Re: [Boost-testing] Re: Re: [thread] test_read_write_mutex breaksregressiontesting

From: Caleb Epstein (caleb.epstein_at_[hidden])
Date: 2005-04-26 16:24:59

On 4/26/05, Gennadiy Rozental <gennadiy.rozental_at_[hidden]> wrote:

> As you could see f.m_file_name is assigned to
> s_log_impl().m_entry_data.m_file in line 274. So the only possible
> explanation is that 2 threads entered this function simultaneously.

Here's are patches which fix the memory corruption problems of
test_mutex and test_read_write_mutex on SunPRO.  Basically they change
the code to avoid calling the Boost.Test code except in the case that
a test has actually failed.  As a result, there will be less progress
reported when --log_level=X is specified, though the checks *are*
being done.

Not sure if these changes are acceptable, but they at least prevent
memory corruption and crashes on this and possibly other platforms.

-- 
Caleb Epstein
caleb dot epstein at gmail dot com