mirror of
https://github.com/nearai/ironclaw.git
synced 2026-09-02 23:56:24 +08:00
* fix(web): keep Routines tab after engine v1 → v2 upgrade (#2982) Users upgrading from a v1 install (e.g. 0.24.0 → 0.26.0) lost the UI affordance to view or manage existing routines: `applyEngineModeToTabs()` and `applyEngineModeUi()` unconditionally hid the v1-only Routines tab whenever ENGINE_V2 was enabled, even though the routines were still in the database and the API still served them. The fix adds a `userHasLegacyRoutines` flag, populated from `/api/routines/summary` on first gateway-status poll. The Routines tab stays visible (and `#/routines/<id>` still resolves to the legacy detail view) when the user has any v1 routines. Also fixes a wire-contract drift in `gateway-tee.js`: it read `data.engine_v2` for the activity store and `data.engine_v2_enabled` for the global, with `applyEngineModeUi()` running before the global was set. Per `.claude/rules/types.md` ("Wire-contract field naming"), the duplicate `engine_v2` field is removed from `GatewayStatusResponse`; the JS now reads the single canonical name once and sets the global before any UI helper consults it. * fix(web): address PR #2992 review notes — race guard, dedup, post-delete refresh Three review-driven hardening tweaks plus expanded Playwright coverage, all on the same #2982 fix: - gateway-tee.js: flip `engineModeApplied = true` synchronously so a second status poll firing while the first refresh is still in flight cannot kick off a duplicate `/api/routines/summary` request. The trailing `.then()` still runs on fetch failure (the `.catch()` chain resolves to undefined), so the UI still settles. - projects.js: route the routines-tab visibility branch through `shouldHideRoutinesTab()` instead of duplicating the predicate inline. Single source of truth for the rule. - routines.js: refresh `userHasLegacyRoutines` after a successful `deleteRoutine` so the v2 user who just removed their last legacy routine sees the tab fall back to hidden without a page reload. Playwright coverage grew from 5 to 11 cases: route-mocked summary, zero-total clears the flag, fetch failure preserves the prior value, post-delete refresh hides the tab, dual back-to-back first polls fan out only one summary fetch, and `restoreFromHash` routes correctly when legacy data exists.
467 lines
16 KiB
Rust
467 lines
16 KiB
Rust
//! End-to-end integration tests for the WebSocket gateway.
|
|
//!
|
|
//! These tests start a real Axum server on a random port, connect a WebSocket
|
|
//! client, and verify the full message flow:
|
|
//! - WebSocket upgrade with auth
|
|
//! - Ping/pong
|
|
//! - Client message → agent msg_tx
|
|
//! - Broadcast AppEvent → WebSocket client
|
|
//! - Connection tracking (counter increment/decrement)
|
|
//! - Gateway status endpoint
|
|
|
|
use std::net::SocketAddr;
|
|
use std::sync::Arc;
|
|
use std::time::Duration;
|
|
|
|
use futures::{SinkExt, StreamExt};
|
|
use tokio::sync::mpsc;
|
|
use tokio::time::timeout;
|
|
use tokio_tungstenite::tungstenite::Message;
|
|
use tokio_tungstenite::tungstenite::client::IntoClientRequest;
|
|
|
|
use ironclaw::channels::IncomingMessage;
|
|
use ironclaw::channels::web::platform::router::start_server;
|
|
use ironclaw::channels::web::platform::state::GatewayState;
|
|
use ironclaw::channels::web::sse::SseManager;
|
|
use ironclaw::channels::web::ws::WsConnectionTracker;
|
|
use ironclaw_common::AppEvent;
|
|
|
|
const AUTH_TOKEN: &str = "test-token-12345";
|
|
const TIMEOUT: Duration = Duration::from_secs(5);
|
|
|
|
/// Start a gateway server on a random port and return the bound address + agent
|
|
/// message receiver.
