David Anderson [:dvander] - inactive, e-mail if emergency Assignee
	Description • 8 years ago

Sample job: https://treeherder.mozilla.org/#/jobs?repo=try&author=danderson@mozilla.com&selectedJob=30215450

Assertion: http://searchfox.org/mozilla-central/rev/f5c9e9a249637c9abd88754c8963ecb3838475cb/gfx/layers/client/TextureClient.cpp#853

	David Anderson [:dvander] - inactive, e-mail if emergency Assignee
	Comment 1 • 8 years ago

I've also seen this assert [1], possibly related. Job is at [2].

Trying some diagnostics: probably this is a TextureClient created before a process restart, and someone is trying to use it again after.

[1] http://searchfox.org/mozilla-central/rev/f5c9e9a249637c9abd88754c8963ecb3838475cb/gfx/layers/client/TextureClient.cpp#863
[2] https://treeherder.mozilla.org/#/jobs?repo=try&author=danderson@mozilla.com&selectedJob=30614139

	David Anderson [:dvander] - inactive, e-mail if emergency Assignee
	Comment 2 • 8 years ago

I can reproduce this now. Happens when canvas is active and the GPU process aborts. TextureClient ends up holding a previous TextureForwarder, and a new one has since been created.

Seems easiest to just early return if IPC isn't open, but I don't know this code well.

With this the content process no longer crashes, but canvas stops rendering until the page is refresh.

	Pulsebot
	Comment 3 • 8 years ago

Pushed by danderson@mozilla.com:
https://hg.mozilla.org/integration/mozilla-inbound/rev/c99f95ab0b90
Fix canvas crashing when the GPU process aborts. (bug 1314427, r=nical)

	Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)
	Comment 4 • 8 years ago
bugherder

https://hg.mozilla.org/mozilla-central/rev/c99f95ab0b90