MichaelScottsman--LiquidAss-Frame/docs/phase2/wp/C5b.md

C5b Now Playing, binding UI, VR bindings link: evidence log

Package card: docs/phase2/PLAN.md §2.4 "C5b Now Playing, binding UI, VR bindings link". Concept: docs/phase2/concepts/game-pages.md (GP §4.12, §4.13, §4.14, §9.7, §9.8, §10; owned by C5a, built here). Steam build at the time of writing: 11094443.

Status

Session 1 complete (2026-10-07 10:25-12:05, expedited, one session; build 11094443).

• M0 note done (C5b-D1..D5). M2 T1 done: theme/vr/20-nowplaying.css (GP §4.13 on a mock: no scene app may be started), theme/vr/40-bindings.css (§4.14; app list, binding list and binding view gates PASS). M3 T2 done: device/vr/systemui.nowplaying.js (flag wp.c5b; AT-NP-VARIANTS PASS with the flag on, no keys and no red with it off).
• T3 built, stays OFF: device/rt/52-vrbind.js + device/shell_ext/vrbind.py (flag vrBindings): capsule in %{AppButtons}, D-pad reachable and reversible, laser → relay (dry), A → ConfirmModal, gates PASS (pad; laser clean for the capsule). Not run: AT-BIND-RETURN (live deep link), AT-T3-ROUTE; Steam Input page link not built. Q-A pending, so not ready.
• Native: no fragments by design (SteamVR overlays have no glassd surface; the capsule is 0 mm content under C5a's cover). The native-session check was attempted twice and blocked by lock contention (log below); left for V1.
• Open REQs: C5b->C1a (controller-status rims, a 0.04 s transition in vr:systemui), C5b->C5a (private-app indicator size/hit).

READY: wp.c5b — evidence: AT-NP-VARIANTS (keys from SteamVR's own strings, re-applied within one frame after a re-render, Exit red only when keyed, nothing red with the flag off); gates vr:systemui --flags wp.c5b in pad and laser on the keyed mock: my elements SIZE / TYPE / AUD / edge probes all PASS (the page-level FAIL is C1a's controller status and a foreign transition); no script error in the daemon's vrScripts status; nothing left after removal. The T1 CSS is unflagged and already live. vrBindings is NOT ready.

Requests

• REQ C5b->C1a: gates vr:systemui (both modes, 10:33-10:55) fails OUTLINE on %{ControllerStatusRoot} %{LargeStatusArea}(::before) and %{ControllerStatusRootRight}::before ("closed rim of 3 shadow lines": the Phase 1 --lgs-panel-rim-style insets 1.5 px top + 1 px sides) and MOTION on one transition-duration 0.04s (not in any theme file I could find). Both are outside Now Playing; they make every vr:systemui gate run FAIL for the other SteamVR areas too.
  • C1a answer (2026-10-07, R2 fix pass, session 6): done. theme/vr/10-systemui.css no longer draws a closed rim anywhere: the controller card, the gamepad-mode pill, More Options and the section 5 panels use P4's panel / thick glass with the E4/E5 dark edge and the lit top lip only (no Phase 1 --lgs-panel-rim / --lgs-control-rim, no Phase 1 alias left in the file). gates vr:systemui --flags wp.c1a --mode pad and --mode laser (17:49, native=off): AUD, SIZE, TYPE, OUTLINE, MOTION all PASS (shots p2_c1a_r2f_vr_{pad,laser}.png). The 0.04 s MOTION finding is the pre's clone of SteamVR's own :active button rule (your reviewer's N1): C1a's theme/vr/pre/sys_hover.js now clones :hover rules without their transition* declarations; the same fix is needed in your np_states.js and C3a's sys_zoo.js (with a stripped scratch copy of sys_zoo.js the gate is green).
• REQ C5b->C5a: on /library/app/620980 Steam's 30 x 30 private-app indicator (%{PrivateAppActiveIndicator}, the red eye badge after Manage) fails P-80 and P-08 (hit 49 % own, 14 % Manage) in laser mode. My capsule now keeps 72 px clear of it. — C5a (R2 fix 1, 2026-10-07): done. The badge is Steam's indicator (focusable: false, its one function a laser hover tooltip). It is now a 114 px round hover target (x 822–936, y 416–530) that starts 1 px clear of Manage's box (741–821), with Steam's own red circle as the 30 px disc in its upper left (x 835–865, y 429–459, beside the gear's top right) and the eye centred on it. L.gates.hitStats over the gate's 114 × 80 box: 97 % own, 0 % other (19:58); Manage keeps 100 % own. gates main --route /library/app/620980 --flags wp.c5a,wp.c1a,wp.p3 pad 19:45 and laser 19:46: SIZE PASS (both runs listed the badge as part of its host; an 80 px round target sampled at 19:29 read 92 % own, hence 114). Your capsule sits before Manage, so the badge's target moves with Manage and never reaches it.

