forked from mediaexplorer74/WPR
-
Notifications
You must be signed in to change notification settings - Fork 1
Update History
Bubbleshum edited this page May 24, 2026
·
17 revisions
Running changelog of WPR fixes / features / compatibility changes. Most recent entries are at the top.
- Compat: Mirror's Edge promoted from Broken → Playable after the
GamerServicesDispatcher.WindowHandlefix below - Fix: Mirror's Edge booted to a flat grey screen (RGB 84,84,84) and never
drew anything else —
LoadContentwas never invoked, so the game cleared to its background colour every frame and sat there. Root cause: Mirror's Edge'sGame.Initialize()override setsGamerServicesDispatcher.WindowHandle, wrapped in a try/catch that only swallowsGamerServicesNotAvailableException,NotSupportedException, andInvalidOperationException. Our shim'sWindowHandleproperty threwNotImplementedExceptionon both get and set, which escaped the game's narrow catch, propagated up into FNA'sGame.DoInitialize, and was swallowed there — leaving the game in a half-initialised state whereLoadContentand the rest of the boot sequence never ran. Fix: inSrc/Core/Microsoft.Xna.Framework.GamerServices/GamerServicesDispatcher.cs, changedWindowHandleto a no-op auto-property (public static IntPtr WindowHandle { get; set; }). We have no Live UI tied to a host window handle anyway, so the stored value is never read back by our stack. Shim-only — no reinstall - Compat: Need For Speed Undercover promoted from Legacy → Partial
(first-hand verification on this fork). Java-midp port wrapped in XNA
(entry point
midp.JavaLibGameinnf9_xna.dll). Engine hardcodes a landscape-right rotation for a phone held physically tilted:midp.Display.ctorunconditionally setsorientation = LANDSCAPE_RIGHTandbackBuffer.isLandscape = true;midp.Graphics.lanscapeTransformisRotZ(π/2) × Translate(width, 0)keyed offPresentationParameters.BackBufferWidth;midp.Graphics.clipRecthardcodes scissor clamps at 480 × 800 portrait logical bounds. Symptom: on a desktop landscape window (no physical tilt) the engine renders portrait UI content rotated 90° CW, so text reads top-to-bottom and the scene appears rotated 90° from what a desktop user expects. Currently launches in a portrait 480 × 800 window with content rotated 90° (sideways-readable); playable in that state if the user tilts their head or rotates their monitor. Investigated four host-side workarounds (per-game rotation setting + RT trampoline;PreparingDeviceSettingsswap to landscape PP dims; PP swap + scissor rotation transform; PP swap + scissor rotation +SpriteBatchtransform override with letterbox fit) — each either failed to progress past the EA splash, hit the engine's hardcoded portrait scissor clamps, left engine-rotated text sideways, or misaligned non-UI draws (cars, background, EA splash) with the letterbox region. All reverted. Net: no clean desktop-landscape interpretation without per-game asset/engine patches; likely affects other Java-midp ports too. No net code changes landed
- Compat: Fable Coin Golf promoted to Playable — plays end-to-end and closes cleanly after the two fixes below
- Fix: Fable Coin Golf aborted during initial content load with a
ContentLoadExceptionwrappingFileNotFoundExceptionforContent/data/Brightwall_Wall_Rock_Snow__0.xnb. Root cause: the XAP ships without that chapter-2 texture, but the obstacle list still references it, and the reference is embedded as an external reference inside the BasicEffect XNB rather than being read via a directOpenStreamcall. FNA'sTitleContainer.OpenStreamcorrectly rethrowsFileNotFoundException(as required by Castlevania Puzzle's localized-asset try/catch fallback — see the 22/05 entry), so the missing asset surfaced as a fatal load failure rather than degrading. Fix: inSrc/ThirdParty/fna/src/Content/ContentReader.cs,ReadExternalReference<T>now catchesContentLoadExceptionwhose inner cause isFileNotFoundExceptionand returnsdefault(T). Embedded external references inside other XNBs degrade gracefully; directOpenStreamcallers (Castlevania Puzzle's localization fallback) still see the exception and can run their own catch. Shim-only — no reinstall - Fix: Fable Coin Golf crashed on window close with a fatal
