https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82338
--- Comment #2 from Vladimir Makarov <vmakarov at gcc dot gnu.org> --- Thank you for reporting this, especially for reducing the test case. The bug is not dangerous, it does not result in wrong code generation but it might result in worse code. I reproduced it. It seams the validity (obsoleteness) of data was not checked first. I'll make a patch and commit it tomorrow. If I am not able to make it on this week, I'll do it in a week.
