[
Date Prev][
Date Next][
Thread Prev][
Thread Next][
Date Index][
Thread Index]
[
List Home]
|
Re: [ptp-dev] Problem with SDM debugger master process startup
|
Dave,
I've just committed a change to the SDM to fix this problem. I can now
debug an MPI program launched by PE. There are still a couple of
problems with breakpoints and displaying threads, but it's looking
good so far. I'm running a build now.
Greg
On Oct 23, 2008, at 11:37 AM, Dave Wootton wrote:
I found a problem with startup of the SDM master process that
appears to
be timing related. My PE proxy code waits for the PE attach.cfg file
to be
created, which means all PE application processes have started, before
attempting to start the SDM master process. Debug startup hangs at
this
point, and if I look at the console log, I see a message 'No port
found
for the sdm child. hostname: k17sf2p03'
If I call 'sleep(5) after the attach.cfg file is cretaed but just
before
invoking the SDM master process, then the debugger starts reliably.
This acts as if there is enough of a delay between when the SDM child
processes are invoked and when they open a port for the sdm master to
connect to them. Should the SDM master code that is trying to
connect to
the child SDM processes have some retry logic with an eventual
timeout if
it cannot connect to the child SDM processes?
Right now, I have kept the 5 second sleep() invocation in the code
in the
patch I just submitted for bug 251853 so that my proxy can start the
debugger.
Dave
_______________________________________________
ptp-dev mailing list
ptp-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/ptp-dev
