Crash on IDaaas using single crystal large machines

[davidvoneshen]

User has reported crashes when trying to do a tobyfit simulation of their phonon spectra. Running their code it works just fine on powder and single crystal machines but falls over at the first step on single crystal large machines producing 24 matlab error reporters. It has also on one occasion caused the machine interface to fail its health check. The cut is of modest size, 6.3M qpoints, phonopy calculation, MC points set to 2 and LET as the instrument. Example script available but don't want to post it publicly.

[mducle]

After some investigations it seems that the issue is in the Euphonic C-module. It's causing a segfault when the chunk size is larger than $\approx 1$ M q-points.

Since the code uses pointer arithmetic I suspect that we're having an signed integer overflow error somewhere as we use an int to index into all the arrays.