Update settings_overlay.cpp

Add early Linux linker dependency probe (#255 )
add bltl (#254 )
2026-09-26 07:01:08 -04:00 · 2026-09-24 22:02:20 +02:00 · 2026-09-24 18:53:56 +02:00 · 2026-09-24 18:06:32 +02:00 · 2026-09-24 13:50:33 +02:00 · 2026-09-23 19:27:51 +02:00
150 changed files with 10598 additions and 2350 deletions
@@ -3,3 +3,4 @@

 # Patch files must stay LF: git apply matches context bytes against LF upstream sources
 *.patch -text
+translator/tests/Translator.Tests/TestAssets/**/*.bin binary
@@ -14,6 +14,10 @@ concurrency:
  cancel-in-progress: true

 jobs:
+  recompilation:
+    name: Recompilation test
+    uses: ./.github/workflows/recomp-test.yml
+
  translator:
    name: Translator (build + test)
    runs-on: windows-latest
@@ -26,6 +30,14 @@ jobs:
        with:
          dotnet-version: '8.0.x'

+      - name: Cache NuGet packages
+        uses: actions/cache@v5
+        with:
+          path: ~/.nuget/packages
+          key: ${{ runner.os }}-nuget-${{ hashFiles('translator/Translator.sln', '**/*.csproj', '**/*.props', '**/*.targets', '**/packages.lock.json', 'global.json', 'NuGet.config', 'nuget.config') }}
+          restore-keys: |
+            ${{ runner.os }}-nuget-
+
      - name: Restore
        run: dotnet restore translator/Translator.sln

@@ -34,46 +46,3 @@ jobs:

      - name: Test
        run: dotnet test translator/Translator.sln -c Release --no-build --verbosity normal
-
-  ios-crosscompile:
-    name: iOS arm64 (cross-compile, no Xcode)
-    runs-on: ubuntu-24.04
-    env:
-      IOS_SDK_DIR: ${{ github.workspace }}/iOS-SDKs/iPhoneOS26.5.sdk
-    steps:
-      - uses: actions/checkout@v7
-        with:
-          persist-credentials: false
-
-      # The runner images carry clang up to 18; the iOS toolchain file is tested
-      # against upstream 19, so take it from apt.llvm.org rather than the archive.
-      - name: Install clang 19, lld and ninja
-        run: |
-          wget -qO /tmp/llvm.sh https://apt.llvm.org/llvm.sh
-          chmod +x /tmp/llvm.sh
-          sudo /tmp/llvm.sh 19
-          sudo apt-get install -y lld-19 ninja-build
-
-      # The only Apple piece in the build. Sparse-cloned so the checkout is one
-      # SDK (~50 MB) rather than every SDK the repository carries.
-      - name: Fetch the iPhoneOS SDK
-        run: |
-          git clone --depth 1 --filter=blob:none --sparse \
-              https://github.com/xybp888/iOS-SDKs.git "${{ github.workspace }}/iOS-SDKs"
-          git -C "${{ github.workspace }}/iOS-SDKs" sparse-checkout set --no-cone iPhoneOS26.5.sdk
-
-      # Compile-only: linking a product needs the translated shards, which come
-      # from the user's own Mario Kart Wii dump and are not in this repository.
-      # This proves the runtime's own sources still build for iOS arm64.
-      - name: Configure
-        run: |
-          cmake -S runtime -B build-ios -G Ninja -DCMAKE_BUILD_TYPE=Release \
-              -DCMAKE_TOOLCHAIN_FILE="${{ github.workspace }}/runtime/cmake/ios-linux-toolchain.cmake" \
-              -DIOS_SDK="$IOS_SDK_DIR" \
-              -DMKW_IOS_LLVM_BIN=/usr/lib/llvm-19/bin \
-              -DMKW_BUILD_PRODUCTS=OFF \
-              -DCMAKE_DISABLE_FIND_PACKAGE_absl=TRUE \
-              -DAURORA_DAWN_PROVIDER=package
-
-      - name: Compile the runtime sources for iOS
-        run: cmake --build build-ios --target mkw_macos_native_compile
@@ -20,6 +20,10 @@ concurrency:
  cancel-in-progress: true

 jobs:
+  recompilation:
+    name: Recompilation test
+    uses: ./.github/workflows/recomp-test.yml
+
  linux-appimage:
    name: Linux (AppImage, ${{ matrix.arch }})
    strategy:
@@ -93,7 +97,7 @@ jobs:
  release:
    name: Publish GitHub Release
    if: startsWith(github.ref, 'refs/tags/v')
-    needs: [linux-appimage, windows-installer]
+    needs: [linux-appimage, windows-installer, recompilation]
    runs-on: ubuntu-latest
    permissions:
      contents: write
@@ -0,0 +1,61 @@
+name: Synthetic recompilation
+
+on:
+  workflow_call:
+  workflow_dispatch:
+
+permissions:
+  contents: read
+
+jobs:
+  windows:
+    name: Windows runtime (synthetic DOL)
+    runs-on: windows-latest
+    timeout-minutes: 60
+    steps:
+      - uses: actions/checkout@v7
+        with:
+          persist-credentials: false
+
+      - uses: actions/setup-dotnet@v6
+        with:
+          dotnet-version: '8.0.x'
+
+      # Cache downloads only. Preparation still validates pins, and every run
+      # compiles current Aurora, runtime, and generated sources from scratch.
+      - uses: actions/cache@v5
+        with:
+          path: Launcher/artifacts/downloads
+          key: windows-recomp-downloads-${{ hashFiles('Launcher/Prepare-PortableTools.ps1', 'Launcher/Prepare-Dependencies.ps1') }}
+
+      # Install sccache.
+      - name: Run sccache-action
+        uses: mozilla/sccache-action@v0.0.11
+
+      # Tell CMake to use sccache and use GitHub's API.
+      - name: Configure sccache environment
+        shell: pwsh
+        run: |
+          "SCCACHE_GHA_ENABLED=true" | Add-Content -Path $env:GITHUB_ENV
+          "ACTIONS_CACHE_SERVICE_V2=on" | Add-Content -Path $env:GITHUB_ENV
+          "CMAKE_C_COMPILER_LAUNCHER=$env:SCCACHE_PATH" | Add-Content -Path $env:GITHUB_ENV
+          "CMAKE_CXX_COMPILER_LAUNCHER=$env:SCCACHE_PATH" | Add-Content -Path $env:GITHUB_ENV
+
+      - name: Prepare the shipped Windows toolchain
+        shell: pwsh
+        run: ./Launcher/Prepare-PortableTools.ps1
+
+      - name: Prepare pinned native dependencies
+        shell: pwsh
+        run: ./Launcher/Prepare-Dependencies.ps1
+
+      # Expose cache token context to the build script.
+      - name: Translate, compile the full runtime, and link
+        shell: pwsh
+        run: ./Launcher/Test-Recompilation.ps1 -Parallel 4
+
+      # Print cache results (even if the build fails)
+      - name: Show sccache stats
+        if: always()
+        shell: pwsh
+        run: sccache --show-stats
@@ -26,6 +26,8 @@ Code.pul
 /build/
 /build-*/
 /native-build/
+/native-build-macos/
+/local-products/
 /dist/
 /out/
 [Bb]in/
@@ -70,3 +72,5 @@ project.lock.json
 *.log
 output.txt

+# Operating System
+.DS_Store
@@ -281,7 +281,7 @@ foreach ($required in @('ToolkitFingerprint','TranslationFingerprint','NativeToo

 $manifest = [ordered]@{
    SchemaVersion = 2
-    ProductVersion = '0.2.27'
+    ProductVersion = '0.2.32'
    ExpectedGameId = $pins.GameId
    ExpectedDolSha256 = $pins.DolSha256
    ExpectedRelSha256 = $pins.RelSha256
@@ -117,12 +117,13 @@ function Get-MkwProjectPins([string]$ProjectFile) {
 }

 function Invoke-Checked([string]$FilePath, [string[]]$Arguments, [string]$Description,
-    [string]$LogPrefix = 'MKWCBUILD', [string]$StepId = '') {
+    [string]$LogPrefix = 'MKWCBUILD', [string]$StepId = '', [bool]$WaitForProcessTree = $true) {
    <#
-    Runs a build tool and turns a non-zero exit code into a described failure. Start-Process -Wait
-    is deliberate: it waits for the whole process tree, since a .NET single-file bundle host may
-    hand off to an extracted child that PowerShell's call operator would not wait for. Start-Process
-    doesn't publish $LASTEXITCODE, so this sets it manually for callers that check it.
+    Runs a build tool and turns a non-zero exit code into a described failure. By default,
+    Start-Process -Wait waits for the whole process tree, since a .NET single-file bundle host may
+    hand off to an extracted child that PowerShell's call operator would not wait for. Callers that
+    need to avoid waiting on unrelated descendants can opt into the call-operator path.
+    Start-Process doesn't publish $LASTEXITCODE, so this sets it manually for callers that check it.
    -StepId emits the machine-readable form the installer's progress bar consumes (BuildStepIds in
    WiiCompiled.Setup/InstallProgress.cs); the human sentence stays on the same log line.
    #>
@@ -132,9 +133,14 @@ function Invoke-Checked([string]$FilePath, [string[]]$Arguments, [string]$Descri
        if ($_.Contains('"')) { throw "A native build argument contains an unsupported quote: $_" }
        '"' + $_ + '"'
    })
-    $process = Start-Process -FilePath $FilePath -ArgumentList $quotedArguments `
-        -NoNewWindow -Wait -PassThru
-    $exitCode = $process.ExitCode
+    if ($WaitForProcessTree) {
+        $process = Start-Process -FilePath $FilePath -ArgumentList $quotedArguments `
+            -NoNewWindow -Wait -PassThru
+        $exitCode = $process.ExitCode
+    } else {
+        & $FilePath @Arguments
+        $exitCode = $LASTEXITCODE
+    }
    $global:LASTEXITCODE = $exitCode
    if ($exitCode -ne 0) { throw "$Description failed with exit code $exitCode." }
 }
@@ -39,9 +39,9 @@ $packages = @(
    },
    [pscustomobject]@{
        Name = 'dawn_prebuilt'; File = 'dawn-v20260603.191052-windows-amd64.tar.gz'
-        Uris = @('https://github.com/encounter/dawn-build/releases/download/v20260603.191052/dawn-windows-amd64.tar.gz')
+        Uris = @('https://github.com/theofficialgman/dawn-build/releases/download/v20260603.191052/dawn-windows-amd64.tar.gz')
        Pins = @(@{ File = $auroraCMake; Text = 'set(AURORA_DAWN_VERSION "v20260603.191052"' },
-                 @{ File = $auroraDawn; Text = 'SHA256=7785373d569b3b0237918ec9c523239f7d0667857c5ea8242e3cdfde95e6aeab' })
+                 @{ File = $auroraDawn; Text = 'SHA256=13be9cff8b9b179c42dcd16aeabb6effcc8f0dfdcc14463eda2a5caeda225142' })
    },
    [pscustomobject]@{
        Name = 'fmt'; File = 'fmt-11.1.4.tar.gz'
@@ -88,7 +88,7 @@ assert_file "$ninja_bin" "Portable Ninja"
 assert_file "$cc" "Portable C compiler"
 assert_file "$cxx" "Portable C++ compiler"
 assert_dir "$aurora_source" "aurora-main source tree"
-clang_binary=$(normalize "$toolchain_dir/bin/clang-23")
+clang_binary=$(normalize "$toolchain_dir/bin/clang-22")
 assert_file "$clang_binary" "Portable clang driver binary"

 (( parallel > 0 )) || parallel=$(nproc)
@@ -1,6 +1,6 @@
 # Fails the release build when a fact duplicated across the repo stops agreeing with the copy
 # that owns it (recomp.yml). Scripts read pinned facts through Get-MkwProjectPins, but three
-# consumers can't read YAML (the C++ runtime header, the C# constants, hand-written lists on
+# consumers can't read YAML (the C++ runtime header, the C# constants, shell scripts, and hand-written lists on
 # both sides of the C#/PowerShell boundary), so those are checked here instead.
 [CmdletBinding()]
 param([string]$RepositoryRoot)
@@ -58,6 +58,12 @@ $hostUri = Get-CapturedValue $retroWfcPayload 'CurrentRetroWfcPayloadUri\s*=\s*"
 if ($hostUri -cne $pins.RetroWfcPayloadUri) {
    Add-Failure "InputValidation.CurrentRetroWfcPayloadUri is '$hostUri' but recomp.yml pins '$($pins.RetroWfcPayloadUri)'."
 }
+$macosSetup = Read-SourceFile (Join-Path $launcher 'macos\setup.command') 'macOS setup.command'
+$macosUri = Get-CapturedValue $macosSetup "'([^']*/api/wfc/payload\?g=RMCPD00)'" `
+    'The macOS Retro-WFC endpoint'
+if ($macosUri -cne $pins.RetroWfcPayloadUri) {
+    Add-Failure "macOS setup.command downloads '$macosUri' but recomp.yml pins '$($pins.RetroWfcPayloadUri)'."
+}

 # --- The game identity: the manifest carries it, but the host also compiles a fallback for a
 # --- manifest that predates the field, and that fallback decides which disc is accepted.
@@ -0,0 +1,148 @@
+# Build the real Windows runtime with translated, entirely synthetic PowerPC code.
+# No game dump, game symbol map, REL, mod download, or existing generated/ output is used.
+[CmdletBinding()]
+param(
+    [string]$PortableToolsDirectory = 'Launcher/artifacts/portable-tools',
+    [string]$DependencySourceDirectory = 'Launcher/artifacts/dependencies',
+    [string]$StageDirectory = 'build/recomp-test',
+    [ValidateRange(1, 64)] [int]$Parallel = 3
+)
+
+$ErrorActionPreference = 'Stop'
+Set-StrictMode -Version 3.0
+. (Join-Path $PSScriptRoot 'NativeBuildFlags.ps1')
+
+$repoRoot = [IO.Path]::GetFullPath((Join-Path $PSScriptRoot '..'))
+function Full([string]$Path) {
+    if ([IO.Path]::IsPathRooted($Path)) { return [IO.Path]::GetFullPath($Path) }
+    return [IO.Path]::GetFullPath((Join-Path $repoRoot $Path))
+}
+$portableTools = Full $PortableToolsDirectory
+$dependencies = Full $DependencySourceDirectory
+$stage = Full $StageDirectory
+$dotnet = (Get-Command dotnet -CommandType Application).Source
+$cmake = Join-Path $portableTools 'CMake/bin/cmake.exe'
+$ninja = Join-Path $portableTools 'Ninja/ninja.exe'
+$compilerBin = Join-Path $portableTools 'llvm-mingw/bin'
+Assert-File $cmake 'Pinned CMake (run Prepare-PortableTools.ps1 first)'
+Assert-File $ninja 'Pinned Ninja'
+Assert-File (Join-Path $dependencies 'cppwinrt/winrt/base.h') 'Pinned dependencies (run Prepare-Dependencies.ps1 first)'
+
+# Refuse reuse so a developer's game translation or an earlier build cannot make
+# the test pass. Keep the staging tree after the run for diagnostics.
+if (Test-Path -LiteralPath $stage) { throw "Test stage already exists; choose a fresh -StageDirectory: $stage" }
+[IO.Directory]::CreateDirectory($stage) | Out-Null
+Write-Host "Synthetic recompilation workspace: $stage"
+
+# Copy current sources, including uncommitted edits, but no ignored build output.
+# An isolated workspace preserves the developer's real generated/ directory.
+$sourceFiles = & git -C $repoRoot -c core.quotepath=false ls-files --cached --others --exclude-standard -- runtime aurora-main
+if ($LASTEXITCODE -ne 0) { throw 'Could not enumerate runtime and Aurora sources.' }
+foreach ($relative in $sourceFiles | Sort-Object -Unique) {
+    $destination = Join-Path $stage $relative
+    [IO.Directory]::CreateDirectory([IO.Path]::GetDirectoryName($destination)) | Out-Null
+    Copy-Item -LiteralPath (Join-Path $repoRoot $relative) -Destination $destination
+}
+
+# One synthetic text section: li r3,40; addi r3,r3,2; nop; blr.
+# Native HLE wrappers also call these eight guest symbols directly. Give each
+# its own generated blr function so the real product can link without game code.
+# Keep this list explicit: a new unresolved guest dependency must fail the test.
+[uint32]$entry = 0x80001000L
+[uint32[]]$guestCallbacks = @(
+    0x8012B830L, 0x801A0620L, 0x801A1ED8L, 0x801A961CL,
+    0x801AADE0L, 0x801D8D30L, 0x801D9E94L, 0x8055531CL
+)
+$textSize = [int]($guestCallbacks[-1] - $entry + 4)
+$dataOffset = 0x100 + $textSize
+$dol = [byte[]]::new($dataOffset + 4)
+function Write-BigEndian32([int]$Offset, [uint32]$Value) {
+    $dol[$Offset] = [byte](($Value -shr 24) -band 255)
+    $dol[$Offset + 1] = [byte](($Value -shr 16) -band 255)
+    $dol[$Offset + 2] = [byte](($Value -shr 8) -band 255)
+    $dol[$Offset + 3] = [byte]($Value -band 255)
+}
+Write-BigEndian32 0x00 0x100       # text[0] file offset
+Write-BigEndian32 0x48 $entry      # text[0] guest address
+Write-BigEndian32 0x90 $textSize   # text[0] length (unreachable gaps are zero)
+Write-BigEndian32 0x1C $dataOffset # data[0] file offset
+Write-BigEndian32 0x64 0x80600000L # data[0] guest address
+Write-BigEndian32 0xAC 4           # data[0] length
+Write-BigEndian32 0xD8 0x80601000L # BSS address
+Write-BigEndian32 0xDC 32          # BSS length
+Write-BigEndian32 0xE0 $entry      # entry point
+Write-BigEndian32 0x100 0x38600028 # li r3,40
+Write-BigEndian32 0x104 0x38630002 # addi r3,r3,2
+Write-BigEndian32 0x108 0x60000000 # nop
+Write-BigEndian32 0x10C 0x4E800020 # blr
+foreach ($address in $guestCallbacks) {
+    Write-BigEndian32 ([int](0x100 + $address - $entry)) 0x4E800020
+}
+Write-BigEndian32 $dataOffset 0x12345678
+[IO.File]::WriteAllBytes((Join-Path $stage 'synthetic.dol'), $dol)
+$entryPoints = (@($entry) + $guestCallbacks | ForEach-Object { '0x{0:X8}' -f $_ }) -join ', '
+$functionMap = (@($entry) + $guestCallbacks | ForEach-Object { '{0:X8} func_{0:X8}' -f $_ }) -join "`n"
+[IO.File]::WriteAllText((Join-Path $stage 'synthetic-functions.txt'), $functionMap)
+$manifest = Join-Path $stage 'recomp.yml'
+[IO.File]::WriteAllText($manifest, @"
+schema_version: 1
+workspace_root: .
+project:
+  id: ci-synthetic-dol
+  display_name: CI Synthetic DOL
+memory:
+  base: 0x80000000
+  size: 0x01800000
+  sda_base: 0x80600000
+  sda2_base: 0x80600000
+inputs:
+  dol:
+    path: synthetic.dol
+translation:
+  entry_points: [$entryPoints]
+  function_map:
+    path: synthetic-functions.txt
+  allow_unsupported_instructions: false
+runtime:
+  native_abi_directories: []
+  native_registration_root: runtime/src
+output:
+  root: generated
+"@)
+
+$translatorProject = Join-Path $repoRoot 'translator/src/Translator.Cli/Translator.Cli.csproj'
+Invoke-Checked $dotnet @('build', $translatorProject, '-c', 'Release', '--disable-build-servers') 'Building the translator'
+$translator = Join-Path $repoRoot 'translator/src/Translator.Cli/bin/Release/net8.0/Translator.Cli.dll'
+$metadata = Join-Path $stage 'generated/base_translation_output.json'
+Invoke-Checked $dotnet @($translator, 'translate-recursive', '0x80001000', '--project', $manifest,
+    '--output-metadata', $metadata, '--threads', "$Parallel") `
+    'Translating the synthetic DOL'
+# Function-map seeds can be skipped by discovery; do not accept a partial fixture.
+$translated = Get-Content -LiteralPath $metadata -Raw | ConvertFrom-Json
+foreach ($address in @($entry) + $guestCallbacks) {
+    if ($address -notin $translated.functions.entryPoint) {
+        throw ('Synthetic function 0x{0:X8} was not translated.' -f $address)
+    }
+}
+Invoke-Checked $dotnet @($translator, 'generate-data-init', '--project', $manifest) 'Generating synthetic data and runtime configuration'
+Invoke-Checked $dotnet @($translator, 'emit-build-shards', '--project', $manifest) 'Emitting the production build graph'
+
+$nativeBuild = Join-Path $stage 'native-build'
+$oldPath = $env:PATH
+try {
+    $env:PATH = Get-MkwToolchainPath $portableTools
+    $configure = Get-MkwNativeConfigureArguments -SourceDirectory (Join-Path $stage 'runtime') -BuildDirectory $nativeBuild `
+        -Ninja $ninja -CCompiler (Join-Path $compilerBin 'x86_64-w64-mingw32-clang.exe') `
+        -CxxCompiler (Join-Path $compilerBin 'x86_64-w64-mingw32-clang++.exe') `
+        -ResourceCompiler (Join-Path $compilerBin 'x86_64-w64-mingw32-windres.exe') `
+        -DependenciesDirectory $dependencies -AdditionalArguments @('-DMKW_BUILD_PRODUCTS=ON')
+    Invoke-Checked $cmake $configure 'Configuring the production Windows runtime' `
+        -WaitForProcessTree $false
+    Invoke-Checked $cmake @('--build', $nativeBuild, '--target', 'WiiCompiled', '--parallel', "$Parallel") `
+        'Compiling and linking the synthetic product with the full runtime' `
+        -WaitForProcessTree $false
+    Assert-File (Join-Path $nativeBuild 'WiiCompiled.exe') 'Linked synthetic product'
+} finally {
+    $env:PATH = $oldPath
+}
+Write-Host 'Synthetic recompilation passed (translation, data generation, runtime compilation, and product link).'
@@ -6,7 +6,7 @@
    <Nullable>enable</Nullable>
    <RootNamespace>WiiCompiled.Setup.Common.Cli</RootNamespace>
    <AssemblyName>WiiCompiled.Setup.Common.Cli</AssemblyName>
-    <Version>0.2.22</Version>
+    <Version>0.2.32</Version>
    <Authors>patchzy</Authors>
    <Product>WiiCompiled</Product>
    <Description>Packaging-time helper: resolves (downloading if needed) the nodtool binary bundled by build-appimage.sh and Build-Installer.ps1</Description>
@@ -24,7 +24,7 @@ public static class RetroWfcPayload
    private static readonly TimeSpan RetroWfcDownloadTimeout = TimeSpan.FromSeconds(30);
    private static readonly TimeSpan RetroWfcRetryDelay = TimeSpan.FromSeconds(1);

