1091766 - Drop messages more gently in ContentChild

Status: RESOLVED FIXED

(Core :: DOM: Content Processes, defect)

Description

[Andrew McCreight [:mccr8]]

Currently, a child process quietly exits when a message to or from it gets dropped. This causes us to not run the rest of shut down, including leak logs. Instead of doing this, I would like to just soldier on. The message is still not going to be sent.

Bug 1067633 is on file to track some of the places we send messages too late.

Comment 1

[Andrew McCreight [:mccr8]]

Attachment: Don't exit(0) when a message is dropped in the child process.

We end up printing a message and failing by returning false higher up in the stack.

try runs: https://treeherder.mozilla.org/ui/#/jobs?repo=try&revision=c123693cd56a
https://treeherder.mozilla.org/ui/#/jobs?repo=try&revision=4836d88b6fb5

Comment 2

[Andrew McCreight [:mccr8]]

This appears to be causing a failure in B2G ICS Emulator opt M5 in PNeckoChild::SendPCookieServiceConstructor:

    bool __sendok = (mChannel)->Send(__msg);
    if ((!(__sendok))) {
        NS_RUNTIMEABORT("constructor for actor failed"); <-- failure here
        return nullptr;
    }

I guess before we'd fail to Send(), but just quietly exit(0) there, whereas now we return and NS_RUNTIMEABORT().

Comment 3

[Andrew McCreight [:mccr8]]

There's a lingering test issue or two, as indicated in the blocking bugs, but I think that's just weird tests, so this seems ready for review.
  https://treeherder.mozilla.org/ui/#/jobs?repo=try&revision=4ac4827a2b0d

Comment 4

[Ben Turner (not reading bugmail, use the needinfo flag!)]

Comment on attachment 8514566 [details] [diff] [review]
Don't exit(0) when a message is dropped in the child process.

Review of attachment 8514566 [details] [diff] [review]:
-----------------------------------------------------------------

Let's keep a warning at least?

Comment 5

[Andrew McCreight [:mccr8]]

(In reply to ben turner [:bent] (use the needinfo? flag!) from comm
> Let's keep a warning at least?

We actually print out a message higher up the stack, but yeah I can add in something here, in case something else calls into this.

Comment 6

[Andrew McCreight [:mccr8]]

Bug 1103036 seems like a better approach.

Comment 7

[Andrew McCreight [:mccr8]]

This papers over some shutdown weirdness, even with bug 1103036, so I'm going to go ahead and land it, hopefully today.

Comment 8

[Andrew McCreight [:mccr8]]

e10s try run: https://treeherder.mozilla.org/ui/#/jobs?repo=try&revision=d6b02c33dce7
full try run: https://treeherder.mozilla.org/ui/#/jobs?repo=try&revision=7b13debca7e7

Landed with a warning added.

https://hg.mozilla.org/integration/mozilla-inbound/rev/e097e82fa29b

Comment 9

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

https://hg.mozilla.org/mozilla-central/rev/e097e82fa29b