]> granicus.if.org Git - python/log
python
23 years agoSilence warnings about passing unsigned char** as char**.
Martin v. Löwis [Sat, 8 Sep 2001 16:23:34 +0000 (16:23 +0000)]
Silence warnings about passing unsigned char** as char**.

23 years agoOops, this file is very outdated. Removed.
Jack Jansen [Sat, 8 Sep 2001 13:57:16 +0000 (13:57 +0000)]
Oops, this file is very outdated. Removed.

23 years agoGeneralize operator.indexOf (PySequence_Index) to work with any
Tim Peters [Sat, 8 Sep 2001 04:00:12 +0000 (04:00 +0000)]
Generalize operator.indexOf (PySequence_Index) to work with any
iterable object.  I'm not sure how that got overlooked before!

Got rid of the internal _PySequence_IterContains, introduced a new
internal _PySequence_IterSearch, and rewrote all the iteration-based
"count of", "index of", and "is the object in it or not?" routines to
just call the new function.  I suppose it's slower this way, but the
code duplication was getting depressing.

23 years agoIt appears that unittest was changed to stop hoarding raw exception data,
Tim Peters [Sat, 8 Sep 2001 03:37:56 +0000 (03:37 +0000)]
It appears that unittest was changed to stop hoarding raw exception data,
saving instead a traceback string, but test_support's run_unittest was
still peeking into unittest internals and trying to pick apart unittest's
errors and failures vectors as if they contained exc_info() tuples instead
of strings.
Whatever, when a unittest-based test failed, test_support blew up.  I'm
not sure this is the right way to fix it; it simply gets me unstuck.

23 years agoThe usual post-release fiddling.
Tim Peters [Sat, 8 Sep 2001 01:25:47 +0000 (01:25 +0000)]
The usual post-release fiddling.

23 years agoPyClass_New(): put the extended Don Beaudry hook back in. When one of
Guido van Rossum [Fri, 7 Sep 2001 21:08:32 +0000 (21:08 +0000)]
PyClass_New(): put the extended Don Beaudry hook back in.  When one of
the base classes is not a classic class, and its class (the metaclass)
is callable, call the metaclass to do the deed.

One effect of this is that, when mixing classic and new-style classes
amongst the bases of a class, it doesn't matter whether the first base
class is a classic class or not: you will always get the error
"TypeError: metatype conflict among bases".  (Formerly, with a classic
class first, you'd get "TypeError: PyClass_New: base must be a class".)

Another effect is that multiple inheritance from ExtensionClass.Base,
with a classic class as the first class, transfers control to the
ExtensionClass.Base class.  This is what we need for SF #443239 (and
also for running Zope under 2.2a4, before ExtensionClass is replaced).

23 years agoPySequence_Check(), PyMapping_Check(): only return true if the
Guido van Rossum [Fri, 7 Sep 2001 20:20:11 +0000 (20:20 +0000)]
PySequence_Check(), PyMapping_Check(): only return true if the
corresponding "getitem" operation (sq_item or mp_subscript) is
implemented.  I realize that "sequence-ness" and "mapping-ness" are
poorly defined (and the tests may still be wrong for user-defined
instances, which always have both slots filled), but I believe that a
sequence that doesn't support its getitem operation should not be
considered a sequence.  All other operations are optional though.

For example, the ZODB BTree tests crashed because PySequence_Check()
returned true for a dictionary!  (In 2.2, the dictionary type has a
tp_as_sequence pointer, but the only field filled is sq_contains, so
you can write "if key in dict".)  With this fix, all standalone ZODB
tests succeed.

23 years agoPyType_IsSubtype(): test tp_flags for HAVE_CLASS bit before accessing
Guido van Rossum [Fri, 7 Sep 2001 18:52:13 +0000 (18:52 +0000)]
PyType_IsSubtype(): test tp_flags for HAVE_CLASS bit before accessing
a->tp_mro.  If a doesn't have class, it's considered a subclass only
of itself or of 'object'.

