29609 - Visiting a page with an applet, then pressing "Back" hangs the browser

Status: VERIFIED FIXED

(Core Graveyard :: Java: OJI, defect, P3)

Description

[edburns]

This may have something to do with the recently introduced fix to 27486.  

Or it may have something to do with Stanley's recent JPI workarounds.

Reproducible: Always:

Setup:

1. Mozilla non MOZ_DEBUG build with source from 1:30pm 02/28/00
2. JRE from Stanley Ho, built on 02/25/00

To reproduce:

1. Visit a page with an applet, say 

http://www.prometheus-music.com/gecko/5105/bft_browser_applet.html

2. Press the back button

The browser will hang and you'll have to kill it using the Windows NT Task 
Manager.

Comment 1

[edburns]

This bug manifests when you use ANY way of getting to another page.

Can we put this on PDT?

Ed

Comment 2

[edburns]

Stanley, I find that mozilla does successfully send the Stop() and SetWindow() 
messages to the plugin.

When it freezes, I use the debugger to "break" into the program and see this 
stack trace:

NTDLL! 77f677ef()
KERNEL32! 77f1c7d6()
JVM! 5044771d()
MSVCRT! 7800265a()
KERNEL32! 77f04f2c()

This looks like it's stuck in the JVM, but I can't tell since I only have the 
release build.  I'll do my own build tomorrow, but can you please take a look 
at this one?

Thanks,

Ed

Comment 3

[edburns]

This bug is the result of a non-deterministinc race condition in the JDK.

Comment 4

[stanley.ho]

This is a duplicate of #27486.


*** This bug has been marked as a duplicate of 27486 ***

Comment 5

[edburns]

On 1 March 15:16:57, Stanley Man-Kit Ho wrote:
> Hi all,
> 
> 	Okay, the good news is that I finally manage to figure out what was
> going wrong. Actually, a AWT bug was introduced in the last couple days
> that prevents the plugin to work inside any browser (not just Mozilla).
> The browser will either hang when Java starts up, or it will hang during
> page-switch. Because I usually tested my OJI plug-in with Kestrel RC1
> JRE, the bug did not occur on my machine. 
> 	The bad news is that the JRE build that I gave you guys actually
> contains the latest Kestrel RC2 JRE, and it has the AWT bug. I didn't
> realize this bug until this morning, when our release engineering has
> promoted the first Kestrel RC2 build and the SQE team complains about
> the plug-in not working. I have tracked down the blocking problem, but
> find no stack trace in the plug-in code. And I think the hang is
> originated somewhere inside AWT. We have alerted the AWT team and they
> are looking into it right now. In the meantime, if you want to get the
> plug-in to work, install Kestrel RC1 build (download from the web site),
> and patch it with the latest OJI plugin binaries.
> 

Comment 6

[ekrock's old account (dead)]

Nominating for beta1. Obviously, we can't brag too loudly about supporting Java 
in beta1 if once you visit any page with an applet, the window is then trapped 
on that page until you Exit the browser or crash it by attempting to leave.

Comment 7

[Phil Peterson]

[PDT+] w/b minus on 3/3

Comment 8

[drapeau]

As of right now, I believe this is going to turn into a release note.  The bug
is with the JDK, not with the Mozilla OJI module.  Using the JDK 1.3 Release
Candidate 1 (or JRE1.3RC1) is working for us, with Stanley Ho's patches to the
Java Plug-In.

We will likely deliver the JRE1.3RC1 plus patched Plug-In for delivery in Beta
1.  If our testing holds up today and tomorrow, that's the plan, and this bug
can be resolved.  We may have to note in release notes that customers should use
the JRE that comes bundled with the installer, or make sure that if the customer
had installed Java him/herself previously, that the version they installed was
JRE 1.3 RC 1.  It should also be noted that using any other version of the JRE
(JRE1.2.2, JRE 1.3RC2, for examples) will not work with the Beta browser, and
our intention is not to support Java releases previous to JDK1.3 (when JRE1.3
goes FCS, that will be supported).

Is this helpful, chofmann,ekrock?

Comment 9

[Jim Roskind]

As noted, this is a JRE bug, and can't be the basis for holding beta1
3/3 is gone... so...
PDT- for beta1
PDT+ for beta2

Comment 10

[leger]

Putting on beta2 keyword radar since this was "plus for beta2" during beta1 
triaging

Comment 11

[stanley.ho]

I tested the applet in the URL with RC1 with plug-in patch, and it doesn't hang 
anymore. This bug should have been fixed already.

Comment 12

[drapeau]

Okay, I'm closing this bug as FIXED, as per Stanley's comments.

Comment 13

[John Unruh]

Verified fixed - there is no crash.

Comment 14

[edburns]

Yep Yep, It's still fixed.