[Bug c++/69701] "v.operator T()" incorrectly parsed if "v.T()" present.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=69701
Language Lawyer changed:
What|Removed |Added
CC||language.lawyer at gmail dot
com
--- Comment #7 from Language Lawyer ---
Another example related to CWG2396:
struct B
{
struct S {};
operator S();
};
auto v = B{}.operator struct S(); // error: 'struct B' has no member named
'operator S'
It can be checked that GCC injects unrelated `struct S` into the global
namespace and then tries to find `operator struct ::S` in B.
However, according to P1787R6 [basic.lookup.unqual]/5: An unqualified name that
is a component name of a type-specifier or ptr-operator of a conversion-type-id
is looked up in the same fashion as the conversion-function-id in which it
appears.
In an elaborated-type-specifier of the form "class-key identifier", such as
`struct S`, identifier is a component name ([dcl.type.elab]/1), so S shall be
searched in B first.
[Bug c++/69701] "v.operator T()" incorrectly parsed if "v.T()" present.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=69701 --- Comment #6 from Andrew Pinski --- http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#1291 [Accepted at the November, 2020 meeting as part of paper P1787R6 and moved to DR at the February, 2021 meeting.] So there is a defect report in this area after all
[Bug c++/69701] "v.operator T()" incorrectly parsed if "v.T()" present.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=69701
Andrew Pinski changed:
What|Removed |Added
Keywords|accepts-invalid |
--- Comment #5 from Andrew Pinski ---
So I think clang and MSVC have a bug in this area too.
See basic.lookup.classref/7 :
If the id-expression is a conversion-function-id, its conversion-type-id is
first looked up in the class of the object
expression (11.8) and the name, if found, is used. Otherwise it is looked up in
the context of the entire
postfix-expression. In each of these lookups, only names that denote types or
templates whose specializations
are types are considered. [Example:
struct A { };
namespace N {
struct A {
void g() { }
template operator T();
};
}
int main() {
N::A a;
a.operator A(); // calls N::A::operator N::A
}
— end example]
If we take the example, clang and MSVC calls N::A::operator ::A and not what
the standard says.
GCC I think is still wrong in the original testcase I think because B::A does
not represent a type .
In my example in comment #4, GCC and ICC does the correct thing and selects the
correct type.
[Bug c++/69701] "v.operator T()" incorrectly parsed if "v.T()" present.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=69701
Andrew Pinski changed:
What|Removed |Added
Ever confirmed|0 |1
Last reconfirmed||2021-12-03
Keywords||accepts-invalid
Status|UNCONFIRMED |NEW
--- Comment #4 from Andrew Pinski ---
Here is a different testcase which shows we also have an accepts invalid:
class A {};
class B {
typedef int AT;
public:
operator AT () const noexcept { return {}; }
typedef int A;
};
int main() {
B b;
b.operator A();
}
I have not looked up in the standard which A is supposed to be found here (for
b.operator A) is it B::A or ::A?
[Bug c++/69701] "v.operator T()" incorrectly parsed if "v.T()" present.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=69701 --- Comment #3 from Andrew Pinski --- ICC also rejects this for the same reason as GCC while both MSVC and clang accept it. Maybe there is a defect report about this.
[Bug c++/69701] "v.operator T()" incorrectly parsed if "v.T()" present.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=69701
--- Comment #1 from Andrew Pinski ---
I think this should be diagnose differently in that A changes meanings inside
the class.
That is if we used:
operator A () const noexcept { return {}; }
We get the following error message:
t9.cc:8:12: error: declaration of ‘void B::A()’ [-fpermissive]
void A() {}
^
t9.cc:2:7: error: changes meaning of ‘A’ from ‘class A’ [-fpermissive]
class A {};
^
[Bug c++/69701] "v.operator T()" incorrectly parsed if "v.T()" present.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=69701 --- Comment #2 from Eric Fiselier --- @Andrew The in-class diagnostics are pretty good. My concern is that users outside the class cannot name the conversion operator. I don't think they care that "A" changes meaning *within* B. I don't think any error should be emitted for the reproducer.
