Boost Testing :
From: Jim Douglas (jim_at_[hidden])
Date: 2006-02-24 04:27:30
Gennadiy Rozental wrote:
>>I carried out further investigation and it looks like the problem may be
>>in test rather than graph. If I put some trace cout statements into
>>test_main() in bidir_vec_remove_edge.cpp and run the test I get nothing
>>on the screen.
I was misleading myself - when you run the test with bjam the std output
is redirected to a file... :~(
> There is nothing particularly complex but there is a lot going on before
> test_main invokcation. But I wouldn't expect any of it to hang. You could
> see Boost.Test sources for more detailes.
The test *is* executing and it causes a SIGSEGV in xcopy. The signal is
trapped in execution_monitor.ipp which then throws an exception. At this
point the process blocks against a mutex somewhere in the system. This
only happens with the Dinkumware library and not with the GNU library.
It appears that the monitored/timeout facility built into regression.py
is not sophisticated enough to catch a stalled process(?) Is there any
way the in bidir_vec_remove_edge test can be temporarily bypassed so
that I can resume regression tests?