From mboxrd@z Thu Jan  1 00:00:00 1970
From: erik quanstrom <quanstro@quanstro.net>
Date: Wed, 12 Mar 2014 19:16:34 -0400
To: 9fans@9fans.net
Message-ID: <a1eeecff62602d8040a292007fa2a7d5@ladd.quanstro.net>
MIME-Version: 1.0
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Subject: [9fans] =?utf-8?q?_sort_-df_vs=2E_=CE=BC?=
Topicbox-Message-UUID: c9716a88-ead8-11e9-9d60-3106f5b1d025

i've been fighting a bit with the manual, and found that =CE=BClawdec(1)
was coming in the middle of the ells.  it appears that sort -df is
wrong.  we should not treat mapto[255] like sign extension.

/n/atom/plan9/sys/src/cmd/sort.c:1292,1298 - sort.c:1292,1298
  		/*
  		 * for characters out of range,
  		 * the table does not do Rflag.
- 		 * ignore is based on mapto[255]
+ 		 * ignore is computed directly.
  		 */
  		if(c !=3D 0 && c < nelem(f->mapto)) {
  			c =3D f->mapto[c];
/n/atom/plan9/sys/src/cmd/sort.c:1299,1305 - sort.c:1299,1305
  			if(c =3D=3D 0)
  				continue;
  		} else {
- 			if(f->mapto[nelem(f->mapto)-1] =3D=3D 0)
+ 			if(f->flags & Iflag)
  				continue;
  			/*
  			 * consider building maps as necessary


this seems too obvious, so have i missed anything?

- erik