On Jul 1, 2011, at 1:58 PM, Roy Stogner wrote:

> 
> On Fri, 1 Jul 2011, Derek Gaston wrote:
> 
>> Yes.... periodic BCs have never worked with threading... they deadlock.
> 
> The disgusting thing about threading bugs is that there's no such
> thing as as "never worked".  We do test one periodic BC example, and
> it gets run with 2 threads at least 10 times a day, and I've never
> seen it fail.
> 
> Hmm... that's probably partly a Heisenbug caused by race conditions
> being less likely in slow debug-mode compilations.  Maybe I need to
> add an opt-mode run to our tests; my own opt-mode tests are always
> with one thread.

Interesting... we actually usually see the deadlock in debug mode ;-)

>> We've just never got around to fixing it.  Glad you tracked it down.
> 
> Well, let's hope that that it is it.  I'll ping you when there's a fix
> in so you can test and see if there are any deadlocks still triggered
> by your stuff.

Cool - thanks!

Derek

------------------------------------------------------------------------------
All of the data generated in your IT infrastructure is seriously valuable.
Why? It contains a definitive record of application performance, security 
threats, fraudulent activity, and more. Splunk takes this data and makes 
sense of it. IT sense. And common sense.
http://p.sf.net/sfu/splunk-d2d-c2
_______________________________________________
Libmesh-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/libmesh-devel

Reply via email to