Path: utzoo!telly!attcan!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!YAHI.STANFORD.EDU!tiemann
From: tiemann@YAHI.STANFORD.EDU (Michael Tiemann)
Newsgroups: gnu.g++.bug
Subject: G++ 1.34.2 overloaded virtual functions
Message-ID: <8904271539.AA02711@yahi.stanford.edu>
Date: 27 Apr 89 15:39:52 GMT
References: <715@arisia.Xerox.COM>
Sender: daemon@tut.cis.ohio-state.edu
Reply-To: tiemann@lurch.stanford.edu
Distribution: gnu
Organization: GNUs Not Usenet
Lines: 48


   Date: 26 Apr 89 19:55:32 GMT
   From: arisia!wedekind@lll-winken.llnl.gov  (Jerry Wedekind)
   Organization: Xerox PARC
   Sender: bug-g++-request@prep.ai.mit.edu


   Environment:  g++ 1.34.2, Sun3 OS4.

   G++ seems to have a bug with overloaded virtual functions.  In the test case:

       class A {
       public:
	   virtual void XXX (int);
       };

       class B : public A {
       public:
	   void XXX (int);
	   void XXX (char);
       };

   It complains:

       "test.h:9: conficting specification deriving virtual function 'void B::XXX (float)'

   It should be the case that XXX(int) is virtual in A & B, and XXX is
   overloaded in B.  The situation is clear:  if you have an A.XXX(int)
   use the virtual function table, and if you call B.XXX, you use the
   overloaded function rules & then virtual (if needed).  It doesn't help
   to make A.XXX overloaded, since it still complains about any new
   overloaded functions in B.

   This test case works in g++ 1.32 & Oregon C++.

   Is this illegal (and two other compilers are wrong), or is there a
   workaround in 1.34.2?

   /John

   arpa:  JWenn.esae@Xerox.com
   uucp:  wedekind@arisia.uucp

Some people thought that this should be an error, and cfront
complains, so I made GNU C++ complain.  There will be a flag in
1.35.0, -Woverloaded-virtual which permits this extension.

Michael