[Bug c++/11118] New: Regression in 3.4 (cvs from jun 06) using nested classes in templates

[Bug c++/11118] New: Regression in 3.4 (cvs from jun 06) using nested classes in templates

[lars@trolltech.com]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11118

           Summary: Regression in 3.4 (cvs from jun 06) using nested classes
                    in templates
           Product: gcc
           Version: 3.4
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: c++
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: lars@trolltech.com
                CC: gcc-bugs@gcc.gnu.org
 GCC build triplet: i686-pc-linux-gnu
  GCC host triplet: i686-pc-linux-gnu
GCC target triplet: i686-pc-linux-gnu

The following code compiles fine in all versions < 3.4: 
 
template<typename T> class Foo 
{ 
public: 
    void foo(); 
    class Data { 
    }; 
}; 
 
template<typename T> void Foo<T>::foo() 
{ 
    // compiles 
    Data *y; 
    // doesn't compile 
    Foo::Data *x; 
} 
 
[lars@binky gcc-bugs]$ gcc nested.cpp 
nested.cpp: In member function `void Foo<T>::foo()': 
nested.cpp:14: error: `x' has not been declared 
[lars@binky gcc-bugs]$ gcc -v 
Reading specs from /home/lars/gcc/lib/gcc-lib/i686-pc-linux-gnu/3.4/specs 
Configured with: ../gcc/configure --prefix=/home/lars/gcc --disable-libgcj 
Thread model: posix 
gcc version 3.4 20030605 (experimental) 
 
 
Best regards, 
Lars

[Bug c++/11118] Regression in 3.4 (cvs from jun 06) using nested classes in templates

[pinskia@physics.uc.edu]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11118



------- Additional Comments From pinskia@physics.uc.edu  2003-06-06 21:46 -------
I can compile this if I put typename in front of Foo:Data but I do not know if this is required by the 
standard.

[Bug c++/11118] Regression in 3.4 (cvs from jun 06) using nested classes in templates

[bangerth@dealii.org]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11118


bangerth@dealii.org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|                            |INVALID


------- Additional Comments From bangerth@dealii.org  2003-06-07 00:31 -------
Not a bug. Foo::Data is a template dependent type, so it needs to be prefixed
by "typename".

W.

[Bug c++/11118] Regression in 3.4 (cvs from jun 06) using nested classes in templates

[lars@trolltech.com]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11118



------- Additional Comments From lars@trolltech.com  2003-06-07 08:33 -------
Subject: Re:  Regression in 3.4 (cvs from jun 06) using nested
 classes in templates


Well, then at lest the error message is rather misleading.

Best regards,
Lars

[Bug c++/11118] [3.4 only] missleading error message on missing typename

[pinskia@physics.uc.edu]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11118


pinskia@physics.uc.edu changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |UNCONFIRMED
           Keywords|                            |diagnostic
         Resolution|INVALID                     |
            Summary|Regression in 3.4 (cvs from |[3.4 only] missleading error
                   |jun 06) using nested classes|message on missing typename
                   |in templates                |


------- Additional Comments From pinskia@physics.uc.edu  2003-06-07 15:13 -------
I am reopening this bug because it is a misleading error message with the missing 
typename.

[Bug c++/11118] [3.4 only] missleading error message on missing typename

[pinskia@physics.uc.edu]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11118


pinskia@physics.uc.edu changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned@gcc.gnu.org      |gdr@gcc.gnu.org
             Status|UNCONFIRMED                 |NEW


------- Additional Comments From pinskia@physics.uc.edu  2003-06-07 15:14 -------
Assigning to Diagnostic maintainer.

[Bug c++/11118] [3.4 only] missleading error message on missing typename

[bangerth@dealii.org]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11118


bangerth@dealii.org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |gdr@integrable-solutions.net


------- Additional Comments From bangerth@dealii.org  2003-06-09 01:24 -------
I disagree with the statement that the diagnostics are insufficient, because
there is really not much the compiler can do. See, if no "typename" is
given, the compiler assumes that Foo::Data is a member function/variable.
Since no parentheses are following, it assumes a member variable, and the
star then means multiplication. So far so good, but multiplication with what?
"x" is not a declared variable. That's exactly what the compiler is trying to
tell you.

Honestly, I don't think there is much a compiler can do in this situation if
two-stage name lookup is properly implemented. I leave it to Gaby to decide
what to do with this PR.

Wolfgang

[Bug c++/11118] [3.4 only] missleading error message on missing typename

[giovannibajo@libero.it]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11118


giovannibajo@libero.it changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |DUPLICATE


------- Additional Comments From giovannibajo@libero.it  2003-06-09 11:28 -------
This is a borderline case because the qualified-id is the injected-class-name 
("Foo" is an alias of "Foo<T>"). DR224 clarifies that such names are "non-
dependent" because of the _semantic_ behind it (they can be looked-up at 
template definition time, not instantiation), even if they appear to be 
dependent from a _syntactic_ point of view (the qualified-id contains the 
template parameter T). 

In other words, the proposed code should not compile as per pure C++98, but 
it's been recognized as a defect. You can consult DR224 
(http://anubis.dkuug.dk/jtc1/sc22/wg21/docs/cwg_defects.html#224) for more 
examples.

We already have an open bug (PR9634) to track this, thus I close this bug as a 
duplicate.

*** This bug has been marked as a duplicate of 9634 ***