On Wed, 19 May 2010, Tim Kroeger wrote: > When I now read in results that have been written with the old > version, I get completely wrong results. If I re-run the simulation > and then read in the result, everything is fine again. Hence, I > assume that there has been a format change in the xdr files; can this > be? If yes, is there any way to read in files that have been written > in the old format? That would be very useful for me since otherwise I > would have to repeat quite a lot of simulations.
There's been an ordering change in the xdr files. We ran into a bug in our libHilbert interface (responsible for providing a consistent node/DoF ordering for I/O) a while back which could cause assertion failures in dbg mode or corruption of output files in opt mode. (very infrequently, but of course any probability above 0% is too high for such a nasty bug) So --disable-libHilbert is now the default until we get that fixed. Ben's the one most qualified to fix it, so I've left that on his plate, but he's been swamped lately; apparently the Space Shuttle takes precedence for some reason. ;-) But we should at least get the new code properly reading old files. Would you mind helping out with that, since you've already got a failure case prepared? I'll try to get things working on a small test case myself this week, and I'll let you know when the svn head is in shape to take a crack at your own old files. --- Roy ------------------------------------------------------------------------------ _______________________________________________ Libmesh-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/libmesh-users
