Sun 08 Sep 2013 09:08:36 PM UTC, comment #14:
killing the zombie bug. Let's just hope it's not truly undead...
|
Sun 08 Sep 2013 02:47:47 AM UTC, comment #13:
I don't get the zombie processes either. Ok to close?
|
Thu 29 Aug 2013 02:34:15 AM UTC, comment #12:
Ping. Can someone who had a problem with zombie processes try the development version. Using hg id 91691d74845d I no longer seem to get zombies.
|
Mon 12 Mar 2012 07:53:25 AM UTC, comment #11:
probably this issue is caused by the following commit of mine:
http://hg.savannah.gnu.org/hgweb/octave/rev/2f0d1e12806d?revcount=32
|
Mon 20 Feb 2012 03:02:24 AM UTC, comment #10:
Ben,
your example code does not hang up in my Devian environment....
>> close all
>> graphics_toolkit gnuplot
>> plot (rand (3))
>> graphics_toolkit fltk
>> figure (2)
>> plot (rand (3))
>> print junk.png
|
Sat 18 Feb 2012 05:35:40 PM UTC, comment #9:
The following results in a hang.
I assume waidpid() is "waiting" for gnuplot to terminate.
I'm going to revert this changeset. We need a way to identify the pid's associated with the print() command.
|
Sat 18 Feb 2012 01:10:54 AM UTC, comment #8:
I changed the TODO -> FIXME and pushed the changeset.
http://hg.savannah.gnu.org/hgweb/octave/rev/688e19ca262b
If there are problems this can be reopened
|
Sat 18 Feb 2012 12:05:03 AM UTC, comment #7:
It's probably a good idea to avoid any weirdness on MinGW. Is there any problem with Cygwin as well?
From the Tips and Guidelines appendix, "The preferred comment mark for places that may need further attention is FIXME." Maybe you can change TODO to FIXME in the patch?
|
Fri 17 Feb 2012 11:21:51 PM UTC, comment #6:
I thought it wise to avoid changing the behavior of mingw, so I've modified the changeset. The 2nd version has been attached.
(file #25078)
|
Fri 17 Feb 2012 04:58:26 AM UTC, comment #5:
The patch works for me as well on Kubuntu 10.04 (A Debian derivative).
Ben, to check for zombie processes you should have some sort of process table viewer. Otherwise, you can do it from the command line using the 'ps' command.
For example, pre-patch
|
Fri 17 Feb 2012 04:23:35 AM UTC, comment #4:
the patch works well in my Debian environment.
octave 3.7.0+
Debian 6(i686)
libfltk 1.1
------
|
Fri 17 Feb 2012 04:21:30 AM UTC, comment #3:
t works in my Debian environment.
octave 3.7.0+
Debian 6(i686)
libfltk 1.1
------
|
Fri 17 Feb 2012 12:44:23 AM UTC, comment #2:
I've attached a changeset.
Either I don't know how or I don't see the zombie processes, so i'm not sure this is a working solution. I won't push until I have confirmation.
If someone can tell me how to check on the zombie processes that would helpful as well (I'm on MacOS, but can check on Ubuntu as well)
(file #25073)
|
Wed 26 Oct 2011 03:48:09 AM UTC, comment #1:
This is confirmed on the development branch. Multiple print commands also produce multiple zombies. The proposed solution of using "waitpid (0)" also works.
|
Wed 19 Oct 2011 02:16:30 AM UTC, original submission:
In fltk backend(not gnuplot backend), the "print" command generates a zombie process.
>> plot(1:100);
>> print -depsc2 test.eps
To clear the zombie process, it needs to call the "waidpid" command manually, like this.
>> waitpid(0)
Is this a bug?
|