Bug in zsh-2.6-beta21

Bug in zsh-2.6-beta21

[alain (a.) caron]

I found the following bug in zsh-2.6-beta20 and beta21:

I typed in the following:

exec <3 <(ls -l)

--->  zsh: no such file or directory: 3

So far so good, but after that it seems that zle is broken as the up
arrow does not retrieve the last command entered.  Instead, the cursor
just goes up.  The tab character does not perform completion anymore,
it just inserts a TAB.  This was on a xterm window.

I run zsh on HP-UX 9.01 and compiled it with the HP C compiler.

Regards,

Alain Caron
Nortel Technology, Montr?al, Qu?bec, Canada
email: alainc@nortel.ca
phone: 514-765-7718 or ESN 852-7718

Re: Bug in zsh-2.6-beta21

[Peter Stephenson]

alainc@nortel.ca wrote:
> I found the following bug in zsh-2.6-beta20 and beta21:
> 
> I typed in the following:
> 
> exec <3 <(ls -l)
> 
> --->  zsh: no such file or directory: 3
> 
> So far so good, but after that it seems that zle is broken as the up
> arrow does not retrieve the last command entered.  Instead, the cursor
> just goes up.  The tab character does not perform completion anymore,
> it just inserts a TAB.  This was on a xterm window.

Zsh thinks the <(ls -l) is a real argument, which is not unreasonable,
so it thinks it's doing a real exec.  However, the <3 open fails and
it decides it not going to do a real exec after all and returns from
execcmd().  The problem is that by that time entersubsh() has been
called to set zsh up for becoming an external program, so at the error
return lots of things (zle, monitor, job tables) are set up wrongly.

What's the answer?

1) Do we delay entersubsh() until after the redirections are handled?
I don't see why that shouldn't work.  It might fix similar potential
problems with globlist.  (Remember, it doesn't affect whether the
shell *really* becomes a subshell, just the settings.)

2) Or should this entersubsh() be delayed for safety until immediately
before the external command is executed?

2) Or is the answer to alter the behaviour of the third argument to
entersubsh (the `fake' flag) so that even less happens on an exec?  (I
believe it needs to be called in this case really just to set up the
terminal for an external command and internal settings are irrelevant
--- but in this particular case we shouldn't really be setting
anything.)



By the way, I know you're not interested, but to me `depreciate' has
only ever meant `decrease in [usually monetary] value', whatever the
dictionary says.  I've never in my life heard it used transitively ---
wrong side of the Atlantic, perhaps.

-- 
Peter Stephenson <pws@ifh.de>       Tel: +49 33762 77366
WWW:  http://www.ifh.de/~pws/       Fax: +49 33762 77330
Deutches Electronen-Synchrotron --- Institut fuer Hochenergiephysik Zeuthen
DESY-IfH, 15735 Zeuthen, Germany.

Re: Bug in zsh-2.6-beta21

[Peter Stephenson]

I wrote:
> alainc@nortel.ca wrote:
> > exec <3 <(ls -l)
> > 
> > --->  zsh: no such file or directory: 3
> > 
> > So far so good, but after that it seems that zle is broken as the up
> > arrow does not retrieve the last command entered.
> 
> 1) Do we delay entersubsh() until after the redirections are handled?

Here's a patch for this.  It's fairly simple, so I don't see any
problems with it.

*** Src/exec.c.subsh	Thu Jul  4 15:30:39 1996
--- Src/exec.c	Thu Jul  4 15:39:05 1996
***************
*** 1102,1108 ****
      int save[10];
      int fil, dfil, is_cursh, type, i;
      int nullexec = 0, assign = 0, forked = 0;
!     int is_shfunc = 0, is_builtin = 0;
      /* Various flags to the command. */
      int cflags = 0, checked = 0;
  
--- 1102,1108 ----
      int save[10];
      int fil, dfil, is_cursh, type, i;
      int nullexec = 0, assign = 0, forked = 0;
!     int is_shfunc = 0, is_builtin = 0, is_exec = 0;
      /* Various flags to the command. */
      int cflags = 0, checked = 0;
  
***************
*** 1420,1426 ****
  	/* This is an exec (real or fake) for an external command.    *
  	 * Note that any form of exec means that the subshell is fake *
  	 * (but we may be in a subshell already).                     */
! 	entersubsh(how, 1, 1);
      }
  
      if (!(cflags & BINF_NOGLOB))
--- 1420,1426 ----
  	/* This is an exec (real or fake) for an external command.    *
  	 * Note that any form of exec means that the subshell is fake *
  	 * (but we may be in a subshell already).                     */
! 	is_exec = 1;
      }
  
      if (!(cflags & BINF_NOGLOB))
***************
*** 1551,1556 ****
--- 1551,1563 ----
      }
  
      if (unset(NOEXEC)) {
+ 	/*
+ 	 * We delay the entersubsh() to here when we are exec'ing
+ 	 * the current shell (including a fake exec to run a builtin then
+ 	 * exit) in case there is an error return.
+ 	 */
+ 	if (is_exec)
+ 	    entersubsh(how, 1, 1);
  	if (type >= CURSH) {
  	    static int (*func[]) _((Cmd)) =
  	    {

-- 
Peter Stephenson <pws@ifh.de>       Tel: +49 33762 77366
WWW:  http://www.ifh.de/~pws/       Fax: +49 33762 77330
Deutches Electronen-Synchrotron --- Institut fuer Hochenergiephysik Zeuthen
DESY-IfH, 15735 Zeuthen, Germany.

Re: bug in zsh-2.6-beta21

[Hrvoje Niksic]

Vincent Lefevre (Vincent.Lefevre@lip.ens-lyon.fr) wrote:
> This is OK. But under zsh 2.6-beta21:
> 
> $ var=`echo "ab\ncd\nef"`
> $ echo $var
> ab
> cd
> ef
> $ var=`echo "ab\ncd\nef\n"`
> $ echo $var
> ab
> cd
> ef
> $
> 
> The last "\n" is not printed.

I think this is good behaviour sh, ksh and bash also strip trailing
newlines. It should maybe be documented in the manual. E.g. bash
manual states:
              $(command)
       or
              `command`
       Bash performs  the  expansion  by  executing  command  and
       replacing  the command substitution with the standard out-
       put of the command, with any trailing newlines deleted.

Whereas zsh manual says:
       A  command  enclosed  in  parentheses preceded by a dollar
       sign, like so: $(...) or quoted with grave accents:  `...`
       is replaced with its standard output.  If the substitution
       is not enclosed in double quotes,  the  output  is  broken
       into  words  using  the  IFS  parameter.  The substitution

Removing trailing newlines is *not* specifically mentioned.

-- 
hniksic@srce.hr              |  Student of electrical engineering
hniksic@fly.cc.fer.hr        |  University of Zagreb, Croatia
------------------------------------------------------------------
`VI' - An editor used by those heretics that don't subscribe to
       the Emacs religion.