|
|
Boost Testing : |
From: David Abrahams (dave_at_[hidden])
Date: 2007-03-28 07:58:02
on Wed Mar 28 2007, Martin Wille <mw8329-AT-yahoo.com.au> wrote:
> David Abrahams wrote:
>
>> In fact, these failures look highly suspicious to me. Exit status 139
>> indicates a segmentation fault. I haven't seen these tests actually
>> /crash/ on any system or compiler before, including all viable
>> permutations 32/64-bit linux/windows gcc/msvc. There's definitely
>> something very odd going on here; I hope we can clear it up quickly.
>
> I have seen that before when stuff from different compilers or different
> Python versions gets mixed.
>
> In this case, we have both. The system's Python 2.3 was compiled with
> gcc 3.3.x, while the test's Python 2.4 was compiled with gcc 3.4.x.
> However, the test gets run with the system's Python.
Well, that's interesting, but it doesn't explain much. Compatibility
between python and the test objects we build depends only on the 'C'
ABI, and not even details like the layout of C standard library
structs like FILE. I routinely mix Pythons built by one GCC version
with tests built by another, and *never* have a problem. There must
be something more exotic about what you're doing.
-- Dave Abrahams Boost Consulting www.boost-consulting.com Don't Miss BoostCon 2007! ==> http://www.boostcon.com