Michael,
If I build python 2.4.3 with the default compiler flags under gcc trunk
on MacOS X 10.4.7, I get the following...
Python 2.4.3 (#1, Aug 23 2006, 17:39:08)
[GCC 4.2.0 20060822 (experimental)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> import sys
>>> divmod(-sys.maxint-1, -1)
(0, -2147483648)
>>> (sys.maxint+1, 0)
(2147483648L, 0)
>>> divmod(-sys.maxint-1, -1) == (sys.maxint+1, 0)
False
If I add the -fwrapv flag to BASECFLAGS in the Makefile, I get...
Python 2.4.3 (#2, Aug 23 2006, 17:54:19)
[GCC 4.2.0 20060822 (experimental)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> import sys
>>> divmod(-sys.maxint-1, -1)
(2147483648L, 0L)
>>> (sys.maxint+1, 0)
(2147483648L, 0)
>>> divmod(-sys.maxint-1, -1) == (sys.maxint+1, 0)
True
So your analysis appears to be correct. Even more interesting is
that if I build python 2.4.3 with Apple's gcc from Xcode 2.3, the
correct results are obtained without the need to resort to the
-fwrapv flag. So either we have a regression in gcc trunk or
Apple has a patch in their branch which was never moved over
that resolves this issue. The only difference I can see between
the builds with gcc trunk and Apple's gcc is that I have to
remove the -Wno-long-double -no-cpp-precomp flags the build
with gcc trunk (because they don't exist).
Jack
ps Mike and Geoff, any comments?
