Experimental Physics and
Industrial Control System

Andrew Johnson <[email protected]> · Wed, 9 Dec 2009 16:18:10 -0600

On Wednesday 09 December 2009 15:06:23 Ralph Lange wrote:
>
> I can see the unexpected string type link behavior possibly being
> intended to do something like an auto-save/restore on record C. This
> could be an interesting feature, if (and only if) it is independent from
> data type (obviously) and configurable (which requires a CA protocol
> change to allow the link spec to define target record processing -
> instead of the PP option of the target field now used for that purpose).

I subsequently discovered that the only reason that non-string output links do 
not replicate the string behavior and re-send their last value when the target 
IOC reboots is probably because of a typo made in 1996.  Both branches of a 
conditional contain the line
   plink->value.pv_link.pvlMask |= pvlOptOutString;
but arguably one of them should actually say
    plink->value.pv_link.pvlMask |= pvlOptOutNative;
which would make all output links re-do their puts on reconnect.  There is no 
other purpose for those two bits in the pvlMask.

Also, this behavior only happens when CA reports the native channel type as an 
ENUM and the value being put is a string, so it doesn't currently happen for 
all string outputs.

I do not plan to make the above change, and in fact I will probably comment 
out both lines, because in the future I think we should look at adding a flag 
(ROR = Resend On Reconnect?) to output links to allow users to turn this 
behavior on.  I can see it as being useful in some cases.  I'm not so sure 
about the idea of a put without causing processing though...

Having a related flag to control the reconnect behavior of CP input links 
might be worth considering.

- Andrew  (leaving for home through the snow-storm now)
-- 
The best FOSS code is written to be read by other humans -- Harald Welte

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System