From 43e69d345eeef851d4a5e68512443c4a4d6190aa Mon Sep 17 00:00:00 2001 From: atlas Date: Sat, 10 Oct 2026 19:52:25 +0200 Subject: [PATCH] subagents, ACP: stop with SIGINT only, so claude ends its own tool commands claude's Bash tool starts every command in a new session, so a signal to claude's process group never reaches it. claude stops those commands itself when SIGINT makes it shut down. A SIGKILL leaves them running, re-parented to init. Per the ruling on #4999, every stop is SIGINT and nothing escalates. - Subagent `interrupt` always sends SIGINT. The `force` argument is gone from the MCP tool and from `session::interrupt`. A caller that still sends `force: true` gets SIGINT; the argument is ignored. - `interrupt` used to free the session name before it signalled, so a `continue` could start a second claude on the session while the first was still exiting. The `running` map now holds a three-state `Slot` (starting, running, stopping). An interrupted claude turn is `Stopping` until the turn task's own `finish_turn` clears it. `status` says "stopping", and `start`/`continue` on that name are refused. `interrupt` still returns at once. The ACP arm is unchanged: its name is freed when `session/cancel` is sent. - ACP `Connection` drop sends its process group SIGINT, not SIGKILL, and the spawn no longer sets tokio's `kill_on_drop`, which SIGKILLed the leader after the `Drop` body. The test fixture's background grandchild resets SIGINT to default, since `&` starts it ignored. Tests: a fake claude that records SIGINT, interrupted, fails with `cancel(true)` (the old `force: true` path). An interrupted fake claude that keeps running keeps its name claimed and `continue` is refused, which fails when the slot is freed before the signal. A fake ACP agent that records SIGINT, dropped, fails with SIGKILL in `Drop`, and also with SIGINT in `Drop` but `kill_on_drop(true)` left in place. Not changed: hive-claude's idle watchdog (still SIGKILL, separate hive-claude change), the units' `KillSignal`, `machinectl kill`, and the login helper. --- .../base/skills/claude-subagents/SKILL.md | 5 +- docs/tools/subagent.md | 22 +- hive-runtime/README.md | 10 +- hive-runtime/src/acp/mod.rs | 46 +- hive-runtime/src/acp/rpc.rs | 10 +- hive-subagent-mcp/src/mcp.rs | 12 +- hive-subagent-mcp/src/session.rs | 405 ++++++++++++------ 7 files changed, 340 insertions(+), 170 deletions(-) diff --git a/claude-plugins/plugins/base/skills/claude-subagents/SKILL.md b/claude-plugins/plugins/base/skills/claude-subagents/SKILL.md index dc53f709..d7f37b3d 100644 --- a/claude-plugins/plugins/base/skills/claude-subagents/SKILL.md +++ b/claude-plugins/plugins/base/skills/claude-subagents/SKILL.md @@ -27,7 +27,7 @@ skill's recipe for you: start(name, prompt_file, model?, effort?, trigger?) continue(name, prompt, model?, effort?) status(name) -interrupt(name, force?) +interrupt(name) ``` `start` and `continue` return once the turn is under way, not once it @@ -42,7 +42,8 @@ reports how long since that turn last produced output, so a few seconds means it's working and an age climbing into the minutes means it's wedged and worth an `interrupt`. Reach for `continue` only once you actually have a new instruction for it, since that spends a turn. -`interrupt` genuinely stops a running turn (`force: true` for SIGKILL). +`interrupt` genuinely stops a running turn, with SIGINT so claude stops +the commands it started; `status` says `stopping` until claude has exited. Model choice, prompt hygiene, splitting big batches, verify-then-report — everything else in this skill — applies exactly the same whether you're diff --git a/docs/tools/subagent.md b/docs/tools/subagent.md index 0d59ec8c..b4c76da9 100644 --- a/docs/tools/subagent.md +++ b/docs/tools/subagent.md @@ -166,10 +166,15 @@ from how the child exited. `continue` clears it and hands the session a fresh turn allowance, which is what makes an interrupt a pause you can undo rather than a session you have to abandon. -On `claude`, the signal goes to claude's whole process group, so the -commands the subagent started (a `nix build`, a test run) get it too. The -signal doesn't reach a command that moved itself out of the group -(`setsid`). +On `claude`, the signal is always SIGINT, and it goes to claude's whole +process group. The signal doesn't reach a command that moved itself out of +the group (`setsid`), and claude's Bash tool starts every command that way: +claude stops those commands itself when it shuts down on SIGINT. A SIGKILL +would leave them running, so nothing escalates to one. `interrupt` returns +at once, and the name stays claimed until claude has exited: `status` says +**stopping**, and the daemon refuses `start` and `continue` on that name +until then. A claude that ignores the SIGINT stays stopping until someone kills +it by hand or the unit stops. When the session knows where its report goes — `start`'s `report_file`, or the path the subagent names when it signals — the daemon appends the @@ -345,7 +350,7 @@ page applies as written. On `acp`: - each run spawns its own copy of the parent's ACP agent. The end of the - run kills it and every command it started. The unit loads + run sends SIGINT to it and every command in its process group. The unit loads `backendEnvironmentFile` for it. On the `opencode` preset, the daemon reads a provider key that file leaves unset from the swarm secret store at `swarm/agents//acp-provider`, as the harness does. @@ -354,9 +359,10 @@ On `acp`: agent process, and a `continue` with none recorded fails straight away. - a role goes in front of the session's first prompt, since ACP has no system prompt. -- `interrupt` sends `session/cancel`, and `force` changes nothing. The - runtime kills an agent that ignores the cancel for 10s, with every - command it started. +- `interrupt` sends `session/cancel`. The runtime sends SIGINT to the + process group of an agent that ignores the cancel for 10s, and when the + run ends. Nothing escalates past that SIGINT. The name is free as soon as + `interrupt` sends the cancel. - the runtime sets `model` and `effort` on the session when the agent offers that value for its `model` or `thought_level` config option. Any other value does nothing (`hive_runtime::AcpRuntime`). diff --git a/hive-runtime/README.md b/hive-runtime/README.md index 11183df4..6b6d4025 100644 --- a/hive-runtime/README.md +++ b/hive-runtime/README.md @@ -35,13 +35,15 @@ The crate depends on no hyperhive binary crate, so `hive-agent` and A turn is stopped with `session/cancel`: by the `Canceller` handle, or by the idle watchdog once no `session/update` has arrived for `Config::idle_timeout`. An agent that has not answered the prompt 10s later -is killed, and the next turn respawns it. A watchdog stop fails the turn -with `AcpError::IdleTimeout` (`IdleKilled` if the agent was killed). This is +is sent SIGINT, and the next turn respawns it. A watchdog stop fails the turn +with `AcpError::IdleTimeout` (`IdleKilled` if the agent was sent SIGINT). This is the only way a provider error the agent retries without reporting it, such as an HTTP 429, ends the turn. -The agent leads its own process group. Killing it, and dropping the runtime, -`SIGKILL`s that group, so the commands the agent started die with it. +The agent leads its own process group. Stopping it, and dropping the runtime, +sends that group SIGINT, so the commands the agent started get it too. +Nothing escalates to SIGKILL: an agent that ignores SIGINT keeps running, +and the next turn starts a new one beside it. ## ACP backend: compaction diff --git a/hive-runtime/src/acp/mod.rs b/hive-runtime/src/acp/mod.rs index b93be539..64afa350 100644 --- a/hive-runtime/src/acp/mod.rs +++ b/hive-runtime/src/acp/mod.rs @@ -39,7 +39,7 @@ const SETTLE_QUIET: Duration = Duration::from_millis(500); const SETTLE_MAX: Duration = Duration::from_secs(5); /// How long the agent has to end its turn after `session/cancel` before it is -/// killed. +/// sent SIGINT. const CANCEL_GRACE: Duration = Duration::from_secs(10); /// How long a `compact` command may send nothing before it is cancelled and @@ -120,10 +120,12 @@ pub enum AcpError { EmptyEndTurn, #[error( "ACP agent sent nothing for {silent_secs}s and did not stop within {grace_secs}s of \ - `session/cancel`, so it was killed" + `session/cancel`, so it was sent SIGINT" )] IdleKilled { silent_secs: u64, grace_secs: u64 }, - #[error("ACP agent did not stop within {grace_secs}s of `session/cancel`, so it was killed")] + #[error( + "ACP agent did not stop within {grace_secs}s of `session/cancel`, so it was sent SIGINT" + )] CancelIgnored { grace_secs: u64 }, } @@ -141,8 +143,8 @@ struct CancelState { impl Canceller { /// Ask the turn in flight to stop. The agent is sent `session/cancel`, /// and the turn then ends normally with what it produced so far; an - /// agent that has not ended it within `CANCEL_GRACE` is killed, and the - /// turn fails with [`AcpError::CancelIgnored`]. Returns whether a turn + /// agent that has not ended it within `CANCEL_GRACE` is sent SIGINT, and + /// the turn fails with [`AcpError::CancelIgnored`]. Returns whether a turn /// was in flight; with none, this does nothing. #[must_use] pub fn cancel(&self) -> bool { @@ -1062,9 +1064,14 @@ mod tests { const AGENT: &str = r#" n=0 prompt= model=m/think effort=low printf 'start\n' >> "$1.methods" +# With TRAP_INT set: SIGINT creates .int and exits 0. +if [ -n "$TRAP_INT" ]; then + trap ': > "$1.int"; exit 