]> granicus.if.org Git - python/log
python
24 years agoRenamed 'build_lib' command to 'build_clib':
Greg Ward [Thu, 2 Mar 2000 01:57:12 +0000 (01:57 +0000)]
Renamed 'build_lib' command to 'build_clib':
  * replaced build_lib.py with build_clib.py
  * renamed the class in build_clib.py
  * changed all references to 'build_lib' command in other command classes

24 years agoChanged '__rcsid__' to '__revision__'.
Greg Ward [Thu, 2 Mar 2000 01:49:45 +0000 (01:49 +0000)]
Changed '__rcsid__' to '__revision__'.

24 years agoIf the "build_lib" command built any C libraries, link with them all
Greg Ward [Thu, 2 Mar 2000 01:32:21 +0000 (01:32 +0000)]
If the "build_lib" command built any C libraries, link with them all
  when building extensions (uses build_lib's 'get_library_names()' method).
Ensure that the relative structure of source filenames is preserved in
  the temporary build tree, eg. foo/bar.c compiles to
  build/temp.<plat>/foo/bar.o.

24 years agoAdded command description.
Greg Ward [Thu, 2 Mar 2000 01:27:36 +0000 (01:27 +0000)]
Added command description.
Added 'build_clib' and 'build_temp' options (where to put C libraries
  and where to put temporary compiler by-products, ie. object files).
Moved the call to 'check_library_list()' from 'run()' to 'finalize_options()'
  -- that way, if we're going to crash we do so earlier, and we guarantee
  that the library list is valid before we do anything (not just run).
Disallow directory separators in library names -- the compiled library
  always goes right in 'build_clib'.
Added 'get_library_names()', so the "build_ext" command knows what
  libraries to link every extension with.

24 years agoIn the 'compile()' method: preserve the directory portion of source
Greg Ward [Thu, 2 Mar 2000 01:21:54 +0000 (01:21 +0000)]
In the 'compile()' method: preserve the directory portion of source
filenames when constructing object filenames, even if output_dir given --
eg. "foo/bar.c" will compile to "foo/bar.o" without an output_dir, and to
"temp/foo/bar.o" if output_dir is "temp".

24 years agoSecond attempt to fix the ctermid_r/tmpnam_r warnings: define USE_CTERMID_R
Greg Ward [Wed, 1 Mar 2000 21:51:56 +0000 (21:51 +0000)]
Second attempt to fix the ctermid_r/tmpnam_r warnings: define USE_CTERMID_R
and USE_TMPNAM_R at the top of the file and refer to them later; this
catches a second reference to 'tmpnam_r' that I didn't spot first time around.

24 years agoFix compiler warning: only use "_r" form of 'ctermid()' and 'tmpnam()' when
Greg Ward [Wed, 1 Mar 2000 18:59:47 +0000 (18:59 +0000)]
Fix compiler warning: only use "_r" form of 'ctermid()' and 'tmpnam()' when
building a threaded Python.

24 years agoGot rid of silly "123456789-..." lines in comments.
Guido van Rossum [Wed, 1 Mar 2000 15:06:53 +0000 (15:06 +0000)]
Got rid of silly "123456789-..." lines in comments.

24 years agoIn compile/link methods: ensure that the directory we expect to be writing to
Greg Ward [Wed, 1 Mar 2000 14:43:49 +0000 (14:43 +0000)]
In compile/link methods: ensure that the directory we expect to be writing to
exists before calling the compiler/linker.

24 years agoAdded 'mkpath()' method: convenience wrapper around 'util.mkpath()' that adds
Greg Ward [Wed, 1 Mar 2000 14:43:12 +0000 (14:43 +0000)]
Added 'mkpath()' method: convenience wrapper around 'util.mkpath()' that adds
the compiler objects 'verbose' and 'dry_run' flags.

24 years agoAdded call to 'ensure_ready()' on the command object in
Greg Ward [Wed, 1 Mar 2000 14:42:07 +0000 (14:42 +0000)]
Added call to 'ensure_ready()' on the command object in
'Distribution.find_command_obj()'.

24 years agoAdded 'get_platform()' to construct a string that identifies the current
Greg Ward [Wed, 1 Mar 2000 14:40:15 +0000 (14:40 +0000)]
Added 'get_platform()' to construct a string that identifies the current
platform, using 'os.uname()' or 'sys.platform'.

