Honest green: skipped is a job status, every failure says why, registry hygiene

CHANGELOG.md

@@ -47,6 +47,39 @@ a CI job fails when they disagree.
 - **`explore-night.ts --dry-run`** opens the tunnel, runs the build and time
   checks against the live brain and prints the plan, without enqueueing,
   posting, warming or booting anything.
+- **A job that ran nothing now has a status of its own: `skipped`.** A runner
+  that could not run a job says `ok: true` with an error beginning `skipped:`,
+  and the queue closed that as `done`, so the job list, the status counts and
+  the digest counted it as a pass: `nightly-tvloop-hardware` and three explore
+  jobs a night "passed" in two seconds for two weeks while testing nothing.
+  The queue now closes it as `skipped`, with the reason in `last_error`; a host
+  row of plain `ok: true` over device rows that all skipped counts the same,
+  and a fan-out family where no child ran is `skipped`. `skipped` is accepted
+  by the `status` filter of `GET /api/jobs` and `GET /api/v1/jobs`, counted in
+  the status facets, `GET /metrics` and the Overview (`skipped 24h`, beside
+  done rather than in it), drawn as a grey chip on the dashboard, and given its
+  own line in the morning digest. The v1 summary's verdict, `fleetr`'s exit 4
+  and the MCP verdict were already `skipped` and are unchanged. A database
+  opened by this release is migrated once: every `done` job with a `skipped:`
+  result is re-labelled by the same rule. Result rows are not touched.
+- **A job waiting on a skipped job skips too.** A skipped build produced no
+  artifact, so the install behind it had nothing to install; it used to be
+  promoted and run against whatever was already on the phone. It now skips
+  with `last_error` `dependency <id> skipped`, the whole way down the chain,
+  and a job enqueued behind an already-skipped one is skipped on arrival. A
+  failed dependency still fails its waiters, and wins over a skipped one.
+- **A `schedule-skipping` alert.** A schedule whose last three runs
+  (`FLEET_ALERT_SKIP_STREAK`) all skipped raises one alert naming the reason it
+  keeps giving, for example "nightly-tvloop-hardware has skipped 3 runs in a
+  row: no Roku developer password on this host", and resolves on the first run
+  that is not a skip. The tvloop nightly needed it for two weeks.
+- **Simulator rows are forgotten after a week offline.** The lease sweep deletes
+  a row whose descriptor says `kind` or `soc` is `simulator` once it has been
+  offline for `FLEET_SIMULATOR_TTL_DAYS` (default 7), because simulators are
+  recreated under new UUIDs and never come back; the live brain listed thirteen
+  of them. Real devices and Android emulators are never forgotten, a row that
+  holds a lock is left alone, and results and beacons are kept. `POST
+  /jobs/sweep` reports what it forgot.
 