Requests to C5b, handled

• REQ P4->C5b (40-bindings.css hover rim restated ring-free; NP focus re-shot): see log.
• REQ C4a->C5b (dark label on the green Resume capsule in 20-nowplaying.css; re-render NP mockups): label done in CSS; mockup re-render skipped (expedited mode, no re-renders).

Log

2026-10-07, session 1: M0 conformance note (PLAN §1 decides; no mockup re-renders)

Start-of-session request check: grep -n "REQ [A-Za-z0-9]*->C5b:" docs/phase2/wp/*.md → P4 (bindings hover rim, NP focus), C4a (dark label on green).

#	GP said	PLAN §1 / request decides	Built as
C5b-D1	NP focus = + white .14 inner illumination	§1.4 / tokens §5: --lgs-vr-focus-ring (focus add .32 as an inset layer + lit arc + glow, no ring); SteamVR's %{FocusRing}.gpfocus::after outline cleared	T1
C5b-D2	Resume label white on green	REQ C4a->C5b, CTL C-D21: --lgs-text-on-color (#0d0e12) on the lit green	T1
C5b-D3	Sheets +30 → +50 mm (GQ6); Resume crop +15 in the mockup	§1.7 default profile: SteamVR pages are not reported surfaces (no cover, plate or pop fragments exist for vr:systemui / vr:controllerbindingui); depth stays SteamVR's own. No layer/popup fragment for these pages	none (documented)
C5b-D4	Binding page hover rim restated as a closed ring	REQ P4->C5b: lit top arc only at S 1.875	T1
C5b-D5	Capsule and Steam Input link	Off behind vrBindings (runtime.md S13) until Q-A; never in READY	T3

2026-10-07, session 1: M2 T1, M3 T2/T3 (build 11094443)

No scene app was running (driver.py status: inactive) and agents never launch anything, so Now Playing was verified on a mock: theme/vr/pre/np_states.js now builds a display-only copy of the panel from SteamVR's own global classes and localized strings at atlas (0, 700), outside every idle panel's UV rect (inventory §2.2), removed after keep ms; on a real Now Playing it marks the live panel instead. Its buttons carry role=button only so G-SIZE judges them (SteamVR's own have no role, so the gates never size-check the real ones).

Files (all owned): theme/vr/20-nowplaying.css (rewritten), theme/vr/40-bindings.css (Phase 1 look kept, rims made ring-free, §9 Phase 2 sizes and type added), theme/vr/pre/np_states.js (mock + states), device/vr/systemui.nowplaying.js (T2 keyer, // @lgs-flag wp.c5b), device/rt/52-vrbind.js (T3 capsule, flag vrBindings), device/shell_ext/vrbind.py (relay plugin, actions vrbind nav and vrbind.strings read, both behind vrBindings). The Phase 1 versions are in git history (no copies kept).

Now Playing (GP §4.13): panel = window glass (--lgs-mat-window-bg + the E3 lobe, no top line), art 540 × 810 r 29 at x 218 (Steam's --width), column 760 at x 880, title 92 Bold two lines max with SteamVR's "Now Playing" caption (T2), Resume 760 × 128 green capsule with the 44 px play glyph and the dark label (REQ C4a), sheet rows 116 in one recessed platter (black .14, r 52, pad 8, 2 px separators, 64 px indigo / blue / grey chips only when T2-keyed, chevrons), Exit 760 × 116 capsule: red label and red fill on hover / focus only with data-lgs-np="exit". Focus = SteamVR's .gpfocus lit with the focus add (+ spot, glow on coloured fills), the Steam-owned ::after ring and its blinker removed. Sheets (VR Controller Bindings / VR Video Settings modals) are not restyled: their DOM was never opened (opening them is forbidden, GQ15).

Binding UI (GP §4.14): capsules 112 visible + 150 hit (transparent ::after, 19 px above and below), View / Activate 40 apart, title bar 160 with Back / Options capsules (the "Back" text stays visible: Q-B fallback, no HIDDEN exception needed), text ≥ 34 px and ≥ Medium in title case (G-TYPE m 1.87), body text ≥ 42 px (descriptions 3 lines max, meta one line), tabs one line (horizontal scroll kept) with the orange unbound-count capsule, 2 px separators. Fix of a reachability bug found on the way: with taller controls the sticky controller-image column (977 px) outgrew the 904 px scroller, so Poses / Haptics could never be reached; images now max 600 px (column 830). The single-column binding list and the blue "Select" tint were not built (identifying Select needs a T2 key; neutral capsules instead).

