Re: [boost] Digital Mars C++ regression tests

From: Aleksey Gurtovoy (agurtovoy_at_[hidden])
Date: 2005-06-23 15:49:07

David Abrahams writes:
> Its performance with our tests is not too impressive. Only 4 libraries
> pass. 

Well:

1) At this point the results are lacking _any_ mark up whatsoever,
   which makes the summary report look more pessimistic when it
   actually is. Click through individual library results, and you'll
   see it.

2) A lot of the failures are due to a failure to build
   Boost.Test. Until this fixed, the results will look more
   pessimistic than they actually are.

3) When you first throw your code against a new compiler, there is
   always a chance of coming across a language corner that the
   standard and the compiler disagree about (unless the compiler is
   100% standard compliant, which, as we all know, not a single one
   is), and if that piece of code happens to be in a widely reusable
   library, then again, this is going to shadow the actual compiler
   standing. I suspect we have a few such cases here.

IOW, give things some time to sort themselves out.

> Is it really worth spending resources on testing this compiler?

Despite what your initial impression might be, it's a good compiler,
and in past Walter has been quite responsive to fixing Boost-related
problems. Give it a chance. People are using it, and the C++ community
needs more, not less choice.

-- 
Aleksey Gurtovoy
MetaCommunications Engineering