Crate wayland_server

Expand description

Server-side Wayland connector

Overview

This crate provides the interfaces and machinery to safely create servers for the Wayland protocol. It can be used as either a rust implementatin of the protocol, or as a wrapper around the system-wide libwayland-server.so if you need interoperability with other libraries. This last case is activated by the use_system_lib cargo feature.

The Wayland protocol revolves around the creation of various objects and the exchange of messages associated to these objects. Whenever a client connects, a Display object is automatically created in their object space, which they use as a root to create new objects and bootstrap their state.

Protocol and messages handling model

The protocol being bi-directional, you can send and receive messages. Sending messages is done via methods of Rust objects corresponding to the wayland protocol objects, receiving and handling them is done by providing callbacks.

Resources

The protocol and message model is very similar to the one of wayland-client, with the main difference being that the underlying handles to objects are represented by the Resource<I> type, very similarly to proxies in wayland-client.

These resources are used to send messages to the client (in the Wayland context, these are called “events”). You usually don’t use them directly, and instead call methods on the Rust objects themselves, which invoke the appropriate Resource methods. It is also possible to directly use the Resource::<I>::send(..) method.

There is not a 1 to 1 mapping between Rust object instances and protocol objects. Rather, you can think of the Rust objects as Rc-like handles to a Wayland object. Multiple instances of a Rust object can exist referring to the same protocol object.

Similarly, the lifetimes of the protocol objects and the Rust objects are not tightly tied. As protocol objects are created and destroyed by protocol messages, it can happen that an object gets destroyed while one or more Rust objects still refer to it. In such case, these Rust objects will be disabled and the alive() method on the underlying Resource<I> will start to return false. Events that are subsequently sent to them are ignored.

Filters

Your wayland objects can receive requests from the client, which need to be processed. To do so, you can assign Filters to your object. These are specially wrapped closure so that several objects can be assigned to the same Filter, to ease state sharing between the code handling different objects.

All objects must be assigned to a filter, even if it is for doing nothing. Failure to do will cause a panic!() if a request is dispatched to the faulty object.

A Rust object passed to your implementation is guaranteed to be alive (as it just received a request), unless the exact message received is a destructor (which is indicated in the API documentations).

General structure

The core of your server is the Display object. It represent the ability of your program to process Wayland messages. Once this object is created, you can configure it to listen on one or more sockets for incoming client connections (see the Display docs for details).

wayland-server does not include an event loop, and you are expected to drive the wayland socket yourself using the Display::flush_clients and Display::dispatch methods. The Display::get_poll_fd methods provides you with a file descriptor that can be used in a polling structure to integrate the wayland socket in an event loop.