vphone
  • English

      • iPhone Duo simulator baseline: alignment table

      Ground truth for vphone-duo, collected 2026-10-29 from Apple's iPhone Duo simulator (iPhone19,4, V68, iOS 27.1 24A94425, Xcode 27.1). The simulator folds through SpringBoard's own display-swap coordinator (README.md, "Posture").

      The vphone-duo column was filled on 2026-10-29 from the VM duobase-01 (bundle 2.14.0-local.21dc0c06, upstream main 82afdb9a, two-display Duo path), iPhone19,4 / iOS 27.1 24A94425, with the simulator probes run in the guest through tools/vm-probe.zsh (VM versions probes/vm-*.lldb; why they are needed is in vm-notes.md). Each cell gives the VM value, then match, differs or n/a. Rows that were marked VM differs before the column was filled say whether the measurement confirmed or corrected that note. Method, step list and caveats: vm-notes.md. VM data: vm-*.json, raw transcripts raw/vm/steps/<step>/.

      On 2026-10-30 the rows the duo-sim-align fix can affect were measured again on duobase-01 with bundle 2.14.0-local.9064efca (branch duo-sim-align: V68 product-id, /product/display0+display1 panel nodes, CoreMotion's device-state relay fed through VPhone Host Controls). The pose now reaches SpringBoard through CoreMotion, not through the test-automation override. Those cells give the new value; where it changed, "(was: …)" holds the first pass. Rows not named in "Changes since the first VM pass" were not re-measured unless their cell says so. Records: the vm-*.json records with "bundle": "2.14.0-local.9064efca" and step fx-<label>; transcripts raw/vm/steps/fx-<label>/; method and the two incidents of this pass in vm-notes.md ("2026-10-30").

      Diff summary

      Remaining differs rows, most consequential first (state after the 2026-10-30 pass). Struck items are aligned now.

      1. Auto-lock / AOD. The VM has no always-on display (DeviceSupportsAlwaysOnDisplay NO, BLSHBacklightHost deviceSupportsAlwaysOn NO). Auto-lock in open turns the inner fully off (backlight 4) instead of AOD (3); folding while auto-locked gives FullyClosed with the cover primary on the lock screen but off (4), inner off. The old note "VM lights the folded inner display" did not reproduce: the inner stayed off and blanked.
      2. Inner display identity (not changed by the fix; another session owns the names). CADisplay VPhone Duo Inner / VPhone Duo Inner / displayId 7 (sim LCD-1 / primary-1 / 3); FBS identity Unknown[7] (sim Unknown.main[3]), not UIKitMainLike; UIScreen._isEmbeddedScreen NO; UIScreen display corner radius 0, sRGB. Re-read on 2026-10-30: unchanged.
      3. Inner display mode. Now 2007×2853 @3 like the sim (was: native 2853×2007 at pointScale 1), but 120 Hz (sim 60 Hz), sRGB (sim P3), and FBS/UIScreen still list a second mode, 2853×2007 (951×669 @3). The cover runs 120 Hz (sim 60 Hz).
      4. MobileGestalt identity. ModelNumber VM0001 (sim A3447), RegulatoryModelNumber nil, HWModelStr V68AP (sim iPhone19,4), AOD keys NO. DisplayExtendedProperties nil, logical display count 1: aligned (item 5 of "Changes").
      5. Display power model (not re-measured beyond the posture rows; stateControl still nil). No CADisplayStateControl on any CADisplay. The cover LCD is the one real panel (the composited output) and is never blanked while the guest is awake, also in book/open; the inner window-server display is enabled when unfolded and blanked when closed. The sim blanks the cover when open.
      6. Digitizer (not re-measured). One Virtual USB Touch Screen (USB, builtIn) as mainDisplayDigitizerService, re-routed to the inner UUID while unfolded; the sim has one CoreDevice touchscreen per panel.
      7. Resolver transition style at a SpringBoard start. 2 .continuousLive right after a SpringBoard start in closed (sim 1 .fade until the first swap). After a swap it now matches: 2 in every posture, 3 with Safari in front (was: 0 at start, then 1, 2 or 3 by path).
      8. Per-panel pose orientation when closed. Device landscapeRight (4) gives cover 4 and landscapeLeft (3) cover 3, now computed by CoreMotion (cmMagicPose cover 4 / 3); the sim gives cover 3 / 4. CoreMotion's own table (identical in the sim) maps a closed hinge to the cover unchanged, so the sim's value comes from its synthetic swap-tool pose. Open matches.
      9. Cover corner radii. extendedProperties and MobileGestalt DisplayExtendedProperties say [5,5,59,59] (Apple's device tree); the sim's device type says [8,8,59,59]. Everything else in both dictionaries matches.
      10. Pixel density. Both panels 460×460 ppi (was: inner 0 / nil); the sim reads 316×669 (cover) and 324×654 (inner). The physical sizes are the sim's (3.0391×4.4217 and 4.3630×6.2022 in) with width and height swapped; the sim's ppi is the native pixel size divided by those swapped axes.
      11. Book angle, residual. 99.444° (position 0.552, closed region 0→0.2747, toClosed 0.5525→0.03) against the sim's synthetic 100° (0.556, 0→0.2761, 0.5556→0.03) (was: 90°, 0.500, 0→0.25). The host controls send 100 and the provider's mechanical angle is 100; CoreMotion's angle is 99.444 (= 100 × 179/180). Not investigated further.
      12. Auto-lock default. The VM runs the real Auto-Lock (60 s; timer warn 40 s, total 60 s); the sim holds the SBDisableAutoDim assertion. Configuration, not behaviour.

