[mpich-discuss] Bug: MPICH2 + valgrind fails with VALGRIND_OPTS

[Dave Goodell]

I can reproduce this.  For some reason hydra (or, less likely, the sub-shell?) is stripping out the "=no" in the env var that is propagated to valgrind:

----8<----
% export VALGRIND_OPTS=--memcheck:error-limit=no
% env | grep VALGRIND
VALGRIND_OPTS=--memcheck:error-limit=no
% mpiexec -n 1 env | grep VALGRIND
VALGRIND_OPTS=--memcheck:error-limit
----8<----

We'll take a look and see where this is happening.

-Dave

On Feb 17, 2011, at 4:16 PM CST, Saurabh T wrote:

> 
> 1. mpiexec -np 2 valgrind --memcheck:error-limit=no true
> Works without issue
> 
> What should be identical:
> 2. export VALGRIND_OPTS=--memcheck:error-limit=no
> mpiexec -np 2 valgrind true
> Fails with 
> Bad option --error-limit, aborting
> ONE OF THE PROCESSES TERMINATED BADLY: CLEANING UP
> etc.
> 
> saurabh
> 		 	   		  
> _______________________________________________
> mpich-discuss mailing list
> mpich-discuss at mcs.anl.gov
> https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss