Experimental Physics and
Industrial Control System

Andrew Johnson <[email protected]> · Wed, 18 Dec 2013 11:18:16 -0600

Hi Michael,

On 12/18/2013 01:14 AM, [email protected] wrote:
> BTW, does CA support zero length arrays, I mean with .NORD == 0?
> (Just an idle question, really.  I imagine not, to be honest.)

The IOC does handle zero-length arrays OK, they don't cause it to crash
or anything like that, but it generally expects there to be at least one
element. Several dbAccess APIs to get or put field data take both a
request length and a set of options flags for requesting metadata such
as the units string, and you pass in length=0 to say that you're only
asking for metadata, so they don't return any data in that case. Other
internal code converts a zero length in to a one, but that behaviour may
not be visible outside the IOC.

Zero-length arrays can cause strange behaviour with some CA client
programs though. A waveform record initializes with NORD=0, and if you
do a caget of such a waveform it currently prints a garbage value for
the first array element:

tux% caget anjHost:wf
anjHost:wf 8 2.50321e-308 0 0 0 0 0 0 0

The caget -c option which uses ca_get_callback does see that this is a
zero-length array, but if you're parsing the result you'd better write
it to expect an empty value string because it doesn't print an element
count when it's zero:

tux% caget -c anjHost:wf
anjHost:wf

These are known bugs: https://bugs.launchpad.net/epics-base/+bug/1242919

HTH,

- Andrew
-- 
Advertising may be described as the science of arresting the human
intelligence long enough to get money from it. -- Stephen Leacock

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System