Experimental Physics and
Industrial Control System

Robert Soliday <[email protected]> · Mon, 06 Dec 2004 15:01:30 -0600

We noticed issues when we switched over too. Almost all of the problems
were due to the way we wrote code for 3.13.x which was more forgiving to
incorectly written code. Our main problem was seen in programs that ran
in an internal loop where there might be 60 seconds or so between
iterations. We found out that we could not simply have the code sleep
because the new version of Epics thought we may have lost connection.
Our changes included:

Replacing all ca_pend_io and ca_pend_event commands that had short
polling times with ca_poll.
Replacing sleep commands with ca_pend_event commands because Epics/Base
3.14.x has an inactivity timer that was causing disconnects from PVs
when the log interval time was too large. Note that you have to be
carefull because if you pass zero to ca_pend_event it takes that to mean
infinite. In this case we simply do a ca_poll.
Ensured that all codes included the ca_task_exit at the end because if a
ca_put was done directly prior to exiting there was a good chance that
the buffer would not be flushed upon exit if ca_task_exit was not in the
code.
--Bob
Geoff Savage wrote:

Hi,
We have just finished upgrading about 75% of the control system nodes to
EPICS 3.14.6/vxworks 5.5. The remaining nodes are running EPICS 3.13.4
with some patches/vxworks 5.3.1. We have noticed more channel access
errors with the later version of epics in the system than with an all
EPICS 3.13.4 system. Is this to be expected at this time? Will the
EPICS 3.14.7 release remove the errors? I understand that I don't have
any details but the situatuions encountered are not reporduceable and
everyone can get their work done. Over the past week while we have
installed the new software we have been fielding questions about error
messages appearing that have not been seen before.
Geoff

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System