Experimental Physics and
Industrial Control System

"Mooney, Tim M." <[email protected]> · Thu, 23 Jan 2014 15:40:34 +0000

This looks very like a problem I ran into back in 2010.  Tech-talk messages from 11/1/2010 through 11/10/2010, with the subject "monitors received out of order", describe attempts by me and others to figure out what was going on.  In the end, I was able to get my client to receive monitors from the critical PVs in order by arranging for them to be in the same event queue block (email from me to tech-talk on 11/9/2010).

I only saw this problem on Linux, though I tried very hard to reproduce it on vxWorks.  I saw something similar on solaris (but not Linux or vxWorks) with an early version of base 3.15.  I'll guess this happens only for monitors posted by different threads.

Tim Mooney
________________________________________
From: [email protected] [[email protected]] on behalf of Benjamin Franksen [[email protected]]
Sent: Thursday, January 23, 2014 5:11 AM
To: [email protected]
Subject: Re: SNL monitor and pvPutComplete

On Wednesday, January 22, 2014 16:01:49 Pearson, Matthew R. wrote:
> Now I'm wondering how the callback reply can arrive at the client
> before the monitor.

I can't tell how this is possible, it seems strange, assuming that the
database is always behaving in the way you see when enabling TPRO. Maybe
some artefact of thread scheduling on the IOC (just guessing).

Anyway I agree that using an event flag for the status pv is the
reliable way to get this to work. An explicit get is also a valid
solution (and yes, doing pvGet on a monitored variable is allowed and
sometimes makes sense).

Cheers
Ben
--
"Make it so they have to reboot after every typo." ― Scott Adams

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System