Record test-only main merge in v0.4.43 candidate

Retry transient Windows wireless inventory smoke timeout

app/package-lock.json

@@ -1,12 +1,12 @@
 {
   "name": "frame-control",
-  "version": "0.4.42",
+  "version": "0.4.43",
   "lockfileVersion": 3,
   "requires": true,
   "packages": {
     "": {
       "name": "frame-control",
-      "version": "0.4.42",
+      "version": "0.4.43",
       "license": "MIT",
       "devDependencies": {
         "electron": "^44.4.5",

app/package.json

@@ -1,7 +1,7 @@
 {
   "name": "frame-control",
   "productName": "Frame Control",
-  "version": "0.4.42",
+  "version": "0.4.43",
   "description": "Desktop app for managing a Valve Steam Frame over SSH",
   "private": true,
   "main": "main.js",

docs/evidence/v0443-candidate-2026-10-26.md

@@ -0,0 +1,7 @@
+# v0.4.43 candidate Frame gate, 26 October 2026
+
+The combined code head `232b253e7a9400aa9829f9fea6a5d2011dd2a8f3` (PR #309 after merging main with PR #308) ran `scripts/frame-smoke.sh` under its own `/tmp/frame-test.lock` owner at 07:57 UTC. Its locked preflight found SteamVR activity 3 (standby), proximity 6.25 and battery 99%. On Frame OS 0.5.6 build `20261022.6153909`, it recorded **12 passed / 1 skipped / 0 failed**. The ARM64 and Proton executables launched; x86_64 launch was skipped because Steam Linux Runtime 4 is absent. It removed its test titles, exited 0 and released its lock. The private detailed result is `/tmp/frame-smoke-pr309/2026-10-26T075752Z-20261022.6153909.json` on the test host. A fresh, separately owned read-only lock then found activity 3 and battery 99%, and was released. PR #310 subsequently changed only the Windows wireless CI smoke test; the release branch merged that test change and bumped the app version, with no further headset runtime code change after this Frame smoke.
+
+PR #308's panel code commit `d5e62f643d462b7070f27c2c47b11b4aa9982094` previously passed the temporary Framey sample `scripts/framey-vr-check.py --exercise` machine check, with cleanup and power restoration; see [Web Browser follow-up](framey-web-browser-2026-10-26.md). That test used synthetic OpenVR input and a repository sample, not the official Web Browser plugin or a person wearing the headset. No wake or VR input exercise ran on this combined code head because wearer confirmation was absent. The current smoke made no worn visual, physical-controller, audible-output or official-plugin VR claim.
+
+This is candidate evidence, not a published-release result. Publication still needs all exact-head PR and release checks, ten tagged installers with matching `SHA256SUMS`, package validation, the publisher dry run, and public updater readback. The exact-version official Decky catalog in this candidate is **4 works / 49 partial / 52 broken / 11 N/A**. Volume Boost's works claim covers tested muted sink controls; Junk-Store remains partial, and Web Browser remains broken.

tests/test_wireless.py

@@ -192,6 +192,12 @@ class Probe(unittest.TestCase):
     @unittest.skipUnless(wireless.frame_host.WINDOWS, 'Live read-only Windows inventory runs on Windows CI')
     def test_live_windows_inventory(self):
         r = wireless.diagnose()
+        # Shared Windows runners can spend the entire probe budget starting
+        # PowerShell. Retry only that transient timeout; a second failure must
+        # still fail this live inventory gate.
+        if (r['state'] == 'unavailable' and
+                r.get('message') == 'Windows wireless inventory timed out. No settings were changed.'):
+            r = wireless.diagnose()
         self.assertIn(r['state'], ('ok', 'partial'), r)
         self.assertEqual(r['sources']['adapters'], 'ok', r['sources'])
         self.assertTrue(r['devices'], 'Windows runner should expose at least its virtual network interface')