-Wdtor-name incorrect on inner class

[JVApen]

Bugzilla Link	46979
Version	11.0
OS	Windows NT
CC	@dwblaikie,@zmodem,@JVApen,@poletti-marco,@zygoloid

Extended Description

See https://compiler-explorer.com/z/vY5xo8

// clang++ -Wdtor-name t.cpp
class A
{
    template <typename T>
    class B;
};

template<typename T>
class A::B
{
    ~B();
};

template<typename T>
A::B<T>::~B()
{

}

// Result:

<source>:14:8: warning: ISO C++ requires the name after '::~' to be found in the same scope as the name before '::~' [-Wdtor-name]

A::B<T>::~B()

~~~~~~~^~

       ::B

As far as I can see, this code is correct and this warning is giving a false positive.

[zygoloid]

The diagnostic is correct, per the standard rules as written; strictly-speaking, the C++ rules require this destructor to be written as

template<typename T>
A::B<T>::~B<T>()

Per C++ [basic.lookup.qual]p6:

"""
In a qualified-id of the form:

nested-name-specifier[opt] type-name :: ~ type-name

the second type-name is looked up in the same scope as the first.
"""

This means that the second B is looked up in class A, which finds only the class template B and not the injected-class-name.

This is not an especially useful rule, and likely isn't the intended rule, which is why this diagnostic (along with a bunch of similar diagnostics for plausible but formally incorrect destructor names) is disabled by default but included in -pedantic.

I've messaged the C++ committee to see whether a rule change is warranted here.

[JVApen]

@​RichardSmith If that's indeed the correct way of writing the dtor, the fix-me for this warning should be fixed to suggest replacing ~B by ~B. Do you consider this part of this bug, or should I log a separate one for that?

[JVApen]

@​RichardSmith: Does the same then hold for Ctor as well, and why don't we have a warning on that?

[zygoloid]

The rules for constructors are different; the is not necessary there. It would certainly seem like a good idea to produce a better fixit or no fixit; I don't think we want to encourage people to write A::B::B::~B().

[zmodem]

Is there something to be done here for clang 11?

[zygoloid]

Given that Clang is technically correct and that the warning in question is disabled by default, I don't think doing anything further about this should be considered release-blocking.

[zmodem]

Makes sense. Thanks!

[llvmbot]

How does this interact with DR 2237? Based on this DR,
GCC 11 now rejects something like the example below.
Who is correct about this, is it GCC or Clang?

[llvmbot]

struct Outer
{
  template<typename TYPE>
  struct Inner { ~Inner(); };
};

template<typename TYPE>
Outer::Inner<TYPE>::~Inner<TYPE>() {}

[zabereer]

This also fails in namespace scope:

namespace test
{
    template<typename T> struct A;
}

template<typename T>
struct test::A
{
    A();
    ~A();
};

template<typename T>
test::A<T>::A()
{
}

template<typename T>
test::A<T>::~A()
{
}

( https://godbolt.org/z/WWv3474sP )
I think Clang is incorrect.
Reading https://eel.is/c++draft/basic.lookup.qual I cannot find the referenced section about repeating the template params after the destructor. Also why only the destructor requires a repeat of template args.