Re: Hang with 1.5.18, 1.5.19 snapshot 20051029 (now with 20051117)

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view
|

Re: Hang with 1.5.18, 1.5.19 snapshot 20051029 (now with 20051117)

Peter Rehley

Hi

I did some more testing and looking through source code, and it seems  
that the process that is hanging hasn't initialized yet.  When I do  
the cygwin kill command it returns with "No such process" and based  
on what I see in the code (pinfo.cc, pinfo::init) that can happen if  
the process is still waiting to be initialized.  Tthat's one case,  
the other case is when the process really doesn't exist.  But the  
process does exist, and it shows up in ps command and windows process  
viewer.

It's a guess that the hung process is waiting to be initialized, but  
I did try resuming the process using a program called "process  
explorer", and the hung process continued, finished, and the main  
test_configure also continued.

This implies that wait_sig (sigproc.cc) isn't called sometimes.

Of course this is pure speculation.  I'll keep trying to see if I can  
provide information that is more useful.

Peter

On Oct 31, 2005, at 4:09 PM, Peter Rehley wrote:

> Hi,
>
> I have a problem where a configure script is hanging.  I first saw  
> the behavior in 1.5.18, and it's still there in the latest  
> snapshot. The only machines that we are seeing it hang on are  
> windows 2000 machines, sp4, with duel pentinum 933 mhz processors,  
> and using ssh to login to the machine.  I haven't been able to  
> reproduce the problem on single processor machines or when ssh is  
> not used.
>
> Under 1.5.18, the hang occurred about 1 in ten times in the  
> test_configure script (provided in the bash_test.tar.bz2 file.  
> Under the latest snapshot it's about 1 in 900.
>
> When the hang happens it appears that a process is completed, but  
> still can be found in the process directory.  The cmdline file says  
> <defunct>, but the process still shows up in the process list (ps -
> ef).  If I try to clean up by killing the process, the kill command  
> says that the process doesn't exist.  The only way that I can make  
> the hung process go away is by using the task manager to kill the  
> process.
>
> The simplest test I've gotten down to is:
>
> ### Simple Test
> #! /bin/sh
> # Guess values for system-dependent variables and create Makefiles.
> # Generated by GNU Autoconf 2.59 for expr-configure 1.5.11-1.
> #
> # Report bugs to <cygwin at cygwin dot com>.
> #
> # Copyright (C) 2003 Free Software Foundation, Inc.
> # This configure script is free software; the Free Software Foundation
> # gives unlimited permission to copy, distribute and modify it.
> ## --------------------- ##
> ## M4sh Initialization.  ##
> ## --------------------- ##
>
> set -xv
>
> count=0
> while [ ! -f stop ] ; do
>   as_var=LC_MONETARY
>   if (test -z "`(eval $as_var=C; export $as_var; echo ho) 2>&1`");  
> then
>     echo "hi"
>     eval $as_var=C; export $as_var
>   fi
>   count=`expr $count + 1`
>   echo $count
> done
> ### End simple test
>
> Someplace in the eval line the hang occurs.  Unfortunately I  
> haven't had success when using strace.
>
> If I've missed anything or there are questions about the above,  
> please let me know.
> Peter
>
> <cygcheck.out>
>
> <bash_test.tar.bz2>
>
>
>
>
>
>
>
> --
> Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
> Problem reports:       http://cygwin.com/problems.html
> Documentation:         http://cygwin.com/docs.html
> FAQ:                   http://cygwin.com/faq/


--
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
Problem reports:       http://cygwin.com/problems.html
Documentation:         http://cygwin.com/docs.html
FAQ:                   http://cygwin.com/faq/