --git a/lib/rxspencer/test3.sh b/lib/rxspencer/test3.sh
new file mode 100755 (executable)
index 0000000..3d61c56
--- /dev/null
@@ -0,0 +1,2 @@
+#!/bin/sh
+./re -er < $srcdir/tests
diff --git a/lib/rxspencer/tests b/lib/rxspencer/tests
new file mode 100644 (file)
index 0000000..e4d928d
--- /dev/null
@@ -0,0 +1,477 @@
+# regular expression test set
+# Lines are at least three fields, separated by one or more tabs.  "" stands
+# for an empty field.  First field is an RE.  Second field is flags.  If
+# C flag given, regcomp() is expected to fail, and the third field is the
+# error name (minus the leading REG_).
+#
+# Otherwise it is expected to succeed, and the third field is the string to
+# try matching it against.  If there is no fourth field, the match is
+# expected to fail.  If there is a fourth field, it is the substring that
+# the RE is expected to match.  If there is a fifth field, it is a comma-
+# separated list of what the subexpressions should match, with - indicating
+# no match for that one.  In both the fourth and fifth fields, a (sub)field
+# starting with @ indicates that the (sub)expression is expected to match
+# a null string followed by the stuff after the @; this provides a way to
+# test where null strings match.  The character `N' in REs and strings
+# is newline, `S' is space, `T' is tab, `Z' is NUL.
+#
+# The full list of flags:
+#      -       placeholder, does nothing
+#      b       RE is a BRE, not an ERE
+#      &       try it as both an ERE and a BRE
+#      C       regcomp() error expected, third field is error name
+#      i       REG_ICASE
+#      m       ("mundane") REG_NOSPEC
+#      s       REG_NOSUB (not really testable)
+#      n       REG_NEWLINE
+#      ^       REG_NOTBOL
+#      $       REG_NOTEOL
+#      #       REG_STARTEND (see below)
+#      p       REG_PEND
+#
+# For REG_STARTEND, the start/end offsets are those of the substring
+# enclosed in ().
+
+# basics
+a              &       a       a
+abc            &       abc     abc
+abc|de         -       abc     abc
+a|b|c          -       abc     a
+
+# parentheses and perversions thereof
+a(b)c          -       abc     abc
+a\(b\)c                b       abc     abc
+a(             C       EPAREN
+a(             b       a(      a(
+a\(            -       a(      a(
+a\(            bC      EPAREN
+a\(b           bC      EPAREN
+a(b            C       EPAREN
+a(b            b       a(b     a(b
+# gag me with a right parenthesis -- 1003.2 goofed here (my fault, partly)
+a)             -       a)      a)
+)              -       )       )
+# end gagging (in a just world, those *should* give EPAREN)
+a)             b       a)      a)
+a\)            bC      EPAREN
+\)             bC      EPAREN
+a()b           -       ab      ab
+a\(\)b         b       ab      ab
+
+# anchoring and REG_NEWLINE
+^abc$          &       abc     abc
+a^b            -       a^b
+a^b            b       a^b     a^b
+a$b            -       a$b
+a$b            b       a$b     a$b
+^              &       abc     @abc
+$              &       abc     @
+^$             &       ""      @
+$^             -       ""      @
+\($\)\(^\)     b       ""      @
+# stop retching, those are legitimate (although disgusting)
+^^             -       ""      @
+$$             -       ""      @
+b$             &       abNc
+b$             &n      abNc    b
+^b$            &       aNbNc
+^b$            &n      aNbNc   b
+^$             &n      aNNb    @Nb
+^$             n       abc
+^$             n       abcN    @
+$^             n       aNNb    @Nb
+\($\)\(^\)     bn      aNNb    @Nb
+^^             n^      aNNb    @Nb
+$$             n       aNNb    @NN
+^a             ^       a
+a$             $       a
+^a             ^n      aNb
+^b             ^n      aNb     b
+a$             $n      bNa
+b$             $n      bNa     b
