Experimental Physics and
Industrial Control System

Stephen Lewis <[email protected]> · Tue, 04 Aug 2009 15:25:20 -0700

I consider forwarding a ca_put from a client into a ca_put-notify a

design flaw: behavior would be very different between a local IOC on

the client LAN and a distant IOC behind a gateway.

Fundamentally EPICS is asynchronous and all client applications should

have been and should be written strictly to this assumption when using

ca_put. As I recall, ca_put_notify was added to allow sophisticated

client to build up a type of 'transaction' capability.

In the case of the motor record (the case that triggered all this

investigation), when the monitor on the DMOV field goes 0->1 then you

can trust all the other field values. Everyone up to now must have

known this, otherwise all the instances where there is no gateway

would have given poor results.

On 4 Aug, 2009, at 3:11 PM, Jeff Hill wrote:

 1. PCAS was unable (before the patch) to handle subscriptions

during

the time that ca_put_callback was being processed.