AccessViolationExceptionraised from native FNA3D — uncatchable, took the host process down. Root cause:ResetWprSingletonsinSrc/Core/WPR/ApplicationLaunch.cshad a broad sweep that walked every type in every Default-ALC sibling DLL and calledf.GetValue(null)on its static fields, which implicitly triggered the static constructor of any untouched type. Fable'sCIwAttractor.cctorbuilds aVertexBuffervia the native FNA3D path; at shutdown theGraphicsDevicewas already torn down, so the native call dereferenced freed device state and raised the AV. Fix: removed the broad sweep entirely. Sibling-DLL touches are now restricted to the hand-curated_PerLaunchFlagsToResetallow-list (currently justPressPlay.FFWD.Application.quitNextUpdatefor Tentacles, which Fable doesn't use), so untouched types never get cctor-triggered at shutdown. Shim-only — no reinstall - Compat: Asteroids Deluxe promoted from Broken → Playable after the
Type2.GetTypefix below — clears theKrome.GameRoom.UI.Screens.InGameScreenData.set_Machinecrash recorded in the 22/05 entry - Fix: Asteroids Deluxe (and any Krome-based title where a sibling library
DLL calls
Type.GetType("X, MainUserAssembly")) crashed during the initial screen-state machine transition with aFileLoadExceptionwrappingNotSupportedException: A non-collectible assembly may not reference a collectible assembly. Root cause:Krome.dll(sibling library, loaded into the DefaultAssemblyLoadContextby design to satisfy the non-collectible-cannot-reference-collectible rule) calledType.GetType("AsteroidsDeluxe, ..."), which our patcher routes throughWPR.WindowsCompability.Type2.GetTypeinSrc/Core/WPR.WindowsCompability/Type2.cs. That shim previously delegated straight to the frameworkType.GetType(string), whose binder resolves assemblies through the Default ALC — butAsteroidsDeluxe.dlllives in the per-launch collectible user ALC, so the CLR's cross-ALC collectibility check tripped and the bind failed. Fix:Type2.GetTypenow searchesAssemblyLoadContext.Allfor an assembly whose simple name matches the type's assembly qualifier and resolves the type viaAssembly.GetTypedirectly, bypassing the framework binder entirely. The new path is a pure managed lookup with no cross-ALC collectibility validation, so the binding succeeds. Shim-only — no reinstall - Compat: Tentacles confirmed Playable end-to-end including clean relaunch after the fix below (first launch was already fine; second launch was silently quitting at frame 0)
- Fix: Tentacles silently closed its XNA window immediately on the second
launch in a WPR session — no error, no log line, just an instant quit
after the first frame. First launch was fine, and the symptom only
appeared if the user backed out via the main-menu Back button (which
routes
MainMenu.OnDoQuit → PressPlay.FFWD.Application.Quit) on launch one. Root cause: Tentacles is built on the PressPlay.FFWD mini-engine (PressPlay.FFWD.dll), which ships as a sibling DLL of the mainTentacles.dll. WPR's assembly resolver routes sibling DLLs into the non-collectible DefaultAssemblyLoadContext(to satisfy the "non-collectible assemblies may not reference collectible assemblies" rule), so FFWD's managed statics survive across launches.PressPlay.FFWD.Application.quitNextUpdateis aprivate static boolthatApplication.Quit()flips totrue; the next frameApplication.Updatereads it and callsGame.Exit(). On launch one this works correctly. On launch two, FFWD was already loaded in Default ALC, the newApplicationinstance read the still-trueflag at frame 0 and immediately calledGame.Exit(). Fix: inSrc/Core/WPR/ApplicationLaunch.cs, track sibling DLLs that the Default-ALC resolver pulls from the user install folder (new_DefaultAlcUserSiblingslist) and extendResetWprSingletonsto reset per-launch state on those siblings between launches with a deliberately narrow policy: clear allIList/IDictionarystatic fields in place (mutation, notSetValue, so it works throughstatic readonly), and reset specific known per-launch flag fields by name via an allow-list (_PerLaunchFlagsToReset), currentlyPressPlay.FFWD.Application.quitNextUpdate. Adding fields to the allow-list requires manual confirmation that the field is a plain managed bool/int, not[ThreadStatic]or JIT-inlined — an earlier blanketSetValueon every writable value-type static crashed with a fatal0x80131506 COR_E_EXECUTIONENGINE. Shim-only — no reinstall - Compat: Tiki Towers promoted from Legacy → Playable (first-hand verification on this fork)
