Re: aao record behaviour

[Marty Kraimer <mrk@aps.anl.gov>]

Korhonen Timo wrote:
> 
> Eric,
> 
> thank you for your reply and sorry for my late response.
> 
> On Mon, 1 May 2000, Eric Bjorklund wrote:

> I thought so, too. But looking at the waveform record, the get_array_info
> returns NORD and cvt_dbaddr returns NELM (in a dbAddr struct).
> 
> The Application Developers guide says for get_array_info: "This routine
> returns the current number of elements and the offset of the first value
> of the array." For cvt_dbaddr: "This routine can change any combination of
> the dbAddr fields...". What should be returned in the dbAddr->no_elements
> is not specified, but to me it looks like it should be NELM.
> 
> What should the correct behaviour be?


What causes the confusion is that old database access, which is what CA uses,
only knows how to handle the maximum number of elements.

The correct behavior is that  cvt_dbaddr sets no_elements = NELM, ie the size of
the allocated array and get_array_info returns NORD, i.e. the current actual
number of elements.

> > As I understand things now, record support needs to be the one allocating the
> > buffer space since the device-support record_initialization routine only
> > gets called during pass=1.  This particular bug has not been a problem for
> > us (yet), but I can see where it could be.  I'd be interested in seeing your
> > solution Marty.

My suggestion is to do the following.

Define the DSET to have one additional method, call it init_record_pass0, and
then let the record support init_record call it during pass 0. The device
support must allocate storage for the array when init_record_pass0 is called.

 
Marty Kraimer