[petsc-dev] plex memory leaks in master

Tobin Isaac tisaac at ices.utexas.edu
Mon Feb 23 14:23:15 CST 2015


On Mon, Feb 23, 2015 at 01:54:08PM -0600, Barry Smith wrote:
> 
> 323a324,334
> > [0]Total space allocated 96 bytes
> > [ 0]96 bytes DMPlexCreateConstraintMatrix_Anchors() line 5852 in /usr/home/balay/petsc.clone-2/src/dm/impls/plex/plex.c
> > [1]Total space allocated 160 bytes
> > [ 1]96 bytes DMPlexCreateConstraintMatrix_Anchors() line 5852 in /usr/home/balay/petsc.clone-2/src/dm/impls/plex/plex.c
> > [ 1]64 bytes DMPlexCreateConstraintMatrix_Anchors() line 5852 in /usr/home/balay/petsc.clone-2/src/dm/impls/plex/plex.c
> > [2]Total space allocated 192 bytes
> > [ 2]96 bytes DMPlexCreateConstraintMatrix_Anchors() line 5852 in /usr/home/balay/petsc.clone-2/src/dm/impls/plex/plex.c
> > [ 2]96 bytes DMPlexCreateConstraintMatrix_Anchors() line 5852 in /usr/home/balay/petsc.clone-2/src/dm/impls/plex/plex.c
> > [3]Total space allocated 160 bytes
> > [ 3]96 bytes DMPlexCreateConstraintMatrix_Anchors() line 5852 in /usr/home/balay/petsc.clone-2/src/dm/impls/plex/plex.c
> > [ 3]64 bytes DMPlexCreateConstraintMatrix_Anchors() line 5852 in /usr/home/balay/petsc.clone-2/src/dm/impls/plex/plex.c
> /home/balay/petsc.clone-2/src/dm/impls/plex/examples/tests
> Possible problem with with runex3_nonconforming_tensor_2, diffs above
> 
> 
> http://ftp.mcs.anl.gov/pub/petsc/nightlylogs/archive/2015/02/22/examples_master_arch-freebsd-cxx-pkgs-opt_wii.log

Fixed in master, merged to next
> 
> Remember to always have the environmental variable 
> 
> export PETSC_OPTIONS=-malloc_test
> 
> this will test for memory leaks in debug versions of the code and will be ignored in optimized so you can always have it set and thus fine memory leaks as soon as you introduce them.

I didn't know about this and will use it now, but it wouldn't have
caught this bug. I had a mismatched PetscMalloc1(), PetscMalloc1(),
PetscFree2().  Coalesced memory isn't used with debugging, so there
was no leak in that case; -malloc_test does nothing without debugging.
The only thing that could catch it is -malloc_dump without debugging.

  Toby

> 
>   Barry
> 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 490 bytes
Desc: Digital signature
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20150223/7108dbb6/attachment.sig>


More information about the petsc-dev mailing list