Experimental Physics and
Industrial Control System

Benjamin Franksen <[email protected]> · Wed, 13 Dec 2000 21:57:00 +0100

Jens Eden wrote:
> 
> I am using the CA-Server (EPICS R3.13.3) on Linux and Windows and
> encountered the following problem.
> Reading an array with caget works fine, but trying to read the same
> array via a CDEV/CA-client results in an internal error on the server
> (you can easily verify this with the simple server in the
> src/cas/example/simple-directory).
> 
> ... internal error.
> 
> Maybe a solution could be to differentiate between array and scalar
> dbr_ctrl-prototypes, when registering them in the
> gddApplicationTypeTable?
> My first approach was, to unmark the Managed/Flat-flags of the
> gddContainer, to be able to change bounds and type of the value-gdd. It
> works, but it is a performance impact and I am not really happy with it.
> 
> Has anyone else encountered the same problem, or any suggestions?

Yes I have encountered the same problem and it has driven me mad. It
concerns all the dbr_gr_xxx and dbr_ctrl_xxx requests.

You can't register the compound array requests as different from the
scalar ones, because then they cannot share the same application type.

You could try to reset only the managed flag. The gdd containers *must*
be flat anyway, wether or not the flag is set, else dbMapper breaks
(because it simply assumes that containers are always flat). Maybe it is
possible to reset (and check inside the cas) only the managed flag for
the value gdd, not for the whole one?

My best suggestion is: wait for the new data object that will replace
the gdd. Until then: try to live with the bad performance.

Ben

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System