https://gcc.gnu.org/bugzilla/show_bug.cgi?id=122470
Andrew Pinski <pinskia at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Ever confirmed|0 |1
Status|UNCONFIRMED |NEW
Last reconfirmed| |2025-10-29
--- Comment #3 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Mason from comment #2)
> Questions for WG14 & WG21 experts (all 7 of you)
>
> Why is foo3 handled differently in C & C++ ?
>
> Why is foo2 not equivalent to foo3 in C++ ?
It is handled differently because the front-end handles the assignments
slightly differently. No other reasons.
In the case of foo3 for C++, the C++ front-end produces:
```
<<cleanup_point struct s1 vv = {.f_1=0};>>;
<<cleanup_point <<< Unknown tree: expr_stmt
vv.f_2 = (<unnamed-unsigned:24>) u >>>>>;
<<cleanup_point <<< Unknown tree: expr_stmt
(void) (*NON_LVALUE_EXPR <out> = *(const struct s1 &) &vv) >>>>>;
```
While the C front-end produces:
```
struct s1 vv = {.f_1=0, .f_2=(<unnamed-unsigned:24>) u};
*out = vv;
```
Which gets gimplified slightly differently.
For the C++ one:
vv = {};
_1 = (<unnamed-unsigned:24>) u;
vv.f_2 = _1;
While C gets:
vv.f_1 = 0;
_1 = (<unnamed-unsigned:24>) u;
vv.f_2 = _1;
Which only the C++ IR here produces decent results but only on x86_64 with the
RTL optimizers.
foo3 with the C++ front-end on aarch64 produces:
```
mov w3, 0
mov w2, 0
bfi w2, w3, 0, 8 // still 0
bfi w2, w1, 8, 24 // shift really
str w2, [x0]
```
