clang version 5.0.0 (trunk 305664)
Target: x86_64-unknown-linux-gnu
The following code compiles successfully:
template <int... A>
void f() {
([](auto) {
if constexpr (A == 0)
return 42;
else
return 3.14;
}(0), ...);
}
int main() {
f<0, 1>();
}
... but this one doesn't:
template <int... A>
void f() {
([](auto...) { // Variadic lambda
if constexpr (A == 0)
return 42;
else
return 3.14;
}(), ...); // No argument
}
int main() {
f<0, 1>();
}
... yielding:
<source>:7:13: error: 'auto' in return type deduced as 'double' here but deduced as 'int' in earlier return statement
return 3.14;
^
<source>:3:6: note: in instantiation of function template specialization 'f()::(anonymous class)::operator()<>' requested here
([](auto...) { // Variadic lambda
^
<source>:12:5: note: in instantiation of function template specialization 'f<0, 1>' requested here
f<0, 1>();
^
I wouldn't expect different behaviours between an empty argument pack and a dummy argument.
Is there a reason to this discrepancy, or is this a compiler bug?
I believe this is a clang bug.
The rules in [dcl.spec.auto] are, emphasis mine:
If the declared return type of the function contains a placeholder type, the return type of the function is deduced from non-discarded
return
statements, if any, in the body of the function ([stmt.if]).[...]
If a function with a declared return type that contains a placeholder type has multiple non-discarded
return
statements, the return type is deduced for each such return statement. If the type deduced is not the same in each deduction, the program is ill-formed.
One or the other return
statement in the lambda is discarded (the non-taken branch in if constexpr
is called a discarded statement), which leaves only one non-discarded return statement, so the return type of the lambda should be simply deduced from that one left over.
Moreover, clang is okay with just this:
template <int A>
void f() {
[](auto...) {
if constexpr (A == 0)
return 42;
else
return 3.14;
}();
}
int main() {
f<0>();
f<1>();
}
So it's probably some bad interaction with how lambdas work in pack expressions.
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