Bug 3257 – globus-start-container-detached needs -help

Bug 3257 - globus-start-container-detached needs -help

Summary:

globus-start-container-detached needs -help

Status:	RESOLVED FIXED

Product:	Java WS Core
Component:	globus_wsrf_core
Version:	4.0.0
Platform:	PC Linux

Importance:	P3 normal
Target Milestone:	---
Assigned To:	Mats Rynge

URL:
Keywords:

Depends on:
Blocks:
	Show dependency tree / graph

Reported:	2005-05-02 17:37 by Charles Bacon
Modified:	2005-05-09 10:31 (History)

Attachments
Add an attachment (proposed patch, testcase, etc.)

Note

You need to log in before you can comment on or make changes to this bug.

Description From Charles Bacon 2005-05-02 17:37:11

[bacon@mayed globus-4.0.0]$ echo $GLOBUS_LOCATION

[bacon@mayed globus-4.0.0]$ sbin/globus-start-container-detached 
Segmentation fault
[bacon@mayed globus-4.0.0]$ export GLOBUS_LOCATION=`pwd`
[bacon@mayed globus-4.0.0]$ sbin/globus-start-container-detached 
Globus container started. PID: 10438

Also, it does not respond to -help:
[bacon@mayed globus-4.0.0]$ sbin/globus-start-container-detached -help
Globus container started. PID: 10458

------- Comment #1 From Mats Rynge 2005-05-02 19:23:18 -------

The segfault problem has been fixed in trunk and 4_0_branch.

I'm not sure how to deal with -help, as the arguments are supposed to be passed
to globus-start-container. One possibility would be to intercept only -help, and
show a message saying to run globus-start-container -help for full help. I'll
discuss this with the ws core developers.

------- Comment #2 From Jarek Gawor 2005-05-02 21:57:39 -------

Yes, I had the same suggestion just to check -help argument and do special 
handing in that case.

Also, there is another problem. If GLOBUS_LOCATION points to a non existient 
directory then no error is reported (on the console or the log file). In 
general that program needs better error checking.

------- Comment #3 From Mats Rynge 2005-05-02 23:13:19 -------

Yeah, I wrote that to get the testbed going, and didn't spend much time on it.

Regarding the error checking, I think that it would be good to fork, exec the
child, and in the parent wait a few seconds, and check if there is a process
with the child PID still running. If not, spit out an warning. Not only would it
take care of a faulty GLOBUS_LOCATION, but also other problems, such as the
container stopping directly from for example bad arguments or unable to bind to
a port.

------- Comment #4 From Mats Rynge 2005-05-09 10:31:20 -------

Fix commited to globus_4_0_branch and trunk:

 - Both globus-start-container-detached and
   globus-stop-container-detached now accepts -help

 - After fork/exec, globus-start-container-detached waits for 10
   seconds to make sure the container actually started

 - globus-start-container-detached will not start a container if
   pidfile exists. globus-stop-container-detached removes pidfile
   when the container is shut down.