Two new effects: ProbeGitRoot resolves the draft target's base repo via
/api/gitroot (seq-guarded, a miss unticks the toggle) and LoadBranches
fills the existing-branch picker from /api/branches, both applied back
to the tab that requested them. Branch loads triggered by a successful
probe are served inline so perform() stays recursion-free.
Worktree on by default over a git repository with Enter/Space to spawn
in the checkout itself; base repo inferred from picked projects or a
guarded /api/gitroot probe for raw paths (a miss unticks, web parity);
new-branch name tracking the generated session name until touched;
Left/Right switches to picking an existing branch from GET /api/branches,
with held branches skipped and rendered dimmed beside their holder.
No cleanup option: cleanup is unconditional daemon-side (#56).
The one that loses data: the project picker lists linked worktrees as
pickable rows, so `base_repo` was easily a worktree rather than the main
checkout, and the new placement then nested the tree inside it. Git hides
a registered worktree's contents from the outer `git status`, so the
outer tree read clean and removing it deleted the inner one's uncommitted
work — while that session could still be running. Placement now resolves
to the main checkout via `--git-common-dir`, and removal refuses outright
when `worktree list` reports anything registered underneath.
Also:
- The `worktree_root` override dropped the old UUID without replacing it,
so two repos spawning the same branch name targeted one directory and
the second failed. The repo basename now sits between root and slug.
- Project staleness read `.git/HEAD`'s mtime, which only moves on
checkout. A repo committed to daily from one long-lived branch reported
its last branch switch and got dimmed. Prefer `logs/HEAD`.
- The projects cache TTL (5s) was shorter than its own background refresh
(60s), so nearly every request rescanned all of $HOME under the scanner
lock — and draft panes re-request on unrelated re-renders. Now 180s.
- onSuccess resolved the leaf against the active tab only, so switching
tabs mid-POST orphaned the session. Searches every tab.
Ten findings from reviewing the accumulated blitz/spawn-v2 diff. The two
that could wedge a repo:
- The cleanup guard refused any worktree whose commits were on no other
ref, which is every ordinary unpushed branch. Paths are deterministic
now, so the leftover tree blocked the next spawn and each branch became
a one-shot. `worktree remove` keeps refs/heads/<branch>, so committed
work is never at risk: guard on an unclean working tree alone.
- A worktree whose directory had vanished failed the git probe and was
reported as unsaved work, stranding the registration forever. Prune it
instead, and prune before `worktree add` tests the target path.
Also:
- The CLI hardcoded create_branch: true, so the daemon's existing-branch
path was unreachable. Added --existing-branch.
- Pastes were forwarded without their DECSET 2004 markers, so a remote
shell ran a multi-line paste line by line.
- spawnForm emitted before its seed was restored and after destroy(), so
a rebuilt draft blanked its cwd and a dead form clobbered the live one.
- showRawCwd and the probe's worktree untick never reached the leaf.
- A late autocomplete fetch reopened the popup under a blurred input.
- The "remove worktree when killed" checkbox promised something the
daemon no longer honours; removed.
forward_to_session dropped Paste(_) on the floor, so pastes into a
session vanished once terminal.rs enabled bracketed paste. Split the
arm: Paste forwards the payload verbatim (the remote readline
reassembles it), only Mouse events have no wire encoding.
Action::OpenSpawnModal becomes NewDraftPane: pressing s turns the empty
pane into an inline draft form (or opens a new tab when the pane shows a
session). The form is a pane-level widget with its own field focus; a
successful submit replaces the draft with the new session in place,
matching the web client's draft panes. SpawnModal and OverlayKind::Spawn
are removed; kill-confirm and help overlays stay.
The umbrella `maestro` binary had zero coverage of its feature-gate
combinations (cli / daemon / registry / tui). Add:
- Unit tests in `main.rs` for `is_cli_subcommand` classification (CLI vs
server subcommands, plus the always-false non-`cli` stub) and the
`-v` verbosity -> tracing-directive mapping.
- A `verbosity_directive` helper, extracted from `cli_filter` purely to
make the verbosity mapping testable without touching the process-global
`RUST_LOG` (behavior unchanged).
- An integration test (`tests/cli.rs`) driving the real built binary via
`CARGO_BIN_EXE_maestro`, asserting the `--help`/`--version`/subcommand
dispatch surface. Per-subcommand checks are `#[cfg]`-gated so the suite
is meaningful under every feature subset.
- A CI feature-gate matrix step: build + test each of cli/daemon/registry/
tui in isolation, and assert the no-feature build is rejected.
