Re: issue with either camonitor or pcaspy in 3.15?

[Wang Xiaoqiang <xiaoqiang.wang@psi.ch>]

Hi Jamie,

The UDF (UnDeFined) alarm and INVALID severity is to indicate the values 
are default values, have not got updated by the driver. This is the same 
behavior as the EPICS database gets initialized on start-up.
Calling setParamStatus to clear the alarm status is all correct. It is 
then necessary to call driver.updatePVs() to inform ca clients with a 
DBE_ALARM event update. Otherwise this update event will be delayed 
until the next driver.updatePVs() call. And in PCASpy 0.6+, every caput 
triggers a call to driver.updatePVs(). That explains seeing all four 
updates together, in which channel A gives a value update and B,C,D give 
an alarm update.
To conclude, it is an incompatible change introduced in PCASpy 0.6. I 
created an issue to track this.
https://github.com/paulscherrerinstitute/pcaspy/issues/48

In any case, it is best to explicitly call driver.updatePVs() after 
calling setParam, setParamStatus, setParamEnums, setParamInfo. I will 
add this the docs.
Best
Xiaoqiang

On 11/19/2017 03:47 AM, Jameson Graef Rollins wrote:
> On Fri, Nov 03 2017, Wang Xiaoqiang <xiaoqiang.wang@psi.ch> wrote:
> > Here is my test on linux-x86_64 arch, using pcaspy 0.7.0 anaconda
> > package from https://anaconda.org/paulscherrerinstitute/pcaspy
> >
> > It does not reproduce the problem. Any chance that you could test using
> > the same setup?
> Hi, Xiaoqiang.  Thanks for confirming for me.  I finally figured out
> what the issue was.  In my test pcaspy IOC, on init I am attempting to
> clear the "UDF INVALID" flags with the following:
>
>      for k in pvdb:
>          driver.setParamStatus(k, pcaspy.Severity.NO_ALARM, pcaspy.Severity.NO_ALARM)
>
> In earlier pcaspy versions this was sufficient to clear the flags
> without causing any additional odd behavior.  I see now that with 0.6
> and later the issue I need to additionally call updatePVs() after the
> above:
>
>      driver.updatePVs()
>
> Is this the right thing to do?  Is there a cleaner way to clear the
> initial "UDF INVALID" flags?  And why is the "UDF INVALID" set
> initially anyway?
>
> Thanks so much for the help.
>
> jamie.