Re: mbboDirect record initialization

[Michael Davidsaver <mdavidsaver@bnl.gov>]

I've come upon the problem which I think you are describing before.  I 
opted to fix it in device support.
http://epics.hg.sourceforge.net/hgweb/epics/mrfioc2/file/ecd5a49f66db/mrfCommon/src/devMbboDirectSoft.c#l58

On 10/25/2012 2:26 PM, Rod Nussbaumer wrote:
> Hi all.
>
> I'm using modbusAsyn with an mbboDirect record. At IOC init time, the
> record correctly reads back from the modbus server (a Moxa Digital IO
> box, but that doesn't seem relevant). This accomplishes initialization
> of the .VAL field of the mbboDirect record, as expected. However, I
> think that for this record type, the initialization should also be
> applied to the input fields of the record. The basis for this is that
> when the record subsequently processes due to a write/put to an input
> field, the record will then re-compose the word to write on its OUT
> link. Since the input bits will not have been initialized, they will go
> out to the respective hardware bits as zeros, overwriting what was there
> at startup.
>
> I've already had a brief dialog with Eric Norum and mark rivers on 
> this subject, and at least two possible courses have been mentioned. 
> One is to initialize the mbboDirect at the record level, and another 
> (which I've already tried quickly; seems to work) at the asyn level 
> (devAsynUInt32Digital.c).
> Is there anyone using this record type who thinks that changing the 
> record behavior to initialize the input bits 'Bx' from RVAL would 
> adversely affect their existing work? Any other records that may also 
> benefit from changes?
> Thanks.
>
> Rod Nussbaumer
> ISAC Controls, TRIUMF
> Vancouver, Canada