894701 - Disable Watchdog thread entirely on b2g18

Status: RESOLVED WONTFIX

(Core :: XPConnect, defect)

Description

[Bobby Holley (:bholley)]

sicking indicated in bug 870043 comment 31 that we'd probably prefer to just have no watchdog thread at all on b2g.

I've got some patches to re-architect the watchdog thread and make it prefable, but that's not going to be landable on b2g18. In the mean time, I think we can just do the dumb thing and #ifdef it out.

Comment 1

[Bobby Holley (:bholley)]

https://tbpl.mozilla.org/?tree=Try&rev=2fc45eab6230

Comment 2

[Bobby Holley (:bholley)]

Attachment: Disable the watchdog thread on b2g. v1

This seems to have some odd failures on the try push, but it also seems like
nobody is pushing b2g18 stuff to try, since my try push seemed to bundle along
more than 2000 changesets to the try repo. Are these failures expected?

Comment 3

[Ryan VanderMeulen [:RyanVM]]

In general, pushing to b2g18 is a minefield, yes.

Comment 4

[Bobby Holley (:bholley)]

https://hg.mozilla.org/releases/mozilla-b2g18/rev/13a54f1eac4f

I realized after pushing that I probably should have gotten approval or something. But dougt says I should just leave it. needinfo-ing release mgmt to make sure they're in the loop here.

Comment 5

[Ryan VanderMeulen [:RyanVM]]

blocking-b2g18:leo+ is the approval you're looking for :)

Comment 6

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/releases/mozilla-b2g18_v1_1_0_hd/rev/13a54f1eac4f

Comment 7

[Bobby Holley (:bholley)]

In a nutshell, the issue here is that we have a thread which, unless the phone is completely idle, wakes up once per second in order to (potentially) trigger the slow script dialog. This has been reported by partner QA as waking up very regularly when a music app is running, which in general is a battery drain.

In bug 870043 comment 31, sicking explains why we probably don't even want the watchdog at all on B2G:
* The parent process should never trigger the slow script dialog, because we wrote all the JS that runs there.
* All untrusted content runs out-of-process, which means that the user can still manually interrupt (and kill) and apps that hang.

Comment 8

[Bobby Holley (:bholley)]

(Note that I've rewritten the watchdog and made it pref-controlled in bug 870043, but it's not the kind of low-risk fix we'd take on b2g18. The patch here simply removes the machinery at compile-time for Gonk builds).

Comment 9

[Lukas Blakk [:lsblakk] use ?needinfo]

we should take this to v1.1, blocking as uplift approval.

Comment 10

[Lukas Blakk [:lsblakk] use ?needinfo]

(In reply to Ryan VanderMeulen [:RyanVM UTC-4] from comment #6)
> https://hg.mozilla.org/releases/mozilla-b2g18_v1_1_0_hd/rev/13a54f1eac4f

Ryan - why did this get uplifted to v1.1hd without being hd+?

Comment 11

[Ryan VanderMeulen [:RyanVM]]

Anything that lands on b2g18 is automatically merged to v1.1hd?

Comment 12

[Doug Turner (:dougt)]

lsblakk,
it was uplifted prematurely and by accident.  bholley gave reasons why we want this.  i needed info on akeybl asking if we should back out.

Comment 13

[Bobby Holley (:bholley)]

We decided in an email thread that this wasn't worth the risk, and that it would be better to let it ride the trains.

Backing out for b2g18.

https://hg.mozilla.org/releases/mozilla-b2g18/rev/35cb56244e74

Comment 14

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/releases/mozilla-b2g18_v1_1_0_hd/rev/35cb56244e74