[mpich-discuss] Crash with hydra on MPI_Finalize

Jain, Rohit Rohit_Jain at mentor.com
Mon Mar 7 19:05:41 CST 2011


Thanks,
Rohit


-----Original Message-----
From: mpich-discuss-bounces at mcs.anl.gov
[mailto:mpich-discuss-bounces at mcs.anl.gov] On Behalf Of Pavan Balaji
Sent: Monday, March 07, 2011 4:02 PM
To: mpich-discuss at mcs.anl.gov
Subject: Re: [mpich-discuss] Crash with hydra on MPI_Finalize


You don't need this option to debug your application. It's only to debug

inside MPICH2. Totalview support for Hydra is automatically always
enabled.

http://wiki.mcs.anl.gov/mpich2/index.php/Using_the_Hydra_Process_Manager
#Debugging

  -- Pavan

On 03/07/2011 05:47 PM, Dave Goodell wrote:
> I suspect that determining which processes in a job are on which hosts
is enabled, as well as message queue inspection.  But I'll admit that I
don't specifically know all of the features that are enabled by this
option.  Perhaps someone else on the list who is more familiar with the
debugger interface (Bill, Pavan?) can chime in.
>
> -Dave
>
> On Mar 7, 2011, at 5:40 PM CST, Jain, Rohit wrote:
>
>> I have filed ticket.
>>
>> Can you please explain when this option needs to be used?
>> 	--enable-debuginfo      Enable support for debuggers
>>
>> What features of totalview debugger won't be enabled, if this option
is
>> not used?
>>
>> Regards,
>> Rohit
>>
>>
>> -----Original Message-----
>> From: mpich-discuss-bounces at mcs.anl.gov
>> [mailto:mpich-discuss-bounces at mcs.anl.gov] On Behalf Of Dave Goodell
>> Sent: Monday, March 07, 2011 11:11 AM
>> To: mpich-discuss at mcs.anl.gov
>> Subject: Re: [mpich-discuss] Crash with hydra on MPI_Finalize
>>
>> On Mar 7, 2011, at 1:04 PM CST, Jain, Rohit wrote:
>>
>>> Yes, seg fault.
>>>
>>> # Attempting stack trace sig 11
>>> # Signal caught: signo [11]
>>>
>>> It crashes with stacktrace that I showed below. There are no error
>>> messages. Crash occurs in all the processes.
>>>
>>> I recently started using --enable-debuginfo, as I read instruction
>>> somewhere (mpich2 docs or totalview) that this option is required to
>> use
>>> mpich2 with totalview. My intention is to be able to debug my
>>> application with totalview, not inside mpich2. But, it seem
totalview
>>> requires some visibility inside mpich2 to make some of its feature
>> work.
>>>
>>> And, you are right. mpich2-1.2.1p1 with mpd also crashes in similar
>> way.
>>>
>>> So, any chances of getting it fixed?
>>
>> Please file a bug report for the issue at
>> http://trac.mcs.anl.gov/projects/mpich2/newticket.
>>
>> If you have a small test program that we can use to reproduce the
>> problem, that will be helpful and will increase the chances of
getting
>> the bug fixed sooner.  Please also note whether it is happening only
>> when you actually are attached with the debugger or if it will also
>> happen without the debugger's involvement.
>>
>> We'll take a look at it as soon as we reasonably can, but I can't
make
>> any promises that it will be fixed very soon.
>>
>> -Dave
>>
>> _______________________________________________
>> mpich-discuss mailing list
>> mpich-discuss at mcs.anl.gov
>> https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss
>> _______________________________________________
>> mpich-discuss mailing list
>> mpich-discuss at mcs.anl.gov
>> https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss
>
> _______________________________________________
> mpich-discuss mailing list
> mpich-discuss at mcs.anl.gov
> https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss

-- 
Pavan Balaji
http://www.mcs.anl.gov/~balaji
_______________________________________________
mpich-discuss mailing list
mpich-discuss at mcs.anl.gov
https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss


More information about the mpich-discuss mailing list