24 years agoBuild reorg:
Greg Ward [Wed, 1 Mar 2000 01:43:28 +0000 (01:43 +0000)]
Build reorg:
  * 'build_dir' -> 'build_lib', which by default takes its value
    straight from 'build_lib' in the 'build' command
  * added 'build_temp' and 'inplace' options
  * change 'build_extensions()' to put object files (compiler turds) in
    'build_temp' dir
  * complicated the name-of-extension-file shenanigans in
    'build_extensions()' to support "in-place" extension building, i.e.
    put the extension right into the source tree (handy for developers)
  * added 'get_ext_fullname()', renamed 'extension_filename()' to
    'get_ext_filename()', and tweaked the latter a bit -- all to support
    the new filename shenanigans

24 years agoBuild reorg:
Greg Ward [Wed, 1 Mar 2000 01:26:45 +0000 (01:26 +0000)]
Build reorg:
  * 'build_lib' -> 'build_purelib'
  * new 'build_lib' and 'build_temp' options
  * use 'get_platform()' to initialize 'build_platlib' and 'build_temp'

24 years agoBuild reorg: change 'build_dir' option to 'build_lib'.
Greg Ward [Wed, 1 Mar 2000 01:19:37 +0000 (01:19 +0000)]
Build reorg: change 'build_dir' option to 'build_lib'.

24 years agoPatch by Fred Gansevles (the module's original author).
Guido van Rossum [Tue, 29 Feb 2000 15:52:40 +0000 (15:52 +0000)]
Patch by Fred Gansevles (the module's original author).

This patch fixes 3 small problems.
1) If a map is used which is generated with 'makedbm -a',
   the trailing '\0' is now handled correctely.
2) The nis.maps() function skipped the first map in the output list.
3) The library '-lnsl' is added in Setup.in (needed on Linux glibc2 and
   Solaris systems. Maybe on other systems too?)

[I note that this still doesn't work when you are using NIS+ --GvR]

24 years agoMassive patch by Skip Montanaro to add ":name" to as many
Guido van Rossum [Tue, 29 Feb 2000 13:59:29 +0000 (13:59 +0000)]
Massive patch by Skip Montanaro to add ":name" to as many
PyArg_ParseTuple() format string arguments as possible.

24 years agoRemove some redundant logic from walk() -- there's no need to check
Guido van Rossum [Tue, 29 Feb 2000 13:31:16 +0000 (13:31 +0000)]
Remove some redundant logic from walk() -- there's no need to check
for "." and "..", since listdir() no longer returns those.

24 years agoCorrect type error in getopt.error handling code.
Guido van Rossum [Tue, 29 Feb 2000 13:08:44 +0000 (13:08 +0000)]
Correct type error in getopt.error handling code.

24 years agoScript by Tim Peters to discover illegal append() calls.
Guido van Rossum [Tue, 29 Feb 2000 13:05:49 +0000 (13:05 +0000)]
Script by Tim Peters to discover illegal append() calls.

24 years agoFix multi-arg append().
Guido van Rossum [Tue, 29 Feb 2000 13:00:16 +0000 (13:00 +0000)]
Fix multi-arg append().

24 years agoUpdate portability information for the spawn*() functions and related
Fred Drake [Tue, 29 Feb 2000 05:19:38 +0000 (05:19 +0000)]
Update portability information for the spawn*() functions and related
constants.  Someone will need to fill in information on the spawn*()
functions that aren't listed.

24 years agoFix a typo in a comment
Andrew M. Kuchling [Tue, 29 Feb 2000 00:10:24 +0000 (00:10 +0000)]
Fix a typo in a comment

24 years agoFix comments relating to the specific regexs used to parse section and
Fred Drake [Mon, 28 Feb 2000 23:23:55 +0000 (23:23 +0000)]
Fix comments relating to the specific regexs used to parse section and
option names; errors noted by Greg Stein <gstein@lyra.org>.

24 years agoPatch by Piers Lauder, who writes:
Guido van Rossum [Mon, 28 Feb 2000 22:37:30 +0000 (22:37 +0000)]
Patch by Piers Lauder, who writes:

This patch is re: Lucas.Dejonge@awtpl.com.au: [Python-bugs-list] imaplib -
not complying with RFC (PR#218)

Lucas de Jonge reported that the code in imaplib that detects a read-write
to read-only change doesn't comply with RFC 2060.

