Re: snprintfv 1.1 pre-release

[Dave Butenhof <David dot Butenhof at hp dot com>]

Bruce Korb wrote:

> "Joseph S. Myers" wrote:
>  
>
> > On Thu, 24 Jul 2003, Paolo Bonzini wrote:
> >    
> >
> > > * there is support for width/precision positional parameters (%NNN$s was
> > > supported in 1.0 too), like
> > >
> > >    printf ("%2$*1$s", "fubar", -10);
> > >      
> > This, of course, being an erroneous example, taking "fubar" as the width
> > and -10 as the string to be formatted.
> >    
> Well, I confess I don't know what is right.  What you say, Joseph, corresponds
> to my reading of the standard.  However, in experiments with Sun's libc and
> Linux's glibc:
I'm not sure why this got sent to the Austin group list; and perhaps I'm 
missing some important context, but...

> > $ cat xx.c
> > #include <stdio.h>
> >
> > int
> > main( int argc, char** argv )
> > {
> >    printf( "%%2$*1$d:  ==%2$*1$d==\n", 11, 22 );
> >    printf( "%%1$*2$d:  ==%2$*1$d==\n", 11, 22 );
> >    printf( "%%*d:      ==%2$*1$d==\n", 11, 22 );
Excuse me, but why are all the actual formatting strings identical? It 
appears that it's doing exactly what you told it to do. I think you must 
have INTENDED to do this:

   #include <stdio.h>
   int
   main( int argc, char** argv )
   {
       printf( "%%2$*1$d:  ==%2$*1$d==\n", 11, 22 );
       printf( "%%1$*2$d:  ==%1$*2$d==\n", 11, 22 );
       printf( "%%*d:      ==%*d==\n", 11, 22 );
       return 0;
   }

That is, on each line REPORT and USE the same formatting. Yes?

> >    return 0;
> > }
> > $ cc    -o xx xx.c 
> > $ ./xx
> > %2$*1$d:  ==         22==
> > %1$*2$d:  ==         22==
> > %*d:      ==         22==
> > $ gcc -o yy xx.c
> > $ ./yy
> > %2$*1$d:  ==         22==
> > %1$*2$d:  ==         22==
> > %*d:      ==         22==
> >    
> I'm pretty sure that something has to be wrong.  What?  Can we say
> that both are in non-compliance and the results should be this?
>
> > %2$*1$d:  ==         22==
> > %1$*2$d:  ==                    11==
> > %*d:      ==         22==
> >    
On Tru64, when I build and run the program you posted I get exactly what 
you report Solaris and Linux doing. But when I correct the program to do 
what it says, I get what you expect. I suspect you'll find the same 
results. (I certainly do on our ancient Solaris 2.6 system; I didn't 
bother trying on Linux.)

--
/--------------------[ David.Butenhof@hp.com ]--------------------\
| Hewlett-Packard Company       Tru64 UNIX & VMS Thread Architect |
|     My book: http://www.awl.com/cseng/titles/0-201-63392-2/     |
\----[ http://homepage.mac.com/dbutenhof/Threads/Threads.html ]---/