Experimental Physics and
Industrial Control System

"Kasemir, Kay" <[email protected]> · Tue, 03 Feb 2009 10:20:06 -0500

Hi:

I ran into a problem with not getting the expected CA monitors
from a calc record.
It's a simple calc record meant to count whenever it's processed.
To reset the counter, we write 0 to its VAL field:

record (calc, "XYZ_Cnt")
{
    field (INPA, "XYZ_Cnt")
    field (CALC, "A+1")
}

Many times, that works OK. But at least with R3.14.8.2,
it omits the monitor for the initial count from 0 to one
after a reset.

Example 'camonitor' output:
# After IOC start, the record is invalid
XYZ_Cnt                        <undefined> 0 UDF INVALID

# After "caput XYZ_Cnt.PROC 1", it counts up to 1
XYZ_Cnt                        2009-02-03 09:33:01.570935 1

# With "caput XYZ_Cnt.VAL 0" it resets to 0 and sends a monitor for that
XYZ_Cnt                        2009-02-03 09:33:01.570935 0

# BUT: When checking dbpr on the IOC, I finds MLST: 1,
# so the record thinks that the monitor sent out was 1,
# even though I just got the value 0?!
# When I issue another "caput XYZ_Cnt.PROC 1", dbpr will
# show VAL:1, so the record counts OK, but it fails
# to send a monitor, which makes sense with MLST: 1.

# After another "caput XYZ_Cnt.PROC 1", it counts again,
# and this time also sends a monitor
XYZ_Cnt                        2009-02-03 09:33:23.019859 2

In general, simply performing a "caput XYZ_Cnt.VAL 0"
seems to leave the calc's MLST where it was.
I assume that's because Channel Access & the database code
directly update the VAL field without triggering the CALC's
process routine.

The "fix" for me is to set MDEL=-1, always getting monitors
even though the record's idea of the last value, MLST,
is wrong.

Thanks,
-Kay

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System