Re: Bug#448732: zsh: printf %g -0 outputs 0 instead of -0

Re: Bug#448732: zsh: printf %g -0 outputs 0 instead of -0

[Clint Adams]

I believe the gist is that since %g calls for a float, -0 should be
parsed as negative zero.

On Wed, Oct 31, 2007 at 01:19:11PM +0100, Vincent Lefevre wrote:
> vin:~> printf "%g %g\n" -0 -0.
> 0 -0
> 
> whereas with the printf from the coreutils on etch:
> 
> courge:~> /usr/bin/printf "%g %g\n" -0 -0.
> -0 -0
> 
> (this will not work on sid due to the bug 448723, but this is a
> different bug since the zsh behavior is the same on any OS).
> 
> Discussion:
> https://www.opengroup.org/sophocles/show_mail.tpl?CALLER=show_archive.tpl&source=L&listname=austin-group-l&id=11150
> https://www.opengroup.org/sophocles/show_mail.tpl?CALLER=show_archive.tpl&source=L&listname=austin-group-l&id=11151
> https://www.opengroup.org/sophocles/show_mail.tpl?CALLER=show_archive.tpl&source=L&listname=austin-group-l&id=11156
> https://www.opengroup.org/sophocles/show_mail.tpl?CALLER=show_archive.tpl&source=L&listname=austin-group-l&id=11159
> https://www.opengroup.org/sophocles/show_mail.tpl?CALLER=show_archive.tpl&source=L&listname=austin-group-l&id=11160

Re: Bug#448732: zsh: printf %g -0 outputs 0 instead of -0

[Peter Stephenson]

On Wed, 31 Oct 2007 12:38:38 -0400
Clint Adams <schizo@debian.org> wrote:
> I believe the gist is that since %g calls for a float, -0 should be
> parsed as negative zero.

I tried the following patch, but on my system (Fedora 7 with glibc
glibc-2.6-4) strtod() returns the double value 0 (not -0) when parsing
the input "-0".  I confirmed this with a standalone programme.  It seems
pointless working around this in the shell.

===================================================================
RCS file: /cvsroot/zsh/zsh/Src/builtin.c,v
retrieving revision 1.180
diff -u -r1.180 builtin.c
--- Src/builtin.c	6 Jul 2007 21:52:39 -0000	1.180
+++ Src/builtin.c	6 Nov 2007 10:39:29 -0000
@@ -4162,9 +4162,25 @@
 			    break;
 		    case 2:
 			if (curarg) {
-			    mnumval = matheval(curarg);
-			    doubleval = (mnumval.type & MN_FLOAT) ?
-			    	mnumval.u.d : (double)mnumval.u.l;
+			    char *eptr;
+			    /*
+			     * First attempt to parse as a floating
+			     * point constant.  If we go through
+			     * a math evaluation, we can lose
+			     * mostly unimportant information
+			     * that people in standards organizations
+			     * worry about.
+			     */
+			    doubleval = strtod(curarg, &eptr);
+			    /*
+			     * If it didn't parse as a constant,
+			     * parse it as an expression.
+			     */
+			    if (*eptr != '\0') {
+				mnumval = matheval(curarg);
+				doubleval = (mnumval.type & MN_FLOAT) ?
+				    mnumval.u.d : (double)mnumval.u.l;
+			    }
 			} else doubleval = 0;
 			if (errflag) {
 			    doubleval = 0;


-- 
Peter Stephenson <pws@csr.com>                  Software Engineer
CSR PLC, Churchill House, Cambridge Business Park, Cowley Road
Cambridge, CB4 0WZ, UK                          Tel: +44 (0)1223 692070

Re: Bug#448732: zsh: printf %g -0 outputs 0 instead of -0

[Vincent Lefevre]

On 2007-11-06 10:46:20 +0000, Peter Stephenson wrote:
> On Wed, 31 Oct 2007 12:38:38 -0400
> Clint Adams <schizo@debian.org> wrote:
> > I believe the gist is that since %g calls for a float, -0 should be
> > parsed as negative zero.
> 
> I tried the following patch, but on my system (Fedora 7 with glibc
> glibc-2.6-4) strtod() returns the double value 0 (not -0) when parsing
> the input "-0".  I confirmed this with a standalone programme.  It seems
> pointless working around this in the shell.

The bug in glibc's strtod() was fixed on 2007-08-03:

revision 1.25
date: 2007/08/03 16:45:24;  author: drepper;  state: Exp;  lines: +4 -3
(____STRTOF_INTERNAL): Properly handle -0.

So, zsh has to fix only its bug.

-- 
Vincent Lefèvre <vincent@vinc17.org> - Web: <http://www.vinc17.org/>
100% accessible validated (X)HTML - Blog: <http://www.vinc17.org/blog/>
Work: CR INRIA - computer arithmetic / Arenaire project (LIP, ENS-Lyon)

Re: Bug#448732: zsh: printf %g -0 outputs 0 instead of -0

[Peter Stephenson]

Vincent Lefevre wrote:
> On 2007-11-06 10:46:20 +0000, Peter Stephenson wrote:
> > On Wed, 31 Oct 2007 12:38:38 -0400
> > Clint Adams <schizo@debian.org> wrote:
> > > I believe the gist is that since %g calls for a float, -0 should be
> > > parsed as negative zero.
> > 
> > I tried the following patch, but on my system (Fedora 7 with glibc
> > glibc-2.6-4) strtod() returns the double value 0 (not -0) when parsing
> > the input "-0".  I confirmed this with a standalone programme.  It seems
> > pointless working around this in the shell.
> 
> The bug in glibc's strtod() was fixed on 2007-08-03:
> 
> revision 1.25
> date: 2007/08/03 16:45:24;  author: drepper;  state: Exp;  lines: +4 -3
> (____STRTOF_INTERNAL): Properly handle -0.

Thanks, I committed the change in zh so the problem should go away
eventually.

-- 
Peter Stephenson <pws@csr.com>                  Software Engineer
CSR PLC, Churchill House, Cambridge Business Park, Cowley Road
Cambridge, CB4 0WZ, UK                          Tel: +44 (0)1223 692070