PolyORB - building and applications

["Dr. Adrian Wrigley" <amtw@linuxchip.demon.co.uk.uk.uk>]

Hi!

I've written here before about my problems getting a robust system
running Annex E with GLADE.

So far, I've found GLADE under 3.15 to be problematic with
what appear to be race hazards with calls via Remote Access to Class-Wide
types (RACW).  Using a recent GLADE under GNAT GPL 2006 and
gcc 4.1.1, I've had the problem of breaking with nested
inter-partition calls.  I'm looking into work-arounds for this,
which *might* give a solution.  But it's worrying to me that
I can't get any affordable compiler to run even some simple
test cases with Ada+Annex E.  I have tried three different
machines and Linux OSs (Debian, FC5, Red Hat).  And several
compilers (GPL 2006, 4.1.0, 4.1.1).

So I'm left with the stark choices:

1) Drop Annex E code altogether - don't use distributed code :(
2) Use Annex E but drop GARLIC, replace with PolyORB DSA
3) Use distributed code, but not Annex E (eg use CORBA)
4) Debug the failing test-cases in GLADE
5) Put up with the failures.  Hope things'll get better.
6) Something else(?)

Choice (1) would involve programming alternate communication
mechanisms to get the components of the program talking to each other -
perhaps using pipes or files?  This is quite a lot of work :(

Choice (2) looks promising, but I have never managed to compile
PolyORB DSA.  What does this need?

Choice (3) also looks promising, but involves quite a lot of additional
work building the interfaces needed. It would be much less
elegant, I think - and might require learning the intricacies of CORBA(?)

Choice (4) would be very frustrating - I've tried already, and
I'm not sure I'd succeed.

Choice (5) isn't very satisfactory, but involves least effort.
Perhaps Annex E is considered obsolescent and will be dropped at
some point in the future?

Q:  Has anyone here used PolyORB DSA with GLADE?  Is it reliable?

Q:  Which versions of the compiler can build PolyORB DSA?  It seems
to need a newer compiler than I can get hold of!  Is good Annex E
support an important goal of PolyORB?

Q:  What is the simplest approach to migrating client-server code
from Annex E to an application personality well supported by PolyORB?
Which personality might be best?  Are there any real eaxamples
of applications using PolyORB out there?

I'd like to get on with the next stages of code development, but
I'm unable to proceed without a reliable base to work from :(

Thanks for more ideas!
--
Adrian