[section.label] Clarify how class rule has to be applied to class templates so that IFNDR can be distinguished from IF for template instantiations

[ranaanoop]

Full name of submitter: Anoop Rana

Reference (section label): [class.copy.ctor]

Issue description: Consider the following code that used to compile with clang but not with gcc and msvc: Demo

template<typename T>
struct C {
    C() {}
    C(C<bool>);  
};
C<bool> c;    //compiles with clang but not with gcc and msvc

We do have [class.copy.ctor] that seems to make it ill-formed:

A declaration of a constructor for a class X is ill-formed if its first parameter is of type cv X and either there
are no other parameters or else all other parameters have default arguments

Note that there is also [temp.res](temp.res#5 in particular) dealing with template instantiations. Now, either the rule from [class.copy.ctor] doesn't apply and the program is well-formed or it does apply to the instantiation and then it is ill-formed and the only question is whether it is IFNDR. I think it is more likely not to apply interpreting the exact wording, although it is probably intended to apply. I guess one could argue that technically a hypothetical instantiation of A<bool> would be ill-formed in the latter case due to a non-dependent construct, so that it is IFNDR. But that wouldn't really be intended I would say. [temp.res]/5 doesn't seem well-worded to me for this particular case.

[frederick-vs-ja]

Clang was buggy before. The bug was fixed in Clang 21 (llvm/llvm-project#130866).

I guess one could argue that technically a hypothetical instantiation of A<bool> would be ill-formed in the latter case due to a non-dependent construct, so that it is IFNDR.

That bullet is in a note, so it doesn't override the normative wording. I believe "a construct ([...]) that does not depend on a template parameter" should be interpreted as "a construct ([...]) that whose well-formedness does not depend on a template parameter".

Also, such reading is unacceptable, as it would make too many conditionally ill-formed usages of class templates IFNDR.

I guess we can reword that bullet editorially.