24 years ago(Finally!) Changes related to the ConfigParser/INI-file topics
Fred Drake [Mon, 28 Feb 2000 20:59:03 +0000 (20:59 +0000)]
(Finally!)  Changes related to the ConfigParser/INI-file topics
discussed on c.l.py last January.  Specifically:
  - more characters allowed in section & option names
  - if '=' is used to separate the option & value, the value can be
    followed by a comment of the form '\s;'

24 years agoDrop the support level for stdwin a notch -- the contents of
Guido van Rossum [Mon, 28 Feb 2000 19:52:53 +0000 (19:52 +0000)]
Drop the support level for stdwin a notch -- the contents of
lib-stdwin is no longer installed.

Increase the support level for other obsolete modules a bit: install
lib-old by default.  It still isn't in the path by default, but at
least it's easier to add to your $PYTHONPATH this way.  (This makes
sense because in 1.6 we're much more aggressive with declaring modules
obsolete.)

24 years agoFix use of list.append() with multiple arguments
Andrew M. Kuchling [Mon, 28 Feb 2000 17:25:14 +0000 (17:25 +0000)]
Fix use of list.append() with multiple arguments

24 years agoSimple changes by Gerrit Holl - move author acknowledgements out of
Guido van Rossum [Mon, 28 Feb 2000 15:12:25 +0000 (15:12 +0000)]
Simple changes by Gerrit Holl - move author acknowledgements out of
docstrings into comments.

24 years agoPatch by Mozhe Zadka, for __contains__ (overloading 'in'). This adds
Guido van Rossum [Mon, 28 Feb 2000 15:03:15 +0000 (15:03 +0000)]
Patch by Mozhe Zadka, for __contains__ (overloading 'in').  This adds
an instance method instance_contains as sq_contains.  It looks for
__contains__ and if not found falls back to previous behaviour.
Done.

24 years agoPatch by Mozhe Zadka, for __contains__ (overloading 'in'). This
Guido van Rossum [Mon, 28 Feb 2000 15:01:46 +0000 (15:01 +0000)]
Patch by Mozhe Zadka, for __contains__ (overloading 'in').  This
patches PySequence_Contains() to check for a valid sq_contains field.
More to follow.

24 years agoPatch by Mozhe Zadka, for __contains__ (overloading 'in'). This adds
Guido van Rossum [Mon, 28 Feb 2000 15:00:40 +0000 (15:00 +0000)]
Patch by Mozhe Zadka, for __contains__ (overloading 'in').  This adds
a new proc type (objobjproc), a new slot sq_contains to
PySequenceMethods, and a new flag Py_TPFLAGS_HAVE_SEQUENCE_IN to
Py_TPFLAGS_DEFAULT.  More to follow.

24 years agoPatch by Gerrit Holl to avoid doing two stat() calls in a row in walk().
Guido van Rossum [Mon, 28 Feb 2000 14:27:07 +0000 (14:27 +0000)]
Patch by Gerrit Holl to avoid doing two stat() calls in a row in walk().

24 years agoFix a typo in Barry's checkin.
Guido van Rossum [Sun, 27 Feb 2000 15:35:47 +0000 (15:35 +0000)]
Fix a typo in Barry's checkin.
Reported both by Gerrit Holl and Mark Favas.

24 years agoFixed a multi-arg append() call, discovered by Mark Favas.
Guido van Rossum [Sun, 27 Feb 2000 15:34:29 +0000 (15:34 +0000)]
Fixed a multi-arg append() call, discovered by Mark Favas.
Also removed some unnecessary backslases (inside parens).

24 years agoTwo buglet fixes. Peter Funk caught the bug in make_escapes:
Barry Warsaw [Sun, 27 Feb 2000 14:30:48 +0000 (14:30 +0000)]
Two buglet fixes.  Peter Funk caught the bug in make_escapes:

    This will fold all ISO 8859 chars from the upper half of the
    charset into the lower half, which is ...ummm.... unintened.

The second is a typo in the reference to options.escape in main().

24 years agoChanges submitted by Peter Funk (some fixes/additions by B.Warsaw) to
Barry Warsaw [Sat, 26 Feb 2000 20:56:47 +0000 (20:56 +0000)]
Changes submitted by Peter Funk (some fixes/additions by B.Warsaw) to
make pygettext more compatible with GNU xgettext, specifically:

Added -E/--escape for allowing pass-thru of iso8859-1 characters above
7 bits.

Added -o/--output option for renaming the output file from
messages.pot (there's overlap with -d/--default-domain, but GNU
xgettext has them both).