- Compat: Storm in a Teacup tested on this fork — marked Partial. Boots and is playable end-to-end but exhibits major graphics issues during gameplay
- Compat: Asteroids Deluxe marked Broken on this fork — crashes during the
initial screen-state machine transition
(
Krome.GameRoom.UI.Screens.InGameScreenData.set_Machine) before gameplay starts, surfacing asFileLoadExceptionwrappingNotSupportedException: A non-collectible assembly may not reference a collectible assembly. Root cause:InGameScreenDatacallsType.GetType("AsteroidsDeluxe, ...")to reflect into its own user assembly, which is redirected throughWPR.WindowsCompability.Type2.GetTypeatSrc/Core/WPR.WindowsCompability/Type2.cs.Type2.GetTypedelegates straight to the frameworkType.GetType(string), which resolves assemblies via the defaultAssemblyLoadContext— butApplicationLaunchloads every user/game assembly into a per-launch collectible ALC (_CurrentUserAlcinSrc/Core/WPR/ApplicationLaunch.cs) so launches can unload cleanly. The CLR refuses to bind a non-collectible static reference (WPR.WindowsCompability) to a collectible assembly (AsteroidsDeluxe.dll) and throws. Likely fix is shim-only: routeType2.GetTypethrough the active user ALC — either via theType.GetType(string, Func<AssemblyName, Assembly>, …)overload with a resolver that delegates to_CurrentUserAlc.LoadFromAssemblyName, or by splitting the type name and resolving the assembly explicitly. Not yet landed - Compat: Guitar Hero 5 Mobile promoted from Legacy → Playable (first-hand verification on this fork)
- Compat: Final Fantasy 3 demoted from Legacy → Broken — first-hand
verification on this fork. Square Enix's NDS → WP7 port ships its own
in-process Nitro/G3X-to-XNA renderer inside
syrcusW.dll, and the NDS background tile loaders (GXS_LoadBG0Char,GXS_LoadBG0Scr,GXS_LoadBGPltt,GXS_LoadOAM,GX_LoadTexPltt,GX_SetCapture, and ~25 siblings) ship as empty-bodied stubs.Game.Tickticks normally, presents happen at 800x480 and no exceptions are thrown, but with the BG palette/char/screen upload path no-op'd the DS background tiles never become XNA textures — maps and menus render as a black backdrop with only the foreground sprite layer and font glyphs visible (~3-4 draw calls/frame). Not fixable shim-side: WPR's existing patcher entries and FNA shims are sufficient, the bug is inside the game's own port code. Would require binary-patchingsyrcusW.dllto translate NDS BG palette/char/screen memory layout into XNA texture data — multi-day RE job, not a shim addition - Compat: The Sims Medieval promoted from Legacy → Playable (first-hand verification on this fork)
- Compat: Angry Birds promoted from Legacy Broken → Playable (first-hand verification on this fork)
- Compat: Max and the Magic Marker promoted from Legacy → Playable (end-to-end after the three fixes below)
- Compat: Pac-Man promoted from Legacy → Playable (first-hand verification on this fork)
- Fix: Max and the Magic Marker crashed on window close with an
AccessViolationExceptionoriginating inFNA3D_AddDisposeTexturefrom~Texture(). Root cause: theTexturefinalizer ran afterGraphicsDevice.Disposehad already destroyed the FNA3D device, so the native dispose call dereferenced freed device state. Fix: guard the native dispose call inSrc/ThirdParty/fna/src/Graphics/Texture.cson!GraphicsDevice.IsDisposed. Shim-only — no reinstall - Fix: Real touchscreen input never registered in Max and the Magic Marker
(and any other XNA game polling SDL touch). Root cause: SDL2's touch
device IDs are 64-bit (
SDL_TouchID), butTouchPanel.LastActiveTouchIdwas declaredint, so the value was truncated as it round-tripped throughTouchPaneland the polling path calledSDL_GetNumTouchFingers(bogusId)— which always returned 0. Widened the field tolonginSrc/ThirdParty/fna/src/Input/Touch/TouchPanel.csand dropped the(int)casts inSrc/ThirdParty/fna/src/FNAPlatform/SDL2_FNAPlatform.cs. Shim-only — no reinstall - Fix: Single-finger drawing in Max and the Magic Marker refused to start a
stroke — one real finger was being reported as two simultaneous touches.