-    public const string CurrentRetroWfcPayloadUri = "http://nas.play.rwfc.net/payload?g=RMCPD00";
+    public const string CurrentRetroWfcPayloadUri = "https://rwfc.net/api/wfc/payload?g=RMCPD00";
    private static readonly string RetroWfcOfflinePayloadFile =
        Path.Combine("binary", "payload.RMCPD00.bin");

@@ -5,7 +5,7 @@
    <Nullable>enable</Nullable>
    <RootNamespace>WiiCompiled.Setup.Common</RootNamespace>
    <AssemblyName>WiiCompiled.Setup.Common</AssemblyName>
-    <Version>0.2.22</Version>
+    <Version>0.2.32</Version>
    <Authors>patchzy</Authors>
    <Product>WiiCompiled</Product>
    <Description>Shared nodtool/Retro-WFC-payload logic used by both the Windows and Linux installers</Description>
@@ -13,7 +13,8 @@ internal static class BuildRunner
        string workspace, string profile, string outputDir, string? baseOutputDir,
        string? retroDir, string? retroWfcOfflineDir, bool skipRetroWfcPayload,
        bool forceCleanBuild, string? translatorBin, string? ccBin, string? cxxBin, string? fuseLd,
-        string? cmakeBin, string? ninjaBin, string? nativePrebuiltDir, IInstallReporter reporter,
+        string? cmakeBin, string? ninjaBin, string? nativePrebuiltDir, string? sysroot,
+        IInstallReporter reporter,
        CancellationToken cancellationToken)
    {
        var script = Path.Combine(workspace, "Launcher", "local-build.sh");
@@ -77,6 +78,10 @@ internal static class BuildRunner
        {
            startInfo.ArgumentList.Add("--native-prebuilt-dir"); startInfo.ArgumentList.Add(nativePrebuiltDir);
        }
+        if (!string.IsNullOrEmpty(sysroot))
+        {
+            startInfo.ArgumentList.Add("--sysroot"); startInfo.ArgumentList.Add(sysroot);
+        }

        using var process = new Process { StartInfo = startInfo };
        var window = new BuildProgressWindow(reporter, InstallStages.Build, start: 6, end: 96);
@@ -3,7 +3,7 @@ namespace WiiCompiled.Setup.Linux;
 internal static class ProductInfo
 {
    public const string Name = "WiiCompiled";
-    public const string Version = "0.2.27";
+    public const string Version = "0.2.32";
 }

 /// <summary>One installed product's record inside install-state.json.</summary>
@@ -143,6 +143,15 @@ internal static class Program
            retroWfcOfflineDir = cacheDir;
        }

+        var sysroot = flags.GetValueOrDefault("sysroot");
+        // --sysroot explicitly provided (even as bare flag at end of argv, which ParseArgs
+        // stores as null) must carry a path; omitting --sysroot entirely is fine (local-build.sh
+        // adds -UCMAKE_SYSROOT to clear any stale cached value from a prior configure).
+        if (flags.ContainsKey("sysroot") && string.IsNullOrWhiteSpace(sysroot))
+        {
+            throw new ArgumentException("--sysroot requires a non-empty directory path.");
+        }
+
        await BuildRunner.RunAsync(
            workspace, profile, installDir, baseInstallDir,
            retroDir,
@@ -156,6 +165,7 @@ internal static class Program
            flags.GetValueOrDefault("cmake"),
            flags.GetValueOrDefault("ninja"),
            flags.GetValueOrDefault("native-prebuilt-dir"),
+            sysroot,
            reporter, token);

        reporter.Progress(InstallStages.Shortcuts, "Creating shortcuts", 98);
@@ -324,7 +334,7 @@ internal static class Program
                  {--download-retro-wfc-payload | --skip-retro-wfc-payload}]
                  [--force-clean-build] [--translator-bin PATH] [--disc-tool-bin PATH]
                  [--cc PATH] [--cxx PATH] [--fuse-ld NAME_OR_PATH] [--cmake PATH] [--ninja PATH]
-                  [--native-prebuilt-dir DIR] [--progress-json] [--workspace DIR]
+                  [--native-prebuilt-dir DIR] [--sysroot PATH] [--progress-json] [--workspace DIR]
          uninstall
          launch-base
          launch-retro
@@ -6,7 +6,7 @@
    <Nullable>enable</Nullable>
    <AssemblyName>WiiCompiled.Setup.Linux</AssemblyName>
    <RootNamespace>WiiCompiled.Setup.Linux</RootNamespace>
-    <Version>0.2.22</Version>
+    <Version>0.2.32</Version>
    <Authors>patchzy</Authors>
    <Product>WiiCompiled</Product>
    <Description>Command-line installer and launcher for WiiCompiled on Linux</Description>
@@ -121,7 +121,7 @@ internal static class PlatformChecks
 internal static class ProductInfo
 {
    public const string Name = "WiiCompiled";
-    public const string Version = "0.2.27";
+    public const string Version = "0.2.32";

    /// <summary>
    /// The setup executable is copied into the installation under this name. It is the launcher and
@@ -7,7 +7,7 @@
    <AssemblyName>WiiCompiled.Setup</AssemblyName>
    <RootNamespace>WiiCompiled.Setup.Windows</RootNamespace>
    <ApplicationManifest>app.manifest</ApplicationManifest>
-    <Version>0.2.27</Version>
+    <Version>0.2.32</Version>
    <Authors>patchzy</Authors>
    <Product>WiiCompiled</Product>
    <Description>Command-line installer and launcher for WiiCompiled</Description>
@@ -65,6 +65,7 @@ translator_dll_override=""
 translator_bin_override=""
 fuse_ld_override=""
 native_prebuilt_dir=""
+sysroot=""

 usage() {
    cat <<'EOF'
@@ -87,6 +88,8 @@ Usage: local-build.sh --output-dir DIR [options]
  --translator-bin PATH           Self-contained Translator.Cli executable (skips building AND needs no dotnet at all)
  --native-prebuilt-dir DIR       Precompiled aurora/third-party package (see Prepare-NativePrebuilt.sh);
                                   skips compiling aurora-main from source entirely
+  --sysroot PATH                   Passed to CMake as -DCMAKE_SYSROOT: where the compiler resolves
+                                   standard headers/startup files
 EOF
 }

@@ -110,6 +113,7 @@ while [[ $# -gt 0 ]]; do
        --translator-dll) translator_dll_override=$2; shift 2 ;;
        --translator-bin) translator_bin_override=$2; shift 2 ;;
        --native-prebuilt-dir) native_prebuilt_dir=$2; shift 2 ;;
+        --sysroot) sysroot=$2; shift 2 ;;
        -h|--help) usage; exit 0 ;;
        *) fail "unknown argument: $1" ;;
    esac
@@ -186,6 +190,30 @@ assert_file "$project" "Translation project"
 assert_file "$assets/main.dol" "Extracted main.dol (see translator/README.md - owning the game is required)"
 assert_file "$assets/StaticR.rel" "Extracted StaticR.rel (see translator/README.md - owning the game is required)"

+# The AppImage bundles Clang, but Linux startup objects and the C/C++ link runtimes
+# still come from the host. Check them before the expensive translation so a missing
+# development package produces a useful error instead of CMake's generic exit 1.
+link_probe_dir=$(mktemp -d)
+link_probe_flags=()
+[[ -z "$sysroot" ]] || link_probe_flags+=(--sysroot="$sysroot")
+[[ -z "$fuse_ld_override" ]] || link_probe_flags+=(-fuse-ld="$fuse_ld_override")
+printf 'int main(void) { return 0; }\n' > "$link_probe_dir/probe.c"
+cat > "$link_probe_dir/probe.cpp" <<'EOF'
+#include <vector>
+int main() { std::vector<int> values{1}; return values.front() - 1; }
+EOF
+if ! "$cc_bin" "${link_probe_flags[@]}" "$link_probe_dir/probe.c" -o "$link_probe_dir/probe-c" > "$link_probe_dir/error" 2>&1; then
+    cat "$link_probe_dir/error" >&2
+    rm -rf "$link_probe_dir"
+    fail "The C compiler cannot link a test program. Linux needs C development files (glibc startup objects and a compiler runtime) in addition to bundled Clang. Install your distribution's development packages, or on SteamOS run WiiCompiled through the Wheel Wizard Flatpak."
+fi
+if ! "$cxx_bin" "${link_probe_flags[@]}" "$link_probe_dir/probe.cpp" -o "$link_probe_dir/probe-cxx" > "$link_probe_dir/error" 2>&1; then
+    cat "$link_probe_dir/error" >&2
+    rm -rf "$link_probe_dir"
+    fail "The C++ compiler cannot link a test program. Install your distribution's C++ development packages, or on SteamOS run WiiCompiled through the Wheel Wizard Flatpak."
+fi
+rm -rf "$link_probe_dir"
+
 # Literal line matching against the manifest's fixed shape, not a YAML dependency - the same
 # approach NativeBuildFlags.ps1's Get-MkwProjectPins uses on Windows, kept here only for the one
 # field this script actually needs from the manifest.
@@ -414,6 +442,14 @@ fi
 if [[ -n "$native_prebuilt_dir" ]]; then
    configure_args+=(-DMKW_NATIVE_PREBUILT_DIR="$native_prebuilt_dir")
 fi
+if [[ -n "$sysroot" ]]; then
+    configure_args+=(-DCMAKE_SYSROOT="$sysroot")
+else
+    # Explicitly clear any cached CMAKE_SYSROOT from a prior configure so an
+    # incremental build that transitions from one sysroot to none does not
+    # silently keep the stale cached path.
+    configure_args+=(-UCMAKE_SYSROOT)
+fi

 log_step configure-native "Configuring the native toolchain"
 "$cmake_bin" "${configure_args[@]}"
@@ -445,7 +481,9 @@ publish_built_product() {
    local exe=$build/$target
    assert_file "$exe" "Locally compiled game executable"
    cp -f "$exe" "$destination/$target"
-    for name in dsp_coef.bin initial_pipeline_cache.db; do
+    # cacert.pem is the TLS root bundle the mbed TLS path looks up beside the executable
+    # (runtime/src/hle/net/network_ssl.cpp); without it HTTPS fails at runtime.
+    for name in dsp_coef.bin initial_pipeline_cache.db cacert.pem; do
        [[ -f "$build/$name" ]] && cp -f "$build/$name" "$destination/"
    done
    [[ -d "$build/wii_bootstrap" ]] && cp -rf "$build/wii_bootstrap" "$destination/"
@@ -27,7 +27,7 @@ done
 [[ "$product" == WiiCompiled || "$product" == RetroRewind ]] || fail '--product must be WiiCompiled or RetroRewind'
 for tool in codesign ditto install_name_tool otool; do command -v "$tool" >/dev/null || fail "required macOS tool is unavailable: $tool"; done
 [[ -x "$build_dir/$product" ]] || fail "missing compiled product: $build_dir/$product"
-for asset in dsp_coef.bin initial_pipeline_cache.db wii_bootstrap; do [[ -e "$build_dir/$asset" ]] || fail "missing runtime asset: $build_dir/$asset"; done
+for asset in dsp_coef.bin initial_pipeline_cache.db cacert.pem wii_bootstrap; do [[ -e "$build_dir/$asset" ]] || fail "missing runtime asset: $build_dir/$asset"; done

 app="$output_dir/$product.app"
 macos="$app/Contents/MacOS"
@@ -52,7 +52,7 @@ cat > "$app/Contents/Info.plist" <<EOF
 </dict></plist>
 EOF
 ditto "$build_dir/$product" "$macos/$product"
-for asset in dsp_coef.bin initial_pipeline_cache.db wii_bootstrap; do
+for asset in dsp_coef.bin initial_pipeline_cache.db cacert.pem wii_bootstrap; do
    ditto "$build_dir/$asset" "$resources/$asset"
    ln -s "../Resources/$asset" "$macos/$asset"
 done
@@ -90,7 +90,7 @@ if [[ -n "$retro_dir" ]]; then
        trap 'rm -rf "$payload_stage"' EXIT
        /usr/bin/curl --fail --silent --show-error --connect-timeout 10 --max-time 30 \
            --retry 1 --output "$temporary_payload" \
-            'http://nas.play.rwfc.net/payload?g=RMCPD00' || fail 'could not download the Retro-WFC payload needed for online play'
+            'https://rwfc.net/api/wfc/payload?g=RMCPD00' || fail 'could not download the Retro-WFC payload needed for online play'
        "$translator" validate-retro-wfc-payload --directory "$payload_stage" || \
            fail 'downloaded Retro-WFC payload failed signature validation'
        mkdir -p "$retro_wfc_dir/binary"
@@ -25,7 +25,7 @@ set -euo pipefail
 script_dir=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
 workspace=$(cd "$script_dir/.." && pwd)

-llvm_version=23.1.0
+llvm_version=22.1.8
 cmake_version=4.3.3
 ninja_version=1.13.2
 destination="$script_dir/artifacts/portable-tools"
@@ -52,13 +52,13 @@ done

 case "$arch" in
    x86_64) llvm_release_arch=X64; target_triple=x86_64-unknown-linux-gnu
-            llvm_release_sha256=18da30f77f475688a18f7704d23f9f155ae007ed9922dbed6850a9419d9fec8c
+            llvm_release_sha256=fccecb1906e7ddf5ec040aec5b646b650e2daaafa4423b41341c4717db5bdec0
            cmake_release_arch=x86_64
            cmake_sha256=927b2368a946c37269c3a66225ab00544e756459cdd0b5d0da438694fb9ff802
            ninja_asset=ninja-linux.zip
            ninja_sha256=5749cbc4e668273514150a80e387a957f933c6ed3f5f11e03fb30955e2bbead6 ;;
    aarch64) llvm_release_arch=ARM64; target_triple=aarch64-unknown-linux-gnu
-            llvm_release_sha256=cfb31bfc713ef453248bf5bd026312f838ad6c52c25623e987cb6a340f3050d4
+            llvm_release_sha256=d431eff9f064c86ee7c4c94af570a8f74fcccd1f74c6f0da3af32ce34a1e1b05
            cmake_release_arch=aarch64
            cmake_sha256=9ea38356dbd3e32e51029a3e09a0f2f8e117ef4fbcaad7a21ffb36409bbd5cb4
            ninja_asset=ninja-linux-aarch64.zip
@@ -101,11 +101,14 @@ rm -rf "$work"
 mkdir -p "$work/bin" "$work/lib/$target_triple" "$work/include/$target_triple/c++/v1"

 # --- clang/lld/llvm-ar, pruned from the official LLVM release ---
-
-llvm_archive_name="LLVM-$llvm_version-Linux-$llvm_release_arch.tar.xz"
+# built from PR https://github.com/llvm/llvm-project/pull/222821 on official LLVM Github Actions Runner
+# only switch to an official stable LLVM release again once:
+# - this PR has merged https://github.com/llvm/llvm-project/pull/221365 and been backported to LLVM stable branch
+# - this bug has been fixed with a workaround in the Wiicompiled translator https://github.com/patchzyy/Wiicompiled/issues/208 or in LLVM and been backported to LLVM stable branch
+llvm_archive_name="LLVM-PR222821-5ae1c7c43a11b4cdc5ce4dd483c28357bab7dae2-Linux-$llvm_release_arch.tar.xz"
 llvm_archive="$downloads/$llvm_archive_name"
 download_verified "$llvm_archive" \
-    "https://github.com/llvm/llvm-project/releases/download/llvmorg-$llvm_version/$llvm_archive_name" \
+    "https://github.com/theofficialgman/llvm-project/releases/download/llvmorg-22.1.8-patched/$llvm_archive_name" \
    "$llvm_release_sha256"

 extract_root="$script_dir/artifacts/.extract-clang-$arch"
@@ -113,16 +116,16 @@ rm -rf "$extract_root"
 mkdir -p "$extract_root"
 echo "prepare-portable-tools.sh: extracting $llvm_archive_name (this is the full ~1.9 GiB release; only a fraction is kept)..."
 tar -xf "$llvm_archive" -C "$extract_root"
-src="$extract_root/LLVM-$llvm_version-Linux-$llvm_release_arch"
+src="$extract_root/${llvm_archive_name%.tar.xz}"
 [[ -d "$src" ]] || { echo "prepare-portable-tools.sh: unexpected archive layout, expected $src" >&2; exit 1; }

 echo "prepare-portable-tools.sh: pruning to the minimal compile+link toolchain..."

