Re: When an external device does not provide termination on TCP ASCIIstream replies

[Dirk Zimoch <dirk.zimoch@psi.ch>]

Allison, Stephanie wrote:
> Hi Dirk,
>
> Thanks for your reply.
>
> > Have you tried to set ReadTimeout=0 ?
>
> Yes, setting ReadTimeout to 0 works but I haven't yet tested it on a busy network.  I fear that if I set ReadTimeout to 0 and the network is a little slow between the time I send the request and the time the reply is ready, the read will return an error.

This is not a problem because StreamDevice uses two different timeouts: 
ReplyTimeout and ReadTimeout. ReplyTimeout is the time between the request and 
the reply which may be long any traffic dependent. But ReadTimeout is the time 
between the individual pieces of a message. If the message is guaranteed to come 
in one piece, 0 is fine here.

> > Can you give an example of such a message? Any constant string at the end can be
> > used as a terminator, not just CR or LF.
>
> I send "MEAS:VOLT?;"  and I receive "00.009" with no terminator (see below for asyn output).  This is an example of a fixed-length reply where I can use MaxInput of 6 and the reply read will not timeout.  But not all replies are fixed length like this one.

This is really bad. Maybe not so bad on TCP connections where there are internal 
methods to check the integrity of a message. But on RS232 you would not notice 
when you get only half a message (maybe because of bad connection) and end up 
with "00.0". This is where the ReadTimeout would trigger.

Thus I think it's safe to use terminator and ReadTimeout>0 on RS232 and no 
terminator and ReadTimeout=0 on TCP. It's a pity that one has to change the 
protocol depending on the bus type.

> 2009/07/15 13:45:26.642 118-PSD5 flush
> 2009/07/15 13:45:26.642 134.79.35.216:8003 flush
> 2009/07/15 13:45:26.642 134.79.35.216:8003 write.
> 2009/07/15 13:45:26.642 134.79.35.216:8003 write 11
> MEAS:VOLT?;
> 2009/07/15 13:45:26.642 118-PSD5 addr -1 queueRequest priority 0 from lockHolder
> 2009/07/15 13:45:26.642 118-PSD5 schedule queueRequest timeout
> 2009/07/15 13:45:26.642 118-PSD5 callback
> 2009/07/15 13:45:26.642 134.79.35.216:8003 read.
> 2009/07/15 13:45:26.727 134.79.35.216:8003 read 6
> 00.009
> 2009/07/15 13:45:26.727 118-PSD5 read
> 00.009
> 2009/07/15 13:45:26.727 134.79.35.216:8003 read.
>
> Note the last "read" above is not necessary but is done anyway since there's no terminator.

This is harmless. With timeout 0 it consumes (almost) no time.


Best regards,
Dirk



> Thank you,
> Stephanie
>
> > Allison, Stephanie wrote:
> > > Hi Mark,
> > >
> > > Thanks for your reply.  I'm currently using stream version 2.2 and asyn version 4.9 on
> > epics base 3.14.8.2.  I'm unable to get stream to work without timeout when the
> > InTerminator is set to "" (meaning no terminator) and MaxInput=0.  Stream/asyn always
> > waits ReadTimeout even though it has received the reply (it just doesn't know it has the
> > *complete* reply).   If I set maxInput to a large number, then it also waits ReadTimeout
> > waiting for extra input it'll never receive.
> > > I would really like to use stream instead of writing my own support on top of asyn just for
> > this case.  Does anybody know how can I set up stream so that the InTerminator is "",
> > maxInput is 100, but there is no need to wait for anything more after something...anything >
> > 0 chars is received?   I still want to use a reasonable ReadTimeout for when things slow
> > down a bit on the network.  Maybe I can use devGpib for this case, though stream would be
> > better.
> > > Thank you,
> > > Stephanie
> > >
> > > > -----Original Message-----
> > > > From: Mark Rivers [mailto:rivers@cars.uchicago.edu]
> > > > Sent: Tuesday, July 14, 2009 10:16 AM
> > > > To: Allison, Stephanie; Epics Questions
> > > > Subject: RE: When an external device does not provide termination on TCP
> > ASCIIstream
> > > > replies
> > > >
> > > > Hi Stephanie,
> > > >
> > > > I think you should be fine.  Many TCP devices don't sent terminators,
> > > > unlike their serial counterparts.  In your st.cmd file you just don't
> > > > define an input or output EOS for that driver, and you don't put a
> > > > terminator in your protocol file.  The TCP driver does not loop waiting
> > > > for maxChars, it just reads the message.  TCP delivers "complete"
> > > > messages, so you will get the complete response, not just a partial
> > > > response.
> > > >
> > > > Mark
> > > >
> > > >
> > > > -----Original Message-----
> > > > From: tech-talk-bounces@aps.anl.gov
> > > > [mailto:tech-talk-bounces@aps.anl.gov] On Behalf Of Allison, Stephanie
> > > > Sent: Tuesday, July 14, 2009 11:58 AM
> > > > To: Epics Questions
> > > > Subject: When an external device does not provide termination on TCP
> > > > ASCIIstream replies
> > > >
> > > > Hello,
> > > >
> > > > We're testing a Genesys power supply (provided by Lambda-TDK) to which
> > > > the vendor has added a TCP interface (I believe specifically by request
> > > > of EPICS sites).  Burkhard Kolb at GSI provided me the protocol file for
> > > > stream/asyn support for this type of device (thanks!).  However, the
> > > > firmware has changed recently and one change is to remove the terminator
> > > > from the reply strings.  Some replies are fixed-length so I can change
> > > > the protocol file for these (to use maxInput).  However some replies are
> > > > not fixed-length and I need to find support for those - I do not want to
> > > > rely on a timeout to know that the reply is finished.  The vendor has
> > > > suggested that (1) I assume the reply will never get fragmented and just
> > > > allocate 100 bytes and take whatever I get and (2) block on the receive
> > > > (which is not an option for an IOC).
> > > >
> > > > Does anybody have experience with such devices (TCP ASCII SCPI commands
> > > > that have no reply terminators and no fixed-length reply strings) and
> > > > can recommend a solution?
> > > >
> > > > Thank you,
> > > > Stephanie Allison
> > > > SLAC National Accelerator Lab
> > --
> > Dr. Dirk Zimoch
> > Paul Scherrer Institut, WBGB/006
> > 5232 Villigen PSI, Switzerland
> > Phone +41 56 310 5182

--
Dr. Dirk Zimoch
Paul Scherrer Institut, WBGB/006
5232 Villigen PSI, Switzerland
Phone +41 56 310 5182