This one fix is enough to prevent the ExtensionClass test suite from
dumping core, but that doesn't say much (it's a rather small test
suite).  Also note that for ExtensionClass-defined types, a different
subclass test may be needed.  But I haven't checked whether
PyType_IsSubtype() is actually used in situations where this matters
-- probably it doesn't, since we also don't check for classic classes.

23 years agoBumping version numbers.
Barry Warsaw [Fri, 7 Sep 2001 18:23:30 +0000 (18:23 +0000)]
Bumping version numbers.

23 years agoMerging in removal of this file from branch to trunk.
Barry Warsaw [Fri, 7 Sep 2001 18:20:04 +0000 (18:20 +0000)]
Merging in removal of this file from branch to trunk.

23 years agoMerging 2.2a3 branch changes back into trunk
Barry Warsaw [Fri, 7 Sep 2001 18:13:44 +0000 (18:13 +0000)]
Merging 2.2a3 branch changes back into trunk

23 years agoPatch #438790: Add additional mappings.
Martin v. Löwis [Fri, 7 Sep 2001 16:49:12 +0000 (16:49 +0000)]
Patch #438790: Add additional mappings.
Also remove mappings that are not registered with IANA, and not extensions.

23 years agoPatch #450702: allow threads when calling into zlib, protect usage of
Martin v. Löwis [Fri, 7 Sep 2001 16:27:31 +0000 (16:27 +0000)]
Patch #450702: allow threads when calling into zlib, protect usage of
the module in multiple threads with a global lock.

23 years agoFix compiler warnings. This closes some of the #458880 problem.
Martin v. Löwis [Fri, 7 Sep 2001 16:10:00 +0000 (16:10 +0000)]
Fix compiler warnings. This closes some of the #458880 problem.

23 years ago(Slightly modified) patch by Steve Spicklemire to make Python build
Jack Jansen [Fri, 7 Sep 2001 14:25:12 +0000 (14:25 +0000)]
(Slightly modified) patch by Steve Spicklemire to make Python build
out of the box on OSX 10.1. Untested by me (except for not having adverse
effects on 10.0.4) but it looks good, for now. Eventually we should not
trigger on the darwin version but test for something, but until I have
the time to install 10.1 myself I have no clue what to test on.

It would be nice if this got in to the 2.2a3 distribution.

23 years agoFix bug #410274 - sys.prefix isn't always set.
Mark Hammond [Fri, 7 Sep 2001 14:08:01 +0000 (14:08 +0000)]
Fix bug #410274 - sys.prefix isn't always set.

If after calculating sys.path we do not have sys.prefix set, we loop over all path entries checking if one can point to our home directory.

23 years agoPatch by Mark Day to allow from __future__ imports. Looks harmless
Jack Jansen [Fri, 7 Sep 2001 10:11:31 +0000 (10:11 +0000)]
Patch by Mark Day to allow from __future__ imports. Looks harmless
enough, but may have side-effects because it preallocates a single
codeop.Compiler() to compile all statements the user enters.

Just: please review and retract/modify if necessary.

23 years agoSF bug [#458941] Looks like a unary minus bug.
Tim Peters [Fri, 7 Sep 2001 08:45:55 +0000 (08:45 +0000)]
SF bug [#458941] Looks like a unary minus bug.
com_factor():  when a unary minus is attached to a float or imaginary zero,
don't optimize the UNARY_MINUS opcode away:  the const dict can't
distinguish between +0.0 and -0.0, so ended up treating both like the
first one added to it.  Optimizing UNARY_PLUS away isn't a problem.

(BTW, I already uploaded the 2.2a3 Windows installer, and this isn't
important enough to delay the release.)

23 years agoFix typo in error reporting. This doesn't need to go into the release
Tim Peters [Fri, 7 Sep 2001 00:47:00 +0000 (00:47 +0000)]
Fix typo in error reporting.  This doesn't need to go into the release
branch (if it ever gets to the typo, the test is failing anyway).

