324751 - Java Applet using LiveConnect crashes FF 1.5

Status: RESOLVED INCOMPLETE

(Core Graveyard :: Java: Live Connect, defect)

Description

[Mark Cave-Ayland]

User-Agent:       Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.8) Gecko/20051111 Firefox/1.5
Build Identifier: Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.8) Gecko/20051111 Firefox/1.5

I have a web application that I am working on which uses a combination of FF/Java 1.5 update 6/Javascript and LiveConnect which consistently crashes my FF 1.5 browser. Unfortunately due to the code being under an NDA it is practically impossible for me to post the source of the code that works, however there is perhaps the possibility of making a compiled version available if there is enough here for developers to work with.

Note: I have tried this under IE6 and while it works 95% of the time, there still appears to be a window in which IE will also crash. However this is 3-4 times a day as opposed to every time I launch the applet.

Here roughly in pseudo-code is what the page containing the applet does:

1. Use window.setInterval to poll documents.applets[0].isActive every 500ms to detect whether the applet has loaded.

2. When the above is set to true, invoke the Open method on the Java Applet.

3. After calling some JNI routines (which load images in the background into a Java Image object) the result is display on screen.

4. Once the images have loaded, a LiveConnect call is made back to the web page to allow the application to continue. Uncommenting this LiveConnect call prevents the crash, however the web page doesn't continue to execute as it never receives the LiveConnect notification.



Reproducible: Always

Steps to Reproduce:
As above.




Since providing access to the application is not possible, I compiled a debug version of the FF 1.5 source code using MingW/cygwin and managed to extract a backtrace using GDB.

Comment 1

[Mark Cave-Ayland]

Attachment: This is the GDB backtrace from a specially compiled FF 1.5

Comment 2

[Mark Cave-Ayland]

I have been working mostly on this application mostly over a slow 512kbps DSL connection with VPN, however today I was in the office on the same LAN as the web application and the number of crashes was much less. So I wonder if it is some sort of timing issue, and having the slower connection makes it almost certain to occur.

Comment 3

[timeless]

afaik that specific assert is covered in some other bug, and i don't think it's the one that matters. try continuing?

Comment 4

[Mark Cave-Ayland]

(In reply to comment #3)
> afaik that specific assert is covered in some other bug, and i don't think it's
> the one that matters. try continuing?

Yes you're right; if I keep hitting Ignore enough times it eventually goes away. Unfortunately the mingw build doesn't seem to be able to launch the java plugin so taking advice from some of the guys on IRC, I rebuilt my debug build with MSVC (which took the best part of a day). Fortunately I am now able to get a back trace from the code.

Comment 5

[Mark Cave-Ayland]

Attachment: This is WinDbg backtrace from FF 1.5

Comment 6

[Mark Cave-Ayland]

I've managed to produce a standalone test applet for this bug which I have uploaded to http://www.webbased.co.uk/mca/mozcrash/. Note that the applet is signed because it downloads a test image from one of our image servers.

The way to provoke the crash is as follows: close down all FF windows, then start FF and visit the above URL. As soon as the page appears (and while the applet is still loading), keep quickly moving the mouse in circles over the applet panel until the crash occurs.

It is possible that it is some form of event overflow condition exacerbated by the long start time of the applet, and the fact that the viewer has a MouseMotionListener attached to it that raises a JS event back to the web page whenever the mouse is moved.

Comment 7

[timeless]

in case you weren't aware, http://www.mozilla.org/build/windbgdlg.html should make life better for such things.

i wonder if this happens on osx...

Comment 8

[Steven Michaud [:smichaud] (Retired)]

> i wonder if this happens on osx...

It doesn't ... at least not when using the Java Embedding Plugin :-)

The JEP doesn't crash on the URL at bug 264062, either
(http://www.spurgius.com/colorpicker/colorpicker.html).

So this bug may be related to bug 264062 (and bug 260478, of which bug 264062
has been made a duplicate).

The "MRJ Plugin JEP" (descended from the MRJ Plugin) required a massive number
of bugfixes, particularly of its LiveConnect code, before it became truly
usable.  But now that it's been fixed up, I find that its LiveConnect code is
sometimes much more efficient than Sun's (as found in Sun's "Java Plugin" on
Windows).

(The "MRJ Plugin JEP" is part of the Java Embedding Plugin, of which I'm the
author.  The Java Embedding Plugin is now bundled with recent Mozilla.org Mac
OS X browsers.  For more info see http://javaplugin.sourceforge.net/)

Comment 9

[Mark Cave-Ayland]

(In reply to comment #7)
> in case you weren't aware, http://www.mozilla.org/build/windbgdlg.html should
> make life better for such things.

Ah OK perhaps I haven't been as clear as I have been; I only saw the Assertions appear with my MingW build; as soon as I recompiled with VC they seemed to go away so I assumed that it was somehow related to a MingW compiled FF trying to invoke a VC compiled Java Plugin.

However, I still manage to get the crash consistently here on WinXP, FF 1.5, JRE 1.5.05 - can anyone else on a PC confirm the backtrace above, and whether it is in fact a FF problem? I see in Steven's comments a reference to bug 260478 which claims to have been fixed a similar problem here (http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=6254466). But I've been seeing this bug with both Java 1.4 and 1.5 for quite a while now so I don't believe this is the case :(

Comment 10

[Mark Cave-Ayland]

Has anyone had a chance to confirm that this is definitely not a FF bug? I'd be happy to chase this up with Sun myself if it does turn out to be a plugin issue.

Comment 11

[Alfred Peng]

Mark, is it still reproducible for FF2.0 and Java1.6.0? The link you provided is obsolete.

Kenneth, any chance to take a look?

Comment 12

[Kenneth Russell]

We can take a look if the test case comes back on-line.

Comment 13

[Gili Tzabari]

Is this a duplicate of #375070?

Comment 14

[Wayne Mery (:wsmwk)]

(In reply to comment #11)
> Mark, is it still reproducible for FF2.0 and Java1.6.0? The link you provided
> is obsolete.

reporter email address appears to be dead (considering that the webspace is gone) and reporter hasn't responded

=> incomplete