Added -p/--output-dir for specifying the output directory for
messages.pot.

Added -V/--version for printing the version number.

Added -w/--width for specifying the output page width (this is because
now pygettext, like GNU xgettext will put several locations on the
same line to cut down on vertical space).

Added -x/--exclude-file for specifying a list of strings that are not
to be extracted from the input files.

Bumped version number to 1.0

Try to import fintl and use fintl.gettext as _ if available.  Fall
back is to use identity definition of _().

Moved the escape creation to a function make_escapes() so that its
behavior can be controlled by the -E option.

__openseen(): Support the -x option.

write(): Support -w option and vertical space preserving feature.

main(): Support new options.

24 years agoThe "Distributing Python Modules" manual; currently empty (I'll get to it
Greg Ward [Sat, 26 Feb 2000 00:52:48 +0000 (00:52 +0000)]
The "Distributing Python Modules" manual; currently empty (I'll get to it
Real Soon Now(TM).

24 years agoTry to deal with pre-1.5.2 IOError exception objects.
Greg Ward [Sat, 26 Feb 2000 00:49:40 +0000 (00:49 +0000)]
Try to deal with pre-1.5.2 IOError exception objects.

24 years agoUnfinished, untested implementation of the lovely baroque installation scheme
Greg Ward [Sat, 26 Feb 2000 00:49:04 +0000 (00:49 +0000)]
Unfinished, untested implementation of the lovely baroque installation scheme
cooked up by Fred Drake and me.  Only saved for posterity (whoever posterity
is), as it is about to be ditched in favour of GvR's much simpler design.

24 years agoBetter fix for mpzmodule problem reported by Peter Funk: just use
Andrew M. Kuchling [Fri, 25 Feb 2000 22:23:31 +0000 (22:23 +0000)]
Better fix for mpzmodule problem reported by Peter Funk: just use
    mp_bits_per_limb with GMP 2.0

24 years agoChanges inspired by Randall Hooper to allow callbacks when an
Barry Warsaw [Fri, 25 Feb 2000 21:54:19 +0000 (21:54 +0000)]
Changes inspired by Randall Hooper to allow callbacks when an
OptionMenu is modified.  Somewhat rewritten and elaborated by myself.

class _setit: The constructor now takes an optional argument
`callback' and stashes this in a private variable.  If set, the
__call__() method will invoke this callback after the variable's value
has changed.  It will pass the callback the value, followed by any
args passed to __call__().

class OptionMenu: The constructor now takes keyword arguments, the
only one that's legally recognized is `command', which can be set to a
callback.  This callback is invoked when the OptionMenu value is set.
Any other keyword argument throws a TclError.

24 years agoPat Knight:
Guido van Rossum [Fri, 25 Feb 2000 19:26:31 +0000 (19:26 +0000)]
Pat Knight:

Solaris 2 has stub implementations of the POSIX thread functions such as
pthread_detach in libc. This means that configure tries to use them without
-lpthread, then the test of pthread_create fails and the configuration
falls back to the Solaris thread library. This patch moves the test for
pthread_create in -lpthread ahead of the test for pthread_detach in libc.
The patch also ensures that -lpthread is at the start of the library list
when linking, to pick up POSIX thread semantics for fork (see below).

Justification.
Use of POSIX threads on Solaris ensures that the fork() call only runs the
thread that called fork() in the child. This is desirable to prevent (for
example) parent server or database threads running in the child. Sun's
-lthread library uses a traditional fork() which replicates all the
parent's threads in the child. I find this undesirable.

Digression.
The configure.in seems to always test for -lthread even if a POSIX library
is found. I'm not enough of a configure.in wizard to decide whether this is
desirable or how to fix it. It is also irrelevant to this patch - I just
spotted it while testing.
End of Digression.

24 years agoTest case for fork1() behavior.
Guido van Rossum [Fri, 25 Feb 2000 19:25:05 +0000 (19:25 +0000)]
Test case for fork1() behavior.
Only the main thread should survive in the child after a fork().

24 years agoDelete references to _SC_AIO_LIST_MAX; it appears that that symbol was
Guido van Rossum [Fri, 25 Feb 2000 17:51:00 +0000 (17:51 +0000)]
Delete references to _SC_AIO_LIST_MAX; it appears that that symbol was
just a typo in some Linux header; the real symbol is
_SC_AIO_LISTIO_MAX.

24 years agoGerrit forgot to remove the "import string".
Guido van Rossum [Fri, 25 Feb 2000 16:34:11 +0000 (16:34 +0000)]
Gerrit forgot to remove the "import string".