 # clang: the real driver executable plus the clang/clang++ symlinks CMake/local-build.sh invoke.
 # Stripped: debug symbols are dead weight for a bundled compiler nobody will debug.
-cp -a "$src/bin/clang-23" "$work/bin/"
-strip "$work/bin/clang-23"
-ln -s clang-23 "$work/bin/clang"
+cp -a "$src/bin/clang-22" "$work/bin/"
+strip "$work/bin/clang-22"
+ln -s clang-22 "$work/bin/clang"
 ln -s clang "$work/bin/clang++"

 # lld: linked via -fuse-ld=lld, which clang resolves by looking for ld.lld next to itself first -
@@ -165,7 +168,8 @@ rm -rf "$cmake_extract_root"
 mkdir -p "$cmake_extract_root"
 echo "prepare-portable-tools.sh: extracting $cmake_archive_name..."
 tar -xzf "$cmake_archive" -C "$cmake_extract_root"
-cmake_src="$cmake_extract_root/cmake-$cmake_version-linux-$cmake_release_arch"
+
+cmake_src="$cmake_extract_root/${cmake_archive_name%.tar.gz}"
 [[ -d "$cmake_src" ]] || { echo "prepare-portable-tools.sh: unexpected archive layout, expected $cmake_src" >&2; exit 1; }

 mkdir -p "$work/share/cmake-$cmake_share_version"
@@ -201,10 +205,10 @@ Ninja $ninja_version
  Apache License 2.0
 EOF

-echo "prepare-portable-tools.sh: smoke-testing the toolchain..."
-smoke_dir=$(mktemp -d)
-trap 'rm -rf "$smoke_dir"' EXIT
-cat > "$smoke_dir/t.cpp" <<'EOF'
+echo "prepare-portable-tools.sh: testing the toolchain..."
+test_dir=$(mktemp -d)
+trap 'rm -rf "$test_dir"' EXIT
+cat > "$test_dir/t.cpp" <<'EOF'
 #include <vector>
 #include <cstdio>
 int main() {
@@ -214,24 +218,25 @@ int main() {
    return sum == 6 ? 0 : 1;
 }
 EOF
-"$work/bin/clang++" -std=c++20 -fuse-ld=lld "$smoke_dir/t.cpp" -o "$smoke_dir/t"
-"$smoke_dir/t"
+"$work/bin/clang++" -std=c++20 -fuse-ld=lld "$test_dir/t.cpp" -o "$test_dir/t"
+"$test_dir/t"

 # Also exercised together through CMake+Ninja, exactly how local-build.sh drives them - a plain
 # clang++ invocation above would not catch a broken CMAKE_ROOT (Modules/Templates) or a Ninja that
 # can't find the compiler.
-cat > "$smoke_dir/CMakeLists.txt" <<'EOF'
+cat > "$test_dir/CMakeLists.txt" <<'EOF'
 cmake_minimum_required(VERSION 3.16)
-project(smoke CXX)
-add_executable(smoke t.cpp)
+project(test CXX)
+add_executable(test t.cpp)
 EOF
-"$work/bin/cmake" -S "$smoke_dir" -B "$smoke_dir/build" -G Ninja \
+"$work/bin/cmake" -S "$test_dir" -B "$test_dir/build" -G Ninja \
    -DCMAKE_MAKE_PROGRAM="$work/bin/ninja" -DCMAKE_CXX_COMPILER="$work/bin/clang++" >/dev/null
-"$work/bin/cmake" --build "$smoke_dir/build" >/dev/null
-"$smoke_dir/build/smoke"
+"$work/bin/cmake" --build "$test_dir/build" >/dev/null
+"$test_dir/build/test"

-rm -rf "$smoke_dir"
+rm -rf "$test_dir"
 trap - EXIT

+rm -rf "$toolchain_dir"
 mv "$work" "$toolchain_dir"
 echo "prepare-portable-tools.sh: toolchain ready at $toolchain_dir ($(du -sh "$toolchain_dir" | cut -f1))"
@@ -1,6 +1,19 @@
+<img width="4190" height="1232" alt="wiicomplogofinalfinalfinalev2MADEBY_INKWRECK_plzcredit" src="https://github.com/user-attachments/assets/df7a3f2e-5336-479a-b4c0-968dd578726d" />

 # WiiCompiled

+<p align="center">
+  <a href="https://github.com/patchzyy/Wiicompiled/releases"><img alt="Windows 10 / 11, x64" src="https://img.shields.io/badge/Windows-10%20%2F%2011%20%C2%B7%20x64-0078D4"></a>
+  <a href="https://github.com/patchzyy/Wiicompiled/releases"><img alt="Linux, x64 / ARM64" src="https://img.shields.io/badge/Linux-x64%20%2F%20ARM64-FCC624?logo=linux&amp;logoColor=white"></a>
+  <a href="https://github.com/patchzyy/Wiicompiled/releases"><img alt="macOS 14+, Apple Silicon" src="https://img.shields.io/badge/macOS-14%2B%20%C2%B7%20Apple%20Silicon-0A84FF?logo=apple&amp;logoColor=white"></a>
+</p>
+<p align="center">
+  <a href="#building-from-source"><img alt="PowerPC static recompilation" src="https://img.shields.io/badge/PowerPC-static%20recompilation-FF9F0A"></a>
+  <a href="#retro-rewind"><img alt="Retro Rewind supported" src="https://img.shields.io/badge/Retro%20Rewind-supported-FF375F"></a>
+  <a href="https://github.com/TeamWheelWizard/WheelWizard/releases"><img alt="Install with Wheel Wizard" src="https://img.shields.io/badge/install%20with-Wheel%20Wizard-8B5CF6"></a>
+  <a href="LICENSE"><img alt="License: GPLv3" src="https://img.shields.io/badge/license-GPLv3-2EA44F?logo=gnu&amp;logoColor=white"></a>
+</p>
+
 A native PC port of Mario Kart Wii, made with static recompilation.

 There's no emulator in the loop, no interpreter, no JIT, no PowerPC
@@ -46,35 +59,26 @@ Press **F10** while the game window has focus:
 - Internal resolution
 - FPS counter
 - Controller assignment for all four ports
- Full per-controller button mapping
+- Full per-controller button mapping, including the bumpers
+- Dolphin-syntax input expressions and GCPadNew.ini import
+- Controller vibration on/off
 - Volume, instant mute, and the music ducking toggle

 Everything you change is saved to `Config.toml` on the spot and restored next launch.

-**Real controller support.** 
-Controllers are fed to the game as a GameCube controller.
-Mappings are positional (`south`, `east`, `west`, `north`) rather than Xbox-labelled, so the
-same config makes sense on Xbox, PlayStation, Nintendo and generic SDL pads alike, and extra
-inputs like paddles, touchpads and share buttons show up when the hardware reports them.
+**Dolphin-compatible input expressions.** 
+Each GameCube control can carry an expression in Dolphin's input syntax, with the same operators
+and the same functions.
+A Dolphin `GCPadNew.ini` can be imported directly from the F10 bar.
+
+**Vibration toggle.** 
+Force feedback can be turned off for every port at once.
 The official Wii U / Switch GameCube adapter (WUP-028) works too; as with Dolphin, on Windows the
 adapter must be switched to the WinUSB driver once (Zadig).

 **Real Wii Remotes over Bluetooth.**
 Pair a Wii Remote with Windows (Settings > Bluetooth > Add device, press 1+2 or SYNC, leave the
-PIN empty) and the game reads it as an actual Wii Remote through KPAD: Wii Remote icons and
-prompts, Wii Wheel tilt steering, wheelies and tricks all come from the game's own motion code.
-Nunchuk and Classic Controller are real Wii extensions too: the game gets the Nunchuk's stick,
-C/Z and accelerometer, and the Classic Controller through `KPADGetUnifiedWpadStatus` with its own
-layout and icons, so its buttons do what the game says they do and no mapping is involved. Plug an
-extension in or pull it out mid-game and the game switches control scheme like on the console
-(the runtime patches SDL's Wii driver, which otherwise loses the remote for good on an extension
-change). Only the Wii U Pro Controller, which has no Wii-era equivalent, is fed to the game as a
-GameCube pad with Nintendo's layout. If a remote drops out or was switched on after launch, the
-runtime keeps rescanning Bluetooth until it comes back (F10 > Controller settings > Wii Remotes). SDL's read of
-the remote's factory accelerometer calibration often times out over Bluetooth (`console.log`
-then says "Using fallback accelerometer calibration") and it falls back to a nominal zero point,
-so the same menu has a one-button calibration (remote flat, buttons up) that removes the small
-tilt offset some remotes show.
+PIN empty)

 Known limitations of the Wii Remote path:
 - No IR pointer yet: menus are navigated with the D-pad and A (the game treats the remote as
@@ -115,92 +119,6 @@ Wheel Wizard downloads the setup tool from this repo and walks you through insta
 launching. The backend itself is deliberately command-line only, Wheel Wizard is a wrapper around it.


-### iOS
-
-Requires macOS with Xcode. Build the `WiiCompiled` target for iOS arm64; the build writes
-`WiiCompiled-unsigned.ipa` next to the app bundle.
-
-```sh
-cmake -S runtime -B build-ios -G Ninja -DCMAKE_BUILD_TYPE=Release \
-    -DCMAKE_SYSTEM_NAME=iOS -DCMAKE_SYSTEM_PROCESSOR=arm64 \
-    -DCMAKE_OSX_ARCHITECTURES=arm64 -DCMAKE_OSX_SYSROOT=iphoneos \
-    -DCMAKE_OSX_DEPLOYMENT_TARGET=17.0 \
-    -DCMAKE_FIND_ROOT_PATH_MODE_PACKAGE=BOTH \
-    -DCMAKE_FIND_ROOT_PATH_MODE_LIBRARY=ONLY \
-    -DCMAKE_FIND_ROOT_PATH_MODE_INCLUDE=ONLY \
-    -DCMAKE_DISABLE_FIND_PACKAGE_absl=TRUE \
-    -DAURORA_DAWN_PROVIDER=package
-cmake --build build-ios --target WiiCompiled
-```
-
-#### iOS from Linux
-
-The same .ipa builds on a Linux host with upstream clang and lld; no Theos, xtool or Xcode. Tested
-on Debian 13 with `clang-19 lld-19 llvm-19 cmake ninja-build git`. The only Apple piece is the
-iPhoneOS SDK, and [xybp888/iOS-SDKs](https://github.com/xybp888/iOS-SDKs) carries current ones:
-
-```sh
-git clone --depth 1 --filter=blob:none --sparse https://github.com/xybp888/iOS-SDKs.git /opt/iOS-SDKs
-git -C /opt/iOS-SDKs sparse-checkout set --no-cone iPhoneOS26.5.sdk
-```
-
-Pass that directory as `IOS_SDK` (or copy `iPhoneOS.sdk` out of a Mac's Xcode to `/opt/iPhoneOS.sdk`,
-the default). The translated shard manifest under `generated/` records absolute paths from the
-machine that ran the translator, so either translate on the Linux host or symlink that path to your
-checkout.
-
-```sh
-cmake -S runtime -B build-ios -G Ninja -DCMAKE_BUILD_TYPE=Release \
-    -DCMAKE_TOOLCHAIN_FILE=cmake/ios-linux-toolchain.cmake -DIOS_SDK=/opt/iOS-SDKs/iPhoneOS26.5.sdk \
-    -DCMAKE_DISABLE_FIND_PACKAGE_absl=TRUE -DAURORA_DAWN_PROVIDER=package
-cmake --build build-ios --target WiiCompiled
-```
-
-#### iOS from Windows
-
-The same toolchain file works on Windows with [llvm-mingw](https://github.com/mstorsjo/llvm-mingw),
-which is what the Windows build already uses. Three things differ from a Linux host:
-
- llvm-mingw does not ship `ld64.lld.exe` or `llvm-install-name-tool.exe`; LLVM tools dispatch on
-  their file name, so copy `ld.lld.exe` to `ld64.lld.exe` and `llvm-objcopy.exe` to
-  `llvm-install-name-tool.exe` in its `bin` directory.
- Git for Windows checks the SDK repo's symlinks out as small text files unless `core.symlinks`
-  is on (Developer Mode). Either enable that before cloning, or replace each placeholder with a
-  copy of its target; `libSystem.tbd` is one of them and the link fails without it.
- Pass the toolchain file as an absolute path, and `MKW_IOS_LLVM_BIN` as the llvm-mingw `bin`
-  directory with forward slashes.
-
-```powershell
-cmake -S runtime -B build-ios -G Ninja -DCMAKE_BUILD_TYPE=Release `
-    -DCMAKE_TOOLCHAIN_FILE=C:/src/Wiicompiled/runtime/cmake/ios-linux-toolchain.cmake `
-    -DIOS_SDK=C:/iOS-SDKs/iPhoneOS26.5.sdk -DMKW_IOS_LLVM_BIN=C:/llvm-mingw/bin `
-    -DCMAKE_DISABLE_FIND_PACKAGE_absl=TRUE -DAURORA_DAWN_PROVIDER=package
-cmake --build build-ios --target WiiCompiled
-```
-
-Install it with AltStore or SideStore, which sign with your own Apple ID. The app needs the
-`com.apple.developer.kernel.increased-memory-limit` entitlement or it exits at startup;
-[GetMoreRam](https://github.com/hugeBlack/GetMoreRam) grants it with a free Apple ID.
-GetMoreRam grants `com.apple.developer.kernel.extended-virtual-addressing` at the same time, which
-matters on devices with less than 4 GB of RAM: the runtime reserves a 4 GiB guest address space at
-startup and that is right on the limit iOS allows without it. It makes no difference on the 8 GB
-devices this was developed on.
-
-`runtime/cmake/ios/WiiCompiled.entitlements` is a template for signing by hand, with placeholders to
-replace; the sideloaders build their own and ignore it. It carries both entitlements, so the
-provisioning profile you sign with needs both capabilities enabled.
-
-Copy `Config.toml` and an extracted `DATA` directory into the app's Documents folder, using Files on
-the device or the Finder with it connected. Keep `dvd_root` relative:
-
-```toml
-[paths]
-dvd_root = "DATA"
-```
-
-Touch controls appear when no gamepad is attached; tap the FPS readout for settings.
-
-
 > [!CAUTION]
 > Only take builds from this repository's
 > [Releases](https://github.com/patchzyy/Wiicompiled/releases) page. If someone's sharing an
@@ -239,7 +157,9 @@ The default test suite needs no binaries and no host C++ compiler, so you can ha
 translator without any game data around.

 For everything beyond that, feeding in your own `main.dol`/`StaticR.rel`, running the
-translation, generating the manifest and build graph, and compiling. see [`translator/README.md`](translator/README.md).
+translation, generating the manifest and build graph, and compiling, see [`translator/README.md`](translator/README.md).
+
+For a step-by-step guide on compiling both WiiCompiled and Retro Rewind from source on macOS (Apple Silicon), see the [macOS Build Guide](docs/building-macos.md).

 ## FAQ

@@ -288,7 +208,7 @@ AI coding tools were used during development of this project.
 All translated output is verified against real hardware behavior and most importantly, physics accuracy is proven synced across Wii, Dolphin, and WiiCompiled (see FAQ). 

 ## Credits
-
+- **inkwreck** - making the logo
 - **[aurora](https://github.com/encounter/aurora)** - the GX rendering/windowing backend this
  project's whole graphics layer sits on. MIT licensed.
 - **[Dawn](https://dawn.googlesource.com/dawn)** - Google's WebGPU implementation, powering
@@ -114,14 +114,16 @@ Source: <https://github.com/higan-emu/libco>. Full license text:

 ## Fetched at build time and redistributed in release builds

-These are pinned in `aurora-main/extern/CMakeLists.txt`, `aurora-main/CMakeLists.txt` and
-`aurora-main/cmake/AuroraDawnProvider.cmake`. They are not stored in this repository; the build
-downloads them, and release installers carry the resulting binaries. Their license texts are
+These are pinned in `aurora-main/extern/CMakeLists.txt`, `aurora-main/CMakeLists.txt`,
+`aurora-main/cmake/AuroraDawnProvider.cmake`, and (for Mbed TLS) `runtime/CMakeLists.txt`. They are
+not stored in this repository; the build downloads them - each fetch is pinned to an exact version
+with a checked SHA-256 - and links or redistributes the resulting binaries. Their license texts are
 included in the installer's `licenses/` folder. The Windows installer bundles the pinned source
 trees themselves (fetched by `Launcher/Prepare-Dependencies.ps1`) so end-user builds run offline.

 | Component | Version | License | Upstream |
 | --- | --- | --- | --- |
+| Mbed TLS | 3.6.7 | Apache-2.0 / GPL-2.0-or-later | <https://github.com/Mbed-TLS/mbedtls> |
 | Dawn (WebGPU) | `v20260603.191052` prebuilt | BSD-3-Clause | <https://dawn.googlesource.com/dawn> |
 | Tint (part of Dawn) | with Dawn | BSD-3-Clause | <https://dawn.googlesource.com/dawn> |
 | DirectXShaderCompiler (`dxcompiler.dll`) | with Dawn | NCSA / University of Illinois Open Source | <https://github.com/microsoft/DirectXShaderCompiler> |
@@ -151,15 +151,36 @@ elseif (_aurora_dawn_provider STREQUAL "package")
      endif ()
    endif ()
    set(AURORA_DAWN_PACKAGE_URL
-      "https://github.com/encounter/dawn-build/releases/download/${AURORA_DAWN_VERSION}/dawn-${_dawn_system}-${_dawn_arch}.tar.gz")
+      "https://github.com/theofficialgman/dawn-build/releases/download/${AURORA_DAWN_VERSION}/dawn-${_dawn_system}-${_dawn_arch}.tar.gz")

    # A release asset is mutable: the same tag has already served two different windows-amd64 archives,
    # and a cached extraction is never re-verified. Pin the digest for the combinations we ship.
-    if (NOT AURORA_DAWN_PACKAGE_URL_HASH
-        AND AURORA_DAWN_VERSION STREQUAL "v20260603.191052"
-        AND _dawn_system STREQUAL "windows" AND _dawn_arch STREQUAL "amd64")
-      set(AURORA_DAWN_PACKAGE_URL_HASH
-        "SHA256=7785373d569b3b0237918ec9c523239f7d0667857c5ea8242e3cdfde95e6aeab")
+    if (NOT AURORA_DAWN_PACKAGE_URL_HASH AND AURORA_DAWN_VERSION STREQUAL "v20260603.191052")
+      if (_dawn_system STREQUAL "windows" AND _dawn_arch STREQUAL "amd64")
+        set(AURORA_DAWN_PACKAGE_URL_HASH
+          "SHA256=13be9cff8b9b179c42dcd16aeabb6effcc8f0dfdcc14463eda2a5caeda225142")
+      elseif (_dawn_system STREQUAL "windows" AND _dawn_arch STREQUAL "arm64")
+        set(AURORA_DAWN_PACKAGE_URL_HASH
+          "SHA256=bf2d921110f14a1d6553f673c5597988e66c02af5587e4a1fee167937d247734")
+      elseif (_dawn_system STREQUAL "linux" AND _dawn_arch STREQUAL "x86_64")
+        set(AURORA_DAWN_PACKAGE_URL_HASH
+          "SHA256=7adcf241bb2a24ec0c576609f2d67203e0e65db9c5a286ca2bbb6281fa644b35")
+      elseif (_dawn_system STREQUAL "linux" AND _dawn_arch STREQUAL "aarch64")
+        set(AURORA_DAWN_PACKAGE_URL_HASH
+          "SHA256=2415e253d46f91b2d72fc73bf6055fb31b98b67c773dc546e1991b1cf019732f")
+      elseif (_dawn_system STREQUAL "darwin" AND _dawn_arch STREQUAL "arm64")
+        set(AURORA_DAWN_PACKAGE_URL_HASH
+          "SHA256=0a8ea8eb0159fc0ba1083c52155d9376fb173cffe690b400464a6ad8881bb461")
+      elseif (_dawn_system STREQUAL "darwin" AND _dawn_arch STREQUAL "x86_64")
+        set(AURORA_DAWN_PACKAGE_URL_HASH
+          "SHA256=5fe2c7a2a8b4cb82acee4af16779a83ae333c7657b9dc1a5008f5fd1f5ad5f80")
+      elseif (_dawn_system STREQUAL "ios" AND _dawn_arch STREQUAL "arm64")
+        set(AURORA_DAWN_PACKAGE_URL_HASH
+          "SHA256=f97701d26fd1f25bbcc260b4c31736ede134c730c12556029e2470fde967f424")
+      elseif (_dawn_system STREQUAL "android" AND _dawn_arch STREQUAL "aarch64")
+        set(AURORA_DAWN_PACKAGE_URL_HASH
+          "SHA256=0e63e8cbf53551f703f582d1306f4257c0380353f66b53369d96952ce6d9f934")
+      endif ()
    endif ()
  endif ()
  message(STATUS "aurora: Fetching prebuilt Dawn package from ${AURORA_DAWN_PACKAGE_URL}")
@@ -127,12 +127,20 @@ typedef struct {
  const char* pipelineCachePath;
 } AuroraConfig;

+typedef enum {
+  AURORA_INITIALIZATION_SUCCESS = 0,
+  AURORA_INITIALIZATION_GRAPHICS_UNAVAILABLE = 1,
+} AuroraInitializationStatus;
+
 typedef struct {
  AuroraBackend backend;
  const char* userPath;
  const char* cachePath;
  SDL_Window* window;
  AuroraWindowSize windowSize;
+  AuroraInitializationStatus initializationStatus;
+  // On failure, owned by SDL on the calling thread. Copy before another SDL call.
+  const char* initializationError;
 } AuroraInfo;

 AuroraInfo aurora_initialize(int argc, char* argv[], const AuroraConfig* config);
@@ -171,6 +171,22 @@ typedef struct PADButtonMapping {
  PADButton padButton;
 } PADButtonMapping;

+// Explicitly disabled, unlike INVALID which permits default L/R trigger input.
+#define PAD_NATIVE_BUTTON_DISABLED 0xfffffffeu
+
+// Axis-to-button bindings share the persisted nativeButton field without
+// changing the binary layout of existing controller mapping files.
+constexpr u32 PADEncodeAxisButton(u32 axis, bool negative, u32 threshold = 50) {
+  return 0x10000u | axis | (negative ? 0x80u : 0u) | (threshold << 8);
+}
+constexpr bool PADIsAxisButton(u32 binding) { return (binding & 0xffff0000u) == 0x10000u; }
+constexpr u32 PADAxisButtonThreshold(u32 binding) { return (binding >> 8) & 0xffu; }
+constexpr u32 PADAxisButtonAxis(u32 binding) { return binding & 0x7fu; }
+constexpr bool PADAxisButtonNegative(u32 binding) { return (binding & 0x80u) != 0; }
+constexpr u32 PADAxisButtonIdentity(u32 binding) {
+  return PADIsAxisButton(binding) ? (binding & ~0xff00u) : binding;
+}
+
 typedef struct PADAxisMapping {
  PADSignedNativeAxis nativeAxis;
  s32 nativeButton;
@@ -229,7 +245,6 @@ s32 PADGetNativeButtonPressed(u32 port);
 PADSignedNativeAxis PADGetNativeAxisPulled(u32 port);
 void PADRestoreDefaultMapping(u32 port);
 void PADBlockInput(bool block);
-bool PADIsInputBlocked(void);

