Semantics of constrained-type-specifiers in the return type of abbreviated functions

[Tom Honermann]

The Concepts TS (N4641) states the following in 7.1.6.4 [dcl.spec.auto]
p4 (this section is now 10.1.7.4 in the WP):

"A constrained-type-specifier C1 within the declared return type of an
abbreviated function template declaration does not designate a
placeholder if its introduced constraint-expression (7.1.6.4.2) is
determined to be equivalent, using the rules in 14.6.6.1 for comparing
expressions, to the introduced constraint-expression for a
constrained-type-specifier C2 in the parameter-declaration-clause of
that function declaration. Instead, C1 is replaced by the template
parameter invented for C2 (8.3.5)."

The text includes the following example:

C const& f1(C); // has one template parameter and no deduced return type

The TS is not explicit that this behavior also applies to placeholders
specified in a trailing-return-type:

auto f1(C) -> C const&;

gcc 7.1 exhibits different behavior for placeholders specified in the
declared return type vs a trailing-return-type:

$ cat t.cpp
template<typename> concept bool C = true;
template<C> class ct {};
ct<C> af1(C) { return ct<int>{}; }
auto af2(C) -> ct<C> { return ct<int>{}; }
auto v1 = af1('c');// Rejected by gcc
auto v2 = af2('c'); // Accepted by gcc

$ g++ -c -fconcepts t.cpp
t.cpp: In instantiation of ‘ct<C> af1(auto:1) [with auto:1 = char]’:
t.cpp:5:18: required from here
t.cpp:3:31: error: could not convert ‘ct<int>{}’ from ‘ct<int>’ to
‘ct<char>’
ct<C> af1(C) { return ct<int>{}; }
^

My expectation is that the call to af2() should be similarly rejected.

Additionally, gcc does not appear to enforce the same-type constraint
when the return type consists solely of a constrained-type-specifier:

$ cat t2.cpp
template<typename> concept bool C = true;
class c {};
const C af3(C) { return 3; }
auto af4(C) -> const C { return 4; }
auto v1 = af3(c{}); // Accepted by gcc
auto v2 = af4(c{}); // Accepted by gcc

$ g++ -c -fconcepts t2.cpp
<no error>

My expectation is that both of the calls to af3() and af4() should be
rejected here.

Could someone clarify what the intent is in the Concepts TS?

Tom.

[Hubert Tong]

On Sat, Jun 24, 2017 at 7:46 AM, Tom Honermann <t...@honermann.net> wrote:

The Concepts TS (N4641) states the following in 7.1.6.4 [dcl.spec.auto] p4 (this section is now 10.1.7.4 in the WP):

"A constrained-type-specifier C1 within the declared return type of an abbreviated function template declaration does not designate a placeholder if its introduced constraint-expression (7.1.6.4.2) is determined to be equivalent, using the rules in 14.6.6.1 for comparing expressions, to the introduced constraint-expression for a constrained-type-specifier C2 in the parameter-declaration-clause of that function declaration. Instead, C1 is replaced by the template parameter invented for C2 (8.3.5)."

The text includes the following example:

C const& f1(C); // has one template parameter and no deduced return type

The TS is not explicit that this behavior also applies to placeholders specified in a trailing-return-type:

auto f1(C) -> C const&;

The TS is explicit enough; [dcl.spec.auto]/2 says:
If the function declarator includes a trailing-return-type, that trailing-return-type specifies the declared return type of the function.

 

gcc 7.1 exhibits different behavior for placeholders specified in the declared return type vs a trailing-return-type:

I suggest a bug report.

[Bjarne Stroustrup]

Andrew reported that bug some time ago.

--
You received this message because you are subscribed to the Google Groups "SG8 - Concepts" group.
To unsubscribe from this group and stop receiving emails from it, send an email to concepts+u...@isocpp.org.
To post to this group, send email to conc...@isocpp.org.
Visit this group at https://groups.google.com/a/isocpp.org/group/concepts/.

[Tom Honermann]

On 06/24/2017 10:54 AM, Bjarne Stroustrup wrote:

On 6/24/2017 10:41 AM, Hubert Tong wrote:

On Sat, Jun 24, 2017 at 7:46 AM, Tom Honermann <t...@honermann.net> wrote:

The Concepts TS (N4641) states the following in 7.1.6.4 [dcl.spec.auto] p4 (this section is now 10.1.7.4 in the WP):

"A constrained-type-specifier C1 within the declared return type of an abbreviated function template declaration does not designate a placeholder if its introduced constraint-expression (7.1.6.4.2) is determined to be equivalent, using the rules in 14.6.6.1 for comparing expressions, to the introduced constraint-expression for a constrained-type-specifier C2 in the parameter-declaration-clause of that function declaration. Instead, C1 is replaced by the template parameter invented for C2 (8.3.5)."

The text includes the following example:

C const& f1(C); // has one template parameter and no deduced return type

The TS is not explicit that this behavior also applies to placeholders specified in a trailing-return-type:

auto f1(C) -> C const&;

The TS is explicit enough; [dcl.spec.auto]/2 says:
If the function declarator includes a trailing-return-type, that trailing-return-type specifies the declared return type of the function.

Thanks, I was just having a hard time squaring gcc's behavior with the TS.

 

gcc 7.1 exhibits different behavior for placeholders specified in the declared return type vs a trailing-return-type:

I suggest a bug report.

Andrew reported that bug some time ago.

Hmm, I had looked for a bug report and still can't find one; neither submitted by Andrew, nor linked from the Concepts meta-bug [1], and nothing with "trailing return type" in the subject looks relevant.  My Google searches aren't turning anything up either.

I filed a new bug [2].

It looks like Andrew filed a bug [3] earlier this week for the second example in my original email.

Tom.

[1]: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67491
[2]: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81198
[3]: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81139