23 years agoReverting to rev 1.2. Apparently gcc doesn't use the extended-precision
Tim Peters [Thu, 6 Sep 2001 23:00:21 +0000 (23:00 +0000)]
Reverting to rev 1.2.  Apparently gcc doesn't use the extended-precision
capabilities of the Pentium FPU, so what should have been (and were on
Windows) exact results got fuzzy.  Then it turns out test_support.fcmp()
isn't tolerant of tiny errors when *one* of the comparands is 0, but
test_complex's old check_close_real() is.  Rather than fix gcc <wink>,
easier to revert this test and revisit after the release.

23 years agoRewrite to use test_support's fine fcmp instead -- I didn't know that
Tim Peters [Thu, 6 Sep 2001 22:07:50 +0000 (22:07 +0000)]
Rewrite to use test_support's fine fcmp instead -- I didn't know that
existed when I wrote this test.

23 years agoAdded some underflow-to-0.0 long/long true division tests.
Tim Peters [Thu, 6 Sep 2001 22:03:36 +0000 (22:03 +0000)]
Added some underflow-to-0.0 long/long true division tests.

23 years agoRename 'getset' to 'property'.
Guido van Rossum [Thu, 6 Sep 2001 22:02:58 +0000 (22:02 +0000)]
Rename 'getset' to 'property'.

23 years agolong_true_divide: reliably force underflow to 0 when the denominator
Tim Peters [Thu, 6 Sep 2001 21:59:14 +0000 (21:59 +0000)]
long_true_divide:  reliably force underflow to 0 when the denominator
has more bits than the numerator than can be counted in a C int (yes,
that's unlikely, and no, I'm not adding a test case with a 2 gigabit
long).

23 years agoRename 'getset' to 'property'.
Guido van Rossum [Thu, 6 Sep 2001 21:56:42 +0000 (21:56 +0000)]
Rename 'getset' to 'property'.

23 years agoDo not rebuild html-$(RELEASE).tar every time we need to use it.
Fred Drake [Thu, 6 Sep 2001 19:35:02 +0000 (19:35 +0000)]
Do not rebuild html-$(RELEASE).tar every time we need to use it.

23 years agoBump version number.
Fred Drake [Thu, 6 Sep 2001 19:28:06 +0000 (19:28 +0000)]
Bump version number.

23 years agoUpdate link to the PyOpenGL project in the "gl" module docs.
Fred Drake [Thu, 6 Sep 2001 19:23:03 +0000 (19:23 +0000)]
Update link to the PyOpenGL project in the "gl" module docs.
This closes SF bug #459256.

23 years agoAdd missing period in docstring.
Fred Drake [Thu, 6 Sep 2001 19:13:14 +0000 (19:13 +0000)]
Add missing period in docstring.
(Steve, can you add this to the PyUnit repository as well?)

23 years agoDocument the built-in iter() function.
Fred Drake [Thu, 6 Sep 2001 19:04:29 +0000 (19:04 +0000)]
Document the built-in iter() function.

23 years agoBump version number.
Fred Drake [Thu, 6 Sep 2001 19:02:57 +0000 (19:02 +0000)]
Bump version number.

23 years agoAdded an additional link to NIST information on secure hashing.
Fred Drake [Thu, 6 Sep 2001 18:59:43 +0000 (18:59 +0000)]
Added an additional link to NIST information on secure hashing.
This closes SF bug #458785.

23 years agoClarified the interaction between string literals and continuation lines.
Fred Drake [Thu, 6 Sep 2001 18:41:15 +0000 (18:41 +0000)]
Clarified the interaction between string literals and continuation lines.
Fixes bug reported as SF bug #453728.

23 years agoMake the examples for "Default Argument Values" more presentable and
Fred Drake [Thu, 6 Sep 2001 18:21:30 +0000 (18:21 +0000)]
Make the examples for "Default Argument Values" more presentable and
less hostile to newbie use at the interactive prompt.
This is in response to SF bug #458654.

23 years agoFix parameter for PyInt_Check().
Fred Drake [Thu, 6 Sep 2001 18:06:46 +0000 (18:06 +0000)]
Fix parameter for PyInt_Check().
Add refcount information for other recently documented APIs.

