https://gcc.gnu.org/bugzilla/show_bug.cgi?id=121603
Bug ID: 121603
Summary: When z is a variable of type _BitInt(64), gcc -m32
evaluates _Alignof(z) to 8
Product: gcc
Version: 15.2.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c
Assignee: unassigned at gcc dot gnu.org
Reporter: pascal_cuoq at hotmail dot com
Target Milestone: ---
My colleague Antonin Decimo has noticed a difference between Clang and GCC on
the example below. And we think that Clang might be correct.
_BitInt(64) z;
int alignof_bi64 = _Alignof(_BitInt(64));
int alignof_z = _Alignof(z);
Using GCC 15.1, "gcc -O2 -std=c23 -m32" generates:
alignof_z:
.long 8
alignof_bi64:
.long 4
With the same options, Clang 20.1.0 generates:
alignof_bi64:
.long 4
alignof_z:
.long 4
CE link: https://gcc.godbolt.org/z/KGnxMPhsh
We expect the value 4 for alignof_bi64 (recall that when generating IA-32,
_Alignof(long long) is 4 for historical reasons, and we think that
_Alignof(_BitInt(64)) has no reason to be different).
Although applying _Alignof to an lvalue is a GCC extension, we do not
understand why it would produce the different result of 8. If this is intended,
we can report it on the Clang side as a difference between the two compilers.
