546902 - Camino-2.1a1pre-1.9.2.2pre.app crash in [@ objc_msgSend | IdleTimerVector ] or [@ objc_msgSend | HIToolbox@0x155fef ]

Status: RESOLVED FIXED

(Camino Graveyard :: General, defect)

Description

[Lars Sonchocky-Helldorf]

User-Agent:       Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10.5; en; rv:1.9.2.2pre) Gecko/20100214 Camino/2.1a1pre (like Firefox/3.6.2pre)
Build Identifier: Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10.5; en; rv:1.9.2.2pre) Gecko/20100214 Camino/2.1a1pre (like Firefox/3.6.2pre)

see attached crash.log

Reproducible: Didn't try




Mac OS X 10.5.8

Comment 1

[Lars Sonchocky-Helldorf]

Attachment: crash.log

Comment 2

[Smokey Ardisson (offline for a while; not following bugs - do not email)]

The good news is that early indications are that the patch for bug 533001 fixes this crash, and we'll spin a new experimental build with that patch in the next few days.

Thanks for the report, but please note that we're trying not to file bugs against the experimental builds (since they're not official builds) but instead reporting issues in this forum thread: http://forums.mozillazine.org/viewtopic.php?f=12&t=1754775

Comment 3

[Steven Michaud [:smichaud] (Retired)]

mem0's comment dated February 16th, 2010, 12:36 pm at
http://forums.mozillazine.org/viewtopic.php?f=12&t=1754775
(specifically the crash log this comment refers to at
http://www.file-upload.net/download-2266768/Camino_2010-02-16-201317_memos-macbook.txt.html)
seems to indicate that my patch for bug 509130 (currently on the 1.9.2
branch and the trunk) can itself cause IdleTimerVector crashes in
Camino:

That crash log doesn't include anything from DivX in its list of
modules.  And in Firefox (1.9.0 and 1.9.1 branches), IdleTimerVector
crashes are normally caused by a bad interaction between Carbon-mode
WebKit (used/loaded by the DivX plugin) and Gecko's two-event-queue
model of event handling.  (See bug 509130 for more info --
particularly bug 509130 comment #35 and bug 509130 comment #44.)

So in your tests, Smokey, you should really 1) try to find some way to
use the DivX plugin to trigger IdleTimerVector crashes in 1.9.0-branch
and/or 1.9.1-branch Camino, and 2) see if my patch for bug 533001
fixes those crashes.

Comment 4

[Steven Michaud [:smichaud] (Retired)]

> the DivX plugin

Specifically, the DivXBrowserPlugin module.

Comment 5

[Smokey Ardisson (offline for a while; not following bugs - do not email)]

(In reply to comment #3)
> seems to indicate that my patch for bug 509130 (currently on the 1.9.2
> branch and the trunk) can itself cause IdleTimerVector crashes in
> Camino:

That was my theory, too.

> So in your tests, Smokey, you should really 1) try to find some way to
> use the DivX plugin to trigger IdleTimerVector crashes in 1.9.0-branch
> and/or 1.9.1-branch Camino, and 2) see if my patch for bug 533001
> fixes those crashes.

As I mentioned in bug 536684 comment 27, I've had an old (2007) DivX Browser Plug-in installed since 2007, and I've never had any crashes with 1.9.0-based Caminos (that's not to say that bug 509130 itself doesn't exist in 1.9.0-based Caminos--it does and is a Camino topcrash--but IdleTimerVector crashes *without visiting DivX content* do not appear to exist in 1.9.0-based Caminos).  

However, as soon as I started a 1.9.2-based Camino with my regular profile, Camino crashed reliably during launch (tab restoration).  Once I removed the DivX Browser Plug-in, I was able to start Camino, but I'd get still get IdleTimerVector crashes randomly about 6 times a day (which again seems to support the theory that the patch for bug 509130 causes IdleTimerVector crashes in Camino, as DivX was no longer installed and I wasn't viewing any plug-in content at all when the random IdleTimerVector crashes happened).

Since applying the patch for bug 533001 to my 1.9.2-based Camino, I've not had a single IdleTimerVector crash in over 24 hours (fingers crossed).

Yesterday I also tested both 1.9.2-based builds, unpatched and patched, with my normal profile and with the DivX Browser Plug-in re-installed.  Although I could not crash the unpatched version as reliably/automatically as I could the other day, I was able to trigger the crash, with some effort (switching tabs and windows while tab restoration was in progress) every time.  I was unable to trigger the crash with the patched version any time.  

This data seems to support the theory that the patch in bug 533001 will prevent the IdleTimerVector crashes that the patch in bug 509130 appears to trigger.

Comment 6

[Steven Michaud [:smichaud] (Retired)]

> This data seems to support the theory that the patch in bug 533001
> will prevent the IdleTimerVector crashes that the patch in bug
> 509130 appears to trigger.

Yes.  But you should also test with current versions of the DivX
plugin -- even the most recent one triggers IdleTimerVector crashes in
1.9.0-branch and 1.9.1-branch Firefox, and in 1.9.0-branch Camino.

I'm quite confident my patch for bug 533001 will clear up these
crashes.  But it's nice to get independent support for this.

Comment 7

[Eiichi]

I have not installed DivX Browser Plugin, but I have seen this crash in [@ objc_msgSend | IdleTimerVector ] more than two or three times a day on 10.4.11.

So I made my trial build with the patch for bug 533001, and since then I have not seen this crash for 72 hours.

I agree with Smokey about effectiveness of the patch for bug 533001.

Comment 8

[Steven Michaud [:smichaud] (Retired)]

This should be fixed by my patch for bug 533001, which just landed on the 1.9.2 branch.

Comment 9

[Smokey Ardisson (offline for a while; not following bugs - do not email)]

(So far all the 1.9.2 crashes we're seeing in crash-stats for these signatures have been from builds that people grabbed before we announced them, i.e., before the fix for bug 533001 was in.  That--plus the fact no one reported this crash in the experimental builds that included the patch for bug 533001 for ~3mos--makes me feel good that the fix is effective.)

Comment 10

[Al Billings [:abillings - ex-MoCo]]

Nothing for QA to do with 1.9.2 for this bug.

Comment 11

[Smokey Ardisson (offline for a while; not following bugs - do not email)]

Yeah, this was just the Camino placeholder for getting bug 533001 landed on 1.9.2--and to have a bug on record for the crash/regression from bug 509130, since the crashes in Camino were different/triggered differently from the other crashes bug 533001 fixed.