0' INT +fi # With GRANDCHILD set: a background `sleep`, its pid in .grandchild. +# `&` starts it with SIGINT ignored; `env` puts the default back. if [ -n "$GRANDCHILD" ]; then - sh -c 'echo $$ >"$1.grandchild"; exec sleep 300' sh "$1" >/dev/null 2>&1 & + sh -c 'echo $$ >"$1.grandchild"; exec env --default-signal=INT sleep 300' sh "$1" >/dev/null 2>&1 & fi update() { printf '{"jsonrpc":"2.0","method":"session/update","params":{"sessionId":"%s","update":%s}}\n' "$1" "$2" @@ -1529,6 +1536,33 @@ done assert_gone(grandchild).await; } + /// SIGINT rather than SIGKILL: an agent only stops the commands it + /// started itself if it gets to shut down. + #[tokio::test] + async fn dropping_the_runtime_lets_its_agent_shut_down_on_sigint() { + let dir = tempfile::tempdir().unwrap(); + let runtime = agent( + dir.path(), + "ok", + &[("TRAP_INT", "1")], + PercentPolicy::default(), + ); + runtime + .run(&config(dir.path()), "one", &NoopSink) + .await + .unwrap(); + + drop(runtime); + let marker = dir.path().join("prompts.int"); + for _ in 0..250 { + if marker.exists() { + break; + } + tokio::time::sleep(Duration::from_millis(20)).await; + } + assert!(marker.exists(), "the agent never got SIGINT"); + } + fn sent(session: &str, text: &str) -> (String, String) { (session.to_owned(), text.to_owned()) } diff --git a/hive-runtime/src/acp/rpc.rs b/hive-runtime/src/acp/rpc.rs index 65681f02..6d07d054 100644 --- a/hive-runtime/src/acp/rpc.rs +++ b/hive-runtime/src/acp/rpc.rs @@ -31,9 +31,10 @@ type Reply = Result; type Pending = Arc)>>>; /// The agent runs as the leader of its own process group; dropping the -/// connection `SIGKILL`s that group, so the commands the agent started die -/// with it. A descendant that left the group (`setsid`, `setpgid`) is not -/// reached. +/// connection sends that group SIGINT, so the agent can stop the commands +/// it started, and they get the signal too. Nothing escalates: an agent +/// that ignores SIGINT keeps running. A descendant that left the group +/// (`setsid`, `setpgid`) is not reached. pub(super) struct Connection { child: Child, /// The agent's pid, which is its group's id. Read at spawn: a reaped @@ -66,7 +67,6 @@ impl Connection { .stdout(Stdio::piped()) .stderr(Stdio::piped()) .process_group(0) - .kill_on_drop(true) .spawn() .map_err(|source| AcpError::Spawn { program: command.command.clone(), @@ -182,7 +182,7 @@ impl Drop for Connection { // SAFETY: plain `kill(2)`; `-pgid` targets the agent's own // process group, which `process_group(0)` made it lead. unsafe { - libc::kill(-pgid, libc::SIGKILL); + libc::kill(-pgid, libc::SIGINT); } } } diff --git a/hive-subagent-mcp/src/mcp.rs b/hive-subagent-mcp/src/mcp.rs index 8035dba1..6d24e0c5 100644 --- a/hive-subagent-mcp/src/mcp.rs +++ b/hive-subagent-mcp/src/mcp.rs @@ -187,10 +187,6 @@ struct StatusArgs { struct InterruptArgs { /// The running session's name to signal. name: String, - /// `true` sends SIGKILL immediately; `false` (default) sends SIGINT, - /// letting claude shut down cleanly if it's already mid-response. - #[serde(default)] - force: bool, } #[derive(Clone)] @@ -276,11 +272,13 @@ impl SubagentMcp { re-prompted toward it afterwards, and `continue` is what restarts it. Only works \ once it's actually running — a `start`/`continue` still in its brief window before \ the process is confirmed spawned refuses interrupt too (nothing to signal yet; \ - retry shortly), same as a name with nothing tracked at all. `force: true` for \ - SIGKILL, otherwise SIGINT." + retry shortly), same as a name with nothing tracked at all. On claude it sends \ + SIGINT, which claude answers by stopping the commands it started and exiting; \ + nothing escalates past it. Returns at once; the name stays claimed until claude \ + has exited, and `status` says `stopping` until then." )] fn interrupt(&self, Parameters(args): Parameters) -> String { - match session::interrupt(&self.state, &args.name, args.force) { + match session::interrupt(&self.state, &args.name) { Ok(msg) => msg, Err(e) => format!("interrupt error: {e:#}"), } diff --git a/hive-subagent-mcp/src/session.rs b/hive-subagent-mcp/src/session.rs index 8411e45a..4a276e82 100644 --- a/hive-subagent-mcp/src/session.rs +++ b/hive-subagent-mcp/src/session.rs @@ -5,8 +5,8 @@ //! continuation loop records. //! //! **No task files, no restart recovery.