1247122 - Intermittent 1228456.html,205735-1.xhtml | application crashed [@ mozilla::CycleCollectedJSRuntime::CycleCollectedJSRuntime(JSRuntime*, unsigned int, unsigned int)]

Status: RESOLVED FIXED

(Core :: DOM: Workers, defect)

Description

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

https://treeherder.mozilla.org/logviewer.html#?job_id=21354740&repo=mozilla-inbound

Comment 1

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

https://treeherder.mozilla.org/logviewer.html#?job_id=21345673&repo=mozilla-inbound

Comment 2

[Andrew McCreight [:mccr8]]

It looks like this is crashing on this MOZ_CRASH:

  mJSRuntime = JS_NewRuntime(aMaxBytes, aMaxNurseryBytes, aParentRuntime);
  if (!mJSRuntime) {
    MOZ_CRASH();
  }

I think that indicates that Firefox ran out of memory while trying to create a new JS runtime.

Comment 3

[Andrew McCreight [:mccr8]]

I'm moving this to DOM:Workers because it looks like the test creates 99 SharedWorkers, which seems like it could cause an OOM.

Comment 4

[Kyle Huey (Exited; not receiving bugmail, old account, do not use)]

Yeah ... this seems like something likely to OOM.

Comment 5

[Andrew McCreight [:mccr8]]

As noted in the summary, there are also a few (4 at current count) crashes in dom/xbl/crashtests/205735-1.xhtml. That test looks quite benign, but it is in fact the test that is run immediately after 1228456.html so presumably we are just running out of memory slightly later.

Comment 6

[Intermittent Failures Robot]

23 automation job failures were associated with this bug yesterday.

Repository breakdown:
* mozilla-inbound: 15
* fx-team: 5
* mozilla-central: 3

Platform breakdown:
* linux32: 23

For more details, see:
https://brasstacks.mozilla.com/orangefactor/?display=Bug&bugid=1247122&startday=2016-02-10&endday=2016-02-10&tree=all

Comment 7

[Intermittent Failures Robot]

25 automation job failures were associated with this bug yesterday.

Repository breakdown:
* fx-team: 10
* mozilla-inbound: 9
* mozilla-central: 6

Platform breakdown:
* linux32: 25

For more details, see:
https://brasstacks.mozilla.com/orangefactor/?display=Bug&bugid=1247122&startday=2016-02-11&endday=2016-02-11&tree=all

Comment 8

[Andrea Marchesini [:baku]]

Attachment: cycle.patch

This patch propagates the error in case we have allocation issues.

Comment 9

[Kyle Huey (Exited; not receiving bugmail, old account, do not use)]

Comment on attachment 8718920 [details] [diff] [review]
cycle.patch

Review of attachment 8718920 [details] [diff] [review]:
-----------------------------------------------------------------

::: js/xpconnect/src/XPCJSRuntime.cpp
@@ +3539,2 @@
>  
>      if (self                                    &&

remove the null check for self while you're here.

Comment 10

[Intermittent Failures Robot]

18 automation job failures were associated with this bug yesterday.

Repository breakdown:
* fx-team: 11
* mozilla-central: 5
* mozilla-inbound: 2

Platform breakdown:
* linux32: 18

For more details, see:
https://brasstacks.mozilla.com/orangefactor/?display=Bug&bugid=1247122&startday=2016-02-12&endday=2016-02-12&tree=all

Comment 11

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/ac6fc75a900c

Comment 12

[Intermittent Failures Robot]

74 automation job failures were associated with this bug in the last 7 days.

Repository breakdown:
* mozilla-inbound: 33
* fx-team: 26
* mozilla-central: 15

Platform breakdown:
* linux32: 73
* windows7-32: 1

For more details, see:
https://brasstacks.mozilla.com/orangefactor/?display=Bug&bugid=1247122&startday=2016-02-08&endday=2016-02-14&tree=all

Comment 13

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/ac6fc75a900c

Comment 15

[Andrea Marchesini [:baku]]

Comment on attachment 8718920 [details] [diff] [review]
cycle.patch

Approval Request Comment
[Feature/regressing bug #]: Cycle Collected JSRuntime initialization / workers
[User impact if declined]: Rarely a crash if the initialization of the worker JSRuntime fails.
[Describe test coverage new/current, TreeHerder]: none
[Risks and why]: the patch moves some code in order to return an error value in case the initialization of a new JS runtime fails.
[String/UUID change made/needed]: none

Comment 16

[Andrew McCreight [:mccr8]]

This needs beta approval, not Aurora. (This is needed because the crash is apparently impacting e10s, as seen in bug 1259187).

Comment 17

[Andrew McCreight [:mccr8]]

Comment on attachment 8718920 [details] [diff] [review]
cycle.patch

See comment 15. This is already on Aurora, but not Beta. It is needed to reduce the OOM crash rate, which is unusually high with e10s.

Comment 18

[Liz Henry (:lizzard) (relman/hg->git project)]

Comment on attachment 8718920 [details] [diff] [review]
cycle.patch

If this landed Tuesday or Wednesday this week in 46 beta 6, that would give us maybe a couple of days of e10s experiment information. I think we can wait for 47 here as it is not a high volume crash and seems to be often e10s related.