fix(#540): pin weston socket + inject WAYLAND_DISPLAY globally
Services started by systemd in a gui-enabled container didn't have WAYLAND_DISPLAY set, so Wayland clients couldn't find the compositor. libwayland would fall back to a headless display or error out, leaving apps running invisibly while the VNC session showed a blank weston desktop (the double-screen problem). Fix in weston-vnc.nix: - Pass --socket=wayland-0 to weston so the socket name is deterministic (weston normally picks any free wayland-N name). - Set WAYLAND_DISPLAY=wayland-0 and XDG_RUNTIME_DIR=/run/user/0 as global environment.variables gated on hyperhive.gui.enable, so every service in the container inherits them automatically. - Update gui.json to include wayland_display for tooling that reads it. Update docs/gotchas.md with the rationale and pointer to #540.
This commit is contained in:
parent
3b76842853
commit
8a97277f20
2 changed files with 39 additions and 4 deletions
|
|
@ -264,9 +264,19 @@ connects to the compositor at `127.0.0.1:<vnc_port>`.
|
|||
`[15900, 16799]`. Mirrors the agent web-UI port pattern from
|
||||
`docs/gotchas.md::Web UI ports collide on hash` — same FNV-1a
|
||||
constant, different range. The compositor's startup script writes
|
||||
`/etc/hyperhive/gui.json = {"vnc_port":N,"auth":"none"}` so the
|
||||
harness reads the port at runtime; no nix-side / harness-side hash
|
||||
`/etc/hyperhive/gui.json = {"vnc_port":N,"auth":"none","wayland_display":"wayland-0"}`
|
||||
so the harness reads the port at runtime; no nix-side / harness-side hash
|
||||
duplication.
|
||||
- **Fixed Wayland socket name (`--socket=wayland-0`)**: weston is
|
||||
launched with `--socket=wayland-0` so the socket path is
|
||||
deterministic. `harness-base.nix` exports `WAYLAND_DISPLAY=wayland-0`
|
||||
and `XDG_RUNTIME_DIR=/run/user/0` as global system environment
|
||||
variables (gated on `hyperhive.gui.enable`) so every systemd service
|
||||
in the container inherits them. Without this, services starting
|
||||
Wayland clients could not find the compositor — libwayland falls
|
||||
back to a headless display or errors out, the app "works" on a
|
||||
second invisible display, and the VNC session shows a blank weston
|
||||
desktop (#540 double-screen).
|
||||
- **VNC bind address**: weston's VNC backend has no CLI
|
||||
bind-address flag (unlike the RDP backend's `--address`), so the
|
||||
listener binds `0.0.0.0`. The harness relay only connects via
|
||||
|
|
|
|||
|
|
@ -78,9 +78,11 @@
|
|||
done
|
||||
VNC_PORT=$((15900 + hash % 900))
|
||||
|
||||
# Marker file the harness reads at startup.
|
||||
# Marker file the harness reads at startup. Also records the
|
||||
# fixed Wayland socket name (`wayland-0`) so other tooling can
|
||||
# read it without inspecting the env-var injection below.
|
||||
${pkgs.coreutils}/bin/mkdir -p /etc/hyperhive
|
||||
${pkgs.coreutils}/bin/printf '{"vnc_port":%d,"auth":"none"}\n' \
|
||||
${pkgs.coreutils}/bin/printf '{"vnc_port":%d,"auth":"none","wayland_display":"wayland-0"}\n' \
|
||||
"$VNC_PORT" > /etc/hyperhive/gui.json || true
|
||||
|
||||
# --disable-transport-layer-security: skips the VeNCrypt TLS
|
||||
|
|
@ -90,11 +92,21 @@
|
|||
WESTON_INI=$(${pkgs.coreutils}/bin/mktemp /tmp/weston-XXXXXX.ini)
|
||||
${pkgs.coreutils}/bin/printf '[core]\nidle-time=0\n\n[vnc]\nauth-method=none\n' > "$WESTON_INI"
|
||||
|
||||
# --socket=wayland-0: pin the compositor's Wayland socket name
|
||||
# to `wayland-0` (weston default is to pick any free name such
|
||||
# as `wayland-1`). Pinning lets the WAYLAND_DISPLAY=wayland-0
|
||||
# global env injection below (see environment.variables) take
|
||||
# effect unconditionally — any Wayland client launched by any
|
||||
# systemd service in this container automatically connects to
|
||||
# this compositor instead of failing or starting a second
|
||||
# isolated display. Closes #540 (double-screen: VNC shows blank
|
||||
# weston desktop while services render on a different seat).
|
||||
exec ${pkgs.weston}/bin/weston \
|
||||
--config="$WESTON_INI" \
|
||||
--backend=vnc-backend.so \
|
||||
--renderer=pixman \
|
||||
--port="$VNC_PORT" \
|
||||
--socket=wayland-0 \
|
||||
--disable-transport-layer-security
|
||||
'';
|
||||
Restart = "on-failure";
|
||||
|
|
@ -102,6 +114,19 @@
|
|||
};
|
||||
};
|
||||
|
||||
# Expose the compositor's socket to every process in the container
|
||||
# so Wayland clients started by any systemd service (e.g. bitburner
|
||||
# launched via a `systemd.services.*` declaration in agent.nix) can
|
||||
# find the compositor without per-service wiring. Without these vars
|
||||
# a service starting a Wayland client would either fail to connect
|
||||
# (libwayland falls back to creating a headless display) or open a
|
||||
# second compositor entirely — both cause the VNC session to show a
|
||||
# blank desktop while apps appear to "work" elsewhere (#540).
|
||||
environment.variables = {
|
||||
WAYLAND_DISPLAY = "wayland-0";
|
||||
XDG_RUNTIME_DIR = "/run/user/0";
|
||||
};
|
||||
|
||||
# weston on the agent's interactive PATH so claude can run Wayland
|
||||
# clients / weston-info against the compositor.
|
||||
environment.systemPackages = [ pkgs.weston ];
|
||||
|
|
|
|||
Loading…
Reference in a new issue