879800 - WebGL abort on OSX 10.8 with OMTC

Status: RESOLVED WORKSFORME

(Core :: Graphics: CanvasWebGL, defect)

Description

[(no longer active)]

STR:  Run the game and shoot around for a bit.

Stack:

Program received signal SIGABRT, Aborted.
[Switching to process 47342 thread 0x871b]
0x00007fff8dcd4212 in __pthread_kill ()
(gdb) bt
#0  0x00007fff8dcd4212 in __pthread_kill ()
#1  0x00007fff990f4b54 in pthread_kill ()
#2  0x00007fff99138dce in abort ()
#3  0x00007fff99114905 in szone_error ()
#4  0x00000001223cf9e9 in gleRemoveHashName ()
#5  0x00000001222b9be7 in gleDeleteHashNameAndObject ()
#6  0x00000001222e1434 in glDeleteSync_Exec ()
#7  0x00000001029f77a1 in mozilla::gl::SharedSurface_GLTexture::WaitSync ()
#8  0x00000001029f895f in mozilla::gfx::SurfaceStream::SwapConsumer ()
#9  0x00000001029d77d9 in mozilla::layers::SurfaceStreamHostOGL::Lock ()
#10 0x00000001029bafe2 in mozilla::layers::ImageHostSingle::Composite ()
#11 0x0000000102996ba5 in mozilla::layers::CanvasLayerComposite::RenderLayer ()
#12 0x00000001029ab0f0 in mozilla::layers::ContainerRender<mozilla::layers::ContainerLayerComposite> ()
#13 0x00000001029ab0f0 in mozilla::layers::ContainerRender<mozilla::layers::ContainerLayerComposite> ()
#14 0x00000001029bf802 in mozilla::layers::LayerManagerComposite::Render ()
#15 0x00000001029bf48e in mozilla::layers::LayerManagerComposite::EndTransaction ()
#16 0x00000001029bf3fe in mozilla::layers::LayerManagerComposite::EndEmptyTransaction ()
#17 0x00000001029a7df0 in mozilla::layers::CompositorParent::Composite ()
#18 0x0000000102908404 in MessageLoop::DeferOrRunPendingTask ()
#19 0x00000001029086fa in MessageLoop::DoWork ()
#20 0x000000010290b009 in base::MessagePumpDefault::Run ()
#21 0x0000000102907d3e in MessageLoop::Run ()
#22 0x0000000102919389 in base::Thread::ThreadMain ()
#23 0x0000000102926e5a in ThreadFunc ()
#24 0x00007fff990f37a2 in _pthread_start ()
#25 0x00007fff990e01e1 in thread_start ()

Vlad tried this demo on 10.7 and the abort doesn't seem to be happening there.

Comment 1

[(no longer active)]

This seems to be OMTC only.

Comment 2

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

I'm not crashing on 10.7 with OMTC enabled either.  ATI HD 6330M for me.

Comment 3

[Kelsey Gilbert [:jgilbert]]

Well, it looks like we're managing to crash in the driver. It looks like it's not handling our OMTC usecase properly.

One thing we can do is to stop deleting the Sync object eagerly, instead only doing so lazily.

Before we do this, can we try to repro in a DEBUG build, so we have assert coverage? (unless this was already a DEBUG build)

Comment 4

[(no longer active)]

(In reply to comment #3)
> Well, it looks like we're managing to crash in the driver. It looks like it's
> not handling our OMTC usecase properly.
> 
> One thing we can do is to stop deleting the Sync object eagerly, instead only
> doing so lazily.
> 
> Before we do this, can we try to repro in a DEBUG build, so we have assert
> coverage? (unless this was already a DEBUG build)

Already did, and couldn't...

Comment 5

[Matt Woodrow (:mattwoodrow)]

Can you get a stack trace for the main thread during this crash?

Comment 6

[(no longer active)]

Hmm, I can't seem to reproduce it any more.  Can somebody else please try?

Comment 7

[Kelsey Gilbert [:jgilbert]]

Is it possible you caught the tail end of bug 874823?

Comment 8

[(no longer active)]

(In reply to Jeff Gilbert [:jgilbert] from comment #7)
> Is it possible you caught the tail end of bug 874823?

I'm not the best person to answer that question!  Somebody who actually knows this stuff should verify that.

Comment 9

[Milan Sreckovic [:milan] (needinfo for best results)]

So... this is works for me?