Fred Drake [Thu, 1 Jan 2004 07:22:46 +0000 (07:22 +0000)]
in the section "The interpreter stack":
- rearranged a bit to avoid duplicated information
- provide more complete (and hopefully less confusing) descriptions of
the return values for most of these functions
(close SF bug #563298)
Anthony Baxter [Thu, 18 Dec 2003 09:43:33 +0000 (09:43 +0000)]
Backport 1.13 through 1.15, after getting a begging letter from a certain
BDFL <wink>
The fullmodname() function chopped off the first character if the
module existed in the current directory.
Open results files, which contain binary pickles, in binary mode.
Remove fallback code that tries to read marshal data from a results
file, since this module never writes marshal data.
SF bug #855317: unittest: 5.3.9 Getting Extended Error Information
The example code did not work and could not easily be made to work.
Since the docs were already complex and the feature was not used
(it took two years for the errors to surface), we decided to dedocument
it entirely, leaving unittest cleaner than before.
Fred Drake [Fri, 5 Dec 2003 16:44:01 +0000 (16:44 +0000)]
Anthony started backporting the changes needed to remove version numbers
from the Doc/ tree; this finishes the job. Version information is now
extracted from Python's Include/patchlevel.h.
backport:
Reduce the size of Big String and Big Binary tests to 2**14 (minus one
for Big String). This should make the tests pass on Win98SE. Note
that the docs only promise lengths up to 2048. Unfortunately this no
longer tests for the segfault I was seeing earlier, but I'm confident
I've nailed that one. :-) Fixes SF 852281.
To find the curses extension as a DLL (on OS/2), we need to adjust the
library search path to include the extension directory. Without this,
the curses_panel extension can't find the curses extension/DLL, which
exports some curses symbols to it.
Guido van Rossum [Sun, 30 Nov 2003 22:07:34 +0000 (22:07 +0000)]
Backport:
Remove all uses of alloca() from this module. The alloca() return value
isn't checked, and it *is* possible that a very large alloca() call is
made, e.g. when a large registry value is being read. I don't know if
alloca() in that case returns NULL or returns a pointer pointing outside
the stack, and I don't want to know -- I've simply replaced all calls to
alloca() with either PyMem_Malloc() or PyString_FromStringAndSize(NULL,)
as appropriate, followed by a size check. This addresses SF buf 851056.
Thomas Heller [Fri, 28 Nov 2003 19:35:58 +0000 (19:35 +0000)]
See SF #848614: distutils' msvccompiler now tries to detect that MSVC6
is installed but the registry settings are incomplete because the gui
has never been run.
Jack Jansen [Thu, 27 Nov 2003 23:20:06 +0000 (23:20 +0000)]
Fix (workaround, actually) for bug #844676: deselecting "show hidden" can
cause an index error. We now select the first package if this threatens
to happen. Backport of 1.17.
Jack Jansen [Thu, 27 Nov 2003 23:12:37 +0000 (23:12 +0000)]
Package Mnager error dialogs could refer to hidden packages, which was
confusing. To be on the safe side we always show hidden packages before
showing error dialogs. Backport of 1.16.
Kurt B. Kaiser [Mon, 24 Nov 2003 02:34:01 +0000 (02:34 +0000)]
Backport:
- After an exception, run.py was not setting the exception vector. Noam
Raphael suggested correcting this so pdb's postmortem pm() would work.
IDLEfork Patch 844675
Update NEWS and include some items missed in IDLE1.0b2.
Tim Peters [Sat, 22 Nov 2003 03:46:30 +0000 (03:46 +0000)]
Indented and repaired the maze of #ifdefs setting USE_RECURSION_LIMIT.
Indented because it was incomprehensible. "Repaired" means someone
checked in a change that screwed up the multiple nesting levels, causing
USE_RECURSION_LIMIT to stop getting defined on all non-LP64 boxes other
than FreeBSD. Tried to repair that in a more-robust way. That error
in turn caused a bogus change to get checked in to test_re.py, which I
repaired earlier.
This needs fresh testing on all non-Win32 platforms (Win32 never used
USE_RECURSION_LIMIT, and still doesn't). Running the standard test_re.py
is an adequate test.
Tim Peters [Sat, 22 Nov 2003 03:25:40 +0000 (03:25 +0000)]
Reverted from rev 1.45.6.1 to rev 1.45: deep recursion is still part of
sre in 2.3, and the backport of the 2.4 version of the tests should not
have been done. It got confused because someone else checked a bad
change into _sre.c that caused the tests that are *supposed* to raise
a recursion exception to stop doing so on some (most?) platforms.
test_re passes again on Windows now. Until the bad change to _sre gets
fixed, it will fail on platforms other than Windows and FreeBSD, either
by "Test Failed", or by bad consequences of C stack overflow.