Hmmm....
You have multiple threads calling the same Client object.
Yes, the Client object is not thread safe, it is not even ok to call
invoke() while the service hasn't returned a response -- which comes into
play when you deal with async services.
Client uses Instance variables to keep track of it's state... For example:
public Object[] invoke(OperationInfo op, Object[] params) throws
Exception
{
try
{
OutMessage msg = new OutMessage(url);
msg.setBody(params);
msg.setChannel(getOutChannel());
context = new MessageContext();
context.setService(service);
context.setXFire(xfire);
context.setBinding(binding);
context.setProperty(CLIENT_MODE, Boolean.TRUE);
context.setClient(this);
Context is such an instance variable... Every invocation of invoke creates a
new MessageContext, so the still running call get's it's message context
changed while in situ....
There is a JIRA Issue to rearchitect the Client side, but we have some way
to go, so don't expect something in Xfire 1.1.
Try creating a fresh Client for every Thread, if that doesn't show this
behaviour, it's the above described.
Mika
-----Ursprüngliche Nachricht-----
Von: Sascha Sadat-Guscheh [mailto:[EMAIL PROTECTED]
Gesendet: Donnerstag, 30. März 2006 15:00
An: [email protected]
Betreff: Re: AW: [xfire-user] strange xfire (concurrency?) behaviour
add-on the second:
it seems that there's an issue with threading, NOT with concurrency
(sound stoopid, i know)
if i run the threaded test (which i posted before) i get all these
strange errors. when i run the same test, i.e..
run the callMethod 1000 times from 3 different machines against the
server, everything is ok.
so maybe the client classes are not threadsafe?
sorry for spamming the list, but there's a big dark deadline hanging
over us ..
