RE: Output record with TSE=-2 and info(asyn:READBACK, "1") returns <undefined> time.

[Yann Mandza via Tech-talk <tech-talk at aps.anl.gov>]

Mark,

 

> I have looked at devAsynFloat64 and it never sets the record .time field for ao records, only for ai records.  So you have found a bug, and TSE=-2 does not currently work for output records.  I will fix that.  It is almost certainly also a problem with asynInt32, asynIUnt32Digital and asynOctet device support as well.

 

Oh well! then, it was worth reporting my issue to the community.

 

> Meanwhile you could put an ai RBV record with SCAN=I/O Intr which should get your hardware's timestamp for that same parameter

 

Yes, I indeed provide  I/O Intr Input records for those output records that are not using the ‘info(asyn:READBACK, "1")’ and the timestamp is correct there.

though I did use this format ‘pasynUser->timestamp = time’ in the poller thread. Gonna test with setTimeStamp and see.

 

Thanks again for your assistance. I’ll wait for when you would update Asyn.

 

Regards,

 

Yann

From: Mark Rivers <rivers at cars.uchicago.edu>
Sent: Saturday, May 11, 2024 8:06 PM
To: Yann Mandza <yann.mandza at ess.eu>; Torsten Bögershausen <tboegi at edom.se>; tech-talk at aps.anl.gov
Subject: Re: Output record with TSE=-2 and info(asyn:READBACK, "1") returns <undefined> time.

 

Hi Yann,

 

I have looked at devAsynFloat64 and it never sets the record .time field for ao records, only for ai records.  So you have found a bug, and TSE=-2 does not currently work for output records.  I will fix that.  It is almost certainly also a problem with asynInt32, asynIUnt32Digital and asynOctet device support as well.

 

Meanwhile you could put an ai RBV record with SCAN=I/O Intr which should get your hardware's timestamp for that same parameter.

 

Mark

 

---

From: Yann Mandza <yann.mandza at ess.eu>
Sent: Saturday, May 11, 2024 12:46 PM
To: Mark Rivers <rivers at cars.uchicago.edu>; Torsten Bögershausen <tboegi at edom.se>; tech-talk at aps.anl.gov <tech-talk at aps.anl.gov>
Subject: RE: Output record with TSE=-2 and info(asyn:READBACK, "1") returns <undefined> time.

 

Mark,

 

I meant that a added some debug print lines to see the content of that time variable in devAsynFloat64 as:

static const char *functionName="processCallbackOutput";

printf( "\033[0;33m%s : ( %s ) : time: '%u'\n\033[0m", epicsThreadGetNameSelf( ), __FUNCTION__, pPvt->pasynUser->timestamp.nsec );

 

and inside my driver writeFloat64 as below:

epicsTimeGetCurrent( &timeStamp  );

setTimeStamp( &timeStamp );

printf( "\033[0;33m%s : ( %s ) : time: '%u'\n\033[0m", epicsThreadGetNameSelf( ), __FUNCTION__, timeStamp.nsec );

 

just mentioned it so that the logs I shared would make sense. Not consequential though.

 

I also tried this method.

asynRegisterTimeStampSource("ETHHAL ", "myTimeStampSource") after registering a timestamp method and adding these lines

to my write method.

 

updateTimeStamp (&timeStamp);

setTimeStamp( &timeStamp );

 

here are some logs.

ETHHAL : ( processCallbackOutput ) : time: '0'

ETHHAL : ( myTimeStampSource ) : time: '813505109'

ETHHAL : ( writeFloat64 ) : time: '813505109'

2024/05/11 19:41:12.814 ETHHAL:writeFloat64: function=0, value=4.000000

2024/05/11 19:41:12.814 DRVASYN:MOD0-VoltageRamp devAsynFloat64::processCallbackOutput process value 4.000000

 

Still undefined time.

 

Regards,

 

Yann

 

From: Mark Rivers <rivers at cars.uchicago.edu>
Sent: Saturday, May 11, 2024 7:33 PM
To: Yann Mandza <yann.mandza at ess.eu>; Torsten Bögershausen <tboegi at edom.se>; tech-talk at aps.anl.gov
Subject: Re: Output record with TSE=-2 and info(asyn:READBACK, "1") returns <undefined> time.

 

·                // Added a print line in devAsynFloat64 processCallbackOutput and writeFloat64 for some info on the time variable.

 

Please show those lines of code, since I don't know what you mean by the "time variable".

---

From: Yann Mandza <yann.mandza at ess.eu>
Sent: Saturday, May 11, 2024 12:22 PM
To: Mark Rivers <rivers at cars.uchicago.edu>; Torsten Bögershausen <tboegi at edom.se>; tech-talk at aps.anl.gov <tech-talk at aps.anl.gov>
Subject: RE: Output record with TSE=-2 and info(asyn:READBACK, "1") returns <undefined> time.

 

Hey Mark,

> What do you observe with the code you showed in the last message?

