https://gcc.gnu.org/bugzilla/show_bug.cgi?id=126362
Bug ID: 126362
Summary: Negative array-new size in requires expression
Product: gcc
Version: 16.1.1
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c++
Assignee: unassigned at gcc dot gnu.org
Reporter: gongke at ios dot ac.cn
Target Milestone: ---
Consider this code (godbolt: https://godbolt.org/z/cPxooPE1d)
```cpp
template <typename T>
concept SmallType = requires { new int[-(int)sizeof(T) + 3]; };
int main() {
static_assert(!SmallType<int>);
return 0;
}
```
GCC produces "static assertion failed" while Clang accepts it.
[expr.new]/8:
> If the expression in a noptr-new-declarator is present, it is implicitly
> converted to std::size_t. The value of the expression is invalid if
> (8.1) the expression is of non-class type and its value before converting to
> std::size_t is less than zero;
> ...
> If the value of the expression is invalid after converting to std::size_t:
> (8.5) if the expression is a potentially evaluated core constant expression,
> the program is ill-formed;
So a negative constant expression size of an array-new expression should be
ill-formed. It should not be checked only after converting to `std::size_t`.
Therefore, either this new expression should produce an error, or this
`requires` expression should evaluate to `false` with `T=int` (I'm not sure
what kind of ill-formed code can and what cannot be tested by `requires`). But
GCC seems to make this evaluate to `true`.