|
|
async fn start_test_server() -> (
|
|
SocketAddr,
|
|
Arc<GatewayState>,
|
|
mpsc::Receiver<IncomingMessage>,
|
|
) {
|
|
let (agent_tx, agent_rx) = mpsc::channel(64);
|
|
|
|
let state = Arc::new(GatewayState {
|
|
msg_tx: tokio::sync::RwLock::new(Some(agent_tx)),
|
|
sse: Arc::new(SseManager::new()),
|
|
workspace: None,
|
|
workspace_pool: None,
|
|
multi_tenant_mode: false,
|
|
session_manager: None,
|
|
log_broadcaster: None,
|
|
log_level_handle: None,
|
|
extension_manager: None,
|
|
tool_registry: None,
|
|
store: None,
|
|
settings_cache: None,
|
|
job_manager: None,
|
|
prompt_queue: None,
|
|
scheduler: None,
|
|
owner_id: "test-user".to_string(),
|
|
shutdown_tx: tokio::sync::RwLock::new(None),
|
|
ws_tracker: Some(Arc::new(WsConnectionTracker::new())),
|
|
llm_provider: None,
|
|
llm_reload: None,
|
|
llm_session_manager: None,
|
|
config_toml_path: None,
|
|
skill_registry: None,
|
|
skill_catalog: None,
|
|
auth_manager: None,
|
|
chat_rate_limiter: ironclaw::channels::web::platform::state::PerUserRateLimiter::new(
|
|
30, 60,
|
|
),
|
|
oauth_rate_limiter: ironclaw::channels::web::platform::state::PerUserRateLimiter::new(
|
|
20, 60,
|
|
),
|
|
webhook_rate_limiter: ironclaw::channels::web::platform::state::RateLimiter::new(10, 60),
|
|
registry_entries: Vec::new(),
|
|
cost_guard: None,
|
|
routine_engine: Arc::new(tokio::sync::RwLock::new(None)),
|
|
startup_time: std::time::Instant::now(),
|
|
active_config: Arc::new(tokio::sync::RwLock::new(
|
|
ironclaw::channels::web::platform::state::ActiveConfigSnapshot::default(),
|
|
)),
|
|
secrets_store: None,
|
|
db_auth: None,
|
|
pairing_store: None,
|
|
oauth_providers: None,
|
|
oauth_state_store: None,
|
|
oauth_base_url: None,
|
|
oauth_allowed_domains: Vec::new(),
|
|
near_nonce_store: None,
|
|
near_rpc_url: None,
|
|
near_network: None,
|
|
oauth_sweep_shutdown: None,
|
|
frontend_html_cache: std::sync::Arc::new(tokio::sync::RwLock::new(None)),
|
|
tool_dispatcher: None,
|
|
});
|
|
|
|
let auth = ironclaw::channels::web::auth::MultiAuthState::single(
|
|
AUTH_TOKEN.to_string(),
|
|
"test-user".to_string(),
|
|
);
|
|
let addr: SocketAddr = "127.0.0.1:0".parse().unwrap();
|
|
let bound_addr = start_server(addr, state.clone(), auth.into())
|
|
.await
|
|
.expect("Failed to start test server");
|
|
|
|
(bound_addr, state, agent_rx)
|
|
}
|
|
|
|
/// Connect a WebSocket client with auth token in query parameter.
|
|
async fn connect_ws(
|
|
addr: SocketAddr,
|
|
) -> tokio_tungstenite::WebSocketStream<tokio_tungstenite::MaybeTlsStream<tokio::net::TcpStream>> {
|
|
let url = format!("ws://{}/api/chat/ws?token={}", addr, AUTH_TOKEN);
|
|
let mut request = url.into_client_request().unwrap();
|
|
// Server requires an Origin header from localhost to prevent cross-site WS hijacking.
|
|
request.headers_mut().insert(
|
|
"Origin",
|
|
format!("http://127.0.0.1:{}", addr.port()).parse().unwrap(),
|
|
);
|
|
let (stream, _response) = tokio_tungstenite::connect_async(request)
|
|
.await
|
|
.expect("Failed to connect WebSocket");
|
|
stream
|
|
}
|
|
|
|
/// Read the next text frame from the WebSocket, with a timeout.
