Read SteamVR activity from the status row

Record v0.4.45 release verification

docs/evidence/framey-vr-activity-row-2026-10-26.md

@@ -0,0 +1,28 @@
+# Read-only Framey activity gate, 26 October 2026
+
+At 11:12 UTC, a read-only probe acquired `/tmp/frame-test.lock` by atomic
+`mkdir` with a random owner token before reading the Frame. SteamVR
+returned two `vrcmd --stats` rows: performance metrics first and the status
+row second. The status row reported activity level **1**. The original
+`framey-worn-check.py` looked only at row zero and reported unknown activity;
+`framey-vr-check.py` used the same row-zero assumption for its wake gate. The
+fixed gates select the unique status row and fail closed if it is missing,
+duplicated or malformed.
+
+The first locked read found 90% battery, proximity raw 222.25, no active
+managed headset panel service and no managed Decky plugin. A subsequent locked
+read found activity 1 and proximity raw 223.625. No reliable wearer
+confirmation was available. No wake, service install, settings write or input
+was attempted.
+
+While activity was 1, a separate owned-lock read captured four seconds from
+`/dev/video99` into a private 640×360 H.264 MP4. It decoded successfully and
+contained active video imagery. The inspected sample did **not** establish
+room passthrough, Frame Control panel visibility, ProtonDB Badges UI or
+controller interaction. The private recording and decoded frame were removed
+after inspection; no room or media images are committed. Each read released
+only its owner-token lock. The hourly worn-check schedule remains enabled.
+
+The local focused safety suite covers a metrics row before the status row,
+awake activity, ambiguous/missing status and foreign-lock refusal. A genuine
+worn visual and physical-controller check remains outstanding.

docs/evidence/v0445-release-2026-10-26.md

@@ -0,0 +1,61 @@
+# v0.4.45 tagged release, 26 October 2026
+
+The annotated `v0.4.45` tag dereferences to main merge commit
+`212f0af35c86eb651d5d23bfe0b1e6bf50c528b8` from PR #320. The app and
+lockfile both report 0.4.45. [Post-merge main checks](https://github.com/saphid/frame-control/actions/runs/40309949732)
+passed all seven jobs, including iOS and fake Frame end-to-end tests. The
+[tagged release workflow](https://github.com/saphid/frame-control/actions/runs/40310897979)
+passed its draft, Linux, Windows, Apple Silicon Mac, Intel Mac and checksum
+jobs. The tagged Linux builder launched its AppImage under isolated Xvfb;
+that does not verify a physical display, FUSE mount or headset connection.
+
+## Frame and package gates
+
+The [candidate evidence](v0445-candidate-2026-10-26.md) records two
+shared-lock real Frame smokes: one on the combined main code and one after
+the 0.4.45 version bump. Each finished **12 passed / 1 optional x86 launch
+skipped / 0 failed**, with owned cleanup and lock release. The optional
+x86_64 launch was skipped because Steam Linux Runtime 4 is absent. The
+candidate also records the unworn sample Framey exercise on the combined
+code. Synthetic OpenVR input, setting persistence and scroll-button
+movement passed; native synthetic wheel scrolling did not. The only change
+after the version-bumped smoke was its evidence paragraph.
+
+The draft had ten installers plus `SHA256SUMS`, each with a GitHub SHA-256
+digest. A fresh downloaded checksum list contained exactly those ten names
+and matching digests; its bytes matched GitHub's digest
+`d1d84f03c09881417610337cc48f0025535c8b0d227d55147f1f2b99592625fb`.
+`scripts/publish-release.sh --dry-run v0.4.45` passed the exact remote tag,
+installer and checksum gates after the release notes were attached.
+
+Freshly downloaded packages passed these independent checks:
+
+| Package | Result |
+| --- | --- |
+| Linux x86_64 AppImage | SHA-256 `478ea6535979f049930351d5c0abba14e3bcc4bcb5e695ee7ee945a5fd5bf72f`; launched under isolated Xvfb, bundled server returned HTTP 200 with `FrameControl/1`, and stopped on exit. |
+| Mac arm64 ZIP | SHA-256 `0204b445bc35596170d85f7a1e05e793f1c1b247f3129fadad6f8c92730cf3e2`; ZIP CRC valid, bundle version 0.4.45, main Mach-O CPU arm64. |
+| Mac x64 ZIP | SHA-256 `88ded8bc492501f467a9c51156661f3a25b0dc6347660caf0d2e230bda0f984e`; ZIP CRC valid, bundle version 0.4.45, main Mach-O CPU x86_64. |
+| Windows x64 Setup | SHA-256 `c512d322b2022a233ed26aaa03e038b9a735a2df654ec7a3ebefc1ba86b04334`; PE/Nullsoft installer format. The tagged Windows CI also checked its packaged files and Python libraries. |
+| Linux arm64 DEB | SHA-256 `d08cc96dd18b1a602bd0af33a8e6d1f380d94195aec0f83b2898098956aeedc3`; package version 0.4.45 and architecture arm64. |
+
+Tagged Mac and Windows packages were not launched on physical machines in
+this release gate. Earlier Intel Mac acceptance covers older bytes.
+
+## Publication and limits
+
+`scripts/publish-release.sh v0.4.45` published
+[v0.4.45](https://github.com/saphid/frame-control/releases/tag/v0.4.45) at
+2026-10-26 11:02:08 UTC. GitHub's latest-release API returned
+`draft=false`, `prerelease=false`, the expected tag, ten installers,
+`SHA256SUMS` and `update.json`. A fresh public latest-download of
+`update.json` reported version 0.4.45, the stable release page, ten hashed
+assets and the worn-VR limits in its notes.
+
+The official exact-version Decky catalog remains **4 works / 49 partial /
+52 broken / 11 N/A** out of 116. AlarMe's tested silent timer, preset and
+stopwatch workflows do not establish its audio, reminders, sleep controls,
+closed-panel notifications or VR panel. The real Frame exercise used the
+repository sample, not an official plugin. No wearer confirmation was
+available. Physical controller use, optical readability, official-plugin
+VR rendering, audible output and the VR keyboard remain unverified. The
+hourly shared-lock worn-check schedule stays enabled.