24 years agoGerrit Holl's patch to move attribution from the docstring to a
Fred Drake [Fri, 25 Feb 2000 16:14:08 +0000 (16:14 +0000)]
Gerrit Holl's patch to move attribution from the docstring to a
comment.  <gerrit@nl.linux.org>

24 years agoFix a couple broken append() calls, spotted by Tim.
Guido van Rossum [Fri, 25 Feb 2000 11:48:42 +0000 (11:48 +0000)]
Fix a couple broken append() calls, spotted by Tim.

24 years agoFix a couple dozen broken ci.append(x, y, z) calls, spotted by Tim.
Guido van Rossum [Fri, 25 Feb 2000 11:46:50 +0000 (11:46 +0000)]
Fix a couple dozen broken ci.append(x, y, z) calls, spotted by Tim.

24 years agoFix a broken r.append(name, value) call, spotted by Tim.
Guido van Rossum [Fri, 25 Feb 2000 11:44:03 +0000 (11:44 +0000)]
Fix a broken r.append(name, value) call, spotted by Tim.

24 years agoOKOK, Greg's right, I should've used the :name option in the argument
Guido van Rossum [Thu, 24 Feb 2000 21:48:29 +0000 (21:48 +0000)]
OKOK, Greg's right, I should've used the :name option in the argument
format strings.

24 years agoPeter Funk:
Guido van Rossum [Thu, 24 Feb 2000 15:26:30 +0000 (15:26 +0000)]
Peter Funk:

This patch allows building the Python 'mpzmodule' under SuSE Linux
without having to install the source package of the GMP-libary.
The gmp-mparam.h seems to be an internal header file.  The patch
shouldn't hurt any other platforms.

24 years agoMade all list methods use PyArg_ParseTuple(), for more accurate
Guido van Rossum [Thu, 24 Feb 2000 15:23:03 +0000 (15:23 +0000)]
Made all list methods use PyArg_ParseTuple(), for more accurate
diagnostics.

*** INCOMPATIBLE CHANGE: This changes append(), remove(), index(), and
*** count() to require exactly one argument -- previously, multiple
*** arguments were silently assumed to be a tuple.

24 years agoMoshe Zadka:
Guido van Rossum [Thu, 24 Feb 2000 15:01:43 +0000 (15:01 +0000)]
Moshe Zadka:
Added docstrings to tkSimpleDialog.py

24 years agoFix from est@hyperreal.org: missing initialize in 'find_defaults()'.
Greg Ward [Thu, 24 Feb 2000 03:17:43 +0000 (03:17 +0000)]
Fix from est@hyperreal.org: missing initialize in 'find_defaults()'.

24 years agoMark Favas discovered this: getatime() accidentally returned the MTIME!
Guido van Rossum [Thu, 24 Feb 2000 02:26:51 +0000 (02:26 +0000)]
Mark Favas discovered this: getatime() accidentally returned the MTIME!
This fixes PR#211.

24 years agoPiers Lauder:
Guido van Rossum [Thu, 24 Feb 2000 02:24:50 +0000 (02:24 +0000)]
Piers Lauder:

A change in my last patch could, under certain circumstances,
cause a loop if the connection to the server dropped while
waiting for a command completion. I've changed the code to
re-raise the error after possible debugging output.

24 years agoAdd tests to exercise sequence operations (multiplication, indexing,
Andrew M. Kuchling [Wed, 23 Feb 2000 22:23:17 +0000 (22:23 +0000)]
Add tests to exercise sequence operations (multiplication, indexing,
    slicing) using long integers

24 years agoAllow using long integers as arguments to PyObject_GetItem(), _SetItem(),
Andrew M. Kuchling [Wed, 23 Feb 2000 22:21:50 +0000 (22:21 +0000)]
Allow using long integers as arguments to PyObject_GetItem(), _SetItem(),
    and _DelItem().
In sequence multiplication by a long, only call PyErr_Occurred() when the
    value returned is -1.

24 years agoAllow using long integers as slice indexes
Andrew M. Kuchling [Wed, 23 Feb 2000 22:18:48 +0000 (22:18 +0000)]
Allow using long integers as slice indexes

24 years agodict_has_key(): Accept only one parameter. PR#210 reported by
Fred Drake [Wed, 23 Feb 2000 15:47:16 +0000 (15:47 +0000)]
dict_has_key():  Accept only one parameter.  PR#210 reported by
                 Andreas Jung <ajung@sz-sb.de>.