      Aligned or reclassified by the 2026-10-30 pass (details in the next section): CoreMotion is absent; inner extendedProperties carry only DeviceSurfaceType; MobileGestalt DisplayExtendedProperties nil / display count 1; hinge HID: one unnamed 0xFF0C/0x21 service; book at 90°; inner portrait-shaped statusBarFrame 0-rect; transition style 1/2/3 by path after a swap; lock-screen clock height: personalization (the VM's lock screen uses another poster, v68.colorB, with a taller adaptive time; see "System chrome").

      Corrected notes: the VM does turn the cover landscape for Safari (678×466, C/C, statusBarFrame 0×0, like the sim); the CoreMotion tabletop and A=7 table rows match (they are CoreMotion's own class methods, not FoldDisplay); the folded inner does not light after auto-lock.

      Changes since the first VM pass

      Bundle 2.14.0-local.21dc0c06 → 2.14.0-local.9064efca, same VM, same iOS.

      #itemfirst passnowsim
      1pose sourcetest-automation override (display.pose)CoreMotion: SBFDevicePoseProvider reads CMAngleManager / CMMagicPoseManager; display.posture replies source: coremotionswap tool writes the provider
      2book angle / position90° / 0.50099.444° / 0.552 (uiHingeStatus 99.444, cmAngle 99.444, mechanical 100)100° / 0.556
      3hinge regions in bookclosed 0→0.25, toClosed 0.50→0.03closed 0→0.2747, toClosed 0.5525→0.030→0.2761, 0.5556→0.03
      4CoreMotion managers+isAvailable NO ×3, sharedManager nil, provider managers nil, _accelerometer BKSAccelerometer+isAvailable YES ×3; sharedManager non-nil and not the provider's (isSharedManager NO); provider _cmMagicPoseManager registered with both handlers, _cmAngleManager active at 0.01 s; _accelerometer nil; currentDeviceState livethe same objects, frozen at boot values; _accelerometer nil
      5MobileGestalt screenDisplayExtendedProperties nil, MGGetLogicalDeviceDisplayCount 12 entries equal to the sim's except the cover corners; count 22 entries; 2
      6extendedProperties{DeviceSurfaceType 1} / {DeviceSurfaceType 2}full dictionaries: DisplayRotation 0 / 270, DeviceSurfaceType and kMGQDeviceSurfaceType 1 / 2, DisplayScale 3, ChromeIdentifier phone15 / phone14, FramebufferIdentifier as the sim, corners [5,5,59,59] / 55×4the same, cover corners [8,8,59,59]
      7inner physical size, ppi0×0, 0 / pixelDensity nil6.2022×4.3630 in, 460×460 (FBS 460×460)4.3630×6.2022, 324×654
      8inner mode2853×2007 pointScale 1 + 951×669 mode, 60 Hz669×951@3x (2007×2853/3) current and preferred, + 951×669@3x; 120 Hz2007×2853 @3, 60 Hz, one mode
      9hinge HID servicesone unnamed Virtual 0xFF0C/0x21CoreMotion_DeviceStateRelay_Crane 0xFF0C/0x21 and _DeviceState 0xFF0C/0x22 (Virtual), plus VPhone Host Controls 0xFF61/0x5B (Virtual)the two relay services
      10inner portrait-shaped status bar (io 1)0-rect, one capture{592,0,77,951}, two captures (25 s after the rotation, and again after a round trip through io 2){592,0,77,951}
      11transition style0 at start; 1 / 2 / 3 by path2 at start; 2 after every swap; 3 with Safari1 at boot; 2; 3 with Safari
      12per-panel pose, closedcover = device orientation (override)the same, now from CoreMotion's table: reclassified as a sim artefactcover = opposite landscape (synthetic)
      13lock-screen clock"differs"personalization: poster v68.colorB vs celosia.timeOfDay

      Row 8's refresh rate comes from upstream's "run the inner display at the panel's 120 Hz" (in this branch), not from the alignment change.

