Skip to content

linux.md

Latest commit

 

History

History
106 lines (72 loc) · 4.33 KB

linux.md

File metadata and controls

106 lines (72 loc) · 4.33 KB

Linux Support

Linux is a partial platform today. Device enumeration works; capture (getUserMedia) does not yet.

What works

MediaDevices.EnumerateDevices()

Returns MediaDeviceInfo[] for:

  • Video inputs: one per /dev/video* node. DeviceId is the device path, Label is the node filename, Kind = "videoinput", GroupId = "v4l2".
  • Audio inputs: one per ALSA card in /proc/asound/cards. DeviceId is hw:N (ALSA hardware reference), Label is "<cardId> (<cardDescription>)", Kind = "audioinput", GroupId = "alsa".

No V4L2 ioctl required for enumeration — just filesystem reads. Works in any process with read access to /dev and /proc/asound.

What doesn't work yet

GetUserMedia(...) — throws PlatformNotSupportedException

Actual frame / sample capture needs:

  1. V4L2 ioctl via P/Invoke to libc — roughly: VIDIOC_QUERYCAP, VIDIOC_ENUM_FMT, VIDIOC_S_FMT, VIDIOC_REQBUFS, VIDIOC_QBUF / VIDIOC_DQBUF, VIDIOC_STREAMON. Plus memory-mapping the driver's frame buffers.
  2. PulseAudio simple API (libpulse-simple) via P/Invoke — pa_simple_new, pa_simple_read, pa_simple_free. ALSA PCM (libasound) is the fallback when PulseAudio isn't running.
  3. Format conversion from V4L2 pixel formats (YUYV, NV12, MJPEG) into the library's common VideoPixelFormat enum, and PCM16/Float32 audio frames.

Estimated scope: ~500 lines of P/Invoke + buffer management + format conversion per platform component. Landed as a separate phase.

GetDisplayMedia(...) — throws PlatformNotSupportedException

Screen capture on Linux needs either:

  • XDG portal (org.freedesktop.portal.ScreenCast) — the modern / Wayland-compatible path. D-Bus method calls.
  • wlroots screencopy — Wayland-native on wlroots-based compositors.
  • X11 XShm + XDamage — X11-session fallback.

Phase 5 work.

Testing against WSL2

Development on a Windows host can exercise the Linux impl via WSL2. Two things to wire up:

USB cameras via usbipd

# On Windows (PowerShell, admin). Install once:
winget install usbipd
# List USB devices, find your camera:
usbipd list
# Bind + attach the device to WSL:
usbipd bind --busid 1-4
usbipd attach --wsl --busid 1-4

Inside the WSL2 distro:

# Verify the device is visible:
ls -l /dev/video*
# Should show at least /dev/video0 for the attached camera.
# Current user needs to be in the 'video' group:
sudo usermod -aG video $USER  # logout/login for effect

Microphone via PulseAudio WSL interop

WSL2 ships a PulseAudio server that proxies to the Windows host's audio stack. ALSA enumeration through /proc/asound/cards should already work without additional setup.

# Verify ALSA sees at least one card:
cat /proc/asound/cards
# Typical output when PulseAudio WSL interop is active:
#  0 [UltraSpeakers ]: HDA-Intel - HDA Intel PCH
#                      HDA Intel PCH at 0xa3240000 irq 172

Running the enumeration test

From inside the WSL distro with the repo checked out at the same path (via /mnt/d/users/tj/Projects/SpawnDev.MultiMedia):

cd /mnt/d/users/tj/Projects/SpawnDev.MultiMedia/SpawnDev.MultiMedia/SpawnDev.MultiMedia.DemoConsole
dotnet run --no-build -c Release -- "DesktopTests.EnumerateDevices_ReturnsNonNullArray"
dotnet run --no-build -c Release -- "DesktopTests.EnumerateDevices_DevicesHaveKind"

These two existing tests exercise the enumeration path platform-agnostically — the Linux impl hooks in via MediaDevices.EnumerateDevices() without any test-side changes.

Alternative: Docker

A minimal Dockerfile for CI Linux enumeration checks:

FROM mcr.microsoft.com/dotnet/sdk:10.0
WORKDIR /app
COPY . .
# Mount a loopback V4L2 device into the container for enumeration tests (host needs v4l2loopback):
# docker run --device /dev/video0:/dev/video0 ...
RUN dotnet build -c Release
CMD dotnet test -c Release --no-build

Docker can't access host hardware by default; for capture testing you need --device /dev/video0 + --group-add video at run time. For enumeration-only tests the /proc/asound/cards file is synthesized from the container's view and works out of the box.

macOS

Not supported today. Phase 5 items are AVFoundation (video capture) + CoreAudio (audio capture/playback). No dev environment for this yet (no macOS hardware on TJ's setup); contributions welcome.