Re: [Opalvoip-devel] H323 New Incoming Connection Deadlock

[Robert J. <ro...@vo...>]

I have checked in the removal of the pthread_exit(), I hope this allows you
to move forward. There is more than just new features in later versions.
Usually there are many interoperability issues fixed as well.


Robert Jongbloed
OPAL/OpenH323/PTLib Architect and Co-founder.


> -----Original Message-----
> From: Jam...@gd... [mailto:Jam...@gd...]
> Sent: Friday, 25 May 2012 1:19 AM
> To: ro...@vo...; opa...@li...
> Subject: RE: [Opalvoip-devel] H323 New Incoming Connection Deadlock
> 
> Haha, I like your comment about spare time.
> 
> I should have also mentioned that when the application crashes, the
> console outputs "terminate called without an active exception".
> Anyways, I did what you suggested and removed the call to
> pthread_exit.
> This worked!
> 
> I then took a look back to compare the latest stable version of
> tlibthrd.cxx with the one found in 2.6.2 since I did not see this
> problem in Opal 3.6.2/Ptlib 2.6.2.  I saw that the only change in
> PX_ThreadStart was for tracing.  Reverting the changes in this one
> function also fixed the issue.  I then narrowed it down to the calling
> of PTimer::Tick(). If this call is omitted, then everything works as
> well.
> 
> I am happy with any of these changes as they are much better than the
> horrid sleep fix I put in.  Perhaps I just changed the timing by
> removing this code, but we have been pretty stable with Ptlib
> 2.6.2/Opal
> 3.6.2 since removing the extra PProcess instance I stupidly
> overlooked.
> I think we'll be in good shape if we just revert the tracing changes.
> Can you explain any of this or do you think the problem is still
> lurking?
> 
> Thanks,
> James