Not quite (Cygwin build_alias bug hunt)

Bogdan Vacaliuc bvacaliuc at ngit.com
Mon Sep 20 06:17:01 UTC 2004


Scott,

Trapping this 'hang' is really challenging for me, even with the automated scripts.  I have a non-stripped version of cygwin1.dll
and bash.exe running, but no sign of hanging after the one that got away earlier.

Pierre was looking at a signal delivery problem, I confirmed it, and its been fixed in the 09/20/04 snapshot of cygwin1.dll.

http://cygwin.com/ml/cygwin/2004-09/msg00926.html
http://www.cygwin.com/snapshots/

I cannot say (given the lack of any clues for the 'hang' bug) whether this would affect our build hangs.  I'm running the new .dll
against the non-stripped bash.exe for now.  Anywone with build 'hang' issues could try it (its fairly easy to switch), and let the
list know.

If anyone *does* manage to catch it, please do the following:

cat /proc/<pid>/cmdline for all the processes in the process group (PGID) of the build shell.

Those running with a non-stripped cygwin1.dll or bash.exe (you know who you are) can also try:

gdb --pid=<pid>
(gdb) info threads
(gdb) bt
(gdb) thread <n-1>
(gdb) bt
...
(gdb) thread 1
(gdb) bt

To dump the backtrace of all the threads.


Thanks!

-bogdan


> -----Original Message-----
> From: Scott Newell
> Sent: Saturday, September 18, 2004 4:38 PM
> To: rtems-users at rtems.com
> Subject: RE: Not quite (Cygwin build_alias bug hunt)
> 
> 
> At 02:31 PM 9/18/2004 , Bogdan Vacaliuc wrote:
> >
> >So if you notice a test script 'hung' again (and if you have 
> the extra
> time), see if you can isolate the hung process and attach an
> >strace to it 'strace --pid=<pid>' or 'gdb --pid=<pid>'.  I 
> had the one
> hang this AM, but I botched it.  It hasn't happened to me
> >again all day...
> 
> I don't think this will help, but here's a hang after 216 iterations:
> 
> $ ps ax
>       PID    PPID    PGID     WINPID  TTY  UID    STIME COMMAND
>       900       1     900        900  con  500 14:47:52 /usr/bin/bash
>      1984       1    1984       1984  con  500 15:00:01 /usr/bin/bash
>      1160     900    1160        472  con  500 15:03:31 /usr/bin/sh
>      2548    1160    1160       3224  con  500 15:21:40 /usr/bin/bash
>      2944    2548    1160       2944  con  500 15:21:44 /usr/bin/bash
> I    2276    2548    1160       3392  con  500 15:21:44 /usr/bin/sort
>      3396    2944    1160       3396  con  500 15:21:44 /usr/bin/bash
>      3076    1984    3076       3228  con  500 15:23:49 /usr/bin/ps
>      
>      
> $ strace --pid=2276
> Attached to pid 2276 (windows pid 3392)
>     4       4 [unknown (0xC88)] sort 2276 _cygtls::remove: wait 0x0
>   302     306 [unknown (0xC88)] sort 2276 _cygtls::remove: removed
> 0x296F140 element 1     
> 
> 
> $ strace --pid=2944
> Attached to pid 2944 (windows pid 2944)
>     4       4 [unknown (0xB4C)] bash 2944 _cygtls::remove: wait 0x0
>   253     257 [unknown (0xB4C)] bash 2944 _cygtls::remove: 
> removed 0xF4F140
> element 1
>                                      
>                                      
> $ strace --pid=2548
> Attached to pid 2548 (windows pid 3224)
>     4       4 [unknown (0xB64)] bash 2548 _cygtls::remove: wait 0x0
>   232     236 [unknown (0xB64)] bash 2548 _cygtls::remove: removed
> 0x164F140 element 2                                     
> 
> $ strace --pid=1160
> Attached to pid 1160 (windows pid 472)
>     4       4 [unknown (0x854)] sh 1160 _cygtls::remove: wait 0x0
>   221     225 [unknown (0x854)] sh 1160 _cygtls::remove: 
> removed 0xAAF140
> element 1
> 
> $ strace --pid=900
> Attached to pid 900 (windows pid 900)
>     7       7 [unknown (0xA2C)] bash 900 _cygtls::remove: wait 0x0
>   252     259 [unknown (0xA2C)] bash 900 _cygtls::remove: 
> removed 0x162F140
> element 2
> 
> 
> --
> newell
> 
> 




More information about the users mailing list