Hi, I did a quick test for exceptions on -current (which worked---that was expected with both the _much_ newer GCC including Macie's and Kalvis's work) and then generated some more tests (claude opus) as well as ran exception-relevant ATF tests (/usr/tests/usr.bin/c++/t_call_once2 with call_once2 call_once2_pic call_once2_pie call_once2_static and /usr/tests/atf/atf-c++/macros_test with require_throw require_throw_re) so we'd close this very old PR. While my manual test with a simple exception worked, the claude-generated tests showed that we're leaking stack space in a corner case when we're catching exceptions in a loop. (That seems to be "fixed" by RETurning from the function, though.) I'l attach the small test cases, build/run by `make run`. Shall we try to fix this remaining issue? Or just ignore it as an unexpected corner case? MfG, JBG --
Attachment:
pr44155-tests.tar.gz
Description: application/gzip
Attachment:
signature.asc
Description: PGP signature