lesavka/client/src/input/camera/capture_pipeline.rs

impl CameraCapture {
    pub fn new(device_fragment: Option<&str>, cfg: Option<CameraConfig>) -> anyhow::Result<Self> {
        Self::new_with_capture_profile(device_fragment, cfg, None)
    }

    pub fn new_with_capture_profile(
        device_fragment: Option<&str>,
        cfg: Option<CameraConfig>,
        capture_profile_override: Option<(u32, u32, u32)>,
    ) -> anyhow::Result<Self> {
        gst::init().ok();

        // Select source: V4L2 device or test pattern
        let (src_desc, dev_label, allow_mjpg_source) = match device_fragment {
            Some(fragment)
                if fragment.eq_ignore_ascii_case("test")
                    || fragment.eq_ignore_ascii_case("videotestsrc") =>
            {
                let pattern =
                    std::env::var("LESAVKA_CAM_TEST_PATTERN").unwrap_or_else(|_| "smpte".into());
                (
                    format!("videotestsrc is-live=true pattern={pattern}"),
                    format!("videotestsrc:{pattern}"),
                    false,
                )
            }
            Some(path) if path.starts_with("/dev/") => (
                format!("v4l2src device={path} do-timestamp=true"),
                path.to_string(),
                true,
            ),
            Some(fragment) => {
                let dev = Self::find_device(fragment)
                    .with_context(|| format!("requested camera '{fragment}' was not found"))?;
                (format!("v4l2src device={dev} do-timestamp=true"), dev, true)
            }
            None => {
                let dev = "/dev/video0".to_string();
                (format!("v4l2src device={dev} do-timestamp=true"), dev, true)
            }
        };

        let output_mjpeg = cfg.map_or_else(
            || {
                std::env::var("LESAVKA_CAM_CODEC").ok().is_some_and(|v| {
                    matches!(v.to_ascii_lowercase().as_str(), "mjpeg" | "mjpg" | "jpeg")
                })
            },
            |cfg| matches!(cfg.codec, CameraCodec::Mjpeg),
        );
        let jpeg_quality = env_u32("LESAVKA_CAM_JPEG_QUALITY", 85).clamp(1, 100);
        let capture_profile = capture_profile_override.unwrap_or_else(|| resolved_capture_profile(cfg));
        let (capture_width, capture_height, capture_fps) = capture_profile;
        let (width, height, fps) = resolved_output_profile(cfg, capture_profile);
        let keyframe_interval = env_u32("LESAVKA_CAM_KEYFRAME_INTERVAL", fps.min(5)).clamp(1, fps);
        let source_profile = camera_source_profile(allow_mjpg_source);
        let use_mjpg_source = source_profile == CameraSourceProfile::Mjpeg;
        let passthrough_mjpg_source =
            use_mjpg_source && capture_profile == (width, height, fps);
        let (enc, kf_prop) = if use_mjpg_source && !output_mjpeg {
            ("x264enc", Some("key-int-max"))
        } else {
            Self::choose_encoder()
        };
        match source_profile {
            CameraSourceProfile::Mjpeg if !output_mjpeg => {
                tracing::info!("📸 using MJPG source with software encode");
            }
            CameraSourceProfile::AutoDecode => {
                tracing::info!("📸 using auto-decoded webcam source (raw/MJPEG accepted)");
            }
            _ => {}
        }
        let enc_opts = Self::encoder_options(enc, kf_prop, keyframe_interval);
        if output_mjpeg {
            tracing::info!("📸 outputting MJPEG frames for UVC (quality={jpeg_quality})");
        } else {
            tracing::info!("📸 using encoder element: {enc}");
        }
        #[cfg(not(coverage))]
        let have_nvvidconv = gst::ElementFactory::find("nvvidconv").is_some();
        let preenc = match enc {
            // ───────────────────────────────────────────────────────────────────
            // Jetson (has nvvidconv)      Desktop (falls back to videoconvert)
            // ───────────────────────────────────────────────────────────────────
            #[cfg(not(coverage))]
            "nvh264enc" if have_nvvidconv =>
                format!(
                    "nvvidconv ! video/x-raw(memory:NVMM),format=NV12,width={width},height={height},framerate={fps}/1 !"
                ),
            #[cfg(not(coverage))]
            "nvh264enc" /* else */ =>
                format!(
                    "videoconvert ! video/x-raw,format=NV12,width={width},height={height},framerate={fps}/1 !"
                ),
            #[cfg(not(coverage))]
            "vaapih264enc" =>
                format!(
                    "videoconvert ! video/x-raw,format=NV12,width={width},height={height},framerate={fps}/1 !"
                ),
            _ =>
                format!(
                    "videoconvert ! video/x-raw,width={width},height={height},framerate={fps}/1 !"
                ),
        };

