------- Comment #7 from jv244 at cam dot ac dot uk 2007-04-26 05:54 ------- (In reply to comment #6)
> Regarding the original problem with CP2k: A dependence on the optimization > options usually indicates a middle-end problem rather than a (Fortran) > front-end problem. As it occurred around the time I got core-dumps for the > Polyhedron test cases, the easiest is probably to wait until PR 31699 and PR > 31697 are fixed. If we are lucky, CP2k is then fixed as well. If not, well, > then one can still dig for the problem. I'm rather sure the CP2K problem is identical to your PR 31699 and PR 31697. So, from this point of view you can even close this PR what concerns me, and add 31699 and 31967 as blockers for PR 29975 You can also leave this open so that gfortran, one day, produces the NAG error for the unallocated ncoset, but I think that this error is low priority, in this case you can remove that it blocks PR 29975 Furthermore, I think you should consider to reopen PR 31688. There really should be no such warnings about frontend generated variables, this is different from false positives for user variables -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=31683