struct X
{
enum class E
{
A,B
};
static constexpr X A()
{
return X{E::A};
}
static constexpr X B()
{
return X{E::B};
}
constexpr operator E() const
{
return a;
}
E a;
};
template <typename T>
struct Y
{
void f()
{
// without this line clang errs
// const auto & x = this->x;
switch(x)
{
case X::A():
case X::B():
default: return;
}
}
X x = X::A();
};
int main()
{
Y<int>{}.f();
}
Without the marked line in the snippet clang gives the following error:
error: case value is not a constant expression case
X::B():
However I tried gcc and it compiled fine. Anybody knows if gcc is being lenient or clang has some bug?
See on godbolt (clang 8.0.0): https://godbolt.org/z/ETe5WQ However (gcc 8.3) compiles fine (also on godbolt) and tried other versions of gcc and were also fine
Update:
opened a bug
Clang (8.0.0) has a bug here.
If you write constexpr auto A = X::A();
and use case A:
in your switch
statement, you get the same compile error (saying that A
is not a constant expression).
If you remove the cases, however, it compiles fine (which implies that A
is a valid constexpr
=> a contradiction to the previous error).
Moreover, switch(x)
fails while switch(this->x)
succeeds. Since x == this->x
in your case, this is definitely a bug.
As chtz mentioned, clang (5/6) seem to work just fine. That's not an argument, but an apparent regression.
Update: As mentioned by the OP, they filed a bug report.
If you love us? You can donate to us via Paypal or buy me a coffee so we can maintain and grow! Thank you!
Donate Us With