Root cause: SDL synthesises mouse events from touch events by default
(
SDL_HINT_TOUCH_MOUSE_EVENTS = 1); we already filtered the synthetic events in our event pump, but the mouse-as-touch poll still read the pollutedSDL_GetMouseState, writing a phantom finger into slot 7 alongside the real one. Max's drawing logic treats a 2-touch frame as a multi-finger gesture and refuses to start a stroke. Fix: setSDL_HINT_TOUCH_MOUSE_EVENTS = "0"in the Windows init block ofSrc/ThirdParty/fna/src/FNAPlatform/SDL2_FNAPlatform.csso SDL stops synthesising mouse events from touch entirely. Shim-only — no reinstall - Compat: Z0MB1ES (on teh ph0ne) confirmed Playable
- Compat: I Dig It promoted from Broken → Playable (end-to-end after the two fixes below)
- Fix: I Dig It crashed at frame 1 of
IDigItApp.UpdatewithFileLoadException(0x80131515)wrappingNotSupportedException: A non-collectible assembly may not reference a collectible assemblyon its first reference to its bundledChipmunk.dllphysics wrapper. Root cause: theAssemblyLoadContext.Default.Resolvinghandler inApplicationLaunchloaded every sibling DLL it found in the install dir into the collectible user ALC and handed the resulting assembly back to its Default-ALC requestor. The CLR then refused to bind a non-collectible static reference to that collectible assembly, so the JIT failed at the IDigItApp.Update call site. Fix: split the handler's two cases — the main game DLL still routes into the user ALC (FNA'sContentTypeReaderManager.Type.GetTypelegitimately needs the collectible type back, and loading it into Default would re-trigger the launch-1-statics contamination bug guarded by the existing comment), but sibling DLLs now load into the Default ALC so non-collectible static refs bind cleanly. Trade-off: those siblings persist across launches; for self-contained native-wrapper libs like Chipmunk that's harmless. Added a_CurrentMainAssemblyNamefield to drive the discrimination. Shim-only — no reinstall - Fix: After clearing the Chipmunk crash, I Dig It got past the splash but
still wedged just before the menu rendered, this time with
MissingMethodException: LeaderboardReader.get_TotalLeaderboardSize(). Our shim hadTotalLeaderboardSizedeclared as a method; XNA exposes it as anintproperty. Adding only that surfaced the next missing member (get_PageStart), so the full property surface (CanPageDown,CanPageUp,IsDisposed,IsSynchronizedWithLiveServer,Leaderboard,PageSize,PageStart,TotalLeaderboardSize) plusEndPageDown/EndPageUp/EndRead-
PageDown/PageUp+IDisposable.Disposewas filled in in one pass. All stubs report "no leaderboard available" (false / 0 / empty / fresh reader), so the game treats the leaderboard screen as empty rather than crashing — appropriate given there's no live Xbox LIVE backend. Shim-only — no reinstall
-
- Compat: Castlevania Puzzle promoted from Partial → Playable (story / arcade mode now start; previously tapping a save slot silently no-op'd)
- Fix: Castlevania Puzzle's save-slot screen rendered correctly but tapping a
slot to start the game did nothing. The slot tap fires
DashResourceProvider.getLocalizedBinary, which uses the standard WP7 localization-fallback idiomtry { OpenStream("dat\\data\\items.en-GB") } catch { OpenStream("dat\\data\\items") }. Our FNATitleContainer.OpenStreamwas catchingFileNotFoundException(and every other exception), logging it to[wpr-ex], and returningnullinstead of rethrowing. The game'scatchtherefore never fired, the fallback to the unsuffixed file never ran, and the subsequentnew BinaryReader(null)threwArgumentNullExceptioninsideGame.Update— whichGame.Tickswallows. The save-slot UI stayed visible, so the user perceived the "Start Game" button as missing. Fix: rethrow at the end of both catch blocks inTitleContainer.OpenStreamafter theContent/Scenes/fallback has been given a chance to run.[wpr-ex]diagnostics still log on the way out. Likely had silent secondary effects in any other game using the same localization-fallback pattern. Shim-only — no reinstall - Compat: Assassin's Creed: Altaïr's Chronicles promoted from Partial →
Playable (end-to-end after
SignedInGamer.SignedIndeferred-invoke fix below) - Fix: Assassin's Creed crashed in
