1427088 - Crash in mozilla::gl::GLContext::InitWithPrefixImpl

Status: RESOLVED FIXED

(Core :: Graphics, defect)

Description

[[:philipp]]

This bug was filed from the Socorro interface and is
report bp-5bf45d47-9d8f-41e2-b7af-4bb7a0171223.
=============================================================

Top 10 frames of crashing thread:

0 libxul.so mozilla::gl::GLContext::InitWithPrefixImpl gfx/gl/GLContext.cpp:867
1 libxul.so mozilla::gl::GLContext::InitWithPrefix gfx/gl/GLContext.cpp:347
2 libxul.so mozilla::gl::GLContextEGL::Init gfx/gl/GLContextProviderEGL.cpp:284
3 libxul.so mozilla::gl::GLContextEGL::CreateGLContext gfx/gl/GLContextProviderEGL.cpp:541
4 libxul.so mozilla::gl::GLContextEGL::CreateEGLPBufferOffscreenContext gfx/gl/GLContextProviderEGL.cpp:942
5 libxul.so mozilla::gl::GLContextProviderEGL::CreateHeadless gfx/gl/GLContextProviderEGL.cpp:960
6 libxul.so mozilla::widget::GfxInfo::GLStrings::EnsureInitialized widget/android/GfxInfo.cpp:76
7 libxul.so mozilla::widget::GfxInfo::EnsureInitialized widget/android/GfxInfo.cpp:37
8 libxul.so mozilla::widget::GfxInfo::GetFeatureStatusImpl widget/android/GfxInfo.cpp:403
9 libxul.so mozilla::widget::GfxInfoBase::GetFeatureStatus widget/GfxInfoBase.cpp:650

=============================================================

this fennec crash signature is regressing in volume in firefox 57 (but has been around before that already).
most of the crash reports come with MOZ_RELEASE_ASSERT(!err) added in bug 1339256.

Comment 1

[Nevin Chen(Not active on Bugzilla)]

Please help advise

Comment 2

[Blake Wu [:bwu][:blakewu]]

Jeff, 
Per comment 0, is this a regression from bug 1339256?

Comment 3

[Kelsey Gilbert [:jgilbert]]

Uh, yeah, but that should indeed never ever be crashing.

Comment 4

[Kelsey Gilbert [:jgilbert]]

This is the assert:
https://dxr.mozilla.org/mozilla-central/rev/351c75ab74c9a83db5c0662ba271b49479adb1f1/gfx/gl/GLContext.cpp#867
We can remove it, but the only time I can see it being hit is:
* Sudden context loss
* A bug we're not picking up on

Comment 5

[Kelsey Gilbert [:jgilbert]]

So at least the ff57 reports have 0x501 in r0, which is INVALID_VALUE. (ff58 has r0:0x0, but that's probably a (bad) optimization)
My guess is that we're querying GL_NUM_EXTENSIONS because we see we have fGetStringi, but that entrypoint isn't actually supposed to be used for some reason. (es2 contexts?) This would cause INVALID_VALUE.

Comment 6

[Kelsey Gilbert [:jgilbert]]

Attachment: Bug 1427088 - Don't rely on glGetStringi just because it exists. -

Review commit: https://reviewboard.mozilla.org/r/211102/diff/#index_header
See other reviews: https://reviewboard.mozilla.org/r/211102/

Comment 7

[Daosheng Mu[:daoshengmu]]

Comment on attachment 8940854 [details]
Bug 1427088 - Don't rely on glGetStringi just because it exists. -

https://reviewboard.mozilla.org/r/211102/#review216984

lgtm

Comment 8

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/mozilla-inbound/rev/722f664d592a
Don't rely on glGetStringi just because it exists. - r=daoshengmu

Comment 9

[Eliza Balazs [:ebalazs_]]

https://hg.mozilla.org/mozilla-central/rev/722f664d592a

Comment 10

[Sylvestre Ledru [:Sylvestre]]

it is something we want to uplift to 58?

Comment 11

[Kelsey Gilbert [:jgilbert]]

(In reply to Sylvestre Ledru [:sylvestre] from comment #10)
> it is something we want to uplift to 58?

This seems pretty low-volume, android-only, so probably not. If I'm reading the crash volume wrong, I'm willing to uplift it into late beta, but otherwise let's just let this already-on-release bug ride out 59.