We have some library RPC calls, generated a while ago for some C/Motif
programs. These calls are basic functions, stored in a dynamic library.

After some work, we got them working with 1.0.2, where the code was sync, so
multiple threads were not executing the same library calls, since the RPC
calls are not thread safe. We have been using them in JDK1.0.2 since last
sept, if not earlier.

With jdk1.1, we moved the native code that calls these RPC routines into the
JNI1.1 functionality. That worked fine; in fact, it ran faster. JDK 1.1,
able to execute these RPC calls going through the JNI interface (with some
native code we wrote for the access points.) Note that JDK 1.1 also accepted
the original 1.0.2 native routines as well, but we figured we would convert
the entire ball of wax to 1.1.

With jdk 1.1.1, the original native routines that called these RPC routines
would not work anymore. (Whether we compiled the JNI code with JDK 1.1, or
with JDK 1.1.1.) What happens is not a Java error, but an RPC error. The RPC
call fails when it's executing with the following error:

fec-a-dv: RPC: Miscellaneous tli error - No message currently available

(Note that fec-a-dv is our RPC host in this case.) It appears that some new
changes directly affect the standard RPC lib on Solaris boxes, or the
libraries are colliding with standard solaris libraries.

The same code works fine with JDK 1.1. There have been reports on
comp.lang.java.* groups about the same condition with other folks as well,
specifically the 'programmer' group.

At this point, we can not migrate to JDK 1.1.1 until this is working, since we
have a huge quantity of RPC's that we want to move to Java from C. (We access
our mainframe via RPC)

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=

The attached tar file to reproduce the problem comes from a different
customer, but demonstrates the problem with JDK 1.1.1 and RPC.