https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106692
--- Comment #14 from Paul Thomas <pault at gcc dot gnu.org> --- (In reply to anlauf from comment #12) > This very local, hackish patch marks the Cray pointers as volatile only > for comparisons, fixes the testcase and does not regress with the analyzer > testcases. Not tested further. > > What do others think? Hmmm! To be honest, I think that the fix is too hackish by half. I presume that making the pointer TREE_VOLATILE runs the risk of unwanted behaviour further down stream from the conditional expression? For the record, the last version of gfortran that I have, which succeeds with the testcase, is GNU Fortran (GCC) 6.4.1 20180703. It fails with GNU Fortran (GCC) 7.4.1 20191027. Paul