      Data files

      These files were deleted on 2026-10-30, after the values worth keeping were moved into this table, the domain notes and Research/Guest/iphone_duo_simulator_reference.md. Rerunning the probes recreates them (README.md).

      filerecordsscopenotes
      smoke.json60harness smoke run
      display.json + display.md46085backboardd CADisplay / CAWindowServer / BKS, SpringBoard UIScreen / FBS display configurations, digitizersgetters enumerated at runtime; extra field capture
      springboard.json + springboard.md37408content-mode coordinator and resolver, window scenes, traits pipeline, continuous fold, backlight, auto-lock45 steps, all four device orientations
      sensors.json + sensors.md10932CMMagicPose / CMDeviceStateEvent tables, hinge HID, SBFDevicePoseProvider, UIDevice, 527 MobileGestalt keyssim_note marks 2578 sim-only or synthetic values; do not align to them
      appkit.json + appkit.md20829Preferences, MobileSafari, VPAppKitProbe: scenes, geometry, traits, safe area, migration callbacksprobe app source in src/appkit/probe-app
      chrome.json + chrome.md127674status bar, icon grid, dock, quick actions, notification list, clock, wallpapermeta.skipped_getters lists getters that throw
      vm-springboard.json, vm-display.json, vm-sensors.json, vm-chrome.json, vm-appkit.json + vm-notes.mdsee vm-notes.mdthe same probes on duobase-01meta.vm, meta.steps (per-step display.pose, orientation, screenshot size and lit/dark, SpringBoard pid); extra fields probe, step. Second pass (2026-10-30): records with bundle 2.14.0-local.9064efca and step fx-<label>, described in meta.fix

      Every record has api, kind, process, posture, orientation, call, value, error. posture includes the path, e.g. open(after-closed), because many values depend on history. Display UUIDs change on every boot, so join on name or identity.

      # every value of one call, by posture
jq -c '.records[] | select(.call | test("primaryWindowScene")) | {posture, orientation, value}' springboard.json

      Posture and policy

      posturehingepolicyprimary scenecontent mode cover / innerbacklight cover / innervphone-duo
      closed0°, .closedFullyClosedSB-main (cover)2 / 1on / off0°, .closed, FullyClosed, SB-main, 2 / 1, on (1) / off (4); screenshot 1398×2034 lit; the same after a SpringBoard start, after open and after book (2026-10-30, CoreMotion) — match
      book100°, .partiallyOpenFoldingDualcom.apple.springboard.bringup (inner)3 / 2off / on99.444° (0.552), .partiallyOpen, FoldingDual, bringup, 3 / 2, off (4) / on (1); screenshot 2034×1398 lit; after closed and after open — differs (angle residual 0.556°) (was: 90°, 0.500)
      open180°, .openFlatFoldingDualinner3 / 2off / on180°, .openFlat, FoldingDual, bringup, 3 / 2, off / on; screenshot 2034×1398 lit; after book and after closed (2026-10-30) — match

      Content mode: 1 disabled, 2 primary, 3 auxiliary. The CA display state is 1 / 0 when closed and 0 / 1 when open, and both panels keep power state 1. Book and open are identical across the display stack. They differ only in the scene region innerCover (active in book only) and in the hinge state.

      VM: the same history paths (after-open, after-closed, after-book, after a SpringBoard start) give the same rows. There is no CADisplayStateControl on the VM; the window server shows cover LCD enabled and unblanked in every awake posture and the inner enabled when unfolded, blanked when closed (differs, Diff summary 5). Book and open differ only in innerCover and the hinge state on the VM too.

      2026-10-30: one exception seen once, outside the measured steps. After an unplanned SpringBoard restart had dropped the idle-timer assertion, the guest auto-locked in closed; an unfold then woke it to the lock screen and the resolver's prevailing policy read Yoink(inner) instead of FoldingDual in open and in the following book (raw/vm/steps/discarded/fx-*-sb1064-autolock-on). With the screen kept lit the same paths give FoldingDual. The sim has no reference for an unfold from a dark, locked cover.