XNAGame..ctorwith aNullReferenceExceptioninside its ownSignedInhandler. Root cause:SignedInGamer.SignedIn'saddaccessor took two paths — a deferred 2sTask.Delayfor the very first subscriber in the session, and a synchronous invoke for any later subscriber (gated by a staticFirstSignInSessionDoneflag). Two compounding problems: (1)SignedInGamer.Reset()was called afterActivator.CreateInstance(mainType!)inApplicationLaunch.Start, so the flag set by a previous game launch in the same WPR session carried into the new game's ctor; (2) the synchronous path then fired the just-attached handler during+=, while the game's ctor was still mid-construction, and the immediate-fire branch had no try/catch wrapper (unlike the deferred branch). Assassin'sXNAGame.aaccessed athis-field its ctor hadn't initialised yet, NRE'd, and the exception escaped up throughActivator.CreateInstance. Fix: collapse the two paths into one — every subscription dispatches onTask.Run(delay 0 for late subscribers, 2s for the first), goes through the same_SignInGatesemaphore + try/catch, and never runs synchronously from+=. Also movedSignedInGamer.Reset()to run beforeActivator.CreateInstance. Shim-only — no reinstall - Compat: Plants vs Zombies promoted from Broken → Playable (end-to-end after achievement-seed + SpriteBatch tolerance fixes below)
- Fix: PvZ launched to a black screen because
Lawn.AchievementsWidget.DrawNRE'd insideAchievements.GetAchievementItem— the game's staticgAchievementListwas empty, so the lookup returned null and the NRE escaped mid-SpriteBatch.Begin, wedging every subsequent frame on"Begin has been called before calling End". Root cause: PvZ stores its 18 achievement keys in an inlineAchievements.ACHIEVEMENT_KEYS[]static array (built in the cctor), so none ofXnaAchievementCodeExtractor's three sources —Content/xml/socialnetworks.xml.xnb, ILldstrnearAwardAchievement*callsites, orContent/Achievements/*.xnbfilenames — recovered any keys at install time. The seeder wrote 0 rows;BeginGetAchievementsreturned 0 rows; the game'sGetAchievementsCallbackadded 0 items togAchievementList. Added a Source D (KnownProductCatalogues) keyed by ProductId that returns hardcoded keys recovered via decompilation, threadedproductIdthroughXnaAchievementCodeExtractor.ExtractRichandXnaAchievementSeeder.SeedAsync. Install-time change — affected games need reinstalling. Now seeds 18 rows andBeginGetAchievements: 18 rowsconfirms it - Fix: After the achievement seed worked the game flickered on main menu and
in level —
Lawn.GameSelector.DrawOverlaystill NRE'd every frame on some other null reference, but the deeper problem was that PvZ'sSexy.Graphicslayer tracks its ownspritebatchBeganflag separately from FNA'sbeginCalled, and the two desync once any Draw exception leaves the game's flag stale. The result was a within-frame double-Begin viaSetupDrawMode → EndFrame → EndDrawImageTransformed → BeginFrame, throwingInvalidOperationExceptionand dropping every other frame. We can't reach the game's private flag from the shim. Made FNA'sSpriteBatch.Begin/Endtolerant of out-of-order calls: a secondBeginwithout an interveningEndsoft-resets and starts a fresh batch (drops the queued sprites of the discarded batch); a strayEndwithout a matchingBeginno-ops. First 5 of each are logged so we can spot if it's firing in production. Shim-only — no reinstall - Feat: Diagnostic
[wpr-trace] BeginGetAchievements: N rows for <ProductId>log inSignedInGamer.BeginGetAchievementsconfirms whether the install- time seed populated for a given game on first launch - Feat: Keyboard accelerometer simulator. New Controls page in the desktop
sidebar binds four keys (defaults WASD) to tilt directions; readings flow
through the existing
