[Carpet] potential integer overflow in Grid setup

[Thomas Radke]

Erik Schnetter wrote:
> Interesting.  This test checks whether one edge of the box is longer  
> than INTMAX/2.  Either this machine uses 4 bit integers, or the  
> corresponding STL entries to determine the limits of data types are  
> broken.
> 
> There is a bit of a complication since the minimum of an integer  
> quantity is a large negative number (the smallest integer), whereas  the 
> minimum of a real quantity is a very small positive number (the  
> smallest POSITIVE real).  It could thus be that the distinction  between 
> integers and reals doesn't work.  I don't assume there is a  core file...?

There is one in 
peg304:/scratch/tradke/Cactus/TEST/PublicThorns/CarpetIOHDF5/coredir.0/ 
but it's not usable; the stacktrace is only one level deep and ends in 
pthread_kill().

I think the easiest is to build a small test configuration 
(MakeThornList 
arrangements/Carpet/CarpetIOHDF5/par/CarpetWaveToyCheckpointTest.par) on 
Pelican and play with that.

-- 
Cheers, Thomas.