On Mon, 2011-07-18 at 09:11 -0300, Lucas Meneghel Rodrigues wrote: > On Mon, Jul 18, 2011 at 9:01 AM, James Laska <[email protected]> wrote: > > Greetings, > > > > While running autotest-0.13.0, I've come across the following error as > > reported in the scheduler.log. Before I start digging in, I thought I'd > > ask if anyone has already seen this issue? I'm wondering if this is > > related to the stream file handle that is removed when running the > > scheduler forks from the SysV initscript. > > Well, need to check, but I've noticed you are using py 2.6 there. The > guys at the chrome OS team @google uncovered a bug that is a python > 2.6 bug: > > http://autotest.kernel.org/changeset/5489 > > This bugfix didn't get into autotest 0.13.0, so would you check if > this solves your problem? you can easily download this patch to try it > out: > > http://patchwork.test.kernel.org/patch/3542/mbox/ > > Thanks and let me know how it goes,
I saw the original bug, and the above patch on the list when packaging +testing autotest-0.13.0 for Fedora. The autotest RPM's I created for Fedora [1] include that patch. One thing I noticed that's different with the current upstream RH-style initscript included in autotest, the scheduler/babysitter stdout handling after forking doesn't seem correct. Meaning, output still is sent to the tty after starting the scheduler. I'm not sure if if this is at all related. Also, I'm seeing that the babysitter seems to be dead, while monitor_db.py continues. Still poking around ... Thanks, James [1] http://repos.fedorapeople.org/repos/fedora-qa/autoqa/
signature.asc
Description: This is a digitally signed message part
_______________________________________________ Autotest mailing list [email protected] http://test.kernel.org/cgi-bin/mailman/listinfo/autotest
