[Bug target/93494] [8/9/10 Regression] wrong code or ICE with __builtin_add_overflow

jakub at gcc dot gnu.org gcc-bugzilla@gcc.gnu.org
Wed Jan 29 17:29:00 GMT 2020


https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93494

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jakub at gcc dot gnu.org

--- Comment #2 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
This started to ICE with
r10-5922-g3d77686d2eddf76d3498169d0ca5653db45a8662
and doesn't ICE anymore since
r10-6129-gac68e287fc2e939ae6b45ba7ff04e493982b7f62
but already the *.original dump is different,
-  int x = a = REALPART_EXPR <SAVE_EXPR <.ADD_OVERFLOW (0, 0, 0)>>;, (int)
(_Bool) IMAGPART_EXPR <SAVE_EXPR <.ADD_OVERFLOW (0, 0, 0)>>;;
+  int x = a = REALPART_EXPR <SAVE_EXPR <.ADD_OVERFLOW (-(0, 0), 0)>>;, (int)
(_Bool) IMAGPART_EXPR <SAVE_EXPR <.ADD_OVERFLOW (-(0, 0), 0)>>;;
so I'd expect the wrong-code is since then, latent before.


More information about the Gcc-bugs mailing list