602650 - Plugin container keeps running, spewing log after killing frozen Firefox

Status: RESOLVED INCOMPLETE

(Core Graveyard :: Plug-ins, defect)

Description

[s.marshall]

User-Agent:       Mozilla/5.0 (Macintosh; Intel Mac OS X 10.6; rv:2.0b6) Gecko/20100101 Firefox/4.0b6
Build Identifier: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.6; rv:2.0b6) Gecko/20100101 Firefox/4.0b6

I can make the plugin container not die (and continue spewing frequent log output for at least a minute or so, possibly forever) on my Mac.

Versions-

Firefox 4 beta 6.
OS X 10.6.4
Java(TM) SE Runtime Environment (build 1.6.0_21-b07-298-10M3239a)
^^ this is latest java developer preview, I haven't tried with retail version java because there is no way to uninstall DPs on OS X - I accept this may be unreliable version but the point of plugin container is that it behaves better when plugins are buggy, right?


Reproducible: Always

Steps to Reproduce:
1) Go to this buggy java applet

http://live.leafdigital.com/temp/bugzilla/loadtest.html

2) Make it crash browser by clicking the 'go' link, wait a second, click 'go' again (browser beachballs)

3) Kill browser with cmd+opt+esc

4) Open Console, observe zillions (about 25 per second) of the following messages spewing into log

07/10/2010 21:59:33	[0x0-0x43a43a].org.mozilla.firefox	###!!! [Child][AsyncChannel] Error: Channel error: cannot send/recv

5) ~ $ ps -ax | grep iref
46668 ??         0:01.18 /Applications/Firefox.app/Contents/MacOS/plugin-container.app/Contents/MacOS/plugin-container /System/Library/Java/Support/Deploy.bundle/Contents/Resources/JavaPlugin2_NPAPI.plugin -omnijar /Applications/Firefox.app/Contents/MacOS/omni.jar 46663 plugin gecko-crash-server-pipe.46663

Actual Results:  
as above

Expected Results:  
(Note: The browser crash is fine, I don't have a problem with that, it's just the log junk.)

Expected result would be that the plugin container outputs approximately one such log message and then exits itself, instead of outputting 25 per second and keeping running.

I noticed a few bugs that might be related:

bug 584626 - plugin container process survives after shutdown?  
bug 577157 - [OOPP] Plugin container will not close  

This didn't seem exactly the same (first one is Vista specific, second is not about closing the browser) so I thought it was safer to file a new bug and let somebody more knowledgeable work out if it is a duplicate.

Comment 1

[s.marshall]

To confirm, I was now able to reproduce this on a second Mac (found one at work) which I assume has the current standard (shipping) Java version. 1.6.0_20 something.

Comment 2

[Q]

I'm able to reproduce similar behavior with Flash. Open a flash process (#1). Open a second flash process that crashes Firefox 4.0. Flash process #1 hangs around in the plugin container until I manually kill it.

Firefox 4.0 (also RC 2)
Mac OS X 10.6.6

I can provide more information on the flash process that crashes Firefox if needed.

Comment 3

[Q]

In my above comment, I tried the same steps in Firefox 3.6.15, and the Plugin Process was killed when force quitting Firefox. But on upgrade to Firefox 4.0, the Plugin Process was no longer killed on force quitting Firefox.

(In reply to comment #2)
> I'm able to reproduce similar behavior with Flash. Open a flash process (#1).
> Open a second flash process that crashes Firefox 4.0. Flash process #1 hangs
> around in the plugin container until I manually kill it.
> 
> Firefox 4.0 (also RC 2)
> Mac OS X 10.6.6
> 
> I can provide more information on the flash process that crashes Firefox if
> needed.

Comment 4

[Benjamin Smedberg]

Resolving old bugs which are likely not relevant any more, since NPAPI plugins are deprecated.