http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49479
--- Comment #1 from Joost VandeVondele <Joost.VandeVondele at pci dot uzh.ch> 2011-06-20 20:04:47 UTC --- the difference (I think) seems to be in the present check that has become if (data != 0B && (integer(kind=4)[0:] * restrict) data->data != 0B) instead of the original if (data != 0B)