+a*(^b$)c*      -       b       b
+a*\(^b$\)c*    b       b       b
+
+# certain syntax errors and non-errors
+|              C       EMPTY
+|              b       |       |
+*              C       BADRPT
+*              b       *       *
++              C       BADRPT
+?              C       BADRPT
+""             &C      EMPTY
+()             -       abc     @abc
+\(\)           b       abc     @abc
+a||b           C       EMPTY
+|ab            C       EMPTY
+ab|            C       EMPTY
+(|a)b          C       EMPTY
+(a|)b          C       EMPTY
+(*a)           C       BADRPT
+(+a)           C       BADRPT
+(?a)           C       BADRPT
+({1}a)         C       BADRPT
+\(\{1\}a\)     bC      BADRPT
+(a|*b)         C       BADRPT
+(a|+b)         C       BADRPT
+(a|?b)         C       BADRPT
+(a|{1}b)       C       BADRPT
+^*             C       BADRPT
+^*             b       *       *
+^+             C       BADRPT
+^?             C       BADRPT
+^{1}           C       BADRPT
+^\{1\}         bC      BADRPT
+
+# metacharacters, backslashes
+a.c            &       abc     abc
+a[bc]d         &       abd     abd
+a\*c           &       a*c     a*c
+a\\b           &       a\b     a\b
+a\\\*b         &       a\*b    a\*b
+a\bc           &       abc     abc
+a\             &C      EESCAPE
+a\\bc          &       a\bc    a\bc
+\{             bC      BADRPT
+a\[b           &       a[b     a[b
+a[b            &C      EBRACK
+# trailing $ is a peculiar special case for the BRE code
+a$             &       a       a
+a$             &       a$
+a\$            &       a
+a\$            &       a$      a$
+a\\$           &       a
+a\\$           &       a$
+a\\$           &       a\$
+a\\$           &       a\      a\
+
+# back references, ugh
+a\(b\)\2c      bC      ESUBREG
+a\(b\1\)c      bC      ESUBREG
+a\(b*\)c\1d    b       abbcbbd abbcbbd bb
+a\(b*\)c\1d    b       abbcbd
+a\(b*\)c\1d    b       abbcbbbd
+^\(.\)\1       b       abc
+a\([bc]\)\1d   b       abcdabbd        abbd    b
+a\(\([bc]\)\2\)*d      b       abbccd  abbccd
+a\(\([bc]\)\2\)*d      b       abbcbd
+# actually, this next one probably ought to fail, but the spec is unclear
+a\(\(b\)*\2\)*d                b       abbbd   abbbd
+# here is a case that no NFA implementation does right
+\(ab*\)[ab]*\1 b       ababaaa ababaaa a
+# check out normal matching in the presence of back refs
+\(a\)\1bcd     b       aabcd   aabcd
+\(a\)\1bc*d    b       aabcd   aabcd
+\(a\)\1bc*d    b       aabd    aabd
+\(a\)\1bc*d    b       aabcccd aabcccd
+\(a\)\1bc*[ce]d        b       aabcccd aabcccd
+^\(a\)\1b\(c\)*cd$     b       aabcccd aabcccd
+
+# ordinary repetitions
+ab*c           &       abc     abc
+ab+c           -       abc     abc
+ab?c           -       abc     abc
+a\(*\)b                b       a*b     a*b
+a\(**\)b       b       ab      ab
+a\(***\)b      bC      BADRPT
+*a             b       *a      *a
+**a            b       a       a
+***a           bC      BADRPT
+
+# the dreaded bounded repetitions
+{              &       {       {
+{abc           &       {abc    {abc
+{1             C       BADRPT
+{1}            C       BADRPT
+a{b            &       a{b     a{b
+a{1}b          -       ab      ab
+a\{1\}b                b       ab      ab
+a{1,}b         -       ab      ab
+a\{1,\}b       b       ab      ab
+a{1,2}b                -       aab     aab
+a\{1,2\}b      b       aab     aab
+a{1            C       EBRACE
+a\{1           bC      EBRACE
+a{1a           C       EBRACE
+a\{1a          bC      EBRACE
+a{1a}          C       BADBR
+a\{1a\}                bC      BADBR
+a{,2}          -       a{,2}   a{,2}
+a\{,2\}                bC      BADBR
+a{,}           -       a{,}    a{,}
+a\{,\}         bC      BADBR
+a{1,x}         C       BADBR
+a\{1,x\}       bC      BADBR
