On Tue, 10 Mar 2009, Tim Kroeger wrote:

> Talking about asserts, what do you think about the attached patch?

A very good idea - I'll commit it now.

> (I'm now using METHOD=devel and try to catch my crash using asserts, and I 
> suspect now that it's in the ghosted part of PetscVector somewhere.)
>
>> libMesh makes it easy enough to keep simultaneous
>> debug and optimized builds around; IMHO every software package should
>> do the same.
>
> Well, our software package tries to make that easy as well.  It's just so 
> huge a package that it's still difficult to work with several versions.

Understandable.  Again, I can't recommend distcc and "make -j 50"
highly enough.  (as long as you're using a single compiler, anyway -
distcc is a little tricky to get working with multiple installed
versions of the same compiler)

> Anyway, METHOD=devel works well with the non-debug version of our software.

Sorry I never mentioned that.  I always find myself wanting debugging
stuff turned all the way on or all the way off, but John swears by
METHOD=devel, and I suppose it can be pretty useful.

> Actuall, it causes my application to crash at a completely different point, 
> even before any NumericVectors are used.  Would you please have a look at the 
> attached test.cpp; it creates a grid, partially refines 5 times and then 
> completely coarsens it again.  To actually delete the elements that are not 
> required, it calls Mesh::contract(), which crashes for me with METHOD=devel. 
> (It runs well with METHOD=opt, though.)

Thanks!  This is crashing for me, too, and it definitely shouldn't be.
Looks like something going wrong with find_neighbors and subactive
elements, which is 99% likely to be my fault - I only have a little
time to check it out today, but I'll give it a shot tonight and
tomorrow as well if I can't find the bug right away.
---
Roy

------------------------------------------------------------------------------
_______________________________________________
Libmesh-devel mailing list
Libmesh-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/libmesh-devel

Reply via email to