1143856 - (apz-linux) Enable APZ on Linux on nightly

Status: RESOLVED FIXED

(Core :: Panning and Zooming, defect)

Description

[Botond Ballo [:botond]]

This bug is for getting APZ to work on desktop linux.

Comment 1

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

Note: This bug is to track enabling APZ on Linux on nightly. The only patch that lands as part of this bug should be the patch that actually flips the pref. All bugs that block this MUST be fixed in order to enable APZ. Bugs that are "nice to have" but not strictly required to enable APZ on Linux should block something else instead (most likely bug 1013364).

Comment 2

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

Attachment: Patch

Try push with all the test fixes:

https://treeherder.mozilla.org/#/jobs?repo=try&revision=4384db0b25e3

Comment 3

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

New and improved try push:

https://treeherder.mozilla.org/#/jobs?repo=try&revision=9493331c4453

Comment 4

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

One more to hopefully address the frequent M-e10s-1 failure which I filed bug 1208072 for:

https://treeherder.mozilla.org/#/jobs?repo=try&revision=d30dc41fa715

Comment 5

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

Comment on attachment 8665048 [details] [diff] [review]
Patch

I won't land this until everything is green, but might as well flag it for review since we're getting to the finish line here.

Comment 6

[Botond Ballo [:botond]]

Comment on attachment 8665048 [details] [diff] [review]
Patch

Review of attachment 8665048 [details] [diff] [review]:
-----------------------------------------------------------------

::: browser/app/profile/firefox.js
@@ +1890,5 @@
>  pref("browser.tabs.remote.autostart.2", true);
>  #endif
>  
>  #ifdef NIGHTLY_BUILD
> +#ifndef MOZ_MULET

Sweet!

Comment 7

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/dfdc5964965d

Comment 8

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

Couple of linux m-e10s(1) failures popped up after this landed:
https://treeherder.mozilla.org/logviewer.html#?job_id=14723678&repo=mozilla-inbound
https://treeherder.mozilla.org/logviewer.html#?job_id=14723703&repo=mozilla-inbound

Backed out in https://hg.mozilla.org/integration/mozilla-inbound/rev/2bdfbf53e401 to see if they go away.

Comment 9

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

Ugh looks like all my try pushes where linux32 and these failures are on linux64. I'll see if I can track them down.

Comment 10

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

Attachment: Crash stack

Both the linux64 debug and linux64 asan failures have a crash stack in them (attached). I'm not sure why this is happening only on linux64 and not linux32. The error message has something to do with calling IsCallerChrome without having a AutoJSAPI on the stack, and on IRC ehsan said bholley would be the person to ask - bholley, do you know what this error means and what level in the stack has the fault? Full log is at http://ftp.mozilla.org/pub/mozilla.org/firefox/tinderbox-builds/mozilla-inbound-linux64-debug/1443203735/mozilla-inbound_ubuntu64_vm-debug_test-mochitest-e10s-1-bm121-tests1-linux64-build198.txt.gz

Comment 11

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

Doh, forgot to actually needinfo ^

Comment 12

[Bobby Holley (:bholley)]

Sure. Can you file a separate bug for this crash and I'll write a patch for it? That patch can land underneath your stuff.

Comment 13

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

I filed bug 1208622 for it and assigned it to you. Thanks!

Comment 14

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

Bug 1208622 doesn't seem to happen any more on latest master, so we should be good to land this now. One more try push for good luck:

https://treeherder.mozilla.org/#/jobs?repo=try&revision=a42a2440d351

Comment 15

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

Looks like scroll-inactive-layers-2.html is failing R-e10s but only on opt builds. This wasn't happening on the last try push I did [1] which means this was regressed by something in the range [2].

As this is probably going to keep happening I might just disable the test and land APZ on Linux to prevent further regressions, and then work on getting the test re-enabled. Feels kinda underhanded but I worry that if I don't do this we're going to keep getting more regressions since we're not exercising APZ in tests on any other platform.

[1] https://treeherder.mozilla.org/#/jobs?repo=try&revision=b0b4f3859cac&group_state=expanded
[2] http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=031db40e2b55&tochange=ccee6614fd9d

Comment 16

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

I did some retriggers, some of which came back green. I also did another push based on latest m-i [1] which is mostly green. Still, the scroll-inactive-layers-2.html is frequent enough that I'm going to disable the test for now. I can reproduce it locally (intermittently) and should be able to debug it and re-enable it.

[1] https://treeherder.mozilla.org/#/jobs?repo=try&revision=ad0b8f273c6f

Comment 17

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/16003c3d5474
https://hg.mozilla.org/integration/mozilla-inbound/rev/30742281c223

Comment 18

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/16003c3d5474
https://hg.mozilla.org/mozilla-central/rev/30742281c223