I just spent some time tracking down a pretty sneaky issue I had, and am
curious of the official policy with QC.
Is it safe to, on multiple threads, concurrently load QCCompositions from disk,
and init QCRenderers?
I saw what looked like machine specific deadlocks* in QC when I had a few
render threads all attempt to init multiple QC Renderers at the same time. Only
when I specifically serialized initialization did the deadlock go away. I would
never get past initting a QCRenderer, and the specific threads loading
QCRenderers would sit deadlocked on
830 -[QCComposition _makePatch]
830 QCPatchFromCompositionWithOptions
830 _PatchFromComposition
830 -[QCPatch(Override) setState:]
830 -[GFGraph setState:]
830 -[GFGraph(Private) _loadState:]
830 -[QCRenderInImage setState:]
830 -[QCPatch(Override) setState:]
830 -[GFGraph setState:]
830 -[GFGraph(Private) _loadState:]
830 +[GFNodeManager
instantiateNodeWithClassName:identifier:]
830 -[QCNodeManager
instantiateNodeWithName:]
830 -[GFNodeManager
instantiateNodeWithName:]
830 pthread_mutex_lock
830 semaphore_wait_signal_trap
(in multiple threads).
I could never re-create this lock on my two development machines, but a slower
Mac Mini was able to deadlock every time.
Is it safe to init concurrently like this? While i've "solved" my issue via
dispatch_semaphore (so useful, thanks again Chris for the pointer), its not
clear this should be a problem. More curious than anything else, and maybe a
warning for those looking to do the same or running into similar issues.
Thank you.
* of course, it could just be that I never got the timing just right on my test
systems... _______________________________________________
Do not post admin requests to the list. They will be ignored.
Quartzcomposer-dev mailing list ([email protected])
Help/Unsubscribe/Update your Subscription:
http://lists.apple.com/mailman/options/quartzcomposer-dev/archive%40mail-archive.com
This email sent to [email protected]
