Screen share 2: Spike — LiveKit video publish/receive and codec facts #149

Closed
opened 2026-11-12 02:06:53 +00:00 by charles · 0 comments
Owner

Files:

  • Create: crates/fluxer-voice/tests/livekit_video_spike.rs, docs/superpowers/research/2026-11-11-screenshare-spikes.md (§1)
  • Modify: nothing else (production video code is Screen share 16/17)

Interfaces:

  • Produces facts, not APIs: in spikes §1 record, for the pinned SDK and the prebuilt webrtc-a545de4 on Linux and (from the Windows runner's build log) Windows: the defines in webrtc.ninja (WEBRTC_USE_H264, RTC_USE_LIBAOM_AV1_ENCODER, RTC_DAV1D_IN_INTERNAL_DECODER_FACTORY, WEBRTC_USE_PIPEWIRE); VideoEncoderBackend::list_available() and video_encoder_codecs(backend) for each backend; whether the PeerConnectionFactory is reachable through livekit's public API (and how) for receiver capabilities; which codecs a second session actually decodes (frames arrive) for VP8, VP9, AV1, H264; CPU time per second of encoding 1280×720/30 and 1920×1080/30 for each software encoder on the runner; whether building with libva-dev installed enables VAAPI (build log line) and what list_available then says; the encoder implementation name in the sender's stats.

  • The test (#[ignore] unless FLUXER_LIVEKIT_TESTS=1) builds its I420 frames inline (moving bars) rather than using fluxer-video, so the spike does not wait for Task 1.

  • Step 1: Write the spike test (it uses livekit directly, not the session, which has no video yet): two sessions on the local server (tools/livekit-test-server.sh), session A publishes a 1280×720/30 screencast source with TrackSource::Screenshare, simulcast: false, for each codec in turn; session B subscribes and counts decoded frames and their size for 5 s; print a table. Also decode_probe (what B could decode) and encoder_backends (lists).

  • Step 2: Run it locally with and without libva-dev headers; record the tables in spikes §1 with the commands used.

  • Step 3: Commit test(voice): LiveKit video spike and codec facts. Acceptance: spikes §1 answers every question above or names why it could not.

Depends on: nothing (can start at once)

Plan: https://git.forked.io/charles/fluxer-native/src/branch/main/docs/superpowers/plans/2026-11-11-screenshare.md
Spec: https://git.forked.io/charles/fluxer-native/src/branch/main/docs/superpowers/specs/2026-11-11-fluxer-native-screenshare-design.md

**Files:** - Create: `crates/fluxer-voice/tests/livekit_video_spike.rs`, `docs/superpowers/research/2026-11-11-screenshare-spikes.md` (§1) - Modify: nothing else (production video code is Screen share 16/17) **Interfaces:** - Produces facts, not APIs: in *spikes §1* record, for the pinned SDK and the prebuilt `webrtc-a545de4` on Linux and (from the Windows runner's build log) Windows: the defines in `webrtc.ninja` (`WEBRTC_USE_H264`, `RTC_USE_LIBAOM_AV1_ENCODER`, `RTC_DAV1D_IN_INTERNAL_DECODER_FACTORY`, `WEBRTC_USE_PIPEWIRE`); `VideoEncoderBackend::list_available()` and `video_encoder_codecs(backend)` for each backend; whether the `PeerConnectionFactory` is reachable through `livekit`'s public API (and how) for receiver capabilities; which codecs a second session actually decodes (frames arrive) for VP8, VP9, AV1, H264; CPU time per second of encoding 1280×720/30 and 1920×1080/30 for each software encoder on the runner; whether building with `libva-dev` installed enables VAAPI (build log line) and what `list_available` then says; the encoder implementation name in the sender's stats. - The test (`#[ignore]` unless `FLUXER_LIVEKIT_TESTS=1`) builds its I420 frames inline (moving bars) rather than using `fluxer-video`, so the spike does not wait for Task 1. - [ ] **Step 1:** Write the spike test (it uses `livekit` directly, not the session, which has no video yet): two sessions on the local server (`tools/livekit-test-server.sh`), session A publishes a 1280×720/30 screencast source with `TrackSource::Screenshare`, `simulcast: false`, for each codec in turn; session B subscribes and counts decoded frames and their size for 5 s; print a table. Also `decode_probe` (what B could decode) and `encoder_backends` (lists). - [ ] **Step 2:** Run it locally with and without `libva-dev` headers; record the tables in *spikes §1* with the commands used. - [ ] **Step 3:** Commit `test(voice): LiveKit video spike and codec facts`. Acceptance: *spikes §1* answers every question above or names why it could not. Depends on: nothing (can start at once) Plan: https://git.forked.io/charles/fluxer-native/src/branch/main/docs/superpowers/plans/2026-11-11-screenshare.md Spec: https://git.forked.io/charles/fluxer-native/src/branch/main/docs/superpowers/specs/2026-11-11-fluxer-native-screenshare-design.md
charles added this to the Screen share milestone 2026-11-12 02:06:53 +00:00
Sign in to join this conversation.
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
charles/fluxer-native#149
No description provided.