Re: EPICS latest stable version

[Ralph Lange via Tech-talk <tech-talk at aps.anl.gov>]

On Tue, 16 Feb 2021 at 14:55, Delahaye Olivier <olivier.delahaye at ganil.fr> wrote:

So If I understand well, we should better switch on 7.0 release. I would be quite confident about the stability since you mentionned half of the installations works now with 7.x version.

Careful. I wrote "... have EPICS 7 running somewhere". Which could be in production, could be on a handful of beamlines, could be in the lab. Some installations are in the process of upgrading. Some are planning to upgrade. A few have upgraded. But they are all running EPICS 7 IOCs, getting more and more confident.

1. What do you mean by tedious small changes ? We've got about 46 IOCs running on LINUX + 20 running on VxWorks. How long would you estimate for the upgrade ?

It's not so much the number of IOCs, more the number of EPICS support and IOC applications that comprise your EPICS software base.

There are some changes in the EPICS (make based) build system that may require patching a Makefile here and there.

The database parser (on the IOC) has become stricter and now fails to load some misconfigured databases that it was silently correcting before.

There were (backward compatible) changes in the function definitions of the Record Support and Device Support interfaces that will generate a load of warnings for all locally developed Device Support and Record Support modules.

Things like that. Not fatal, not especially hard to fix, but - depending on the number and quality of your existing sources and database files - possibly a large number of little things that need fixing.

The estimate for such an upgrade depends on the number of systems, their OS, the number of applications, the complexity of those applications, the number of people that would work on this, their level of EPICS knowledge, and - for existing installations sometimes the most limiting factor - the amount of testing time available. Could be anything, say, between "5 people, 6 months" and "1 person, forever".

Start with one subsystem, preferably one with Linux IOCs (less frustration). Take notes while you're migrating. Extrapolate.

2. You mentionned pvAcess protocol has been reviewed on 7.0 version. Our equipments rely on Channel Acess protocol. Is it the same one ?

As Timo wrote: pvAccess is new, powerful and interesting. Channel Access has not changed. Separate protocols on separate TCP/UDP ports. All Channel Access traffic is expected to work as before.

The tedious changes mentioned above are in the applications that build your IOCs (Linux and VxWorks) and databases. The Channel Access clients will hardly need any change. (The flexibility of the CA protocol is the glue and key to a step-by-step migration strategy.)

ITER's CODAC Core System has been using EPICS 7 for three years. Some systems in production are still on 3.15, others on 7.0. We are starting to make use of the pvAccess protocol.

Cheers,
~Ralph