All four subsets and the default build pass fmt, clippy, and tests.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The subtree search returned early whenever it found *any* group member,
so an in-tree stage that was busy or sleeping on a timer masked a
reparented member genuinely parked on the tty.
Checking every member instead would have meant a /proc scan per session
per tick whenever nothing was parked — the common case. ProcGroups
snapshots /proc once per activity pass and indexes by process group, so
the cost is one walk however many sessions the daemon owns, and the
subtree walk disappears: the index is a superset of it.
Also fixes the background-helper test, which had stopped discriminating.
It polled until awaiting_input was true and caught a moment before the
helper spawned, so it passed against a deepest-descendant walk too. It now
waits for the helper to exist first. Both regression tests were re-checked
by patching each old implementation back in.
Reported in review on #51.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
A member of the foreground group can sit outside the pane's subtree: a
stage that forks a terminal-reading child and exits leaves that child
reparented to init while it keeps the foreground pgid. The subtree search
finds nothing and the session reads as idle.
An empty subtree result now falls back to scanning /proc for the group.
The scan only runs when the cheap path found nothing, so the common case
still costs a handful of reads.
Tested at the level it can honestly be tested: scan_for_pgid against this
process's own group, and against a group that cannot exist. Reproducing
the reparenting scenario end to end is not practical — setpgid across
sessions is EPERM, and from a shell the window closes as soon as the shell
reclaims the terminal.
Reported in review on #51.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
tpgid names a process *group*, not a process. Its leader can exit while
another member still holds the terminal — a foreground pipeline whose
first stage finished — and then /proc/<tpgid> does not exist, so reading
it directly reported the session idle while it sat waiting on the user.
awaiting_input now searches the pane's subtree for live processes whose
pgrp matches, and asks whether any is parked on the tty. Members stay in
that subtree because their parent shell is alive, so they never reparent.
Reported in review on #51. The regression test needs an interactive shell
— job control is what puts a pipeline in its own process group — and waits
for the leader to actually be gone before asserting, otherwise it passes
against the old lookup while the shell is still foreground.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
A plain deadline loop says the same thing without catching a panic to
turn it back into an assert.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
awaiting_input walked `children` to the deepest descendant, guessing that
the newest child is the foreground one. An agent that starts a background
helper and then blocks on read landed on the helper, so a session genuinely
waiting on the user was reported as idle.
tpgid in /proc/<pid>/stat is the kernel's own answer — the foreground
process group of the controlling terminal — and the group leader's pid is
the group id. One field read replaces a 64-deep walk.
Adds tests/activity.rs: the first end-to-end coverage of this pipeline,
which had none. The background-helper case fails against the old walk and
passes against tpgid.
Also makes the attach handler subscribe before upgrading. Checking
existence and then subscribing inside the callback left a window for the
liveness watcher to reap in between, which attach::serve compensated for
with a clean close; acquiring the subscription up front means the 404 and
the subscription are one step and the window does not exist.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
SPA attach had no reconnect at all. Any dropped frame closed the socket,
`onClosed` fired, and the pane was destroyed permanently — the TUI
reconnects forever on the same endpoint. Session death also reached the UI
on two uncoordinated channels: the row from the events socket, the pane
from `ws.onclose`, with nothing reconciling them, so either could happen
without the other.
The catalog is now the authority. `shouldRetryClose` decides from session
status: Running means the link dropped, so retry with 500ms-to-10s
backoff; Dead means the PTY stream was sealed and nothing more is coming;
absent means it is gone. Application closes (4000+, a relayed daemon
rejection) and clean closes are never retried. The shell also closes any
pane whose session has left the catalog. The policy lives in reconnect.ts
so it is tested without a DOM.
kill_session checked `has()` before `kill()`, which only opened a window
for the liveness watcher to reap in between — turning a 404 into a 502.
`kill` reports the miss itself.
with_spawn_history logged and carried on when `Arc::get_mut` failed,
producing a daemon whose recent-dirs picker is permanently empty with
nothing to show for it. It panics now: that is a construction-order
mistake, at startup, in code that runs once.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
promote() early-returned for anything that was not a Click, leaving the
pending click intact — so click, drag, click at the same cell within
400ms was reported as a double-click. Same for a scroll or a keystroke
between the two.
Any non-click input now clears the pending click. Two clicks with
something in between are two clicks, however fast.
Reported in review on #48.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The keybinding parser validated strings, not whether any producer could
emit the event — so the shipped default `mouse:dblclick(sidebar-row)`
could never fire, and nothing failed.