** The daemon's only state is an -//! in-memory `name -> Option` map (see `State`'s own doc for the -//! rest of the maps and for what the `None`/`Some` split is for) — all of +//! in-memory `name -> Slot` map (see `State`'s own doc for the rest of the +//! maps, and `Slot`'s for what each state of a claimed name means) — all of //! it living for exactly as long as the process is. A daemon restart means //! whatever was running gets killed //! with it (`tokio`'s own child-process drop semantics), not adopted, and @@ -230,11 +230,33 @@ enum SubagentRuntime { enum Stopper { /// Signals the turn's `claude` process. Claude(Cancel), - /// Sends the ACP agent `session/cancel`; one that ignores it is killed - /// by the runtime after its own grace period. + /// Sends the ACP agent `session/cancel`; the runtime sends SIGINT to one + /// that ignores it for its own grace period. Acp(Canceller), } +/// What `running` holds for a claimed name. +enum Slot { + /// Claimed, with no process confirmed under it: a `start`/`continue` + /// that hasn't spawned yet, or a goal run between turns. + Starting, + /// A turn is in flight, and this is how `interrupt` reaches it. + Running(Stopper), + /// `interrupt` has signalled a claude turn and its process hasn't exited + /// yet. Nothing escalates past that SIGINT, so a claude that ignores it + /// stays here until it exits; the turn's own `finish_turn` frees the + /// name. + Stopping, +} + +/// Which [`Slot`] a claimed name is in, without its cancel handle. +#[derive(Debug, Clone, Copy, PartialEq, Eq)] +enum Occupancy { + Starting, + Running, + Stopping, +} + /// What settled a resumed turn's bounded wait — the one thing `continue` /// blocks on before it answers. enum ResumeVerdict { @@ -333,13 +355,13 @@ fn searched_location(config: &Config) -> Option { /// and where to push the completion todo. `Arc`-wrapped so the background /// task driving a run outlives the tool call that started it. /// -/// The map value is `Option`: `None` means `name` is reserved for -/// an in-flight `start`/`continue` that hasn't reached a confirmed -/// spawn yet; `Some(stopper)` means a real turn is tracked and -/// interruptible. The `None` state exists to close a real TOCTOU window a -/// reviewer caught in the original check-then-insert version: checking "is -/// `name` free" and committing to it are two different lock acquisitions -/// unless the check *is* the reservation — see `reserve`. +/// The map value is a `Slot`: `Starting` means `name` is reserved for an +/// in-flight `start`/`continue` that hasn't reached a confirmed spawn yet; +/// `Running` means a real turn is tracked and interruptible; `Stopping` +/// means it was interrupted and its process is still winding down. Each +/// claims the name. `Starting` closes a TOCTOU window: checking "is `name` +/// free" and committing to it are two different lock acquisitions unless +/// the check *is* the reservation — see `reserve`. /// /// `killed` is the other half of that map's story: `running` says what is in /// flight *now*, `killed` remembers the names whose last turn ended on a @@ -359,7 +381,7 @@ fn searched_location(config: &Config) -> Option { /// name across directories at your own risk, the tool doesn't disambiguate /// it (flagged in review when `dir` was added). pub struct State { - running: Mutex>>, + running: Mutex>, dirs: Mutex>, killed: Mutex>, last_event: Mutex>, @@ -547,17 +569,17 @@ impl State { } } - /// `Some(true)` — a live process is tracked, interruptible. `Some(false)` - /// — the name is claimed but no process is confirmed under it: an - /// in-flight start/continue that hasn't spawned yet, or the gap between - /// one continued turn's child exiting and the next one spawning (see - /// `between_turns`). `None` — nothing tracked under `name` at all. - fn occupancy(&self, name: &str) -> Option { + /// Which [`Slot`] `name` is in; `None` with nothing tracked under it. + fn occupancy(&self, name: &str) -> Option { self.running .lock() .unwrap_or_else(PoisonError::into_inner) .get(name) - .map(Option::is_some) + .map(|slot| match slot { + Slot::Starting => Occupancy::Starting, + Slot::Running(_) => Occupancy::Running, + Slot::Stopping => Occupancy::Stopping, + }) } /// Atomically claim `name` for an in-flight start/continue: check @@ -572,21 +594,21 @@ impl State { if running.contains_key(name) { return false; } - running.insert(name.to_owned(), None); + running.insert(name.to_owned(), Slot::Starting); true } /// Release a reservation that never made it to a real spawn (an error /// on the slow path between `reserve` and `Claude::spawn` succeeding). - /// A no-op if the entry was already upgraded to `Some` — this only ever - /// clears a still-`None` placeholder, never a live process. + /// A no-op if the entry was already upgraded to `Running` — this only + /// ever clears a still-`Starting` placeholder, never a live process. /// /// The signal token the failed call minted goes with it: no process ever /// read that URL, and a token outliving the call that minted it is the /// one way a route could resolve to a session that isn't there. fn release_reservation(&self, name: &str) { let mut running = self.running.lock().unwrap_or_else(PoisonError::into_inner); - let released = matches!