recent zsh-4.3.0 snapshot hangs

recent zsh-4.3.0 snapshot hangs

[Alexey Tourbin]

[-- Attachment #1: Type: text/plain, Size: 424 bytes --]

Hello,

Recent zsh snashot hangs and fails to produce a prompt.
When started as 'zsh -x', the following lines are executed
right before the hang occures:

+compdump:95> read -rA _d_line
+compdump:101> print
+compdump:106> _d_als=+compdump:1> typeset +fm '_*'

I've got a few my own patches to zsh, but they are least suspectable
at the moment.  Can anybody reproduce this?

-- 
Alexey Tourbin
ALT Linux Team

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: recent zsh-4.3.0 snapshot hangs

[Alexey Tourbin]

[-- Attachment #1: Type: text/plain, Size: 849 bytes --]

On Mon, Jul 18, 2005 at 12:49:00PM +0400, Alexey Tourbin wrote:
> Recent zsh snashot hangs and fails to produce a prompt.
> When started as 'zsh -x', the following lines are executed
> right before the hang occures:
> 
> +compdump:95> read -rA _d_line
> +compdump:101> print
> +compdump:106> _d_als=+compdump:1> typeset +fm '_*'

[test@solemn test]$ zsh -cx 'HOSTNAME=`/bin/hostname`'
+/etc/zshenv:9> [ -f /etc/profile.d/tmpdir.sh ']'
+/etc/zshenv:9> . /etc/profile.d/tmpdir.sh
+/etc/profile.d/tmpdir.sh:3> [ -z /home/test/tmp ']'
+/etc/profile.d/tmpdir.sh:3> [ ! -d /home/test/tmp ']'
+/etc/zshenv:10> TMPPREFIX=/home/test/tmp/zsh
+zsh:1> HOSTNAME=+zsh:1> /bin/hostname
[... zsh hangs ...]

This is 20050716 snapshot, the last change was made in
Completion/Unix/Command/_python.

I had no such problem with 20050609 snapshot.

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: recent zsh-4.3.0 snapshot hangs

[Peter Stephenson]

Alexey Tourbin wrote:
> [test@solemn test]$ zsh -cx 'HOSTNAME=3D`/bin/hostname`'
> +/etc/zshenv:9> [ -f /etc/profile.d/tmpdir.sh ']'
> +/etc/zshenv:9> . /etc/profile.d/tmpdir.sh
> +/etc/profile.d/tmpdir.sh:3> [ -z /home/test/tmp ']'
> +/etc/profile.d/tmpdir.sh:3> [ ! -d /home/test/tmp ']'
> +/etc/zshenv:10> TMPPREFIX=3D/home/test/tmp/zsh
> +zsh:1> HOSTNAME=3D+zsh:1> /bin/hostname
> [... zsh hangs ...]

So it looks like it's independent of startup files:  are you able to
confirm this?  Does it still happen after a rebuild from scratch?

The latest code is working OK on Solaris 8 on SPARC and Fedora Core 3 on
x86.  At worst you might be stuck with a binary search of the archive.

-- 
Peter Stephenson <pws@csr.com>                  Software Engineer
CSR PLC, Churchill House, Cambridge Business Park, Cowley Road
Cambridge, CB4 0WZ, UK                          Tel: +44 (0)1223 692070


**********************************************************************
This email and any files transmitted with it are confidential and
intended solely for the use of the individual or entity to whom they
are addressed. If you have received this email in error please notify
the system manager.

**********************************************************************

Re: recent zsh-4.3.0 snapshot hangs

[Alexey Tourbin]

[-- Attachment #1: Type: text/plain, Size: 2289 bytes --]

On Mon, Jul 18, 2005 at 10:47:29AM +0100, Peter Stephenson wrote:
> Alexey Tourbin wrote:
> > [test@solemn test]$ zsh -cx 'HOSTNAME=3D`/bin/hostname`'
> > +/etc/zshenv:9> [ -f /etc/profile.d/tmpdir.sh ']'
> > +/etc/zshenv:9> . /etc/profile.d/tmpdir.sh
> > +/etc/profile.d/tmpdir.sh:3> [ -z /home/test/tmp ']'
> > +/etc/profile.d/tmpdir.sh:3> [ ! -d /home/test/tmp ']'
> > +/etc/zshenv:10> TMPPREFIX=3D/home/test/tmp/zsh
> > +zsh:1> HOSTNAME=3D+zsh:1> /bin/hostname
> > [... zsh hangs ...]
> 
> So it looks like it's independent of startup files:  are you able to
> confirm this?  Does it still happen after a rebuild from scratch?