Microsoft.Devices.Sensors.Accelerometerso games see them without any per-game shim work. Sensitivity slider, master toggle, in-game tilt-overlay (Avalonia for Silverlight host, FNADrawableGameComponentfor XNA), and a live-preview dial on the Controls page. Orientation-aware: the screen-relative key intent gets rotated into the device-portrait frame the WP7 sensor contract expects, so a landscape game (W = "tilt up the screen") produces the correct device-X tilt the game interprets as steer-left. Desktop orientation is inferred from the back-buffer aspect because FNA'sWindow.CurrentOrientationonly updates from SDL display-rotation events that never fire on the desktop.Accelerometer.CurrentValue/IsDataValid/TimeBetweenUpdateswere also added so games that poll instead of subscribing toReadingChangedsee live readings too - Compat: Hydro Thunder GO promoted from Partial → Playable (steerable via the keyboard accelerometer)
- Compat: Uno confirmed Playable
- Fix: Sonic 4 Episode I self-paused on every Update tick. The game's
AppMain.isForegroundflag is flipped true only by itsGame.Activatedhandler, andActivatednever fired on WPR becauseINTERNAL_isActivewas initialisedtrue(suppressing the BeforeLoop setter's no-op transition to avoid Asphalt 5's pre-Initialize KeyNotFoundException). Net effect:isForegroundstayed false, the game's pause condition (!isForegroundORed into the trigger) re-armed every frame, and pause reasserted itself immediately after dismissal. Fix: defer a one-shotOnActivatedto the end of the firstGame.Tick, where every game's per-Update state (including Asphalt 5's) is already populated - Fix: Window background/foreground transitions now fire
Deactivated/Activatedcorrectly on desktop and Android. Restored theSDL_WINDOWEVENT_FOCUS_LOST → IsActive=falsebranch inSDL2_FNAPlatform.PollEventsthat had been commented out under a//RnDmarker. SymmetricFOCUS_GAINEDwas already wired - Feat: Multi-touch input is now additive with the mouse-as-touch shim.
UpdateTouchPanelStatepreviously branched either-or — ifTouchPanel.MouseAsTouchwas on (the default for XNA games on WPR) the real-finger poll loop was skipped entirely, capping all input at a single touch even on multi-touch hardware. New shape: real fingers fill slots0..MAX_TOUCHES-2fromSDL_GetTouchFinger; whenMouseAsTouchis on the mouse takes the last slot with synthetic finger IDint.MaxValueso it can never collide with a real finger's ID. Sonic 4's "hold D-pad + tap jump" gameplay now works - Fix: Asphalt 5 splash → menu tap-to-continue now registers. The
per-
Game.TickFrameworkDispatcher.Update()call added on 21/05 was redundant — stock FNA'sGame.Updatealready pumps the dispatcher at its end. Pumping twice per tick madeTouchPanel.Updaterun twice in close succession; the second run promotedtouches[0]fromPressedtoMovedbeforeCGame1.g()could read it, soh2.b(press handler) never recorded the finger andh2.d'sif (i > 0)release-guard always failed. Splash-statelt.b()waiting onbe.ey.fm != 0was the surfaced case. Removed the redundant pump in FNAGame.Tick - Compat: Asphalt 5 promoted from Partial → Playable
- Compat: Sonic 4 Episode I confirmed Playable (pause loop fixed, multi-touch routed)
- Compat: Tentacles promoted from Broken → Playable (end-to-end after the fixes below)
- Fix: FNA's
ContentTypeReaderManagercouldn't resolveReflectiveReader<PressPlay.FFWD.Scene>becauseType.GetType()doesn't see types in the user game's collectible ALC, and the existingstring.Split(',')fallback broke on the generic argument[[PressPlay.FFWD.Scene, PressPlay.FFWD, …]](the comma between the inner type and its assembly hint was indistinguishable from the outer delimiter). Replaced withType.GetType's resolver-callback overload that walksAppDomain.CurrentDomain.GetAssemblies()— which returns every loaded assembly across every ALC — so generic readers parameterised by user-ALC types resolve correctly. Tentacles' Preloader scene XNB now deserialises - Fix: FFWD-based games call
Application.LoadLevel("X")with bare scene names even though every level XNB ships underContent/Scenes/. ContentManager constructedContent/X.xnb, the file wasn't there,AssetHelper.Load<T>silently swallowed the failure and returneddefault(T), and the loading screen hung forever waiting onloadingProgress == 1.0f.TitleContainer.OpenStreamnow retriesContent/<name>.xnbasContent/Scenes/<name>.xnbwhen the original is missing and has no subdirectory hint. Safe for non-FFWD games (fallback file simply won't exist either) - Fix:
