From nobody Wed Nov 30 23:49:26 2005
From: Paul Eggert <eggert@CS.UCLA.EDU>
Subject: Re: gawk: length return incorrect value when MB_CUR_MAX > 1
To: Hirofumi Saito <hi_saito@yk.rim.or.jp>
Cc: bug-gawk@gnu.org, KIMURA Koichi <kimura.koichi@canon.co.jp>
Date: Wed, 30 Nov 2005 13:39:56 -0800

Hirofumi Saito <hi_saito@yk.rim.or.jp> writes:

> And then, I tried to use gawk 3.1.5 which I build with sarge.
>
>  $ LANG=ja_JP.utf8 gawk 'BEGIN {print length("abc\0def")}'
>  7
>  $ LANG=ja_JP.eucJP gawk 'BEGIN {print length("abc\0def")}'
>  3

Very strange.  I don't get this result with Debian sarge x86; instead,
I get 3 in both cases.  And that is what I would expect to get, given
the source code.  Perhaps your locales weren't all built?  (Also, I
set LC_ALL rather than LANG; that's safer.)

> By the way, I patched Kimura's patch, then:

Yes, his patch should work.

Here's a slightly more-efficient patch:

--- gawk-3.1.5.orig/node.c-bak	2005-07-26 11:07:43 -0700
+++ gawk-3.1.5/node.c	2005-11-30 13:33:44 -0800
@@ -749,9 +749,10 @@ str2wstr(NODE *n, size_t **ptr)
 		switch (count) {
 		case (size_t) -2:
 		case (size_t) -1:
-		case 0:
 			goto done;
 
+		case 0:
+			count = 1;
 		default:
 			*wsp++ = wc;
 			src_count -= count;


_______________________________________________
bug-gnu-utils@gnu.org
http://lists.gnu.org/mailman/listinfo/bug-gnu-utils