24 years agoAdded entry for libtabnanny.tex.
Fred Drake [Wed, 23 Feb 2000 15:44:58 +0000 (15:44 +0000)]
Added entry for libtabnanny.tex.

24 years agoAdded dependency on libtabnanny.tex.
Fred Drake [Wed, 23 Feb 2000 15:44:34 +0000 (15:44 +0000)]
Added dependency on libtabnanny.tex.

24 years agoInitial version from Peter Funk <pf@artcom-gmbh.de>, with minor
Fred Drake [Wed, 23 Feb 2000 15:44:08 +0000 (15:44 +0000)]
Initial version from Peter Funk <pf@artcom-gmbh.de>, with minor
consistency changes and a note that the API may change in the future.

24 years agoDeleting tabpolice.py; it was superceded by tabnanny.py long ago.
Guido van Rossum [Wed, 23 Feb 2000 15:34:43 +0000 (15:34 +0000)]
Deleting tabpolice.py; it was superceded by tabnanny.py long ago.

24 years agoMoved tabnanny.py to standard library status.
Guido van Rossum [Wed, 23 Feb 2000 15:33:52 +0000 (15:33 +0000)]
Moved tabnanny.py to standard library status.

24 years agoAdded tabnanny.py, by Tim Peters, formerly from Tools/scripts, to the
Guido van Rossum [Wed, 23 Feb 2000 15:32:19 +0000 (15:32 +0000)]
Added tabnanny.py, by Tim Peters, formerly from Tools/scripts, to the
standard library.  Added some comments:

# XXX Note: this is now a standard library module.
# XXX The API needs to undergo changes however; the current code is too
# XXX script-like.  This will be addressed later.

24 years agoThis branch is for 1.6 & forward.
Fred Drake [Tue, 22 Feb 2000 18:20:03 +0000 (18:20 +0000)]
This branch is for 1.6 & forward.

24 years agoAdded a new command: Check module (Alt-F5) It does a full syntax check
Guido van Rossum [Tue, 22 Feb 2000 00:19:58 +0000 (00:19 +0000)]
Added a new command: Check module (Alt-F5) It does a full syntax check
of the current module.  It also runs the tabnanny to catch any
inconsistent tabs.

Also did a little bit of refactoring: added an errorbox() method to
simplify the display of error dialogs.

24 years agosatisfy the tabnanny (thanks to MH for noticing the problem)
Jeremy Hylton [Mon, 21 Feb 2000 22:46:00 +0000 (22:46 +0000)]
satisfy the tabnanny (thanks to MH for noticing the problem)

24 years agoWhat used to be tp_xxx4 is now tp_flags; set it to Py_TPFLAGS_DEFAULT.
Fred Drake [Mon, 21 Feb 2000 18:19:06 +0000 (18:19 +0000)]
What used to be tp_xxx4 is now tp_flags; set it to Py_TPFLAGS_DEFAULT.

24 years agoRemove comment that Guido agree's doesn't make sense:
Fred Drake [Mon, 21 Feb 2000 17:59:48 +0000 (17:59 +0000)]
Remove comment that Guido agree's doesn't make sense:
PyEval_EvalCode() is *not* a "backward compatible interface", it's the
one to use!

24 years agoScript to help locate markup that LaTeX2HTML has a problem with.
Fred Drake [Mon, 21 Feb 2000 17:20:16 +0000 (17:20 +0000)]
Script to help locate markup that LaTeX2HTML has a problem with.

24 years agoMark pointed out a buglet in his patch: i < _sys_nerr isn't strong
Guido van Rossum [Mon, 21 Feb 2000 16:50:31 +0000 (16:50 +0000)]
Mark pointed out a buglet in his patch: i < _sys_nerr isn't strong
enough, it could be negative.  Add i > 0 test.  (Not i >= 0; zero isn't
a valid error number.)

24 years agoAdded FreeMem, MaxBlock and CompactMem calls. The values returned by these are lower...
Jack Jansen [Mon, 21 Feb 2000 11:07:37 +0000 (11:07 +0000)]
Added FreeMem, MaxBlock and CompactMem calls. The values returned by these are lower bounds in the Python case (as malloc doesn't return memory to the heap) but they can be used to decide when to give low-memory warnings.

