Experimental Physics and
Industrial Control System

"Mark Rivers" <[email protected]> · Fri, 29 Jun 2007 08:06:20 -0500

Heinrich,

> Why does a longout record call the drivers READ command at startup?
> Doesn't make sense 2 me.

asyn device support does 1 read at startup for output records.  It does this so that the EPICS output PVs correctly display the current state of the device, without having to change the actual state of the device (sometimes called bumpless reboot).  This is usually what is desired, so that, for example, you can reboot an EPICS IOC and have the ao record controlling a power supply be set to the actual current value of the power supply without changing it.

However, there are times when this is not the desired behavior, i.e. if the device is known to be in a "not initialized" state.  In that case it is the responsibility of your asyn driver to return an error status on that initial read.  If it returns an error status then the longout record will not be set to the value from the initial read.

We may want to think of a way to disable the initial read with some record field in the database, but I think Marty Kraimer thought of that and did not come up with a solution.

Mark

________________________________

From: [email protected] on behalf of Heinrich du Toit
Sent: Fri 6/29/2007 7:37 AM
To: TechTalk EPICS
Subject: asyn record support - output

Hi

I've created an asyn driver that provides a asynInt32 multi_device
interface. Actually the addr is simply different "registers" on the same
device.
Everything works fine - well mostly.
The registers are multi-directional in the driver.
ie. you can write them and you can read them. - Interrupts also works
Offcourse some registers are for reading and other for writing.

Now on one such register for writing I've put a longout record.
As this register will configure the device I want to put in a default
value in the .db file.
so I added the DOL field to the longout register. (this will set the
default VAL of the record and it does)
But when the system initializes the thing goes and READ this register
using the driver. This not only overwrites the default value. It also
gives some garbage as the device can be in unknown state.

I've made a workaround using DrvUser Interface and passing a startup
value that way - but this just doesn't seem like a good solution.
Why does a longout record call the drivers READ command at startup?
Doesn't make sense 2 me.

-Heinrich

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System