 /**
 * Set the default controller mapping used.
@@ -225,7 +225,7 @@ enum class ImGuiFramePolicy {
 bool begin_frame_impl(bool pumpEvents, ImGuiFramePolicy imguiPolicy = ImGuiFramePolicy::Immediate,
                      bool* imguiNewFrameOwed = nullptr) noexcept;
 bool begin_frame_render_state_impl(ImGuiFramePolicy imguiPolicy, bool* imguiNewFrameOwed) noexcept;
-void end_frame_impl(bool pumpEvents, bool drainFifo) noexcept;
+void end_frame_impl(bool pumpEvents, bool drainFifo);

 // The two publication points of a frame-worker cycle, cleared together under `mutex`. Sealed:
 // producer-shared renderer state is free again. Done: slots encoded, presented, ImGui restarted.
@@ -689,15 +689,23 @@ AuroraInfo initialize(int argc, char* argv[], const AuroraConfig& config) noexce
  const AuroraBackend requestedBackend = config.desiredBackend;
  AuroraBackend selectedBackend = requestedBackend;
  bool windowCreated = false;
+  std::string firstGraphicsError;
+  const auto rememberGraphicsError = [&] {
+    if (firstGraphicsError.empty() && SDL_GetError()[0] != '\0') {
+      firstGraphicsError = SDL_GetError();
+    }
+  };
  if (selectedBackend != BACKEND_AUTO) {
    Log.info("Requested graphics backend: {}", backend_name(selectedBackend));
    if (window::create_window(selectedBackend)) {
      if (webgpu::initialize(selectedBackend)) {
        windowCreated = true;
      } else {
+        rememberGraphicsError();
        window::destroy_window();
      }
    } else {
+      rememberGraphicsError();
      Log.error("Failed to create a window for backend {}: {}", backend_name(selectedBackend),
                SDL_GetError());
    }
@@ -714,18 +722,28 @@ AuroraInfo initialize(int argc, char* argv[], const AuroraConfig& config) noexce
    for (const auto backendType : PreferredBackendOrder) {
      selectedBackend = backendType;
      if (!window::create_window(selectedBackend)) {
+        rememberGraphicsError();
        continue;
      }
      if (webgpu::initialize(selectedBackend)) {
        windowCreated = true;
        break;
      } else {
+        rememberGraphicsError();
        window::destroy_window();
      }
    }
  }

-  ASSERT(windowCreated, "Error creating window: {}", SDL_GetError());
+  if (!windowCreated) {
+    if (firstGraphicsError.empty()) firstGraphicsError = "No supported graphics backend is available";
+    SDL_SetError("%s", firstGraphicsError.c_str());
+    Log.error("Graphics initialization failed: {}", firstGraphicsError);
+    return {
+        .initializationStatus = AURORA_INITIALIZATION_GRAPHICS_UNAVAILABLE,
+        .initializationError = SDL_GetError(),
+    };
+  }
  if (requestedBackend != BACKEND_AUTO && selectedBackend != requestedBackend) {
    Log.error("Graphics backend fallback in effect: video.graphics_api requested {}, "
              "running on {}",
@@ -1661,7 +1679,7 @@ bool run_frame_worker_cycle(gfx::SealedFrame& sealedFrame) noexcept {

 // Synchronous frame submission: seal, encode and present inline on the calling thread. Used when
 // the frame worker is disabled (RenderDoc captures) and on the boot path.
-void end_frame_impl(bool pumpEvents, bool drainFifo) noexcept {
+void end_frame_impl(bool pumpEvents, bool drainFifo) {
  ZoneScoped;
 #ifdef AURORA_ENABLE_GX
  webgpu::fail_if_device_lost();
@@ -1671,11 +1689,9 @@ void end_frame_impl(bool pumpEvents, bool drainFifo) noexcept {
  gfx::SealedFrame sealedFrame;
  SealedFrameContext ctx;
  std::vector<PresentationJob> presentationJobs;
+  if (drainFifo) gx::fifo::drain();
  {
    std::lock_guard gpuLock(g_rendererGpuMutex);
-    if (drainFifo) {
-      gx::fifo::drain();
-    }
    seal_frame_locked(sealedFrame, ctx);
    presentationJobs = encode_sealed_frame(sealedFrame, ctx);
  }
@@ -1752,7 +1768,7 @@ bool begin_frame() noexcept {
  return prepared;
 }

-void end_frame() noexcept {
+void end_frame() {
 #ifdef AURORA_ENABLE_GX
  webgpu::fail_if_device_lost();
 #endif
@@ -1768,10 +1784,7 @@ void end_frame() noexcept {

  // Seal all current GX work on the CPU while the renderer is known ready.
  // Later FIFO writes belong exclusively to the next frame.
-  {
-    std::lock_guard gpuLock(g_rendererGpuMutex);
-    gx::fifo::drain();
-  }
+  gx::fifo::drain();
  {
    std::lock_guard lock(g_frameWorker.mutex);
    g_frameWorker.framePrepared = false;
@@ -1797,6 +1810,10 @@ bool wait_for_frame_worker_for(std::chrono::microseconds timeout) noexcept {
  return wait_for_frame_worker_private_for(FrameWorkerPhase::Done, timeout);
 }
 std::recursive_mutex& renderer_gpu_mutex() noexcept { return g_rendererGpuMutex; }
+void submit_staging_commands(const wgpu::CommandBuffer& commands) {
+  std::lock_guard submitLock(g_queueSubmitMutex);
+  webgpu::g_queue.Submit(1, &commands);
+}
 } // namespace aurora

 // C API bindings
@@ -1859,10 +1876,6 @@ bool aurora_flush_efb_copies_to_ram() {
  if (!aurora::gfx::efb_ram::has_pending()) {
    return true;
  }
-  if (!aurora::gfx::efb_ram::prepare_downloads()) {
-    return false;
-  }
-
  // This finalizes the frame still being recorded, on the producer thread, so join the whole cycle
  // first: the encode phase owns the previous passes, EFB targets and image pool.
  aurora::wait_for_frame_worker();
@@ -1870,6 +1883,7 @@ bool aurora_flush_efb_copies_to_ram() {
  // suffix cannot safely be replayed against the same mutable EFB resources.
  aurora::gx::mark_frame_interpolation_replay_unsafe();
  aurora::gx::fifo::drain();
+  if (!aurora::gfx::efb_ram::prepare_downloads()) return false;
  const wgpu::CommandEncoderDescriptor encoderDescriptor{
      .label = "GX CPU-visible EFB copy encoder",
  };
@@ -1895,8 +1909,7 @@ bool aurora_flush_efb_copies_to_ram() {
 }
 bool aurora_flush_efb_copy_to_ram(void* dest) {
 #ifdef AURORA_ENABLE_GX
-  if (dest == nullptr || !aurora::gfx::efb_ram::has_pending(dest) ||
-      !aurora::gfx::efb_ram::prepare_downloads(dest)) {
+  if (dest == nullptr || !aurora::gfx::efb_ram::has_pending(dest)) {
    return false;
  }

@@ -1907,6 +1920,7 @@ bool aurora_flush_efb_copy_to_ram(void* dest) {
  // image instead of replaying this split frame.
  aurora::gx::mark_frame_interpolation_replay_unsafe();
  aurora::gx::fifo::drain();
+  if (!aurora::gfx::efb_ram::prepare_downloads(dest)) return false;
  const wgpu::CommandEncoderDescriptor encoderDescriptor{
      .label = "GX demanded EFB copy encoder",
  };
@@ -2,12 +2,43 @@

 #import <Foundation/Foundation.h>
 #include <SDL3/SDL_metal.h>
+#include <SDL3/SDL_properties.h>
+#include <SDL3/SDL_video.h>

 namespace aurora::webgpu::utils {
+namespace {
+constexpr const char* MetalViewProperty = "aurora.window.metal_view";
+
+void SDLCALL DestroyMetalView(void*, void* value) {
+  SDL_Metal_DestroyView(value);
+}
+} // namespace
+
 std::shared_ptr<wgpu::ChainedStruct> SetupWindowAndGetSurfaceDescriptorCocoa(SDL_Window* window) {
-  SDL_MetalView view = SDL_Metal_CreateView(window);
-  std::shared_ptr<wgpu::SurfaceSourceMetalLayer> desc = std::make_shared<wgpu::SurfaceSourceMetalLayer>();
+  const auto properties = SDL_GetWindowProperties(window);
+  if (!properties) {
+    return nullptr;
+  }
+  auto view = SDL_GetPointerProperty(properties, MetalViewProperty, nullptr);
+  if (!view) {
+    view = SDL_Metal_CreateView(window);
+    if (!view) {
+      return nullptr;
+    }
+    // Own one view per window, not per WebGPU surface. Surface recovery must
+    // preserve the UIKit root and its controls (and the Cocoa Metal subview).
+    // SDL cleans window properties before destroying its native window.
+    // The cleanup callback also runs if setting the property fails.
+    if (!SDL_SetPointerPropertyWithCleanup(properties, MetalViewProperty, view, DestroyMetalView, nullptr)) {
+      return nullptr;
+    }
+  }
+  auto desc = std::make_shared<wgpu::SurfaceSourceMetalLayer>();
  desc->layer = SDL_Metal_GetLayer(view);
-  return std::move(desc);
+  if (!desc->layer) {
+    SDL_ClearProperty(properties, MetalViewProperty);
+    return nullptr;
+  }
+  return desc;
 }
 } // namespace aurora::webgpu::utils
@@ -520,9 +520,11 @@ void GXCopyTex(void* dest, GXBool clear) {
    clearState.clearAlpha = clear && alphaUpdate;
  }
  const auto copyFilter = combined_copy_filter_coefficients(g_gxState.copyFilterVFilter);
-  // Skip only recurring color copies so one-shot copies are never lost.
-  const bool producedConsecutively = handle.revision != 0 && currentFrame - handle.lastProducedFrame <= 1;
-  const bool persistentCopy = !aurora::gx::is_depth_format(texCopyFmt) && !producedConsecutively;
+  // Every GXCopyTex is observable texture data. Reusing a destination in this
+  // or the previous frame does not guarantee another redraw: menu thumbnail
+  // scratch targets can be reused and then retained. Depth copies have the
+  // same requirement. Only display presentation may skip unfinished draws.
+  const bool persistentCopy = true;
  aurora::gfx::resolve_pass(handle.handle, rect, clearState.clearColor, clearState.clearAlpha, clearState.clearDepth,
                            clearState.clearColorValue, aurora::gx::clear_depth_value(), resolveFmt,
                            &sourceRect.sampleRect, g_gxState.texCopyHalfScale, &copyFilter, forceOpaqueAlpha,
@@ -319,6 +319,18 @@ std::array<bool, PAD_CHANMAX> g_suppressLeftTrigger{};
 std::array<bool, PAD_CHANMAX> g_suppressRightTrigger{};

 bool is_mouse_scancode(const s32 scancode) { return scancode < PAD_KEY_INVALID; }
+bool is_native_binding_pressed(SDL_Gamepad* gamepad, u32 binding) {
+  if (PADIsAxisButton(binding)) {
+    const u32 axis = PADAxisButtonAxis(binding);
+    const u32 threshold = PADAxisButtonThreshold(binding);
+    if (axis >= SDL_GAMEPAD_AXIS_COUNT || threshold < 1 || threshold > 100) return false;
+    int value = SDL_GetGamepadAxis(gamepad, static_cast<SDL_GamepadAxis>(axis));
+    if (PADAxisButtonNegative(binding)) value = -value;
+    return value > 0 && value * 100 >= static_cast<int>(threshold) * 32767;
+  }
+  return binding < SDL_GAMEPAD_BUTTON_COUNT &&
+         SDL_GetGamepadButton(gamepad, static_cast<SDL_GamepadButton>(binding));
+}
 bool is_mouse_button_pressed(const s32 scancode) {
  const int32_t buttonNum = -(scancode + 1);
  if (buttonNum < 1 || buttonNum > 5) {
@@ -724,10 +736,10 @@ u32 PADRead(PADStatus* status) {
    }

    status[i].err = PAD_ERR_NONE;
-    if (g_keyboardBindings[i].m_mappingsSet) {
+    if (g_keyboardBindings[i].m_mappingsSet && SDL_GetKeyboardFocus() != nullptr) {
      std::ranges::for_each(
-          g_keyboardBindings[i].m_buttonMapping, [&kbState, &i, &status](const PADKeyButtonBinding& mapping) {
-            if (mapping.scancode > PAD_KEY_INVALID && kbState[mapping.scancode]) {
+          g_keyboardBindings[i].m_buttonMapping, [&kbState, &numKeys, &i, &status](const PADKeyButtonBinding& mapping) {
+            if (mapping.scancode > PAD_KEY_INVALID && mapping.scancode < numKeys && kbState[mapping.scancode]) {
              status[i].button |= mapping.padButton;
            } else if (is_mouse_scancode(mapping.scancode) && is_mouse_button_pressed(mapping.scancode)) {
              status[i].button |= mapping.padButton;
@@ -788,7 +800,7 @@ u32 PADRead(PADStatus* status) {
      status[i].triggerRight = static_cast<u8>(std::min(static_cast<int>(status[i].triggerRight) + tr, 255));
    }

-    if (controller) {
+    if (controller && !g_keyboardBindings[i].m_mappingsSet) {
      EnsureMappingLoaded(controller);

      // Wii U Pro Controller raw D-pad fallback. SDL's HIDAPI Wii driver posts
@@ -835,7 +847,7 @@ u32 PADRead(PADStatus* status) {
      bool rightTriggerSet = false;
      std::ranges::for_each(controller->m_buttonMapping, [&controller, &i, &status, &leftTriggerSet,
                                                          &rightTriggerSet](const auto& mapping) {
-        if (SDL_GetGamepadButton(controller->m_controller, static_cast<SDL_GamepadButton>(mapping.nativeButton))) {
+        if (is_native_binding_pressed(controller->m_controller, mapping.nativeButton)) {
          status[i].button |= mapping.padButton;
        }

@@ -852,7 +864,7 @@ u32 PADRead(PADStatus* status) {
        if (mapping.nativeButton == PAD_NATIVE_BUTTON_INVALID) {
          return;
        }
-        if (SDL_GetGamepadButton(controller->m_controller, static_cast<SDL_GamepadButton>(mapping.nativeButton))) {
+        if (is_native_binding_pressed(controller->m_controller, mapping.nativeButton)) {
          status[i].button |= mapping.padButton;
        }

@@ -946,6 +958,17 @@ u32 PADRead(PADStatus* status) {
      Sint16 tl = std::max(static_cast<Sint16>(0), _get_axis_value(controller, PAD_AXIS_TRIGGER_L));
      Sint16 tr = std::max(static_cast<Sint16>(0), _get_axis_value(controller, PAD_AXIS_TRIGGER_R));

+      // Games can read either the digital L/R bits or their analog pressure.
+      // An explicit button binding must drive both, otherwise the original
+      // L2/R2 axis still activates L/R even when it was rebound to L1/R1.
+      // Real GC pads retain independent analog travel and end-stop clicks.
+      if (!(controller->m_isGameCube ||
+            (SDL_GetGamepadType(controller->m_controller) == SDL_GAMEPAD_TYPE_NINTENDO_SWITCH_PRO &&
+             controller->m_pid == 0x2073))) {
+        if (leftTriggerSet) tl = (status[i].button & PAD_TRIGGER_L) != 0 ? 32767 : 0;
+        if (rightTriggerSet) tr = (status[i].button & PAD_TRIGGER_R) != 0 ? 32767 : 0;
+      }
+
      if (controller->m_deadZones.emulateTriggers) {
        if (!leftTriggerSet && tl > controller->m_deadZones.leftTriggerActivationZone) {
          status[i].button |= PAD_TRIGGER_L;
@@ -990,12 +1013,13 @@ void PADControlMotor(const u32 chan, const u32 cmd) {
  }

  if (controller->m_isGameCube) {
-    if (cmd == PAD_MOTOR_STOP) {
-      aurora::input::controller_rumble(instance, 0, 1, 0);
+    if (cmd == PAD_MOTOR_STOP || cmd == PAD_MOTOR_STOP_HARD) {
+      // Use an unambiguous motor-off request. The (0, 1) coast encoding
+      // requires SDL's GameCube brake mode; other backends or an overridden
+      // hint interpret it as rumble and can leave the controller vibrating.
+      aurora::input::controller_rumble(instance, 0, 0, 0);
    } else if (cmd == PAD_MOTOR_RUMBLE) {
      aurora::input::controller_rumble(instance, 1, 1, 0);
-    } else if (cmd == PAD_MOTOR_STOP_HARD) {
-      aurora::input::controller_rumble(instance, 0, 0, 0);
    }
  } else {
    if (cmd == PAD_MOTOR_STOP) {
@@ -1278,6 +1302,11 @@ BOOL PADSetKeyButtonBindings(const u32 port, PADKeyButtonBinding bindings[PAD_BU
 }

 PADKeyButtonBinding* PADGetKeyButtonBindings(const u32 port, u32* buttonCount) {
+  PADInit();
+  if (!g_keyboardBindingsLoaded) {
+    g_keyboardBindingsLoaded = true;
+    load_keyboard_bindings();
+  }
  if (port >= PAD_MAX_CONTROLLERS || !g_keyboardBindings[port].m_mappingsSet) {
    return nullptr;
  }
@@ -1627,8 +1656,6 @@ void PADBlockInput(const bool block) {
  }
 }

-bool PADIsInputBlocked() { return g_blockPAD.load(std::memory_order_acquire); }
-

 SDL_Gamepad* PADGetSDLGamepadForIndex(const u32 index) {
  const auto* ctrl = __PADGetControllerForIndex(index);
@@ -7,11 +7,13 @@
 #include <algorithm>
 #include <atomic>
 #include <optional>
+#include <mutex>

 namespace aurora::vi {
 std::optional<GXRenderModeObj> g_renderMode;
 namespace {
 std::atomic<float> g_presentAspectCorrection{1.f};
+std::mutex g_renderModeMutex;

 float calculate_present_aspect_correction(const GXRenderModeObj& rm) noexcept {
  if (rm.viWidth == 0 || rm.viHeight == 0) {
@@ -29,9 +31,8 @@ float calculate_present_aspect_correction(const GXRenderModeObj& rm) noexcept {
  const float verticalFill = static_cast<float>(rm.viHeight) / nominalActiveHeight;
  return horizontalFill / verticalFill;
 }
-} // namespace

-Vec2<uint32_t> render_mode_size() noexcept {
+Vec2<uint32_t> render_mode_size_locked() noexcept {
  if (!g_renderMode) {
    return {640, 528};
  }
@@ -40,18 +41,31 @@ Vec2<uint32_t> render_mode_size() noexcept {
  return {std::max<uint32_t>(g_renderMode->fbWidth, 640), std::max<uint32_t>(g_renderMode->efbHeight, 528)};
 }

+} // namespace
+
+Vec2<uint32_t> render_mode_size() noexcept {
+  std::lock_guard lock(g_renderModeMutex);
+  return render_mode_size_locked();
+}
+
 void configure(const GXRenderModeObj* rm) noexcept {
-  const auto oldSize = render_mode_size();
-  if (rm == nullptr) {
-    g_renderMode.reset();
-  } else {
-    g_renderMode = *rm;
-    g_presentAspectCorrection.store(calculate_present_aspect_correction(*rm), std::memory_order_release);
+  bool sizeChanged = false;
+  {
+    std::lock_guard lock(g_renderModeMutex);
+    const auto oldSize = render_mode_size_locked();
+    if (rm == nullptr) {
+      g_renderMode.reset();
+    } else {
+      g_renderMode = *rm;
+      g_presentAspectCorrection.store(calculate_present_aspect_correction(*rm), std::memory_order_release);
+    }
+    if (rm == nullptr) {
+      g_presentAspectCorrection.store(1.f, std::memory_order_release);
+    }
+    sizeChanged = render_mode_size_locked() != oldSize;
  }
-  if (rm == nullptr) {
-    g_presentAspectCorrection.store(1.f, std::memory_order_release);
-  }
-  if (render_mode_size() != oldSize) {
+  // Never hold the mode lock across a resize request or a renderer callback.
+  if (sizeChanged) {
    window::request_frame_buffer_resize();
  }
 }
@@ -61,6 +75,7 @@ Vec2<uint32_t> configured_fb_size() noexcept {
 }

 Vec2<uint32_t> visible_fb_size() noexcept {
+  std::lock_guard lock(g_renderModeMutex);
  if (!g_renderMode) {
    return {640, 528};
  }
@@ -1,4 +1,5 @@
 #include "common.hpp"
+#include "staging_map.hpp"
 #include "../gx/shader_info.hpp"

 #include "clear.hpp"
@@ -36,10 +37,13 @@ using webgpu::g_device;
 using webgpu::g_instance;
 using webgpu::g_queue;

