[Bug libgcj/7938] java-rmi-problem

["norbertf at gmx dot net" <gcc-bugzilla at gcc dot gnu dot org>]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=7938



------- Additional Comments From norbertf at gmx dot net  2003-08-02 18:50 -------
Subject: Re:  java-rmi-problem

hi andrew,

i have just checked out the latest gcc/gcj sources (3.4) from cvs and
compiled them on my mandrake 9.1 system. I recompiled my tests, using
the Makefile i provided in the last mail. 

Now the tests work!!!

But there are two minor issues remaining:

1) There is no "java.net.BindException"  when calling
LocateRegistry.createRegistry(port) on a port where another RMI-Registry
already sits.

you can test it by starting the server twice:

./testsrv &
./testsrv

The problem with this behaviour is: An application might believe that it
is the initial creator and owner of the registry - which is not true in
this case. Without knowing, it might use the registry served by another
application using the same port, "confusing" the other application.

Having a quick look at the code, i think the problem might be in the
code of the constructor in class UnicastConnectionManager, where the
IOException is just "swallowed". 

this is what the JDK reports - i think GCJ should have a similar
behaviour:

----------------------------
Exception in thread "main" java.rmi.server.ExportException: Port already
in use: 7777; nested exception is:
        java.net.BindException: Address already in use
----------------------------

2) The other issue is that JDK and GCJ are not compatible. And i don't
understand why. I have used the same rmic (gcj) for generating the stubs
and skeletons. When i mix my gcj-RMI-client and RMITestServer invoked
with the JDK, things don't work properly (for instance in this case only
the method calls "status" and "com3" work.)

regards,
norbert