[Buildbot-devel] Zombie process warning
l10n.moz at googlemail.com
Tue Sep 23 10:18:54 UTC 2008
I see this consistantly on at least mac and linux when using twisted 8.1.
Actually, I wonder if that isn't even intentional, because you don't
want to see the child process die from the starter, right?
I guess the right way to fix this is to explicitly handle that
exception and silence it.
2008/9/22 Christian Lins <Christian.Lins at sun.com>:
> Brian Warner schrieb:
>> Weird, I haven't seen that problem before. What platform is this on? And what
>> reactor is it using? (grep for "reactor class" in twistd.log).
> I am using Buildbot/0.7.9 on Ubuntu/8.04 (x86).
> cat twistd.log | grep reactor
> 2008-09-17 11:40:32+0200 [-] reactor class: <class
> 2008-09-22 10:18:30+0200 [-] reactor class: <class
>> When the process starts, a reactor is installed (something like
>> selectreactor, pollreactor, glibreactor, etc). When the reactor starts up,
>> it's supposed to register a signal handler for the SIGCHLD signal: this is
>> used to catch the SIGCHLD that occurs when a child process finishes.
>> The 'buildbot restart' command spawns a child process to watch the buildbot's
>> twistd.log: it looks for errors and warnings, and waits until it sees the
>> line that indicates the buildmaster/buildslave has started properly, then it
> In fact the message appears on "buildbot start" as well.
> ../bin/buildbot start .
> Following twistd.log until startup finished..
> PotentialZombieWarning: spawnProcess called, but the SIGCHLD handler is
> not installed. This probably means you have not yet called reactor.run,
> or called reactor.run(installSignalHandler=0). You will probably never
> see this process finish, and it may become a zombie process.
> 2008-09-22 10:18:30+0200 [-] Log opened.
> 2008-09-22 10:18:30+0200 [-] twistd 8.1.0 (/usr/bin/python 2.5.2)
> starting up
> Perhaps this was all my fault as I noticed here that I had the twistd
> 8.1 installed which is from Ubuntu/8.10 not 8.04, sorry for that ;)
> A reinstalled twistd 2.5 and the message disappeared...
>> hope that helps,
> Yes, very interesting explanation. Thank you :)
More information about the devel