Experimental Physics and
Industrial Control System

"Pearson, Matthew R." <[email protected]> · Fri, 11 Oct 2013 17:17:10 -0400

Could it be that the device is actually sending back 32 bytes of data? If the input buffer size is <=32 bytes, then I think the writeRead will succeed as soon as the buffer is full. But if you define a larger buffer it will wait until the timeout before returning with the timeout error code.

Did you define end-of-stream characters using setInputEos and setOutputEos?

Cheers,
Matt

On Oct 11, 2013, at 4:58 PM, Mark Rivers <[email protected]> wrote:

> Something does not seem right here, I don't understand why it is failing if nbt > 32.
> 
> Unfortunately I have no experience with VXI11, and I don't have any VXI11 devices to test with.  Hopefully someone else on tech-talk can help you with this.
> 
> When the writeRead fails, what is the error return?
> 
> What do you see if you set asynTraceMask=255 in the underlying VXI11 port (L0x)?
> 
> Mark
> 
> 
> -----Original Message-----
> From: Zenon Szalata [mailto:[email protected]] 
> Sent: Friday, October 11, 2013 3:34 PM
> To: Mark Rivers; [email protected]
> Subject: VXI11 - GPIB
> 
> Hi Mark,
> I have written a device driver for a GPIB device that is using VXI11 
> protocol.
> The device driver is sub-classed from the asynPortDriver class.
> In the st.cmd file I configure the VXI11 object as follows:
> vxi11Configure( "L0x","$(ETHER)",0,0.0,"gpib0",0,0)
> 
> Then in the constructor I connect:
> 
> stat=pasynOctetSyncIO->connect( _ioPort,_addr,&_apvt,0);
> 
> finally in IO routine I call:
> 
> stat=pasynOctetSyncIO->writeRead( 
> _apvt,pcs,ncb,_buf,nbt,1,(size_t*)nio,&nbi,&eom);
> 
> _buf is defined as char _buf[2048]
> 
> I have discovered by trial and error that the writeRead succeeds for nbt 
> <= 32 and fails otherwise.
> I do not understand this.  I think that nbt is supposed to be the size 
> of the read buffer (in my case _buf) simply telling the routine, which 
> ultimately does the device IO, to put up to nbt bytes of data into the 
> read buffer.  Now, it happens that in my test case the device returns 
> only one byte of data, that is the status byte. So, why should it fail 
> when I give the writeRead a ridiculously large buffer to read into?
> 
> Then what if I poke the device and it chooses to reply by sending 1024 
> bytes of data?  Do I need to worry about that and issue the writeRead 
> number of times with 32 byte long read buffer until all bytes of data 
> are read in?  Since I don't know before hand how much data will be 
> returned, it is not obvious to me how this could be done efficiently.
> Thanks Mark,
> Zen
> 

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System