fix(robomp): harden lifecycle contract review cases

python/robomp/tests/test_status_contract.py

@@ -1,8 +1,13 @@
-"""Contract test for the robomp server API response, producing status-contract.json."""
+"""Contract test for the robomp server API response vs status-contract.json.
+
+Regenerate the committed fixture with:
+    ROBOMP_UPDATE_STATUS_CONTRACT=1 pytest tests/test_status_contract.py
+"""
 
 from __future__ import annotations
 
 import json
+import os
 from pathlib import Path
 from typing import Any
 
@@ -14,10 +19,11 @@ from robomp.db import get_database
 from robomp.server import create_app
 
 
-# Runtime/timestamp fields vary every run; normalize them so an ordinary pytest
-# run regenerates a byte-identical committed fixture instead of dirtying it.
+# Runtime/timestamp fields vary every run; normalize them so the live payload
+# can be compared against (or regenerated into) a byte-stable committed fixture.
 _VOLATILE_TS_KEYS = {"received_at", "started_at", "last_tool_ts", "updated_at"}
 _FIXED_TS = "2024-01-01T00:00:00+00:00"
+_UPDATE_ENV = "ROBOMP_UPDATE_STATUS_CONTRACT"
 
 
 def _normalize_for_fixture(value: Any) -> Any:
@@ -191,12 +197,16 @@ def test_status_contract(settings: Settings) -> None:
         # check runtime:
         assert data["runtime"]["repo_allowlist"] == ["octo/widget"]
 
-        # Write fixture:
+        # Compare the normalized payload against the committed fixture. Normal
+        # pytest runs assert equality; regenerate only when ROBOMP_UPDATE_STATUS_CONTRACT=1.
         fixture_path = Path(__file__).parent.parent / "web/test/fixtures/status-contract.json"
-        fixture_path.parent.mkdir(parents=True, exist_ok=True)
-        fixture_path.write_text(
-            json.dumps(_normalize_for_fixture(data), indent=2), encoding="utf-8"
-        )
+        actual = _normalize_for_fixture(data)
+        if os.environ.get(_UPDATE_ENV) == "1":
+            fixture_path.parent.mkdir(parents=True, exist_ok=True)
+            fixture_path.write_text(json.dumps(actual, indent=2), encoding="utf-8")
+        else:
+            expected = json.loads(fixture_path.read_text(encoding="utf-8"))
+            assert actual == expected
 
 
 def _enable_replay(monkeypatch: pytest.MonkeyPatch) -> str:

python/robomp/web/src/types.ts

@@ -8,9 +8,11 @@ export type IssueState =
   | "new"
   | "reproducing"
   | "fixing"
+  | "reviewing"
   | "opened"
   | "merged"
   | "closed"
+  | "needs_info"
   | "abandoned";
 
 export interface RuntimeInfo {
@@ -37,7 +39,7 @@ export interface IssueRow {
   number: number;
   branch: string | null;
   pr_number: number | null;
-  state: IssueState | string;
+  state: IssueState;
   classification: string | null;
   updated_at: string;
   latest_event: LatestEvent | null;

python/robomp/web/src/work-items.test.ts

@@ -470,8 +470,15 @@ describe("buildWorkItems", () => {
     expect(stageOrdinal("nonsense")).toBe(0);
     expect(stageOrdinal("new")).toBe(0);
     expect(stageOrdinal("reproducing")).toBe(1);
+    // needs_info is set by mark_unable_to_reproduce during reproduction and
+    // cleared back to reproducing by repro_record in host_tools; it stays in
+    // the reproduction phase (ordinal 1), not advanced past fixing.
+    expect(stageOrdinal("needs_info")).toBe(1);
     expect(stageOrdinal("fixing")).toBe(2);
+    // reviewing is PR review work (tasks.py sets it on PR-opened issues); it
+    // lives on the PR step, same stage as opened (ordinal 3).
     expect(stageOrdinal("opened")).toBe(3);
+    expect(stageOrdinal("reviewing")).toBe(3);
     expect(stageOrdinal("merged")).toBe(4);
   });
 
@@ -570,4 +577,41 @@ describe("buildWorkItems", () => {
     expect(items[0].latestEvent?.state).toBe("running");
   });
 
+  test("live running event outranks a newer queued latest_event for the same issue", () => {
+    const items = buildWorkItems(
+      status({
+        issues: [
+          issue({
+            key: "owner/repo#22",
+            number: 22,
+            latest_event: latestEvent({
+              delivery_id: "queued-newer",
+              state: "queued",
+              received_at: "2026-06-17T00:09:00Z",
+            }),
+          }),
+        ],
+        running_events: [
+          runningEvent({
+            delivery_id: "live-older-3",
+            issue_key: "owner/repo#22",
+            received_at: "2026-06-17T00:01:00Z",
+            started_at: "2026-06-17T00:02:00Z",
+          }),
+        ],
+      }),
+    );
+    expect(items).toHaveLength(1);
+    expect(items[0]).toMatchObject({
+      key: "owner/repo#22",
+      bucket: "running",
+      deliveryId: "live-older-3",
+      inflightOnly: false,
+      error: null,
+    });
+    expect(items[0].live?.delivery_id).toBe("live-older-3");
+    expect(items[0].latestEvent?.state).toBe("running");
+    expect(items[0].latestEvent?.delivery_id).toBe("live-older-3");
+  });
+
 });

python/robomp/web/src/work-items.ts

@@ -38,8 +38,10 @@ export const SIMPLE_CLASSIFICATIONS: ReadonlySet<string> = new Set([
 const STATE_ORDINAL: Record<string, number> = {
   new: 0,
   reproducing: 1,
+  needs_info: 1,
   fixing: 2,
   opened: 3,
+  reviewing: 3,
   merged: 4,
   closed: 4,
   abandoned: 4,
@@ -78,10 +80,10 @@ export function buildWorkItems(status: StatusResponse): WorkItem[] {
     const latest = issue.latest_event;
 
     // A matching live running_events entry is authoritative over the issue's
-    // own latest_event, which may be a newer failed/done row that the live run
-    // has not yet superseded. Render the live run so the card stays running,
-    // cancel-capable (deliveryId from the live delivery), and free of the
-    // stale failure. Non-live rows keep latest_event authority below.
+    // own latest_event. That summary row can be a newer queued retry/comment,
+    // or a failed/done row the live run has not superseded yet. Render the live
+    // run so the card stays running, cancel-capable (deliveryId from the live
+    // delivery), and free of stale latest-event state.
     if (live) {
       items.push({
         key,