Re: [Boost-users] [Thread] Timed join returning true before thread terminated

Subject: Re: [Boost-users] [Thread] Timed join returning true before thread terminated
From: John Rocha (jrr_at_[hidden])
Date: 2012-03-16 15:00:01

Nice! Thanks! And NUTS!!!!

It appears that 1.43 is very different from the version that the patch was 
created for. in 1.43 wait is a template, while in the patch wait is an inline. 
We probably need to move up to boost 1.47, but I doubt our QA division would 
accept that since we're suppose to be close to done.

It's nice to know that I'm not nuts!

-=John

On 3/16/2012 11:41 AM, Nathan Crookston wrote:
> Hi John,
>
> On Fri, Mar 16, 2012 at 12:00 PM, John Rocha<jrr_at_[hidden]>  wrote:
>> I followed your advice to verify where the threads where blocked. I analyzed
>> the GDB data and I found that the LU thread was blocked on a join() for
>> worker
>> thread #7 (having sent it an interrupt) and that worker thread #7 was block
>> on
>> a boost::condition_variable_any.
> This bug: https://svn.boost.org/trac/boost/ticket/4978 bit us at my
> work -- it was merged in 1.47.  SInce you're using 1.43, perhaps it's
> the same issue?
>
>> I then carefully looked at my boost thread source code, for version 1.43 of
>> boost, and I believe there is a race condition between the boost thread's
>> interrupt() and convar::wait() methods. Its a very small window, but I've
>> been
>> able to reproduce the issue with logs.
>>
>> Can you please read the following to find any flaws with my logic, or
>> confirm
>> my suspicion? Also, if this is a defect with the library, what should I do
>> next? Do I raise a defect somewhere?
> The patch was pretty trivial for us to apply, for what that's worth.
>
> HTH,
> Nate
> _______________________________________________
> Boost-users mailing list
> Boost-users_at_[hidden]
> http://lists.boost.org/mailman/listinfo.cgi/boost-users