Can't construct int32_t vector with 0xFFFFFFFF

[Nemi26]

#include <stdint.h>

using ivec3 = int32_t __attribute__((ext_vector_type(3)));

void ok() {

    const int32_t x = 0xFFFFFFFF; // ok
    const ivec3 mask1 = { 0, 0xFFFFFFFF, 0 }; // error!!!

    const ivec3 mask2 = { 0, x, 0 }; // ok
}

https://godbolt.org/z/z165s9GMG

[keinflue]

This is not specific to the vector extension.

In C++11 and later narrowing conversions in braced initialization is not permitted. 0xFFFFFFFF isn't representable by int32_t, so the conversion to it is narrowing.

The same can be observed with

    const int32_t x = { 0xFFFFFFFF };

In other words, I do not see anything wrong here, it looks consistent with initialization in other C++ contexts.

[llvmbot]

@llvm/issue-subscribers-clang-frontend

Author: Nemi (Nemi26)

```cpp #include <stdint.h>

using ivec3 = int32_t attribute((ext_vector_type(3)));

void ok() {

const int32_t x = 0xFFFFFFFF; // ok
const ivec3 mask1 = { 0, 0xFFFFFFFF, 0 }; // error!!!

const ivec3 mask2 = { 0, x, 0 }; // ok

}

https://godbolt.org/z/z165s9GMG
</details>

[Nemi26]

I stand corrected