Experimental Physics and
Industrial Control System

"J. Lewis Muir via Tech-talk" <[email protected]> · Tue, 14 Jan 2020 11:01:27 -0600

On 01/14, Kasemir, Kay via Tech-talk wrote:
> > One of the hassles with Windows is that you cannot delete or replace an executable or dll if any process is using it.
> > If an EPICS client or IOC is run on Windows it will launch a caRepeater process that runs in the background.
> > If you want to update your EPICS installation you must kill that caRepeater process before you can do so.
> 
> The Java CA client implementation has the same issue:
> Run any program that uses the CA client for the first time, and it launches a caRepeater Java process.
> Closing all clients leaves that process running.
> 
> For the Java CA client, there's a property
>    -DCA_DISABLE_REPEATER=true
> that will keep it from launching the caRepeater.
> Most members of the general public will never notice that they don't have a CA repeater.
> Worst that can happen is that after an IOC reboots, you need to close/reopen some display.
> The automatic re-connect supported by the caRepeater is essential in the control room, but for people in their office who start a tool, look at it, close it, the dangling caRepeater instance may be more bug than feature.
> Is it worthwhile to have the C implementation also look for a CA_DISABLE_REPEATER type setting?

Hi, Kay!

+1.  I'd love to have a way to disable the C implementation's automatic
starting of the caRepeater process.  The automatic starting also has the
unfortunate consequence that, if caRepeater is not intentionally started
at system boot time, whichever user happens to be the first to run a CA
program that automatically starts caRepeater will have that caRepeater
process running under their user ID.  That's undesirable from a security
standpoint.

I'm not a big fan of the Java property name, though.
CA_DISABLE_REPEATER makes it sound like it doesn't use the caRepeater,
even if it is running.  Perhaps a name like CA_REPEATER_AUTO_START would
be better.  But if it's too late to change it, then for consistency,
maybe CA_DISABLE_REPEATER as an environment variable is OK.

BTW, someone else ran into an issue related to the automatic starting of
caRepeater in October 2019:

  https://epics.anl.gov/tech-talk/2019/msg01548.php

Regards,

Lewis

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System