Re: [Boost-testing] [graph][QNX/QCC][ test] New bidir* regression testshang my PC

From: Gennadiy Rozental (gennadiy.rozental_at_[hidden])
Date: 2006-02-23 09:36:01

>I have not been able to complete a regression test run since Monday when
> the two new Boost.graph bidir* tests were introduced. Under QNX my PC
> hangs during the 'bidir_vec_remove_edge' test.
>
> The system diagnostics tell me that the process is blocked against a
> mutex, but I am unable to ascertain which other process/thread also
> shares this mutex. Does this test/library explicitly use mutexes?

Boost.Test used shared_ptr which may use mutexes. Other then that no.

> I have tried to use regression.py with the the options "--monitored
> --timeout=15" but this does not cancel the test. What mechanism is used
> to cancel a wayward test?
>
> If I attach gdb to the process, the last point in the source code
> appears to be in "boost::execution_monitor::catch_signals()" before it
> (possibly) enters a system library routine.

Could you run the test with detailed log (--log_level=all) to see where it 
hangs?

> I also found that I could not kill the process with a SIGABRT (to force
> a core dump). I could only use SIGKILL, at which point the regression
> test sequence resumed.
>
> Does anyone have any ideas?
>
> Jim

Gennadiy