24 years agoadd TODO section
Greg Stein [Sat, 19 Feb 2000 13:36:23 +0000 (13:36 +0000)]
add TODO section

24 years agoPatch from Vladimir Marangozov <marangoz@python.inrialpes.fr>
Andrew M. Kuchling [Fri, 18 Feb 2000 19:16:45 +0000 (19:16 +0000)]
Patch from Vladimir Marangozov <marangoz@python.inrialpes.fr>
The same problem (mixed mallocs) exists for the pcre stack.
The buffers md->... are allocated via PyMem_RESIZE in grow_stack(),
while in free_stack() they are released with free() instead of
PyMem_DEL().

24 years agoPatch from Vladimir Marangozov <marangoz@python.inrialpes.fr>:
Andrew M. Kuchling [Fri, 18 Feb 2000 18:30:01 +0000 (18:30 +0000)]
Patch from Vladimir Marangozov <marangoz@python.inrialpes.fr>:
The buffers self->regex and self->regex_extra are allocated in
pcre_compile() and pcre_study() via pcre_malloc, but are released
via free() instead of pcre_free.

24 years agoreduce fsimp coupling: provide fs_imp param to ImportManager constructor,
Greg Stein [Fri, 18 Feb 2000 13:04:10 +0000 (13:04 +0000)]
reduce fsimp coupling: provide fs_imp param to ImportManager constructor,
  add clsFilesystemImporter class attribute, alter handling of suffix list
convert suffix importers to funcs rather than instances
remove backwards compat code: Importer.install and 2-tuple get_code()
  result values

24 years agoshift demo importers to importers.py (destined for Demo/ ?)
Greg Stein [Fri, 18 Feb 2000 12:03:40 +0000 (12:03 +0000)]
shift demo importers to importers.py (destined for Demo/ ?)
switch to isinstance() rather than direct type comparisons
removing chaining concept
update ImportManager.install() to take an optional namespace to install
  itself in. this will be useful for setting up rexec environments.
minor comment nits

24 years agoChanged all references to command methods 'set_default_options()' and
Greg Ward [Fri, 18 Feb 2000 00:36:20 +0000 (00:36 +0000)]
Changed all references to command methods 'set_default_options()' and
'set_final_options()' to 'initialize_options()' and 'finalize_options()'.

24 years agoRenamed 'set_default_options()' to 'initialize_options()', and
Greg Ward [Fri, 18 Feb 2000 00:35:22 +0000 (00:35 +0000)]
Renamed 'set_default_options()' to 'initialize_options()', and
'set_final_options()' to 'finalize_options()'.

24 years agoChanged references to the command class 'options' attribute to 'user_options'.
Greg Ward [Fri, 18 Feb 2000 00:26:23 +0000 (00:26 +0000)]
Changed references to the command class 'options' attribute to 'user_options'.
Related docstring changes.
Unrelated comment changes.

24 years agoRenamed all 'options' class attributes to 'user_options'.
Greg Ward [Fri, 18 Feb 2000 00:25:39 +0000 (00:25 +0000)]
Renamed all 'options' class attributes to 'user_options'.

24 years agoCommand classes are now named identically to their commands, so reflect this
Greg Ward [Fri, 18 Feb 2000 00:14:21 +0000 (00:14 +0000)]
Command classes are now named identically to their commands, so reflect this
in 'find_command_class()' method.

24 years agoRenamed all command classes so they're exactly the same as the name of the
Greg Ward [Fri, 18 Feb 2000 00:13:53 +0000 (00:13 +0000)]
Renamed all command classes so they're exactly the same as the name of the
command itself: no more of this "FooBar class for foo_bar command"
silliness.

24 years agoChanged 'dist' to 'sdist'.
Greg Ward [Fri, 18 Feb 2000 00:11:52 +0000 (00:11 +0000)]
Changed 'dist' to 'sdist'.

24 years agoThe 'sdist' command to create a source distribution. This is derived from the
Greg Ward [Thu, 17 Feb 2000 23:56:15 +0000 (23:56 +0000)]
The 'sdist' command to create a source distribution.  This is derived from the
old 'dist' command, but the code for dealing with manifests is completely
redone -- and renaming the command to 'sdist' is more symmetric with the
soon-to-exist 'bdist' command.

24 years agoThe 'dist' command is dead -- long live the 'sdist' command!
Greg Ward [Thu, 17 Feb 2000 23:54:55 +0000 (23:54 +0000)]
The 'dist' command is dead -- long live the 'sdist' command!

