bug in zsh 2.6 beta 11: xkeysymdb

bug in zsh 2.6 beta 11: xkeysymdb

[kpc]

If this is a mailing list, I do not subscribe to it (but follow the
instructions in mailing to it), and would appreciate a CC.

I am upgrading from 2.3.1.  The reason I am not upgrading from a later
version is that no other version compiled and ran correctly on both
Solaris and SunOS.  Everything works fine with 2.3.1 but 2.6 beta 11
fails to read my dot files correctly.

beta11 compiles cleanly but has the following bug and complains of a
syntax error in my .zshrc that I cannot locate.  It says ".zshrc:
parse error near `)' [307]".  I will spare you my .zshrc but hope that
future versions can locate syntax errors better.  Note that 2.3.1 does
not complain at all.  I have no idea where the error is or whether
there is one.  All parentheses and braces and brackets match that I
know of.

Here is the bug.  .zslow is a file that I load only once from my
.zshenv file.  It works with 2.3.1.

% . .zslow
.zslow: no such file or directory: usr/local/X.V11R5/lib/XKeysymDB [291]
.zslow: no such file or directory: usr/local/X.V11R5/lib/XKeysymDB [561]
.zslow: no such file or directory: usr/local/X.V11R5/lib/XKeysymDB [831]
.zslow: no such file or directory: usr/local/X.V11R5/lib/XKeysymDB [1101]
.zslow: no such file or directory: usr/local/X.V11R5/lib/XKeysymDB [1371]
.zslow: no such file or directory: usr/local/X.V11R5/lib/XKeysymDB [1641]
.zslow: no such file or directory: usr/local/X.V11R5/lib/XKeysymDB [1911]
.zslow: no such file or directory: usr/local/X.V11R5/lib/XKeysymDB [2181]
.zslow: no such file or directory: usr/local/X.V11R5/lib/XKeysymDB [2451]
.zslow: no such file or directory: usr/local/X.V11R5/lib/XKeysymDB [2721]
.zslow: no such file or directory: usr/local/X.V11R5/lib/XKeysymDB [2991]
.zslow: no such file or directory: usr/local/X.V11R5/lib/XKeysymDB [3261]
.zslow: no such file or directory: usr/local/X.V11R5/lib/XKeysymDB [3531]
.zslow: no such file or directory: usr/local/X.V11R5/lib/XKeysymDB [3801]
.zslow: no such file or directory: usr/local/X.V11R5/lib/XKeysymDB [4071]
.zslow: no such file or directory: usr/local/X.V11R5/lib/XKeysymDB [4341]
.zslow: no such file or directory: usr/local/X.V11R5/lib/XKeysymDB [4611]
.zslow: no such file or directory: usr/local/X.V11R5/lib/XKeysymDB [4881]

usr/local/X.V11R5/lib/XKeysymDB exists.  I see 3 errors:

	o zsh cares whether it exists
	o zsh says it does not exist
	o zsh infinitely recurses or something

Here is the file.  Hope it helps.

#
#.zslow
#
#this file is dotted by .zshenv
#
#everything here is inherited by subshells or subprocesses.
#

#ugh, recursion
typeset -x zshenvloadedp=t
#touch ~/.beginning-zslow

#!!!!see also .zshrc
umask 007

#typeset -x USER=kpc@ptolemy.arc.nasa.gov
typeset -x homehost=phenotype.arc.nasa.gov			#!!!!
typeset -x homeuid=kpc			#!!!!

typeset -x XKEYSYMDB=/usr/local/X.V11R5/lib/XKeysymDB
#typeset -x LD_LIBRARY_PATH=/usr/local/X.V11R5/lib
#this causes problems with programs copied from sunos
#typeset -x LD_LIBRARY_PATH=/usr/lib:/usr/local/X.V11R5/lib:/usr/ucblib
#works typeset -x LD_LIBRARY_PATH=/usr/local/X.V11R5/lib:/usr/ucblib
#typeset -x LD_LIBRARY_PATH=/usr/local/X.V11R5/lib:/usr/4lib:/usr/ucblib
#typeset -x LD_LIBRARY_PATH=/usr/local/X.V11R5/lib
typeset -x LD_LIBRARY_PATH=/usr/local/lib/motif/usr/lib:/usr/local/X.V11R5/lib

#
#paths
#

#path=($path /usr/ucb /usr/bin /usr/etc)
#:/usr/local/X.V11R5/bin/motif1.2\

#/usr/local/sunos-bin contains binaries from sunos that might be
#slower under solaris due to sunos compatibility library.  i don't
#know anything about this, however.

