What is the tf threading model?

edit

A TransformListener class is continuously listening for incoming coordinate transforms from TransformBroadcasters. Any blocking call would severely disrupt the flow of information to the listener. Therefore, a transform listener, by default, spins its own thread. This particularly helps with calls to functions like tf::Transformer::waitForTransform which can block callbacks while it is waiting to hear back from tf.

• The API for tf::Transformer, tf::TransformListener and tf::TransformSender all use mutex locks to be threadsafe.
• tf::Transformer::waitForTransform requires a dedicated listening thread (provided by default by tf::TransformListener)

If TransformListener spins a thread it will only service callbacks in the custom callback queue for the tf topic.