Date: Sun, 28 Mar 2004 16:19:41 +0200 (CEST) From: Jean-Baptiste Quenot <jb.quenot@caraldi.com> To: FreeBSD-gnats-submit@FreeBSD.org Cc: mark@grondar.za Subject: ports/64845: Par must exclude non-breaking space from the class of space chars Message-ID: <20040328141941.C538ABD@caraldi.com> Resent-Message-ID: <200403281420.i2SEKLTc034627@freefall.freebsd.org>
next in thread | raw e-mail | index | archive | help
>Number: 64845
>Category: ports
>Synopsis: Par must exclude non-breaking space from the class of space chars
>Confidential: no
>Severity: serious
>Priority: medium
>Responsible: freebsd-ports-bugs
>State: open
>Quarter:
>Keywords:
>Date-Required:
>Class: update
>Submitter-Id: current-users
>Arrival-Date: Sun Mar 28 06:20:21 PST 2004
>Closed-Date:
>Last-Modified:
>Originator: Jean-Baptiste Quenot
>Release: FreeBSD 5.1-CURRENT i386
>Organization:
>Environment:
System: FreeBSD watt.intra.caraldi.com 5.1-CURRENT FreeBSD 5.1-CURRENT #6: Tue Oct 14 19:03:28 CEST 2003 jbq@watt.intra.caraldi.com:/usr/obj/usr/src/sys/WATT i386
>Description:
Par 1.52 on FreeBSD does not work as expected by the upstreams author. On
FreeBSD, the isspace() system call returns true for the non-breaking space
character 0xA0, but this is an unintended side effect.
Quoting a message from the upstreams author:
--------------------------------------------------------------------------------
From: "Adam M. Costello" <amc+0zjyiz+@nicemice.net>
Date: Tue, 2 Dec 2003 21:19:10 +0000
To: Jean-Baptiste Quenot <jb.quenot@caraldi.com>
User-Agent: Mutt/1.5.4i
> on FreeBSD, the locales definitions include non-breaking space in the
> list of spaces, thus isspace(160) is true, and as a result all my
> nbsps are filtered out, and lines are broken on them.
>
> I noticed that the GNU libc has removed 0xA0 from spaces on purpose.
> But the BSD guys seem to have another approach, as this kind of stuff
> is "implementation specific".
That's interesting. This was not an issue in Par 1.51, because it
didn't call setlocale(), so only ASCII characters were recognized
by isspace(), isalnum(), islower(), etc. In par 1.52, a call to
setlocale() was added so that non-ASCII letters and digits would be
recognized for the purpose of the g,B,P,Q options.
An unforseen side effect is that non-ASCII white-space characters are
now recognized.
--------------------------------------------------------------------------------
Here is the fragment declaring SPACE and BLANK for the ISO Latin 1 locale on
FreeBSD:
/*
* Standard LOCALE_CTYPE for the ISO 8859-1 Locale
*
* $FreeBSD: src/share/mklocale/la_LN.ISO8859-1.src,v 1.3 2001/11/30 05:05:53 ache Exp $
*/
...
SPACE 0x09 - 0x0d ' ' 0xa0
UPPER 'A' - 'Z' 0xc0 - 0xd6 0xd8 - 0xde
XDIGIT '0' - '9' 'a' - 'f' 'A' - 'F'
BLANK ' ' '\t' 0xa0
>How-To-Repeat:
Set your locale settings to an 8 bit character set like ISO8859-1. Insert
non-breaking spaces in a text, and notice how par converts them to spaces, and
even wrapping the lines on them.
>Fix:
Apply the following patch:
--------------------------------------------------------------------------------
--- par.c.orig Sun Mar 28 16:00:15 2004
+++ par.c Sun Mar 28 16:04:00 2004
@@ -403,7 +403,8 @@
}
continue;
}
- if (isspace(c)) ch = ' ';
+ // Exclude non-breaking space from the class of space chars
+ if (isspace(c) && c != 0xA0) ch = ' ';
else blank = 0;
additem(cbuf, &ch, errmsg);
if (*errmsg) goto rlcleanup;
--------------------------------------------------------------------------------
Thanks in advance,
--
Jean-Baptiste Quenot
http://caraldi.com/jbq/
>Release-Note:
>Audit-Trail:
>Unformatted:
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20040328141941.C538ABD>