      Display identity

      itemcoverinnervphone-duo
      CADisplay name / deviceName / displayIdLCD / primary / 1, mainDisplayLCD-1 / primary-1 / 3cover LCD / primary / 1, mainDisplay; inner VPhone Duo Inner / VPhone Duo Inner / 7 — differs (unchanged on 2026-10-30; names owned by another session)
      CADisplay tag0655360 / 65536 — match
      native size, scale, refresh1398×2034 @3, 60 Hz, P32007×2853 @3, 60 Hzcover 1398×2034 @3, 120 Hz, P3; inner FBS current/preferred mode 669×951@3x (2007×2853/3), UIScreen currentMode 2007×2853, scale 3, 120 Hz, sRGB; FBS/UIScreen still add a 951×669@3x (2853×2007) mode; CADisplay currentMode 2853×2007 — differs (refresh, gamut, extra mode) (was: inner 2853×2007, pointScale 1, 60 Hz, extra 951×669 mode at /2)
      CAWindowServerDisplay orientationrot0 / portraitrot90 / landscapeLeft, from bootcover rot0 / portrait; inner rot90 / landscapeLeft from SpringBoard start — match
      extendedPropertiesDisplayRotation 0, DeviceSurfaceType 1, corners 8,8,59,59DisplayRotation 270, DeviceSurfaceType 2, corners 55×4cover DisplayRotation 0, DeviceSurfaceType 1, kMGQDeviceSurfaceType 1, DisplayScale 3, ChromeIdentifier …chrome.phone15, FramebufferIdentifier 1C896A2B-…, corners 5,5,59,59; inner DisplayRotation 270, DeviceSurfaceType 2, kMGQDeviceSurfaceType 2, DisplayScale 3, …chrome.phone14, BF0DC480-…, corners 55×4 — equal to the sim's dictionaries key by key except the cover corners — differs (cover corner radii only: Apple's device tree vs the sim device type) (was: {DeviceSurfaceType 1} / {DeviceSurfaceType 2} only)
      BKSDisplayService traitsForDisplayUUID: surfaceType12inner 2 (also on 2026-10-30); cover not queried (its UIScreen._displayID is nil, so the probe has no cover UUID) — match (inner), cover n/a
      UIScreenindex 0, Main, 466×678, _isMainScreen YESindex 1, Unknown.main[3], _isMainScreen NO, _isEmbeddedScreen YEScover index 0, Main, 466×678, YES; inner index 1, Unknown[7], _isMainScreen NO, _isEmbeddedScreen NO, _isMainLikeScreen NO, corner radius 0, sRGB — differs (unchanged on 2026-10-30)
      FBSDisplayIdentitytype 0, isMain, isMainRoot, isRoottype 7, displayID 3, isRoot, UIKitMainLike, alwaysConnectedcover type 0, isMain, isMainRoot, isRoot; inner type 7, displayID 7, isRoot, not UIKitMainLike, alwaysConnected — differs (unchanged on 2026-10-30)
      FBS config name / boundsMain / 466×678Unknown / 669×951 (never rotates)Main / 466×678; Unknown / 669×951, unchanged through all four inner orientations — match
      _exclusionArea{399,30,37,37}the same {399,30,37,37}, copied from the cover{399,30,37,37} on both — match
      pixelsPerInch316×669324×654cover 460×460; inner CADisplay 460×460, FBS pixelDensity 460×460; physical size cover 4.4217×3.0391 in, inner 6.2022×4.3630 in (the sim's 3.0391×4.4217 and 4.3630×6.2022 with the axes swapped; the sim's ppi divides the native pixels by those swapped axes) — differs (was: inner 0×0, pixelDensity nil, physical size 0)
      sbh_screenType313231 / 32 — match
      digitizerCoreDevice touchscreen(<cover UUID>), main display digitizerCoreDevice touchscreen(<inner UUID>)one Virtual USB Touch Screen (USB, builtIn, authenticated) = mainDisplayDigitizerService, mapped to the inner UUID while unfolded and to the main display when closed — differs

      CAWindowServer lists five displays: LCD, TVOut, LCD-1, Wireless, Resizable. VM: seven, LCD, Wireless, Wireless-1…Wireless-4, VPhone Duo Inner (no TVOut, no Resizable).