(running.get(name), Some(None)); + let released = matches!(running.get(name), Some(Slot::Starting)); if released { running.remove(name); } @@ -742,14 +764,14 @@ impl State { } } - /// Upgrade `name`'s `None` reservation to a real, interruptible process. - /// Same key as the reservation, so there is no window in which `name` - /// reads as unoccupied between the two. + /// Upgrade `name`'s `Starting` reservation to a real, interruptible + /// process. Same key as the reservation, so there is no window in which + /// `name` reads as unoccupied between the two. fn track(&self, name: &str, stopper: Stopper) { self.running .lock() .unwrap_or_else(PoisonError::into_inner) - .insert(name.to_owned(), Some(stopper)); + .insert(name.to_owned(), Slot::Running(stopper)); } /// Drop `name`'s cancel handle but keep the name claimed — the moment @@ -763,7 +785,7 @@ impl State { self.running .lock() .unwrap_or_else(PoisonError::into_inner) - .insert(name.to_owned(), None); + .insert(name.to_owned(), Slot::Starting); } /// Set (or clear) `name`'s goal and start its turn budget over at turn @@ -1286,7 +1308,7 @@ pub async fn start(state: &Arc, req: StartRequest) -> anyhow::Result) -> String { /// them is a fact about the same session. struct StatusFacts<'a> { name: &'a str, - occupancy: Option, + occupancy: Option, killed: Option, session_exists: bool, last_event_age: Option, @@ -2224,7 +2248,7 @@ fn describe_status(facts: &StatusFacts<'_>) -> anyhow::Result { let name = facts.name; let turns = describe_turns(facts.turns); match facts.occupancy { - Some(true) => { + Some(Occupancy::Running) => { return Ok(format!( "subagent `{name}` is running — its turn is still in flight, so there's nothing \ to do but let it work: the daemon pushes a todo when the run ends, or \ @@ -2233,7 +2257,7 @@ fn describe_status(facts: &StatusFacts<'_>) -> anyhow::Result { describe_pending_signal(facts.stop.as_ref()), )); } - Some(false) => { + Some(Occupancy::Starting) => { return Ok(format!( "subagent `{name}` is starting — the name is claimed but no process is confirmed \ under it yet, either because a `start`/`continue` hasn't spawned one or because \ @@ -2241,6 +2265,15 @@ fn describe_status(facts: &StatusFacts<'_>) -> anyhow::Result { shortly rather than starting anything else under this name.{turns}" )); } + Some(Occupancy::Stopping) => { + return Ok(format!( + "subagent `{name}` is stopping — `interrupt` sent it SIGINT and its process \ + hasn't exited yet. The name stays claimed until it does, so `start` and \ + `continue` are refused meanwhile. Nothing escalates past that SIGINT: one that \ + never exits needs killing by hand.{turns}{}", + describe_liveness(facts.last_event_age), + )); + } None => {} } if let Some(signal) = facts.killed { @@ -2327,57 +2360,60 @@ fn describe_stopped(name: &str, stop: &StopReason, turns: &str) -> String { } } -/// Signal `name`'s running process — `force` picks SIGKILL over SIGINT (see -/// `hive_claude::Cancel::cancel`). Refuses a name with nothing running: no -/// entry at all, or one still in the brief not-yet-spawned window (nothing -/// to signal yet — the reservation is put back so a concurrent -/// `start`/`continue` for the same name still gets refused). +/// Stop `name`'s running turn: SIGINT to claude's process group (see +/// `hive_claude::Cancel::cancel`), or `session/cancel` on ACP. Nothing +/// escalates to SIGKILL: each command claude's Bash tool starts leads its +/// own session, and only claude shutting down on SIGINT stops it. /// -/// This stops a goal run, not just the turn in it — and it does so by -/// **recording `StopReason::Cancelled`**, not by relying on how the child -/// dies. +/// Returns without waiting. A claude turn's name stays claimed as +/// `Slot::Stopping` until the turn's own `finish_turn` clears it, so no +/// second claude starts on the session while the first is winding down. /// -/// That distinction is the entire fix. The signal alone is not enough: the -/// default is SIGINT, claude handles SIGINT by emitting its terminal -/// `result` event and exiting 0, and a zero exit is `TurnEnd::Complete` — -/// so `spawn_and_track`'s killed-turn early return never fires and the loop -/// goes straight on to spawn turn three of five. The recorded reason is what -/// `plan_after_turn` reads, and it is checked before the goal, so the run -/// stops whichever way the child ends up exiting. `force`'s SIGKILL does -/// produce a `TurnEnd::Killed`; recording the reason anyway costs nothing -/// and keeps one path, and `status` still prefers the kill record it -/// already had. -/// -/// The reason