Question

I was just over specializing std::hash for a user-defined type using:

template<>
struct hash<...> {...};

When VC10 greeted me with the warning:

warning C4099: ‘std::hash<_Kty>’: type name first seen using ‘class’
now seen using ‘struct’

and I found out that its standard library declares std::hash as class, whereas the standard (or the latest free draft I have) declares it as struct.

Well, of course I know that a struct isn’t any different from a class (except for the different default access and inheritance types). But my questions are:

1. Does VC10 violate the standard here or is it free to exchange structs for classs in any standard library components (as long as the required access types for members stay consistent, of course)?
2. Is it legal to specialize a template class as a struct and vice versa or does this bring problems with name resolution and the like (at least VC10 thinks it’s worth a warning).

Answer 1

First off, here is the answer to 2. taken from 14.5.1 [temp.class] paragraph 4:

In a redeclaration, partial specialization, explicit specialization or explicit instantiation of a class template, the class-key shall agree in kind with the original class template declaration.

However, struct and class are referring to the same class-key according to 7.1.6.3 [dcl.type.elab] paragraph 3 last sentence:

The class-key or enum keyword present in the elaborated-type-specifier shall agree in kind with the declaration to which the name in the elaborated-type-specifier refers. […] Thus, in any elaborated-type-specifier, the enum keyword shall be used to refer to an enumeration, the union class-key shall be used to refer to a union, and either the class or struct class-key shall be used to refer to a class declared using the class or struct class-key.

Trying g++, clang, and EDG all agree that it is possible to specialize a template declared as struct as a class. However, clang warns about having changed from struct to class or vice versa. Based on this, the standard library is free to choose whatever keyword it sees fit for the definition. Obviously, if the compiler rejects the code as a result something is seriously broken but I’d think it is the compiler rather than the library which is at error in this case.