209559 - Applet calls JavaScript and Mozilla crashes on reloading the page

Status: RESOLVED INCOMPLETE

(Core Graveyard :: Java: OJI, defect)

Description

[ewlieber]

User-Agent:       Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.0.1) Gecko/20020830
Build Identifier: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.0.1) Gecko/20020830

I load a simple java applet which in the start method just calls a javascript
function which counts to some large number then returns.  After the function
returns, the applet thread sleeps for some amount of time, then calls the same
javascript function again.  When you hit the reload button in mozilla, then
almost all of the time mozilla crashes.  I have tried this in 1.0.1 as well as
1.3.1 and they both seem to crash.

Reproducible: Always

Steps to Reproduce:
1. Open up the testcase.html
2. Wait for it to load and start running
3. Hit Mozilla's reload button

Actual Results:  
Mozilla starts to reload the page, but then just crashes entirely.

Expected Results:  
Reload the page without crashing, or at least display some error message saying
what went wrong.

Comment 1

[ewlieber]

Attachment: html/javascript for the test case

test case that exhibits the problem...works along with Test.java

Comment 2

[ewlieber]

Attachment: Java Applet that shows the bug happening

Comment 3

[Phil Schwartau]

---> OJI

Comment 4

[Hermann Schwab]

I just wanted to test some old bugs, either if they are happening on Windows
too, or if they are Windows bugs, if they got solved by brandnew Java 1.4.2_02
I don´t know how to test yours:
did a download of both files to a directory, and loaded the html from it
then started the html directly from the attachment.
In both cases, I get an error message in the status line of the browser:
Applet not inited.
Looking now, it is even translated into german (had been english before)
Can you give advice how to test?

Comment 5

[Kyle Yuan]

Test with jre1.5, I got this exception when reload:

netscape.javascript.JSException: Native DOM Object has been released
	at sun.plugin.javascript.navig5.JSObject.invoke(JSObject.java:131)
	at sun.plugin.javascript.navig5.JSObject.call(JSObject.java:175)
	at Test.start(Test.java:8)
	at sun.applet.AppletPanel.run(Unknown Source)
	at java.lang.Thread.run(Unknown Source)

mozilla doesn't crash anymore. There is still some minor issues when quitting
mozilla, though it's not critical.

Comment 6

[Beat Zahnd]

original testcase did not work here, but I have my own at
https://mefisto1.unibe.ch/~bz/lc.html and Java applet source at
https://mefisto1.unibe.ch/~bz/src/LiveConnectTest.java

Appplet creates a thread which calls
win.eval("document.getElementById(\"testtext\").firstChild." +
	"nodeValue = \"run() loop: " + Integer.toString(i++) + "\"");

If the dellay in the loop is increased to 1000 and the reload button hit between
eval() calls no crashes occur

Symptoms on Firefox 1.0.2 (Linux) and JRE 1.4.2_08:

Page loads and applets runs without problems. As soon as page is left or
reloaded firefox crashes. Talkbak ID: TB5194787X

Comment 7

[Kyle Yuan]

Please try it again using the most recent jre, say jre 1.5.0_02. We fixed a lot
of liveconnect bugs in 1.5.

Comment 8

[Beat Zahnd]

Yes, I tried it yesterday using jre 1.5.0_2. No crashes so far with Firefox
1.0.2 and Mozilla 1.7.7 both on WinXP an Linux.

There seems to be a memory leak in jre. The java_vm process constantly increases
memory using my example.

Comment 9

[Kyle Yuan]

(In reply to comment #8)
> There seems to be a memory leak in jre. The java_vm process constantly increases
> memory using my example.

We fixed a memory leak problem in jre 1.5.0_02. the version number of jre can be
found from the first line of java console.

Comment 10

[Beat Zahnd]

(In reply to comment #9)
> We fixed a memory leak problem in jre 1.5.0_02. the version number of jre can be
> found from the first line of java console.

I hava jre 1.5.0_02 on windows and linux. There must be still a leak somewhere ...

Comment 11

[Pete Zha]

mass reassign to Alfred

Comment 12

[Benjamin Smedberg]

Mass-closing bugs in the "OJI" component: OJI plugin integration was replaced with npruntime long ago, and these bugs appear to be irrelevant now. If there is in fact a real bug that remains, please file it new in the "Core" product, component "Plug-ins".