is written while the `running` lock is still held, so there is -/// no instant in which the name has lost its cancel handle but not yet -/// gained its stop reason — a turn ending of its own accord in that gap -/// would find nothing to stop it and start the next one. +/// It stops the goal run, not just the turn, by **recording +/// `StopReason::Cancelled`**: claude answers SIGINT by emitting its +/// terminal `result` event and exiting 0, which is `TurnEnd::Complete`, so +/// only the recorded reason, which `plan_after_turn` checks before the +/// goal, keeps the loop from re-prompting. It is written while the +/// `running` lock is held, so a turn that ends on its own in between still +/// finds it. /// /// # Errors /// -/// An invalid name, nothing tracked under `name`, or nothing to signal right -/// now (a spawn or ACP turn not yet live, or a goal run between turns). -pub fn interrupt(state: &State, name: &str, force: bool) -> anyhow::Result { +/// An invalid name, nothing tracked under `name`, nothing to signal right +/// now (a spawn or ACP turn not yet live, or a goal run between turns; the +/// reservation is put back, so a concurrent `start`/`continue` is still +/// refused), or a claude already signalled and not yet exited. +pub fn interrupt(state: &State, name: &str) -> anyhow::Result { validate_name(name)?; let mut running = state.running.lock().unwrap_or_else(PoisonError::into_inner); + let mut claimed = ""; match running.remove(name) { None => anyhow::bail!("no subagent named `{name}` is currently running"), - Some(None) => { - running.insert(name.to_owned(), None); + Some(Slot::Starting) => { + running.insert(name.to_owned(), Slot::Starting); return Err(still_starting(name)); } - Some(Some(Stopper::Claude(cancel))) => { - state.record_stop(name, StopReason::Cancelled); - drop(running); - cancel.cancel(force); + Some(Slot::Stopping) => { + running.insert(name.to_owned(), Slot::Stopping); + anyhow::bail!( + "subagent `{name}` was already interrupted and its process hasn't exited yet — \ + `status` says when it has" + ); } - // `force` has no stronger form on ACP: the runtime already kills an - // agent that ignores `session/cancel`. `cancel()` is `false` when no - // turn is in flight (the agent not prompted yet, or a turn just - // ended), and then nothing was cancelled. The stop goes in before the - // cancel and comes out under the same `stops` lock, so - // `plan_after_turn` never reads a cancel that didn't happen. - Some(Some(Stopper::Acp(canceller))) => { + Some(Slot::Running(Stopper::Claude(cancel))) => { + state.record_stop(name, StopReason::Cancelled); + running.insert(name.to_owned(), Slot::Stopping); + drop(running); + cancel.cancel(false); + claimed = " The name stays claimed until claude has exited; `status` says `stopping` \ + until then."; + } + // `cancel()` is `false` when no turn is in flight (the agent not + // prompted yet, or a turn just ended), and then nothing was + // cancelled. The stop goes in before the cancel and comes out under + // the same `stops` lock, so `plan_after_turn` never reads a cancel + // that didn't happen. + Some(Slot::Running(Stopper::Acp(canceller))) => { let mut stops = state.stops.lock().unwrap_or_else(PoisonError::into_inner); let prior = stops.insert(name.to_owned(), StopReason::Cancelled); if !canceller.cancel() { @@ -2386,7 +2422,7 @@ pub fn interrupt(state: &State, name: &str, force: bool) -> anyhow::Result stops.remove(name), }; drop(stops); - running.insert(name.to_owned(), Some(Stopper::Acp(canceller))); + running.insert(name.to_owned(), Slot::Running(Stopper::Acp(canceller))); return Err(still_starting(name)); } } @@ -2394,10 +2430,24 @@ pub fn interrupt(state: &State, name: &str, force: bool) -> anyhow::Result anyhow::Error { + if state.occupancy(name) == Some(Occupancy::Stopping) { + anyhow::anyhow!( + "subagent `{name}` is stopping — it was interrupted and its process hasn't exited \ + yet; the name is free once it has (`status` says when)" + ) + } else { + anyhow::anyhow!("subagent `{name}` is already running — {hint}") + } +} + /// `interrupt`'s refusal for a name that is claimed but has nothing to /// signal yet. fn still_starting(name: &str) -> anyhow::Error { @@ -2539,7 +2589,7 @@ mod tests { state.reserve("n"); assert_eq!