23 years agoDocument the PyMethod_* type object, functions, and macros.
Fred Drake [Thu, 6 Sep 2001 17:12:44 +0000 (17:12 +0000)]
Document the PyMethod_* type object, functions, and macros.

23 years agoMinimal instructions for using the Makefile here. Also a clear
Jack Jansen [Thu, 6 Sep 2001 16:36:42 +0000 (16:36 +0000)]
Minimal instructions for using the Makefile here. Also a clear
statement that this is a proof-of-concept meant for people to
experiment with, nothing more.

23 years agoAdded targets to install the application, and to install mac-additions
Jack Jansen [Thu, 6 Sep 2001 16:33:57 +0000 (16:33 +0000)]
Added targets to install the application, and to install mac-additions
in Python.

23 years agoBump Windows buildno (installer changes were made earlier).
Tim Peters [Thu, 6 Sep 2001 16:33:17 +0000 (16:33 +0000)]
Bump Windows buildno (installer changes were made earlier).

23 years agoDocument the rule that Python.h must be included before any standard
Fred Drake [Thu, 6 Sep 2001 16:30:30 +0000 (16:30 +0000)]
Document the rule that Python.h must be included before any standard
headers.  This is the final checkin for SF bug #458768.

23 years agoDo not #include <stdio.h> since Python.h already does that.
Fred Drake [Thu, 6 Sep 2001 16:20:33 +0000 (16:20 +0000)]
Do not #include <stdio.h> since Python.h already does that.

23 years agoUse the standard argument convention for main(), and conform to the
Fred Drake [Thu, 6 Sep 2001 16:17:24 +0000 (16:17 +0000)]
Use the standard argument convention for main(), and conform to the
Python/C style guide.

23 years agoTwo small changes to the resource usage option:
Guido van Rossum [Thu, 6 Sep 2001 16:09:41 +0000 (16:09 +0000)]
Two small changes to the resource usage option:

(1) Allow multiple -u options to extend each other (and the initial
    value of use_resources passed into regrtest.main()).

(2) When a test is run stand-alone (not via regrtest.py), needed
    resources are always granted.

23 years agoAdded note of unittest.py changes that fixed bug 451309
Steve Purcell [Thu, 6 Sep 2001 16:05:17 +0000 (16:05 +0000)]
Added note of unittest.py changes that fixed bug 451309

23 years agoSeveral updates to cover omissions noted by Rich Salz.
Fred Drake [Thu, 6 Sep 2001 15:51:56 +0000 (15:51 +0000)]
Several updates to cover omissions noted by Rich Salz.
This closes SF bug #458771.

23 years agoDocument the "unixfrom" attribute of the rfc822.Message class.
Fred Drake [Thu, 6 Sep 2001 15:07:55 +0000 (15:07 +0000)]
Document the "unixfrom" attribute of the rfc822.Message class.
Based on a patch by Skip Montanaro, this closes SF bug #458885.

23 years agoAdded tests for key deletion for both Weak*Dictionary flavors.
Fred Drake [Thu, 6 Sep 2001 14:52:39 +0000 (14:52 +0000)]
Added tests for key deletion for both Weak*Dictionary flavors.
This covers regression on SF bug #458860.

23 years agoAdd __delitem__() support for WeakKeyDictionary.
Fred Drake [Thu, 6 Sep 2001 14:51:01 +0000 (14:51 +0000)]
Add __delitem__() support for WeakKeyDictionary.
This closes SF bug #458860.

23 years agoSkip instead of fail this test if the socket module has no ssl
Guido van Rossum [Thu, 6 Sep 2001 09:54:47 +0000 (09:54 +0000)]
Skip instead of fail this test if the socket module has no ssl
support.

23 years agoReport patch #416079 changes.
Martin v. Löwis [Thu, 6 Sep 2001 08:54:16 +0000 (08:54 +0000)]
Report patch #416079 changes.