|
|
async fn recv_text(
|
|
stream: &mut (impl StreamExt<Item = Result<Message, tokio_tungstenite::tungstenite::Error>> + Unpin),
|
|
) -> String {
|
|
let msg = timeout(TIMEOUT, stream.next())
|
|
.await
|
|
.expect("Timed out waiting for WS message")
|
|
.expect("Stream ended")
|
|
.expect("WS error");
|
|
match msg {
|
|
Message::Text(text) => text.to_string(),
|
|
other => panic!("Expected Text frame, got {:?}", other),
|
|
}
|
|
}
|
|
|
|
// ============================================================================
|
|
// Tests
|
|
// ============================================================================
|
|
|
|
#[tokio::test]
|
|
async fn test_ws_ping_pong() {
|
|
let (addr, _state, _agent_rx) = start_test_server().await;
|
|
let mut ws = connect_ws(addr).await;
|
|
|
|
// Send ping
|
|
let ping = r#"{"type":"ping"}"#;
|
|
ws.send(Message::Text(ping.into())).await.unwrap();
|
|
|
|
// Expect pong
|
|
let text = recv_text(&mut ws).await;
|
|
let parsed: serde_json::Value = serde_json::from_str(&text).unwrap();
|
|
assert_eq!(parsed["type"], "pong");
|
|
|
|
ws.close(None).await.unwrap();
|
|
}
|
|
|
|
#[tokio::test]
|
|
async fn test_ws_message_reaches_agent() {
|
|
let (addr, _state, mut agent_rx) = start_test_server().await;
|
|
let mut ws = connect_ws(addr).await;
|
|
|
|
// Send a chat message
|
|
let msg = r#"{"type":"message","content":"hello from ws","thread_id":"t42"}"#;
|
|
ws.send(Message::Text(msg.into())).await.unwrap();
|
|
|
|
// Verify it arrives on the agent's msg_tx
|
|
let incoming = timeout(TIMEOUT, agent_rx.recv())
|
|
.await
|
|
.expect("Timed out waiting for agent message")
|
|
.expect("Agent channel closed");
|
|
|
|
assert_eq!(incoming.content, "hello from ws");
|
|
assert_eq!(incoming.thread_id.as_ref().map(|t| t.as_str()), Some("t42"));
|
|
assert_eq!(incoming.channel, "gateway");
|
|
assert_eq!(incoming.user_id, "test-user");
|
|
|
|
ws.close(None).await.unwrap();
|
|
}
|
|
|
|
#[tokio::test]
|
|
async fn test_ws_broadcast_event_received() {
|
|
let (addr, state, _agent_rx) = start_test_server().await;
|
|
let mut ws = connect_ws(addr).await;
|
|
|
|
// Give the connection a moment to fully establish
|
|
tokio::time::sleep(Duration::from_millis(50)).await;
|
|
|
|
// Broadcast an event (simulates agent sending a response)
|
|
state.sse.broadcast(AppEvent::Response {
|
|
content: "agent says hi".to_string(),
|
|
thread_id: "t1".to_string(),
|
|
});
|
|
|
|
// The WS client should receive it
|
|
let text = recv_text(&mut ws).await;
|
|
let parsed: serde_json::Value = serde_json::from_str(&text).unwrap();
|
|
assert_eq!(parsed["type"], "event");
|
|
assert_eq!(parsed["event_type"], "response");
|
|
assert_eq!(parsed["data"]["content"], "agent says hi");
|
|
|
|
ws.close(None).await.unwrap();
|
|
}
|
|
|
|
#[tokio::test]
|
|
async fn test_ws_thinking_event() {
|
|
let (addr, state, _agent_rx) = start_test_server().await;
|
|
let mut ws = connect_ws(addr).await;
|
|
tokio::time::sleep(Duration::from_millis(50)).await;
|
|
|
|
state.sse.broadcast(AppEvent::Thinking {
|
|
message: "analyzing...".to_string(),
|
|
thread_id: None,
|
|
});
|
|
|
|
let text = recv_text(&mut ws).await;
|
|
let parsed: serde_json::Value = serde_json::from_str(&text).unwrap();
|
|
assert_eq!(parsed["type"], "event");
|
|
assert_eq!(parsed["event_type"], "thinking");
|
|
assert_eq!(parsed["data"]["message"], "analyzing...");
|
|
|
|
ws.close(None).await.unwrap();
|
|
}
|
|
|
|
#[tokio::test]
|
|
async fn test_ws_connection_tracking() {
|
|
let (addr, state, _agent_rx) = start_test_server().await;
|
|
let tracker = state.ws_tracker.as_ref().unwrap();
|
|
|
|
assert_eq!(tracker.connection_count(), 0);
|
|
|
|
// Connect first client
|
|
let ws1 = connect_ws(addr).await;
|
|
tokio::time::sleep(Duration::from_millis(50)).await;
|
|
assert_eq!(tracker.connection_count(), 1);
|
|
|
|
// Connect second client
|
|
