Date: Thu, 11 Aug 1994 09:17:51 -0700 (PDT)
From: Gordon Hogenson <ghogenso AT u DOT washington DOT edu>
Subject: Re: Libg++ woes
To: Paul Harness <gapa83 AT udcf DOT gla DOT ac DOT uk>
Cc: djgpp <djgpp AT sun DOT soe DOT clarkson DOT edu>

On Thu, 11 Aug 1994, Paul Harness wrote:

> of 1.12, this didn't occur. Anyway, I found that it could be solved by
> putting:
> 
> #pragma implementation "_string.h"
> 
> in my program, or by commenting out `#pragma interface' in _string.h. We
> shouldn't have to do this, should we? 

Hmmmm, I tried this and now I have the opposite problem.  Instead of
undefined references to String methods, I get multiple definitions of
_IO_stdout_ , _IO_stderr_, and _IO_listall_ ...

Same program I posted earlier, same compilation line.  Doing either of
the things you suggest produces the same behavior.  Paul: can you try my
test program and see whether it works for you?

#pragma implementation "_string.h"
#include <_string.h>

main()
{
	String s("Hello?");
}

gcc strtst.cc -o strtst -lgpp -v

This gives:

stdstrbufs.cc(.data+0x0): multiple definition of `_IO_stdin_'
stdfiles.c(.data+0x0): first defined here
stdstrbufs.cc(.data+0x50): multiple definition of `_IO_stdout_'
stdfiles.c(.data+0x50): first defined here
stdstrbufs.cc(.data+0xa0): multiple definition of `_IO_stderr_'
stdfiles.c(.data+0xa0): first defined here
stdstrbufs.cc(.data+0x1ec): multiple definition of `_IO_list_all'
stdfiles.c(.data+0xf0): first defined here

If the "test release" of libg++ 2.6.0 doesn't have this problem,
perhaps we should go back to it??  (Is that at ftp.cygnus.com ?)

Gordon