      Orientation

      itemsimvphone-duo
      inner interface orientation for device portrait / landscapeLeft / landscapeRight / upside-down4 / 1 / 2 / 34 / 1 / 2 / 3 (display.rotation 90 / 0 / 180 / 270 in open) — match
      inner acceptsall four orientations, home screen (mask 30) and appsall four on the home screen, HomeScreen participant mask 30 — match
      cover with home screen or lock screen in frontstays portrait (participant mask 2), device orientation still changeshome screen stays portrait (mask 2): device 4, systemArbitratedDeviceOrientation 4, interface 1; vphoned reports "The device did not apply the requested rotation" — match
      cover with a landscape-capable app in front (Safari, VPAppKitProbe)goes landscape: 678×466, size classes C/C, statusBarFrame 0×0Safari: 678×466, interface 4 (display.rotation 90) or 3 (270), C/C, statusBarFrame 0×0, safe area 0,0,34,84 / 0,84,34,0, SwitcherLiveOverlay mask 26 — match. The old note "VM refuses cover landscape outright" is corrected: only the home screen refuses, as in the sim
      cover, portrait-only app (Preferences)stays portraitstays portrait 466×678 with UIDevice.orientation 4 — match
      UIDevice.orientationfollows the lit panel; on the cover the name is the opposite of the requested one (landscapeLeft → 4)follows the lit panel (open: 3 for landscapeLeft; closed: equals the cover's orientation) — match; the cover mapping is the next row
      per-panel pose orientationlandscapeLeft: cover 4 / inner 3; landscapeRight: cover 3 / inner 42026-10-30, through CoreMotion: open, display.rotation 0 → device 3, cover 4 / inner 3; 180 → device 4, cover 3 / inner 4 — match. Closed with Safari on the cover: display.rotation 90 → device 4, cover 4, inner 1 (cmMagicPose 4 / 4 / 0); 270 → device 3, cover 3, inner 1 (cmMagicPose 3 / 3 / 0); Safari interface 4 / 3, 678×466 — differs (sim synthetic): CoreMotion's own table (the "propertyA table" row, identical in the sim) maps a closed hinge to the cover unchanged, so the VM value is CoreMotion's and the sim's cover 3 for device 4 comes from its swap tool. Upside-down not measured (the host controls have no upside-down orientation) (was: the same cover values from the override)

      History-dependent values

      itemsimvphone-duo
      inner UIScreen bounds669×951, orientation 0, size classes 0/0 at boot; afterwards it follows the inner's orientation (951×669 for 3/4)669×951 at boot (observed before this run); 951×669 at orientation 4 or 3, 669×951 at 1 or 2, Regular/Regular after an unfold — match (boot size classes not measured)
      inner SBWindowScene.interfaceOrientation1 from boot until the first rotation reaches the inner panel, then 4 in portrait1 from SpringBoard start (two starts: pid 1440, 2384) through every fold and unfold, then 1/2/3/4 with the rotations and 4 in portrait — match
      cover UIScreen after a landscape appkeeps 678×466 while the inner is in usecover rotated with Safari, then unfolded: 678×466, orientation 4, also after the inner returned to portrait — match
      resolver _preferredTransitionStyle.fade until the first swap, then .continuousLive (.continuousSnapshot with Safari)2026-10-30, after a forced respring in closed: 2 before any swap (sim 1), then 2 after every fold and unfold (book, open, closed, both history orders, and after rotations), 3 with Safari in front (closed landscape, book, open) — differs at the start only (was: 0 after a SpringBoard start; 1 after the first fold; 2 after book→open; 1 or 3 with Safari/Preferences by path)
      app client orientation1 if launched on the cover and then unfolded; 4 if launched on the inner, and stays 4 after foldingSafari and Preferences launched on the cover, unfolded: 1; Preferences launched on the inner: 4, still 4 on the cover after folding — match
      hinge regions (hysteresis)closed region 0→0.47 at boot, 0→0.28 after book, 0→0.11 after a fold2026-10-30: 0→0.47 after a SpringBoard start (closed), after open and after every fold; 0→0.2747 after book (toClosed 0.5525→0.03, partiallyOpen 0.2747→0.97), from the 99.444° angle — match within the angle residual (sim 0→0.2761, toClosed 0.5556→0.03) (was: 0→0.25, toClosed 0.50→0.03). The sim's own springboard.json also reads 0→0.47 after folds; 0→0.11 is from sensors.md