Microsoft.Phone.Shell.StandardTileData.Countwasintin our shim but the WP7 SDK usesint?. Tentacles' live-tile updater callsset_Count(int?)every frame from a component Update and was trippingMissingMethodException~once per tick - Fix:
FNA.Game.RunLoopnow wraps the finalOnExiting(this, EventArgs.Empty)in try/catch + log. Tentacles'MetricsSender.CreateTearDownExtendedKeysNREs onGlobalManager.Instance.currentProfilewhen the user closes during early boot (currentProfile is null until the preloader finishes loading it). Can't fix the game code, but the host no longer surfaces its "unexpected error" dialog on close - Fix:
SignedInGamer.SignedInhandler invocations now serialise behind aSemaphoreSlim. Tentacles'Game1.Initializeregisters the same callback twice; both invocations were firing in parallel via separateTask.Delay.ContinueWiths, racing on the sharedAchievementContext.Currentand tripping EF Core'sConcurrencyDetectorinsideGamer.GetProfile. Handlers still get their independent ~2 s delay; only the synchronous Invoke runs single-file - Fix: Mouse-as-touch clicks now produce
GestureType.Tapgestures on desktop.SDL_MOUSEMOTIONwas forwarded toINTERNAL_onTouchEventunconditionally, including hover (no button held). Hover motion ranGestureDetector.OnMovedwhich setactiveFingerId = 1; the nextMOUSEBUTTONDOWN'sOnPressed(1)then sawactiveFingerId != NO_FINGERand routed into pinch-init (state = PINCHING). The matchingMOUSEBUTTONUPranOnReleased_Pinchinstead of Tap detection — so Tap-gated screens (Tentacles'LemmyTravelScreen, "tap to continue" prompts) never advanced from mouse but worked fine from real touch (no hover). Fix:SDL2_FNAPlatformskips the synthesised Moved event whenevt.motion.state == 0. Drag (button-held motion) still goes through, so swipe/pinch/drag from mouse continue to work - Feat: Periodic FFWD loading-state heartbeat in FNA
Game.Tick([wpr-heartbeat], fires ~every 2 s past the first-30-ticks verbose trace cap). ReflectsPressPlay.FFWD.ApplicationandPressPlay.Tentacles.Scripts.LevelHandlerstatic fields plus the active screen stack and the current level identity — turned a "stuck on loading screen" symptom into a precise gate-by-gate diagnosis. Best-effort, swallows all reflection errors,[Conditional("DEBUG")]-gated - Feat:
ContentManager.Load<T>now logs the underlying exception with type/HResult/inner/stack before rethrowing ([wpr-content]). FFWD'sAssetHelper.Loadsilently catches ContentManager.Load failures and returnsdefault(T); without the trace, a missing ContentTypeReader or malformed XNB was invisible. Trace lets the caller's swallow stand — only adds visibility - Feat: Assembly-resolver diagnostics in
ApplicationLaunch—[wpr-resolve-user]/[wpr-resolve-default]lines log every Resolving probe with full exception details whenLoadFromAssemblyPathfails
- Fix: Asphalt 5 sat on a blank loading screen —
StartupModeenum integer values now match the WP7 SDK (Launch=1,Activate=2), and FNA'sGame.IsActiveno longer fires a spuriousActivatedevent at first frame - Fix: Tentacles crashed at exit on a null
ApplicationCurrentMemoryUsage—Microsoft.Phone.Info.DeviceExtendedPropertiesnow returns the WP7 memory counters (ApplicationCurrentMemoryUsage,ApplicationPeakMemoryUsage,ApplicationMemoryUsageLimit) - Fix: XNA games that wait on
MediaPlayer/ song-finished callbacks before advancing past their splash now progress —Game.Tickauto-pumpsFrameworkDispatcher.Update()once per update, matching WP7 XNA 4.0 behaviour - Feat: Per-game
wpr_game_debug.logand the in-engine[wpr-trace]output are now#if DEBUG-gated viaWprDebugTrace— Release builds elide the trace formatting and file listener entirely (no log spam, no per-frame cost) - Feat: Richer assembly-resolver diagnostics in
ApplicationLaunch—Resolvingfailures log the underlying exception type, HResult and inner exception instead of the opaque "Operation is not supported" surface error - Compat: Asphalt 5 promoted from Broken → Partial (loads splash, gameplay TBD)
- Compat: Final Fantasy promoted to Playable
- Compat: Fruit Ninja confirmed Playable (no longer blocked on
GamerProfile.GetGamerPicture)
- Fix: Small bug on second launch of XNA games, where resources arent released fully
- Feat: Added Windows based notifications
- Feat: Added game icon as window icon for XNA