Re: [boost] Re: Boost 1.31 release?

From: Martin Wille (mw8329_at_[hidden])
Date: 2003-08-10 15:11:31

Beman Dawes wrote:

> Assuming I'm release manager for 1.31.0, I'm going to publish explicit 
> release criteria for key platform/compiler pairs. Basically, the release 
> criteria will be 100% accounting for all failures on those 
> platform/compiler pairs.
> 
> I assume that Linux/GCC will be one of those platform/compiler pairs. I 
> need input from Linux/GCC experts as to which version of GCC should be 
> used for the release criteria.
> 
> I see GCC 3.3.1 has just been released, and so will be switching the 
> Windows/GCC tests to use that version. Unless anyone objects, it will be 
> one of the Windows release criteria compilers.

I added gcc-3.3.1 to the Linux tests for CVS HEAD.

Test failures have been down to 1% for gcc versions 3.2.3 and 3.3 a
few weeks ago. I think 3.2.3 and 3.3.1 would be good candidates for
being release criteria.
6 tests fail for 3.2.3 and 6 tests fail for 3.3.1

 From the list I recently posted 1 failure has been fixed and
1 failure is due to a compiler error and not considered harmful.

That leaves

     function::sum_avg_portable
     math::octonion_test
     math::quaternion_test
     test::errors_handling_test

to be fixed or to be explained. Note, all these tests fail for
both CVS HEAD and RC_1_30_0.

Regards,
m

PS: The one additional error gcc 3.2.3 produces over 3.3.1 is in
     crc_test, btw.