Hi Michael,
could you also cherry-pick 4ef64ad1a to releases/gcc-16? The gcc-16 branch was cut on April 22nd, seven weeks before the commit landed on master, so the 16 series is the one gap: gcc 15.3 has the fix, master (17) has it, but 16.1/16.2 ship without microblaze signal-frame unwinding. Anyone moving from 15.3 to 16.x gets a regression -- pthread cancellation cleanup stops working. Verified with gcc 16.2.0 against the uClibc-ng test suite: 12 tests fail (tst-cancelx*/tst-oncex*, "cleanup handler not called"), and all pass again with the commit cherry-picked onto 16.2.0. best regards Ramin
