Date: Sat, 1 Sep 2012 13:40:11 GMT From: Steffen "Daode" Nurpmeso <sdaoden@gmail.com> To: freebsd-bugs@FreeBSD.org Subject: Re: bin/170651: On 9.0-RELEASE#0 and master sh(1) gobbles high bit at first Message-ID: <201209011340.q81DeBj2013615@freefall.freebsd.org>
next in thread | raw e-mail | index | archive | help
The following reply was made to PR bin/170651; it has been noted by GNATS.
From: Steffen "Daode" Nurpmeso <sdaoden@gmail.com>
To: bug-followup@FreeBSD.org
Cc: eadler@FreeBSD.org, jilles@stack.nl
Subject: Re: bin/170651: On 9.0-RELEASE#0 and master sh(1) gobbles high bit at first
Date: Sat, 01 Sep 2012 15:33:41 +0200
Steffen "Daode" Nurpmeso <sdaoden@gmail.com> wrote:
|Jilles Tjoelker <jilles@stack.nl> wrote:
||This PR bin/170651 does in fact look valid. It looks like libedit only
||partially picks up an LC_CTYPE change. Whether it shows '\OOO' or real
||characters with bit 7 set changes immediately; whether it accepts
||characters with bit 7 set changes only after doing something that causes
||libedit to be reinitialized (such as 'set +o').
||
||This is probably a bug in libedit and/or sh's use of it, although there
||are bigger bugs in libedit and sh's use of it.
||
||--
||Jilles Tjoelker
|
|As all locale changes effectively boil down in \
|var.c:{update,init}charset(),
|either call histedit() in updatecharset(), which is what the first (-easy)
|diff does, or introduce a new function which only performs the necessary
|libedit call-in to update the mapping (second patch, -harder).
|Since histedit() always causes a reparse of editrc(5) (and AFAIU), i would
|vote for the harder patch :)
Hmm, looking deeper at the code as such i indeed see some more
issues. E.g., bindings get all lost if a set command occurs, as
in '$ bind ^H sh-complete', '$ set +o' and it's gone ... and
everything around in there.
So i think this should encapsulate itself further than what
currently happens and what i've sent to you already, and instead
act upon real changes on the actual iflag/Vflag/Eflag combination
only.
I've still not wrapped my head around that entirely, but i (also
still) wonder why there is not a single histedit_init() in main(),
and maximally a histedit_destroy() after an exec (forgotten
where), as *iflag* is a pretty constant condition!? Then a single
histedit_on_change() could be called and conditionalize further
execution on 'is_init && STATE_CHANGE_OCCURRED'.
I would like to give that a try now..
I'll append a patch which does some cleanup on really ugly code.
Thanks and ciao,
--steffen
diff --git a/bin/sh/input.c b/bin/sh/input.c
index 12f285f..1da03c6 100644
--- a/bin/sh/input.c
+++ b/bin/sh/input.c
@@ -59,8 +59,10 @@ __FBSDID("$FreeBSD$");
#include "error.h"
#include "alias.h"
#include "parser.h"
-#include "myhistedit.h"
#include "trap.h"
+#ifndef NO_HISTORY
+# include "myhistedit.h"
+#endif
#define EOF_NLEFT -99 /* value of parsenleft when EOF pushed back */
@@ -102,8 +104,6 @@ static struct parsefile *parsefile = &basepf; /* current input file */
int init_editline = 0; /* editline library initialized? */
int whichprompt; /* 1 == PS1, 2 == PS2 */
-EditLine *el; /* cookie for editline package */
-
static void pushfile(void);
static int preadfd(void);
static void popstring(void);
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201209011340.q81DeBj2013615>