Skip to content

CONTRIBUTING.md

Latest commit

 

History

History
61 lines (48 loc) · 3.37 KB

CONTRIBUTING.md

File metadata and controls

61 lines (48 loc) · 3.37 KB
Raw

Contributing to Nib

Thanks for considering a contribution. Nib is a small project with two codebases in one repo — a Rust/GTK4 host app and a Flutter client — so a few conventions keep them consistent.

Getting set up

Follow Building & Running in the README for prerequisites and how to run the host app (cargo run) and client app (flutter run).

Where things live

  • src/daemon/ — the streaming pipeline: portal/D-Bus session setup (virtual_monitor.rs), GStreamer pipeline construction (screencast.rs), the binary input protocol (input_injector.rs), ADB transport (adb_transport.rs), and the cursor-keepalive overlay workaround (cursor_keepalive.rs). daemon/mod.rs orchestrates all of these via NibDaemon.
  • src/ui/ — GTK4/Libadwaita widgets (device list, gesture help, display settings).
  • src/i18n.rs — the app's translation lookup (tr(key)); see its doc comment for why this exists instead of gettext.
  • nib_client/ — the Flutter/Dart companion app; see its own README.md if present, and nib_client/android/.../MainActivity.kt for the Android MediaCodec decoder that the iOS side still needs an equivalent for.

Conventions

  • Errors: daemon-side fallible functions return Result<_, daemon::DaemonError> (see src/daemon/error.rs), not String. Prefer ? with DaemonError's From impls (ashpd::Error, zbus::Error, std::io::Error) over hand-rolled .map_err(|e| e.to_string()). If you need to detect a specific failure (e.g. a cancelled portal request), add a method like DaemonError::is_cancelled() rather than matching on the message string.
  • Panics: avoid .unwrap()/.expect()/panic!() on anything reachable at runtime (portal calls, D-Bus replies, GStreamer state changes, mutex locks). It's fine in main()-time setup that truly can't recover, and in #[cfg(test)] code.
  • Docs: every pub item should have a /// doc comment explaining why, not just what — run RUSTFLAGS="-W missing_docs" cargo check to catch gaps. Comments on tricky bits (portal quirks, sandbox workarounds, GStreamer caps negotiation) should explain the constraint that forced the code to look the way it does, since that context isn't visible from the diff alone.
  • Sandboxing quirks: this app has been bitten repeatedly by differences between running under Flatpak confinement and running natively (cargo run). If you touch sandbox detection (ScreencastPipeline::is_running_confined) or anything encoder/PipeWire-related, test both paths if you can, and leave a comment explaining what broke and why if you're fixing a regression — see the git history around screencast.rs for the level of detail expected.

Before opening a PR

CI runs cargo fmt --check, cargo clippy -- -D warnings, cargo build, and lints the Flatpak manifest. Run these locally first:

cargo fmt
cargo clippy --all-targets -- -D warnings
cargo test

If you change dependencies in Cargo.toml, cargo-sources.json (used by the Flatpak build) needs to be regenerated to match the new Cargo.lock. There's no committed generator script yet; the standard tool is flatpak-cargo-generator.py.

Code of Conduct

This project follows the GNOME Code of Conduct.