Experimental Physics and
Industrial Control System

On Jun 15, 2021, at 12:28 PM, Michael Davidsaver via Core-talk <core-talk at aps.anl.gov> wrote:

On 6/15/21 10:25 AM, Mark Rivers wrote:

The report I received (indirectly) from the NSLS-II controls group was:


"The new EPICS version on the IOC server will not process
CA_PROTO_SEARCH request without first seeing a CA_PROTO_VERSION
request (CA protocol actually requires this). The IDL packets do not
contain the required CA_PROTO_VERSION (16 bytes)."

Does this make sense in terms of IDL being built with an old version of base?

Yes.

Does this mean that IDL is using libca circa Base 3.13?
Or does it have an independent implementation of the CA protocol?

(obviously I've never looked at IDL before)

Mark


________________________________
From: Core-talk <core-talk-bounces at aps.anl.gov> on behalf of Michael Davidsaver via Core-talk <core-talk at aps.anl.gov>
Sent: Tuesday, June 15, 2021 11:20 AM
To: Zimoch Dirk (PSI)
Cc: core-talk at aps.anl.gov
Subject: Re: CA incompatibilities?

On 6/15/21 8:39 AM, Zimoch Dirk (PSI) via Core-talk wrote:

One of my customers complained that his EPICS 3.13.2 client cannot connect any longer since the IOC has been upgraded
from 3.14.12 to 7.0.5.

I checked and 'caget' from the current 3.14 works with 'softIoc' from current 7.0.

(Don't tell me that  he should upgrade, please.)

I won't tell you that he _must_ upgrade, but I will certainly say that he _should_.

Searching the release notes, I found that support for clients < V4 (< EPICS R3.12.0-beta1) are not supported any longer
since 3.16.1. But the client is question reports V4.8.
Was there another CA compatibility drop since 3.16.1?

No.  Though, my change which dropped support for < 3.12 actually seems
to have dropped support for < 3.13.  I have no way to test anything
before 3.14, so I'm not certain of this.

https://github.com/epics-base/epics-base/pull/141#discussion_r602439829

While I don't anticipate any further deprecation of older CA protocol
versions, it should be noted that interoperability with older versions
is getting little to no testing.  It is inevitable that regressions
will creep in unless those with a continuing interest regularly test
the latest developments with these oldest versions.