An application with the ask close policy could decline every close, and
nothing could end it. TERMINATE_APP now takes force, which terminates
the session without asking. The taskbar offers it as Force quit once an
app has had longer than the close timeout to answer.
The close interaction now comes before Running -> Stopping, as blueprint
6.2 requires. The session stays running while the application answers,
and a declined close no longer moves Stopping back to Running. Stop
requests reach the supervisor on a channel, so a repeated or forced
request during the close wait is seen without registering again. A
declining window is brought to the front, so its prompt is visible. A
clean exit after a close request is recorded as a close on request. A
decline report is recognised after page output that lacks a newline.
Notes keeps a pending Save and close when asked again. It saves text
typed during that save before closing, and reports when it cannot save
because it is not connected.
A package declares its unsaved-change policy in [ui].close. With
"ask", the app shell dispatches a cancelable weftclose event before
the window closes and waits up to 2 s for the page. A page that
cancels the close keeps running, and the app shell reports this to
weft-appd with its readiness token, which returns the session to
running. The page closes itself later with window.close().
Notes declares "ask". With unsaved changes it offers Save and close,
Close without saving, and Cancel. It closes after a save only once the
save is acknowledged. The taskbar shows a declined close as running
again.
A repeated TERMINATE_APP during the close wait no longer removes the
session from the registry. The close wait also watches the runtime and
the IPC relay, and reports a failed app shell exit as a failure rather
than a close. A session being closed is not saved for restore, only the
32 most recent stopped sessions stay registered, and the taskbar shows
an entry as closing until its session stops.
TERMINATE_APP, which the taskbar sends, killed the session's processes
at once, so an application had no chance to finish, for example to
save. A running session is now asked to close through the compositor
(AppCloseRequest, which sends xdg_toplevel close to its windows) and
the app shell is given five seconds to exit by itself before what
remains is terminated. The stop reason records whether the app closed
or was terminated after the timeout. TERMINATE_APP replies stopping and
the session stays registered until it settles; a starting session is
stopped at once, and appd shutting down terminates sessions without a
close.
check_counter.py requires the compositor to ask the session's window
to close and the app shell to exit with status 0 on the close; a unit
test requires an app that ignores the close to be terminated after the
timeout and not before.
Session clients could still create layer-shell surfaces, including
full-screen overlays with exclusive keyboard input, and bind the input
method, which bypassed the refusal of shell roles. Both globals are now
hidden from clients bound to a session.
The compositor IPC connection is tracked by its read source, so a write
error can no longer let a second peer feed the same buffers; a failed
write shuts the connection down. An undecodable or oversized frame ends
the connection, since descriptors could otherwise pair with the wrong
messages, and the descriptor limit counts only descriptors no message
claims, so a burst of launches is not taken for abuse. When appd's
connection ends, the compositor closes the clients of the sessions it
attached.
weft-appd notices when the compositor closes the connection, resends a
message whose write failed after reconnecting, and refuses a launch
while the compositor is not connected instead of starting an app shell
that can never connect.
The compositor stored appd's (pid, session) announcements and never used
them, so no surface could be attributed to a session and any client
could claim any app ID or the panel role.
weft-appd now creates a connected socket pair for each session, gives
the app shell one end as standard input with WAYLAND_SOCKET=0 and no
WAYLAND_DISPLAY, and sends the other end to the compositor with the
session and app ID over the compositor IPC socket (SCM_RIGHTS). The
compositor inserts it as a client bound to that session: holding the
connection is the session's identity. A session's zweft_shell windows
carry its app ID, the panel and overlay roles are refused to sessions
with a new role_not_permitted error, the session's first toplevel is
reported to appd, an ended session's client is closed, and a client's
disconnection is reported. The compositor accepts one appd connection
at a time.
The counter desktop check now requires the compositor to bind the app's
client to its session and to close it when the session is terminated.
The runtime, the app shell and grant derivation each searched the
package stores and parsed wapp.toml on their own, so they could use
different roots; the runtime ignored [runtime].module, the image loader
looked for companion files the builder never wrote, and a failed image
mount fell back to the mutable directory install.
weft-appd now resolves a launch once: a verified image wins over a
directory in the same store, and when any image file exists the image
must mount or the launch is refused. The manifest must declare the
requested ID, and both entries must be plain relative paths to regular
files without links. The runtime receives --module and the app shell
--ui; neither looks for the package. Grants come from the same
manifest, and the mount belongs to the session.