 ### Changed
 
@@ -160,6 +193,37 @@ a CI job fails when they disagree.
   `fleet-explore-tv` night "passed" in about a minute. A tvOS simulator is
   now `tvos-26.5` (watchOS and visionOS likewise). No TV night has run with
   the fix yet.
+- **A failing `ui-test` now says which step failed.** `nightly-fleet-ui-smoke`
+  failed on the Galaxy S8+ every night from 10-10 with `failed: 1` and no
+  error, while `Assertion is false: "Fleet Runner" is visible` sat in the JUnit
+  report it had uploaded. The executor now reads the failing flow and its
+  assertion out of the report into the device row's `error` and
+  `test.failing_flow`; when Maestro produces no report, the error is its exit
+  code and last stderr line, never its command line, which carries the test
+  account's password.
+- **No host result says `ok: false` with nothing beside it.** Every host
+  workload's rows go through one wrapper in the executor: a device row with
+  only counts says so ("1 of 1 tests failed; the runner reported no failing
+  step"), and a final host row with no error of its own gets the failing
+  devices' reasons. A subprocess failure nobody caught is reported as its exit
+  code and last stderr line instead of Node's message. The collector also
+  copies a failure's reason into the job's `last_error`, so the job list and
+  the `job-failed` alert show it.
+- **`GET /api/artifacts/<sha256>` answered "no such endpoint".** It is where
+  `DELETE` and `/pin` already live, so it is the URL a person reaches for, and
+  the 404 read as a missing artifact: the ui-smoke JUnit report was in the
+  store all along. It now redirects (308) to
+  `/api/v1/artifacts/<sha256>/content`, and says plainly when no artifact by
+  that hash was ever uploaded.
+- **The executor registered Rokus as `platform: android`.** Its Roku
+  descriptor said `os: roku-15.3.4` but no platform, and the collector's
+  fallback for anything that is not iOS is Android. It now sends
+  `platform: "roku"`; `os ~ 'roku'` targeting and the tvloop workload are
+  unaffected.
+- **A Roku seen first by address and later by serial was listed twice.** A
+  sweep where the box answered SSDP but not device-info left a
+  `roku-ip-192-168-50-218` row offline beside `roku-X01600EMFRJ9`. A
+  serial-named registration now retires the IP-named row at the same address.
 
 ## [0.8.0] — 2026-10-15
 

collector/dash/src/api.ts

@@ -166,7 +166,9 @@ export type Job = {
   job_id: string;
   workload: string;
   executor: "device" | "host";
-  status: "waiting" | "queued" | "claimed" | "done" | "failed" | "cancelled";
+  /** `skipped`: claimed and closed without running anything (no device matched,
+   *  the app was not installed, a dependency skipped). Not a pass. */
+  status: "waiting" | "queued" | "claimed" | "done" | "failed" | "cancelled" | "skipped";
   /** Job ids this one waits for; it stays `waiting` until they are all done. */
   depends_on?: string[] | null;
   /** Whether a higher-priority job may ask this one to checkpoint and step aside. */
@@ -214,6 +216,7 @@ export type Overview = {
     claimed: number;
     done_24h: number;
     failed_24h: number;
+    skipped_24h: number;
     last_attempt: number;
     oldest_queued_age_s: number | null;
   };

collector/dash/src/deviceKind.ts

@@ -7,7 +7,8 @@
  * The model string comes first, because it is the one field that is right
  * about every device on the shelf today. `platform` is not: agents that
  * predate the field are reported as ios/android by the collector's fallback,
- * so a Roku arrives as `android` and an Apple TV or Apple Watch as `ios`.
+ * so an Apple TV or Apple Watch arrives as `ios`, and a Roku registered by an
+ * executor older than 0.9 as `android` (newer ones say `roku`).
  * `kind` is the agent naming its own shape and is believed — but only to pick
  * the family; the model still chooses which member of it (a `tv` that says
  * "Roku Streaming Stick" is drawn as the stick, not a bare panel).

collector/dash/src/pages/JobDetail.tsx

@@ -87,7 +87,7 @@ function DependsOn({ job }: { job: Detail }) {
       .{" "}
       <span class="faint">
         {statusOf(job) === "waiting"
-          ? "It is queued the moment the last of them is done, with any ${jobs.…} references in its spec filled in from their final rows. If one fails or is cancelled, this job fails with it rather than waiting forever."
+          ? "It is queued the moment the last of them is done, with any ${jobs.…} references in its spec filled in from their final rows. If one fails or is cancelled, this job fails with it rather than waiting forever; if one skips, this job skips too, since there is nothing for it to run on."
           : "The chain has already resolved; this is what it hung off."}
       </span>
     </p>

collector/dash/src/pages/Jobs.tsx

@@ -132,7 +132,7 @@ export function Jobs() {
                   // 'waiting' leads because it is the one status a person goes
                   // looking for: a chain that has not started is invisible in
                   // every other view, including the overview's queue depth.
-                  options={["waiting", "queued", "claimed", "done", "failed", "cancelled"].map((s) => ({
+                  options={["waiting", "queued", "claimed", "done", "failed", "skipped", "cancelled"].map((s) => ({
                     value: s,
                     label: `${s} (${d.status_counts[s] ?? 0})`,
                   }))}