23 years agoPatch #416079: fix the debug string output when receiving telnet commands.
Martin v. Löwis [Thu, 6 Sep 2001 08:51:38 +0000 (08:51 +0000)]
Patch #416079: fix the debug string output when receiving telnet commands.
added all the telnet options known to arpa/telnet.h
added all the options registered with IANA as of today
added the possibility for the user to have it's own option negotiation callback

23 years agoChanged TestResult to store only the text representation of an error.
Steve Purcell [Thu, 6 Sep 2001 08:24:40 +0000 (08:24 +0000)]
Changed TestResult to store only the text representation of an error.

This patch is similar to that proposed by Jeremy. The proposed patch altered
the interface of TestResult such that it would be passed the error
information as a string rather than an exc_info() tuple.

The implemented change leaves the interface untouched so that TestResults
are still passed the tracebacks, but stor them in stringified form for
later reporting.

Notes:
- Custom subclasses of TestResult written by users should be unaffected.
- The existing 'unittestgui.py' will still work with this module after the
  change.
- Support can later be added to pop into the debugger when an error occurs;
  this support should be added to a TestRunner rather than to TestCase itself,
  which this change will enable.

(Jeremy, Fred, Guido: Thanks for all the feedback)

23 years agoRevert parts of patch #453627, documenting the resulting test failures
Martin v. Löwis [Thu, 6 Sep 2001 08:16:17 +0000 (08:16 +0000)]
Revert parts of patch #453627, documenting the resulting test failures
instead.

23 years agoBase address updates for bug #442142 - DLL base assignments need update
Mark Hammond [Thu, 6 Sep 2001 06:42:00 +0000 (06:42 +0000)]
Base address updates for bug #442142 - DLL base assignments need update

23 years agoFirst part of fix for bug #442142 - DLL base assignments need update
Mark Hammond [Thu, 6 Sep 2001 06:39:39 +0000 (06:39 +0000)]
First part of fix for bug #442142 - DLL base assignments need update

23 years agoDubious assumptions:
Tim Peters [Thu, 6 Sep 2001 01:17:45 +0000 (01:17 +0000)]
Dubious assumptions:

1. That seeking beyond the end of a file increases the size of a file.
2. That files so extended are magically filled with null bytes.

I find no support for either in the C std, and #2 in particular turns out
not to be true on Win32 (you apparently see whatever trash happened to be
on disk).  Left #1 intact, but changed the test to check only bytes it
explicitly wrote.  Also fiddled the "expected" vs "got" failure reports
to consistently use repr (%r) -- they weren't readable otherwise.

23 years agoEnable large file support on Win32 systems.
Tim Peters [Thu, 6 Sep 2001 00:32:15 +0000 (00:32 +0000)]
Enable large file support on Win32 systems.
Curious:  the MS docs say stati64 etc are supported even on Win95, but
Win95 doesn't support a filesystem that allows partitions > 2 Gb.

test_largefile:  This was opening its test file in text mode.  I have no
idea how that worked under Win64, but it sure needs binary mode on Win98.
BTW, on Win98 test_largefile runs quickly (under a second).

23 years agoBetter error msg for 3-arg pow with a float argument.
Tim Peters [Wed, 5 Sep 2001 23:49:24 +0000 (23:49 +0000)]
Better error msg for 3-arg pow with a float argument.

23 years agoFix for bug #442374 - Modulefinder registry support broken
Mark Hammond [Wed, 5 Sep 2001 23:42:36 +0000 (23:42 +0000)]
Fix for bug #442374 - Modulefinder registry support broken

23 years agoAdd PyMethod_Function(), PyMethod_Self(), PyMethod_Class() back.
Guido van Rossum [Wed, 5 Sep 2001 22:52:50 +0000 (22:52 +0000)]
Add PyMethod_Function(), PyMethod_Self(), PyMethod_Class() back.
While not even documented, they were clearly part of the C API,
there's no great difficulty to support them, and it has the cool
effect of not requiring any changes to ExtensionClass.c.

23 years agoRework the way we try to check for libm overflow, given that C99 no longer
Tim Peters [Wed, 5 Sep 2001 22:36:56 +0000 (22:36 +0000)]
Rework the way we try to check for libm overflow, given that C99 no longer
requires that errno ever get set, and it looks like glibc is already
playing that game.  New rules:

+ Never use HUGE_VAL.  Use the new Py_HUGE_VAL instead.