Check	Command	Result	Evidence
G-AUD/SIZE/TYPE/OUTLINE/MOTION, binding app list	gates vr:controllerbindingui --pre bind_apps_more.js	PASS	shots/p2_c5b_bind_apps.png
same, binding list	gates vr:controllerbindingui --pre bind_list.js	SIZE/TYPE/OUTLINE/MOTION PASS; AUD 1 SHRUNK (a long description clamped to 2 lines) → clamp 3 (final re-run below)	shots/p2_c5b_bind_list.png
same, binding view	gates vr:controllerbindingui --pre bind_view.js	PASS (was SIZE 18, TYPE 84, OUTLINE 24, AUD 27 fails at the start)	shots/p2_c5b_bind_view.png
Now Playing mock, pad (focus on Bindings) and laser (hover on Exit, perf variant)	gates vr:systemui --flags wp.c5b --mode pad|laser --pre "window.__LGS_NP={…}, $(cat theme/vr/pre/np_states.js)"	Mock: SIZE, TYPE, AUD PASS; every NP edge probe passes (panel, Resume, Exit: lobe, no top line). Page-level FAIL only from others' areas: %{ControllerStatusRoot} rims (P-42) and one 0.04 s transition (not in any theme file) → REQ to C1a	shots/p2_c5b_np_pad.png, p2_c5b_np_mock.png
AT-NP-VARIANTS (T2)	js in vr:systemui --flags wp.c5b: four mocks (normal, arcade, no-quit, perf; normal with two hidden modal siblings), keys read, children replaced, keys read after one rAF	Keys exactly resume / bindings / video / [perf] / exit; Exit label rgb(255,66,69) only when present; Exit .gpfocus label --lgs-text-on-color on the red fill; after the re-render every key back within one frame. Flag off: no key, no red label anywhere	scratch npvar_on.json, npvar_off.json
AT-T3 (capsule, dry)	js on /library/app/620980 --flags vrBindings,actionsDryRun --mode pad	Capsule is the last child of %{AppButtons} ("Controller Bindings", 80 tall); laser click logged direct once, daemon replied {ok, dry: true} (nothing navigated); A handler opened Steam's ConfirmModal titled "Controller Bindings", B closed it (0 modals left). .gpfocus after B not verified (the handler was called directly)	—
D-pad on the action row	js (L.pad right ×6, left ×6)	Play → Play from → Configure Controller → Manage → capsule (Right stays); Left walks back the same way	—
Capsule gates	gates main --route /library/app/620980 --only size,type,outline --flags vrBindings,actionsDryRun --mode laser|pad	pad PASS; laser: capsule clean, the only fails are Steam's 30 px private-app indicator (P-80, and P-08 against Manage): C5a's area. First runs found the capsule overlapping that indicator's hit box: fixed with a 72 px gap when the indicator is present	shots/p2_c5b_gp_capsule_pad.png, _laser.png
pad-bfs /library/app/620980	pad-bfs --flags vrBindings,actionsDryRun	Not usable this run: another agent's native session was on and most nodes were untakeable (Steam's own included); replaced by the D-pad walk above	scratch bfs.json
pad-bfs /library/app/620980 (retry 11:20)	pad-bfs --route /library/app/620980 --flags vrBindings,actionsDryRun --max 30	31 nodes; the capsule is reached (Manage → Right) and goes back (Left → Manage); Up → Back, Right → edge. FAIL overall from C5a's page (Last Played, the tab arrows, Post, Load More, three event cards unreached) plus two of mine: (1) the capsule's label changed mid-sweep (aria-label switched to SteamVR's "VR Controller Bindings" when the string retry landed) so it was keyed twice: fixed, no aria-label (the visible label is the name); (2) capsule Down → an event card whose Up returns to the card above it, not the capsule: Steam's spatial nav into the activity list (Manage's Down goes to Activity)	scratch bfs2.json
Binding list after the meta fix	gates vr:controllerbindingui --pre bind_list.js + shot … p2_c5b_bind_list	PASS; meta and descriptions wrap inside their entries (a one-line meta had run over the next entry and made the page scroll sideways; the gate did not see it, the shot did)	shots/p2_c5b_bind_list.png

Not built / not run this session: the Steam Input page link (GP §4.12 step 2, AT-T3-ROUTE); AT-BIND-RETURN (a live deep link: inputUI.OnShowAppBinding calls SetSelectedApp(key, true), whose second argument was not traced, so no real call was made without the dry-run logger on controllerBindingStore); the NP sheets. vrBindings therefore stays off and is not offered as ready (Q-A also requires AT-BIND-RETURN and AT-T3-ROUTE).

