From:	Ned Ulbricht <nedu AT ee DOT washington DOT edu>
Newsgroups:	comp.os.msdos.djgpp
Subject:	Re: printf 'g' conversion
Date:	Sun, 01 Mar 1998 16:54:31 -0800
Organization:	University of Washington
Lines:	31
Message-ID:	<34FA0346.33@ee.washington.edu>
References:	<Pine DOT SUN DOT 3 DOT 91 DOT 980301131518 DOT 13547I-100000 AT is>
NNTP-Posting-Host:	cs236-16.student.washington.edu
Mime-Version:	1.0
To:	djgpp AT delorie DOT com
DJ-Gateway:	from newsgroup comp.os.msdos.djgpp

Eli Zaretskii wrote:
> 
> On Fri, 27 Feb 1998, Andrew Gibson wrote:
> 
> > From Linux this is what I get. From DJGPP I get 9 decimal places and
> > only 8 sig figs.
> 
> I'm not sure that this is a bug.  My ANSI C references seem to be not
> definitive on this.  Can someone with clearer understanding of what the
> standard says clarify this?

The Working Draft, 97-11-21, WG14/N794 J11/97-158, p.290 (&cf p.287),
seems to be a little bit ambiguous about this, but it says under 'g,G'
"the number is converted in style f or e (...), with the precision
specifying the number of significant digits." I interpret the "with" as
saying that the g,G format always applies its own semantics to precision
irrespective of the "style" used for the output. I.e. it consistently
takes over the meaning of precision.

I can read K&RII either way, so no help there...

Harbison & Steele (pre-ANSI reference for compiler writers) give a
definitive algorithm for g,G which supports the behavior reported under
Linux gcc.  Also a couple of fairly recent MS-DOS compilers also
generate that behavior.

So, IMHO, it is a bug in djgpp (also I've confirmed the behavior on my
libc.a installation as well).
--
Ned Ulbricht
mailto:nedu AT ee DOT washington DOT edu

- Raw text -