[omniORB] Single thread hanging...

[Mark Zimmerman]

On Thu, Sep 05, 2002 at 06:14:21PM +0100, Marcus MacWilliam wrote:
> Hello,
> 
> I am using the following software:
> 
> Solaris 8, gcc v3.1, gdb 5.2, python 1.5.2, OmniORB v3.0.5.
> 
> I initialise my own server POA to have SINGLE_THREAD_MODEL
> as its thread policy, all the others are as the "RootPOA".
> 
> OK, the problem is when I pass a CORBA object as a parameter to
> a method, and attempt to call a method on the passed object. The server
> just freezes. (This works fine on NT).
> 
> Example:
> 
> void CMyObject::doSomething( MySecondObject_ptr obj2 )
> {
>     obj2->method(); // Call to method() freezes the server.
> }
> 
> The server hangs trying to call the method() on the passed object..
> 
> If I change the thread policy to ORB_CTRL_MODEL, it works fine,
> but defeats the object. The CORBA server MUST be single threaded,
> because the
> CORBA server is acting as a client to another server using sockets. And
> that
> other server is single threaded. (As is the client code within the CORBA
> server).
> 
> Does anyone know why invoking a method on a CORBA object from within my
> server causes the server to freeze. The output from gdb, is as follows:
> 
> where:
> #0  0xfeb19c64 in _lwp_sema_wait () from /usr/lib/libc.so.1
> #1  0xfe9497fc in _park () from /usr/lib/libthread.so.1
> #2  0xfe9494d8 in _swtch () from /usr/lib/libthread.so.1
> #3  0xfe947c2c in cond_timedwait () from /usr/lib/libthread.so.1
> #4  0xfe9479d8 in pthread_cond_timedwait () from /usr/lib/libthread.so.1
> 
--snip--

Try adding /usr/lib/lwp to your LD_LIBRARY_PATH. This will give you a
better thread library. If you search deep enough in the Sun problem
reports you will find that they have abandoned all hope of fixing all
of the problems in the /usr/lib/libthread.so.1 version on Solaris 8.
The new design in /usr/lib/lwp/libthread.so.1 is the default thread
library in Solaris 9.

-- Mark