The manifest parser and the image file names move to weft-ipc-types,
shared with weft-pack, whose check now applies the same entry rules.
Mapper devices are named after their per-launch mountpoint, so
concurrent sessions no longer collide.
appd created $XDG_RUNTIME_DIR/weft with the umask mode when it opened
its own socket, wrote its port or system token, or opened a portal
socket, so the 0700 mode requested for session sockets never applied.
All of them now use one helper that creates the directory 0700 and
tightens an existing one.
Session sockets are named by session ID and every test registry starts
at 1, so test processes running at the same time shared one socket path
in the common test runtime directory and broke each other's sessions.
Each test process now uses its own directory under the system temporary
directory, and directories of ended test processes are removed. The
test of a closed IPC connection connects to the socket itself instead
of relying on python3 in a child script.
A session whose component closes its IPC connection during startup now
stops before it is reported ready. The weft runtime directory is
created with mode 0700, and the socket is removed if its permissions
cannot be restricted.
IPC_FORWARD held the registry lock while awaiting a send to the
component; with a full queue it stalled appd, and once the relay was
closed it locked the registry again and deadlocked. It now sends without
waiting and reports an error when the session has no relay or its queue
is full.
A relay whose component disconnected left the session running with no
way to exchange messages. The session now stops when its IPC connection
closes. The relay releases its task and socket when dropped, so a
cancelled supervisor leaks neither, and the socket is created with mode
0600.
New tests fail against the previous code: IPC_FORWARD against a missing
relay and a full queue must answer at once, and a runtime that closes
its IPC connection while still running must end its session. The test
runtime directory lives under target/ instead of one directory per run.
dispatch opened each session's IPC socket and registered its sender
before the supervisor existed, and nothing released them: a relay whose
runtime never connected waited forever with its socket file in place,
and every stopped session left its sender registered, so app messages
for it were still accepted and queued.
The supervisor now opens the relay after its early checks, and the
session owns it: every exit path stops the relay task, removes the
socket and unregisters the sender. A session whose socket cannot be
opened does not start. The relay accepts a single connection.
A dispatch-level test launches a package whose runtime reports ready
and exits without connecting; it requires the stopped session to leave
no sender and no socket, and fails against the previous code. The
Counter and Notes checks pass end to end.
Every host import was registered and callable whatever the package
declared, read-only directory capabilities were preopened read-write,
and appd launched packages whose capabilities it could not satisfy. The
fetch import also could not be linked: the host returned a tuple where
the WIT declares the response record, so any component importing
weft:app/fetch failed to instantiate.
One capability vocabulary now lives in weft_ipc_types::capability and
is used by weft-pack, weft-appd and weft-runtime. weft-appd derives a
session's grants once, before the session exists, and refuses the launch
with error 403 when a capability is unknown, unsupported (hw:gpu:*) or
unsatisfiable (documents without an XDG documents directory). The
documents directory now comes from user-dirs.dirs instead of a guessed
~/Documents.
The runtime receives grants explicitly (--grant CAP, --preopen
HOST::GUEST::ro|rw), preopens read-only directories read-only, and checks
the grant on every notifications, clipboard and fetch call. Fetch accepts
only http and https to granted hosts, returns redirects and HTTP error
statuses as responses instead of following or failing them, and bounds
connect time, total time and response size. The file portal takes
read-only directories with --allow-read and refuses writes there.
tests/runtime/check_grants.py runs the real runtime on a probe component
under each kind of grant and checks the outcome of every operation; it
fails when read-only data is writable or an import ignores its grant.
The appd dispatch tests that read WEFT_RUNTIME_BIN now hold the env lock
shared with the tests that set it.
The browser-facing WebSocket accepted requests from any local page, so a
page could launch or terminate apps and inject messages into any session.
Every connection now completes the upgrade and sends HELLO within five
seconds. The system role authenticates with a token appd writes to
$XDG_RUNTIME_DIR/weft/appd.systoken (mode 0600) and may manage sessions,
but never sees or sends application messages. The app role authenticates
with a per-session bridge token passed to weft-app-shell in
WEFT_BRIDGE_TOKEN and exchanges APP_MESSAGE envelopes with its own
session only; it is disconnected when that session stops.
weft-servo-shell hands the system endpoint only to the system UI
document and denies navigation away from it. weft-app-shell installs the
bridge only in top-level documents inside the application's UI
directory and denies navigation outside it. The launcher renders package
names as text.
tests/frame/check_counter.py drives the Counter demo end to end through
the compositor, appd, the Wasm runtime and the app shell.
weft-appd marked a session Running and broadcast APP_READY as soon as the
runtime printed READY, then started weft-app-shell with its output
discarded and never watched it. A session with no UI, a UI host that
failed to start, or one that crashed was still reported running. Early
failure paths skipped parts of the cleanup: the image mount was not
unmounted, the compositor association was not released, and the abort
sender leaked on spawn failure. Readiness was any stdout line equal to
READY, which a Wasm guest (inherited stdout) or a page (window.alert
writes to the app shell's stdout) could print.
