On Mon, Jun 28, 2010 at 03:24:54PM +0200, Werner Fink wrote:
> On Fri, May 28, 2010 at 02:27:26PM +0200, Werner Fink wrote:
> > On Fri, May 28, 2010 at 02:00:36PM +0200, Werner Fink wrote:
> > > /suse/werner> ksh -c 'echo ${.sh.version}'
> > > Version JM 93t+ 2010-05-24
> > > /suse/werner> ksh ksh93/src/cmd/ksh93/tests/sjis.sh
> > > sjis.sh[61]: Shift-JIS encoding failed
> > >
> > > it seems that the `test' builtin fails on the characters
> > >
> > > \x81\x5c
> > > \x81\x7c
> > >
> > > the scriptlet for testing this can be found in the
> > > attachment.
> > >
> > >
> > > Werner
> >
> > Beside this I've found that the program pty does not exit
> > and that timetype.sh may missing an
> >
> > export LC_TIME=en_US
> >
> > or similar as
> >
> > test timetype.sh begins at 2010-05-28+11:01:48
> > timetype.sh[61]: tt[0] is not 2pm
> > timetype.sh[75]: tt[2pm] is not 2pm
> >
> > does fail as well.
>
> A few facts:
>
> * the error occurs only on line using the external
> /usr/bin/printf program and does not happen if
> the builtin is used.
>
> * if set
>
> set -o noglob
>
> the error also does not happen.
>
> * For /usr/bin/printf '\x81\x5c' the result depends
> on how the variables are handled. That is that
>
> test $($printf "\x81\x$second") != ${mbchar}
> test $(${printf} "\x81\x$second") != ${mbchar}
> test $($printf "\x81\x${second}") != ${mbchar}
> test $(${printf} "\x81\x${second}") != ${mbchar}
> test $($printf "\x81\x$second") != ${mbchar}
>
> show different results.
>
> Werner
Now it looks like that this change cause the trouble:
--- ksh93.2010-03-09/src/cmd/ksh93/sh/macro.c 2007-12-20 18:50:28.000000000
+0100
+++ ksh93.2010-05-27/src/cmd/ksh93/sh/macro.c 2007-12-20 18:50:28.000000000
+0100
[...]
@@ -2022,7 +2052,7 @@ static void comsubst(Mac_t *mp,register
/* read command substitution output and put on stack or here-doc */
sfpool(sp, NIL(Sfio_t*), SF_WRITE);
sh_offstate(SH_INTERACTIVE);
- while((str=(char*)sfreserve(sp,SF_UNBOUND,0)) && (c = sfvalue(sp))>0)
+ while((str=(char*)sfreserve(sp,SF_UNBOUND,0)) &&
(c=bufsize=sfvalue(sp))>0)
{
#if SHOPT_CRNL
/* eliminate <cr> */
... that is that sfvalue() sometimes returns the value 2 or 3 which
then cause to switch into the boolean
if(++c < bufsize)
str[c] = 0;
else
{
[...]
of comsubst() in ksh93/sh/macro.c and now the multi byte character
is splitted which is a problem for Shift_JIS as now the second
byte seems to be handled as a single byte (that is that 0x5c and 0x7c
aka `\' and `|') and maybe are detected as special shell characters.
Changing the `++c' into `c++' seems to avoid the split and does
solve the problem. The question rises if this is correct?
Werner
--
"Having a smoking section in a restaurant is like having
a peeing section in a swimming pool." -- Edward Burr
_______________________________________________
ast-developers mailing list
[email protected]
https://mailman.research.att.com/mailman/listinfo/ast-developers
