Experimental Physics and
Industrial Control System

[email protected] (Jeff Hill) · Thu, 3 Sep 1998 13:07:49 -0600

Jeff

This is a known problem.

All of the following must be true prior to experiencing this problem:
o Running on a little endian architecture with signed characters (e.i. Intel
	and probably also VAX)
o Running a CA client that fetches a 2 byte data type such as DBR_SHORT
o The most significant bit of the data is set
o Running a beta release of EPICS R3.13 and perhaps earlier releases also

I have created a potential fix, but not committed it to CVS at this time. A patch will
be installed into the final release of 3.13.

Sorry, I have procrastinated on the bug report for this while finalizing the patch.

An easy workaround for this problem would be to fetch the data as a DBR_LONG.

Thanks for bringing this to my attention, and thanks also to Mark Rivers and Ken 
Evans for initially experiencing (and reporting) the problem.

Jeff

On Thursday, September 03, 1998 12:08 PM, Jeffrey McDonald [SMTP:[email protected]] wrote:
> Hi,
> 
> 	Using a slightly modified version of caExample (which comes with
> makeBaseApp), I download 200 parameters to a waveform record using a
> ca_array_put command.  
> 
> The record is defined as: 
> 
> record(waveform, "$(user):waveForm1")
> {
>         field(DESC, "waveform record")
>         field(NELM, "200")
>         field(DTYP, "Soft Channel")
>         field(FTVL, "USHORT")
> }
> 
> 
> from caExample.c, I do: 
> 	
>     SEVCHK(ca_array_put(DBR_SHORT,200,mychid,(void *)indata),"ca_array_put
> failure");
>     SEVCHK(ca_pend_io(5.0),"ca_pend_io failure"); 
>     SEVCHK(ca_array_get(DBR_SHORT,200,mychid,(void
> *)outdata),"ca_array_get failure");
>     SEVCHK(ca_pend_io(5.0),"ca_pend_io failure");
> 
> 	
> 	I set the data equal to the array index, e.g. outdata[50] =50.  I
> have defined the indata and outdata to be pointers to short.  
> 
> 	When I read-back the data,  on the sgi OS, I get back exactly
> what I expect.  When I read-back data on Linux (Red Hat 5.0), I get the
> correct answer for the first 128 elements, for elements following 128, I
> get:
> 
> 	outvalue = invalue | 0xff00
> 	e.g. for outdata[199]= 65479 = 0xffc7
> 	(for those not hex-speaking  0xc7 is 199)
> 
> 	I have checked to see that the correct values are in the waveform
> record, and they are (the ca_array_put command works correctly). However,
> when the ca_array_get is called, the values for the elements greater than  
> 128 are corrupted when they are returned on the Linux box -- but NOT on
> the sgi.  
> 
> 	Any ideas?  
> 
> 	Thanks,
> 
> 	Jeff
> 
> _____________________________________________________
> Jeffrey McDonald, Florida State   [email protected]
> FNAL MS 352			  phone   630-840-3983
> Batavia, IL 60510                 fax     630-840-8886
> 
> 

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System