[omniORB] omniORB_2.7.1 changes for AIX 4.3 + xlc 3.6.4.0

[Gary D. Duzan]

In Message <372F129C.469F431A@lincoln.midcoast.com> ,
   "Mark D. Landry" <mdlandry@lincoln.midcoast.com> wrote:

=>b.keeping@ic.ac.uk wrote:
=>
=>> A pedantic point: surely there should be a
=>> mk/platforms/powerpc_aix_4.3.mk file for you to use on 4.3 rather than
=>> putting things into the 4.2 version that might not work (and a
=>> 4.3_egcs as well I guess!).
=>
=>Yes, certainly. I copied powerpc_aix_4.2.mk to powerpc_aix_4.3.mk before
=>making any changes. (Developers:  how should changes like this be
=>incorporated into the "official" omniORB source?)

   One note on this point. I believe the changes Mark is making are
fine (for a powerpc_aix_4.3.mk config) with one exception: the
makeC++SharedLib_r -> makeC++SharedLib change. AIX 4.2.1 requires the
_r version and 4.3.x requires the version without it. It should
probably be turned into a variable or a macro and defined in the config
file. As for the nsleep() change, AIX 4.2.1 should support it, but I
haven't tried it.

=>> I will send again this week with news on whether your proposed changes
=>> were needed and/or usable on 4.2. We can't change to 4.3 right now
=>> unfortunately!

   We are in the same boat. Our primary customer has a large installed
base of AIX 4.2.1 boxes and isn't likely to upgrade any time soon, even
though IBM is dropping support for it next year. There are also
significant enough differences (X11, Motif, pthreads, etc.) that would
make it something of an investment to shift development to 4.3 that
make it less than attractive from our perspective as well.

=>What problems are you having with 4.2 and what compiler are you using?

   Good question. If you are running gcc/egcs, you can pretty much
forget it. The only version that works is a patched EGCS 1.0, and
that's it.  Otherwise exception handling in the presense of threads is
non-functional, which pretty much rules out OmniORB. You'll want xlC
3.1.4.7+ or 3.6.4.x if you want to work with OmniORB on AIX. I really
wish egcs would be fixed, though; xlC is a really dated compiler, and
IBM isn't likely to support it too much longer.

					Gary Duzan
					GTE Laboratories