Hello,
In some multibyte encodings, such as Shift_JIS[1], the trailing bytes of
a multibyte character may contain 7-bit character. That could confuse
the 'read' command on ksh.
$ /bin/echo -e '\x95\x5c' > x
$ LANG=ja_JP.SJIS ksh -c 'read a' < x
zsh: exit 1 LANG=ja_JP.SJIS ksh -c 'read a' < x
Note that the byte sequence '\x95\x5c' represents a multibyte character
U+8868, while '\x5c' is a backslash when interpreted as a single byte
character.
The attached patch tries to fix this.
Index: ast-ksh.2014-09-29/src/cmd/ksh93/bltins/read.c
===================================================================
--- ast-ksh.2014-09-29.orig/src/cmd/ksh93/bltins/read.c
+++ ast-ksh.2014-09-29/src/cmd/ksh93/bltins/read.c
@@ -728,6 +728,9 @@ int sh_readline(register Shell_t *shp,ch
del = 0;
while(1)
{
+#if SHOPT_MULTIBYTE
+ ssize_t z;
+#endif /* SHOPT_MULTIBYTE */
switch(c)
{
#if SHOPT_MULTIBYTE
@@ -864,11 +867,31 @@ int sh_readline(register Shell_t *shp,ch
}
/* skip over word characters */
wrd = -1;
+#if SHOPT_MULTIBYTE
+ /* skip a preceding multibyte character if any */
+ if(c==0 && (z=mbsize(cp-1))>1)
+ cp += z - 1;
+#endif /* SHOPT_MULTIBYTE */
while(1)
{
+#if !SHOPT_MULTIBYTE
while((c=shp->ifstable[*cp++])==0)
if(!wrd)
wrd = 1;
+#else /* !SHOPT_MULTIBYTE */
+ while(1)
+ {
+ if((c=shp->ifstable[*cp])!=0)
+ {
+ cp++;
+ break;
+ }
+ if((z=mbsize(cp))>=1)
+ cp += z;
+ if(!wrd)
+ wrd = 1;
+ }
+#endif /* SHOPT_MULTIBYTE */
if(inquote)
{
if(c==S_QUOTE)
Footnotes:
[1] https://en.wikipedia.org/wiki/Shift_JIS
Regards,
--
Daiki Ueno
_______________________________________________
ast-developers mailing list
[email protected]
http://lists.research.att.com/mailman/listinfo/ast-developers
