630002 - Firefox 4: MobileMe iDisk has JavaScript error that breaks functionality

Status: RESOLVED DUPLICATE of bug 594999

(Core :: XPConnect, defect)

Description

[Marcus Münch]

I saw this first in the beta version of Firefox 4, that enabled the new JavaScript engine, but I am not really shure if this is a Mozilla bug or a bug within the Sproutcore framework:

When using Firefox 4 (current beta, but also every earlier version with new Javascript engine enabled) with the MobileMe iDisk webinterface, I get this error when clicking on a folder in the interface:

Permission denied to access property 'isObject' @ https://www.me.com/my/sproutcore/de/wpw/javascript.js:29

This leads to a complete broken functionality in the MobileMe webinterface, the only way to get it back working is to reload the page. This is the Sproutcore version Apple uses on the website:

var Prototype={Version:"1.6.0.3"

Firefox 3.x works fine, all other browsers too. 

Steps to reproduce:
- Log in to your MobileMe account. 
- Select "iDisk"
- Click on a folder in the iDisk, i.e. "Documents"
- You get the error: Permission denied to access property 'isObject'
-> The interface is now broken and does not respond anymore. Every click now leads to this error:

this.parentNode.parentNode is null @ https://www.me.com/my/shared/de/1sj2/javascript.js:13

I see this on OSX and Windows, I do not know about the Linux version.

Comment 1

[Marcus Münch]

Additional information: iDisk is the only webapp on the MobileMe site, that shows this error.

Comment 2

[Andreas Gal :gal]

This is an exception thrown by a security wrapper.

Comment 3

[Boris Zbarsky [:bzbarsky]]

Marcus, would you be willing to get a stack trace for that exception in Firebug?

Comment 4

[Marcus Münch]

Shure, what do I have to do? Firebug 1.7x.a9 is installed.

Comment 5

[Boris Zbarsky [:bzbarsky]]

Hmm.  My experience with Firebug is about nil, but I can look up exact steps if you're unfamiliar with it.  The basic idea would be to break on exception and then look at the callstack...  Let me know if you do want me to look up more detailed steps?

Comment 6

[Marcus Münch]

I try my best when I'm back home from work. Normally I', used to Firebug, but not the stacktrace feature, just simple Javascript debugging and CSS stuff.

Comment 7

[Marcus Münch]

Attachment: Firebug trace with options DBG2FIREBUG, ERRORS and FBS_ERRORS active.

Creating a stacktrace on error was not as easy as I thought. Instead I traced everything after the click on the folder object, I hope this helps you. If I should activate more options in the trace utility, just tell me.

One note: I do not really know, what lead to this, but with the live trace window of Firebug open and some tab switching magic, I twice got the code on  MobileMe running without an error. Looks like a timing problem for me, maybe some code is running parallel, which should not run parallel? I'm not sure, my knowledge of the new JavaScript engine is not good. And I really cannot reproduce this.

Comment 8

[Boris Zbarsky [:bzbarsky]]

Marcus, thanks!  That trace does have the information I wanted; unfortunately it's dozens of stack frames all on one line, including the entry point.  :(

Another thing to try...  Would you be willing to test nightly builds to narrow down when this broke?

Comment 9

[Marcus Münch]

Sure. As far as I remember, it broke with the beta release that was the first to have the new JavaScript engine enabled (don't know, beta 6 or so). Just tell me, which nightlies I should try.

Comment 10

[Boris Zbarsky [:bzbarsky]]

If you go into about:config and flip the "javascript.options.methodjit.content" to false, does the problem go away?

No matter what the answer to that question is, flip the pref back.

You should be able to find nightlies at <http://ftp.mozilla.org/pub/mozilla.org/firefox/nightly/2010/>.  You want to test the "tracemonkey" nightlies, I suspect.  Maybe start with the 2010-10-01 tracemonkey nightly, and see if it shows the bug.  If so, go back in time.  If it does, then pick a nightly halfway between then and a build you know is bad and test that?

Thanks a ton for helping narrow this down.

Comment 11

[Josh Matthews [:jdm]]

This is a dup of bug 627305 which is a dup of bug 594999, right?

Comment 12

[Boris Zbarsky [:bzbarsky]]

Oh, hmm.  Yes, looks likely....

Marcus, want to just try today's tracemonkey nightly and see whether your issue is fixed, in that case?

Comment 13

[Marcus Münch]

Ok, did it. Flipping javascript.options.methodjit.content did not change anything, but today's tracemonkey works fine again. Seems to be fixed, thank you!

Comment 14

[Boris Zbarsky [:bzbarsky]]

Thanks for checking that!  Sorry I didn't find the duplicate bug.  :(