24 years agochanges to _lookupName
Jeremy Hylton [Thu, 17 Feb 2000 22:58:54 +0000 (22:58 +0000)]
changes to _lookupName
- removed now (happily) unused second arg
- need to verify results of [].index are correct; for building consts,
  need to have same value and same type, e.g. 2 not the same as 2L

24 years agothe previous quick hack to fix def foo((x,y)) failed on some cases
Jeremy Hylton [Thu, 17 Feb 2000 22:09:35 +0000 (22:09 +0000)]
the previous quick hack to fix def foo((x,y)) failed on some cases
(big surprise).  new solution is a little less hackish.

Code gen adds a TupleArg instance in the argument slot. The tuple arg
includes a copy of the names that it is responsble for binding.  The
PyAssembler uses this information to calculate the correct argcount.

all fix this wacky case: del (a, ((b,), c)), d
which is the same as: del a, b, c, d
(Can't wait for Guido to tell me why.)

solution uses findOp which walks a tree to find out whether it
contains OP_ASSIGN or OP_DELETE or ...

24 years agoadd varargs and kwargs flags to Lambda nodes
Jeremy Hylton [Thu, 17 Feb 2000 22:06:20 +0000 (22:06 +0000)]
add varargs and kwargs flags to Lambda nodes

24 years agosatisfy the tabnanny
Jeremy Hylton [Thu, 17 Feb 2000 17:56:29 +0000 (17:56 +0000)]
satisfy the tabnanny
fix broken references to filename var in generateXXX methods

24 years agoUpdate the description of int() to include the radix parameter;
Fred Drake [Thu, 17 Feb 2000 17:45:52 +0000 (17:45 +0000)]
Update the description of int() to include the radix parameter;
omission noted on c.l.py by Aahz Maruch.

Swapped the order of the descriptions of int() and intern() so that
int() comes first (the functions are in alphabetic order).

24 years agoTypo in a comment: "wheter" --> "whether"
Fred Drake [Thu, 17 Feb 2000 17:30:40 +0000 (17:30 +0000)]
Typo in a comment:  "wheter" --> "whether"

24 years agoPatches by Piers Lauder.
Guido van Rossum [Thu, 17 Feb 2000 17:12:39 +0000 (17:12 +0000)]
Patches by Piers Lauder.

Reasons for patches:

1st patch (15,21):
version change

2nd patch (66,72):
This is a patch I found in a Zope product release (quite by accident!).
It relaxes the conditions for matching a literal. I've looked over the
logic, and tested it, and it seems sensible.

3rd patch (117,123):
It appears the quoting matcher was too general, and that the IMAP4
protocol requires characters like ':' in commands to be unquoted.
(This is the patch already sent to Guido.)

4th patch (699,705):
Spelling correction in comment.

5th patch (753,761):
Another patch from the Zope product. It seems that some IMAP4 servers
produce unexpected responses in the middle of valid command/response
sequences. This patch ignores the unexpected responses in this
situation. (How I wish users would send me bug reports with examples!).

last 2 patches: (1015,1028) (1038,1044):
Minor improvements to test code.

24 years agoAdd primitive customization of window size and font.
Guido van Rossum [Thu, 17 Feb 2000 16:14:16 +0000 (16:14 +0000)]
Add primitive customization of window size and font.
A few alternative selections can be made by changing "if 0" to "if 1".

24 years agoPatch by Mark Hammond:
Guido van Rossum [Thu, 17 Feb 2000 15:19:15 +0000 (15:19 +0000)]
Patch by Mark Hammond:

* Changes to a recent patch by Chris Tismer to errors.c.  Chris' patch
always used FormatMessage() to get the error message passing the error code
from errno - but errno and FormatMessage use a different numbering scheme.
The main reason the patch looked OK was that ENOFILE==ERROR_FILE_NOT_FOUND -
but that is about the only shared error code :-).  The MS CRT docs tell you
to use _sys_errlist()/_sys_nerr.  My patch does also this, and adds a very
similar function specifically for win32 error codes.

24 years agoChanges by Mark Hammond related to the new WindowsError exception.
Guido van Rossum [Thu, 17 Feb 2000 15:18:10 +0000 (15:18 +0000)]
Changes by Mark Hammond related to the new WindowsError exception.

24 years agoDeclarations related to new WindowsError exception.
Guido van Rossum [Thu, 17 Feb 2000 15:17:18 +0000 (15:17 +0000)]
Declarations related to new WindowsError exception.