Add a function to recalculate all nodes associated with a driver by
iterating the graph for each driver node. We used to do this in an
incremental way, which is easy to join graph but expensive to
split.
A full scan simplifies some things and we can't avoid it when we
need to calculate latencies later. It will also simplifies assigning
master and slave roles to drivers when the graphs are joined/split.
When we link/unlink or add/remove nodes, recalc the graph.
Pass a message around to make things more extensible later.
Keep fds per message if we ever want to write individual
messages.
Pass number of fds in the message header. We might need this to
close the fds when the proxy is gone.
Add a node subscribe params method that automatically emits the new
params when they change so that we can avoid an enum_params.
Use this in the stream object.
Remove the control messages to update controls in stream, use the
set_param. This is more overhead but allows for notifications to other
clients.
Make it possible to update many controls in one go.
Make a get node method that binds to the server side node of the
client-node immediately. use this in the remote_export and always
return a node proxy.
Use the node proxy to get property updates and signal those in the
stream.
Output controls can be linked to many input controls and many input
controls can receive input from many output controls. Keep the control
link information inside the link.
Destroying a local proxy will also attempt to destroy the remote
resource unless the server already did that and told us to remove
the proxy.
Fix some cleanups.
Just prepare the output on the port and signal ready. When the graph
completes we will be signaled again to recycle the buffer and
prepare more output if we can.
Improve the bookkeeping a little when activating nodes.
Fix race with moving nodes between drivers.
When we have a client driver node, it will have prepared the io
areas for us before sending the ready signal. We then need to run
the converters before signaling ready.
It is assumed that the output io area is ready when the ready
callback is called so that the next elements in the graph can be
triggered immediately. When the graph finishes, the node that
triggered the ready (the driver) is scheduled to recycle and complete
the graph.
