[Vm-dev] Intermittent failures on InterpreterVM build

David T. Lewis lewis at mail.msen.com
Sat Feb 16 14:34:58 UTC 2013


On Sat, Feb 16, 2013 at 08:59:29AM -0500, David T. Lewis wrote:
>  
> On Sat, Feb 16, 2013 at 11:48:14AM +0000, Frank Shearar wrote:
> >  
> > http://build.squeak.org/job/InterpreterVM/200/console shows a
> > successful build of r2688
> > http://build.squeak.org/job/InterpreterVM/201/console shows a failed
> > build of r2688
> > 
> > The failure was because the watchdog timer killed the build. I don't
> > know the ins and outs of either the script driving it (I've glanced at
> > it, but haven't taken the time to understand it) nor of VMMaker, but
> > when I've had builds hang it's normally for the extremely unhelpful
> > reason of a dialog popping up on a screen noone will see, either
> > notifying of an error or asking the user a question.
> 
> This is due to some kind of intermittent issue in CommandShell. I've
> never been able to track it down to root cause, but I suspect that
> it's a missed event notification or something related to process
> scheduling. In any case, I put a watchdog into the script so that
> if one of the CommandShell procedures becomes unresponsive, the image
> won't just sit out there forever.
>

Hmmm ... Frank, you just gave me an idea. I suspect that the problem
may be associated with missed aio event notifications (these drive the
CommandShell output). I put another background process into the
VMUnixBuild.st script to tickle the semaphores every once in a while.
Let's see if that makes a difference. If the build script runs for
another week or so without failure, that will confirm my theory and
I'll find a way to fix it in OSProcess/CommandShell.

Thanks,
Dave
 


More information about the Vm-dev mailing list