scripts/framey-vr-check.py

@@ -90,6 +90,15 @@ def settings(state):
     return pause, timeout
 
 
+def steamvr_activity(rows):
+    """Read the unique status row; metrics can precede it in --stats output."""
+    if not isinstance(rows, list):
+        return None
+    levels = [row.get('activity_level') for row in rows
+              if isinstance(row, dict) and row.get('operation') == 'status']
+    return levels[0] if len(levels) == 1 and type(levels[0]) is int else None
+
+
 def diagnostic_source():
     return (ROOT / 'scripts/framey_openvr.py').read_text() + '''
 import json
@@ -183,9 +192,7 @@ def main(argv=None):
         if not before['battery'] or min(before['battery']) < 20:
             raise RuntimeError('Battery unknown or below 20%')
         if args.wake or args.exercise:
-            stats = before.get('stats')
-            if (not isinstance(stats, list) or not stats or
-                    not isinstance(stats[0], dict) or stats[0].get('activity_level') != 3):
+            if steamvr_activity(before.get('stats')) != 3:
                 raise RuntimeError('Frame is awake or SteamVR activity is unknown; active session untouched')
         if args.wake or args.exercise:
             # Independent systemd timer restores power even if the controller is

scripts/framey-worn-check.py

@@ -44,8 +44,11 @@ for item in pathlib.Path('/sys/class/power_supply').glob('*/capacity'):
             power.append({'name':item.parent.name,'percent':int(item.read_text().strip())})
     except (OSError,ValueError): pass
 stats=subprocess.run(['/opt/steamvr/bin/linuxarm64/vrcmd','--stats'],capture_output=True,text=True,timeout=5)
-try: activity=json.loads(stats.stdout)[0]['activity_level'] if stats.returncode == 0 else None
-except (ValueError,KeyError,IndexError,TypeError): activity=None
+try:
+    rows=json.loads(stats.stdout) if stats.returncode == 0 else None
+    levels=[row.get('activity_level') for row in rows if isinstance(row,dict) and row.get('operation')=='status'] if isinstance(rows,list) else []
+    activity=levels[0] if len(levels)==1 and type(levels[0]) is int else None
+except (ValueError,TypeError): activity=None
 panels=subprocess.run(['/opt/steamvr/bin/linuxarm64/vrcmd','--overlays'],capture_output=True,text=True,timeout=5)
 panel=next((line.strip() for line in panels.stdout.splitlines() if 'valve.steam.desktopgame.2001947813' in line),None)
 base=pathlib.Path.home()/'.local/share/frame-control/decky/plugins'