+ Never believe errno.  If overflow is the only thing you're interested in,
  use the new Py_OVERFLOWED(x) macro.  If you're interested in any libm
  errors, use the new Py_SET_ERANGE_IF_OVERFLOW(x) macro, which attempts
  to set errno the way C89 said it worked.

Unfortunately, none of these are reliable, but they work on Windows and I
*expect* under glibc too.

23 years agoRudimentary makefile for building the executable to go into a
Jack Jansen [Wed, 5 Sep 2001 22:09:50 +0000 (22:09 +0000)]
Rudimentary makefile for building the executable to go into a
fullblown OSX application. It is starting to work, but building
the application bundle is still handwork, and we need a minimal
readme file too.

23 years agoChanges to make these work under OSX as the main program for a
Jack Jansen [Wed, 5 Sep 2001 22:07:52 +0000 (22:07 +0000)]
Changes to make these work under OSX as the main program for a
fullblown drag and drop application. To my surprise it is starting
to work already: Python actually executes a script dropped on it.

To be done:
- Make sure this still works in MacPython
- Don't lose argv[0] in the process
- Applet support

23 years agoPython is a Shell, not a Viewer.
Jack Jansen [Wed, 5 Sep 2001 22:04:25 +0000 (22:04 +0000)]
Python is a Shell, not a Viewer.

23 years agoLongReprTest fails on the Mac because it uses filenames with more than
Jack Jansen [Wed, 5 Sep 2001 20:08:07 +0000 (20:08 +0000)]
LongReprTest fails on the Mac because it uses filenames with more than
32 characters per component. This makes mkdir() calls and such fail with EINVAL.

For now I am disabling the test on the Mac, and I'll open a bugreport.

23 years agoUse the build directory by default, and update the version.
Guido van Rossum [Wed, 5 Sep 2001 19:51:08 +0000 (19:51 +0000)]
Use the build directory by default, and update the version.

23 years agoClass FieldStorage: add two new methods, getfirst() and getlist(),
Guido van Rossum [Wed, 5 Sep 2001 19:45:34 +0000 (19:45 +0000)]
Class FieldStorage: add two new methods, getfirst() and getlist(),
that provide a somewhat more uniform interface to getting values.

This is from SF patch #453691.

23 years agoclass Listbox: add itemcget, to satisfy SF patch #457713.
Guido van Rossum [Wed, 5 Sep 2001 19:29:56 +0000 (19:29 +0000)]
class Listbox: add itemcget, to satisfy SF patch #457713.
Fix up docstring for itemconfigure.

23 years agotypo...
Skip Montanaro [Wed, 5 Sep 2001 19:27:13 +0000 (19:27 +0000)]
typo...

23 years agoSF bug #427073: DLINCLDIR defined incorrectly (Skip Montanaro).
Guido van Rossum [Wed, 5 Sep 2001 19:13:16 +0000 (19:13 +0000)]
SF bug #427073: DLINCLDIR defined incorrectly (Skip Montanaro).

I don't know what difference it makes, but '/' indeed makes less sense
as an include dir than '.', so I'm changing the default.  Just so I
can close the bug. ;-)

23 years agoChange the date field to use $Date$ so it won't be outrageously out of
Guido van Rossum [Wed, 5 Sep 2001 18:57:51 +0000 (18:57 +0000)]
Change the date field to use $Date$ so it won't be outrageously out of
date.

23 years agoDocument -Q. Move arguments around to be in strict alphabetical
Guido van Rossum [Wed, 5 Sep 2001 18:55:34 +0000 (18:55 +0000)]
Document -Q.  Move arguments around to be in strict alphabetical
order.  Add breaks in SYNOPSIS.

23 years agoDescribe -E (which was added to 2.2a2).
Guido van Rossum [Wed, 5 Sep 2001 18:43:35 +0000 (18:43 +0000)]
Describe -E (which was added to 2.2a2).

