1339097 - Crash in mozilla::PresShell::Paint

Status: RESOLVED FIXED

(Core :: Web Painting, defect)

Description

[Marcia Knous [:marcia]]

This bug was filed from the Socorro interface and is 
report bp-dadb16ad-c025-44fc-9426-19a8e2170213.
=============================================================

Seen while looking at nightly crash stats - crashes started with 20170211030205: http://bit.ly/2lbNn78.

Possible regression range: https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=25a94c1047e793ef096d8556fa3c26dd72bd37d7&tochange=855e6b2f6199189f37cea093cbdd1735e297e8aa. Maybe Bug 1331294? ni on :heycam

Comment 1

[Timothy Nikkel (:tnikkel)]

I doubt it is Cameron's bug. More likely its the graphics branch merge to m-c.

Comment 2

[Kartikaya Gupta (email:[e-mail])]

It shouldn't be the graphics branch merge. 99% of that code isn't even activated unless Firefox is built with --enable-webrender, and that's not yet the case on nightly builds. I'm pretty sure we didn't touch anything in the crash stack. Putting needinfo back to heycam but I'm doubtful it's from his bug either.

Comment 3

[Timothy Nikkel (:tnikkel)]

Whatever caused this is unlikely to have changed code in the crash stack as the layer manager on a widget (the crash is on a null layer manager from the widget) is set up (hopefully) long before anything in this stack. I didn't see anything else likely on a quick scan of the pushlog.

Comment 4

[Kartikaya Gupta (email:[e-mail])]

Hm. I misread the code the first time, sorry. I guess this might be a regression from https://hg.mozilla.org/projects/graphics/rev/b7619d5557e0, I can take a look.

Comment 5

[Kartikaya Gupta (email:[e-mail])]

So this is almost certainly a result of PuppetWidget::GetLayerManager returning null if it doesn't have a mTabChild. I'd really like to know what kind of scenario results in the creation and painting of a PuppetWidget which doesn't have an associated TabChild, but we don't have STR and none of the crashes have any comments.

So for now I think it's better to just fall back to using the gfxVars::UseWebRender if we don't have a TabChild. We already basically concluded that trying to mix compositors is not really going to go well. We might want to get rid of the CompositorOptions entirely but there might still be some value in keeping it for now. I'm working on a patch.

Comment 6

[Kartikaya Gupta (email:[e-mail])]

https://treeherder.mozilla.org/#/jobs?repo=try&revision=efcc2a285b0d3872a59ae1abe613f302a5707761&group_state=expanded

Comment 7

[Kartikaya Gupta (email:[e-mail])]

Attachment: Bug 1339097 - Allow PuppetWidget::GetLayerManager to return a layer manager even if it doesn't have a tabchild.

Review commit: https://reviewboard.mozilla.org/r/112420/diff/#index_header
See other reviews: https://reviewboard.mozilla.org/r/112420/

Comment 8

[David Anderson [:dvander] - inactive, e-mail if emergency]

Comment on attachment 8837250 [details]
Bug 1339097 - Allow PuppetWidget::GetLayerManager to return a layer manager even if it doesn't have a tabchild.

https://reviewboard.mozilla.org/r/112420/#review113818

Comment 9

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/ab41a32432a3
Allow PuppetWidget::GetLayerManager to return a layer manager even if it doesn't have a tabchild. r=dvander

Comment 10

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/ab41a32432a3