796706 - "Communications has crashed" then re-dial of previous call

Status: RESOLVED WORKSFORME

(Core :: DOM: Device Interfaces, defect, P3)

Description

[GH to BZ]

[GitHub issue by overholt on 2012-09-21T20:48:34Z, https://github.com/mozilla-b2g/gaia/issues/5043]
I'm not sure if this will ever be reproducible, but here's what happened:

- I made a phone call to person A (I think my cheek hit the hang-up button in the middle but other than that disconnect things were okay)
- I hung up
- I made a phone call to person B
- I hung up
- I looked at the phone and it said "Communications has crashed" along the bottom
- then I could hear person A saying "Hello?" even though it did not appear a call was in progress
- I couldn't hang up so person A hung up
- I tried to make a call to person C but it never completed

I reset the phone since I didn't know what would happen next :)

Comment 1

[GH to BZ]

[GitHub comment by etiennesegonzac on 2012-09-23T07:13:25Z]
Since the Dialer app is finally **really** OOP (wasn't when we still had the in process background service) it can be killed, leaving the RIL with an ongoing call.

I don't the details of the memory management strategy in gecko but if possible we should probably try to avoid killing an app with an attention screen open (ie. with *a* window active, but the main window hidden).

Ping @cgjones

Comment 2

[Vivien Nicolas (:vingtetun) (:21) - (NOT reading bugmails, needinfo? please)]

Chris,

I don't know what is the exact process for killing an application but will it be possible to fire something on the application to let it know it will be killed? That will at least live a change to the communications app to stop the call.

Also is there a way from the front-end to adjust the oom_score?

Comment 3

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

We never kill foreground apps.  Are you talking about closing apps through the task switcher?  Or maybe this was crash?

We know in the telephony DOM impl when a client has exited unexpectedly, so it's probably best to do cleanup there.

Comment 4

[:Margaret Leibovic]

What's there to do here? Is this reproducible?

(In reply to Chris Jones [:cjones] [:warhammer] from comment #3)
> We never kill foreground apps.  Are you talking about closing apps through
> the task switcher?  Or maybe this was crash?

If the "Communications has crashed" banner showed up, that means that there was a fatal mozbrowsererror.

> We know in the telephony DOM impl when a client has exited unexpectedly, so
> it's probably best to do cleanup there.

If that's the case should this bug be moved over to a different component?

Comment 5

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

(In reply to Margaret Leibovic [:margaret] from comment #4)
> What's there to do here? Is this reproducible?
> 
> (In reply to Chris Jones [:cjones] [:warhammer] from comment #3)
> > We never kill foreground apps.  Are you talking about closing apps through
> > the task switcher?  Or maybe this was crash?
> 
> If the "Communications has crashed" banner showed up, that means that there
> was a fatal mozbrowsererror.

For OOP content, that means a crash.

> 
> > We know in the telephony DOM impl when a client has exited unexpectedly, so
> > it's probably best to do cleanup there.
> 
> If that's the case should this bug be moved over to a different component?

Sure.  Core -> DOM:Device Interfaces should be fine.

Comment 6

[:Margaret Leibovic]

Okay, I'll move this over.

Comment 7

[Doug Turner (:dougt)]

do you think you have a few minutes to look into this?

Comment 8

[Dave Hylands [:dhylands]]

If we have an active call, then we should probably change the oom_adj/oom_score_adj in a manner that makes the Dialer less likely to be chosen.

Comment 9

[Jonas Sicking (:sicking) No longer reading bugmail consistently]

Moving this to blocking-. We've only seen this problem once and that didn't result in enough actionable data.

If this is happening more often then surely we'll find it through dogfooding.

Comment 10

[Nick Davidson]

Unable to repro this issue on Unagi

Build ID: 20130208070202
Kernel: Dec 5
Gecko  http://hg.mozilla.org/releases/mozilla-b2g18/rev/4518b78583d0
Gaia   7e54ca673277b20b1d91d18477dc44d6ad226761

Comment 11

[Geo Mealer [:geo] -- This account is inactive after 2015-07-07]

Looks like we tried to repro but couldn't in comment 10. Removing qawanted. If the problem reappears go ahead and re-add the flag.

Comment 12

[Hsin-Yi Tsai (she/her) [:hsinyi]]

Havne't seen this for over a year. Going to close it.