feat-ish: Grocery bag of improvements

[urwrstkn8mare]

this is kind of a grocery bag of stuff i've been using locally:

• splits Moonlight input types across separate ENet channels and makes control parsing more tolerant
• drains the gamepad feedback queue properly and gives input channels more room
• makes DualSense touchpad events use the actual touch lifecycle instead of pressure, which made the touchpad feel better for me
• trims some packetizer/control-stream overhead
• moves Moonshine virtual inputs onto their own seat so they don't leak into the host desktop

Fixes #38.

personally tried this out on Spider-Man, and the touchpad felt better as well as responsiveness.

[Copilot]

Pull request overview

Note

Copilot was unable to run its full agentic suite in this review.

This PR optimizes video shard accumulation and control-stream responsiveness by reducing copies/allocations, caching AES-GCM state across packets, and improving input/touch handling and ENet channel configuration.

Changes:

• Replace shard-batch extension with an ownership-based API to enable move/append semantics and remove packetizer timing instrumentation.
• Update control-stream encryption/decryption to use a cached AES-128-GCM cipher keyed by key rotation id; increase ENet channel count and channel buffer sizes.
• Improve PS5 touchpad handling by using explicit Moonlight touch lifecycle events and track active touch pointers.

Reviewed changes

Copilot reviewed 6 out of 7 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
moonshine-core/src/session/stream/video/shard_batch.rs	Switch shard batch extension to move/append to avoid copies in common cases.
moonshine-core/src/session/stream/video/packetizer.rs	Use new shard batch extension API; remove per-block timing instrumentation.
moonshine-core/src/session/stream/control/mod.rs	Cache GCM cipher, increase channel_limit, drain feedback bursts, and adjust peer targeting + parse behavior.
moonshine-core/src/session/stream/control/input/mod.rs	Increase channel capacity to reduce input/control backpressure.
moonshine-core/src/session/stream/control/input/gamepad.rs	Use touch lifecycle event types and track active touch pointers for CancelAll.
moonshine-core/src/crypto.rs	Introduce GcmCipher to reuse AES-GCM initialization across packets and encrypt/decrypt in place.
dist/60-moonshine.rules	Change udev rules for Moonshine devices, including permissions/seat environment.

Comments suppressed due to low confidence (1)

moonshine-core/src/session/stream/control/mod.rs:1

• The loop no longer records which peer connected/disconnected, but later sends control messages (feedback/HDR/termination) anyway. Combined with the PeerId(0) hardcode, this makes outgoing delivery depend on an implementation detail (peer slot assignment). Track the peer_id from Event::Connect (and clear it on disconnect) and use that for outbound sends.

use std::net::SocketAddr;

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[urwrstkn8mare]

will wait for #116

[urwrstkn8mare]

still in draft until i can fully test it myself. there seems to be 2 regressions from main with detecting HDR (at least on Spiderman 2) and Forza Horizon 6 just being a black screen - might have something to do with #93 ). until i get that sorted out and am able to use Moonshine for a bit I can't really 100% back it. (most of the commits r fine just some that effect input handling I want to see how it compares to main).