setpath () {
	if [ -d $1 -a -x $1 ]
	then
		PATH=$PATH:$1
	fi
}

setlocalpaths () {		#local paths in my home directory
	uid=${1:?}

	setpath ~"$uid"/local/user-host/bin
	setpath ~"$uid"/local/host/bin
	setpath ~"$uid"/local/arch/bin
	setpath ~"$uid"/local/site/bin
	setpath ~"$uid"/local/user/bin
}

#The first item in the path should not start with a colon, to prevent ::
#typeset -x PATH=/dev/null/xyzzy-nonexistent-bin
typeset -x PATH=~/sbin
setlocalpaths $LOGNAME
if [ "$LOGNAME" != "$homeuid" ]
then
	setpath ~"$homeuid"/sbin
	setlocalpaths $homeuid
fi

#
#standard paths
#

#
#zsh 2.3.1 at least sometimes prints not a directory if there is a
#non-executable file in path.  also ~user fails if user is a program
#in path.
#

setpath /bin
setpath /usr/bin

setpath /usr/local2/GNU/bin
setpath /usr/local/GNU/bin
setpath /usr/local2/GNU/sparc-sun-sunos4.1.3/bin
setpath /usr/local2/GNU/sparc-sun-solaris2.3/bin
setpath /usr/local2/GNU/sparc-sun-solaris2.4/bin
setpath /usr/local2/GNU/sparc-sun-solaris2.5/bin
setpath /usr/local/GNU/sparc-sun-sunos4.1.3/bin
setpath /usr/local/GNU/sparc-sun-solaris2.3/bin
setpath /usr/local/GNU/sparc-sun-solaris2.4/bin
setpath /usr/local/GNU/sparc-sun-solaris2.5/bin
setpath /usr/gnu/bin
setpath /usr/fsf/bin
setpath /opt/gnu/bin

setpath /usr/local2/bin
setpath /usr/local/bin
setpath /usr/local/packages	#my thing for things like glimpse binaries
setpath /usr/local/share/bin
setpath /usr/share/bin
setpath /usr/5bin
setpath /usr/local/bin/sparc-sun-sunos4.1.3
setpath /usr/local/bin/sparc-sun-solaris2.3
setpath /usr/local/bin/sparc-sun-solaris2.4
setpath /usr/local/bin/sparc-sun-solaris2.5
setpath /usr/local/bin/Sparc10

#phenotype bw motif
setpath /usr/local/lib/motif/usr/bin/X11
setpath /usr/local/X.V11R5/bin
#4.1.3 network
setpath /usr/local/X.V11R5/bin/motif
#setpath /usr/X.V11R5/bin
#setpath /usr/X.V11R5/bin/motif
#setpath /usr/local/X11/bin
#setpath /usr/X11/bin
#setpath /usr/local/X.V11R4/bin
#setpath /usr/local/bin/X.V11R4
#setpath /usr/local/X.V11R6/bin
#setpath /usr/local/X.V11R6/bin/motif
#setpath /usr/X.V11R6/bin

setpath /etc
setpath /usr/etc
setpath /usr/local/etc
setpath /usr/lib
setpath /sbin
setpath /usr/sbin
setpath /usr/kvm
setpath /usr/demo
setpath /usr/games
setpath /usr/sccs
setpath /usr/local/bin/news
setpath /usr/local/lib/news
setpath /usr/lib/news
setpath /usr/ncd/sparc
setpath /usr/diag/sundiag
setpath /usr/diag
setpath /usr/ccs/bin
setpath /opt/SUNWabe/bin
setpath /opt/SUNWcg12/bin
setpath /opt/SUNWdiag/bin
setpath /opt/SUNWdxlib/bin
setpath /opt/SUNWgt/bin
setpath /opt/SUNWits/bin
setpath /opt/SUNWleo/bin
setpath /opt/SUNWrtvc/bin
setpath /opt/SUNWtcx/bin
setpath /usr/ucb
setpath /usr/local/sunos-bin
setpath /usr/lib/sa
setpath /usr/lib/saf
setpath /usr/lib/acct
setpath /usr/lib/nfs
if [ ! -e ~/local/host/.standalone ]
then
	touch ~/.put-foreign-paths-in-zslow
	#stuff mounted on megawatt
	setpath /usr/ccf/bin
	setpath /usr/ccf/X11R5/bin
	setpath /usr/ccf/etc
	setpath /usr/ccf/perl5/bin
	setpath /usr/ccf/perl/bin
	setpath /usr/ccf/graphics/bin
	setpath /usr/ccf/hotjava/bin
	setpath /usr/ccf/texmaker/bin
	setpath /usr/ccf/texmaker/bin
	setpath /usr/ccf/openwin/bin
	setpath /usr/ccf/X11R6/bin
	#stuff on pluto or ptolemy copernicus kronos
	setpath /mnt/pluto/local/GNU/bin
	setpath /mnt/pluto/local/bin
	setpath /mnt/bin
	setpath /usr/openwin/bin
	setpath /usr/annex
	setpath /usr/lib/annex
fi