      Hinge, pose and identity

      itemsimvphone-duo
      hinge HID serviceCoreMotion_DeviceStateRelay_Crane, 0xFF0C/0x21 (sibling _DeviceState 0xFF0C/0x22), frozen at boot values2026-10-30 (vphoned device.input_devices): CoreMotion_DeviceStateRelay_Crane 0xFF0C/0x21 and CoreMotion_DeviceStateRelay_DeviceState 0xFF0C/0x22, Transport Virtual, published by locationd's relay, plus the feed VPhone Host Controls 0xFF61/0x5B (Virtual, from libvphonehinge in backboardd); values live, not frozen — match (names, usages, transport). systemEventOfType:0x2c not re-read: the backboardd probe that reads it wedged locationd's relay (vm-notes.md) (was: one unnamed Virtual 0xFF0C/0x21, no 0x22 sibling)
      CoreMotion device statefrozen at boot: the sim fold bypasses HID and CoreMotion and writes into SBFDevicePoseProvider2026-10-30: +isAvailable YES for CMDeviceStateManager, CMMagicPoseManager, CMAngleManager (SpringBoard and Safari); the provider's _cmMagicPoseManager (registered, client and internal handlers set) and _cmAngleManager (isAngleActive YES, interval 0.01) are CoreMotion objects as in the sim; _accelerometer nil (sim nil); currentDeviceState live (open: propertyA 1, propertyB 3); cmMagicPose follows the hinge (1 / 2 / 3), the sim's stays at boot values; internal.isOnVM still YES — match (was: +isAvailable NO, managers nil, pose from vphoned's override)
      CMMagicPoseManager sharedManagernon-nil2026-10-30: non-nil, its _manager a CMDeviceStateManager with no handlers, and not the provider's instance (isSharedManager NO, as in the sim) — match (was: nil, +isAvailable NO). +orientationForMagicPose:deviceSurface: is CoreMotion's own class method and runs normally (next row)
      orientationForMagicPose:deviceSurface:equals propertyAForConfigType:surface propertyA:orientation propertyB:hinge propertyC:tabletop for every inputidentical to the sim on all 896 sim inputs; tabletop 0..11 probed — match
      propertyA table, tabletop 0surface 1: hinge 1 → identity, hinge 2/3 → 0,1,2,4,3,6,5,0; surface 2: hinge 1 → 0, hinge 2/3 → 0,1,2,3,4,5,6,0; hinge 0/4/5/6 → 0identical on all 1120 sim inputs — match
      propertyA, tabletop 1..10 with hinge 2/3constants (surface 1: 3,4,1,2,6,3,4,5,6,1; surface 2: 0,0, then 1,2,3,4,6,6,5,1)surface 1: 3,4,1,2,6,3,4,5,6,1; surface 2: 0,0,1,2,3,4,6,6,5,1 — match (corrects "FoldDisplay does not model tabletop": the table is CoreMotion's)
      propertyA, A=7 with hinge 2/300 — match (corrects "not covered")
      hinge enumsSpringBoard uiHingeStatus 1/2/3; app _UIHingeState 1/3/4; public UIHinge status 1/2/3, angle in radians2026-10-30: uiHingeStatus 1/2/3 (angle 0 / 99.444 / 180); Safari _UIHingeState closed / partiallyOpen (99.4) / openFlat (180), scene _UISceneHingeState closed / partiallyOpen / fullyOpen — match (was: angle 90 in book); public UIHingeInteraction not measured (VPAppKitProbe not installed) — n/a
      hinge state deliveryonly to scenes whose client sets wantsSceneHingeState; FBSSceneSettings.hingeAngle/hingeStatus/magicPose stay 02026-10-30: Safari (wantsSceneHingeState YES) gets it closed / book / open; FBSSceneSettings.hingeAngle and hingeStatus 0 — match; Preferences (nil in the first pass) not re-measured
      MobileGestalt identityiPhone19,4, DeviceClass iPhone, A3447, "iPhone Duo", HomeButtonType 2, DeviceCornerRadius 59, DynamicIsland, AlwaysOnDisplay, EnhancedMultitasking falseiPhone19,4, iPhone, VM0001 (RegulatoryModelNumber nil), "iPhone Duo", 2, 59, DynamicIsland YES, AlwaysOnDisplay NO, EnhancedMultitasking NO; HWModelStr V68AP (sim iPhone19,4) — differs (re-read on 2026-10-30: unchanged)
      MobileGestalt screenmain-screen 1398×2034 @3, pitch 460 (the cover); DisplayExtendedProperties has 2 entries; logical display count 22026-10-30: 1398×2034 @3, pitch 460; DisplayExtendedProperties 2 entries, equal to the sim's except the cover's DeviceCornerRadius [5,5,59,59] (sim [8,8,59,59]); MGGetLogicalDeviceDisplayCount 2 — differs (cover corner radii only) (was: DisplayExtendedProperties nil, count 1)
      UIDeviceno hinge, tabletop, fold or posture getter existsnone exists — match

