Lesson learned: LINR, LINEAR, SLOPE, R3.13, R3.14, breakpoint tables

[Kay-Uwe Kasemir <kasemirk@ornl.gov>]

Hi:

I spent some time yesterday debugging a previously unknown
application, and couldn't find any obvious hint about the issue
in the release notes or tech talk.

When you use breakpoint tables, you have to first
define the break point table.
Then, according to the R3.14.8 IOC Appl. Dev. Guide:
* Copy the menuConvert.dbd file from EPICS base
* Add definitions for new breakpoint tables to the end
...

That's what got us!

The symptom:
An application that worked fine under R3.13 suddenly
stopped honoring LINR=LINEAR, EGUL=..., EGUF=... conversions.
We spent hours trying to figure out why some driver/device support's
special() routine didn't get invoked,
and why ESLO always ended up at 1.

The reason:
That application's menuConvert.dbd, based on the R3.13 original,
started with "NO CONVERSION", "LINEAR", "typeKdegF", ...

But for R3.14, it has to be "NO CONVERSION", "SLOPE", "LINEAR", ...

So while "dbpr" always indicated that our records were
configured for "LINR=LINEAR", they were in fact set to
LINR==1, which the R3.14 EPICS base code considers to be SLOPE,
not LINEAR.

==> Beware about application-specific menuConvert files
when moving to R3.14.

Well, I guess you all knew this, but this way I can
find it the next time around by searching tech-talk...

-Kay