typeset -x MANPATH
MANPATH=.
MANPATH=${MANPATH}:man
MANPATH=${MANPATH}:/usr/local2/man
MANPATH=${MANPATH}:/usr/local/man
MANPATH=${MANPATH}:/usr/gnu/man
MANPATH=${MANPATH}:/opt/gnu/man
MANPATH=${MANPATH}:/usr/man
MANPATH=${MANPATH}:/usr/share/man
MANPATH=${MANPATH}:/usr/local/man/X.V11R5
MANPATH=${MANPATH}:/usr/ccf/man

typeset -x NAME=${LOGNAME:-?}		      #for gnus
typeset -x TEXEDIT="emacs +%d %s"
typeset -xr FCEDIT=emacs EDITOR=emacs VISUAL=emacs 
typeset -x PAGER=less
typeset -x LESS=iMnQsC
#typeset -x WEBSTERHOST=ames.arc.nasa.gov
typeset -x GZIP=--fast

#!!live dangerously (and try to avoid emacs crashes)
#unlimit filesize datasize stacksize coredumpsize resident descriptors
unlimit
#limit -h coredumpsize 1
#limit coredumpsize 1
#limit -h core 1
#limit core 1
#ulimit?
#soft limits do not seem to be useful for emacs at least

#
#executable paths
#

fix () {
	if [ -x $2 ]
	then
		echo $2
	else
		#echo $1 #!!!!
		whence -p $1
	fi
}

#there is a major problem with this: aliases get overridden.  it seems
typeset -x du=du ls=ls pa=pa man=man id=id

typeset -x cat=cat
typeset -x cut=cut
typeset -x wc=wc
typeset -x sed=sed
typeset -x rm=/bin/rm

typeset -x pk=~/sbin/pk
typeset -x web=~/sbin/web
typeset -x news=~/sbin/news
typeset -x save=~/sbin/save
typeset -x lisp=~/sbin/lisp
typeset -x z=~/sbin/z
typeset -x into=~/sbin/into

typeset -x grep=mygrep
typeset -x mygrepgrep=`fix grep /usr/local/bin/grep`
#!!!!inconsisntency; should be myfgrep shell script.  however, I
#only ever use fgrep for fgrep -vxhf, so the bugs fixed by
#mygrep should not appear.
typeset -x fgrep=`fix fgrep /usr/local/bin/fgrep`

typeset -x rsh=`fix rsh /usr/ucb/rsh`
typeset -x find=`fix find /usr/local/bin/find`
typeset -x gawk=`fix gawk /usr/local2/GNU/bin/gawk`
typeset -x gunzip=`fix gunzip /usr/local2/bin/gunzip`
typeset -x gzip=`fix gzip /usr/local2/bin/gzip`
typeset -x tail=tail
typeset -x diff=`fix diff /usr/local/sunos-bin/diff`
typeset -x head=`fix head /usr/local/sunos-bin/head`
typeset -x join=`fix join /usr/local2/bin/join`
typeset -x sort=`fix sort /usr/local2/bin/sort`
typeset -x tr=`fix tr /usr/ucb/tr`
typeset -x uniq=`fix uniq /usr/local2/bin/uniq`
typeset -x xargs=`fix xargs /usr/local/bin/xargs`

#functions are not exported and variables cannot contain spaces
case `uname -r` in
4.1.3) 	#typeset -x localls='/bin/ls -g'
	typeset -x localpsall='/bin/ps -auxw'
	typeset -x localid='/bin/id'
	typeset -x localdu='/bin/du'
	typeset -x localdf='/bin/df'
	typeset -x localman='man'
;;
5.3|5.4|5.5|*) 	#typeset -x localls='/bin/ls'
	#in 5.2, ps -l shows size
	typeset -x localpsall='/bin/ps -cefj'
	typeset -x localid='/bin/id -a'
	typeset -x localman='/bin/man -a'
	typeset -x localdu='/usr/ucb/du'
	typeset -x localdf='/usr/ucb/df'
	#typeset -x CC=/usr/local/bin/cc
;;
esac

export df=$localdf
export du=$localdf

touch ~/.loaded-zslow

---
kpc@ptolemy.arc.nasa.gov.  AI, multidisciplinary neuroethology, info filtering.
Down with Caps-Lock!  Boycott keyboard makers when it displaces Control & Meta!

