341592 - Crash on first start of Sunbird - "Failed to get the script language"

Status: VERIFIED FIXED

(Core :: DOM: Core & HTML, defect)

Description

[Matthew (lilmatt) Willis]

On first launch of Sunbird, it crashes with the following:

###!!! ASSERTION: preserved wrapper table not empty at shutdown: 'sPreservedWrapperTable.ops == 0', file /Users/mwillis/moz/tree2/mozilla/dom/src/base/nsDOMClassInfo.cpp, line 3683
WARNING: nsExceptionService ignoring thread destruction after shutdown: file /Users/mwillis/moz/tree2/mozilla/xpcom/base/nsExceptionService.cpp, line 191
Failed to load jar:file:///Users/mwillis/moz/tree2/mozilla/obj-i386-apple-darwin8.6.2/dist/SunbirdDebug.app/Contents/MacOS/chrome/calendar.jar!/content/calendar/hiddenWindow.js
###!!! ASSERTION: Failed to get the script language: 'Error', file /Users/mwillis/moz/tree2/mozilla/dom/src/base/nsDOMScriptObjectFactory.cpp, line 187
###!!! ASSERTION: Failed to hold the script object: 'Error', file ../../../../dist/include/dom/nsDOMScriptObjectHolder.h, line 80
###!!! ASSERTION: You can't dereference a NULL nsCOMPtr with operator->().: 'mRawPtr != 0', file ../../../../dist/include/xpcom/nsCOMPtr.h, line 849

Comment 1

[Matthew (lilmatt) Willis]

Attachment: gdb bt and debugging information

copy/paste of bz's debugging

Comment 2

[Mark Hammond [:markh] [:mhammond]]

Attachment: Patch as described.

The problem here is that we attempt to get a script-language during shutdown, for the purposes of freeing/unrooting a JS object.

I'm attaching a patch that addresses this.  It involves not relying on do_GetService(), but instead keeping a reference to (the singleton) nsDOMScriptObjectFactory, and cleaning this up in the (new, static) Shutdown method.

Note that this patch does not touch the other cleanup done by the xpcom shutdown observer.  The fact this code attempts to use services implies that it may not be working as expected - however, I am leaving that alone and only addressing the specific error identified in this report.

Comment 3

[Mark Hammond [:markh] [:mhammond]]

Attachment: Updated patch that does threads better and avoids a static nsCOMPtr

Comment 4

[Boris Zbarsky [:bzbarsky]]

Um...  So _do_ we need the extra ref?  The old code cleared mLanguageArray in a shutdown observer; does just moving that to the destructor, without the rest of this patch, not help?

Comment 5

[Mark Hammond [:markh] [:mhammond]]

Attachment: new patch after being hit with the clue-stick

Comment 6

[Boris Zbarsky [:bzbarsky]]

Comment on attachment 225678 [details] [diff] [review]
new patch after being hit with the clue-stick

Ok, good.  ;)

Comment 7

[Mark Hammond [:markh] [:mhammond]]

Checked in - lilmatt, please verify if you can.

Checking in nsDOMScriptObjectFactory.cpp;
new revision: 1.13; previous revision: 1.12
Checking in nsDOMScriptObjectFactory.h;
new revision: 1.5; previous revision: 1.4

Comment 8

[Matthew (lilmatt) Willis]

We're still seeing this.

Comment 9

[Stephen Donner [:stephend] Not actively reading bugmail]

Bug 341595 is probably related.

Comment 10

[Jesse Ruderman]

This is a topcrash for Firefox trunk (bug 342025).

Comment 11

[Boris Zbarsky [:bzbarsky]]

I backed out that patch to fix bug 341700.  Please see comments there explaining what the issue was....

Comment 12

[Mark Hammond [:markh] [:mhammond]]

I just checked in a patch in bug 342025 which should hopefully stop this crash.  Please confirm (or deny) that you no longer see crashes here.

Comment 13

[Matthew (lilmatt) Willis]

(In reply to comment #12)
> I just checked in a patch in bug 342025 which should hopefully stop this crash.
>  Please confirm (or deny) that you no longer see crashes here.

Confirming!

Thank you!

-> FIXED (based on patch from bug 342025)

Comment 14

[Matthew (lilmatt) Willis]

Verified fixed in
Mozilla/5.0 (Macintosh; U; PPC Mac OS X Mach-O; en-US; rv:1.9a1) Gecko/20060718 Sunbird/0.3a2+

-> VERIFIED