@botbuddy/cli 1.61.6 → 1.61.8

package/daemon-release/darwin-arm64.json

@@ -1,7 +1,7 @@
 {
-  "release_version": "1.61.6",
-  "source_commit": "1e8d5f51d3f9be35c44e793551defaeaf8ae2d23",
-  "source_sha256": "0156e8ade4a62f2a6c4cb29d57fa0f5c23600ba264093198900c22e4c6fc0d68",
+  "release_version": "1.61.8",
+  "source_commit": "c781ee6542c93c05a5ec08e7348fddc0330f33d8",
+  "source_sha256": "b1b8c8cfaa0f4c91f023a78360d2039089885a65367182f00df7dd4312865608",
   "target": "darwin-arm64",
-  "sha256": "8bc316711530d7900ff4186b4eb114cf48ce255344d5d3436962c8d6bee46810"
+  "sha256": "cdaab9fc0b5116cbed342c31c5e7c6544022d8c927cd987daa6d9c6986cc99c4"
 }

package/daemon-release/linux-x64.json

@@ -1,7 +1,7 @@
 {
-  "release_version": "1.61.6",
-  "source_commit": "1e8d5f51d3f9be35c44e793551defaeaf8ae2d23",
-  "source_sha256": "0156e8ade4a62f2a6c4cb29d57fa0f5c23600ba264093198900c22e4c6fc0d68",
+  "release_version": "1.61.8",
+  "source_commit": "c781ee6542c93c05a5ec08e7348fddc0330f33d8",
+  "source_sha256": "b1b8c8cfaa0f4c91f023a78360d2039089885a65367182f00df7dd4312865608",
   "target": "linux-x64",
-  "sha256": "3bf5b7a6396da374f12abada7e42685ad8f783366c5c4de885da21d9749b257b"
+  "sha256": "4047d1635cace367c281d3d8c68268d45ffba24cbae48ff07f08cb384985ca03"
 }

package/daemon-release/manifest.json

@@ -1,21 +1,21 @@
 {
   "schema_version": 1,
-  "release_version": "1.61.6",
-  "source_commit": "1e8d5f51d3f9be35c44e793551defaeaf8ae2d23",
-  "source_sha256": "0156e8ade4a62f2a6c4cb29d57fa0f5c23600ba264093198900c22e4c6fc0d68",
+  "release_version": "1.61.8",
+  "source_commit": "c781ee6542c93c05a5ec08e7348fddc0330f33d8",
+  "source_sha256": "b1b8c8cfaa0f4c91f023a78360d2039089885a65367182f00df7dd4312865608",
   "targets": {
     "darwin-arm64": {
       "goos": "darwin",
       "goarch": "arm64",
       "binary": "daemon/bin/darwin-arm64/botbuddy-daemon",
-      "sha256": "8bc316711530d7900ff4186b4eb114cf48ce255344d5d3436962c8d6bee46810",
+      "sha256": "cdaab9fc0b5116cbed342c31c5e7c6544022d8c927cd987daa6d9c6986cc99c4",
       "mode": "0755"
     },
     "linux-x64": {
       "goos": "linux",
       "goarch": "amd64",
       "binary": "daemon/bin/linux-x64/botbuddy-daemon",
-      "sha256": "3bf5b7a6396da374f12abada7e42685ad8f783366c5c4de885da21d9749b257b",
+      "sha256": "4047d1635cace367c281d3d8c68268d45ffba24cbae48ff07f08cb384985ca03",
       "mode": "0755"
     }
   }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@botbuddy/cli",
-  "version": "1.61.6",
+  "version": "1.61.8",
   "description": "BotBuddy \u2014 Swarm coordination CLI for multi-agent workflows",
   "type": "module",
   "bin": {

package/src/test-lane.mjs

@@ -9,8 +9,13 @@
 // the lane keeps executing detached. A worker (later phase) tails the reporter's
 // NDJSON and flushes per-case verdicts through `sync_test_run_cases`.
 //
-// Backend down? The lane still runs; the JSON line says telemetry:"unavailable",
+// Test-run telemetry down (create_test_run fails) but the durable launcher
+// accepts? The lane still runs; the JSON line says telemetry:"unavailable",
 // wait:null, and the child exit code is preserved.
+//
+// Launcher refused (BOT-2685)? Nothing runs. The line carries the launcher's
+// own outcome/error/recovery and exit code, wait:null, and the pre-created run
+// is closed `completed` with metadata.launch_failed so its waiters wake.
 