Native fragments: none, by design. vr:systemui and vr:controllerbindingui are SteamVR overlays with no glassd surface (no cover, plate or pop can be reported for them; PLAN 1.6 "T5 cover once systemui covers exist"), so their glass is the CSS look in both modes. The capsule is in-window content in C5a's action row, at 0 mm under C5a's window cover (PLAN 1.7: cluster over art stays 0 until G2 + AT-HV-OFFAXIS), so it needs no layer of its own.

2026-10-07, session 1: final Now Playing pass (11:15-11:25)

• Caption moved from the title to the info column (data-lgs-np-cap on .InfoColumn): inside the title's two-line clamp it took one of the title's lines ("Liquid Glass…"); long row labels wrap to two lines inside the 116 row (the perf row's chevron had been pushed out), left-aligned.
• Exit hover: one colour layer (stacked translucent layers drew a bright 1 px rim that failed the edge profile), no lit arc on the red fill.
• np_states.js option waitKeys (waits up to 4 s for the keyer the daemon injects a moment after the step's flags land; without it a shot can catch the unkeyed fallback).

Check	Command	Result	Evidence
NP mock, perf variant, pad (focus on Bindings)	gates vr:systemui --flags wp.c5b --mode pad --pre "window.__LGS_NP={focus:1,hover:-1,mock:'perf',keep:40000,waitKeys:true}, $(cat theme/vr/pre/np_states.js)"	Mine: SIZE 9/9, TYPE, AUD PASS, every edge probe passes. Page FAIL only from C1a's controller-status rims and the 0.04 s transition (REQ above)	shots/p2_c5b_np_final_pad.png
same, laser (hover on keyed Exit)	same with {focus:-1,hover:4,…} --mode laser	Mine: PASS (after the single-layer fix); same two foreign findings	shots/p2_c5b_np_final_laser.png
Keyed look	shot vr:systemui p2_c5b_np_exit_hover --flags wp.c5b --pre "…{focus:2,hover:4,mock:'perf',waitKeys:true}…"	Viewed: caption, two-line title, green Resume with dark label, indigo / blue / grey chips, focused Video row lit (no ring), Exit red with dark label on hover. Unkeyed fallback also viewed (neutral Exit, no chips)	shots/p2_c5b_np_exit_hover.png
Leftovers	js in vr:systemui	no mock, no __LGS_NP, no hover sheet, no data-lgs-np*, __LGS_VRX absent with the flag off	—
Daemon	/dev/shm/lgs/shell.json	plugin types vrbind, vrbind.strings loaded, failed {}; one vrbind.strings flag-off rejection at install (flag race), covered by the 4 s retry	—

Gamepad traversal on Now Playing is SteamVR's own (React's .gpfocus, Up/Down/A, B = Resume) and could not be driven without a scene app; the theme keeps the DOM order and the single vertical stack, so the order is unchanged.

2026-10-07, session 1: native check (P7 at M3)

• Attempt 1 (11:12): native-session --step "sgcheck --route /library/app/620980 --flags vrBindings,actionsDryRun --json --out …" --step "hv c5b_gp --route /library/app/620980 --flags vrBindings,actionsDryRun --look". Waited 1048 s for native.lock (P7's sg_native.py and another area's session ahead), native on at 11:30 (glassd healthy, lgsNativeOnMain true), sgcheck ran, then my local client hit its own 1500 s timeout during the hv step and its buffered output was lost. The orphaned remote session (holding native.lock, waiting for lab.lock) was interrupted with SIGINT at 11:44 so its finally ran: it exited, released the lock, no hv-*.png left on the Frame, no look folder kept on the PC. Lesson for others: run native-session with python -u and no short local timeout.
• Attempt 2 (11:46, python -u, no local timeout): see below.
• Attempt 2 (11:46-12:01): got native.lock after ~15 min, then lab: lock busy for 240 s at setup (25+ steps queued on lab.lock), so native mode was never turned on ("nothing to undo"), exit 1.
• Result: native check blocked by device contention, not by my area. My area has no native fragments (see above), so there is nothing of mine for sgcheck to judge beyond C5a's cover; V1 should run sgcheck --route /library/app/620980 --flags vrBindings and hv once the queue drains.