On Mon, Dec 29, 2008 at 11:28:07PM +0100, Anders Logg wrote: > On Mon, Dec 29, 2008 at 02:35:58PM +0100, Johan Hake wrote: > > On Monday 29 December 2008 14:26:56 Anders Logg wrote: > > > On Mon, Dec 29, 2008 at 02:23:55PM +0100, Johan Hake wrote: > > > > On Monday 29 December 2008 14:11:17 Anders Logg wrote: > > > > > This demo works for me, both with and without DOLFIN_NOPLOT. > > > > > I had to run instant-clean to make it work (don't know why). > > > > > > > > > > Try and see if that helps. > > > > > > > > I have done that several times with no luck. I have tried valgrind, with > > > > no directly usefull output (found some MPI memory leaks though) and also > > > > tried gdb, but the segfaults comes from within some python library code > > > > and I cannot find any usefull there. > > > > > > > > I am compiling swig from svn, but I assume that the buildbot is using a > > > > release swig and it also gets the segfault. > > > > > > > > I am running hardy with python version 2.5.2, numpy 1.1.1 from Andrew > > > > Straws repositories and gcc-4.2.4. > > > > > > > > Johan > > > > > > Can you isolate the step that goes wrong in the demo? (Comment out > > > everything and gradually uncomment.) > > > > If I uncomment > > > > PIv = interpolate(v, P1) > > > > and the related plot command it does not segfaults. A print statement in > > the > > end of the script will print, so it does not segfaults in our code, but we > > could of course be the one to blame. I therefore suspect that it could be > > some memory that we haven't freed properly. > > > > Johan > > I also get a segmentation fault now, but it only appears the first > time I run the demo after running instant-clean. If I run it again it > works...
ok, this should be fixed now. It was a bug in the interpolation between function spaces which lead to us writing out of memory. See if it works. I expect (hope) the buildbots to go green in a few minutes. -- Anders
signature.asc
Description: Digital signature
_______________________________________________ DOLFIN-dev mailing list [email protected] http://www.fenics.org/mailman/listinfo/dolfin-dev
