Closed Bug 1176548 Opened 8 years ago Closed 8 years ago

[e10s] crash in sse2_composite_src_x888_8888 with HWA disabled


(Core :: Graphics: Layers, defect)

41 Branch
Windows 7
Not set





(Reporter: over68, Assigned: bas.schouten)



(Keywords: crash, regression, topcrash-win)

Crash Data

The nightly build crash with hardware acceleration disabled.

Crash report: bp-3fd77e2a-e63d-4c52-a869-b84682150620
Blocks: 1172014
Crash Signature: [@ sse2_composite_src_x888_8888 ]
Flags: needinfo?(bas)
tracking-e10s: --- → ?
Component: Untriaged → Graphics: Layers
Product: Firefox → Core
Summary: [e10s] Crash in sse2_composite_src_x888_8888 → [e10s] crash in sse2_composite_src_x888_8888
Keywords: regression
The crash happens when start the browser with HWA disabled and enable e10s.
Also crashing on WinXP, not crashing in todays Aurora.

Steps to reproduce:
1. open about:crashes
2. load crash report
Flags: needinfo?(bas)
Found it after create of a new clean profile. On clean profiles is E10s enabled and HWA off here. 

unexpected software exception 0x80000003 in plugin-container.exe at .....

Deleted Pluginreg.dat to get firefox to start and disabled E10s to fix this.
Bas, Jeff, Milan: This bug has caused Nightly crash volume to go up 10x. Could you please take a look urgently? Thanks.
Flags: needinfo?(milan)
Flags: needinfo?(jmuizelaar)
This is a startup crash that has gone off the charts for Nightly builds of 41, 2015062003,  on Windows.
Keywords: crash, topcrash-win
Very interesting, I wonder if nightly is having some sort of alignment issue on certain systems.. what's the OS breakdown here?
Assignee: nobody → bas
We can just back bug 1172014 out.
(In reply to Jeff Muizelaar [:jrmuizel] from comment #9)
> We can just back bug 1172014 out.

Yeah, I'm fine with that too, although it would stop the supply of data, then again, we might already have the data we need.
Thanks Jeff. Last I checked the OS breakdown looked pretty typical.
Flags: needinfo?(milan)
Oh and it doesn't smell like alignment to me. A lot of the addresses end in 000 which makes me suspect a bad buffer where the read went off the end of a page.
Summary: [e10s] crash in sse2_composite_src_x888_8888 → [e10s] crash in sse2_composite_src_x888_8888 with HWA disabled
worksforme now
not crashing since update to yesterdays nightly 20150622052959
FIXED per backout. So far no reports of this crash on Nightly build of 20150623.
Closed: 8 years ago
Resolution: --- → FIXED
no more crashes with this signature on Nightly builds since the backout landed.
tracking-e10s: ? → ---
You need to log in before you can comment on or make changes to this bug.