Experimental Physics and
Industrial Control System

Andrew Johnson <[email protected]> · Thu, 22 Apr 2010 14:08:12 -0500

Hi Dennis,

On Thursday 22 April 2010 11:40:11 Dennis Nicklaus wrote:
> We've encountered an odd behavior the last couple days that has us
> pretty confused.
> Can someone please explain this and/or suggest a fix?
> When I set a certain PV going through an EPICS gateway, the setting
> doesn't "work*" entirely.
> If set through a non-gateway-ed connection, everything works fine.

One significant difference between doing a caput directly and via the CA 
gateway is that the gateway converts your ca_put() into a ca_put_callback() to 
the IOC.  You can ask caput to do the same thing directly by giving it a -c 
option.

Why haven't you noticed this before now?  Because for most records the 
ca_put_callback() will complete almost immediately and you won't notice the 
difference.  However the busy record type is deliberately designed to not 
report completion until all of the underlying operations kicked off by your 
put have completed, and in your case that probably means until the camera's 
exposure has completed, which I'm guessing is taking longer than the default 1 
second time-out that caput() waits for the CA operation to complete.  You can 
increase that time-out using the -w option.

One more thing to say about the gateway: while a channel has a put_callback() 
operation outstanding (i.e. until the busy record goes back to state Done) you 
won't be able to write to that channel through the gateway at all.

I think the combination of these should explain the funny behaviors that 
you're seeing, which I believe are interactions between deliberate design 
choices.

One other question you might ask is why the gateway converts your ca_put() 
into a ca_put_callback()?  That's because the CAS library that implements the 
CA server doesn't tell the gateway code whether your caput request asked to do 
a ca_put() or ca_put_callback() operation (although internally the CAS does 
know that), so the gateway has to use the more conservative operation.

- Andrew
-- 
The best FOSS code is written to be read by other humans -- Harald Welte

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System