+struct DebugFrameData {
 #ifdef AURORA_GFX_DEBUG_GROUPS
-std::vector<std::string> g_debugGroupStack;
-std::vector<std::string> g_debugMarkers;
+  std::vector<std::string> groups;
+  std::vector<std::string> markers;
 #endif
+};
+DebugFrameData g_debugFrame;

 constexpr uint64_t StagingBufferSize = UniformBufferSize + VertexBufferSize + IndexBufferSize + StorageBufferSize +
                                       (UseTextureBuffer ? TextureUploadSize : 0);
@@ -128,12 +132,7 @@ wgpu::Buffer g_storageBuffer;
 constexpr size_t FrameSlotCount = 3;
 static std::array<wgpu::Buffer, FrameSlotCount> g_stagingBuffers;
 static size_t currentStagingBuffer = 0;
-enum class BufferMapState {
-  Unmapped,
-  Mapping,
-  Mapped,
-};
-static std::atomic s_mappingState{BufferMapState::Unmapped};
+static StagingMapState s_mappingState;
 static wgpu::Limits g_cachedLimits;
 // Advanced once per logical frame in the seal prologue, under the renderer GPU mutex and with the
 // producer blocked, so every later reader sees a value that no longer moves.
@@ -168,7 +167,12 @@ struct RenderPass {
  Range resolveUniformRange;
  std::array<u32, 3> resolveCopyFilterCoefficients{0, 64, 0};
  Vec4<float> clearColorValue{0.f, 0.f, 0.f, 0.f};
-  float clearDepthValue = 1.f;
+  // 1.f is the forward-Z "farthest" clear value; under UseReversedZ farthest is 0.f instead (see
+  // gx::clear_depth_value(), which the main render pass explicitly overrides this default with -
+  // any OTHER pass that keeps this default, e.g. an offscreen render-to-texture pass composited
+  // later, needs the same reversed-Z-aware value or its depth buffer starts "already nearest",
+  // failing every subsequent depth test and making whatever's drawn into it vanish).
+  float clearDepthValue = gx::UseReversedZ ? 0.f : 1.f;
  CommandList commands;
  bool clearColor = true;
  bool clearDepth = true;
@@ -229,6 +233,8 @@ static void recycle_render_passes(std::vector<RenderPass>& passes) noexcept {
 }

 struct SealedFrameData {
+  depth_peek::FrameMapping depthMapping;
+  DebugFrameData debug;
  std::vector<RenderPass> passes;
 };

@@ -250,6 +256,51 @@ static std::atomic_bool g_inOffscreen{false};
 static std::optional<RenderPass> g_suspendedEfbPass;
 static Viewport g_suspendedEfbViewport;
 static ClipRect g_suspendedEfbScissor;
+// Prefix referenced by a suspended EFB pass. Preserve its offsets across an
+// offscreen split, without rendering it before the bake it may sample finishes.
+static StagingSizes g_suspendedEfbBytes{};
+static constexpr StagingSizes PhysicalStagingCapacity{
+    VertexBufferSize, UniformBufferSize, IndexBufferSize, StorageBufferSize};
+static StagingSizes g_stagingCapacity = PhysicalStagingCapacity;
+static uint64_t g_stagingEpoch = 0;
+static uint64_t g_stagingSplitCount = 0;
+static StagingSizes g_stagingHighWater{};
+
+StagingSizes staging_usage() noexcept {
+  return {g_verts.size(), g_uniforms.size(), g_indices.size(), g_storage.size()};
+}
+StagingSizes staging_high_water() noexcept { return g_stagingHighWater; }
+uint64_t staging_epoch() noexcept { return g_stagingEpoch; }
+uint64_t staging_split_count() noexcept { return g_stagingSplitCount; }
+uint64_t staging_uniform_bytes(uint64_t bytes) {
+  return staging_padded(bytes, g_cachedLimits.minUniformBufferOffsetAlignment);
+}
+uint64_t staging_storage_bytes(uint64_t bytes) {
+  return staging_padded(bytes, g_cachedLimits.minStorageBufferOffsetAlignment);
+}
+void set_staging_capacity_limits_for_testing(const StagingSizes& limits) {
+  for (unsigned i = 0; i < limits.size(); ++i) {
+    if (limits[i] > PhysicalStagingCapacity[i])
+      throw StagingCapacityError("Test staging capacity exceeds physical buffer");
+  }
+  g_stagingCapacity = limits;
+  g_stagingHighWater = {};
+}
+bool staging_has_space(const StagingSizes& demand) {
+  // Async readback preparation runs in the worker's noexcept seal prologue.
+  // Reserve all 32 slots plus the uniform binding's 3840-byte trailing window.
+  const StagingSizes tail{0, gx::MaxUniformSize + efb_ram::MaxAsyncReadbackSlots * staging_uniform_bytes(48), 0, 0};
+  const StagingSizes retained = g_suspendedEfbPass ? g_suspendedEfbBytes : StagingSizes{};
+  if (!staging_fits(retained, demand, tail, g_stagingCapacity))
+    throw StagingCapacityError("GPU operation exceeds staging capacity including retained EFB data");
+  return staging_fits(staging_usage(), demand, tail, g_stagingCapacity);
+}
+void ensure_staging_space(const StagingSizes& demand) {
+  if (staging_has_space(demand)) return;
+  split_staging_batch();
+  if (!staging_has_space(demand))
+    throw StagingCapacityError("GPU operation still exceeds staging capacity after submission");
+}

 static void discard_suspended_efb_pass() noexcept {
  if (g_suspendedEfbPass) {
@@ -274,7 +325,8 @@ static size_t g_recordingSnapshotSlot = 0;
 static TextureHandle new_resolve_source_snapshot(wgpu::Extent3D size, wgpu::TextureFormat format) noexcept {
  const wgpu::TextureDescriptor textureDescriptor{
      .label = "GX Copy Source Snapshot",
-      .usage = wgpu::TextureUsage::TextureBinding | wgpu::TextureUsage::CopyDst,
+      .usage = wgpu::TextureUsage::TextureBinding | wgpu::TextureUsage::CopySrc |
+               wgpu::TextureUsage::CopyDst,
      .dimension = wgpu::TextureDimension::e2D,
      .size = size,
      .format = format,
@@ -420,7 +472,7 @@ static inline void push_command(CommandType type, const Command::Data& data) {
  g_renderPasses[g_currentRenderPass].commands.push_back({
      .type = type,
 #ifdef AURORA_GFX_DEBUG_GROUPS
-      .debugGroupStack = g_debugGroupStack,
+      .debugGroupStack = g_debugFrame.groups,
 #endif
      .data = data,
  });
@@ -480,6 +532,7 @@ void set_scissor(const ClipRect& cmd) noexcept {
 template <>
 void push_draw_command(clear::DrawData data) {
  if (data.uniformRange.size == 0) {
+    ensure_staging_space({0, staging_uniform_bytes(16), 0, 0});
    const std::array clearUniform{
        std::clamp(data.depth, 0.f, 1.f),
        0.f,
@@ -506,6 +559,7 @@ void resolve_pass(TextureHandle texture, ClipRect rect, bool clearColor, bool cl
    Log.warn("Dropping resolve pass without an active render pass");
    return;
  }
+  ensure_staging_space({0, 2 * staging_uniform_bytes(48), 0, 0});
  auto& prevPass = g_renderPasses[g_currentRenderPass];
  const auto targetWidth = static_cast<int32_t>(prevPass.targetSize.width);
  const auto targetHeight = static_cast<int32_t>(prevPass.targetSize.height);
@@ -538,7 +592,7 @@ void resolve_pass(TextureHandle texture, ClipRect rect, bool clearColor, bool cl
    sourceRect = {srcLeft, srcTop, std::max(srcRight - srcLeft, 1.0f), std::max(srcBottom - srcTop, 1.0f)};
  }
  prevPass.resolveTarget = std::move(texture);
-  prevPass.requireReadyPipelines = persistentCopy;
+  prevPass.requireReadyPipelines |= persistentCopy;
  prevPass.resolveRect = rect;
  prevPass.resolveSourceRect = sourceRect;
  prevPass.resolveFormat = resolveFormat;
@@ -734,6 +788,7 @@ void begin_offscreen(uint32_t width, uint32_t height) {
  if (!g_inOffscreen) {
    auto& currentPass = g_renderPasses[g_currentRenderPass];
    if (!currentPass.resolveTarget) {
+      g_suspendedEfbBytes = staging_usage();
      g_suspendedEfbPass = std::move(currentPass);
      g_renderPasses.pop_back();
      --g_currentRenderPass;
@@ -757,7 +812,9 @@ void begin_offscreen(uint32_t width, uint32_t height) {
      .targetSize = {width, height, 1},
      .msaaSamples = 1,
      .clearColorValue = {0.f, 0.f, 0.f, 0.f},
-      .clearDepthValue = 1.f,
+      // See the RenderPass::clearDepthValue default's comment: this offscreen pass gets its own
+      // depth buffer, and the farthest clear value is 0.f, not 1.f, under UseReversedZ.
+      .clearDepthValue = gx::UseReversedZ ? 0.f : 1.f,
      .clearColor = true,
      .clearDepth = true,
  };
@@ -844,7 +901,7 @@ void initialize() {
                 label.c_str());
  }
  currentStagingBuffer = 0;
-  s_mappingState.store(BufferMapState::Unmapped, std::memory_order_release);
+  s_mappingState.reset();
  map_staging_buffer();

  {
@@ -950,6 +1007,8 @@ void shutdown() {
  g_uniformBuffer = {};
  g_indexBuffer = {};
  g_storageBuffer = {};
+  // Invalidate outstanding callbacks before releasing their buffers.
+  s_mappingState.reset();
  g_stagingBuffers.fill({});
  for (auto& pool : g_resolveSourceSnapshotPools) {
    pool.entry.reset();
@@ -968,37 +1027,36 @@ void shutdown() {
  g_inOffscreen = false;
  g_frameIndex = UINT32_MAX;
  currentStagingBuffer = 0;
-  s_mappingState.store(BufferMapState::Unmapped, std::memory_order_release);
 }

 void map_staging_buffer() {
-  auto expected = BufferMapState::Unmapped;
-  if (!s_mappingState.compare_exchange_strong(expected, BufferMapState::Mapping, std::memory_order_acq_rel,
-                                              std::memory_order_acquire)) {
+  const auto generation = s_mappingState.request();
+  if (generation == 0) {
    return;
  }

  g_stagingBuffers[currentStagingBuffer].MapAsync(
      wgpu::MapMode::Write, 0, StagingBufferSize, wgpu::CallbackMode::AllowSpontaneous,
-      [](wgpu::MapAsyncStatus status, wgpu::StringView message) {
+      [generation](wgpu::MapAsyncStatus status, wgpu::StringView message) {
+        const auto result = status == wgpu::MapAsyncStatus::Success
+            ? BufferMapState::Mapped : BufferMapState::Unmapped;
+        if (!s_mappingState.complete(generation, result)) return;
        if (status == wgpu::MapAsyncStatus::CallbackCancelled || status == wgpu::MapAsyncStatus::Aborted) {
          Log.warn("Buffer mapping {}: {}", magic_enum::enum_name(status), message);
-          s_mappingState.store(BufferMapState::Unmapped, std::memory_order_release);
          return;
        }
        ASSERT(status == wgpu::MapAsyncStatus::Success, "Buffer mapping failed: {} {}", magic_enum::enum_name(status),
               message);
-        s_mappingState.store(BufferMapState::Mapped, std::memory_order_release);
      });
 }

-static bool begin_frame_impl(bool clearEfb) {
+static bool begin_frame_impl(bool clearEfb, bool capacityResume = false) {
  ZoneScoped;
  {
    ZoneScopedN("Wait for buffer map");
    map_staging_buffer();
    while (true) {
-      const auto mappingState = s_mappingState.load(std::memory_order_acquire);
+      const auto mappingState = s_mappingState.state();
      if (mappingState == BufferMapState::Mapped) {
        break;
      }
@@ -1014,8 +1072,11 @@ static bool begin_frame_impl(bool clearEfb) {
        return false;
      }
      g_instance.ProcessEvents();
+      webgpu::fail_if_device_lost();
+      s_mappingState.wait_for_progress();
    }
  }
+  ++g_stagingEpoch;
  g_recordingSnapshotSlot = currentStagingBuffer;
  size_t bufferOffset = 0;
  const auto& stagingBuf = g_stagingBuffers[currentStagingBuffer];
@@ -1040,7 +1101,7 @@ static bool begin_frame_impl(bool clearEfb) {
    gx::begin_frame_interpolation();
  }
  discard_suspended_efb_pass();
-  webgpu::clear_present_source_override();
+  if (!capacityResume) webgpu::clear_present_source_override();

  push_render_pass(RenderPass{});
  set_efb_targets(g_renderPasses[0]);
@@ -1079,12 +1140,12 @@ void abort_frame() noexcept {
    g_textureUploads.clear();
    g_textureUpload.release();
  }
-  if (s_mappingState.load(std::memory_order_acquire) == BufferMapState::Mapped) {
+  if (s_mappingState.state() == BufferMapState::Mapped) {
    // Pending interpolation tasks hold raw pointers into the mapped staging
    // range; they must be dropped before the buffer is unmapped and rotated.
    gx::drop_pending_frame_interpolation_uniforms();
    g_stagingBuffers[currentStagingBuffer].Unmap();
-    s_mappingState.store(BufferMapState::Unmapped, std::memory_order_release);
+    s_mappingState.reset();
    currentStagingBuffer = (currentStagingBuffer + 1) % g_stagingBuffers.size();
    map_staging_buffer();
  }
@@ -1101,7 +1162,7 @@ void abort_frame() noexcept {

 static void end_batch_impl(const wgpu::CommandEncoder& cmd, bool advanceFrame) {
  ZoneScoped;
-  ASSERT(!g_inOffscreen, "end_frame called while offscreen rendering is active");
+  ASSERT(!advanceFrame || !g_inOffscreen, "end_frame called while offscreen rendering is active");
  if (advanceFrame) {
    gx::finalize_frame_interpolation();
  } else {
@@ -1110,6 +1171,8 @@ static void end_batch_impl(const wgpu::CommandEncoder& cmd, bool advanceFrame) {
    gx::drop_pending_frame_interpolation_uniforms();
  }
  g_uniforms.append_zeroes(gx::MaxUniformSize); // Pad the end of the buffer
+  const auto used = staging_usage();
+  for (unsigned i = 0; i < used.size(); ++i) g_stagingHighWater[i] = std::max(g_stagingHighWater[i], used[i]);
  uint64_t bufferOffset = 0;
  const auto writeBuffer = [&](ByteBuffer& buf, wgpu::Buffer& out, uint64_t size, std::string_view label) {
    const auto writeSize = buf.size(); // Only need to copy this many bytes
@@ -1121,7 +1184,7 @@ static void end_batch_impl(const wgpu::CommandEncoder& cmd, bool advanceFrame) {
    return writeSize;
  };
  g_stagingBuffers[currentStagingBuffer].Unmap();
-  s_mappingState.store(BufferMapState::Unmapped, std::memory_order_release);
+  s_mappingState.reset();
  g_stats.drawCallCount = g_drawCallCount;
  g_stats.mergedDrawCallCount = g_mergedDrawCallCount;
  g_stats.lastVertSize = writeBuffer(g_verts, g_vertexBuffer, VertexBufferSize, "Vertex");
@@ -1164,6 +1227,68 @@ void end_frame(const wgpu::CommandEncoder& cmd) { end_batch_impl(cmd, true); }

 void end_batch(const wgpu::CommandEncoder& cmd) { end_batch_impl(cmd, false); }

+void split_staging_batch() {
+  // Never called under the decoder's renderer lock: the worker needs that lock
+  // to reach DONE. FIFO admission yields its unconsumed command first.
+  aurora::wait_for_frame_worker();
+  std::lock_guard gpuLock(aurora::renderer_gpu_mutex());
+  if (!has_current_render_pass())
+    throw StagingCapacityError("Cannot split staging outside an active render pass");
+  gx::mark_frame_interpolation_replay_unsafe();
+  const bool offscreen = g_inOffscreen;
+  const auto viewport = g_cachedViewport;
+  const auto scissor = g_cachedScissor;
+  const auto renderViewport = gx::g_gxState.renderViewport;
+  const auto renderScissor = gx::g_gxState.renderScissor;
+  const auto& active = g_renderPasses[g_currentRenderPass];
+  RenderPass continuation{
+      .colorView = active.colorView, .resolveView = active.resolveView,
+      .depthView = active.depthView, .copySourceTexture = active.copySourceTexture,
+      .copySourceView = active.copySourceView, .copySourceDepthView = active.copySourceDepthView,
+      .targetSize = active.targetSize, .msaaSamples = active.msaaSamples,
+      .clearColor = false, .clearDepth = false,
+      .requireReadyPipelines = active.requireReadyPipelines || offscreen,
+  };
+  auto suspended = std::move(g_suspendedEfbPass);
+  g_suspendedEfbPass.reset();
+  std::array<std::vector<uint8_t>, 4> retained;
+  std::array<ByteBuffer*, 4> buffers{&g_verts, &g_uniforms, &g_indices, &g_storage};
+  if (suspended) {
+    for (unsigned i = 0; i < buffers.size(); ++i) {
+      if (g_suspendedEfbBytes[i])
+        retained[i].assign(buffers[i]->data(), buffers[i]->data() + g_suspendedEfbBytes[i]);
+    }
+  }
+  // GXCopyTex can capture the ordinary EFB as well as an explicit offscreen
+  // target. Its command may arrive in the next batch, after this prefix has
+  // already been submitted. Preserve every prefix; target kind cannot tell
+  // us whether the guest will later retain these pixels in a texture.
+  for (auto& pass : g_renderPasses) pass.requireReadyPipelines = true;
+  auto encoder = g_device.CreateCommandEncoder();
+  end_batch(encoder);
+  render(encoder);
+  aurora::submit_staging_commands(encoder.Finish());
+  after_submit();
+  if (!begin_frame_impl(false, true))
+    throw StagingCapacityError("Staging remap failed after capacity submission");
+  recycle_render_passes(g_renderPasses);
+  push_render_pass(std::move(continuation));
+  g_currentRenderPass = 0;
+  g_suspendedEfbPass = std::move(suspended);
+  for (unsigned i = 0; i < buffers.size(); ++i) {
+    if (!retained[i].empty()) buffers[i]->append(retained[i].data(), retained[i].size());
+  }
+  g_inOffscreen = offscreen;
+  g_cachedViewport = viewport;
+  g_cachedScissor = scissor;
+  gx::g_gxState.renderViewport = renderViewport;
+  gx::g_gxState.renderScissor = renderScissor;
+  gx::g_gxState.stateDirty = true;
+  push_command(CommandType::SetViewport, Command::Data{.setViewport = viewport});
+  push_command(CommandType::SetScissor, Command::Data{.setScissor = scissor});
+  ++g_stagingSplitCount;
+}
+
 uint32_t current_frame() noexcept { return g_frameIndex; }

 // The only place that erases from g_cachedBindGroups, whose handles the frame being encoded still
@@ -1196,10 +1321,10 @@ static const char* render_pass_label(u32 index) noexcept {
 }

 static void render_pass_impl(const wgpu::RenderPassEncoder& pass, const std::vector<RenderPass>& passes, u32 idx,
-                             int32_t interpolatedFrame);
+                             int32_t interpolatedFrame, DebugFrameData& debugFrame);

 static void render_impl(std::vector<RenderPass>& renderPasses, wgpu::CommandEncoder& cmd, int32_t interpolatedFrame,
-                        bool finalize) {
+                        bool finalize, DebugFrameData& debugFrame, const depth_peek::FrameMapping& depthMapping) {
  ZoneScoped;
  // Palette conversions, MSAA resolves and EFB copies depend on sealed frame state, not on the
  // interpolation weight, so encode them on the native render and let replay slots sample them.
@@ -1249,11 +1374,11 @@ static void render_impl(std::vector<RenderPass>& renderPasses, wgpu::CommandEnco
    };

    auto pass = cmd.BeginRenderPass(&renderPassDescriptor);
-    render_pass_impl(pass, renderPasses, i, interpolatedFrame);
+    render_pass_impl(pass, renderPasses, i, interpolatedFrame, debugFrame);
    pass.End();

    if (finalize && i == renderPasses.size() - 1) {
-      depth_peek::encode_frame_snapshot(cmd, passInfo.copySourceDepthView, passInfo.targetSize, passInfo.msaaSamples);
+      depth_peek::encode_frame_snapshot(cmd, passInfo.copySourceDepthView, passInfo.targetSize, passInfo.msaaSamples, depthMapping);
    }

    if (passInfo.resolveTarget) {
@@ -1327,20 +1452,21 @@ static void render_impl(std::vector<RenderPass>& renderPasses, wgpu::CommandEnco
  }