+a{1,x          C       EBRACE
+a\{1,x         bC      EBRACE
+a{300}         C       BADBR
+a\{300\}       bC      BADBR
+a{1,0}         C       BADBR
+a\{1,0\}       bC      BADBR
+ab{0,0}c       -       abcac   ac
+ab\{0,0\}c     b       abcac   ac
+ab{0,1}c       -       abcac   abc
+ab\{0,1\}c     b       abcac   abc
+ab{0,3}c       -       abbcac  abbc
+ab\{0,3\}c     b       abbcac  abbc
+ab{1,1}c       -       acabc   abc
+ab\{1,1\}c     b       acabc   abc
+ab{1,3}c       -       acabc   abc
+ab\{1,3\}c     b       acabc   abc
+ab{2,2}c       -       abcabbc abbc
+ab\{2,2\}c     b       abcabbc abbc
+ab{2,4}c       -       abcabbc abbc
+ab\{2,4\}c     b       abcabbc abbc
+((a{1,10}){1,10}){1,10}        -       a       a       a,a
+
+# multiple repetitions
+a**            &C      BADRPT
+a++            C       BADRPT
+a??            C       BADRPT
+a*+            C       BADRPT
+a*?            C       BADRPT
+a+*            C       BADRPT
+a+?            C       BADRPT
+a?*            C       BADRPT
+a?+            C       BADRPT
+a{1}{1}                C       BADRPT
+a*{1}          C       BADRPT
+a+{1}          C       BADRPT
+a?{1}          C       BADRPT
+a{1}*          C       BADRPT
+a{1}+          C       BADRPT
+a{1}?          C       BADRPT
+a*{b}          -       a{b}    a{b}
+a\{1\}\{1\}    bC      BADRPT
+a*\{1\}                bC      BADRPT
+a\{1\}*                bC      BADRPT
+
+# brackets, and numerous perversions thereof
+a[b]c          &       abc     abc
+a[ab]c         &       abc     abc
+a[^ab]c                &       adc     adc
+a[]b]c         &       a]c     a]c
+a[[b]c         &       a[c     a[c
+a[-b]c         &       a-c     a-c
+a[^]b]c                &       adc     adc
+a[^-b]c                &       adc     adc
+a[b-]c         &       a-c     a-c
+a[b            &C      EBRACK
+a[]            &C      EBRACK
+a[1-3]c                &       a2c     a2c
+a[3-1]c                &C      ERANGE
+a[1-3-5]c      &C      ERANGE
+a[[.-.]--]c    &       a-c     a-c
+a[1-           &C      ERANGE
+a[[.           &C      EBRACK
+a[[.x          &C      EBRACK
+a[[.x.         &C      EBRACK
+a[[.x.]                &C      EBRACK
+a[[.x.]]       &       ax      ax
+a[[.x,.]]      &C      ECOLLATE
+a[[.one.]]b    &       a1b     a1b
+a[[.notdef.]]b &C      ECOLLATE
+a[[.].]]b      &       a]b     a]b
+a[[:alpha:]]c  &       abc     abc
+a[[:notdef:]]c &C      ECTYPE
+a[[:           &C      EBRACK
+a[[:alpha      &C      EBRACK
+a[[:alpha:]    &C      EBRACK
+a[[:alpha,:]   &C      ECTYPE
+a[[:]:]]b      &C      ECTYPE
+a[[:-:]]b      &C      ECTYPE
+a[[:alph:]]    &C      ECTYPE
+a[[:alphabet:]]        &C      ECTYPE
+[[:alnum:]]+   -       -%@a0X- a0X
+[[:alpha:]]+   -       -%@aX0- aX
+[[:blank:]]+   -       aSSTb   SST
+[[:cntrl:]]+   -       aNTb    NT
+[[:digit:]]+   -       a019b   019
+[[:graph:]]+   -       Sa%bS   a%b
+[[:lower:]]+   -       AabC    ab
+[[:print:]]+   -       NaSbN   aSb
+[[:punct:]]+   -       S%-&T   %-&
+[[:space:]]+   -       aSNTb   SNT
+[[:upper:]]+   -       aBCd    BC
+[[:xdigit:]]+  -       p0f3Cq  0f3C
+a[[=b=]]c      &       abc     abc
+a[[=           &C      EBRACK
+a[[=b          &C      EBRACK
+a[[=b=         &C      EBRACK
+a[[=b=]                &C      EBRACK
+a[[=b,=]]      &C      ECOLLATE
+a[[=one=]]b    &       a1b     a1b
+
+# complexities
+a(((b)))c      -       abc     abc
+a(b|(c))d      -       abd     abd
+a(b*|c)d       -       abbd    abbd
+# just gotta have one DFA-buster, of course
+a[ab]{20}      -       aaaaabaaaabaaaabaaaab   aaaaabaaaabaaaabaaaab
+# and an inline expansion in case somebody gets tricky
+a[ab][ab][ab][ab][ab][ab][ab][ab][ab][ab][ab][ab][ab][ab][ab][ab][ab][ab][ab][ab]      -       aaaaabaaaabaaaabaaaab   aaaaabaaaabaaaabaaaab
+# and in case somebody just slips in an NFA...
