Experimental Physics and
Industrial Control System

Dirk Zimoch <[email protected]> · Wed, 10 Apr 2013 10:33:10 +0200

Hi Alan

It is the "flush". Before each out command, I flush the input in order

not to get any old unprocessed input in the queue. However, I found that

when I call flush(), "I/O Intr" records that might be interested in that

input don't get it. Thus I emulated the flush() by reading until the

input queue is empty (and feeding the input to any registered "I/O Intr"

record). I did not think about device that must not be read. I will

check that.

Which asyn port drive do you use? Something from asyn or something

self-written? In the second case, please send me your code.

Dirk

On 08.04.2013 18:40, Alan Greer wrote:

Thank you for the suggestion, I will give it a try.

Kind regards, Alan

On 04/08/2013 04:59 PM, Eric Norum wrote:

Perhaps this could be related to the patch:
StreamDevice-2-6-patch20121113
<http://epics.web.psi.ch/software/streamdevice/StreamDevice-2-6-patch20121113>

Try applying that to the 2-6 code and see if it fixes things.

On Apr 8, 2013, at 8:51 AM, Alan Greer <[email protected]
<mailto:[email protected]>> wrote:

Hi, I have been carrying out some additional tests.

I checked the flush operation in my driver and it does print "flush".
I then started looking at versions. I have discovered that it works
as expected if I use stream device version 2-5 instead of 2-6. In
this situation I see a flush followed by a write.

Kind regards, Alan

On 04/08/2013 04:35 PM, Mark Rivers wrote:

I can't think of any reason why it would be doing a read. asyn
device support does attempt a read at iocInit on many output
records, but not on stringout records.

Could this be from a "flush" operation? Your driver is printing the
message "device read"?

Mark

--
Eric Norum
[email protected] <mailto:[email protected]>

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System