 #if defined(AURORA_GFX_DEBUG_GROUPS)
-  if (finalize && !g_debugGroupStack.empty()) {
-    for (auto& it : std::ranges::reverse_view(g_debugGroupStack)) {
+  if (finalize && !debugFrame.groups.empty()) {
+    for (auto& it : std::ranges::reverse_view(debugFrame.groups)) {
      Log.warn("Debug group was not popped at end of frame: {}", it);
    }
-    g_debugGroupStack.clear();
+    debugFrame.groups.clear();
  }

-  if (finalize && g_debugMarkers.size() > 0) {
-    g_debugMarkers.clear();
+  if (finalize && debugFrame.markers.size() > 0) {
+    debugFrame.markers.clear();
  }
 #endif
 }

 void seal_frame(SealedFrame& out) noexcept {
+  out.data().depthMapping = depth_peek::capture_frame_mapping();
  ZoneScoped;
  // The encode that could still have been holding these has completed: the
  // producer joins the worker's DONE phase before it seals another frame.
@@ -1350,15 +1476,24 @@ void seal_frame(SealedFrame& out) noexcept {
  // capacity included, back to the producer.
  recycle_render_passes(passes);
  passes.swap(g_renderPasses);
+#ifdef AURORA_GFX_DEBUG_GROUPS
+  // Marker indices and unmatched-group warnings belong to these detached passes.
+  // The next producer frame must not modify strings still read by this encoder.
+  auto& debug = out.data().debug;
+  debug.groups.clear();
+  debug.markers.clear();
+  debug.groups.swap(g_debugFrame.groups);
+  debug.markers.swap(g_debugFrame.markers);
+#endif
  g_currentRenderPass = UINT32_MAX;
 }

 void render(SealedFrame& frame, wgpu::CommandEncoder& cmd, int32_t interpolatedFrame, bool finalize) {
-  render_impl(frame.data().passes, cmd, interpolatedFrame, finalize);
+  render_impl(frame.data().passes, cmd, interpolatedFrame, finalize, frame.data().debug, frame.data().depthMapping);
 }

 void render(wgpu::CommandEncoder& cmd, int32_t interpolatedFrame, bool finalize) {
-  render_impl(g_renderPasses, cmd, interpolatedFrame, finalize);
+  render_impl(g_renderPasses, cmd, interpolatedFrame, finalize, g_debugFrame, depth_peek::capture_frame_mapping());
  if (finalize) {
    g_currentRenderPass = UINT32_MAX;
    expire_bind_group_cache();
@@ -1376,7 +1511,7 @@ void after_submit() noexcept {
 }

 static void render_pass_impl(const wgpu::RenderPassEncoder& pass, const std::vector<RenderPass>& renderPasses, u32 idx,
-                             int32_t interpolatedFrame) {
+                             int32_t interpolatedFrame, DebugFrameData& debugFrame) {
  // Per-invocation, not per-process: two encoders can be recording at once.
  gx::DrawEncodeState encodeState{};
  encodeState.boundTextureBindGroup = gx::g_emptyTextureBindGroup.Get();
@@ -1410,10 +1545,19 @@ static void render_pass_impl(const wgpu::RenderPassEncoder& pass, const std::vec
    switch (cmd.type) {
    case CommandType::SetViewport: {
      const auto& vp = cmd.data.setViewport;
-      // WebGPU requires 0 <= minDepth <= maxDepth <= 1, and the guest's (near, far) order is already
-      // reproduced in clip space. Passing the raw swapped pair diverged per backend in release builds.
-      const float minDepth = std::clamp(std::min(vp.znear, vp.zfar), 0.0f, 1.0f);
-      const float maxDepth = std::clamp(std::max(vp.znear, vp.zfar), 0.0f, 1.0f);
+      // WebGPU requires 0 <= minDepth <= maxDepth <= 1. vp.znear/vp.zfar are in GX's own distance
+      // terms (0 = near); under UseReversedZ the host depth-buffer storage direction is flipped
+      // (near = 1, far = 0), so this range has to be remapped through 1-x the same way the
+      // projection matrix, depth compare function, and clear value all are - a plain min/max clamp
+      // (the previous code here) maps a *restricted* range (e.g. a viewport deliberately narrowed
+      // to force something to draw "in front of everything") to the wrong end of the buffer: what
+      // should land near the near-storage-extreme (1.0) instead lands near the far-storage-extreme
+      // (0.0), so anything else drawn afterward at its true depth wins the compare test and the
+      // "in front" geometry silently vanishes. A full [0,1] viewport is unaffected either way,
+      // which is why this only broke specific elements, not the whole scene. Matches upstream
+      // aurora's apply_viewport (lib/gfx/encoding.cpp) exactly.
+      const float minDepth = gx::UseReversedZ ? 1.0f - vp.zfar : vp.znear;
+      const float maxDepth = gx::UseReversedZ ? 1.0f - vp.znear : vp.zfar;
      pass.SetViewport(vp.left, vp.top, vp.width, vp.height, minDepth, maxDepth);
    } break;
    case CommandType::SetScissor: {
@@ -1447,7 +1591,7 @@ static void render_pass_impl(const wgpu::RenderPassEncoder& pass, const std::vec
    } break;
    case CommandType::DebugMarker: {
 #if defined(AURORA_GFX_DEBUG_GROUPS)
-      pass.InsertDebugMarker(wgpu::StringView(g_debugMarkers[cmd.data.debugMarkerIndex]));
+      pass.InsertDebugMarker(wgpu::StringView(debugFrame.markers[cmd.data.debugMarkerIndex]));
 #endif
    } break;
    }
@@ -1470,8 +1614,8 @@ bool bind_pipeline(PipelineRef ref, const wgpu::RenderPassEncoder& pass, Pipelin
  if (!skip_unready_pipelines()) {
    pipelineReady = wait_pipeline(ref, pipeline);
  } else if (requireReady) {
-    // The pass resolves into a persistent texture (a one-shot bake such as MKW's minimap), so a
-    // skipped draw would never be re-issued. These run behind loads, not mid-race.
+    // Texture copies and capacity prefixes must retain complete draw results.
+    // A future display frame cannot repair a texture that already captured them.
    pipelineReady = wait_pipeline_for_persistent_pass(ref, pipeline);
  } else {
    pipelineReady = try_pipeline(ref, pipeline);
@@ -1600,8 +1744,8 @@ uint32_t align_uniform(uint32_t value) { return AURORA_ALIGN(value, g_cachedLimi

 void insert_debug_marker(std::string label) {
 #if defined(AURORA_GFX_DEBUG_GROUPS)
-  auto idx = g_debugMarkers.size();
-  g_debugMarkers.emplace_back(std::move(label));
+  auto idx = g_debugFrame.markers.size();
+  g_debugFrame.markers.emplace_back(std::move(label));
  push_command(CommandType::DebugMarker, {.debugMarkerIndex = idx});
 #endif
 }
@@ -1610,22 +1754,22 @@ void insert_debug_marker(std::string label) {

 void aurora::gfx::push_debug_group(std::string label) {
 #if defined(AURORA_GFX_DEBUG_GROUPS)
-  g_debugGroupStack.push_back(std::move(label));
+  g_debugFrame.groups.push_back(std::move(label));
 #endif
 }
 void aurora_push_debug_group(const char* label) {
 #ifdef AURORA_GFX_DEBUG_GROUPS
-  aurora::gfx::g_debugGroupStack.emplace_back(label);
+  aurora::gfx::g_debugFrame.groups.emplace_back(label);
 #endif
 }
 void aurora_pop_debug_group() {
 #ifdef AURORA_GFX_DEBUG_GROUPS
-  if (aurora::gfx::g_debugGroupStack.empty()) {
+  if (aurora::gfx::g_debugFrame.groups.empty()) {
    aurora::gfx::Log.error("Debug group stack underflowed!");
    return;
  }

-  aurora::gfx::g_debugGroupStack.pop_back();
+  aurora::gfx::g_debugFrame.groups.pop_back();
 #endif
 }

@@ -1,4 +1,5 @@
 #pragma once
+#include "staging_capacity.hpp"

 #include "../internal.hpp"
 #include "../webgpu/gpu.hpp"
@@ -394,6 +395,20 @@ wgpu::BindGroup& find_bind_group(BindGroupRef id);
 wgpu::Sampler& sampler_ref(const wgpu::SamplerDescriptor& descriptor);

 uint32_t align_uniform(uint32_t value);
+uint64_t staging_uniform_bytes(uint64_t bytes);
+uint64_t staging_storage_bytes(uint64_t bytes);
+// Admission does not allocate. A false result requires a producer-side split.
+// Oversized operations fail before mutating the current draw/pass.
+bool staging_has_space(const StagingSizes& demand);
+void ensure_staging_space(const StagingSizes& demand);
+void split_staging_batch();
+uint64_t staging_epoch() noexcept;
+StagingSizes staging_usage() noexcept;
+StagingSizes staging_high_water() noexcept;
+uint64_t staging_split_count() noexcept;
+// Internal integration-test seam: never increases the physical allocation.
+void set_staging_capacity_limits_for_testing(const StagingSizes& limits);
+

 Vec2<uint32_t> get_render_target_size() noexcept;
 // Same value as get_render_target_size() outside a render pass, but never
@@ -92,7 +92,7 @@ struct Params {

 constexpr std::string_view ReversedZBody = R"(
 fn gx_z24(depth: f32) -> u32 {
-    return min(u32(clamp(depth, 0.0, 1.0) * 16777216.0), 0x00ffffffu);
+    return min(u32(clamp(1.0 - depth, 0.0, 1.0) * 16777215.0 + 0.5), 0x00ffffffu);
 }
 )"sv;

@@ -196,7 +196,8 @@ wgpu::BindGroupLayout create_bind_group_layout(const char* label) {
  return g_device.CreateBindGroupLayout(&descriptor);
 }

-Params make_params(wgpu::Extent3D sourceSize, Vec2<uint32_t> dstSize) noexcept {
+Params make_params(wgpu::Extent3D sourceSize, const FrameMapping& mapping) noexcept {
+  const auto dstSize = mapping.logicalSize;
  Params params{
      .dstWidth = dstSize.x,
      .dstHeight = dstSize.y,
@@ -204,16 +205,16 @@ Params make_params(wgpu::Extent3D sourceSize, Vec2<uint32_t> dstSize) noexcept {
      .srcHeight = sourceSize.height,
  };

-  if (gx::g_gxState.viewportPolicy == AURORA_VIEWPORT_NATIVE) {
+  if (mapping.viewportPolicy == AURORA_VIEWPORT_NATIVE) {
    return params;
  }

-  const auto logicalSize = vi::configured_fb_size();
+  const auto logicalSize = mapping.logicalSize;
  if (logicalSize.x == 0 || logicalSize.y == 0 || sourceSize.width == 0 || sourceSize.height == 0) {
    return params;
  }

-  const bool stretch = gx::g_gxState.viewportPolicy == AURORA_VIEWPORT_STRETCH;
+  const bool stretch = mapping.viewportPolicy == AURORA_VIEWPORT_STRETCH;
  const float scaleX = static_cast<float>(sourceSize.width) / static_cast<float>(logicalSize.x);
  const float scaleY = static_cast<float>(sourceSize.height) / static_cast<float>(logicalSize.y);
  const float scale = std::min(scaleX, scaleY);
@@ -336,8 +337,12 @@ void poll() noexcept {
  }
 }

+FrameMapping capture_frame_mapping() noexcept {
+  return {vi::configured_fb_size(), gx::g_gxState.viewportPolicy};
+}
+
 void encode_frame_snapshot(const wgpu::CommandEncoder& cmd, const wgpu::TextureView& depthView,
-                           wgpu::Extent3D sourceSize, uint32_t msaaSamples) noexcept {
+                           wgpu::Extent3D sourceSize, uint32_t msaaSamples, const FrameMapping& mapping) noexcept {
  ZoneScoped;
  const auto now = Clock::now();
  {
@@ -349,7 +354,7 @@ void encode_frame_snapshot(const wgpu::CommandEncoder& cmd, const wgpu::TextureV
    g_nextSnapshotTime = now + SnapshotInterval;
  }

-  const auto dstSize = vi::configured_fb_size();
+  const auto dstSize = mapping.logicalSize;
  if (!depthView || dstSize.x == 0 || dstSize.y == 0 || sourceSize.width == 0 || sourceSize.height == 0) {
    return;
  }
@@ -357,7 +362,7 @@ void encode_frame_snapshot(const wgpu::CommandEncoder& cmd, const wgpu::TextureV
    Log.fatal("Depth Peek from multisampled EFB targets is not supported");
  }

-  const Params params = make_params(sourceSize, dstSize);
+  const Params params = make_params(sourceSize, mapping);
  wgpu::Buffer storageBuffer;
  wgpu::Buffer readbackBuffer;
  wgpu::Buffer paramsBuffer;
@@ -1,6 +1,7 @@
 #pragma once

 #include "common.hpp"
+#include <dolphin/gx/GXAurora.h>

 #include <vector>

@@ -13,8 +14,15 @@ void request_snapshot() noexcept;
 bool read_latest(uint16_t x, uint16_t y, uint32_t& z) noexcept;
 void poll() noexcept;

+// Captured before SEALED; the producer may configure the next frame during encode.
+struct FrameMapping {
+  Vec2<uint32_t> logicalSize{};
+  AuroraViewportPolicy viewportPolicy = AURORA_VIEWPORT_FIT;
+};
+FrameMapping capture_frame_mapping() noexcept;
+
 void encode_frame_snapshot(const wgpu::CommandEncoder& cmd, const wgpu::TextureView& depthView,
-                           wgpu::Extent3D sourceSize, uint32_t msaaSamples) noexcept;
+                           wgpu::Extent3D sourceSize, uint32_t msaaSamples, const FrameMapping& mapping) noexcept;
 void after_submit() noexcept;

 namespace testing {
@@ -8,7 +8,9 @@
 #include <algorithm>
 #include <array>
 #include <cstring>
+#include <memory>
 #include <mutex>
+#include <string>
 #include <unordered_map>
 #include <vector>

@@ -27,7 +29,7 @@ using webgpu::g_instance;
 constexpr size_t kAsyncReadbackMaxBytes = 256;
 // Each destination keeps its readback buffer forever. Only a handful are expected, and the cap
 // stops an unexpected pattern of one-shot destinations from leaking GPU buffers.
-constexpr size_t kMaxAsyncSlots = 32;
+constexpr size_t kMaxAsyncSlots = MaxAsyncReadbackSlots;

 struct PendingCopy {
  void* dest = nullptr;
@@ -37,6 +39,7 @@ struct PendingCopy {
  TextureHandle texture;
  TextureHandle nativeTexture;
  Range nativeBlitUniform;
+  uint64_t nativeUniformEpoch = 0;
 };

 struct Download {
@@ -81,6 +84,7 @@ std::vector<PendingCopy> g_asyncSealed;
 std::mutex g_asyncMutex;
 std::unordered_map<void*, AsyncSlot> g_asyncSlots;
 uint32_t g_asyncMapsInFlight = 0;
+uint64_t g_asyncGeneration = 1;

 uint32_t align_to(uint32_t value, uint32_t alignment) noexcept { return (value + alignment - 1) & ~(alignment - 1); }

@@ -90,10 +94,10 @@ void ensure_native_texture(PendingCopy& pending, TextureHandle* cache = nullptr)
  if (pending.texture->size.width == pending.width && pending.texture->size.height == pending.height) {
    return;
  }
+  if (pending.nativeTexture && pending.nativeUniformEpoch == staging_epoch()) return;
  if (pending.nativeTexture) {
-    return;
-  }
-  if (cache != nullptr && *cache && (*cache)->size.width == pending.width &&
+    // Keep the texture; its old staging range belongs to a submitted batch.
+  } else if (cache != nullptr && *cache && (*cache)->size.width == pending.width &&
      (*cache)->size.height == pending.height) {
    pending.nativeTexture = *cache;
  } else {
@@ -102,10 +106,12 @@ void ensure_native_texture(PendingCopy& pending, TextureHandle* cache = nullptr)
      *cache = pending.nativeTexture;
    }
  }
+  // The shared blit shader clamps Y to flags.z/w; preserve the full source.
  const std::array nativeBlitUniform{
-      0.0f, 0.0f, 1.0f, 1.0f, 0.0f, 64.0f, 0.0f, 0.0f, 0.0f, 1.0f, 0.0f, 0.0f,
+      0.0f, 0.0f, 1.0f, 1.0f, 0.0f, 64.0f, 0.0f, 0.0f, 0.0f, 1.0f, 0.0f, 1.0f,
  };
  pending.nativeBlitUniform = push_uniform(nativeBlitUniform);
+  pending.nativeUniformEpoch = staging_epoch();
 }

 void encode_native_blit(const wgpu::CommandEncoder& encoder, const PendingCopy& pending) noexcept {
@@ -125,16 +131,17 @@ HostPixelOrder texture_pixel_order(const TextureHandle& texture) noexcept {
  return texture->format == wgpu::TextureFormat::BGRA8Unorm ? HostPixelOrder::BGRA : HostPixelOrder::RGBA;
 }

-void complete_async_slot(void* dest, wgpu::MapAsyncStatus status, wgpu::StringView message) noexcept {
+void complete_async_slot(void* dest, uint64_t generation, wgpu::MapAsyncStatus status,
+                         wgpu::StringView message) noexcept {
  std::lock_guard lock{g_asyncMutex};
-  if (g_asyncMapsInFlight > 0) {
-    --g_asyncMapsInFlight;
-  }
+  if (generation != g_asyncGeneration) return;
  const auto it = g_asyncSlots.find(dest);
  if (it == g_asyncSlots.end()) {
    return;
  }
  auto& slot = it->second;
+  if (slot.state != AsyncState::MapPending) return;
+  if (g_asyncMapsInFlight > 0) --g_asyncMapsInFlight;
  if (status == wgpu::MapAsyncStatus::Success) {
    const auto* pixels = static_cast<const uint8_t*>(slot.buffer.GetConstMappedRange(0, slot.bufferSize));
    if (pixels != nullptr) {
@@ -227,7 +234,14 @@ bool has_pending(void* dest) noexcept {
                     [dest](const Download& download) { return download.copy.dest == dest; });
 }

-bool prepare_downloads(void* dest) noexcept {
+bool prepare_downloads(void* dest) {
+  uint64_t copies = 0;
+  for (const auto& pending : g_pending) {
+    if (dest != nullptr && pending.dest != dest) continue;
+    if (pending.texture->size.width != pending.width || pending.texture->size.height != pending.height) ++copies;
+  }
+  // Reserve all copies, even already-prepared ones: a split retires their ranges.
+  ensure_staging_space({0, copies * staging_uniform_bytes(48), 0, 0});
  bool found = false;
  for (auto& pending : g_pending) {
    if (dest != nullptr && pending.dest != dest) continue;
@@ -290,15 +304,35 @@ void encode_downloads(const wgpu::CommandEncoder& encoder, void* dest) noexcept
 bool complete_downloads() noexcept {
  bool success = true;
  for (auto& download : g_downloads) {
-    wgpu::MapAsyncStatus mapStatus = wgpu::MapAsyncStatus::CallbackCancelled;
-    wgpu::StringView mapMessage{};
+    // WaitAny may time out before Dawn delivers cancellation. The callback must
+    // own its result rather than retaining references to this stack frame.
+    struct MapResult {
+      std::mutex mutex;
+      wgpu::MapAsyncStatus status = wgpu::MapAsyncStatus::CallbackCancelled;
+      std::string message;
+    };
+    const auto result = std::make_shared<MapResult>();
    const auto future =
        download.buffer.MapAsync(wgpu::MapMode::Read, 0, download.bufferSize, wgpu::CallbackMode::WaitAnyOnly,
-                                 [&mapStatus, &mapMessage](wgpu::MapAsyncStatus status, wgpu::StringView message) {
-                                   mapStatus = status;
-                                   mapMessage = message;
+                                 [result](wgpu::MapAsyncStatus status, wgpu::StringView message) {
+                                   std::lock_guard lock{result->mutex};
+                                   result->status = status;
+                                   if (message.data != nullptr) {
+                                     size_t length = 0;
+                                     while (length < 512 && length < message.length && message.data[length] != '\0') {
+                                       ++length;
+                                     }
+                                     result->message.assign(message.data, length);
+                                   }
                                 });
    const auto waitStatus = g_instance.WaitAny(future, 5000000000);
+    wgpu::MapAsyncStatus mapStatus;
+    std::string mapMessage;
+    {
+      std::lock_guard lock{result->mutex};
+      mapStatus = result->status;
+      mapMessage = result->message;
+    }
    if (waitStatus != wgpu::WaitStatus::Success || mapStatus != wgpu::MapAsyncStatus::Success) {
      Log.error("EFB RAM readback failed wait={} map={} message={}", magic_enum::enum_name(waitStatus),
                magic_enum::enum_name(mapStatus), mapMessage);
@@ -412,6 +446,7 @@ void after_submit() noexcept {
    void* dest;
    wgpu::Buffer buffer;
    uint64_t bufferSize;
+    uint64_t generation;
  };
  std::vector<PendingMap> pendingMaps;
  {
@@ -422,14 +457,15 @@ void after_submit() noexcept {
      }
      slot.state = AsyncState::MapPending;
      ++g_asyncMapsInFlight;
-      pendingMaps.push_back({dest, slot.buffer, slot.bufferSize});
+      pendingMaps.push_back({dest, slot.buffer, slot.bufferSize, g_asyncGeneration});
    }
  }

  for (const auto& pending : pendingMaps) {
    pending.buffer.MapAsync(wgpu::MapMode::Read, 0, pending.bufferSize, wgpu::CallbackMode::AllowSpontaneous,
-                            [dest = pending.dest](wgpu::MapAsyncStatus status, wgpu::StringView message) {
-                              complete_async_slot(dest, status, message);
+                            [dest = pending.dest, generation = pending.generation](wgpu::MapAsyncStatus status,
+                                                                                 wgpu::StringView message) {
+                              complete_async_slot(dest, generation, status, message);
                            });
  }