23 years ago[ #458701 ] Patch to zipfile.py for Java
Finn Bock [Wed, 5 Sep 2001 18:40:33 +0000 (18:40 +0000)]
[ #458701 ] Patch to zipfile.py for Java

Patch by Jim Ahlstrom which lets java's zipfile classes read zipfiles
create by zipfile.py.

23 years agoRemove a debug print left in the code by Fred.
Guido van Rossum [Wed, 5 Sep 2001 17:52:31 +0000 (17:52 +0000)]
Remove a debug print left in the code by Fred.

23 years agoPatch #449815: Set filesystemencoding based on CODESET.
Martin v. Löwis [Wed, 5 Sep 2001 17:09:48 +0000 (17:09 +0000)]
Patch #449815: Set filesystemencoding based on CODESET.

23 years agoA few more gcc warnings bite the dust.
Jack Jansen [Wed, 5 Sep 2001 15:44:37 +0000 (15:44 +0000)]
A few more gcc warnings bite the dust.

23 years agoMove UnixWare 7 defines to acconfig.h, regenerate pyconfig.h.in.
Martin v. Löwis [Wed, 5 Sep 2001 15:18:00 +0000 (15:18 +0000)]
Move UnixWare 7 defines to acconfig.h, regenerate pyconfig.h.in.

23 years agoChanges to automatically enable large file support on some systems.
Guido van Rossum [Wed, 5 Sep 2001 14:58:11 +0000 (14:58 +0000)]
Changes to automatically enable large file support on some systems.
I believe this works on Linux (tested both on a system with large file
support and one without it), and it may work on Solaris 2.7.

The changes are twofold:

(1) The configure script now boldly tries to set the two symbols that
    are recommended (for Solaris and Linux), and then tries a test
    script that does some simple seeking without writing.

(2) The _portable_{fseek,ftell} functions are a little more systematic
    in how they try the different large file support options: first
    try fseeko/ftello, but only if off_t is large; then try
    fseek64/ftell64; then try hacking with fgetpos/fsetpos.

I'm keeping my fingers crossed.  The meaning of the
HAVE_LARGEFILE_SUPPORT macro is not at all clear.

I'll see if I can get it to work on Windows as well.

23 years agoNote some changes that I need to write about
Andrew M. Kuchling [Wed, 5 Sep 2001 14:53:31 +0000 (14:53 +0000)]
Note some changes that I need to write about

23 years agoPatch #453627: Define the following macros when compiling on a UnixWare 7.x system:
Martin v. Löwis [Wed, 5 Sep 2001 14:45:54 +0000 (14:45 +0000)]
Patch #453627: Define the following macros when compiling on a UnixWare 7.x system:
SCO_ATAN2_BUG, SCO_ACCEPT_BUG, and STRICT_SYSV_CURSES.
Work aroudn a bug in the SCO UnixWare atan2() implementation.

23 years agoPatch #453627: Adds a list of tests that are expected to be skipped for UnixWare...
Martin v. Löwis [Wed, 5 Sep 2001 14:38:48 +0000 (14:38 +0000)]
Patch #453627: Adds a list of tests that are expected to be skipped for UnixWare 7.x systems.

23 years agoPatch #455231: Support ELF properly on OpenBSD.
Martin v. Löwis [Wed, 5 Sep 2001 14:24:43 +0000 (14:24 +0000)]
Patch #455231: Support ELF properly on OpenBSD.

23 years agoPatch #428326: New class threading.Timer.
Martin v. Löwis [Wed, 5 Sep 2001 13:44:54 +0000 (13:44 +0000)]
Patch #428326: New class threading.Timer.

23 years agobuiltin_execfile(): initialize another local that the GCC on leroy
Guido van Rossum [Wed, 5 Sep 2001 13:37:47 +0000 (13:37 +0000)]
builtin_execfile(): initialize another local that the GCC on leroy
found it necessary to warn about.

23 years agoImplement PEP250: Use Lib/site-packages under windows.
Thomas Heller [Wed, 5 Sep 2001 13:00:40 +0000 (13:00 +0000)]
Implement PEP250: Use Lib/site-packages under windows.
bdist_wininst doesn't use the NT SCHEME any more, instead
a custom SCHEME is used, which is exchanged at installation
time, depending on the python version used.