Re: bug in zsh 2.6 beta 11: xkeysymdb

[P.Stephenson]

kpc@ptolemy.arc.nasa.gov wrote:
> beta11 compiles cleanly but has the following bug and complains of a
> syntax error in my .zshrc that I cannot locate.  It says ".zshrc:
> parse error near `)' [307]".  I will spare you my .zshrc but hope that
> future versions can locate syntax errors better.  Note that 2.3.1 does
> not complain at all.  I have no idea where the error is or whether
> there is one.  All parentheses and braces and brackets match that I
> know of.

Maybe you're using the csh-like syntax 'if (...) { ... }', in which
case `setopt cshjunkiesyntax' at the top of .zshrc (or if you're
really set on the syntax, in ~/.zshenv).  This change was forced
because there is no way in general of telling whether the parentheses
represent a C-shell-like test or a Bourne-shell-like subshell and it
was felt the latter should take precedence in case of conflict.

> % . .zslow
> .zslow: no such file or directory: usr/local/X.V11R5/lib/XKeysymDB [291]
> ...

May be problems with the Berkeley compatibility library on Solaris
again, judging from the initial part of the file name missing ---
although it's not actually obvious why that would be relevant, it's
worth checking: it's clearly a system-specific problem.  Here's the
relevant bit from the FAQ.  (As far as I know, this is the only
compatibility problem for Solaris of any version of 2.6.)

  *Note for Solaris 2.2 and 2.3*: The UCB versions of the routines for
  reading directories are not usable (the struct definitions are2
  incompatible with the ones assumed by zsh).  The symptom of this is
  that globbed filenames in the compiled version of zsh will be
  missing the first two letters.  To avoid this, make sure you compile
  zsh without any reference to /usr/ucblib in (e.g.) your
  LD_LIBRARY_PATH.

-- 
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 beta 11

[Peter William Stephenson]

According to kpc:
> % . .zslow
> .zslow: no such file or directory: usr/local/X.V11R5/lib/XKeysymDB [291]
> <and over and over again...>

O.K., I can now confirm this... it's a classic exit-instead-of-_exit
bug AGAIN --- I fixed this just a couple of months ago and the _exit I
put into execcmd() has mysteriously vanished from the face of the
earth.  (I'm pretty sure it was the same bug, apologies if it was
something slightly different which became redundant.)  The subshell
forked to do the `fix ...` was calling exit(), which messed up the
files for the parent shell.  (This is going to cause apparently random
inexplicable bugs on all sorts of SYSV-like machines until it's
re-fixed.)

kpc: the ${EMACS:-} bug you reported by private email I've just sent a
patch to the list for.

If this doesn't fix the problems, holler again.

*** Src/exec.c~	Tue Nov  7 04:43:04 1995
--- Src/exec.c	Tue Nov  7 16:19:25 1995
***************
*** 1514,1520 ****
  		/* only save the history file on a real exec */
  		if ((cmd->flags & CFLAG_EXEC) && unset(NORCS) && interact)
  		    savehistfile(getsparam("HISTFILE"), 1, isset(APPENDHISTORY) ? 3 : 0);
! 		exit(lastval);
  	    }
  
  	    if (!forked && !assign)
--- 1514,1523 ----
  		/* only save the history file on a real exec */
  		if ((cmd->flags & CFLAG_EXEC) && unset(NORCS) && interact)
  		    savehistfile(getsparam("HISTFILE"), 1, isset(APPENDHISTORY) ? 3 : 0);
! 		if (subsh)
! 		    _exit(lastval);
! 		else
! 		    exit(lastval);
  	    }
  
  	    if (!forked && !assign)

-- 
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 beta 11

[Vinnie Shelton]

In message <9511071542.AA01774@sgi.ifh.de>, Peter William Stephenson wrote:
>According to kpc:

>O.K., I can now confirm this... it's a classic exit-instead-of-_exit
>bug AGAIN --- I fixed this just a couple of months ago and the _exit I
>put into execcmd() has mysteriously vanished from the face of the
>earth.  (I'm pretty sure it was the same bug, apologies if it was
>something slightly different which became redundant.)  The subshell
>forked to do the `fix ...` was calling exit(), which messed up the
>files for the parent shell.  (This is going to cause apparently random
>inexplicable bugs on all sorts of SYSV-like machines until it's
>re-fixed.)