@@ -443,9 +479,15 @@ void abort_async() noexcept { g_asyncSealed.clear(); }
 void shutdown() noexcept {
  cancel();
  g_asyncSealed.clear();
-  std::lock_guard lock{g_asyncMutex};
-  g_asyncSlots.clear();
-  g_asyncMapsInFlight = 0;
+  // Retire callbacks before releasing buffers, and release outside their mutex:
+  // destruction may itself deliver an AllowSpontaneous cancellation callback.
+  decltype(g_asyncSlots) retiredSlots;
+  {
+    std::lock_guard lock{g_asyncMutex};
+    ++g_asyncGeneration;
+    retiredSlots.swap(g_asyncSlots);
+    g_asyncMapsInFlight = 0;
+  }
 }

 } // namespace aurora::gfx::efb_ram
@@ -7,9 +7,11 @@

 namespace aurora::gfx::efb_ram {

+inline constexpr size_t MaxAsyncReadbackSlots = 32;
+
 void schedule(void* dest, uint32_t width, uint32_t height, GXTexFmt format, TextureHandle texture) noexcept;
 bool has_pending(void* dest = nullptr) noexcept;
-bool prepare_downloads(void* dest = nullptr) noexcept;
+bool prepare_downloads(void* dest = nullptr);
 void encode_downloads(const wgpu::CommandEncoder& encoder, void* dest = nullptr) noexcept;
 bool complete_downloads() noexcept;
 void cancel() noexcept;
@@ -396,6 +396,7 @@ static PendingPipeline* touch_pending_pipeline(PipelineRef hash, bool prioritize

  g_priorityPipelines.emplace_back(std::move(*backgroundIt));
  g_backgroundPipelines.erase(backgroundIt);
+  g_pipelineCv.notify_all();
  return &g_priorityPipelines.back();
 }

@@ -530,7 +531,8 @@ static PipelineRef find_pipeline_impl(ShaderType type, const PipelineConfig& con
  }

  if (notifyWorker) {
-    g_pipelineCv.notify_one();
+    // Compiler workers and renderer waiters share this condition variable.
+    g_pipelineCv.notify_all();
  }
  if (notifyWaiters) {
    g_pipelineCv.notify_all();
@@ -0,0 +1,33 @@
+#pragma once
+#include <array>
+#include <cstdint>
+#include <limits>
+#include <stdexcept>
+
+namespace aurora::gfx {
+// Byte counts after each allocation's own trailing alignment, in V/U/I/S order.
+using StagingSizes = std::array<uint64_t, 4>;
+class StagingCapacityError : public std::runtime_error {
+public:
+  using std::runtime_error::runtime_error;
+};
+struct StagingBatchFull {};
+inline uint64_t staging_padded(uint64_t bytes, uint64_t alignment) {
+  if (!bytes) return alignment;
+  const auto remainder = alignment ? bytes % alignment : 0;
+  const auto padding = remainder ? alignment - remainder : 0;
+  if (bytes > UINT64_MAX - padding) throw StagingCapacityError("Staging allocation size overflow");
+  return bytes + padding;
+}
+inline bool staging_fits(const StagingSizes& used, const StagingSizes& demand,
+                         const StagingSizes& tail, const StagingSizes& capacity) noexcept {
+  for (unsigned i = 0; i < used.size(); ++i) {
+    const auto limit = capacity[i] < UINT32_MAX ? capacity[i] : UINT32_MAX;
+    // The final GPU copy rounds to four bytes. Subtractions avoid wraparound.
+    const auto alignedLimit = limit & ~uint64_t(3);
+    if (tail[i] > alignedLimit || used[i] > alignedLimit - tail[i] ||
+        demand[i] > alignedLimit - tail[i] - used[i]) return false;
+  }
+  return true;
+}
+} // namespace aurora::gfx
@@ -0,0 +1,61 @@
+#pragma once
+
+#include <chrono>
+#include <condition_variable>
+#include <cstdint>
+#include <mutex>
+
+namespace aurora::gfx {
+
+enum class BufferMapState { Unmapped, Mapping, Mapped };
+
+// The renderer owns request/reset; Dawn may complete a request on another thread.
+// An old callback must never publish readiness for a different staging slot.
+class StagingMapState {
+  mutable std::mutex mutex_;
+  std::condition_variable changed_;
+  uint64_t generation_ = 0;
+  BufferMapState state_ = BufferMapState::Unmapped;
+
+public:
+  uint64_t request() {
+    std::lock_guard lock(mutex_);
+    if (state_ != BufferMapState::Unmapped) return 0;
+    state_ = BufferMapState::Mapping;
+    return ++generation_;
+  }
+
+  bool complete(uint64_t generation, BufferMapState state) {
+    {
+      std::lock_guard lock(mutex_);
+      if (generation != generation_ || state_ != BufferMapState::Mapping) return false;
+      state_ = state;
+    }
+    changed_.notify_all();
+    return true;
+  }
+
+  void reset() {
+    {
+      std::lock_guard lock(mutex_);
+      ++generation_;
+      state_ = BufferMapState::Unmapped;
+    }
+    changed_.notify_all();
+  }
+
+  BufferMapState state() const {
+    std::lock_guard lock(mutex_);
+    return state_;
+  }
+
+  void wait_for_progress() {
+    std::unique_lock lock(mutex_);
+    // ProcessEvents is still serviced between waits for implementations that
+    // need it. A spontaneous completion wakes immediately, without polling.
+    changed_.wait_for(lock, std::chrono::milliseconds(1),
+                      [&] { return state_ != BufferMapState::Mapping; });
+  }
+};
+
+} // namespace aurora::gfx
@@ -137,7 +137,7 @@ fn gx_z24_at_coord(unclamped_coord: vec2i) -> u32 {
    let tex_size = vec2i(textureDimensions(src));
    let coord = clamp(unclamped_coord, vec2i(0), tex_size - vec2i(1));
    let depth = textureLoad(src, coord, 0);
-    return min(u32(clamp(depth, 0.0, 1.0) * 16777216.0), 0x00ffffffu);
+    return min(u32(clamp(1.0 - depth, 0.0, 1.0) * 16777215.0 + 0.5), 0x00ffffffu);
 }
 )"s
                        : R"(
@@ -373,7 +373,7 @@ static wgpu::BindGroupLayout g_depthBindGroupLayout;
 static wgpu::Sampler g_nearestSampler;
 static wgpu::Sampler g_linearSampler;
 static absl::flat_hash_map<GXTexFmt, wgpu::RenderPipeline> g_pipelines;
-static wgpu::RenderPipeline g_blitPipeline;
+static absl::flat_hash_map<wgpu::TextureFormat, wgpu::RenderPipeline> g_blitPipelines;

 static wgpu::RenderPipeline create_pipeline(const ConvPipeline& conv, const std::string_view shaderPreamble,
                                            const wgpu::BindGroupLayout& bindGroupLayout) {
@@ -487,9 +487,12 @@ void initialize() {
  };
  g_depthBindGroupLayout = g_device.CreateBindGroupLayout(&depthBindGroupLayoutDescriptor);

-  g_blitPipeline = create_pipeline(
-      {GX_TF_RGBA8, FragPassthrough, webgpu::g_graphicsConfig.surfaceConfiguration.format, "TexCopyConv Blit"},
-      ShaderPreamble, g_bindGroupLayout);
+  // Native RAM readback uses RGBA even when the EFB/surface uses BGRA.
+  // Build both variants here; frame workers only read the completed map.
+  for (const auto format : {wgpu::TextureFormat::RGBA8Unorm, wgpu::TextureFormat::BGRA8Unorm}) {
+    g_blitPipelines[format] = create_pipeline(
+        {GX_TF_RGBA8, FragPassthrough, format, "TexCopyConv Blit"}, ShaderPreamble, g_bindGroupLayout);
+  }
  for (const auto& conv : ConvPipelines) {
    g_pipelines[conv.fmt] = create_pipeline(conv, ShaderPreamble, g_bindGroupLayout);
    if (conv.outputFormat != to_wgpu(conv.fmt)) {
@@ -520,7 +523,7 @@ void initialize() {

 void shutdown() {
  g_pipelines.clear();
-  g_blitPipeline = {};
+  g_blitPipelines.clear();
  g_bindGroupLayout = {};
  g_depthBindGroupLayout = {};
  g_nearestSampler = {};
@@ -602,6 +605,12 @@ void run(const wgpu::CommandEncoder& cmd, const ConvRequest& req) {
  execute(cmd, req, it->second);
 }

-void blit(const wgpu::CommandEncoder& cmd, const ConvRequest& req) { execute(cmd, req, g_blitPipeline); }
+void blit(const wgpu::CommandEncoder& cmd, const ConvRequest& req) {
+  const auto it = g_blitPipelines.find(req.dst->format);
+  if (it == g_blitPipelines.end()) {
+    Log.fatal("Unsupported blit destination format {}", static_cast<int>(req.dst->format));
+  }
+  execute(cmd, req, it->second);
+}

 } // namespace aurora::gfx::tex_copy_conv
@@ -30,10 +30,11 @@ using IndexBuffer = std::vector<u16>;
 static u32 prepare_idx_template(IndexBuffer& buf, GXPrimitive prim, u16 vtxCount) {
  size_t writePos = 0;
  if (prim == GX_QUADS) {
-    // Retain the existing incomplete-quad behavior: every started group emits a complete six-index quad.
-    buf.resize(((static_cast<u32>(vtxCount) + 3u) / 4u) * 6u);
+    // GX renders a three-vertex remainder as a triangle. One/two are ignored.
+    const u32 completeVertices = static_cast<u32>(vtxCount) & ~3u;
+    buf.resize((completeVertices / 4u) * 6u + (vtxCount % 4u == 3u ? 3u : 0u));

-    for (u16 v = 0; v < vtxCount; v += 4) {
+    for (u32 v = 0; v < completeVertices; v += 4) {
      const u16 idx0 = v;
      const u16 idx1 = static_cast<u16>(v + 1);
      const u16 idx2 = static_cast<u16>(v + 2);
@@ -45,15 +46,21 @@ static u32 prepare_idx_template(IndexBuffer& buf, GXPrimitive prim, u16 vtxCount
      buf[writePos++] = idx3;
      buf[writePos++] = idx0;
    }
+    if (vtxCount % 4u == 3u) {
+      buf[writePos++] = static_cast<u16>(completeVertices);
+      buf[writePos++] = static_cast<u16>(completeVertices + 1u);
+      buf[writePos++] = static_cast<u16>(completeVertices + 2u);
+    }
  } else if (prim == GX_TRIANGLES) {
-    buf.resize(vtxCount);
-    for (u16 v = 0; v < vtxCount; ++v) {
+    const u32 completeVertices = (static_cast<u32>(vtxCount) / 3u) * 3u;
+    buf.resize(completeVertices);
+    for (u32 v = 0; v < completeVertices; ++v) {
      buf[writePos++] = v;
    }
  } else if (prim == GX_TRIANGLEFAN) {
-    const u32 indexCount = vtxCount <= 3 ? vtxCount : 3u + (static_cast<u32>(vtxCount) - 3u) * 3u;
+    const u32 indexCount = vtxCount < 3 ? 0u : (static_cast<u32>(vtxCount) - 2u) * 3u;
    buf.resize(indexCount);
-    for (u16 v = 0; v < vtxCount; ++v) {
+    for (u32 v = 0; indexCount != 0 && v < vtxCount; ++v) {
      if (v < 3) {
        buf[writePos++] = v;
        continue;
@@ -63,9 +70,9 @@ static u32 prepare_idx_template(IndexBuffer& buf, GXPrimitive prim, u16 vtxCount
      buf[writePos++] = v;
    }
  } else if (prim == GX_TRIANGLESTRIP) {
-    const u32 indexCount = vtxCount <= 3 ? vtxCount : 3u + (static_cast<u32>(vtxCount) - 3u) * 3u;
+    const u32 indexCount = vtxCount < 3 ? 0u : (static_cast<u32>(vtxCount) - 2u) * 3u;
    buf.resize(indexCount);
-    for (u16 v = 0; v < vtxCount; ++v) {
+    for (u32 v = 0; indexCount != 0 && v < vtxCount; ++v) {
      if (v < 3) {
        buf[writePos++] = v;
        continue;
@@ -88,6 +95,13 @@ static u32 prepare_idx_template(IndexBuffer& buf, GXPrimitive prim, u16 vtxCount
  return static_cast<u32>(writePos);
 }

+// Empty/incomplete draws consume FIFO bytes but cannot produce a primitive.
+static bool has_complete_primitive(GXPrimitive prim, u16 count) {
+  if (prim == GX_POINTS) return count >= 1;
+  if (prim == GX_LINES || prim == GX_LINESTRIP) return count >= 2;
+  return count >= 3;
+}
+
 // GX FIFO opcodes - use CP_ prefix to avoid clashing with GXCommandList.h macros
 static constexpr u8 CP_CMD_NOP = GX_NOP;
 static constexpr u8 CP_CMD_LOAD_CP_REG = GX_LOAD_CP_REG;
@@ -466,13 +480,14 @@ static void handle_xf(const u8* data, u32& pos, u32 size, bool bigEndian);
 static bool handle_draw(u8 cmd, const u8* data, u32& pos, u32 size, bool bigEndian);
 static bool handle_aurora(const u8* data, u32& pos, u32 size, bool bigEndian);

-void process(const u8* data, u32 size, bool bigEndian) {
+uint32_t process(const u8* data, u32 size, bool bigEndian) {
  ZoneScoped;
  // Everything decoded here mutates renderer state (GX state, the recorded command lists and the mapped staging buffers), so take the renderer GPU mutex once for the whole drain rather than once per draw command.
  std::lock_guard gpuLock(aurora::renderer_gpu_mutex());
  u32 pos = 0;

  while (pos < size) {
+    const u32 commandStart = pos;
    u8 cmd = data[pos++];
    u8 opcode = cmd & CP_OPCODE_MASK;
    // Log.warn("Processing opcode {:02x} at pos {} (size {})", opcode, pos - 1, size);
@@ -551,12 +566,16 @@ void process(const u8* data, u32 size, bool bigEndian) {
      for (int i = GX_VA_POS; i <= GX_VA_TEX7; ++i) {
        g_gxState.arrays[i].cachedRange = {};
      }
+      // A merged draw retains its previous array uploads. Force a new draw so
+      // handle_draw_unmerged observes the invalidation and uploads fresh data.
+      // Pipeline configuration itself did not change.
+      g_gxState.stateDirty = true;
      break;
    }

    case GX_LOAD_AURORA: {
      if (!handle_aurora(data, pos, size, bigEndian)) {
-        return;
+        return size;
      }
      break;
    }
@@ -564,8 +583,10 @@ void process(const u8* data, u32 size, bool bigEndian) {
    default:
      // Draw commands occupy the full 0x80-0xBF range.
      if (is_draw_cmd(cmd)) {
-        if (!handle_draw(cmd, data, pos, size, bigEndian)) {
-          return;
+        try {
+          if (!handle_draw(cmd, data, pos, size, bigEndian)) return size;
+        } catch (const gfx::StagingBatchFull&) {
+          return commandStart;
        }
      } else {
        static u32 unknownLogCount = 0;
@@ -588,6 +609,7 @@ void process(const u8* data, u32 size, bool bigEndian) {
      break;
    }
  }
+  return size;
 }

 // Helper to extract bit fields from a 32-bit register
@@ -1848,6 +1870,10 @@ static u32 calculate_last_vtx_size(GXVtxFmt fmt) {

  g_gxState.lastVtxFmt = fmt;
  g_gxState.lastVtxSize = vtxSize;
+  // The format is selected by the draw opcode, without a register write.
+  // Even equal-stride formats may decode bytes differently, so do not merge
+  // into a draw using the previous format's shader and uniform layout.
+  g_gxState.stateDirty = true;

  return vtxSize;
 }
@@ -2080,6 +2106,22 @@ static const CachedPipelineState& resolve_pipeline_state(GXPrimitive prim, GXVtx
  return state;
 }

+static bool admit_draw(GXPrimitive prim, GXVtxFmt fmt, u16 count, uint32_t vertexBytes, bool merged = false) {
+  const auto& indexTemplate = cached_index_template(prim, count);
+  gfx::StagingSizes demand{vertexBytes, 0, indexTemplate.indices.size() * sizeof(u16), 0};
+  if (merged) return gfx::staging_has_space(demand);
+  const auto& info = resolve_pipeline_state(prim, fmt).shaderInfo;
+  demand[1] = gfx::staging_uniform_bytes(info.uniformSize);
+  if (frame_interpolation_identity_needed() && frame_interpolation_replay_safe())
+    demand[1] *= 1 + MaxInterpolatedFrames;
+  for (int i = GX_VA_POS; i <= GX_VA_TEX7; ++i) {
+    if ((g_gxState.vtxDesc[i] == GX_INDEX8 || g_gxState.vtxDesc[i] == GX_INDEX16) &&
+        g_gxState.arrays[i].cachedRange.size == 0)
+      demand[3] += gfx::staging_storage_bytes(g_gxState.arrays[i].size);
+  }
+  return gfx::staging_has_space(demand);
+}
+
 bool submit_raw_draw(GXPrimitive prim, GXVtxFmt fmt, const uint8_t* vertices, uint16_t vtxCount,
                     uint32_t vertexBytes) {
  ZoneScoped;
@@ -2112,8 +2154,17 @@ bool submit_raw_draw(GXPrimitive prim, GXVtxFmt fmt, const uint8_t* vertices, ui
    return false;
  }

+  if (!has_complete_primitive(prim, vtxCount)) return true;
+
  // This entry point bypasses process(), so it owns the renderer lock itself.
-  std::lock_guard gpuLock(aurora::renderer_gpu_mutex());
+  std::unique_lock gpuLock(aurora::renderer_gpu_mutex());
+  if (!admit_draw(prim, fmt, vtxCount, vertexBytes)) {
+    gpuLock.unlock();
+    gfx::split_staging_batch();
+    gpuLock.lock();
+    if (!admit_draw(prim, fmt, vtxCount, vertexBytes))
+      throw gfx::StagingCapacityError("Raw draw does not fit after capacity submission");
+  }
  const gfx::Range vertRange = gfx::push_verts(vertices, vertexBytes);
  const bool interpolationIdentityActive = frame_interpolation_identity_needed();
  const PnMtxUsage matrixUsage = interpolationIdentityActive
@@ -2151,17 +2202,32 @@ static bool handle_draw(u8 cmd, const u8* data, u32& pos, u32 size, bool bigEndi
  }


-  // Push raw vertex data to buffer
-  const uint8_t* vertices = data + pos;
-  gfx::Range vertRange = gfx::push_verts(vertices, totalVtxBytes);
-  pos += totalVtxBytes;
+  if (!has_complete_primitive(prim, vtxCount)) {
+    pos += totalVtxBytes;
+    return true;
+  }