Now a session becomes Running only after the runtime and then the app
shell report readiness, each within 30 s; WEFT_APP_SHELL_BIN is required
and checked before anything starts. After that, the runtime exiting, the
app shell exiting or a terminate request ends the session. Every exit
path goes through one teardown that kills both children, releases the
compositor association, unmounts the image, stops the file portal and
broadcasts Stopped.
Readiness uses a per-session random token passed to both children in
WEFT_READY_TOKEN. weft-runtime and weft-app-shell print READY <token>,
and appd accepts only a line ending with it; guests receive only the
variables the runtime names and pages cannot read the host environment.
Child output is read in bounded, lossily decoded lines, so invalid UTF-8
or long lines no longer stop the readers. Compositor notifications use
try_send so a disconnected compositor cannot stall teardown, and all
children are killed if the supervisor is dropped.
The IPC relay is not yet owned by the session, and children are killed
without a graceful period.
Validation:
- cargo test -p weft-appd (43 passed); new tests cover: not ready until
the app shell is ready, runtime exit before that, no app shell
configured (runtime never starts), app shell exit ending a running
session with the runtime process gone, and READY lines without the
token; they fail against the previous supervisor
- end to end with weft-compositor (nested), weft-appd, weft-runtime
(wasmtime-runtime) and weft-app-shell: the Counter demo reaches
APP_READY and is on screen; a page that prints a READY line through
alert and never loads is stopped after 30 s without APP_READY;
terminate leaves no child processes
- clippy -D warnings and fmt --check for weft-appd and weft-runtime
When a systemd user manager was detected, weft-appd launched the runtime
with `systemd-run --user --scope --wait ...`. systemd rejects that
combination while parsing options ("--wait may not be combined with
--scope"), so every app launch on such a host exited immediately and the
session stopped without the runtime ever starting.
A scope already runs the command in the foreground: systemd-run registers
the scope and then executes the command in its own process, so stdout and
the exit status still reach appd without --wait.
The options now live in one constant. A Linux test runs the installed
systemd-run with them against an empty runtime directory and bus address
and requires it to reach the manager connection; option and combination
errors stop it earlier. The test does not validate the -p property values
or start a scope. It requires systemd-run to be installed.
This keeps the existing ownership scope for the runtime only; placing the
app shell and helpers under the same owner is separate lifecycle work.
Validation:
- systemd-run 255: the test passes with the new options and fails with
the old --wait (and with --pty) using systemd's own message
- cargo test -p weft-appd (38 passed), clippy -D warnings, fmt --check
The two consecutive registry lock acquisitions after READY left a window
where session state was Running but the AppReady broadcast had not yet
been sent. A concurrent reader winning the lock in that window would
observe Running without a corresponding event. Combining into one
acquisition closes the window.
- supervise: in the stdout-read-error-before-READY case, immediately
kill the child process, tear down the file portal, mark the session
Stopped, and return; previously the function fell through to child.wait()
leaving the session in Starting state with no guaranteed cleanup path
- supervise: restructure app_shell binding as a match expression so the
compiler can verify the initial None value is not silently discarded
- weft-runtime: replace Mutex::lock().unwrap() with unwrap_or_else in
the weft:app/ipc send, recv, and connect host functions so a poisoned
mutex does not panic inside the Wasmtime host-call context
Add weft-app-shell binary: takes <app_id> <session_id> args, connects to
zweft_shell_manager_v1 as an application window, resolves the app UI URL,
and runs a single Servo WebView in an isolated process. Prints READY to
stdout after the window is initialised so weft-appd can track the session
lifecycle.
weft-appd runtime.rs: after weft-runtime emits READY, spawn weft-app-shell
(WEFT_APP_SHELL_BIN env var) alongside it. The app shell is killed when the
session ends via abort or natural runtime exit.
weft-servo-shell: remove in-process app WebView management. The shell now
manages the system UI WebView only; all app rendering happens in dedicated
weft-app-shell processes.
