[Bug c/86308] [7/8/9 Regression] verify_gimple failed in tree-cfg.c

[msebor at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=86308

Martin Sebor <msebor at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|ice-on-valid-code           |ice-on-invalid-code
             Status|NEW                         |ASSIGNED
           See Also|                            |https://gcc.gnu.org/bugzill
                   |                            |a/show_bug.cgi?id=86125,
                   |                            |https://gcc.gnu.org/bugzill
                   |                            |a/show_bug.cgi?id=86202
           Assignee|unassigned at gcc dot gnu.org      |msebor at gcc dot gnu.org

--- Comment #3 from Martin Sebor <msebor at gcc dot gnu.org> ---
The ICE is due to the same "feature" of the C front-end where incompatible
declarations of built-in functions are accepted when they should be rejected. 
The ICE doesn't trigger with -m64 because the incompatible declaration of the
builti-in char* index(const char*, int) function is rejected because
sizeof(int) != sizeof(char*) (i.e., the TYPE_MODE of each type is different).

The code is invalid, of course, because the index() declaration in the program
is incompatible with that of the built-in.  It's a deficiency in
-Wbuiltin-declaration-mismatch that the incompatibility isn't diagnosed (this
is tracked n bug 86125)

In the patch I posted for pr86125
(https://gcc.gnu.org/ml/gcc-patches/2018-06/msg00804.html) I tried to preserve
the dubious feature of the C front-end but I now think it would be better to
reject incompatible declarations regardless of TYPE_MODE.

(Strangely, my bisection points to r240783 (gcc 7.0.0) as the first revision to
ICE, even though that revision looks completely unrelated.)