1336641 - Crash early in aurora and nightly for imap sync helpers that will crash anyway later.

Status: RESOLVED FIXED

(MailNews Core :: Networking: IMAP, defect)

Description

[Kent James (:rkent)]

We're fighting some threading related crashed in IMAP (See bug 1335638 for example), but the null receiver causing the crash is set earlier, and we don't really see the root cause.

Crash earlier to get better feedback. We might change these to MOZ_RELEASE_ASSERT later (or should we just do it now?) if we are convinced that these issues will cause a crash eventually anyway.

Comment 1

[Kent James (:rkent)]

Attachment: crash earlier with null receiver in aurora and nightly

This will crash at the time that the proxy is being created, so that we can get crash stats with a useful history. Otherwise we are just going to crash later I believe when the object is used with null mReceiver.

It would not take much to convince me to make the a MOZ_RELEASE_ASSERT now, since it seems to me that we'll crash anyway, so we might as well do it when it is useful.

Comment 2

[Jorg K (CEST = GMT+2)]

Comment on attachment 8833555 [details] [diff] [review]
crash earlier with null receiver in aurora and nightly

Nice, thanks for looking into it.

On an unrelated note: Aleth is finding those leaks (bug 1334874), one by one, slowly but surely. Three leaks found, 6 of 14 failing tests fixed. So we're really cleaning up some legacy problems here.

Comment 3

[Jorg K (CEST = GMT+2)]

https://hg.mozilla.org/comm-central/rev/584c4519de6bf1521a6eadee0c07d185541c2aea

Comment 4

[Jorg K (CEST = GMT+2)]

Aurora (TB 53):
https://hg.mozilla.org/releases/comm-aurora/rev/2bc44a54c33292cc3c5ed231b7f676e67a3cbafe

Comment 5

[Jorg K (CEST = GMT+2)]

Kent, are you sure you don't want to uplift this to beta and ESR given the useful results we're getting, see bug 1335638 comment #10.

Comment 6

[Kent James (:rkent)]

My fear for uplift to release is that we might start getting lots of unexpected crashes that we were not getting before. In other words, I never fully trust my own analysis "Otherwise we are just going to crash later I believe when the object is used with null mReceiver.".

Uplift will do no good (or have not effect), instead we have to change to a release crash and not a diagnostic crash.