        // let desc = format!(
        //     "v4l2src device={dev} do-timestamp=true ! {raw_caps},width=1280,height=720 ! \
        //     videoconvert ! {enc} key-int-max=30 ! \
        //     h264parse config-interval=-1 ! \
        //     appsink name=asink emit-signals=true max-buffers=60 drop=true"
        // );
        // tracing::debug!(%desc, "📸 pipeline-desc");
        // Build a pipeline that works for any of the three encoders.
        //  * nvh264enc needs NVMM memory caps;
        //  * vaapih264enc wants system-memory caps;
        //  * x264enc needs the usual raw caps.
        let preview_tap_path = camera_preview_tap_path();
        let preview_tap_branch = camera_preview_tap_branch(width, height, fps);
        let source_raw_caps = format!(
            "video/x-raw,width={capture_width},height={capture_height},framerate={capture_fps}/1"
        );
        let raw_source_chain = camera_raw_source_chain(
            &src_desc,
            &source_raw_caps,
            capture_width,
            capture_height,
            capture_fps,
            source_profile,
        );
        let normalized_raw_chain = format!(
            "{raw_source_chain} ! {}",
            camera_output_raw_chain(width, height, fps)
        );
        let desc = if preview_tap_path.is_some() {
            if output_mjpeg {
                if passthrough_mjpg_source {
                    format!(
                        "{src_desc} ! \
                         image/jpeg,width={width},height={height},framerate={fps}/1 ! \
                         tee name=t \
                         t. ! queue max-size-buffers=30 leaky=downstream ! \
                              appsink name=asink emit-signals=true max-buffers=60 drop=true \
                         t. ! queue max-size-buffers=2 leaky=downstream ! jpegdec ! \
                              {preview_tap_branch}"
                    )
                } else {
                    format!(
                        "{normalized_raw_chain} ! \
                         tee name=t \
                         t. ! queue max-size-buffers=30 leaky=downstream ! \
                              videoconvert ! jpegenc quality={jpeg_quality} ! \
                              appsink name=asink emit-signals=true max-buffers=60 drop=true \
                         t. ! queue max-size-buffers=2 leaky=downstream ! \
                              {preview_tap_branch}"
                    )
                }
            } else {
                format!(
                    "{normalized_raw_chain} ! \
                     tee name=t \
                     t. ! queue max-size-buffers=30 leaky=downstream ! \
                          {preenc} {enc_opts} ! \
                          h264parse config-interval=-1 ! video/x-h264,stream-format=byte-stream,alignment=au ! \
                          appsink name=asink emit-signals=true max-buffers=60 drop=true \
                     t. ! queue max-size-buffers=2 leaky=downstream ! \
                          {preview_tap_branch}"
                )
            }
        } else if output_mjpeg {
            if passthrough_mjpg_source {
                format!(
                    "{src_desc} ! \
                     image/jpeg,width={width},height={height},framerate={fps}/1 ! \
                     queue max-size-buffers=30 leaky=downstream ! \
                     appsink name=asink emit-signals=true max-buffers=60 drop=true"
                )
            } else {
                format!(
                    "{normalized_raw_chain} ! \
                     videoconvert ! jpegenc quality={jpeg_quality} ! \
                     queue max-size-buffers=30 leaky=downstream ! \
                     appsink name=asink emit-signals=true max-buffers=60 drop=true"
                )
            }
        } else {
            format!(
                "{normalized_raw_chain} ! \
                 {preenc} {enc_opts} ! \
                 h264parse config-interval=-1 ! video/x-h264,stream-format=byte-stream,alignment=au ! \
                 queue max-size-buffers=30 leaky=downstream ! \
                 appsink name=asink emit-signals=true max-buffers=60 drop=true"
            )
        };
        tracing::info!(
            %enc,
            capture_width,
            capture_height,
            capture_fps,
            output_width = width,
            output_height = height,
            output_fps = fps,
            ?desc,
            "📸 using encoder element"
        );

        let pipeline: gst::Pipeline = gst::parse::launch(&desc)
            .context("gst parse_launch(cam)")?
            .downcast::<gst::Pipeline>()
            .expect("not a pipeline");

        tracing::debug!("📸 pipeline built OK – setting PLAYING…");
        let sink: gst_app::AppSink = pipeline
            .by_name("asink")
            .expect("appsink element not found")
            .downcast::<gst_app::AppSink>()
            .expect("appsink down‑cast");

        spawn_camera_bus_logger(&pipeline, dev_label.clone());
        if let Err(err) = pipeline.set_state(gst::State::Playing) {
            let _ = pipeline.set_state(gst::State::Null);
            return Err(err.into());
        }
        tracing::info!("📸 webcam pipeline ▶️ device={dev_label}");

        let preview_tap_running = if let Some(path) = preview_tap_path {
            let preview_sink = pipeline
                .by_name("preview_sink")
                .context("missing camera preview tap appsink")?
                .downcast::<gst_app::AppSink>()
                .expect("camera preview tap appsink");
            Some(spawn_camera_preview_tap(preview_sink, path))
        } else {
            None
        };

