1480139 - Disable tiling on Linux because of the high crash rate

Status: RESOLVED FIXED

(Core :: Graphics: Layers, defect, P1)

Description

[Ryan Hunt [:rhunt]]

We're getting many crash reports here with no obvious fix.

Comment 1

[Ryan Hunt [:rhunt]]

Attachment: tiling-disable.patch

Comment 2

[Jeff Muizelaar [:jrmuizel]]

Comment on attachment 8996761 [details] [diff] [review]
tiling-disable.patch

Review of attachment 8996761 [details] [diff] [review]:
-----------------------------------------------------------------

Can you put a little more information about the cause of the crashes (perhaps with a link to a bug) so that if we need to look back in the future that information is around.

Comment 3

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/mozilla-inbound/rev/b9b178105f1b
Disable tiling on linux due to high crash rate. r=jrmuizel

Comment 4

[Tiberius Oros[:tiberius_oros]]

Backed out changeset b9b178105f1b (bug 1480139) for background-position-001.xht failures

Backout: https://hg.mozilla.org/integration/mozilla-inbound/rev/ef88a6a743c26fce23c121d5c3372cebc0448032

Push with the failures: https://treeherder.mozilla.org/#/jobs?repo=mozilla-inbound&revision=b9b178105f1bd1d794e53c2fb128bd9e505f8f32

Failure log: https://treeherder.mozilla.org/logviewer.html#?job_id=191403398&repo=mozilla-inbound&lineNumber=5105

Comment 5

[Ryan Hunt [:rhunt]]

Yup, forgot there were reftest annotation changes involved here. Will update.

Comment 6

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/mozilla-inbound/rev/bfa19e6f2407
Disable tiling on linux due to high crash rate. r=jrmuizel

Comment 7

[Noemi Erli[:noemi_erli]]

https://hg.mozilla.org/mozilla-central/rev/bfa19e6f2407

Comment 8

[Eric Rahm [:erahm]]

Disabling also regressed explicit memory [1], which indicates there was a win on the initial landing. There's even larger RSS regression (~40MB) [2]. Ryan were you expecting memory improvements with bug 1471704? Is there a bug on file to re-enable?

[1] https://treeherder.mozilla.org/perf.html#/comparesubtest?originalProject=mozilla-central&newProject=mozilla-inbound&newRevision=bfa19e6f240792564d99c2f85c92354c0f1c84e4&originalSignature=e832acb0cd6ce7001b6d0c4dbb0f38bc6e46f87f&newSignature=e832acb0cd6ce7001b6d0c4dbb0f38bc6e46f87f&framework=4&selectedTimeRange=172800
[2] https://treeherder.mozilla.org/perf.html#/comparesubtest?originalProject=mozilla-central&newProject=mozilla-inbound&newRevision=bfa19e6f240792564d99c2f85c92354c0f1c84e4&originalSignature=9aa2ccc224aa311008091b25c3f052b84247b94d&newSignature=9aa2ccc224aa311008091b25c3f052b84247b94d&framework=4&selectedTimeRange=172800

Comment 9

[Ryan Hunt [:rhunt]]

I was not expecting this to cause memory improvements. We actually got a heap unclassified regression when we enabled tiling on some windows targets, bug 1458480.

There is a bug on file to re-enable, bug 1471704. Unfortunately, no one has time to work on it right now.

Comment 10

[Landry Breuil (:gaston)]

Congrats, somehow you broke nightly on OpenBSD, where we need tiling (cf Bug 1345899).

Bug 1471704 / https://hg.mozilla.org/integration/mozilla-inbound/rev/3e5f747ee223 changed defined(OS_OPENBSD) for defined(MOZ_WIDGET_GTK), and this bug here removed it in https://hg.mozilla.org/mozilla-central/rev/bfa19e6f2407, so now tiling is disabled on OpenBSD, and nightly crashes at startup (i cant even get the profile manager widget).

I'll retry locally to enable tiling for us, and revert back to the state where we were after Bug 1345899.

Comment 11

[Landry Breuil (:gaston)]

Nightly starts again on OpenBSD after reenabling tiling by default (ie reapply https://hg.mozilla.org/mozilla-central/rev/895d9a8826a3 which got lost in the recent commits)

Do i need to file another bug so that this situation is resolved ?

Comment 12

[Ryan Hunt [:rhunt]]

(In reply to Landry Breuil (:gaston) from comment #11)
> Nightly starts again on OpenBSD after reenabling tiling by default (ie
> reapply https://hg.mozilla.org/mozilla-central/rev/895d9a8826a3 which got
> lost in the recent commits)
> 
> Do i need to file another bug so that this situation is resolved ?

Apologies for the error. I'll file a bug and fix it.