102665 - Plugin "zombies" left unreaped

Status: RESOLVED INVALID

(Core Graveyard :: Plug-ins, defect, P3)

Description

[Chris Rankin]

I am using the CrossOver plugin to view QuickTime 5.0 content within Mozilla
0.9.4. This plugin spawns a process called the "pluginserver", whose parent is
Mozilla. When the pluginserver exits, Mozilla does not reap its zombie.

Comment 1

[Francois Gouget]

According to a mail on netscape.public.mozilla.plugins
([e-mail]), this problem also affects RealPlayer.

I'm not sure how to link to that article, but here's enough information to find
it:

Date: Fri, 11 Jan 2002 20:22:17 -0500
From: WD <[e-mail]>
To: [e-mail]
Newsgroups: netscape.public.mozilla.plugins
Subject: Realplayer plugin doesn't clean up after itself?

Comment 2

[av (gone)]

--- Mass reassigning Unix bugs to serge ---

Comment 3

[Gregory McLean]

This happens with any of the plugins and not just the crossover one. Definitly
some reaping should be going on in the mozilla process(s)
For example:
----
 2235 ?        Z      0:00 [pluginserver <defunct>]
 2241 ?        Z      0:00 [netstat <defunct>]
 2363 ?        Z      0:09 [pluginserver <defunct>]
 2378 ?        Z      0:09 [acroread <defunct>]
 2720 ?        Z      0:00 [realplay <defunct>]
----

Those all have a parent of mozilla and get reaped when mozilla exits.

Comment 4

[rubydoo123]

future

Comment 5

[Louis Bavoil]

Attachment: Patch for plugger.c 4.0

Apply this patch to plugger.c and recompile it to fix the bug with Plugger 4.0.

This bug could be fixed in Mozilla 1.0 very simply using the code of my patch.

Comment 6

[serge (gone)]

Thanks for the patch, but we do not control plugger's code,
so we cannot apply this patch for mozilla source tree,
please send an email to plugger author Fredrik Hübinette [e-mail].
In mozilla plugin specific signal handler has to be implemented in the future
releases, this is why TM for this bug is "Future".

Comment 7

[Josh Aas]

This is a plugin issue, not a browser issue.

Comment 8

[Gregory McLean]

Actually it is a browser issue..

When a process ends, all of the memory and resources associated with it are deallocated so they can be used by other processes. However, the process's entry in the process table remains. The parent can read the child's exit status by executing the wait system call, at which stage the zombie is removed. 

If you look at the zombie processes in question they all have the parent of the Mozilla process which is not cleaning up after itself. And it's not specifically the plugger plugin, it affects _all_ plugins.

Comment 9

[Benjamin Smedberg]

It is impolite for our code to wait/reap a process which we did not explicitly create. Presumably the plugin which created these processes should be waiting on them. In any case the Mozilla code is definitely not going to.