expansion bug in 4.0.3?

expansion bug in 4.0.3?

[Danek Duvall]

This example is easily reconstructed on Linux:

    % echo $$
    146
    % print -l ${$(</proc/$$/environ)}
    DISPLAY=:0.0 HOME=/home/duvall LOGNAME=duvall USER=duvall [ ... ]
    % print -l ${$(</proc/146/environ)}
    DISPLAY=:0.0
    HOME=/home/duvall
    LOGNAME=duvall
    USER=duvall
    [ ... ]

The only difference is that in one case the filename is specified with a
parameter, in the other, without.  If, in the former case, I explicitly
split on spaces:

    % print -l ${(s: :)$(</proc/$$/environ)}

then I get the output I want.  Alternatively, if I skip the outer
braces:

    % print -l $(</proc/$$/environ)

then the output also comes out on multiple lines.

(For those not on Linux, this file contains fields separated by null
characters.)

Is this a bug, or is there some behavior I'm not understanding properly?

Thanks,
Danek

PATCH: Re: expansion bug in 4.0.3?

[Bart Schaefer]

On Nov 3, 12:00am, Danek Duvall wrote:
}
}     % echo $$
}     146
}     % print -l ${$(</proc/$$/environ)}
}     DISPLAY=:0.0 HOME=/home/duvall LOGNAME=duvall USER=duvall [ ... ]
}     % print -l ${$(</proc/146/environ)}
}     DISPLAY=:0.0
}     HOME=/home/duvall
}     LOGNAME=duvall
}     USER=duvall
}     [ ... ]
} 
} Is this a bug, or is there some behavior I'm not understanding properly?

It's a bug.  The optimization that causes $(<filename) to be implmented
without forking is not properly saving/restoring some state, such that
the expansion of $$ in the file name causes zsh to behave as if you'd
written it with double quotes, i.e. `print -l ${"$(</proc/$$/environ)"}'.

This bug doesn't exist in 3.0.x because the state that needs to be saved
and restored wasn't introduced until 3.1.9-pws-19.  However, the bug has
probably been present ever since then -- for more than two years.

The following would appear to be the fix.  "make check" still passes.

Index: Src/subst.c
===================================================================
--- Src/subst.c	2001/10/17 14:38:29	1.8
+++ Src/subst.c	2001/11/03 16:34:57
@@ -245,11 +245,13 @@
 mod_export void
 singsub(char **s)
 {
+    int omi = mult_isarr;
     local_list1(foo);
 
     init_list1(foo, *s);
 
     prefork(&foo, PF_SINGLE);
+    mult_isarr = omi;
     if (errflag)
 	return;
     *s = (char *) ugetnode(&foo);


-- 
Bart Schaefer                                 Brass Lantern Enterprises
http://www.well.com/user/barts              http://www.brasslantern.com

Zsh: http://www.zsh.org | PHPerl Project: http://phperl.sourceforge.net   

Re: PATCH: Re: expansion bug in 4.0.3?

[Bart Schaefer]

On Nov 3,  4:43pm, Bart Schaefer wrote:
}
} This bug doesn't exist in 3.0.x because the state that needs to be saved
} and restored wasn't introduced until 3.1.9-pws-19.
                                           ^
                      That should say "3.1.6-pws-19", of course.


-- 
Bart Schaefer                                 Brass Lantern Enterprises
http://www.well.com/user/barts              http://www.brasslantern.com

Zsh: http://www.zsh.org | PHPerl Project: http://phperl.sourceforge.net   

Re: [16215] PATCH: Re: expansion bug in 4.0.3?

[Danek Duvall]

On Sat, Nov 03, 2001 at 04:43:31PM +0000, Bart Schaefer wrote:

> The following would appear to be the fix.  "make check" still passes.

That does the trick.  Thanks!

Danek