      What an app sees

      itemclosed (cover)book / open (inner)vphone-duo
      [UIScreen mainScreen]covercovercover (screens[0] LCD Main) in every posture; 2026-10-30, Safari closed / book / open: the same — match
      [UIScreen screens] count12 (the inner connects and disconnects with the fold)1 / 2; 2026-10-30 Safari: 1 / 2 / 2 — match
      scene bounds466×678951×669466×678 / 951×669 (Safari and Preferences); 2026-10-30 Safari: 466×678 / 951×669 / 951×669 — match
      size classesC/RR/RC/R / R/R; 2026-10-30 Safari: C/R / R/R / R/R — match
      interface orientation (device portrait)portraitlandscapeRight (4)1 / 4; 2026-10-30 Safari: 1 / 4 / 4 — match
      safe area t,l,b,r0,0,34,84 (right status column)0,0,34,84; Safari 82,0,34,00,0,34,84 / 0,0,34,84; Safari 82,0,34,0; 2026-10-30 Safari: the same — match
      idiomPhonePhonePhone (0) — match
      scene cornerRadiusConfiguration8,8,59,5955×48,8,59,59 / 55×4; 2026-10-30 Safari: the same (the scene keeps 8,8,59,59 although extendedProperties now say 5,5,59,59) — match
      ssk_surfaceType12 (4 during launch)1 / 2 (no launch-time capture); 2026-10-30 Safari: 1 / 2 / 2 — match
      scene regionsfold {0,455.5,669,40}, innerCover active in book only, statusBarOcclusion {0,0,124,84}cover: cover, coverCamera, statusBarOcclusion; inner: innerCover, innerRear, inner, fold {0,455.5,669,40} margins 20, innerCamera, statusBarOcclusion; innerCover active YES in book, NO in open; 2026-10-30 Safari: closed cover, coverCamera, statusBarOcclusion active; book innerCover, innerRear, inner, fold active; open innerCover, innerRear, fold inactive — match
      Preferences split columncompact475.67 in book (sidebar ends at the fold), 320 in opencollapsed (466) / 475.67 in book, 320 in open — match

      Migration moves the existing scene: same UIWindowScene, same persistent identifier, foregroundActive throughout, with no lifecycle callbacks.

      • Unfold: hinge 2, then about 110 ms later UIScreenDidConnect → one didUpdateEffectiveGeometry: → trait C→R.
      • Fold: hinge 1, then about 70 ms later trait R→C → geometry → UIScreenDidDisconnect.

      VM: the callback sequence was not measured (it needs VPAppKitProbe). In the app the inner screen's identity is Unknown.main[7] and its _name is VPhone Duo Inner.

      System chrome (points)