Avoid a bogus warning frpom install_lib about installing
into a directory not on sys.path.

23 years ago[Bug #404274] Restore some special-case code for AIX and BeOS under 1.5.2.
Andrew M. Kuchling [Wed, 5 Sep 2001 12:02:59 +0000 (12:02 +0000)]
[Bug #404274] Restore some special-case code for AIX and BeOS under 1.5.2.
   This will have to stay until we decide to drop 1.5.2 compatibility
   completely.

23 years agoShut up many more gcc warnings.
Jack Jansen [Wed, 5 Sep 2001 10:31:52 +0000 (10:31 +0000)]
Shut up many more gcc warnings.

23 years agoOnly output the buffer size error label if it is used.
Jack Jansen [Wed, 5 Sep 2001 10:27:53 +0000 (10:27 +0000)]
Only output the buffer size error label if it is used.
Shuts up another couple of gcc warnings.

23 years agoUse -fPIC instead of -fpic for gcc on HP/UX. Fixes bug #433234.
Martin v. Löwis [Wed, 5 Sep 2001 08:36:52 +0000 (08:36 +0000)]
Use -fPIC instead of -fpic for gcc on HP/UX. Fixes bug #433234.

23 years agoCheck for RFC 2553 API. Fixes bug #454493.
Martin v. Löwis [Wed, 5 Sep 2001 08:22:34 +0000 (08:22 +0000)]
Check for RFC 2553 API. Fixes bug #454493.

23 years agoMake the error msgs in our pow() implementations consistent.
Tim Peters [Wed, 5 Sep 2001 06:24:58 +0000 (06:24 +0000)]
Make the error msgs in our pow() implementations consistent.

23 years agoRepair indentation.
Tim Peters [Wed, 5 Sep 2001 06:24:24 +0000 (06:24 +0000)]
Repair indentation.

23 years agoTry to recover from that glibc's ldexp apparently doesn't set errno on
Tim Peters [Wed, 5 Sep 2001 05:38:10 +0000 (05:38 +0000)]
Try to recover from that glibc's ldexp apparently doesn't set errno on
overflow.  Needs testing on Linux (test_long.py and test_long_future.py
especially).

23 years agologhelper(): Try to nudge the compiler into doing mults in an order that
Tim Peters [Wed, 5 Sep 2001 04:33:11 +0000 (04:33 +0000)]
loghelper():  Try to nudge the compiler into doing mults in an order that
minimizes roundoff error.

23 years agoAnother / that should be a // (previously not caught because of
Guido van Rossum [Wed, 5 Sep 2001 02:27:04 +0000 (02:27 +0000)]
Another / that should be a // (previously not caught because of
incomplete coverage of the test suite).

23 years agoAdd a test for the final branch in repr.Repr.repr1(), which deals with
Guido van Rossum [Wed, 5 Sep 2001 02:26:26 +0000 (02:26 +0000)]
Add a test for the final branch in repr.Repr.repr1(), which deals with
a default repr() that's longer than 20 characters.

23 years agoReturn reasonable results for math.log(long) and math.log10(long) (we were
Tim Peters [Wed, 5 Sep 2001 00:53:45 +0000 (00:53 +0000)]
Return reasonable results for math.log(long) and math.log10(long) (we were
getting Infs, NaNs, or nonsense in 2.1 and before; in yesterday's CVS we
were getting OverflowError; but these functions always make good sense
for positive arguments, no matter how large).

23 years agoMechanical fiddling to make this easier to work with in my editor.
Tim Peters [Tue, 4 Sep 2001 23:17:42 +0000 (23:17 +0000)]
Mechanical fiddling to make this easier to work with in my editor.
Repaired the ldexp docstring (said the name of the func was "ldexp_doc").

23 years agoAdded prototypes to shut gcc -Wstrict-prototypes up.
Jack Jansen [Tue, 4 Sep 2001 22:29:31 +0000 (22:29 +0000)]
Added prototypes to shut gcc -Wstrict-prototypes up.