        Ok(Self {
            pipeline,
            sink,
            preview_tap_running,
            pts_rebaser: crate::live_capture_clock::DurationPacedSourcePtsRebaser::default(),
            frame_duration_us: (1_000_000u64 / u64::from(fps.max(1))).max(1),
        })
    }

    pub fn pull(&self) -> Option<VideoPacket> {
        let sample = self.sink.pull_sample().ok()?;
        let buf = sample.buffer()?;
        let map = buf.map_readable().ok()?;
        let source_pts_us = buf.pts().map(|ts| ts.nseconds() / 1_000);
        let packet_duration_us = buf
            .duration()
            .map(|ts| (ts.nseconds() / 1_000).max(1))
            .unwrap_or(self.frame_duration_us);
        let timing = self.pts_rebaser.rebase_with_packet_duration(
            source_pts_us,
            packet_duration_us,
            crate::live_capture_clock::upstream_source_lag_cap(),
        );
        let pts = timing.packet_pts_us;
        static CAMERA_PACKET_COUNT: std::sync::atomic::AtomicU64 =
            std::sync::atomic::AtomicU64::new(0);
        let packet_index = CAMERA_PACKET_COUNT.fetch_add(1, std::sync::atomic::Ordering::Relaxed);
        log_camera_first_packet(packet_index, map.as_slice().len(), pts);
        log_camera_timing_sample(packet_index, timing, map.as_slice().len());
        Some(VideoPacket {
            id: 2,
            pts,
            data: map.as_slice().to_vec(),
            ..Default::default()
        })
    }
}

/// Resolve the profile requested from the local webcam.
///
/// The server UVC contract is applied after capture. Keeping these separate
/// prevents a browser-facing 640x480/20 gadget mode from forcing a local webcam
/// to expose that exact mode when the selected camera quality is 720p/30.
fn resolved_capture_profile(cfg: Option<CameraConfig>) -> (u32, u32, u32) {
    (
        env_u32("LESAVKA_CAM_WIDTH", cfg.map_or(1280, |cfg| cfg.width)),
        env_u32("LESAVKA_CAM_HEIGHT", cfg.map_or(720, |cfg| cfg.height)),
        env_u32("LESAVKA_CAM_FPS", cfg.map_or(25, |cfg| cfg.fps)).max(1),
    )
}

/// Resolve the profile emitted toward the remote UVC gadget.
fn resolved_output_profile(
    cfg: Option<CameraConfig>,
    capture_profile: (u32, u32, u32),
) -> (u32, u32, u32) {
    match cfg {
        Some(cfg)
            if env_flag_enabled("LESAVKA_CAM_LOCK_TO_SERVER_PROFILE")
                || !env_flag_enabled("LESAVKA_CAM_EMIT_UI_PROFILE") =>
        {
            (cfg.width, cfg.height, cfg.fps.max(1))
        }
        _ => capture_profile,
    }
}

fn env_flag_enabled(name: &str) -> bool {
    std::env::var(name).ok().is_some_and(|value| {
        let trimmed = value.trim();
        !(trimmed.is_empty()
            || trimmed.eq_ignore_ascii_case("0")
            || trimmed.eq_ignore_ascii_case("false")
            || trimmed.eq_ignore_ascii_case("no")
            || trimmed.eq_ignore_ascii_case("off"))
    })
}

fn log_camera_first_packet(packet_index: u64, bytes: usize, pts_us: u64) {
    if packet_index == 0 {
        tracing::info!(bytes, pts_us, "📸 upstream webcam frames flowing");
    }
}

fn should_log_camera_timing_sample(packet_index: u64) -> bool {
    crate::live_capture_clock::upstream_timing_trace_enabled()
        && (packet_index < 10 || packet_index.is_multiple_of(300))
}

fn log_camera_timing_sample(
    packet_index: u64,
    timing: crate::live_capture_clock::RebasedSourcePts,
    bytes: usize,
) {
    if should_log_camera_timing_sample(packet_index) {
        tracing::info!(
            packet_index,
            source_pts_us = timing.source_pts_us.unwrap_or_default(),
            source_base_us = timing.source_base_us.unwrap_or_default(),
            capture_base_us = timing.capture_base_us.unwrap_or_default(),
        capture_now_us = timing.capture_now_us,
        packet_pts_us = timing.packet_pts_us,
        pull_path_delay_us = timing.capture_now_us as i128 - timing.packet_pts_us as i128,
        used_source_pts = timing.used_source_pts,
        lag_clamped = timing.lag_clamped,
        lead_clamped = timing.lead_clamped,
        bytes,
        "📸 upstream webcam timing sample"
    );
}
}