[boost] Re: Regressions on Linux

From: David Abrahams (dave_at_[hidden])
Date: 2003-10-20 10:30:52

Martin Wille <mw8329_at_[hidden]> writes:

> 10 days ago, I wrote
>
>> Currently, these regressions on Linux exist (see
>> http://tinyurl.com/k36t ):
>> CRC
>>   - crc_test regresses for gcc-3.1 and gcc-3.2.3
>> Graph
>>   - graph regresses for gcc-2.95.3
>> Multi_array
>>   - slice, iterators, compare, index_bases, stl_interaction,
>>     constructors, assign, concept_checks, reshape, access,
>>     storage order regress for gcc-2.95.3-stlport-4.5.3
>> Random
>>   - random_test and random_demo regress for gcc-2.95.3
>> Some tests fail for all or most of the compilers:
>> Function
>>   - sum_avg_portable fails for all compilers
>> Math
>>   - octonion_test and quaternion_test fail for all compilers.
>> Type_traits
>>   - tricky_is_enum_test fails for all compilers
>> Test
>>   - errors_handling_test fails for all compilers
>
>
> No changes and no comments from the respective maintainers since then,
> except for Boost.Function (which uses a slightly modified test now).

Well, I don't know what broke tricky_is_enum_test, but it was working
on several compilers for me when I checked it in... and your link
seems to show it working on intel-7.1.

For me it passes vc7, vc7.1, cwpro8, intel5, intel6, intel7, intel8,
and borland.  I don't think it ever passed on GCC.

> Is this kind of report welcome on this list? Or would the
> respective maintainers prefer to be contacted off-list?

On list is fine, but it usually helps to write separate messages for
separate libraries, with subject lines including the lib name.  I
know, it's a pain...

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com