Ron, Tim,
Can you guys add anything to this conversation about the motor record
alarm behaviour?
- Andrew
-------- Forwarded Message --------
Subject: Re: CSS shows "Disconnected" for all PV's with Base 3.15.2 (?)
Date: Fri, 9 Oct 2015 10:43:45 -0400
From: Michael Davidsaver <[email protected]>
To: Torsten Bögershausen <[email protected]>, Johnson,
Andrew N. <[email protected]>
CC: EPICS Core-Talk <[email protected]>, Ralph Lange
<[email protected]>
On 10/09/2015 02:37 AM, Torsten Bögershausen wrote:
Hej,
The IOC does not use autosave yet, and the fact that the variables
are UNDEFINED is probably a distraction from the real problem.
I can move the motor, and the e.g. IOC:m1.RBV goes away from
UNDEFINED; but CSS does never take away the pink Disconnected.
I'm not sure if this has anything to do with the motor record.
Matt's comment makes me think that it might. What he describes sounds
like a missing subscription update (eg. missing call to
db_post_event() in support code).
Of course it could also be that the client logic is ignoring an update
for some reason (non-monotonic timestamp?).
In addition to an OPI screen, can you observe the PV(s) in question
with Probe as well?
I will put that aside that for a moment, in case somebody is
interested:
https://github.com/tboegi/cashark/commit/b367eeb6ab3861379ab558f512175ed13e9d56af
When comparing these two captures I don't see any difference which
would explain what you see. The initial subscription updates are sent
in both cases, and show alarm severity==0.
Thanks for help. /Torsten
On 08/10/15 07:02, Johnson, Andrew N. wrote:
Hi Torsten,
I suspect you would be wasting your time using Wireshark to
investigate this, so far I don't think you've shown us any
evidence that can't be explained as an issue with or a
misunderstanding of what's going on inside the IOC itself, which
has little or nothing to do with the CA protocol and what goes
over the network. The command-line tools caget, caput and
camonitor and the iocsh commands dbpr, dbgf and dbpf are
probably all you need to investigate what's happening inside the
IOC.
Database records start off with an <undefined> or zero time
stamp until they have been processed at least once, so you need
to understand the database design and how the records are
supposed to get processed. The motor record is probably the most
complex type we have, so having an understanding of how it
behaves and when it sets its time stamp is probably very
important. If you should exhaust the subject of the record
support, is your 3.14.12 IOC using autosave perhaps? If so it
might be that the 3.15 version of that isn't behaving the same.
- Andrew