X-Authentication-Warning: delorie.com: mailnull set sender to djgpp-workers-bounces using -f
Date: Tue, 12 Feb 2002 08:30:06 +0200 (WET)
From: Andris Pavenis <pavenis AT lanet DOT lv>
X-Sender: pavenis AT ieva06
To: Richard Dawe <rich AT phekda DOT freeserve DOT co DOT uk>
Cc: djgpp-workers AT delorie DOT com
Subject: Re: GCC-3.1 related patches for DJGPP CVS version
In-Reply-To: <3C682BBD.12967A63@phekda.freeserve.co.uk>
Message-ID: <Pine.A41.4.05.10202120820360.10676-100000@ieva06>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Reply-To: djgpp-workers AT delorie DOT com
Errors-To: nobody AT delorie DOT com
X-Mailing-List: djgpp-workers AT delorie DOT com
X-Unsubscribes-To: listserv AT delorie DOT com
Precedence: bulk



On Mon, 11 Feb 2002, Richard Dawe wrote:

> Hello.
> 
> pavenis AT lanet DOT lv wrote:
> > Tried to compile DJGPP today's CVS version with GCC-3.1 20020210.
> > To get it to compile applied included patches.
> > 
> > 1) internal representation of __builtin_va_list changed to char *, as result
> > many breakages. Mostly added some ugly type casts
> > 
> > 2) warning were generated about snprintf  with empty format.
> > 
> > Perhaps changes which only requires using va_list of something more ugly
> > but not explicit casts from (or to) va_list can be applied. I don't know
> > about other ones though
> 
> These look OK to me, apart from a couple of things. See my comments inline
> below.
> 
> [snip]
> > diff -ur3 djgpp.orig/src/libc/ansi/stdio/fscanf.c
> > djgpp/src/libc/ansi/stdio/fscanf.c
> > --- djgpp.orig/src/libc/ansi/stdio/fscanf.c     Mon Dec 26 22:34:48 1994
> > +++ djgpp/src/libc/ansi/stdio/fscanf.c  Mon Feb 11 17:49:18 2002
> > @@ -9,7 +9,7 @@
> >    int r;
> >    va_list a=0;
> >    va_start(a, fmt);
> 
> I don't think that 'a' needs initialising in its definition.

You're right of course. It somehow slipped in ...

> 
> > -  r = _doscan(f, fmt, a);
> > +  r = _doscan(f, fmt, (void **) a);
> 
> Yuck. Wouldn't it be better to modify _doscan to work on a va_list too?
> 

Perhaps

> >    va_end(a);
> >    return r;
> >  }
> 
> [snip]
> > diff -ur3 djgpp.orig/src/libc/ansi/stdio/scanf.c
> > djgpp/src/libc/ansi/stdio/scanf.c
> > --- djgpp.orig/src/libc/ansi/stdio/scanf.c      Mon Dec 26 22:34:54 1994
> > +++ djgpp/src/libc/ansi/stdio/scanf.c   Mon Feb 11 17:49:38 2002
> > @@ -9,7 +9,7 @@
> >    int r;
> >    va_list a=0;
> >    va_start(a, fmt);
> > -  r = _doscan(stdin, fmt, a);
> > +  r = _doscan(stdin, fmt, (void **) a);
> >    va_end(a);
> >    return r;
> >  }
> 
> See the comments for fscanf.c.
> 
> [snip]
> > diff -ur3 djgpp.orig/src/libc/posix/unistd/confstr.c
> > djgpp/src/libc/posix/unistd/confstr.c
> > --- djgpp.orig/src/libc/posix/unistd/confstr.c  Wed Jun 20 02:00:46 2001
> > +++ djgpp/src/libc/posix/unistd/confstr.c       Mon Feb 11 18:17:28 2002
> > @@ -43,7 +43,9 @@
> >      case _CS_POSIX_V6_ILP32_OFF32_LDFLAGS:
> >      case _CS_POSIX_V6_ILP32_OFF32_LIBS:
> >      {
> > -      out_len = snprintf(buf, len, "");
> > +/*      out_len = snprintf(buf, len, "");  */
> > +      if (len>0) *buf=0;
> > +      out_len=0;
> >        ++out_len;
> >        break;
> >      }
> 
> Why not just set out_len to 1? Also, presumably the comment will be removed
> too?

snprintf should output number of outputted chars not including trailing \0
AFAIK, so 0 is correct. gcc-3.1 20020210 emitted warning about empty
format string. Maybe it will not be so later, so I left snprintf in
comment. We could also revert this change when (if) we'll need some real
snprintf there.

Andris