[openorb-devel] Bug 831680 / ServerManagerImpl

[<Lar...@pp...>]

Michael,

I'm currently investigating Bug 831680 (NotificationServer hangs with more
than 10 consumer admins) and it seems the problem is related to the
maxThreadPoolSize / ServerManagerImpl, see my comments in the tracker.

I might be totally off here (it's the first time I look at the
ServerManagerImpl), but to me the implementation of enqueue_request /
notifyPoolThread looks suspicious. If there is no thread left in the
m_waiting_threads list, an incoming method call is enqued, but no pool
thread is notified. 

Who will ever take care of that request again?

Have you ever stress-tested with more simultanious method calls than
maxThreadPoolSize?

Lars

PS: The current ServerManagerImpl implementation contains several unused
variables and methods. Are you still working on that class or can I delete
stuff that my IDE reports as unused?