collector/dash/src/pages/Overview.tsx

@@ -83,6 +83,9 @@ export function Overview() {
                 <Stat label="running" value={d.queue.claimed} tone={d.queue.claimed > 0 ? "warn" : undefined} />
                 <Stat label="done 24h" value={d.queue.done_24h} tone="ok" />
                 <Stat label="failed 24h" value={d.queue.failed_24h} tone={d.queue.failed_24h > 0 ? "bad" : undefined} />
+                {/* Its own number, never added to done: a skip ran nothing. No tone,
+                    like the chip -- it is not good news and not a failure. */}
+                <Stat label="skipped 24h" value={d.queue.skipped_24h ?? 0} />
                 <Stat label="oldest queued" value={duration(d.queue.oldest_queued_age_s)} />
                 <Stat
                   label="last attempt"

collector/dash/src/style.css

@@ -163,6 +163,10 @@ td.num { text-align: right; font-variant-numeric: tabular-nums; }
 /* Cancelled is deliberately not styled like failed: a person stopped it,
    nothing went wrong. */
 .pill.cancelled { color: var(--ink-faint); border-color: var(--line); text-decoration: line-through; }
+/* Skipped is neither green nor red: the job was claimed and nothing ran, so it
+   says nothing about the code. Grey and solid, so it does not read as a pass at
+   a glance, and does not read as a person's cancellation either. */
+.pill.skipped { color: var(--ink-dim); border-color: var(--line); background: color-mix(in srgb, var(--ink-faint) 12%, transparent); }
 .pill.claimed { color: var(--warn); border-color: color-mix(in srgb, var(--warn) 40%, transparent); }
 .pill.done, .pill.online { color: var(--ok); border-color: color-mix(in srgb, var(--ok) 40%, transparent); }
 .pill.failed, .pill.offline { color: var(--bad); border-color: color-mix(in srgb, var(--bad) 40%, transparent); }

collector/openapi.json

@@ -441,7 +441,7 @@
             "name": "status",
             "in": "query",
             "required": false,
-            "description": "Comma-separated: waiting, queued, claimed, done, failed, cancelled.",
+            "description": "Comma-separated: waiting, queued, claimed, done, failed, cancelled, skipped.",
             "schema": {
               "type": "string"
             }
@@ -2169,7 +2169,8 @@
               "claimed",
               "done",
               "failed",
-              "cancelled"
+              "cancelled",
+              "skipped"
             ]
           },
           "ok": {
@@ -2458,7 +2459,8 @@
               "claimed",
               "done",
               "failed",
-              "cancelled"
+              "cancelled",
+              "skipped"
             ]
           },
           "claimed_by": {
@@ -3571,7 +3573,7 @@
           },
           "status": {
             "type": "string",
-            "description": "What the job would be on enqueue: queued, waiting or failed."
+            "description": "What the job would be on enqueue: queued, waiting, failed (a dependency failed) or skipped (a dependency skipped)."
           },
           "fanout": {
             "type": "array",

collector/schemas/result.schema.json

@@ -218,7 +218,9 @@
       "properties": {
         "passed": { "type": "integer" },
         "failed": { "type": "integer" },
-        "artifacts": { "type": "array", "items": { "type": "string" }, "description": "sha256 refs of screenshots/videos/junit uploaded to the artifact store" }
+        "artifacts": { "type": "array", "items": { "type": "string" }, "description": "sha256 refs of screenshots/videos/junit uploaded to the artifact store" },
+        "skipped": { "type": "integer" },
+        "failing_flow": { "type": "string", "description": "The first failing flow or test case, by name; the step that broke is in the row's error" }
       }
     },
     "shot": {