tests/test_framey_vr_check.py

@@ -15,6 +15,18 @@ spec.loader.exec_module(check)
 
 
 class Safety(unittest.TestCase):
+    def test_status_row_can_follow_metrics_and_ambiguous_status_fails_closed(self):
+        self.assertEqual(check.steamvr_activity([
+            {'operation':'performance','frame_presents':12},
+            {'operation':'status','activity_level':3}]), 3)
+        self.assertEqual(check.steamvr_activity([
+            {'operation':'performance','frame_presents':12},
+            {'operation':'status','activity_level':1}]), 1)
+        self.assertIsNone(check.steamvr_activity([
+            {'operation':'status','activity_level':3},
+            {'operation':'status','activity_level':1}]))
+        self.assertIsNone(check.steamvr_activity([{'activity_level':3}]))
+
     def test_overlay_wait_is_bounded_and_stops_when_visible(self):
         from unittest.mock import Mock
         remote = Mock()
@@ -61,7 +73,7 @@ class Safety(unittest.TestCase):
     def test_power_restore_after_partial_setter_failure(self):
         state = {'battery':[99], 'pause':'power.pauseCompositorOnStandby=true',
                  'timeout':'power.turnOffScreensTimeout=17.000000','dashboardVisible':False,
-                 'stats':[{'activity_level':3}], 'proximity':[6.25]}
+                 'stats':[{'operation':'status','activity_level':3}], 'proximity':[6.25]}
         with tempfile.TemporaryDirectory() as folder, patch.object(check, 'Remote') as factory:
             remote = factory.return_value
             remote.python.return_value = state
@@ -78,7 +90,9 @@ class Safety(unittest.TestCase):
             self.assertIn('rmdir /tmp/frame-test.lock', commands[-1])
 
     def test_awake_or_unknown_activity_refuses_wake_and_install(self):
-        for stats in ([{'activity_level':1}], [{'activity_level':2}], [], None):
+        for stats in ([{'operation':'status','activity_level':1}],
+                      [{'operation':'performance'}, {'operation':'status','activity_level':2}],
+                      [], None):
             state = {'battery':[99], 'pause':'power.pauseCompositorOnStandby=true',
                      'timeout':'power.turnOffScreensTimeout=17.000000',
                      'dashboardVisible':False, 'stats':stats, 'proximity':[0.0]}

tests/test_framey_worn.py

@@ -3,7 +3,9 @@ import sandbox  # noqa: F401
 import contextlib
 import importlib.util
 import io
+import json
 from pathlib import Path
+from types import SimpleNamespace
 import unittest
 from unittest.mock import patch
 
@@ -61,3 +63,22 @@ class WornGate(unittest.TestCase):
 
     def test_screenshot_palette_counts_distinct_colors(self):
         self.assertEqual(worn.color_count('0 #ff00aa 1 #ff00aa 2 #00ff00'), 2)
+
+    def test_probe_reads_activity_from_status_after_metrics(self):
+        rows = [{'operation':'performance','frame_presents':12},
+                {'operation':'status','activity_level':1}]
+        def run(argv, **_):
+            if argv[-1] == '--stats':
+                return SimpleNamespace(returncode=0, stdout=json.dumps(rows))
+            return SimpleNamespace(returncode=1, stdout='')
+        for current, expected in (
+            (rows, 1),
+            ([{'operation':'performance','frame_presents':12}], None),
+            (rows + [{'operation':'status','activity_level':3}], None),
+        ):
+            with self.subTest(rows=current):
+                rows = current
+                out = io.StringIO()
+                with patch('subprocess.run', side_effect=run), contextlib.redirect_stdout(out):
+                    exec(worn.PROBE, {})
+                self.assertEqual(json.loads(out.getvalue())['activity'], expected)