let ws2 = connect_ws(addr).await;
|
|
tokio::time::sleep(Duration::from_millis(50)).await;
|
|
assert_eq!(tracker.connection_count(), 2);
|
|
|
|
// Disconnect first
|
|
drop(ws1);
|
|
tokio::time::sleep(Duration::from_millis(100)).await;
|
|
assert_eq!(tracker.connection_count(), 1);
|
|
|
|
// Disconnect second
|
|
drop(ws2);
|
|
tokio::time::sleep(Duration::from_millis(100)).await;
|
|
assert_eq!(tracker.connection_count(), 0);
|
|
}
|
|
|
|
#[tokio::test]
|
|
async fn test_ws_invalid_message_returns_error() {
|
|
let (addr, _state, _agent_rx) = start_test_server().await;
|
|
let mut ws = connect_ws(addr).await;
|
|
|
|
// Send invalid JSON
|
|
ws.send(Message::Text("not json".into())).await.unwrap();
|
|
|
|
// Should get an error message back
|
|
let text = recv_text(&mut ws).await;
|
|
let parsed: serde_json::Value = serde_json::from_str(&text).unwrap();
|
|
assert_eq!(parsed["type"], "error");
|
|
assert!(
|
|
parsed["message"]
|
|
.as_str()
|
|
.unwrap()
|
|
.contains("Invalid message")
|
|
);
|
|
|
|
ws.close(None).await.unwrap();
|
|
}
|
|
|
|
#[tokio::test]
|
|
async fn test_ws_unknown_type_returns_error() {
|
|
let (addr, _state, _agent_rx) = start_test_server().await;
|
|
let mut ws = connect_ws(addr).await;
|
|
|
|
// Send valid JSON but unknown message type
|
|
ws.send(Message::Text(r#"{"type":"foobar"}"#.into()))
|
|
.await
|
|
.unwrap();
|
|
|
|
let text = recv_text(&mut ws).await;
|
|
let parsed: serde_json::Value = serde_json::from_str(&text).unwrap();
|
|
assert_eq!(parsed["type"], "error");
|
|
|
|
ws.close(None).await.unwrap();
|
|
}
|
|
|
|
#[tokio::test]
|
|
async fn test_gateway_status_endpoint() {
|
|
let (addr, _state, _agent_rx) = start_test_server().await;
|
|
|
|
// Connect a WS client
|
|
let _ws = connect_ws(addr).await;
|
|
tokio::time::sleep(Duration::from_millis(50)).await;
|
|
|
|
// Hit the status endpoint
|
|
let client = reqwest::Client::new();
|
|
let resp = client
|
|
.get(format!("http://{}/api/gateway/status", addr))
|
|
.header("Authorization", format!("Bearer {}", AUTH_TOKEN))
|
|
.send()
|
|
.await
|
|
.expect("Failed to fetch status");
|
|
|
|
assert_eq!(resp.status(), 200);
|
|
|
|
let body: serde_json::Value = resp.json().await.unwrap();
|
|
assert_eq!(body["ws_connections"], 1);
|
|
assert!(body["total_connections"].as_u64().unwrap() >= 1);
|
|
|
|
// Regression for #2982: the response carries the engine flag under a
|
|
// single canonical name. The old duplicate `engine_v2` field led the
|
|
// gateway JS to read the value from one name while writing the other,
|
|
// and dropping it locks in the wire-contract rule from
|
|
// .claude/rules/types.md.
|
|
assert!(
|
|
body.get("engine_v2_enabled").is_some(),
|
|
"expected engine_v2_enabled field on gateway status response"
|
|
);
|
|
assert!(
|
|
body.get("engine_v2").is_none(),
|
|
"duplicate engine_v2 field must not be re-introduced (#2982)"
|
|
);
|
|
}
|
|
|
|
#[tokio::test]
|
|
async fn test_ws_no_auth_rejected() {
|
|
let (addr, _state, _agent_rx) = start_test_server().await;
|
|
|
|
// Try to connect without auth token
|
|
let url = format!("ws://{}/api/chat/ws", addr);
|
|
let request = url.into_client_request().unwrap();
|
|
let result = tokio_tungstenite::connect_async(request).await;
|
|
|
|
// Should fail (401 from auth middleware before WS upgrade)
|
|
assert!(result.is_err());
|
|
}
|
|
|
|
#[tokio::test]
|
|
async fn test_ws_multiple_events_in_sequence() {
|
|
let (addr, state, _agent_rx) = start_test_server().await;
|
|
let mut ws = connect_ws(addr).await;
|
|
tokio::time::sleep(Duration::from_millis(50)).await;
|
|
|
|
// Broadcast multiple events rapidly
|
|
state.sse.broadcast(AppEvent::Thinking {
|
|
message: "step 1".to_string(),
|
|
thread_id: None,
|
|
});
|
|
state.sse.broadcast(AppEvent::ToolStarted {
|
|
name: "shell".to_string(),
|
|
detail: None,
|
|
call_id: Some("call_shell_1".to_string()),
|
|
