[Vm-dev] Re: [Pharo-dev] How do diagnose image locks up (cpu 100%) on save?

Eliot Miranda eliot.miranda at gmail.com
Fri Aug 23 21:29:27 UTC 2013

Hi Norbert,

On Fri, Aug 23, 2013 at 3:45 AM, Norbert Hartl <norbert at hartl.name> wrote:

> strange but true I have a similar problem as of today. I don't have RFB
> installed I just installed zinc and use it. I can reproduce the behavior
> partially:
> Opening the image and saving works. Opening, starting a zinc server does
> as well. But opening, starting the zinc server and issue a request from a
> browser freezes the image when saving it. If I only issue one request from
> a browser the image freezes for something between half a minute and a
> minute. That smells like a timeout problem to me. The issue requested from
> the browser ends in "self halt" so there is an exception going on. I didn't
> switch zinc into debugMode for this.
> I wanted to get some more information in the loop by issuing a USR1 signal
> to the vm when it hangs. But in my case it does not write a dump file into
> my working directory.
> This should be assured behavior that whenever a USR1 signal is received by
> the vm that it always writes a file? I have plenty of space left on my
> device.

Weeelllll....  the problem is that if the signal is received during the
right phase of garbage collection (specifically the pointer-reversal mark
phase) then printing the back-trace can cause a crash.  The VM writes the
stack trace to crash.dmp first, and then writes it to stdout.  So if you
see an incomplete crash.dmp file you can probably safely infer that the VM
crashed when trying to print the stack trace.

The VM could be written to defer the stack trace generation until GC
finishes, which would be safer.  It could perhaps immediately generate "in
GC, stack trace generation deferred" or some such.  Whne time allows I'll
get to this (volunteer efforts welcome).

> Norbert
> Am 23.08.2013 um 02:13 schrieb Paul DeBruicker <pdebruic at gmail.com>:
> > So when you open the image I posted and in the workspace run
> >
> > RFBServer start.
> > Smalltalk snapshot: true andQuit: false.
> >
> >
> > Everything works fine?  It doesn't go to 100% cpu use?
> >
> >
> >
> > --
> > View this message in context:
> http://forum.world.st/How-do-diagnose-image-locks-up-cpu-100-on-save-tp4704639p4704698.html
> > Sent from the Pharo Smalltalk Developers mailing list archive at
> Nabble.com.
> >

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.squeakfoundation.org/pipermail/vm-dev/attachments/20130823/dc743b96/attachment.htm

More information about the Vm-dev mailing list