 import { randomUUID } from "node:crypto";
 import { existsSync, readFileSync } from "node:fs";
@@ -45,6 +50,7 @@ export function parseTestArgs(argv, { env = process.env } = {}) {
     environment: "local",
     ticket: null, pr: null, repo: null,
     laneKind: null,
+    rerunReason: null,
     wait: false, json: false,
   };
   const errors = [];
@@ -66,6 +72,17 @@ export function parseTestArgs(argv, { env = process.env } = {}) {
       case "--pr": { const raw = value(flag, i); i++; opts.pr = raw == null ? null : Number(raw); if (raw != null && !Number.isInteger(opts.pr)) errors.push("--pr must be an integer"); break; }
       case "--repo": opts.repo = value(flag, i); i++; break;
       case "--lane-kind": opts.laneKind = value(flag, i); i++; break;
+      // BOT-2685: the durable launcher refuses an identical repeat without a
+      // reason; validate the shape here, the server validates the code.
+      case "--rerun-reason": {
+        const raw = value(flag, i); i++;
+        if (raw == null) break;
+        let parsed = null;
+        try { parsed = JSON.parse(raw); } catch { /* reported below */ }
+        if (!parsed || typeof parsed !== "object" || Array.isArray(parsed)) errors.push("--rerun-reason must be a JSON object, e.g. '{\"code\":\"flaky_suspected\"}'");
+        else opts.rerunReason = raw;
+        break;
+      }
       case "--wait": opts.wait = true; break;
       case "--json": opts.json = true; break;
       default:
@@ -140,18 +157,44 @@ export function defaultGitInfo({ cwd = process.cwd(), ticket = null, pr = null,
 
 // The production lane launcher: run the lane through the durable `botbuddy run`
 // worker, carrying the telemetry env into its detached child.
-async function defaultLaunchLane({ command, sessionId, sessionToken, environment, expectedDurationSeconds, childEnv, testRun, cwd, call }) {
+async function defaultLaunchLane({ command, sessionId, sessionToken, environment, expectedDurationSeconds, childEnv, testRun, cwd, call, rerunReason }) {
   // BOT-1572/1582: pass --session-id only when a plain id is used; a token-only lane
   // relies on the inherited $BOTBUDDY_AGENT_SESSION_TOKEN (and an explicit --agent-session-token
   // so the child never falls back to a machine credential).
   const argv = [
     ...(sessionId ? ["--session-id", sessionId] : []),
     ...(sessionToken ? ["--agent-session-token", sessionToken] : []),
+    ...(rerunReason ? ["--rerun-reason", rerunReason] : []),
     "--environment", environment, "--category", "validation", "--kind", "full_suite",
     "--expected-duration", String(expectedDurationSeconds ?? 0), "--", ...command,
   ];
   const result = await launchRun(argv, { cwd, call: sessionToken ? null : call, childEnv, testRun });
-  return { exitCode: result.exitCode, runId: result.receipt?.run_id ?? null };
+  return { exitCode: result.exitCode, runId: result.receipt?.run_id ?? null, receipt: result.receipt ?? null };
+}
+
+// BOT-2685: the BOT-816 rerun policy refusals carry no recovery of their own,
+// and they are the usual reason a lane retry on an unchanged commit is refused.
+const REFUSAL_RECOVERY = Object.freeze({
+  rerun_reason_required: "this lane already ran on this commit and environment; retry with --rerun-reason '{\"code\":\"flaky_suspected\"}' (codes: source_changed|environment_changed|focused_reproducer_passed|flaky_suspected|infrastructure_failure|dependency_drift|approval_retry|user_requested|other+detail), or commit a change",
+  // owner_lost: the command registered (so a plain retry is now a repeat) but
+  // its detached worker never started.
+  launcher_failed: "the lane's worker process failed to start; retry with --rerun-reason '{\"code\":\"infrastructure_failure\"}'",
+  full_suite_rerun_blocked: "this full suite already failed twice on this commit; commit a fix, or prove a focused reproducer with `bb run --kind focused` and retry with --rerun-reason '{\"code\":\"focused_reproducer_passed\",\"focused_reproducer_run_id\":\"<run_id>\"}'",
+});
+
+// BOT-2685: a refused launch has no worker, so nothing else will ever complete
+// the run created for it. Close it here, once, with the refusal on the record.
+// A denied or failed update is reported as such — never retried into a claim.
+async function reconcileRefusedRun(update, { testRunId, laneName, runner, outcome, error, exitCode, commandRunId }) {
+  let res;
+  try {
+    res = await update("update_test_run", {
+      test_run_id: testRunId, status: "completed",
+      metadata: { lane: laneName, runner, launch_failed: true, launch_outcome: outcome, launch_error: error, exit_code: exitCode, command_run_id: commandRunId },
+    });
+  } catch (e) { res = { ok: false, error: e instanceof Error ? e.message : String(e) }; }
+  if (res?.ok && !res.isError) return { test_run_reconciled: true };
+  return { test_run_reconciled: false, reconcile_error: res?.data?.error ?? res?.error ?? "update_test_run_failed" };
 }
 
