https://gcc.gnu.org/bugzilla/show_bug.cgi?id=118134
Bug ID: 118134
Summary: std::bit_cast handling of bool anad padding bits is
inconsistent at compile time vs runtime
Product: gcc
Version: 14.2.0
Status: UNCONFIRMED
Keywords: accepts-invalid, wrong-code
Severity: normal
Priority: P3
Component: c++
Assignee: unassigned at gcc dot gnu.org
Reporter: pinskia at gcc dot gnu.org
Target Milestone: ---
Take:
```
#include <bit>
constexpr auto c = (char)('A'&~0x1);
constexpr bool A = std::bit_cast<bool>(c);
static_assert(!A);
bool t = A;
int main()
{
char c1 = c;
bool t1 = std::bit_cast<bool>(c1);
__builtin_printf("%s\n", A ? "true" : "false");
__builtin_printf("%s\n", t ? "true" : "false");
__builtin_printf("%s\n", t1 ? "true" : "false");
}
```
GCC accepts the above even though the bool representation is only 0x01/0x00 are
valid as far as I know. So at constexpr time, GCC thinks bool has padding bits.
While runtime we get a different answer, true.
