Re: [boost] Bug in Boost.Threads?

From: William E. Kempf (williamkempf_at_[hidden])
Date: 2002-07-12 15:55:28

----- Original Message -----
From: "Mark Rodgers" <mark.rodgers_at_[hidden]>
To: <boost_at_[hidden]>
Sent: Friday, July 12, 2002 3:30 PM
Subject: [boost] Bug in Boost.Threads?

> Bill,
>
> We've had a problem with recursive_mutex on Solaris using GCC 2.95.3 using
> Boost 1.27.0.  Occasionally we get the assert failing after the
> pthread_mutex_unlock on line 328 of recursive_mutex.cpp.  On further
> investigation we've found that m_count seems to be getting decremented
below
> zero, so is ending up as 4 billion or so.  Since the mutex in question is
> being used in a condition, we had a look at
>
>     void recursive_mutex::do_unlock(cv_state& state)
>
> and it occurs to us that we should add
>
>     m_count = 0;
>
> at the very end.  Certainly that seems to have fixed the problem.
>
> Is this the right fix, and does something similar need to be applied in
> other places, eg. recursive_try_mutex?
>
> Mark

Thanks for the bug report.  I'm a little swamped right now to actually look
at this immediately, but I'll research this over the weekend and let you
know.

Bill Kempf