Hi all,

On Mon, Jun 08, 2009 at 08:09:43AM -0400, Jeff Squyres wrote:

> In short: perhaps you could setenv OMPI_MCA_disable_memory_allocator to 
> 1, or somesuch.  I can easily provide you with a patch (that we'd then 
> also commit upstream, but you'll need the patch until we include this 
> feature in a release) for such a fix.

I uploaded this change to Debian's openmpi packages yesterday, patch below.
Jeff: will you commit this upstream?

Fakeroot maintainer: will you set OMPI_MCA_disable_memory_allocator to 1
in the fakeroot environment?

Thanks,
-Steve


diff -u -r openmpi-1.3.2/opal/mca/memory/ptmalloc2/hooks.c 
../openmpi-1.3.2/opal/mca/memory/ptmalloc2/hooks.c
--- openmpi-1.3.2/opal/mca/memory/ptmalloc2/hooks.c     2009-04-21 
22:51:03.000000000 -0500
+++ ../openmpi-1.3.2/opal/mca/memory/ptmalloc2/hooks.c  2009-06-17 
21:55:54.000000000 -0500
@@ -725,6 +725,14 @@
     check_result_t lpp = check("OMPI_MCA_mpi_leave_pinned_pipeline");
     bool want_rcache = false, found_driver = false;
 
+    /* Allow user to disable memory allocators.  We check this and
+     * exit early because in some environments (e.g. Debian fakeroot
+     * the stat() calls below will cause a memory allocation, which
+     * is deadly.  See http://bugs.debian.org/531522
+     */
+    if ( RESULT_YES == check("OMPI_MCA_disable_memory_allocator") )
+       return;
+
     /* Look for sentinel files (directories) to see if various network
        drivers are loaded (yes, I know, further abstraction
        violations...).

Attachment: signature.asc
Description: Digital signature

Reply via email to