[Bug c++/15786] New: [regression 3.5/3.4.0] Bad error message for frequently occuring error.

[carlo at gcc dot gnu dot org]

I've run so many times into this now that I feel that it is important to get it
fixed.

The following code snippet,

struct A {
  void foo(bar* p);
};

Gives as error message:

>g++-cvs-3.5 -c test.cc
test.cc:2: error: variable or field `foo' declared void
test.cc:2: error: expected `;' before '(' token

Only when foo is not a member function, there also appears
a MUCH more correct error message:

test.cc:1: error: variable or field `foo' declared void
test.cc:1: error: `bar' was not declared in this scope
test.cc:1: error: `p' was not declared in this scope

If possible, the `bar' was not declared in this scope should
come first - and it should definitely also be printed in
the case of the member function.

3.4.0 does the same unclear thing.

You can consider this a regression because 3.3.3 gives most
sensible output:

>g++-3.3.3 -c test.cc
test.cc:2: error: `bar' was not declared in this scope
test.cc:2: error: `p' was not declared in this scope
test.cc:2: error: invalid data member initialization
test.cc:2: error: (use `=' to initialize static data members)
test.cc:2: error: variable or field `foo' declared void

-- 
           Summary: [regression 3.5/3.4.0] Bad error message for frequently
                    occuring error.
           Product: gcc
           Version: 3.5.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: c++
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: carlo at gcc dot gnu dot org
                CC: gcc-bugs at gcc dot gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15786