Add support for choosing between native-tls and rustls-tls backends
through feature flags, with native-tls as the default for maximum
platform compatibility.
Key changes:
- Add mutually exclusive native-tls and rustls-tls feature flags
- Use conditional compilation to select TLS implementation
- Configure rustls-tls with platform certificate verifier
- Refactor to workspace-based dependency management
- Update CI workflows with improved cross-compilation support
- Add comprehensive TLS backend documentation
The native-tls backend uses system TLS libraries (OpenSSL on Linux,
Secure Transport on macOS, SChannel on Windows) while rustls-tls
provides a pure Rust implementation with platform certificate stores.
- Update MSRV to 1.85 and Rust edition to 2024.
- Refactor all logging macros to use inline argument formatting.
- Fix import order in main.rs and examples.
- Add async environment variable setter to main.rs as safe facade.
- Update many dependencies to latest versions across all crates
- Switch from `once_cell::OnceCell` to `std::sync::OnceLock` where appropriate
- Update OAuth to use stateful `reqwest` for HTTP requests
- Fix Rodio backend to honor the requested sample format
* discovery: use opaque error type for DnsSdError
This helps to decouple discovery and core by not leaking implementation
details of the zeroconf backend into Error conversion impls in core.
* discovery: map all MDNS/DNS-SD errors to DiscoveryError::DnsSdError
previously, libmdns errors would use a generic conversion
from std::io::Error to core::Error
* discovery: use an opaque type for the handle to the DNS-SD service
* discovery: make features additive
i.e. add with-libmdns instead of using not(with-dns-sd).
The logic is such that enabling with-dns-sd in addition to the default
with-libmdns will still end up using dns-sd, as before.
If only with-libmdns is enabled, that will be the default.
If none of the features is enabled, attempting to build a `Discovery`
will yield an error.
* discovery: add --zeroconf-backend CLI flag
* discovery: Add minimal Avahi zeroconf backend
* bump MSRV to 1.75
required by zbus >= 4
* discovery: ensure that server and dns-sd backend shutdown gracefully
Previously, on drop the the shutdown_tx/close_tx, it wasn't guaranteed
the corresponding tasks would continue to be polled until they actually
completed their shutdown.
Since dns_sd::Service is not Send and non-async, and because libmdns is
non-async, put them on their own threads.
* discovery: use a shared channel for server and zeroconf status messages
* discovery: add Avahi reconnection logic
This deals gracefully with the case where the Avahi daemon is restarted
or not running initially.
* discovery: allow running when compiled without zeroconf backend...
...but exit with an error if there's no way to authenticate
* better error messages for invalid options with no short flag
* core: Create credentials from access token via OAuth2
* core: Credentials.username is optional: not required for token auth.
* core: store auth data within session. We might need this later if need to re-auth and original creds are no longer valid/available.
* bin: New --token arg for using Spotify access token. Specify 0 to manually enter the auth code (headless).
* bin: Added --enable-oauth / -j option. Using --password / -p option will error and exit.
* core: reconnect session if using token authentication
Token authenticated sessions cannot use keymaster. So reconnect using the reusable credentials we just obtained. Can perhaps remove this
workaround once keymaster is replaced with login5.
* examples: replace password login with token login
When started at boot as a service discovery may fail due to it
trying to bind to interfaces before the network is actually up.
This could be prevented in systemd by starting the service after
network-online.target but it requires that a wait-online.service is
also enabled which is not always the case since a wait-online.service
can potentially hang the boot process until it times out in certain situations.
This allows for discovery to retry every 10 secs in the 1st 60 secs of uptime
before giving up thus papering over the issue and not holding up the boot process.
This should fix remaining cases of a client connecting, and failing
to start playback from *beyond* the beginning when `librespot` is
still loading that track.
This undoes the `suppress_loading_status` workaround from #430,
under the assumption that the race condition reported there has
since been fixed on Spotify's end.
Include a systemd.service file to run as a user
This new `systemd.service` file allows running and controlling the service as
an unpriviledged user (generally, as part of a user session).