Here are some logs,

 

caput DRVASYN:MOD0-VoltageRamp 4

Old : DRVASYN:MOD0-VoltageRamp 0

New : DRVASYN:MOD0-VoltageRamp 4

// Added a print line in devAsynFloat64 processCallbackOutput and writeFloat64 for some info on the time variable.

ETHHAL : ( processCallbackOutput ) : time: '0'

ETHHAL : ( writeFloat64 ) : time: '169054678'

2024/05/11 19:15:20.169 ETHHAL:writeFloat64: function=0, value=4.000000

2024/05/11 19:15:20.169 DRVASYN:MOD0-VoltageRamp devAsynFloat64::processCallbackOutput process value 4.000000

 

caget -a DRVASYN:MOD0-VoltageRamp

DRVASYN:MOD0-VoltageRamp <undefined> 4

 

Regards,

 

Yann

 

From: Mark Rivers <rivers at cars.uchicago.edu>
Sent: Saturday, May 11, 2024 7:13 PM
To: Yann Mandza <yann.mandza at ess.eu>; Torsten Bögershausen <tboegi at edom.se>; tech-talk at aps.anl.gov
Subject: Re: Output record with TSE=-2 and info(asyn:READBACK, "1") returns <undefined> time.

 

What do you observe with the code you showed in the last message?

 

Mark

 

 

---

From: Yann Mandza <yann.mandza at ess.eu>
Sent: Saturday, May 11, 2024 11:50 AM
To: Torsten Bögershausen <tboegi at edom.se>; tech-talk at aps.anl.gov <tech-talk at aps.anl.gov>; Mark Rivers <rivers at cars.uchicago.edu>
Subject: RE: Output record with TSE=-2 and info(asyn:READBACK, "1") returns <undefined> time.

 

Hey Torsten, Mark pitched in, but I forgot to add you in my response.

> Do you call setTimeStamp() ?

 

Here the implementation for my write method. Testing for normal async with caput first.

 

asynStatus drvAsynIsegHalService::writeFloat64( asynUser *pasynUser, epicsFloat64 value )

{

 

  static const char *functionName = "writeFloat64";

 

  const char *propertyName;

  epicsTimeStamp timeStamp;

  char sVal[WRITE_BUF_LEN];

  int function = pasynUser->reason;

  asynStatus status = asynSuccess;

 

  epicsSnprintf( sVal, WRITE_BUF_LEN, "%f", value );

/*call driver methods and return asynError if write operation fails*/

       

// Get current IOC time.

epicsTimeGetCurrent( &timeStamp  );

       

// tried this without success.

setTimeStamp(&timeStamp);

 

  // update value of parameter

  status = ( asynStatus ) setDoubleParam( function, value );

  status = ( asynStatus ) callParamCallbacks( );

  if( status )

    epicsSnprintf( pasynUser->errorMessage, pasynUser->errorMessageSize,

                   "%s:%s: status=%d, function=%d, value=%f",

                   session_, functionName, status, function, value );

  else

    asynPrint( pasynUser, ASYN_TRACEIO_DEVICE,

               "%s:%s: function=%d, value=%f\n",

              session_, functionName, function, value );

  return status;

}

 

Regards,

 

Yann

-----Original Message-----
From: Torsten Bögershausen <tboegi at edom.se>
Sent: Saturday, May 11, 2024 4:57 PM
To: Yann Mandza <yann.mandza at ess.eu>; tech-talk at aps.anl.gov
Subject: Re: Output record with TSE=-2 and info(asyn:READBACK, "1") returns <undefined> time.

 

 

 

On 2024-05-11 13:39, Yann Mandza via Tech-talk wrote:

> Dear all,

> 

> I am working onan Asyn Port driver using with asynR4-42 on a Linux

> machine. Am trying to add timing support for Output record using

> TSE=-2 and

> 

> and info(asyn:READBACK, "1"). When I caget an output PV with TSE=-2

> after caput to it the time obtained is <undefined> though the value is

> well written to the device.

> 

> A camonitor on a PV with TSE=-2 and info(asyn:READBACK, "1") show an

> <undefined> time too though the read value is correct. I would like to

> associate the device timestamp to the record for output readback, and

> the current time return by epicsTimeGetCurrent to the record after a

> caput. In my poller thread this is how I set the time stamp:

> 

> epicsUInt32 seconds = 0;

> 

> epicsUInt32 microsecs = 0;

> 

> epicsTimeStamp time;

> 

> sscanf( timeStampLastChanged/*hardware time string*, "%u.%u",

> &seconds, &microsecs ) != 2 )

> 

> time.secPastEpoch = seconds - POSIX_TIME_AT_EPICS_EPOCH;

> 

> time.nsec = microsecs * 100000;

> 

> pasynUser->timestamp = time;

 

Do you call setTimeStamp() ?

 

<https://github.com/EuropeanSpallationSource/m-epics-ethercatmc/blob/master/ethercatmcApp/src/ethercatmcIndexer.cpp#L1798>