This is the mail archive of the cygwin mailing list for the Cygwin project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: once more unto the breech - still hanging, but I have a little more info.



On Jan 11, 2006, at 2:19 PM, Peter Rehley wrote:



On Jan 10, 2006, at 6:23 PM, Christopher Faylor wrote:



<snip>

<obligatory-often-ignored-request>
So, please try a snapshot and report problems. Please indicate whether
the problem is a regression from 1.5.18 or previous snapshots and please
respond to this message when reporting a problem. Don't start a new
thread. Please always provide cygcheck output even if you think you've
already done it recently. Please provide exact details to duplicate a
problem - a simple program indicating the problem is ideal.
</obligatory-often-ignored-request>

It's a rather difficult bug to reproduce, but I'm still seeing the hang up with the test_configure script. I seem to be getting it about once every 2000 iterations of the program. The thread where I mentioned this problem is http://cygwin.com/ml/cygwin/2005-11/ msg00000.html


The script can be restarted by using the process.exe program (http://www.beyondlogic.org/solutions/processutil/processutil.htm) using the restart option.

When I run the test_configure script, after about 2000 iterations it will hang. If I do a ps -ef I can see several running sh commands from the script. I pick the one that doesn't have any children and cat /proc/<pid>/cmdline and it show <defunct>. Then I cat the winpid and use the process program to restart. This causes things to start running again.

Occasionally I will see that there is a child process (i.e. sort, not another sh) If I try to cat the cmdline the cat process seems to hang; I can only kill cat with the task manager. However, I can still cat the winpid and when I do the process restart, things start running again. This behavior is fairly new. I have noted that it was the 12/22/2005 snapshot. But I didn't try the 12/21, 12/20 or 12/19 snapshots.


I tried the 20060206 snapshot, and managed to get an strace log when it hung and after I resumed it. I couldn't get gdb to attach to the hung process though so I couldn't get any gdb information.


Also in the snapshot I noticed new behavior. When I ran the test script and did a build (bash conf) in another window, the test script segfaulted. I have the stackdump of that sh.exe plus the addr2line output. In the 20060205 snapshot this didn't occur.

Peter

Attachment: cygcheck.out
Description: Binary data

Attachment: sh.exe.addr.out
Description: Binary data

Attachment: sh.exe.stackdump
Description: Binary data

Attachment: strace_hung.log
Description: Binary data

Attachment: strace_short.log
Description: Binary data




--
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/

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]