The READY timeout branch killed the process and returned Ok(()) without
calling remove_abort_sender(), leaving a spent entry in abort_senders.
Add the call to keep it consistent with all other exit paths.
Previously the supervisor returned Ok(()) without updating the session
state, leaving it in Starting indefinitely. Now it sets state to
Stopped, removes the abort sender, and broadcasts AppState:stopped
before returning, consistent with every other early-exit path.
When a runtime process exits naturally (not via TerminateApp) the
oneshot Sender remained in abort_senders until shutdown_all().
Add remove_abort_sender() and call it at the normal exit path in
supervise() to release the entry immediately.
Runtime spawn failure, READY timeout, and abort-during-startup all
previously returned without killing the portal child process or removing
its socket file. Extracted kill_portal() to consolidate cleanup and
call it in each of those three paths.
spawn_file_portal() checks WEFT_FILE_PORTAL_BIN; if set, creates a
per-session Unix socket at /weft/portal-<id>.sock and
spawns weft-file-portal with --allow args derived from the same host
paths that resolve_preopens() produces for the session.
WEFT_FILE_PORTAL_SOCKET is added to the runtime environment so
weft-runtime (and the WASM app) can locate the socket.
The portal process is killed and the socket cleaned up at the end of
supervise(), after the runtime exits and the mount is torn down.
When WEFT_FILE_PORTAL_BIN is unset the behaviour is unchanged.
Add crates/weft-appd/src/mount.rs with MountOrchestrator.
On each app launch, supervise() calls MountOrchestrator::mount_if_needed
which looks for <app_id>.app.img, <app_id>.hash, and <app_id>.roothash
in the app store roots. If all three exist and weft-mount-helper is
available (WEFT_MOUNT_HELPER env or /usr/lib/weft/ default paths), it:
- creates /tmp/weft-mnt-<session_id>/<app_id>/
- invokes weft-mount-helper mount to set up dm-verity and EROFS mount
- sets WEFT_APP_STORE=/tmp/weft-mnt-<session_id> in the child env so
the runtime resolves the package from the mounted read-only image
After the process exits, umount() invokes weft-mount-helper umount and
removes the temporary directory.
Falls back to directory-based install silently if no image is found,
mount-helper is absent, or the mount fails.
Test: find_image_returns_none_when_absent.
supervise() checks /systemd/private to detect an active
user systemd session. When present (and WEFT_DISABLE_CGROUP is unset),
the runtime binary is launched via:
systemd-run --user --scope --wait --collect --slice=weft-apps.slice -p CPUQuota=200% -p MemoryMax=512M -- <bin> ...
This places each app in a transient weft-apps.slice scope with default
resource limits from the blueprint. The --wait flag keeps systemd-run
alive so child.wait()/child.kill() remain correct.
When no user session is present the command is built directly as before.
WEFT_DISABLE_CGROUP=1 bypasses the wrapping unconditionally.
Add capabilities field to weft-pack PackageMeta (optional Vec<String>).
Print cap: lines in weft-pack info output when capabilities are declared.
In weft-appd:
- Make app_store_roots pub(crate) so runtime.rs can use it.
- Add resolve_preopens(app_id) in runtime.rs: reads wapp.toml from the
package store, extracts capabilities, maps each to a (host, guest) pair:
fs:rw:app-data / fs:read:app-data -> ~/.local/share/weft/apps/<id>/data :: /data
fs:rw:xdg-documents / fs:read:xdg-documents -> ~/Documents :: /xdg/documents
Unknown capabilities are logged at debug level and skipped.
- supervise() calls resolve_preopens() and appends --preopen HOST::GUEST
flags before spawning the runtime binary.
weft-runtime now parses optional flags after <app_id> <session_id>:
--preopen HOST::GUEST pre-opens a host directory at GUEST path in the
WASI filesystem (HOST::GUEST or HOST for same path)
--ipc-socket PATH sets WEFT_IPC_SOCKET env var inside the component
wasmtime-runtime path applies preopened dirs via cap_std and WasiCtxBuilder,
and injects WEFT_IPC_SOCKET when --ipc-socket is present. Stub path ignores
both flags.
weft-appd: SessionRegistry gains ipc_socket field (set to the appd Unix
socket path in run()), extracted alongside compositor_tx in dispatch(), and
forwarded to supervise() as ipc_socket_path. supervise() passes
--ipc-socket <path> to the spawned runtime when present.
cap-std added as optional dep under wasmtime-runtime feature.
