Experimental Physics and
Industrial Control System

[email protected] (Deb Kerstiens) · Fri, 13 May 94 17:05:15 MDT

Paul writes 
>I've been messing around with the little utilities from the 
><epics>/share/src/toolsComm/ca directory and I've noticed that when I:
>
>	caget <recordname>
>
>If the recordname exists or does not exist the caget process will not die.
>The shell returns the prompt to me, but if I look at the process table, there
>caget sits.  
>
>So I wrote my own caget which does the following:
>
>	ca_task_initialize()
>	ca_search()
>	ca_pend_io()
>	ca_get()
>	ca_clear_channel()
>	ca_task_exit()
>
>and I see the same behavior. 

First off, I'm assuming your running on a workstation not an IOC.  
Disregard the rest if this assumption is incorrect.  One could get 
such results if there is not a CArepeater task running on the
workstation when you run a CA client.  In this case the 
CArepeater task is initiated from your process and is named 
whatever your process is named, i.e. caget, dm, etc.  While you're
running the process you can actually see two processes with the 
same name.  When you exit, you'll see one process left.  It is
really the CArepeater.  To eliminate this problem, you can start this 
task before you run CA clients--we add a line to run startCArepeater
when we boot our workstations so we don't have to deal with it. 
By doing this, CArepeater runs as a process with its own name.  I
think Jeff is planning in a future release to have the CArepeater 
get its own name as the process name even when initiated from another
process so this won't happen.

	Deb 

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System