Hi List,the attached test program (bsm-db.cc) always dies in malloc called from opal_free_list (backtrace in error.txt and valgrind (vg*) output can be found in the tar file). It seems that there is an invalid write in ompi_osc_pt2pt_sendreq_send. I checked the derived datatype using the checkdt.plx script and can confirm that there is no overlap and the window on the target side is large enough. The segfault occurs on the origin side.
I would be glad if someone could try to reproduce the problem. I'm using the trunk (revision 20157 from 12/19/08).
Unfortunately the code (which has been extracted automatically from a real-life application) is not really readable, sorry for that ...
Many thanks + Best regards Dorian
ompisigsegv.tar.gz
Description: GNU Zip compressed data