Adds crates/weft-appd/src/compositor_client.rs: async Tokio client that connects to the
compositor's Unix socket (/weft/compositor.sock or WEFT_COMPOSITOR_SOCKET),
retrying every 2s on failure and 500ms on write error. Incoming CompositorToAppd frames are
decoded and logged (SurfaceReady, ClientDisconnected).
Wires compositor_tx into SessionRegistry. The supervise task now sends AppSurfaceCreated
(with child PID) immediately after process spawn, and AppSurfaceDestroyed when the process
exits. All three existing supervisor tests updated to pass None for compositor_tx.
Before this fix, TerminateApp sent while a process was waiting for its
READY signal was not acted on until the 30-second timeout fired.
abort_rx is now included in the tokio::select! that wraps wait_for_ready,
so the child is killed and AppState::Stopped broadcast as soon as the
abort is received, regardless of where in the startup sequence it fires.
test: supervisor_abort_during_startup_broadcasts_stopped
If WEFT_RUNTIME_BIN is set but the binary cannot be spawned (missing,
not executable, etc.), supervise() now transitions the session to
Stopped and broadcasts AppState::Stopped instead of returning an error
that left the session permanently stuck in Starting.
runtime.rs: the READY-timeout early-return path now broadcasts
AppState::Stopped before returning so WebSocket clients see the
session disappear when a module fails to signal readiness within 30s.
runtime.rs: after the child exits (natural exit or abort), supervise()
now broadcasts AppState { session_id, state: Stopped } in the same
lock scope as set_state. WebSocket clients receive the notification
without needing to poll QueryAppState or call TerminateApp.
wait_for_ready() now returns the BufReader<ChildStdout> with the READY
line already consumed. supervise() spawns drain_stdout() on that reader
so any subsequent module output is forwarded to the trace log and the
pipe buffer never fills up.
Without this, a long-running Wasm module that writes to stdout after
printing READY would eventually block waiting on a full pipe.
ipc.rs: AppReady { session_id, app_id: String }.
runtime.rs: supervise() passes app_id (already in scope as parameter)
when building the AppReady broadcast message.
main.rs: supervisor integration test updated to use .. to ignore
app_id in the AppReady pattern match.
SessionRegistry now tracks a oneshot abort sender per active session:
- abort_senders: HashMap<u64, oneshot::Sender<()>> field added.
- register_abort(session_id): creates the channel, stores the sender,
returns the receiver to the supervise task.
- terminate(): removes the session state AND drops the abort sender,
closing the channel and triggering the receiver in supervise.
runtime::supervise() now accepts abort_rx: oneshot::Receiver<()>:
- After the READY signal is received, the process-wait loop uses
tokio::select! on child.wait() vs abort_rx.
- On abort: logs intent, calls child.kill(), then sets state Stopped.
- On natural exit: logs exit status, sets state Stopped.
dispatch::LaunchApp: calls register_abort immediately after launch,
passes the receiver to the spawned supervise task.
Integration test updated to pass the abort receiver.
runtime.rs — process lifecycle manager:
- supervise(session_id, app_id, registry): spawns the weft-runtime child
process identified by WEFT_RUNTIME_BIN env var. If unset, logs debug
and returns immediately (no-op until runtime binary is available).
- Child process invoked as: <WEFT_RUNTIME_BIN> <app_id> <session_id>
with stdout/stderr piped, stdin closed.
- wait_for_ready(): reads stdout line-by-line; returns Ok(()) on first
line matching 'READY'; returns Err if stdout closes without it.
- 30-second READY_TIMEOUT via tokio::time::timeout; on expiry, kills
the child and transitions session to Stopped.
- On success: sets session state to Running, broadcasts AppReady to all
connected WebSocket clients via registry broadcast channel.
- drain_stderr(): async task that forwards child stderr lines to tracing
at WARN level for observability.
- On process exit: sets session state to Stopped regardless of exit code.
main.rs — wiring:
- SessionRegistry now owns broadcast::Sender<Response>; Default creates
the channel internally. Added set_state(), subscribe(), broadcast()
methods. Removed standalone broadcast_tx from run(); WS handlers
subscribe via registry.lock().await.subscribe().
- dispatch::LaunchApp spawns a tokio task calling runtime::supervise
immediately after creating the session. supervise is a no-op when
WEFT_RUNTIME_BIN is unset, so existing tests are unaffected.
Cargo.toml: added tokio 'process' and 'time' features.