Re: c++/8986

Re: c++/8986

[Wolfgang Bangerth]

The following reply was made to PR c++/8986; it has been noted by GNATS.

From: Wolfgang Bangerth <bangerth@ices.utexas.edu>
To: gcc-gnats@gcc.gnu.org
Cc: Giovanni Bajo <giovannibajo@libero.it>
Subject: Re: c++/8986
Date: Tue, 1 Apr 2003 15:50:18 -0600 (CST)

 Giovanni incited me to do something inherently useless ;-) Here's a 
 smaller testcase for this horribly broken report:
 ---------------------------------------
 namespace NS
 {
   struct C {};
 
   namespace NS
   {
     template <typename> struct X
     {
         NS::C foo() const { return NS::C(); }
 
         int size() const;
 
         X bar(int s)
           {
             const int i = s;
             bogus()->bogus = s;
           }
 
         template <class> void baz() {}
 
         X& foobar(int n)
           {
             const int x = n;
             const bool y = n;
           }
     };
   }
 }
 --------------------------------------
 
 It ICEs with present 3.3:
 g/x> /home/bangerth/bin/gcc-3.3-pre/bin/c++ -c x.ii
 x.ii:9: error: syntax error before `(' token
 x.ii:11: error: semicolon missing after declaration of 
 `NS::NS::X<<anonymous 
    template type parameter> >'
 x.ii:11: error: extraneous `int' ignored
 x.ii:11: error: non-member function `NS::NS::X<<anonymous template type 
    parameter> > NS::NS::size()' cannot have `const' method qualifier
 x.ii:13: error: syntax error before `(' token
 x.ii:16: error: ISO C++ forbids declaration of `bogus' with no type
 x.ii: In function `int NS::NS::bogus()':
 x.ii:16: error: parse error before `->' token
 x.ii: In function `void NS::baz()':
 x.ii:21: error: syntax error before `&' token
 x.ii:24: error: `n' undeclared (first use this function)
 x.ii:24: error: (Each undeclared identifier is reported only once for each 
    function it appears in.)
 x.ii:24: internal compiler error: Segmentation fault
 Please submit a full bug report,
 with preprocessed source if appropriate.
 See <URL:http://gcc.gnu.org/bugs.html> for instructions.
 
 I didn't ICE (but spew lots of error messages) in 3.2, so this is a 
 regression. However, given the uselessness of the code (and where it came 
 from) and in particular the fact that it doesn't ICE with present 
 mainline, I only mark it as a regression, but don't increase the priority.
 
 W.
 
 -------------------------------------------------------------------------
 Wolfgang Bangerth              email:            bangerth@ices.utexas.edu
                                www: http://www.ices.utexas.edu/~bangerth/