I do build from scratch (but I have ccache).
Here is strace log for HOSTNAME=`/bin/hostname`:

[pid  3629] uname({sys="Linux", node="solemn.turbinal.org", ...}) = 0
[pid  3629] fstat64(1, {st_mode=S_IFIFO|0600, st_size=0, ...}) = 0
[pid  3629] mmap2(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x4002a000
[pid  3629] write(1, "solemn.turbinal.org\n", 20) = 20
[pid  3629] munmap(0x4002a000, 4096)    = 0
[pid  3629] exit_group(0)               = ?
fcntl64(12, F_GETFL)                    = 0 (flags O_RDONLY)
fstat64(12, {st_mode=S_IFIFO|0600, st_size=0, ...}) = 0
mmap2(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x401ca000
_llseek(12, 0, 0xbffff324, SEEK_CUR)    = -1 ESPIPE (Illegal seek)
read(12, "solemn.turbinal.org\n", 4096) = 20
read(12, "", 4096)                      = 0
close(12)                               = 0
munmap(0x401ca000, 4096)                = 0
rt_sigprocmask(SIG_BLOCK, [CHLD], [CHLD], 8) = 0
kill(3629, SIG_0)                       = 0
rt_sigprocmask(SIG_SETMASK, ~[HUP INT CHLD], [CHLD], 8) = 0
--- SIGCHLD (Child exited) @ 0 (0) ---
rt_sigprocmask(SIG_BLOCK, ~[], ~[HUP INT KILL STOP], 8) = 0
rt_sigprocmask(SIG_SETMASK, ~[HUP INT KILL STOP], ~[KILL STOP], 8) = 0
wait4(-1, [{WIFEXITED(s) && WEXITSTATUS(s) == 0}], WNOHANG|WSTOPPED, {ru_utime={0, 10000}, ru_stime={0, 0}, ...}) = 3629
getrusage(RUSAGE_CHILDREN, {ru_utime={0, 10000}, ru_stime={0, 0}, ...}) = 0
wait4(-1, 0xbfffece0, WNOHANG|WSTOPPED, 0xbfffece4) = -1 ECHILD (No child processes)
sigreturn()                             = ? (mask now ~[HUP INT KILL CHLD STOP])
pause(

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: recent zsh-4.3.0 snapshot hangs

[Bart Schaefer]

On Jul 18,  2:17pm, Alexey Tourbin wrote:
}
} Here is strace log for HOSTNAME=`/bin/hostname`:
} 
} pause(

What platform is this?  (Linux, but can you be more specific?  Is it
multiprocessor or otherwise using an SMP kernel?)

Unless your config.h defines BROKEN_POSIX_SIGSUSPEND, there's only
one call to pause() in the sources, in signal_suspend().

I can't see any recent changes that would be even remotely related to
this, except for zsh-workers/21391, and that's from more than one
snapshot ago.

Re: recent zsh-4.3.0 snapshot hangs

[Alexey Tourbin]

[-- Attachment #1: Type: text/plain, Size: 1211 bytes --]

On Mon, Jul 18, 2005 at 03:29:36PM +0000, Bart Schaefer wrote:
> On Jul 18,  2:17pm, Alexey Tourbin wrote:
> }
> } Here is strace log for HOSTNAME=`/bin/hostname`:
> } 
> } pause(
> 
> What platform is this?  (Linux, but can you be more specific?  Is it
> multiprocessor or otherwise using an SMP kernel?)

$ uname -rms
Linux 2.4.20-alt7-up i686
$ rpm -q glibc
glibc-2.3.3.200406160000-alt1
$

> Unless your config.h defines BROKEN_POSIX_SIGSUSPEND, there's only
> one call to pause() in the sources, in signal_suspend().

One important thing I didn't realize until now: the whole build process
was run under strace(1).  (Don't ask me why.  (In short, this was
build-time dependency analysis.))

So I tend to blame strace now, since it does some clever mangling
which affects process control behaviour, which in turn could affect
configure tests:

	If the parent process decides to wait(2) for a child
	that  is  currently  being  traced,  it is suspended
	until an appropriate child process either terminates
	or  incurs a signal that would cause it to terminate
	(as determined from the child's current signal  dis-
	position).

Now I'm trying to find out the exact reason.

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]