[boost] Re: Remaining GCC failures

From: David Abrahams (dave_at_[hidden])
Date: 2004-01-04 08:34:34

"Gennadiy Rozental" <gennadiy.rozental_at_[hidden]> writes:

>> It wouldn't surprise me if the errors_handling_test and regex failures are
>> a Boost problem with Boost.Test, or at least a problems that are being
>> worsened by the Boost.Test problem. But Gennadiy is (1) on vacation, and
>> (2)  says he needs help with gcc, which he isn't familiar with.
>
> During lifetime of the errors_handling_test test it was reported to fail 3
> different ways
>
> 1. runtime failures
>   I was able to make it work for most windows compilers. I should be able to
> workround it for unix either, but I need pattern file produced by this test
> if you specify option --save_pattern=yes
> 2. Uses a lot of CPU and memory and hungs
>   That is what happening how in my local version with gcc and cygwin. My
> investigation led to that message:
> http://lists.boost.org/MailArchives/boost/msg28497.php
>
> I still  do not know proper resolution.
> 3. Crash
>   In my experiment in above message I was able to reproduce crash in some
> cases. Still no resolution.

It seems to me that the only responsible thing to do with this is to
take a conservative approach which doesn't try to do anything so fancy
on platforms where the test hasn't been proven to work reliably.

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