Marcel -
> Armen Yampolsky wrote:
>
>> Marcel,
>>
>> Thanks for your great summary! BTW, I just wanted to let you know,
>> we've switched to omniORB2/Sun IDL (JDK1.2) instead of Visibroker.
>
> We had at that time a lot of problems with omniORB, threads,
> exceptions, the gcc/egcs couldn't handle
> it very well. Are these problems gone?
So far we did a few highly simplified tests of mutex and exceptions,
with g++ on Linux and Sun's CC on Solaris, all worked as expected. They
have a threading model, it is probably newer than when you last used it,
but I have not had the opportunity to look into it deeply.
> Where are you running JDK 1.2 and SUN IDL, on Linux ? Does this work?
Oh yes, just fine! I add rt.jar to the *end* of my classpath, so that
only the CORBA files are used from it, and swing and everything else
come from the jdk1.1.7 version. I use IBM's jikes to compile, and
Blackdown's JDK1.1.7 java.
> Did you have to port your Java code much to run with Sun IDL?
Not too much. Some things are different, like ORB instantiation
arguments, the singleton argument-less ORB.init() does not return the
full-fledged ORB (don't use it to get the ORB), and there were also a
few things Visigenic does that do not comply to the standard, so that I
had to change. Not too much of an issue. The biggest issue was omniORB's
(well-done) connection management scheme, which Sun just does not deal
well with. One must turn the connection idle timeout *off* on the
omniORB side, until Sun fixes this bug. The trick is too turn it off for
the Naming Service daemon too, and that involved getting into the source
code. They have accepted my suggestion and will add a command line arg
to the Naming Service so that one can shut the idle timeout off easily.
Check out
http://www.orl.co.uk/omniORB/javaORBs.html
For an interesting breakdown. It is a little dated, Sun has fixed
several of the items mentioned.
Cheers,
-Armen
--
Armen Yampolsky
Axiom Software Labs
New York