I was having big problems under Solaris 2.3 until I applied
this patch.  In particular $(whence foo) in my init scripts was
bombing.  Thanks, Peter.

--Vin

Re: bug in zsh 2.6 beta 11

[Richard Coleman]

> O.K., I can now confirm this... it's a classic exit-instead-of-_exit
> bug AGAIN --- I fixed this just a couple of months ago and the _exit I
> put into execcmd() has mysteriously vanished from the face of the
> earth.  (I'm pretty sure it was the same bug, apologies if it was
> something slightly different which became redundant.)  The subshell
> forked to do the `fix ...` was calling exit(), which messed up the
> files for the parent shell.  (This is going to cause apparently random
> inexplicable bugs on all sorts of SYSV-like machines until it's
> re-fixed.)

Well, the disappearance isn't mysterous.  I removed it when I changed
execcmd() to use the CFLAG_FAKE_EXEC flag.  That's why I was making
test releases so we could catch these types of bugs before the release.
But it appears no one tried it out.

Anyway, I'll probably make a beta11a in the next couple of days to
correct some of these problems.

rc

Re: bug in zsh 2.6 beta 11

[kpc]

Thanks for confirming that the problems were not just mine.  Unless
you need me to test it, instead of finding the patch in the list
archives, I will wait for a later release.

I tried the patch you sent, and zsh now works under emacs but does not
work under xterm.

FWIW, I think that I have tried about 5 versions of zsh, including the
most recent release and several recent betas, and that the only one I
have gotten to work is 2.3.1.  Maybe things will stabilize when 2.6
goes "gamma".  Anybody know when 2.6 will likely be released?  I am
just looking for reliability and speed, since I am not a feature
creature and I prefer not to attempt to hack on the zsh source.

---
kpc@ptolemy.arc.nasa.gov.  AI, multidisciplinary neuroethology, info filtering.

Re: bug in zsh 2.6 beta 11

[Richard Coleman]

> Thanks for confirming that the problems were not just mine.  Unless
> you need me to test it, instead of finding the patch in the list
> archives, I will wait for a later release.
> 
> I tried the patch you sent, and zsh now works under emacs but does not
> work under xterm.

In what way doesn't it work under xterm.  On what architecture? What OS?
After the subshell patch, what is going wrong?

> FWIW, I think that I have tried about 5 versions of zsh, including the
> most recent release and several recent betas, and that the only one I
> have gotten to work is 2.3.1.  Maybe things will stabilize when 2.6
> goes "gamma".  Anybody know when 2.6 will likely be released?  I am
> just looking for reliability and speed, since I am not a feature
> creature and I prefer not to attempt to hack on the zsh source.

I've given up trying to estimate when we will release the next production
version of zsh (which will be labeled zsh-3.0).  Since zsh has gotten a
reputation for being buggy, I want to make sure things are ultra-solid
before we make our production release.  Maybe in 3 or 4 months, but no
promises.

rc

Re: bug in zsh 2.6 beta 11

[kpc]

Quoth Richard Coleman on 1995 November 8:
> In what way doesn't it work under xterm.  On what architecture? What OS?
> After the subshell patch, what is going wrong?

In the same way I reported in my previous message: it hangs after
printing the first prompt.  Solaris 2.4, SS20, X11R5.  This is before
and after applying the patch CC'ed to me.

> reputation for being buggy, I want to make sure things are ultra-solid
> before we make our production release.  Maybe in 3 or 4 months, but no
> promises.

Didn't know it had that reputation generally.  I'll probably wait and
use 2.3.1 until then.

2.3.1 works OK for me except for a few scoping bugs which I have
reported, the fact that it hangs when certain pipelines are
interrupted, which I think is known, and very minor bugs.  Just hoping
a newer version will be faster.

---
kpc@ptolemy.arc.nasa.gov.  AI, multidisciplinary neuroethology, info filtering.
Prescriptions are no more moral than ration slips for food, vitamins, or shoes.