`MouseKind::DblClick` appeared only in the enum and in parse/format;
`canonicalise` emits Down/Drag/ScrollUp/ScrollDown. Adds a `ClickTracker`
that promotes a second click at the same cell within 400ms, keeping
`canonicalise` a pure function of one event — the tracker takes `now` as a
parameter, so the timing rule is tested without sleeping.
`Region::Sidebar` and `Region::SplitDivider` are deleted: `region_at` can
only return Modal, Tab, SidebarRow, Pane or Unknown, so the parser was
accepting names the renderer cannot produce. It now rejects them.
Shift-Tab is wired to `overlays::prev_field`, which existed with no
caller, leaving the spawn modal cycleable in one direction only. crossterm
reports it as `KeyCode::BackTab` rather than a modified Tab, so `InputKey`
gains a `BackTab` variant that also forwards `\x1b[Z` to the pane.
Left alone: `Action::FocusPane(Direction)` discards its direction, but
that is v1 having one pane per tab, not a defect.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
There were three answers to "how do I test against a daemon", and the CLI
— the thinnest client — had none, so its formatting and error handling
were exercised by compiling and forking the real binary plus tmux.
Adds crates/maestro-testing with a MockDaemon that serves the real /api/*
surface from maestro-protocol types, backed by a Vec<Session>. It replaces
two ad-hoc stubs: maestro-web's, which returned [{"id":"session-1"}] —
not a valid Session — and maestro-tui's scripted socket with a hardcoded
timeline.
maestro-cli gets its first tests against a daemon: six cases covering the
catalog, that DaemonInfo carries the version and presets its old private
copy dropped, that a structured failure surfaces the daemon's own message,
and that connection-refused stays distinguishable from an HTTP rejection.
maestro-web's proxy suite gains an end-to-end check that an upstream
rejection reaches the browser as a 4502 close carrying the daemon's
message — upstream_close_reason had only unit tests.
e2e keeps the tmux-shaped tests. cli_list_reads_from_daemon still runs the
shipped binary as a subprocess but now against MockDaemon, and
tui_integration.rs is deleted: the real daemon's /api/events contract is
covered by maestro-daemon/tests/local_api.rs and the client reducer by
maestro-tui/tests/catalog_ws.rs. The e2e suite drops from ~32s to ~20s.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
maestro-cli/src/config.rs was two pass-throughs, a newtype error, a
two-field projection of MaestroConfigFile, and one four-line or_else with
a single caller. `resolve_daemon_url` moves onto MaestroConfigFile, next
to the schema it reads, and the module goes; its tests move with it.
`[web].dev_proxy_target` is deleted end to end — schema, env var,
normalisation, three tests, README line, and the startup warning that
admitted the server ignores it.
MaestroConfigFile::save had no callers outside its own tests. Deleting it
also removes every skip_serializing_if, the five is_empty impls, and the
Serialize derives that existed only to make its output tidy.
BroadcastTuning was a struct that was never instantiated: all three uses
were `::DEFAULT.field` in const position and there is no [broadcast]
section. Now two plain consts.
$MAESTRO_CONFIG was honoured twice, by resolve_config_path and by clap's
`env =`. Every binary already gets it through clap, so resolve_config_path
and CONFIG_PATH_ENV go and the TUI falls back to default_config_path like
everything else.
Also collapses three layers of 1:1 forwarding: the umbrella's cli_dispatch
shim, and four near-identical 8-line functions in cli/dispatch.rs that
differed only in which run_* they called.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Each was a typed field that type-checked on both sides and carried
nothing across.
SessionStatus::Dead was never constructed by the daemon, yet the TUI, CLI
and SPA all render it and DESIGN.md specifies the exited badge. Sessions
were removed on death, never marked. They are now marked Dead and kept
until the operator dismisses them, with the PtyBridge sealed rather than
dropped: attached clients still get a Close frame (NAP-226), and the
replay buffer survives so a later attach can read what a crashed agent
printed. Previously that output was destroyed. Dead sessions are
read-only and do not survive a daemon restart.
CreateSessionRequest.env was accepted, stored, echoed back and dropped —
new_session_piped had no env parameter. It now reaches tmux as sorted
`-e KEY=VAL` flags. A key containing `=` or NUL is rejected, since tmux
would parse it as a different assignment than the caller wrote.
PROTOCOL_VERSION was written by the daemon and read by nobody. Adds
`maestro_protocol::is_compatible`, compared on the major component only;
`maestro list` shows the daemon version and the TUI warns on mismatch.