( state.occupancy("n"), - Some(false), + Some(Occupancy::Starting), "reserved-but-not-yet-spawned must read as occupied-but-not-running" ); } @@ -2984,7 +3034,7 @@ mod tests { // has to carry its own explanation — checked one state at a time, // which is all a caller ever gets back. let running = describe_status(&StatusFacts { - occupancy: Some(true), + occupancy: Some(Occupancy::Running), ..StatusFacts::new("n") }) .expect("running status"); @@ -2993,7 +3043,7 @@ mod tests { "a running answer must say the turn is in flight and how to stop it: {running}" ); let starting = describe_status(&StatusFacts { - occupancy: Some(false), + occupancy: Some(Occupancy::Starting), ..StatusFacts::new("n") }) .expect("starting status"); @@ -3238,7 +3288,7 @@ mod tests { #[test] fn a_running_status_reports_the_age_and_an_idle_one_does_not() { let running = describe_status(&StatusFacts { - occupancy: Some(true), + occupancy: Some(Occupancy::Running), last_event_age: Some(Duration::from_secs(4)), ..StatusFacts::new("n") }) @@ -3249,7 +3299,7 @@ mod tests { {running}" ); let wedged = describe_status(&StatusFacts { - occupancy: Some(true), + occupancy: Some(Occupancy::Running), last_event_age: Some(Duration::from_mins(15)), ..StatusFacts::new("n") }) @@ -3281,7 +3331,7 @@ mod tests { // turn can finish between the two. The answer drops the age rather // than inventing one. let raced = describe_status(&StatusFacts { - occupancy: Some(true), + occupancy: Some(Occupancy::Running), ..StatusFacts::new("n") }) .expect("running status"); @@ -3710,9 +3760,7 @@ mod tests { } assert_eq!(turns_spawned(&log), 1, "the first turn never started"); - // An operator's plain `interrupt`, not `force` — SIGINT is the - // default and the case that was broken. - interrupt(&state, "n", false).expect("a running turn is interruptible"); + interrupt(&state, "n").expect("a running turn is interruptible"); // Generous: the fake logs its invocation as its first act, so a // re-prompt shows up here almost immediately once the loop decides @@ -3776,53 +3824,131 @@ mod tests { #[tokio::test] async fn an_interrupt_stops_the_commands_the_subagent_started() { - for force in [false, true] { - let dir = std::env::temp_dir().join(format!( - "hive-subagent-group-{}-{:?}-{force}", - std::process::id(), - std::thread::current().id() - )); - std::fs::create_dir_all(&dir).expect("scratch dir"); - let pidfile = dir.join("grandchild.pid"); - let state = Arc::new(mid_run("n", "carry on indefinitely", 5)); - let config = Config { - program: Some(busy_claude(&dir, &pidfile)), - ..Config::default() - }; - spawn_and_track( - &state, - "n", - &config, - &Attach::Create("n".to_owned()), - "turn one".to_owned(), - None, - ) - .expect("the fake claude spawns"); - let grandchild = read_pid(&pidfile).await; + let dir = scratch_dir("group"); + let pidfile = dir.join("grandchild.pid"); + let state = Arc::new(mid_run("n", "carry on indefinitely", 5)); + let config = Config { + program: Some(busy_claude(&dir, &pidfile)), + ..Config::default() + }; + spawn_and_track( + &state, + "n", + &config, + &Attach::Create("n".to_owned()), + "turn one".to_owned(), + None, + ) + .expect("the fake claude spawns"); + let grandchild = read_pid(&pidfile).await; - interrupt(&state, "n", force).expect("a running turn is interruptible"); + interrupt(&state, "n").expect("a running turn is interruptible"); - let mut alive = is_running(grandchild); - for _ in 0..250 { - if !alive { - break; - } - tokio::time::sleep(Duration::from_millis(20)).await; - alive = is_running(grandchild); + let mut alive = is_running(grandchild); + for _ in 0..250 { + if !alive { + break; } - if alive { - // SAFETY: `kill(2)` on a pid this test started; keeps a failed - // run from leaving a 300s `sleep` behind. - unsafe { - libc::kill(grandchild, libc::SIGKILL); - } - } - std::fs::remove_dir_all(&dir).ok(); - assert!( - !alive, - "grandchild {grandchild} outlived interrupt(force = {force})" - ); + tokio::time::sleep(Duration::from_millis(20)).await; + alive = is_running(grandchild); } + if alive { + // SAFETY: `kill(2)` on a pid this test started; keeps a failed + // run from leaving a 300s `sleep` behind. + unsafe { + libc::kill(grandchild, libc::SIGKILL); + } + } + std::fs::remove_dir_all(&dir).ok(); + assert!(!alive, "grandchild {grandchild} outlived the interrupt"); + } + + /// A stand-in for claude that shuts down on SIGINT the way claude does: + /// it writes `interrupted` once the signal arrives, then exits 0 as soon + /// as `release` exists, or after 10s. It writes `ready` once its trap is + /// set. A SIGKILL leaves no `interrupted` behind. + fn graceful_claude(dir: &Path) -> PathBuf { + let script = dir.join("graceful-claude"); + write_executable( + &script, + &format!( + "#!