thread_id: None,
|
|
});
|
|
state.sse.broadcast(AppEvent::ToolCompleted {
|
|
name: "shell".to_string(),
|
|
success: true,
|
|
error: None,
|
|
parameters: None,
|
|
call_id: Some("call_shell_1".to_string()),
|
|
duration_ms: Some(42),
|
|
thread_id: None,
|
|
});
|
|
state.sse.broadcast(AppEvent::Response {
|
|
content: "done".to_string(),
|
|
thread_id: "t1".to_string(),
|
|
});
|
|
|
|
// Receive all 4 in order
|
|
let t1 = recv_text(&mut ws).await;
|
|
let t2 = recv_text(&mut ws).await;
|
|
let t3 = recv_text(&mut ws).await;
|
|
let t4 = recv_text(&mut ws).await;
|
|
|
|
let p1: serde_json::Value = serde_json::from_str(&t1).unwrap();
|
|
let p2: serde_json::Value = serde_json::from_str(&t2).unwrap();
|
|
let p3: serde_json::Value = serde_json::from_str(&t3).unwrap();
|
|
let p4: serde_json::Value = serde_json::from_str(&t4).unwrap();
|
|
|
|
assert_eq!(p1["event_type"], "thinking");
|
|
assert_eq!(p2["event_type"], "tool_started");
|
|
assert_eq!(p2["data"]["call_id"], "call_shell_1");
|
|
assert_eq!(p3["event_type"], "tool_completed");
|
|
assert_eq!(p3["data"]["call_id"], "call_shell_1");
|
|
assert_eq!(p3["data"]["duration_ms"], 42);
|
|
assert_eq!(p4["event_type"], "response");
|
|
|
|
ws.close(None).await.unwrap();
|
|
}
|
|
|
|
/// Regression test: verify session lock is not held during API handler operations.
|
|
///
|
|
/// This test ensures that concurrent API requests (e.g., listing threads) don't
|
|
/// block the agent loop from processing messages. Previously, chat_threads_handler
|
|
/// and chat_history_handler held session locks during slow DB operations, which
|
|
/// would deadlock the agent loop waiting to resolve sessions for incoming messages.
|
|
///
|
|
/// The test verifies that concurrent access to session state completes quickly
|
|
/// without deadlock. If locks are heavily contended, the test will timeout.
|
|
#[tokio::test]
|
|
async fn test_session_lock_not_held_during_api_operations() {
|
|
use ironclaw::agent::SessionManager;
|
|
|
|
let (_addr, _state, _agent_rx) = start_test_server().await;
|
|
|
|
// Create a session manager and attach it to state
|
|
let session_manager = Arc::new(SessionManager::new());
|
|
|
|
// Note: We can't directly modify state.session_manager in the test due to its type.
|
|
// Instead, we test the session manager directly in isolation to verify lock behavior.
|
|
|
|
// Spawn concurrent operations simulating API handler + agent loop interaction
|
|
let mut handles = vec![];
|
|
|
|
// Simulate API handler threads accessing sessions
|
|
for user_id in 0..5 {
|
|
let sm = session_manager.clone();
|
|
handles.push(tokio::spawn(async move {
|
|
for _ in 0..20 {
|
|
let session = sm.get_or_create_session(&format!("user-{}", user_id)).await;
|
|
// Lock and release quickly (simulating API reading session state)
|
|
{
|
|
let _sess = session.lock().await;
|
|
tokio::time::sleep(Duration::from_micros(100)).await;
|
|
}
|
|
}
|
|
}));
|
|
}
|
|
|
|
// Simulate agent loop thread resolving threads
|
|
let sm = session_manager.clone();
|
|
let agent_handle = tokio::spawn(async move {
|
|
for i in 0..20 {
|
|
let (_session, _thread_id) = sm
|
|
.resolve_thread(&format!("user-{}", i % 5), "gateway", None)
|
|
.await;
|
|
// Should not block waiting for API handler locks
|
|
tokio::time::sleep(Duration::from_micros(100)).await;
|
|
}
|
|
});
|
|
handles.push(agent_handle);
|
|
|
|
// Wait for all tasks to complete within reasonable time
|
|
// If session locks are held during slow operations, this will timeout
|
|
let timeout_duration = Duration::from_secs(5);
|
|
let wait_result = timeout(timeout_duration, async {
|
|
for handle in handles {
|
|
let _ = handle.await;
|
|
}
|
|
})
|
|
.await;
|
|
|
|
assert!(
|
|
wait_result.is_ok(),
|
|
"Concurrent session access deadlocked or timed out. \
|
|
This suggests session locks are held too long during I/O operations."
|
|
);
|
|
}
|