Also stops mapping every non-404 session error to 502: an invalid env key
is a client mistake and now answers 400.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
A background task, a second broadcast channel and a whole module existed
to remap SessionEvent onto CatalogEvent one variant at a time.
SessionManager now emits CatalogEvent directly.
Deletes local_events.rs, the publisher task, the LocalEventBus channel,
enum SessionEvent, LocalApiState::events, and the duplicate capacity
constant — events_capacity was declared twice for one logical stream.
Lag semantics improve. There were two lag points: the publisher falling
behind the manager, which broadcast one Lagged to every client, and each
WS subscriber falling behind the bus. Only the per-subscriber one is left,
which pump_events already handled — a slow client now makes only itself
resync.
Adds the first daemon-level tests for WS /api/events: the Snapshot-then-
deltas contract, and that a late subscriber gets a Snapshot of current
state. Both were described only in a comment.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
dispatch was `async fn -> ()` with reqwest calls and task spawns inlined,
so it could not report intent and the spawn, kill and attach arms had no
test. It now returns Vec<Effect> and App::run performs them.
Two bugs the old shape produced, both fixed and covered:
- The toast overlay could not be dismissed. handle_overlay_input returned
Pass for Toast(_) and nothing cleared it, so "spawn failed: ..." stayed
until another overlay replaced it. It now clears and still returns Pass,
so the keystroke does its normal job instead of being eaten.
- Closing a tab leaked its emulator. Only submit_kill released one, so a
closed tab's WebSocket task and vt100 parser outlived it forever.
close_tab now reports the session it dropped — and holds onto it when
another tab still shows the same one.
AppState::overlay is private behind show_overlay/dismiss_overlay/toast.
Four sites poked the field directly; that is why the toast bug existed.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
SessionManager held a concrete Tmux, so its decisions — spawn rollback
ordering, liveness diffing, activity change-detection — could only be
reached by forking a real tmux server. Inner now holds an Arc<dyn Panes>;
Tmux is the production adapter, FakePanes the one tests drive.
Arc<dyn Panes> rather than a type parameter: a generic would spread
through LocalApiState, attach and cli, and one vtable hop is free next to
a fork+exec.
Adds tests/session_decisions.rs — six cases, no subprocess, 0.24s.
Replaces the chmod-0o500 rollback test, which detected root and returned
early, so on CI it asserted nothing; the failure is now forced by a
regular file sitting where the sidecar directory's parent must be, which
no uid can work around. It also asserts the rollback *order*, not just
the end state.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Killing the last session shuts the server down, and tmux >= 3.7 reports
that to the still-connected client with a phrasing `is_no_server_error`
did not match. `list_sessions` then returned Err, and `run_liveness_pass`
logs a warning and returns — so the pass reaped nothing.
Silent by construction: nothing asserted the phrasing set, and the only
symptom was an integration test that looked flaky. Both now covered.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Restores the harness's deliberate 5s per-request budget, which the shared
client's 30s default had silently replaced, and converts the last two
hand-rolled daemon calls in e2e.rs (recent-dirs, gitroot). Drops reqwest
from maestro-e2e entirely.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Collapsing the health probe and the app client into one lost both
policies: the probe's 500ms was a total request timeout, not a connect
timeout, and the app's calls had 5s rather than the client default 30s.
A daemon that accepts TCP and then stalls held the terminal for 30s.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The CLI, TUI and end-to-end harness each carried their own client: three
reqwest builders with three timeout policies, three copies of the
success-or-body check, and route strings at every call site. Two of them
discarded the daemon's structured error and reported only a status code.
Deletes `maestro-cli/src/http.rs`, both `ws_base` copies in the TUI, the
CLI's partial `DaemonInfo` (it dropped `version` and `presets`), and the
e2e harness's `check_ok`. `maestro-web` gains a protocol dependency and
stops building the error envelope with a format string and reading it back
as untyped JSON; the daemon and the proxy now register routes from the same
constants.
`maestro_web::router` returns `Result` — an unusable daemon URL fails at
startup instead of at first connect.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The crate held the wire types but not the wire surface, so every consumer
re-derived routes, URL construction and the error envelope.
Adds `routes` (path constants, axum patterns, and a single fallible
`ws_base`), the `ApiError` envelope, `DaemonInfo`, and a `DaemonClient`
behind an optional `client` feature. `SpawnPreset` moves here because it
rides `GET /api/info`; `maestro-config` re-exports it so the on-disk
schema and the payload stay one type.
`ws_base` now rejects a base with no recognised scheme. The five previous
copies disagreed — the CLI errored, the web proxy trusted it, the TUI
passed it through.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Miguel Palhas
naps62-yolo