1120252 - startup crash in mozilla::layout::RenderFrameParent::GetApzcTreeManager()

Status: RESOLVED FIXED

(Core :: Panning and Zooming, defect)

Description

[(Away)]

This bug was filed from the Socorro interface and is 
report bp-205720aa-02a2-4baf-bbd1-6c7c52150109.
=============================================================

This startup crash exploded with build 20150109030224 and is currently the top crash in 3-day data on nightly. The crash is Win7SP1 only.

RenderFrameParent::GetApzcTreeManager received a poisoned value from CompositorParent::GetAPZCTreeManager, so it looks like the CompositorParent is dead.

Regression range: https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=70de2960aa87&tochange=b3f84cf78dc2

Bug 1109873 looks the most suspicious. Kats can you confirm? As a persistent startup crash this could cost us users and we should avoid letting it hit aurora even for a day. Please consider a backout if it can't be fixed before merge day.

Comment 1

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

I don't think bug 1109873 is the likely culprit here, as it doesn't affect the codepaths that are being hit in the crash stack. I'll take a closer look later this evening, leaving ni on me for now.

Comment 2

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

This might be the same issue as bug 1107009. It's not clear to me why it spiked suddenly. I'm pretty sure that my patches in bug 1109873 would not have caused this. However since this crash is on Windows only, and APZC is disabled there anyway, we can add an extra check for the APZ pref and prevent going down this code path earlier. That should be a safe and quick fix that we can get in before the merge.

Comment 3

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

Attachment: Patch

This should avoid the crash for now.

Comment 4

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

(Note also that the expected return value of this function is null when APZ is disabled because of the code at http://mxr.mozilla.org/mozilla-central/source/gfx/layers/ipc/CompositorParent.cpp?rev=160acaa4fc44#389)

Comment 5

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

https://hg.mozilla.org/integration/mozilla-inbound/rev/26f5729d5ccc

Comment 6

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/26f5729d5ccc

Comment 8

[(Away)]

No hits so far on the latest nightly. We'd definitely have seen them by now if the issue were still happening. So the fix looks good. Thanks Kats!

Comment 9

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

Based on the crash in bug 1120485 which started on the same nightly as this one and is also windows 7, I suspect the root cause here was bug 1107718.

Comment 10

[Ryan VanderMeulen [:RyanVM]]

Pushed a=bustage to fix mochitest crashes from bug 1107009.

https://hg.mozilla.org/releases/mozilla-beta/rev/1b21115851ef
https://hg.mozilla.org/releases/mozilla-b2g34_v2_1/rev/78c600a5f2fa
https://hg.mozilla.org/releases/mozilla-b2g32_v2_0/rev/8ea706019b32
https://hg.mozilla.org/releases/mozilla-esr31/rev/e224d1b46d56
https://hg.mozilla.org/releases/mozilla-esr31/rev/7ae1b58e8458
https://hg.mozilla.org/releases/mozilla-b2g30_v1_4/rev/76e172ece6de

Comment 11

[Vincent Liu[:vliu]]

https://hg.mozilla.org/releases/mozilla-b2g34_v2_1s/rev/78c600a5f2fa

Comment 12

[[SV Manager] Florin Mezei, QA (:FlorinMezei)]

Socorro [1] shows zero crashes in Firefox 36 and 37 (builds after January 11).

[1] - https://crash-stats.mozilla.com/report/list?product=Firefox&range_unit=days&range_value=28&signature=mozilla%3A%3Alayout%3A%3ARenderFrameParent%3A%3AGetApzcTreeManager%28%29#tab-reports