https://gcc.gnu.org/bugzilla/show_bug.cgi?id=117730
--- Comment #2 from anlauf at gcc dot gnu.org --- (In reply to Paul Thomas from comment #1) > Harald and Andre: If you have any ideas on how to fix it, please let me know > or take it over. The testcase fails on all branches that I have to hand but > works fine with nagfor and ifort. I have no idea, but played around: - changed the main so that the function result is first stored to a variable -> looks like we get random junk as result - removed all PURE attributes and add print statements The critical procedure is child_reset: adding a print there "fixes" the bug - The dump-trees do not seem helpful - the module file module1.mod has an IMPLICIT_PURE for child_reset even if I add an explicit IMPURE! Is that a bug or feature? I need to add a print (helps) or a save statement to a temporary variable (does not help) to get rid of IMPLICIT_PURE - the IMPLICIT_PURE is likely a red herring, as SAVE does not help Strange.
