I should really pay attention to who I'm replying to. Adding the list back
on...


I updated two smokers to CPAN::Reporter 1.1803. Once of them was set to
command_timeout 1000, one had no command_timeout set. I started them running
on the same set of modules. The one with command_timeout 1000 got stuck in
the middle of the TermReadKey-2.30 test script while testing prereqs for
Angel_100. The last output was:
1..8
ok 1
ok 2
ok 3
ok 4
ok 5
ok 6
ok 7

The other smoker, without command_timeout successfully tested
TermReadKey-2.30 and saved a pass report, in which it completed all eight
tests.

The smoker that froze up has three processes that show up for `ps -F -U
perl5101 | grep ptee`, the smoker that is still running shows between three
and five process matching that most of the time.

--Dan

--
DCollins/ST47
Stevens Institute of Technology Class of '13
Technical Staff, DeBaun Auditorium
Lighting Designer, Noises Off, Stevens Dramatic Society 2010 Fall

On Sun, Oct 24, 2010 at 8:23 PM, David Golden <[email protected]> wrote:

> Try it without the command_timeout, please, and let me know if you
> find similar problems.
>
> -- David
>
>
> On Sun, Oct 24, 2010 at 8:21 PM, Dan Collins <[email protected]> wrote:
> > Yep, command_timeout is set to 1000.
> > --Dan
> >
> > On Sun, Oct 24, 2010 at 7:59 PM, David Golden <[email protected]> wrote:
> >>
> >> Are you  using a command timeout?  The big change from 1.1801 to
> >> 1.1802 was eliminating the requirement for Proc::ProcessTable and
> >> relying on a more direct approach (that I had hoped was more
> >> portable).  It's possible that the newer approach is causing problems.
> >>
> >> -- David
> >>
> >> On Sun, Oct 24, 2010 at 2:55 AM, Dan Collins <[email protected]> wrote:
> >> > Greetings list,
> >> >
> >> > I was recently advised to update my smokers to CPAN::Reporter 1.1803,
> to
> >> > allow the truncation limit to increase. I obliged earlier this week.
> >> > Since
> >> > then, all or most of my smokers have apparently stopped automatically
> >> > timing
> >> > out. I will come back to find that the smoker has been on the same
> >> > distribution for many hours, though rarely are they on the same distro
> >> > or in
> >> > the same point in the build cycle - I have observed them stopped in
> the
> >> > middle of a test, or in the middle of compiling an XS module, etc. I
> >> > have
> >> > also observed them freezing up on the very first module to be tested.
> I
> >> > use
> >> > CPAN::Reporter::Smoker on Linux (Ubuntu, some older version) 32 bit,
> >> > with
> >> > perls build in alternate users' home directories and ulimit -v 150000
> >> > set
> >> > and with command_timeout 1000. I have tried increasing the ulimit (to
> >> > 250000) to no avail. I have also repeatedly restarted the smokers
> (using
> >> > killall -u perl5100 perl, as ctrl-C would not work: terminal displayed
> >> > control characters but nothing happened), clearing out the build
> >> > directories, and lowering the system load to only one smoker at a
> time,
> >> > to
> >> > no avail, so I have had to downgrade my smokers to CPAN::Reporter
> >> > 1.1801.
> >> > Doing so appears to have corrected the issue.
> >> >
> >> > I noticed a potentially similar however old bug report on RT,
> bug#61912.
> >> >
> >> > If there is any further information I can provide, please let me know.
> I
> >> > can
> >> > push one of my smokers back to 1.1803 if you'd like to test anything.
> >> >
> >> > --Dan
> >> >
> >> > --
> >> > DCollins/ST47
> >> > Stevens Institute of Technology Class of '13
> >> > Technical Staff, DeBaun Auditorium
> >> > Lighting Designer, Noises Off, Stevens Dramatic Society 2010 Fall
> >> >
> >
> >
>

Reply via email to