https://gcc.gnu.org/bugzilla/show_bug.cgi?id=118742
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|UNCONFIRMED |NEW
Last reconfirmed| |2025-02-04
Ever confirmed|0 |1
--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
Smaller testcase:
_Bool __GIMPLE (ssa,startwith("optimized")) g1(int i)
{
_Bool _1;
_Bool _2;
__BB(2,guessed_local(97603132)):
_1 = i_3(D) == 1;
_2 = ~_1;
return _2;
}
the issue is that we do
case CPP_COMPL:
c_parser_consume_token (parser);
op = c_parser_gimple_postfix_expression (parser);
return parser_build_unary_op (op_loc, BIT_NOT_EXPR, op);
and parser_build_unary_op wraps op. We're defering to the C parser
GENERIC building to emit diagnostics, maybe that's a bad choice here.
This function wraps build_unary_op which has a 'noconvert' argument
which should be set for all invocations from the GIMPLE FE, we can
either pass in an additional arg, somehow extract from state that we're
parsing a GIMPLE function or provide our own wrapper around build_unary_op.