      itemcoverinner, landscape-shaped (orientation 4/3)inner, portrait-shaped (1/2)vphone-duo
      status bar class…AxisAlignableSplitV68Cover…AxisAlignableSplit2007samethe same classes — match
      status bar frameright column {389,0,77,678}statusBarFrame 0-rect; foreground 88×123.67 top right{592,0,77,951}cover {389,0,77,678}; inner landscape 0-rect — match; 2026-10-30 inner portrait-shaped (interface orientation 1, display.rotation 0): {592,0,77,951}, wrapper height 77, in both captures (25 s after the rotation, and after a round trip through orientation 2) — match (was: 0-rect, one capture). At orientation 1 straight after a SpringBoard start, before any rotation, it is 0-rect, as in the sim's open-after-closed and first open-landscapeLeft runs
      icon grid4×6, 60×60 icons, label 12 pt, spacing 28×38.84×6, list 324 wide at x=470 (open) / 512.67 (book)4×6 over 669×951cover spacing 28×38.8, list 466×678; inner list 324×669 at x=470 (open) / 512.67 (book), spacing 28×38; portrait-shaped list 669×951 — match
      dockright side column {376.17,45,84,678}, capacity 1×4right side column {861.17,45,84,669}bottom dock {0,851,669,100}{376.17,45,84,678}; {861.17,45,84,669}; {0,851,669,100} — match (the VM dock holds 4 icons, the sim 1)
      quick action button84×84 at {376,573}84×84 at {861,564}84×84 at {579,846}84×84 at {376,573}, {861,564}, {579,846} — match; the VM shows them (hidden NO) with the real torch and camera glyphs (28.33×22 and 17.33×30.67 in the 48-point icon area), the sim hides them with a placeholder
      notification platter width350 (list 370 at boot, 378 after open)open 450 (list 478, centred); book 350 (list 378, right half)450no notifications on the VM (no simctl push) — n/a
      clock frame372×132.26 at {11,47}open 400×132.26 at {275.5,51.67}; book x=487.25cover 372×240.28 at {11,47}; inner open 400×232.16 at {275.5,51.67} (2026-10-30; first pass 235.61) — personalization (poster: VM com.apple.MercuryPoster v68.colorB, sim celosia.timeOfDay). The height follows the poster's adaptive time: adaptiveTimeTextHeight 213.01 (sim 105), time font .SFAdaptiveSoftNumeric weight wght3200000 at 128.77 pt (sim wght2800000 at 110.03 pt), intrinsicContentSize 272×244.67 (sim 196.67×131.67); the limits are the same (minimum/maximumAdaptiveTimeTextHeight 104.74 / 303.68), as are titleContentStyle (vibrant material) and the glass renderer. Screenshots: raw/vm/steps/fx-closed-lockscreen/shot.jpg (photo poster, tall time) against raw/chrome/shots/closed-after-open-lock-d1.png (abstract time-of-day poster)
      wallpaper windowportrait-anchored 669×951 while the other windows are 951×669cover 466×678; inner 669×951 while the other inner windows are 951×669 — match

      Quick action glyph: button 84, icon area 48 (insets 18), glyph view 27×25, image 19.33×18. Glyph/button is 0.321 × 0.298. The sim hides the quick action buttons and shows the placeholder symbol questionmark.app, so this does not measure the flashlight or camera glyph. The VM's 0.68 glyph scale has no direct sim reference.

      Power and auto-lock

      itemsimvphone-duo
      auto-lock defaultoff (SBDisableAutoDim assertion); real timer dim 40 s, lock 60 son: Auto-Lock 60 s, SBBacklightIdleTimer warn 40 s, total 60 s, no disable assertion — differs (configuration)
      auto-lock in book/openinner goes always-on dimmed and stays primarymeasured in open: inner stays primary but goes off (backlight 4, not 3), both panels off, screenshot black, window server blanks both — differs (no AOD)
      fold after auto-lockFullyClosed; the cover becomes primary, always-on dimmed, lock screen; inner offFullyClosed, cover primary on the lock screen but off (4), inner off (4) and blanked; still so 20 s later; screenshot black — differs (no AOD). Corrected: the folded inner does not light
      fold right after a lock-button pressthe cover turns fully oncover on (1), lock screen, inner off; screenshot lit — match

      Limits

      • The sim fold does not go through HID or CoreMotion, so the VM's HID → CoreMotion → pose provider path has no reference. Only values from SBFDevicePoseProvider onward compare.
      • Values with sim_note (host Mac model, memory, CPU, IsSimulator, the SIMULATOR_* environment) do not exist on the VM.
      • posture.zsh's 3 s readback during a 180° inner rotation once crashed SpringBoard mid-animation (_SBSystemApertureContainerViewRotatingContentView setTransform:). Use a longer settle for rotations.
      • Known harness defects, worked around but not fixed in tools/:
        • POSTURE_OUT defaults to the shared raw/posture.
        • smoke.zsh labels interface orientations 3 and 4 with device-orientation names.
        • posture.zsh unlock releases the testing lock, not the device lock screen.
        • sim.zsh pid fails while lldb is attached.
      • VM side: see vm-notes.md for the lldb traps (EXC_GUARD from more than one cast objc_msgSend per expression, garbage strcmp/strncmp, unlowered PAuth_LR code) and the getters skipped on the VM.
      • 2026-10-30, bundle 2.14.0-local.9064efca: do not run probes/vm-sensors-backboardd.lldb (an IOHIDServiceCopyProperty walk from backboardd's main thread). With the CoreMotion relay present it was interrupted and left the relay wedged until locationd was restarted, which restarted backboardd and SpringBoard too. Read the service list with vphoned's device.input_devices instead (vm-notes.md, "Incident 1").