Catching function types turns into catching their respective pointers

[Alcaro]

template<typename T>
void fn()
{
    try {}
    catch (T) {}
}
int main() { fn<int(int)>(); }

https://godbolt.org/z/K8rdsKbMT

Expected: Error. T is a function type, can't have a variable like that.

Actual: Clang thinks it says catch (T*), as can be seen if you add a second catch. https://godbolt.org/z/q7T43oW6n

(Fun fact: GCC is equally wrong in the opposite direction, by erroring so hard it segfaults. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=120754)

[llvmbot]

@llvm/issue-subscribers-clang-frontend

Author: None (Alcaro)

```c++ template<typename T> void fn() { try {} catch (T) {} } int main() { fn<int(int)>(); } ```

https://godbolt.org/z/K8rdsKbMT

Expected: Error. T is a function type, can't have a variable like that.

Actual: Clang thinks it says catch (T*), as can be seen if you add a second catch. https://godbolt.org/z/q7T43oW6n

(Fun fact: GCC is equally wrong in the opposite direction, by erroring so hard it segfaults. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=120754)

[AaronBallman]

template
void fn()
{
try {}
catch (T) {}
}
int main() { fn<int(int)>(); }

https://godbolt.org/z/K8rdsKbMT

Expected: Error. T is a function type, can't have a variable like that.

Actual: Clang thinks it says catch (T*), as can be seen if you add a second catch. https://godbolt.org/z/q7T43oW6n

Closing as not a bug because Clang's behavior is correct per the standard: https://eel.is/c++draft/except#handle-2

[shafik]

Other implementation also agree w/ clang here: https://godbolt.org/z/vx4c379e4

[Alcaro]

Understood, will update the corresponding GCC bug. Should've checked the spec, not just seen the disagreement and guessed which behavior is more intuitive. Sorry about the noise.

[AaronBallman]

Understood, will update the corresponding GCC bug. Should've checked the spec, not just seen the disagreement and guessed which behavior is more intuitive. Sorry about the noise.

No worries! The spec usually isn't so clear as it happened to be in this case. (My initial intuition was that you were right.) :-D