#!/usr/bin/env bash # Launch headless Sway on the NVIDIA box for the lumen M0 capture spike. # # Runs on the user's *shared* session bus (NOT a private dbus-run-session) so that the # ScreenCast portal (xdg-desktop-portal-wlr) and the lumen host share one bus. After this # is up, run `prepare-session.sh` from a second shell to set the mode + portal env. # # Prereqs (see docs/linux-setup.md / scripts/bootstrap-ubuntu.sh): # - nvidia-drm.modeset=Y # - the NVIDIA GL/EGL userspace (libnvidia-gl-NNN) — provides libEGL_nvidia + the GLVND # vendor JSON; without it wlroots can't init EGL on the GPU and falls back to pixman, # which the ScreenCast portal cannot capture. # - membership in the `render` + `video` groups (re-login after `usermod -aG`). set -euo pipefail HERE="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" # shellcheck disable=SC1091 source "$HERE/env.sh" export DBUS_SESSION_BUS_ADDRESS="${DBUS_SESSION_BUS_ADDRESS:-unix:path=$XDG_RUNTIME_DIR/bus}" mkdir -p "$XDG_RUNTIME_DIR" 2>/dev/null || true chmod 700 "$XDG_RUNTIME_DIR" 2>/dev/null || true echo "starting headless Sway (renderer=$WLR_RENDERER) on bus $DBUS_SESSION_BUS_ADDRESS" echo "from another shell on this user: bash $HERE/prepare-session.sh" echo # wlroots opens the render node (/dev/dri/renderD128, group 'render'); NVIDIA's EGL uses it. # If this login predates `usermod -aG render,video` (groups not yet active), bridge with # `sg render` for this launch. The clean fix is to re-login so the groups apply natively. if id -nG | tr ' ' '\n' | grep -qx render; then exec sway --unsupported-gpu else echo "note: 'render' group not active in this login — bridging via 'sg render' (re-login to avoid)" exec sg render -c 'sway --unsupported-gpu' fi