1242811 - GrBatchAtlas.cpp:243: failed assertion "!plot->unique()"

Status: RESOLVED WORKSFORME

(Core :: Graphics: Canvas2D, defect)

Description

[Jesse Ruderman]

Attachment: testcase

On shutdown:

gfx/skia/skia/src/gpu/GrBatchAtlas.cpp:243: failed assertion "!plot->unique()"

Comment 1

[Jesse Ruderman]

Attachment: stack

Comment 2

[Lee Salzman [:lsalzman]]

I was not able to reproduce this at all. I tried loading the test case then shutting down Firefox in various ways. It all worked fine.

Comment 3

[Lee Salzman [:lsalzman]]

Attachment: discard partial draws before destroying GrContext

While I couldn't manage to reproduce this one, at least it would be a good idea to force the GrContext to discard any remaining unprocessed batched draws. This way we avoid the GrContext trying to draw things during the shutdown process for whatever reason, which could only lead to nothing good.

Comment 4

[Jeff Muizelaar [:jrmuizel]]

Comment on attachment 8712773 [details] [diff] [review]
discard partial draws before destroying GrContext

Review of attachment 8712773 [details] [diff] [review]:
-----------------------------------------------------------------

::: gfx/gl/SkiaGLGlue.cpp
@@ +942,5 @@
>       * through it, uses mGLContext
>       */
> +    if (mGrContext) {
> +      // Discard any partial draws as deleting the GrContext will implicitly flush them.
> +      mGrContext->flush(GrContext::kDiscard_FlushBit);

Is discarding the right thing to do? What if we have a snapshot? What makes sure it gets these deferred updates?

Comment 5

[Jeff Muizelaar [:jrmuizel]]

(In reply to Jesse Ruderman from comment #0)
> Created attachment 8711967 [details]
> testcase

Can you reproduce in rr?

Comment 6

[Jeff Muizelaar [:jrmuizel]]

Comment on attachment 8712773 [details] [diff] [review]
discard partial draws before destroying GrContext

Review of attachment 8712773 [details] [diff] [review]:
-----------------------------------------------------------------

I'm going to r- for now until we better understand what's happening here.

Comment 7

[Ryan VanderMeulen [:RyanVM]]

This no longer reproduces on trunk.

Comment 8

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/mozilla-inbound/rev/fbad23d9e737
Add crashtest. r=me

Comment 9

[Attila Craciun [:arny]]

https://hg.mozilla.org/mozilla-central/rev/fbad23d9e737