Experimental Physics and
Industrial Control System

Hi Ben,

    You guys are all way more qualified than I am to do this stuff! I understand Heinz is on vacation currently, so I hope he will join the conversation when he's back. I'm willing to help in any way I can.

Pete.

Am 07.01.19 um 21:28 schrieb Johnson, Andrew N. via Tech-talk:

The main complication with RTEMS 5.x which Heinz's not-yet-merged
implementation handles is that the OS APIs changed when RTEMS added SMP
support. We wanted to switch to the new Posix APIs for RTEMS 5.x, but
the EPICS build rules make that hard to do without breaking support for
RTEMS 4.x builds.

There is a hint of our final solution in the presentation below, where
on slide 14 you can see there are now 3 source subdirectories: RTEMS,
RTEMS-kernel and RTEMS-posix. I won't explain why, just saying that this
wasn't a trivial thing to do while still supporting older versions of
RTEMS and not completely breaking applications that are using
OS-specific source files in a src/os/RTEMS directory (e.g. iocStats).

Heinz's code can be seen in a couple of git repo's on Launchpad, but it
hasn't been updated to the latest Base releases. It should be possible
to apply the necessary changes to the 3.15 branch instead of to 7.0, but
there will need to be some work done to get anywhere from their current
state (which IIRC pre-dated our recombining libCom into the EPICS core
branch):
  https://code.launchpad.net/~epics-core/epics-base/+git/rtems5
  https://code.launchpad.net/~epics-core/epics-base/+git/rtems5libcom

If others want to help with this I wouldn't expect Heinz to say no.
Please use the core-talk mailing list for any technical questions and
discussions.

We are also very much interested in getting RTEMS5 support into base,
preferably in a clean and backward compatible way. We have previously
offered to help with merging Heinz' changes and the offer still stands.

Cheers
Ben