+a[ab][ab][ab][ab][ab][ab][ab][ab][ab][ab][ab][ab][ab][ab][ab][ab][ab][ab][ab][ab](wee|week)(knights|night)     -       aaaaabaaaabaaaabaaaabweeknights aaaaabaaaabaaaabaaaabweeknights
+# fish for anomalies as the number of states passes 32
+12345678901234567890123456789  -       a12345678901234567890123456789b 12345678901234567890123456789
+123456789012345678901234567890 -       a123456789012345678901234567890b        123456789012345678901234567890
+1234567890123456789012345678901        -       a1234567890123456789012345678901b       1234567890123456789012345678901
+12345678901234567890123456789012       -       a12345678901234567890123456789012b      12345678901234567890123456789012
+123456789012345678901234567890123      -       a123456789012345678901234567890123b     123456789012345678901234567890123
+# and one really big one, beyond any plausible word width
+1234567890123456789012345678901234567890123456789012345678901234567890 -       a1234567890123456789012345678901234567890123456789012345678901234567890b        1234567890123456789012345678901234567890123456789012345678901234567890
+# fish for problems as brackets go past 8
+[ab][cd][ef][gh][ij][kl][mn]   -       xacegikmoq      acegikm
+[ab][cd][ef][gh][ij][kl][mn][op]       -       xacegikmoq      acegikmo
+[ab][cd][ef][gh][ij][kl][mn][op][qr]   -       xacegikmoqy     acegikmoq
+[ab][cd][ef][gh][ij][kl][mn][op][q]    -       xacegikmoqy     acegikmoq
+
+# subtleties of matching
+abc            &       xabcy   abc
+a\(b\)?c\1d    b       acd
+aBc            i       Abc     Abc
+a[Bc]*d                i       abBCcd  abBCcd
+0[[:upper:]]1  &i      0a1     0a1
+0[[:lower:]]1  &i      0A1     0A1
+a[^b]c         &i      abc
+a[^b]c         &i      aBc
+a[^b]c         &i      adc     adc
+[a]b[c]                -       abc     abc
+[a]b[a]                -       aba     aba
+[abc]b[abc]    -       abc     abc
+[abc]b[abd]    -       abd     abd
+a(b?c)+d       -       accd    accd
+(wee|week)(knights|night)      -       weeknights      weeknights
+(we|wee|week|frob)(knights|night|day)  -       weeknights      weeknights
+a[bc]d         -       xyzaaabcaababdacd       abd
+a[ab]c         -       aaabc   abc
+abc            s       abc     abc
+a*             &       b       @b
+
+# Let's have some fun -- try to match a C comment.
+# first the obvious, which looks okay at first glance...
+/\*.*\*/       -       /*x*/   /*x*/
+# but...
+/\*.*\*/       -       /*x*/y/*z*/     /*x*/y/*z*/
+# okay, we must not match */ inside; try to do that...
+/\*([^*]|\*[^/])*\*/   -       /*x*/   /*x*/
+/\*([^*]|\*[^/])*\*/   -       /*x*/y/*z*/     /*x*/
+# but...