 function waitCommand(testRunId, sessionId, sessionToken) {
@@ -288,12 +331,41 @@ export async function launchTestLane(argv, {
   const testRun = testRunId
     ? { test_run_id: testRunId, events_path: eventsPath, lane: laneName, runner: resolved.lane.runner, git_sha: git.sha ?? null, branch: git.branch ?? null }
     : null;
-  const launch = await launchLane({
-    command: [...resolved.lane.command, ...extra],
-    sessionId: opts.sessionId, sessionToken: opts.sessionToken, environment: opts.environment,
-    expectedDurationSeconds: resolved.lane.expected_duration_seconds ?? 0,
-    childEnv, testRun, cwd, call, testRunId, lane: laneName, runner: resolved.lane.runner,
-  });
+  let launch;
+  try {
+    launch = await launchLane({
+      command: [...resolved.lane.command, ...extra],
+      sessionId: opts.sessionId, sessionToken: opts.sessionToken, environment: opts.environment,
+      expectedDurationSeconds: resolved.lane.expected_duration_seconds ?? 0,
+      childEnv, testRun, cwd, call, testRunId, lane: laneName, runner: resolved.lane.runner,
+      rerunReason: opts.rerunReason,
+    });
+  } catch (e) {
+    const error = `launch_exception:${e instanceof Error ? e.message : String(e)}`;
+    launch = { exitCode: EXIT_TEST.INTERNAL, runId: null, receipt: { outcome: "rejected", error, exit_code: EXIT_TEST.INTERNAL } };
+  }
+
+  // BOT-2685: the durable launcher's verdict is authoritative. Anything but OK
+  // means no worker exists — say so with its own reason, never "launched".
+  if (launch.exitCode !== EXIT_TEST.OK) {
+    const receipt = launch.receipt ?? {};
+    const outcome = receipt.outcome === "owner_lost" ? "owner_lost" : "rejected";
+    // owner_lost is the one refusal whose command run was registered (and
+    // already marked owner_lost server-side); any other run id never existed.
+    const commandRunId = outcome === "owner_lost" ? launch.runId ?? null : null;
+    const error = receipt.error
+      ?? (Array.isArray(receipt.errors) ? receipt.errors.join("; ") : outcome === "owner_lost" ? "launcher_failed" : "launch_refused");
+    const reconciled = testRunId
+      ? await reconcileRefusedRun(callWithSessionRetry, { testRunId, laneName, runner: resolved.lane.runner, outcome, error, exitCode: launch.exitCode, commandRunId })
+      : {};
+    const recovery = receipt.recovery ?? REFUSAL_RECOVERY[error] ?? null;
+    const line = {
+      outcome, error, ...(recovery ? { recovery } : {}), exit_code: launch.exitCode,
+      test_run_id: testRunId, ...reconciled, command_run_id: commandRunId, lane: laneName,
+      wait: null, receipt_path: null, waits_url: WAITS_URL,
+    };
+    return { exitCode: launch.exitCode, line: JSON.stringify(line), testRunId, commandRunId, opts, refused: true };
+  }
 
   const commandRunId = launch.runId;
   const line = testRunId
@@ -308,7 +380,7 @@ export async function launchTestLane(argv, {
         receipt_path: commandRunId ? receiptPath(commandRunId) : null, waits_url: WAITS_URL,
       };
 
-  return { exitCode: launch.exitCode, line: JSON.stringify(line), testRunId, commandRunId, opts };
+  return { exitCode: launch.exitCode, line: JSON.stringify(line), testRunId, commandRunId, opts, refused: false };
 }
 
 // AC-9 (--wait foreground convenience): block until the run reaches a terminal
@@ -349,7 +421,7 @@ function laneList(cwd) {
 }
 
 function testHelp() {
-  console.log(`botbuddy test <lane> [--session-id <uuid>] [--agent-session-token <token>] [--environment local] [--ticket <KEY>] [--pr <n>] [--repo <owner/repo>] [--lane-kind <${LANE_KINDS.join("|")}>] [--wait] [--json] [-- <extra args>]
+  console.log(`botbuddy test <lane> [--session-id <uuid>] [--agent-session-token <token>] [--environment local] [--ticket <KEY>] [--pr <n>] [--repo <owner/repo>] [--lane-kind <${LANE_KINDS.join("|")}>] [--rerun-reason <json>] [--wait] [--json] [-- <extra args>]
   --agent-session-token <token>  session credential (env $BOTBUDDY_AGENT_SESSION_TOKEN); --agent-key / --session-token are one-release legacy aliases
 botbuddy test list      List the lanes configured in .botbuddy/lanes.json
 botbuddy test help      Show this help
@@ -370,8 +442,8 @@ export async function cmdTest(args, { cwd = process.cwd() } = {}) {
   process.stdout.write(result.line + "\n");
   // --wait (AC-9): after launch, block until the run completes and exit with the
   // lane's exit code; a timeout exits 2 without killing the lane. Only meaningful
-  // when telemetry is live (there is a run to wait on).
-  if (result.testRunId && result.opts?.wait) {
+  // when telemetry is live (there is a run to wait on) and the lane launched.
+  if (result.testRunId && result.opts?.wait && !result.refused) {
     const wait = await waitForLane({ testRunId: result.testRunId, commandRunId: result.commandRunId });
     process.exitCode = wait.exitCode;
     return;