informations about a core-dumping situation

[Helmut Dipper <helmut dot dipper at sicad dot de>]

Informations about a core-dumping situation with bash and make
==============================================================

I use following bash-version:

bash -version delivers:
GNU bash, version 2.05.0(6)-release (i686-pc-cygwin)
Copyright 2000 Free Software Foundation, Inc.

on Windows-NT 4.0 SP5.

My bash.exe is taken from bash-2.05-6.tar.gz.
sh.exe is a copy of bash.exe.
Make is from make-3.79.1-4.tar.bz2.

During a make in a complicated makefile hierarchy sometimes, but
at varying points in the hierarchy, I get messages like the
following (taken from a part of the make protocol, for a greater
part of the protocol see attachments):

make[3]: Entering directory
`//M/UTX_32A00_010515.1020/utx_v01/src/utb/SICAD/prc'
 >ut_base_syn_create.prc generated< . >ut_base_syn_drop.prc generated< .
>ut_dd_syn_create.prc generated< . >ut_dd_syn_drop.prc generated< . >uth_node_alpha.prc copied< .......................................................................................................................................      0 [main] sh 354 open_stackdumpfile: Dumping stack trace to sh.exe.stackdump
      0 [main] sh 489 open_stackdumpfile: Dumping stack trace to
sh.exe.stackdump
.      0 [main] sh 430 open_stackdumpfile: Dumping stack trace to
sh.exe.stackdump
make[3]: *** [ute_branch_trace] Segmentation fault (core dumped)

Explanations to the current make process:
- 'ute_branch_trace' is a target to be produced from the dependency
  'ute_branch_trace.prc' by the suffix-rule '.prc:'
- the target is used as a 'time-stamp' for the dependency
- all the files *.prc are afterwards combined in a library
- some of the *.prc files are to be generated


In the directory following file is created:
sh.exe.stackdump
with following contents:

Exception: STATUS_ACCESS_VIOLATION at eip=00000000
eax=00000000 ebx=00000007 ecx=61095A9C edx=00000000 esi=610901EC edi=610901F4
ebp=0247F654 esp=0247F638 program=c:\bin\sh.exe
cs=001B ds=0023 es=0023 fs=003B gs=0000 ss=0023
Stack trace:
Frame     Function  Args
End of stack trace


The file tree is presented by a "view" onto one version of
versioned objects (directories and files). The name of the
Configuration Management System is ClearCase from
Rational Software Corporation:
cleartool -version delivers:
ClearCase version 3.2.1 (Thu Oct 1 08:51:50 1998)
clearcase patch p3.2.1.NT-12 (Fri May 7 14:09:34 1999)
clearcase patch p3.2.1.NT-13 (Mon Apr 26 15:54:36 1999)
clearcase patch p3.2.1.NT-14 (Mon Apr 26 15:56:23 1999)
clearcase patch p3.2.1.NT-15 (Mon Jun 7 14:45:07 1999)
@(#) MVFS version 3.2.1+ (Apr 10 1999 13:15:02)
cleartool version 3.2.1+ (Sat May 22 02:45:46  1999)

The ClearCase-vob-server ('vob': versioned object base) is a
UNIX-machine, where uname -a delivers:
SunOS phosphat 5.7 Generic_106541-17 sun4u sparc SUNW,Ultra-60

Both machines are connected via Hummingbird Maestro:
NFS Maestro Client for Windows NT Version 6.1.0.0


In almost all cases the problem occured during the night, when
the production process has been started in the evening. Duration
of the production process in this case (after a complete 'clean')
was from about 20:00 in the evening to about 10:00 in the morning.
Duration during the day is about 6 hours.
In the night the network may have lower performance due to all
the backup-processes for more than 5000 employees.

In all cases there are only a few (at most three) such
core-dumping situations. In the current case it was the only one.

All core-dumping situations occured in different steps of the
production process.


Because the problem occurs only sporadically I cannot provide
a test case but I hope that the informations will help You to fix
the problem. If there is any test I can execute for You in the
environment described please do not hesitate to contact me by email.


Best regards
Helmut Dipper


Attachments:
- makefile currently executed
  XPG_ECHO = echo -e
- part of protocol of production process

*********************************************************
* Helmut Dipper                  | SICAD Utilities GmbH *
*  phone: +49 89 636 41558       |                      *
*    fax: +49 89 636 49613       | Otto-Hahn-Ring 6     *
* e-mail: helmut.dipper@sicad.de | D-81730 Muenchen     *
*********************************************************

makefile

protocol

--
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
Bug reporting:         http://cygwin.com/bugs.html
Documentation:         http://cygwin.com/docs.html
FAQ:                   http://cygwin.com/faq/