+/\*([^*]|\*[^/])*\*/   -       /*x**/y/*z*/    /*x**/y/*z*/
+# and a still fancier version, which does it right (I think)...
+/\*([^*]|\*+[^*/])*\*+/        -       /*x*/   /*x*/
+/\*([^*]|\*+[^*/])*\*+/        -       /*x*/y/*z*/     /*x*/
+/\*([^*]|\*+[^*/])*\*+/        -       /*x**/y/*z*/    /*x**/
+/\*([^*]|\*+[^*/])*\*+/        -       /*x****/y/*z*/  /*x****/
+/\*([^*]|\*+[^*/])*\*+/        -       /*x**x*/y/*z*/  /*x**x*/
+/\*([^*]|\*+[^*/])*\*+/        -       /*x***x/y/*z*/  /*x***x/y/*z*/
+
+# subexpressions
+.*             -       abc     abc     -
+a(b)(c)d       -       abcd    abcd    b,c
+a(((b)))c      -       abc     abc     b,b,b
+a(b|(c))d      -       abd     abd     b,-
+a(b*|c|e)d     -       abbd    abbd    bb
+a(b*|c|e)d     -       acd     acd     c
+a(b*|c|e)d     -       ad      ad      @d
+a(b?)c         -       abc     abc     b
+a(b?)c         -       ac      ac      @c
+a(b+)c         -       abc     abc     b
+a(b+)c         -       abbbc   abbbc   bbb
+a(b*)c         -       ac      ac      @c
+(a|ab)(bc([de]+)f|cde) -       abcdef  abcdef  a,bcdef,de
+# the regression tester only asks for 9 subexpressions
+a(b)(c)(d)(e)(f)(g)(h)(i)(j)k  -       abcdefghijk     abcdefghijk     b,c,d,e,f,g,h,i,j
+a(b)(c)(d)(e)(f)(g)(h)(i)(j)(k)l       -       abcdefghijkl    abcdefghijkl    b,c,d,e,f,g,h,i,j,k
+a([bc]?)c      -       abc     abc     b
+a([bc]?)c      -       ac      ac      @c
+a([bc]+)c      -       abc     abc     b
+a([bc]+)c      -       abcc    abcc    bc
+a([bc]+)bc     -       abcbc   abcbc   bc
+a(bb+|b)b      -       abb     abb     b
+a(bbb+|bb+|b)b -       abb     abb     b
+a(bbb+|bb+|b)b -       abbb    abbb    bb
+a(bbb+|bb+|b)bb        -       abbb    abbb    b
+(.*).*         -       abcdef  abcdef  abcdef
+(a*)*          -       bc      @b      @b
+
+# do we get the right subexpression when it is used more than once?
+a(b|c)*d       -       ad      ad      -
+a(b|c)*d       -       abcd    abcd    c
+a(b|c)+d       -       abd     abd     b
+a(b|c)+d       -       abcd    abcd    c
+a(b|c?)+d      -       ad      ad      @d
+a(b|c?)+d      -       abcd    abcd    @d
+a(b|c){0,0}d   -       ad      ad      -
+a(b|c){0,1}d   -       ad      ad      -
+a(b|c){0,1}d   -       abd     abd     b
+a(b|c){0,2}d   -       ad      ad      -
+a(b|c){0,2}d   -       abcd    abcd    c
+a(b|c){0,}d    -       ad      ad      -
+a(b|c){0,}d    -       abcd    abcd    c
+a(b|c){1,1}d   -       abd     abd     b
+a(b|c){1,1}d   -       acd     acd     c
+a(b|c){1,2}d   -       abd     abd     b
+a(b|c){1,2}d   -       abcd    abcd    c
+a(b|c){1,}d    -       abd     abd     b
+a(b|c){1,}d    -       abcd    abcd    c
+a(b|c){2,2}d   -       acbd    acbd    b
+a(b|c){2,2}d   -       abcd    abcd    c
+a(b|c){2,4}d   -       abcd    abcd    c
+a(b|c){2,4}d   -       abcbd   abcbd   b
+a(b|c){2,4}d   -       abcbcd  abcbcd  c
+a(b|c){2,}d    -       abcd    abcd    c
