Experimental Physics and
Industrial Control System

"J. Lewis Muir via Tech-talk" <[email protected]> · Thu, 9 Jan 2020 16:01:05 -0600

On 01/09, Mark Rivers wrote:
> I am trying to understand your motivation in pushing for ABI compatibility.  Can you please explain clearly how you will use this, and how you think it will save you time and effort?
> 
> As far as I know you are responsible for the control system on a single beamline with a single experimental station.  How much time do you estimate you will save by being able to upgrade EPICS with binary modules, rather than just building base and synApps from source?
> 
> I run 4 beamlines with 5 experimental stations, and the build and deploy process does not take much of my time.

Whenever I upgrade, I spend a fair amount of time downloading all the
distribution files, patching, and getting everything to build correctly.
I wish to have binary packages that I could build once and then install
on Linux and macOS machines.  I was thinking of using pkgsrc so that
I could use one package management system to build and install binary
packages for a number of platforms including Linux and macOS.  Then, for
example, it would be as simple as "pkgin install epics" to install EPICS
on a machine.

The motivation for wishing for ABI compatibility is mainly to make the
package management possible so that binary packages for EPICS and EPICS
modules could be installed and upgraded independently without breaking
anything that depends on them and also to allow having different sets
of packages installed in parallel (e.g., EPICS 3.15 and 7.0 in parallel
along with the EPICS modules built for them).

A secondary motivation for API and ABI compatibility is to make the
upgrade process easier.  If EPICS and all of the EPICS modules cared
about API and ABI compatibility, I suspect they would not break it as
often, and I think this would result in an easier time upgrading EPICS
and EPICS modules in that I wouldn't have to change as much source code
for API changes, and I would have a better chance of avoiding so-called
"dependency hell."

Lewis

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System