/bin/sh\ncd '{}'\ntrap ': > interrupted; i=0; until [ -e release ] || \ + [ $i -ge 200 ]; do sleep 0.05; i=$((i+1)); done; kill $!; exit 0' INT\n\ + sleep 300 >/dev/null 2>&1 &\n: > ready\nwait\n", + dir.display() + ), + ); + script + } + + /// Spawn [`graceful_claude`] as `n`'s turn and wait until its trap is set. + async fn spawn_graceful(state: &Arc, dir: &Path) { + let config = Config { + program: Some(graceful_claude(dir)), + ..Config::default() + }; + spawn_and_track( + state, + "n", + &config, + &Attach::Create("n".to_owned()), + "turn one".to_owned(), + None, + ) + .expect("the fake claude spawns"); + until("the fake claude sets its trap", || { + dir.join("ready").exists() + }) + .await; + } + + #[tokio::test] + async fn an_interrupt_lets_claude_shut_down_on_sigint() { + let dir = scratch_dir("sigint"); + std::fs::write(dir.join("release"), "").expect("release file"); + let state = Arc::new(mid_run("n", "carry on indefinitely", 5)); + spawn_graceful(&state, &dir).await; + + interrupt(&state, "n").expect("a running turn is interruptible"); + + until("the run ends", || state.occupancy("n").is_none()).await; + let interrupted = dir.join("interrupted").exists(); + std::fs::remove_dir_all(&dir).ok(); + assert!( + interrupted, + "claude must get SIGINT, which it shuts its tool commands down on, never SIGKILL" + ); + assert_eq!(state.killed_by("n"), None, "it exited on its own"); + } + + #[tokio::test] + async fn an_interrupted_name_stays_claimed_until_claude_has_exited() { + let dir = scratch_dir("stopping"); + let state = Arc::new(mid_run("n", "carry on indefinitely", 5)); + spawn_graceful(&state, &dir).await; + + interrupt(&state, "n").expect("a running turn is interruptible"); + until("the fake claude gets the SIGINT", || { + dir.join("interrupted").exists() + }) + .await; + + assert_eq!(state.occupancy("n"), Some(Occupancy::Stopping)); + let status = status(&state, "n", None).expect("status"); + assert!(status.contains("is stopping"), "{status}"); + let refused = continue_(&state, "n", "more".to_owned(), None, None, None) + .await + .expect_err("a second claude must not start on the session"); + assert!(refused.to_string().contains("is stopping"), "{refused}"); + assert!(interrupt(&state, "n").is_err(), "it was already signalled"); + + std::fs::write(dir.join("release"), "").expect("release file"); + until("the run ends", || state.occupancy("n").is_none()).await; + std::fs::remove_dir_all(&dir).ok(); + assert!( + state.reserve("n"), + "the name is free once claude has exited" + ); } #[test] @@ -4079,7 +4205,12 @@ mod tests { #[test] fn status_reports_the_turn_counter_in_every_state_a_goal_run_reaches() { - for occupancy in [Some(true), Some(false), None] { + for occupancy in [ + Some(Occupancy::Running), + Some(Occupancy::Starting), + Some(Occupancy::Stopping), + None, + ] { let answer = describe_status(&StatusFacts { occupancy, session_exists: true, @@ -4094,7 +4225,7 @@ mod tests { ); } let goalless = describe_status(&StatusFacts { - occupancy: Some(true), + occupancy: Some(Occupancy::Running), ..StatusFacts::new("n") }) .expect("status"); @@ -4110,7 +4241,7 @@ mod tests { // it is blocked, its turn hasn't ended yet, and plain "running" would // tell the parent to leave it alone. let answer = describe_status(&StatusFacts { - occupancy: Some(true), + occupancy: Some(Occupancy::Running), stop: Some(StopReason::NeedHelp("no credential".to_owned())), ..StatusFacts::new("n") }) @@ -4120,7 +4251,7 @@ mod tests { "both facts are true at once and both have to be in the answer: {answer}" ); let quiet = describe_status(&StatusFacts { - occupancy: Some(true), + occupancy: Some(Occupancy::Running), stop: Some(StopReason::Done), ..StatusFacts::new("n") }) @@ -4633,10 +4764,8 @@ done turns_spawned(&dir.join("claude-turns")) == 1 }) .await; - interrupt(&state, "n", false).expect("a running claude turn is interruptible"); - // `interrupt` drops the tracking entry itself; the run is over once - // `finish_turn` has cleared its liveness clock too. - until("the run ends", || state.last_event_age("n").is_none()).await; + interrupt(&state, "n").expect("a running claude turn is interruptible"); + until("the run ends", || state.occupancy("n").is_none()).await; // This `start` had no goal, so the fake claude's clean exit after // SIGINT (`TurnEnd::Complete`) is exactly the case that used to @@ -4983,7 +5112,7 @@ echo '{"type":"result","subtype":"success","is_error":false}' }) .await; - interrupt(&state, "n", false).expect("a running ACP turn is interruptible"); + interrupt(&state, "n").expect("a running ACP turn is interruptible"); until("the run ends", || state.last_event_age("n").is_none()).await; assert_eq!(lines(&dir.join("prompts")).len(), 1, "no second goal turn"); @@ -5006,7 +5135,7 @@ echo '{"type":"result","subtype":"success","is_error":false}' // The test runtime is single-threaded, so the run's task has not // been polled yet: no turn is in flight. - let err = interrupt(&state, "n", false).expect_err("nothing to cancel yet"); + let err = interrupt(&state, "n").expect_err("nothing to cancel yet"); assert!(err.to_string().contains("still starting"), "{err}"); assert_eq!(state.stop_reason("n"), None);