1249880 - non-e10s generating content process crash reports

Status: RESOLVED WONTFIX

(Core :: DOM: Content Processes, defect)

Description

[Jonathan Howard]

The reporting is mostly but not exclusively limited to telemetry.
bug 1222890 comment 37 spotted the problem.
bug 1219672 comment 20 highlighted small number showing up with socorro.

https://telemetry.mozilla.org/new-pipeline/dist.html#!compare=e10sEnabled&cumulative=0&end_date=2016-02-15&keys=content!pluginhang!plugin!__none__&max_channel_version=beta%252F45&measure=SUBPROCESS_CRASHES_WITH_DUMP&min_channel_version=beta%252F42&processType=false&product=Firefox&sanitize=1&sort_keys=submissions&start_date=2016-01-27&table=1&trim=1&use_submission_date=0

Comment 1

[Jonathan Howard]

I spotted this which looks to be using the wrong argument. Not investigated further.
https://dxr.mozilla.org/mozilla-central/source/dom/ipc/CrashReporterParent.cpp#63

Comment 2

[Robert Kaiser]

Also, note that non-e10s builds actually do have content processes for thumbnailing.

Comment 3

[Bill McCloskey [inactive unless it's an emergency] (:billm)]

(In reply to Jonathan Howard from comment #1)
> I spotted this which looks to be using the wrong argument. Not investigated
> further.
> https://dxr.mozilla.org/mozilla-central/source/dom/ipc/CrashReporterParent.
> cpp#63

Yeah, that is wrong. It's not used by desktop Firefox though.

Comment 4

[Benjamin Smedberg]

I strongly suspect thumbnailing for this, and the limited evidence backs this up: there is no visible way to submit thumbnail content process crashes: you'd have to visit about:crashes to submit them. That's why they show up much higher in telemetry than in crash-stats.

Without more data, I'm not sure what to make of this. We always treated this as low-priority in the past, because a failed thumbnail is just not a big deal.

With e10s enabled, are these thumbnails still in a separate process from the "main" e10s content process? If not, and fiven that e10s is coming, perhaps we should just WONTFIX this?

Comment 5

[Mark Hammond (PTO; back Oct-3) [:markh] [:mhammond]]

(In reply to Benjamin Smedberg  [:bsmedberg] from comment #4)
> I strongly suspect thumbnailing for this, and the limited evidence backs
> this up: there is no visible way to submit thumbnail content process
> crashes: you'd have to visit about:crashes to submit them. That's why they
> show up much higher in telemetry than in crash-stats.

There is explicit handling for process crashing at https://dxr.mozilla.org/mozilla-central/rev/789a12291942763bc1e3a89f97e0b82dc1c9d00b/toolkit/components/thumbnails/BackgroundPageThumbs.jsm#193, so it should be easy to add telemetry or some other action around these crashes if we care enough.

> With e10s enabled, are these thumbnails still in a separate process from the
> "main" e10s content process? If not, and fiven that e10s is coming, perhaps
> we should just WONTFIX this?

I believe it is the same process - we just create a <browser type="content" remote="true"/> which IIUC will end up in the single e10s process.

Comment 6

[Jim Mathies [:jimm]]

We discussed splitting thumbnailing out in bug 1187441 - Use separate content processes for e10s browsing and thumbnail generation.

Comment 7

[Andrew Overholt [:overholt]]

Should we leave this open?