Is template-name a deduced context?

2019-04-07 03:19发布

问题:

[temp.deduct.type] paragraph 8 lists all deduced contexts, but it seems not to include template-name<TT> where template-name refers to a class template and TT refers to a template template argument. Is this a deduced context?

If it is, why?

If not, consider the following code:

template<template<typename> class U, template<typename> class V>
struct foo {};

template<template<typename> class U>
struct foo<U, U> {}; 

int main() {}

This code compiles under Clang 7.0.0 and GCC 8.0.1, which means the compilers consider the partial specialization is more specialized than the primary template, which means U and V in the primary template are successfully deduced against foo<U, U>. Is this a compiler bug?

回答1:

This paragraph is full of problems, including the one you pointed out. Core issue 2328 has a nice list:

The presentation style of 17.9.2.5 [temp.deduct.type] paragraph 8 results in a specification that is unclear, needlessly verbose, and incomplete. Specific problems include:

  • What does it mean for P and A to have one of a set of forms? Do they both have to have that form? (That doesn't happen; typically, only P contains template parameters)

  • In the introductory sentence, aren't T, TT, and i supposed to be the names of template parameters rather than template arguments?

  • In T[i], it appears we can deduce i, but not T (T can only be deduced in the form T[integer-constant])

  • What is an integer-constant supposed to be?

  • What is a cv-list?

  • Why can we not deduce const T from T? (Apparently you only get to deduce if both or neither type have a cv-list, whatever a cv-list is.)

  • We have extreme redundancy because, for instance, there is no way to say “in T (T::*)(T), you can deduce any of those Ts, and it's OK if some of the positions don't have a T”. So we have seven (!) forms of that construct, for all cases except the one where none of the three positions contain a T.

  • We have special case rules for pointers to member functions, even though they're not a special case and should be covered by the rule for pointers to members and the rule for functions.

  • We do not allow deducing a template template parameter's value from a template template argument — there is a TT<T> form, a TT<i> form, a template-name<T> form, and a template-name<i> form, but no TT<TT> form nor template-name<TT> form.

It looks like the editor managed to get rid of cv-list, at least, since the issue was filed. It's now just cv. (cv-list is kind of hilariously wrong, because [syntax] says that the -list suffix is for comma-separated lists...)



回答2:

It's correct, template-name<TT> is not a deduced context. That's not relevant here. Plain TT is a deduced context, and that's what you have here

The deduced context template-name<I> for a non-type template parameter I means that you can deduce 5 when you have argument Foo<5> to parameter Foo<I>.

The reason template-name<TT> is not a deduced context is fairly simple. If template-name<TT> would be legal, then template-name would have to take a template template parameter. That means it itself would have to be a template template template ("TTT"). There's only so much recursion in the language definition.

[edit] In your example, you are deducing V for U to see if it's a specialization. Both are template templates. You are not trying to deduce whether Foo<U> isFoo. Therefore your deduced context isTT, nottemplate-name.`.

To answer point 2 of the comment, TT is a deduced context because that form is listed explicitly. And what template-name<TT> means is that TT would have to be deducable when TT is used as an argument to a known template. But what type of template accepts a template template parameter? That the hypothetical TTT.