+a(b|c){2,}d    -       abcbd   abcbd   b
+a(b+|((c)*))+d -       abd     abd     @d,@d,-
+a(b+|((c)*))+d -       abcd    abcd    @d,@d,-
+
+# check out the STARTEND option
+[abc]          &#      a(b)c   b
+[abc]          &#      a(d)c
+[abc]          &#      a(bc)d  b
+[abc]          &#      a(dc)d  c
+.              &#      a()c
+b.*c           &#      b(bc)c  bc
+b.*            &#      b(bc)c  bc
+.*c            &#      b(bc)c  bc
+
+# plain strings, with the NOSPEC flag
+abc            m       abc     abc
+abc            m       xabcy   abc
+abc            m       xyz
+a*b            m       aba*b   a*b
+a*b            m       ab
+""             mC      EMPTY
+
+# cases involving NULs
+aZb            &       a       a
+aZb            &p      a
+aZb            &p#     (aZb)   aZb
+aZ*b           &p#     (ab)    ab
+a.b            &#      (aZb)   aZb
+a.*            &#      (aZb)c  aZb
+
+# word boundaries (ick)
+[[:<:]]a       &       a       a
+[[:<:]]a       &       ba
+[[:<:]]a       &       -a      a
+a[[:>:]]       &       a       a
+a[[:>:]]       &       ab
+a[[:>:]]       &       a-      a
+[[:<:]]a.c[[:>:]]      &       axcd-dayc-dazce-abc     abc
+[[:<:]]a.c[[:>:]]      &       axcd-dayc-dazce-abc-q   abc
+[[:<:]]a.c[[:>:]]      &       axc-dayc-dazce-abc      axc
+[[:<:]]b.c[[:>:]]      &       a_bxc-byc_d-bzc-q       bzc
+[[:<:]].x..[[:>:]]     &       y_xa_-_xb_y-_xc_-axdc   _xc_
+[[:<:]]a_b[[:>:]]      &       x_a_b
+
+# past problems, and suspected problems
+(A[1])|(A[2])|(A[3])|(A[4])|(A[5])|(A[6])|(A[7])|(A[8])|(A[9])|(A[A])  -       A1      A1
+abcdefghijklmnop       i       abcdefghijklmnop        abcdefghijklmnop
+abcdefghijklmnopqrstuv i       abcdefghijklmnopqrstuv  abcdefghijklmnopqrstuv
+(ALAK)|(ALT[AB])|(CC[123]1)|(CM[123]1)|(GAMC)|(LC[23][EO ])|(SEM[1234])|(SL[ES][12])|(SLWW)|(SLF )|(SLDT)|(VWH[12])|(WH[34][EW])|(WP1[ESN])    -       CC11    CC11
+CC[13]1|a{21}[23][EO][123][Es][12]a{15}aa[34][EW]aaaaaaa[X]a   -       CC11    CC11
+Char \([a-z0-9_]*\)\[.*        b       Char xyz[k      Char xyz[k      xyz
+a?b    -       ab      ab
+-\{0,1\}[0-9]*$        b       -5      -5
+a*a*a*a*a*a*a* &       aaaaaa  aaaaaa
diff --git a/lib/rxspencer/utils.h b/lib/rxspencer/utils.h
new file mode 100644 (file)
index 0000000..1a997ac
--- /dev/null
@@ -0,0 +1,22 @@
+/* utility definitions */
+#ifdef _POSIX2_RE_DUP_MAX
+#define        DUPMAX  _POSIX2_RE_DUP_MAX
+#else
+#define        DUPMAX  255
+#endif
+#define        INFINITY        (DUPMAX + 1)
+#define        NC              (CHAR_MAX - CHAR_MIN + 1)
+typedef unsigned char uch;
+
+/* switch off assertions (if not already off) if no REDEBUG */
+#ifndef REDEBUG
+#ifndef NDEBUG
+#define        NDEBUG  /* no assertions please */
+#endif
+#endif
+#include <assert.h>
+
+/* for old systems with bcopy() but no memmove() */
+#ifdef USEBCOPY
+#define        memmove(d, s, c)        bcopy(s, d, c)
+#endif