+  DrawData* mergeTarget = nullptr;
+  // Decide admission before allocating anything. The merged path needs only
+  // vertices and indices; it must not resolve pipelines or upload arrays.
  // Try to merge with previous draw call
  if (!g_gxState.stateDirty) LIKELY {
    auto* lastDraw = gfx::get_last_draw_command<DrawData>();
-    // Only if the previous draw call was a single instance draw (no lines/points handling)
+    // Expanded lines/points have different vertex interpretation even with one instance.
+    // Triangle-list output has no restart index; index 65535 is usable.
+    // Overflow would address earlier vertices instead of the appended geometry.
    if (lastDraw != nullptr && prim != GX_LINES && prim != GX_LINESTRIP && prim != GX_POINTS &&
-        lastDraw->instanceCount == 1) LIKELY {
+        !lastDraw->expandedPrimitive && lastDraw->instanceCount == 1 &&
+        uint64_t(lastDraw->vtxCount) +
+            vtxCount <= 65536u) LIKELY {
+      mergeTarget = lastDraw;
+    }
+  }
+  if (!admit_draw(prim, fmt, vtxCount, totalVtxBytes, mergeTarget != nullptr)) throw gfx::StagingBatchFull{};
+  const uint8_t* vertices = data + pos;
+  gfx::Range vertRange = gfx::push_verts(vertices, totalVtxBytes);
+  pos += totalVtxBytes;
+  if (auto* lastDraw = mergeTarget) {
      const auto& indexTemplate = cached_index_template(prim, vtxCount);
      const auto indices = offset_index_template(indexTemplate, lastDraw->vtxCount);
      const u32 numIndices = indexTemplate.indexCount;
@@ -2182,7 +2248,6 @@ static bool handle_draw(u8 cmd, const u8* data, u32& pos, u32 size, bool bigEndi
        extend_interpolation_draw(pn_mtx_mask(vertices, vtxCount, vtxSize));
      }
      return true;
-    }
  }

  const bool interpolationIdentityActive = frame_interpolation_identity_needed();
@@ -2278,6 +2343,7 @@ static void handle_draw_unmerged(GXPrimitive prim, GXVtxFmt fmt, u16 vtxCount,
      .vtxCount = vtxCount,
      .indexCount = numIndices,
      .instanceCount = instanceCount,
+      .expandedPrimitive = prim == GX_LINES || prim == GX_LINESTRIP || prim == GX_POINTS,
      .bindGroups = bindGroups,
      .dstAlpha = pipelineState.dstAlpha,
  });
@@ -9,7 +9,7 @@ namespace aurora::gx::fifo {
 void reset_cp_register_cache();

 // Process a buffer of GX FIFO commands
-void process(const uint8_t* data, uint32_t size, bool bigEndian);
+uint32_t process(const uint8_t* data, uint32_t size, bool bigEndian);

 // Submit already-packed direct vertex bytes against the current GX state.
 bool submit_raw_draw(GXPrimitive prim, GXVtxFmt fmt, const uint8_t* vertices, uint16_t vtxCount,
@@ -1,5 +1,6 @@
 #include "fifo.hpp"
 #include "command_processor.hpp"
+#include "../gfx/common.hpp"
 #include "../internal.hpp"

 #include <chrono>
@@ -81,7 +82,18 @@ void drain() {
  if (detail::sBufferSize == 0) {
    return;
  }
-  process(detail::sBufferData, detail::sBufferSize, true);
+  uint32_t consumed = 0;
+  bool retried = false;
+  while (consumed < detail::sBufferSize) {
+    const auto count = process(detail::sBufferData + consumed, detail::sBufferSize - consumed, true);
+    if (count == 0 && retried)
+      throw gfx::StagingCapacityError("FIFO draw does not fit after capacity submission");
+    consumed += count;
+    if (consumed == detail::sBufferSize) break;
+    // process returned with its renderer lock released. No recursive drain.
+    gfx::split_staging_batch();
+    retried = true;
+  }
  detail::sBufferSize = 0;
 }

@@ -143,6 +143,7 @@ private:
 };
 struct FrameTransformSnapshot {
  Mat4x4<float> projection{};
+  HashType viewportIdentity = 0;
  Mat3x4<float> position{};
  Mat3x4<float> normal{};
  uint16_t usedMatrixMask = 1;
@@ -1185,7 +1186,8 @@ void finalize_frame_interpolation() noexcept {
        if ((transform.usedMatrixMask & (1u << slot)) == 0) {
          continue;
        }
-        paletteSlotKeys.push_back({transform.indexedMatrices->slotHash[slot], palette, slot});
+        paletteSlotKeys.push_back({combine_identity(transform.indexedMatrices->slotHash[slot],
+                                                   transform.viewportIdentity), palette, slot});
      }
    }
    std::sort(paletteSlotKeys.begin(), paletteSlotKeys.end(),
@@ -1446,10 +1448,21 @@ void extend_interpolation_draw(uint16_t usedPnMtxMask) noexcept {
 }

 std::array<gfx::Range, MaxInterpolatedFrames> record_interpolation_draw(
-    const FrameInterpolationDrawIdentity& identity, const Mat4x4<float>& projection,
+    const FrameInterpolationDrawIdentity& drawIdentity, const Mat4x4<float>& projection,
    uint16_t usedPnMtxMask, const InterpolatedUniformLayout& uniformLayout) noexcept {
+  // Split-screen cameras can draw identical meshes in unrelated view spaces.
+  // Scope exact, material-only and sibling-palette history to the guest viewport.
+  // Logical coordinates keep render-scale changes out of the camera identity.
+  const auto& viewport = g_gxState.logicalViewport;
+  const std::array viewportValues{viewport.left, viewport.top, viewport.width,
+                                  viewport.height, viewport.znear, viewport.zfar};
+  const HashType viewportIdentity = xxh3_hash_s(viewportValues.data(), sizeof(viewportValues));
+  auto identity = drawIdentity;
+  identity.combined = combine_identity(identity.combined, viewportIdentity);
+  identity.pipeline = combine_identity(identity.pipeline, viewportIdentity);
  FrameTransformSnapshot snapshot{
      .projection = projection,
+      .viewportIdentity = viewportIdentity,
      .usedMatrixMask = usedPnMtxMask,
  };
  if (uniformLayout.indexedMatrices) {
@@ -1416,23 +1416,32 @@ static inline GXBlendFactor remove_dst_alpha_usage(GXBlendFactor fac) {
  }
 }

+// GX_LEQUAL etc. describe "pass if this pixel is closer than/equal to what's stored" in GX's own
+// distance terms, independent of how that distance is encoded as a host depth value. Under
+// UseReversedZ the encoding is flipped (near=1, far=0), so "closer" now corresponds to a *larger*
+// stored value, not a smaller one - the ordered compare functions (LESS/LEQUAL/GREATER/GEQUAL)
+// must invert to match, or the depth test silently runs backwards (verified directly: this was
+// the actual cause of a bug report after the projection/shader half of the reverse-Z fix
+// eliminated the double-negation that used to accidentally keep the unreversed comparisons
+// correct - LEQUAL now needs GreaterEqual, not LessEqual, once the encoding it's testing against
+// is genuinely reversed). Matches upstream aurora's to_compare_function exactly.
 static inline wgpu::CompareFunction to_compare_function(GXCompare func) {
  switch (func) {
    DEFAULT_FATAL("invalid depth fn {}", underlying(func));
  case GX_NEVER:
    return wgpu::CompareFunction::Never;
  case GX_LESS:
-    return wgpu::CompareFunction::Less;
+    return UseReversedZ ? wgpu::CompareFunction::Greater : wgpu::CompareFunction::Less;
  case GX_EQUAL:
    return wgpu::CompareFunction::Equal;
  case GX_LEQUAL:
-    return wgpu::CompareFunction::LessEqual;
+    return UseReversedZ ? wgpu::CompareFunction::GreaterEqual : wgpu::CompareFunction::LessEqual;
  case GX_GREATER:
-    return wgpu::CompareFunction::Greater;
+    return UseReversedZ ? wgpu::CompareFunction::Less : wgpu::CompareFunction::Greater;
  case GX_NEQUAL:
    return wgpu::CompareFunction::NotEqual;
  case GX_GEQUAL:
-    return wgpu::CompareFunction::GreaterEqual;
+    return UseReversedZ ? wgpu::CompareFunction::LessEqual : wgpu::CompareFunction::GreaterEqual;
  case GX_ALWAYS:
    return wgpu::CompareFunction::Always;
  }
@@ -436,6 +436,8 @@ struct GXState {
  u32 pipelineStateGeneration = next_gx_state_epoch();
  std::array<u32, 0x100> bpRegCache = [] {
    std::array<u32, 0x100> regs{};
+    // Force the first GEN_MODE decode without changing its masked reset value.
+    regs[0x00] = 0xFF000000;
    regs[0xFE] = 0x00FFFFFF;
    return regs;
  }();
@@ -485,7 +487,14 @@ const gfx::TextureBind& get_texture(GXTexMapID id) noexcept;
 void resolve_sampled_textures(const ShaderInfo& info) noexcept;

 inline float clear_depth_value() {
-  return std::min(static_cast<float>(g_gxState.clearDepth) / 16777216.f, 16777215.f / 16777216.f);
+  // g_gxState.clearDepth is in GX's own distance terms (0 = near, larger = farther), independent of
+  // how UseReversedZ encodes that as a host depth value - it must be re-mapped the same way the
+  // projection matrix and depth compare function are, or the buffer clears to the wrong extreme
+  // (verified directly: matches upstream aurora's clear_depth_value, which does this same inversion
+  // and was the second missing piece alongside to_compare_function's compare-op inversion).
+  const float normalizedDepth =
+      std::min(static_cast<float>(g_gxState.clearDepth) / 16777216.f, 16777215.f / 16777216.f);
+  return UseReversedZ ? (1.f - normalizedDepth) : normalizedDepth;
 }

 inline bool render_target_has_alpha(GXPixelFmt pixelFmt) noexcept { return pixelFmt == GX_PF_RGBA6_Z24; }
@@ -13,6 +13,7 @@ struct DrawData {
  uint32_t vtxCount;
  uint32_t indexCount;
  uint32_t instanceCount;
+  bool expandedPrimitive;
  GXBindGroups bindGroups;
  uint32_t dstAlpha;
 };
@@ -993,11 +993,13 @@ wgpu::ShaderModule build_shader(const ShaderConfig& config) noexcept {
        "\n    let clip_base = select(clip_a, clip_b, use_b);"
        "\n    out.pos = vec4f(clip_base.xy + offset_ndc * clip_base.w, clip_base.zw);";
  }
-  if constexpr (UseReversedZ) {
-    vtxXfrAttrsPre += "\n    out.pos.z = -out.pos.z;";
-  } else {
-    vtxXfrAttrsPre += "\n    out.pos.z += out.pos.w;";
-  }
+  // The near/far depth correction used to be applied here per-vertex (out.pos.z = -out.pos.z for
+  // reversed, or += out.pos.w for forward), redundantly on top of the same correction already
+  // folded into ubuf.proj by effective_projection() (shader_info.cpp) - applying it twice canceled
+  // out for the common case (any draw where effective_projection() decides to flip), silently
+  // making "reversed" Z behave identically to forward Z. It is now applied exactly once, in the
+  // projection matrix alone (matching upstream aurora commit 1dde08fa: "Move depth correction to
+  // projection matrix"), so nothing needs to happen to out.pos.z here.
  // GX rasterizes at a 7/12 pixel center when antialiasing is disabled, while WebGPU rasterizes at 1/2.
  vtxXfrAttrsPre +=
      "\n    let gx_pixel_center_correction = "
@@ -1465,7 +1467,14 @@ wgpu::ShaderModule build_shader(const ShaderConfig& config) noexcept {
                    textureDependency.texMapId, uvIn);
  }

-  std::string fogDepthExpr = UseReversedZ ? "in.pos.z" : "(1.0 - in.pos.z)";
+  // in.pos.z is the host NDC z (forward: 0=near/1=far; reversed: 1=near/0=far post-fix), but this
+  // expression needs to produce GX's own native distance term (always 0=near/1=far, matching how
+  // g_gxState.clearDepth/clear_depth_value() are interpreted before their own UseReversedZ
+  // inversion) - forward already matches directly; reversed needs the same 1-x flip everything
+  // else reversed-Z-aware uses. This was backwards (verified directly against upstream aurora's
+  // identical expression in build_shader_source), which fed both fog density and the GX_ZT_ADD
+  // z-texture path the wrong distance value.
+  std::string fogDepthExpr = UseReversedZ ? "(1.0 - in.pos.z)" : "in.pos.z";
  std::string fogZCoordExpr =
      fmt::format("u32(round(clamp({}, 0.0, 1.0) * 16777216.0))", fogDepthExpr);
  if (usesZTextureDepth) {
@@ -1498,7 +1507,7 @@ wgpu::ShaderModule build_shader(const ShaderConfig& config) noexcept {
      fragmentFn += fmt::format(
          "\n    let oldZ = u32(round(clamp({0}, 0.0, 1.0) * 16777216.0));"
          "\n    ztexCoord = (ztexCoord + oldZ) & 0x00ffffffu;",
-          UseReversedZ ? "in.pos.z" : "(1.0 - in.pos.z)");
+          UseReversedZ ? "(1.0 - in.pos.z)" : "in.pos.z");
    }
    fragmentFn += "\n    let ztexDepth = f32(ztexCoord) / 16777216.0;";
    fogZCoordExpr = "ztexCoord";
@@ -1639,7 +1648,13 @@ wgpu::ShaderModule build_shader(const ShaderConfig& config) noexcept {
        "    @builtin(frag_depth) depth: f32,\n"
        "};";

-    fragmentFn += fmt::format("\n    let fragDepth = {}ztexDepth;", UseReversedZ ? "" : "1.0 - ");
+    // ztexDepth is in GX's native distance terms (0=near/1=far, see fogDepthExpr's comment above),
+    // but frag_depth must be written in the same host NDC-z convention in.pos.z itself uses -
+    // forward matches directly (no change), reversed needs the same 1-x flip. This was backwards
+    // the same way fogDepthExpr was (verified by the same derivation, since aurora upstream has no
+    // directly equivalent line here to cross-check against - this z-texture-depth-output path
+    // appears to be specific to this fork).
+    fragmentFn += fmt::format("\n    let fragDepth = {}ztexDepth;", UseReversedZ ? "1.0 - " : "");
    fragmentReturnType = "FragmentOutput";
    fragmentReturn =
        "    var out: FragmentOutput;\n"
@@ -1693,8 +1708,22 @@ fn load_u16(p: ptr<storage, array<u32>>, byte_off: u32, le: bool) -> u32 {{
  return bswap16(raw, le);
 }}

+fn load_u24_raw(p: ptr<storage, array<u32>>, byte_off: u32) -> u32 {{
+  let word_idx = byte_off >> 2u;
+  let sub = byte_off & 3u;
+  let word = p[word_idx];
+  // Three bytes at offsets zero or one fit entirely in this word. Do not
+  // access the next word: this attribute may end at the binding boundary.
+  if (sub <= 1u) {{
+    return (word >> (sub * 8u)) & 0x00FFFFFFu;
+  }}
+  let next = p[word_idx + 1u];
+  let shift = sub * 8u;
+  return ((word >> shift) | (next << (32u - shift))) & 0x00FFFFFFu;
+}}
+
 fn load_u24(p: ptr<storage, array<u32>>, byte_off: u32, le: bool) -> u32 {{
-  let raw = load_u32_raw(p, byte_off) & 0x00FFFFFFu;
+  let raw = load_u24_raw(p, byte_off);
  if (le) {{
    return raw;
  }}
@@ -1734,7 +1763,7 @@ fn raw_fetch_u8_2(p: ptr<storage, array<u32>>, byte_off: u32) -> vec2u {{
 }}

 fn raw_fetch_u8_3(p: ptr<storage, array<u32>>, byte_off: u32) -> vec3u {{
-  let raw = load_u32_raw(p, byte_off);
+  let raw = load_u24_raw(p, byte_off);
  return vec3u(
    extractBits(raw, 0u, 8u),
    extractBits(raw, 8u, 8u),
@@ -548,14 +548,22 @@ constexpr size_t kStagedUniformBytes =
    96 + sizeof(Mat4x4<float>) + sizeof(Mat3x4<float>) * (MaxPostexMtx + MaxPnMtx);

 // The host viewport always receives the normalized GX depth window (render_pass_impl clamps to minDepth <= maxDepth).
+//
+// Folds the near/far depth correction the vertex shader used to apply per-vertex directly into the
+// projection matrix instead (matching upstream aurora commit 1dde08fa, "Move depth correction to
+// projection matrix") - valid because the correction is a linear combination of the z/w rows, so
+// applying it once here to the row is equivalent to applying it once per-vertex to the dot product,
+// and it must be applied exactly once: doing it here AND in the shader (the previous bug) canceled
+// the negation out for `flip`, silently making "reversed" Z behave identically to forward Z.
+// `flip` decides which of the two single-application forms this draw needs: true bakes in the
+// reversed-Z inversion (z' = -z), false bakes in the forward-Z near/far combination (z' = z + w) -
+// exactly one always applies, never both, and never neither.
 static Mat4x4<float> effective_projection() noexcept {
  const auto& vp = g_gxState.renderViewport;
  const bool flip = (vp.znear <= vp.zfar) == UseReversedZ;
  Mat4x4<float> proj = g_gxState.proj;
-  if (flip) {
-    for (size_t i = 0; i < 4; ++i) {
-      proj.m2.m[i] = -(proj.m2.m[i] + proj.m3.m[i]);
-    }
+  for (size_t i = 0; i < 4; ++i) {
+    proj.m2.m[i] = flip ? -proj.m2.m[i] : (proj.m2.m[i] + proj.m3.m[i]);
  }
  return proj;
 }
@@ -75,18 +75,30 @@ void initialize() noexcept {

 void shutdown() noexcept {
  ZoneScoped;
-  if (g_useSdlRenderer) {
-    ImGui_ImplSDLRenderer3_Shutdown();
-  } else {
-    ImGui_ImplWGPU_Shutdown();
+  // Startup can fail before either backend initializes. A context alone does
+  // not mean its renderer/platform backend owns resources to release.
+  if (ImGui::GetCurrentContext() != nullptr) {
+    ImGuiIO& io = ImGui::GetIO();
+    if (io.BackendRendererUserData != nullptr) {
+      if (g_useSdlRenderer) {
+        ImGui_ImplSDLRenderer3_Shutdown();
+      } else {
+        ImGui_ImplWGPU_Shutdown();
+      }
+    }
+    if (io.BackendPlatformUserData != nullptr) {
+      ImGui_ImplSDL3_Shutdown();
+    }
+    ImGui::DestroyContext();
  }
-  ImGui_ImplSDL3_Shutdown();
-  ImGui::DestroyContext();
  for (const auto& texture : g_sdlTextures) {
    SDL_DestroyTexture(texture);
  }
  g_sdlTextures.clear();
  g_wgpuTextures.clear();
+  g_useSdlRenderer = false;
+  g_scale = 0.f;
+  g_frameDataBuilt = false;
 }

 void process_event(const SDL_Event& event) noexcept {
@@ -401,6 +401,8 @@ SDL_JoystickID add_controller(SDL_JoystickID which) noexcept {
      return -1;
    }
    controller.m_isGameCube = controller.m_vid == 0x057E && controller.m_pid == 0x0337;
+    const char* serial = SDL_GetGamepadSerial(ctrl);
+    controller.m_gameCubeUseOrdinaryStop = controller.m_isGameCube && serial && "GCP+"sv == serial;
    if (controller.m_isGameCube ||
        (SDL_GetGamepadType(ctrl) == SDL_GAMEPAD_TYPE_NINTENDO_SWITCH_PRO && controller.m_pid == 0x2073)) {
      controller.m_deadZones.emulateTriggers = false;
@@ -481,6 +483,13 @@ bool controller_has_rumble(Uint32 instance) noexcept {
 void controller_rumble(uint32_t instance, uint16_t low_freq_intensity, uint16_t high_freq_intensity,
                       uint16_t duration_ms) noexcept {
  if (auto it = g_GameControllers.find(instance); it != g_GameControllers.end()) {
+    // GC Pocket+ has been observed continuing to vibrate after a hard stop;
+    // an ordinary stop cleared it. With GAMECUBE_RUMBLE_BRAKE enabled, SDL
+    // encodes (0, 1) as adapter command 0, whereas (0, 0) sends command 2.
+    // Apply the workaround here so shutdown uses the same stop as PAD calls.
+    if (it->second.m_gameCubeUseOrdinaryStop && low_freq_intensity == 0 && high_freq_intensity == 0) {
+      high_freq_intensity = 1;
+    }
    SDL_RumbleGamepad(it->second.m_controller, low_freq_intensity, high_freq_intensity, duration_ms);
  }
 }
@@ -17,6 +17,7 @@ extern Module Log;
 struct GameController {
  SDL_Gamepad* m_controller = nullptr;
  bool m_isGameCube = false;
+  bool m_gameCubeUseOrdinaryStop = false;
  Sint32 m_index = -1;
  Sint32 m_playerIndex = -1;
  bool m_hasRumble = false;
@@ -122,7 +122,10 @@ auto underlying(T value) -> std::underlying_type_t<T> {

 #define UNIMPLEMENTED() FATAL("UNIMPLEMENTED: {}", __FUNCTION__)

+namespace wgpu { class CommandBuffer; }
+
 namespace aurora {
+void submit_staging_commands(const wgpu::CommandBuffer& commands);
 extern AuroraConfig g_config;
 extern uint32_t g_sdlCustomEventsStart;
 extern char g_gameName[4];
@@ -570,12 +570,16 @@ bool initialize(AuroraBackend auroraBackend) {
            g_adapter = std::move(adapter);
          } else {
            Log.warn("Adapter request failed: {}", message);
+            const std::string_view reason{message};
+            SDL_SetError("Graphics adapter unavailable: %.*s",
+                         static_cast<int>(std::min<size_t>(reason.size(), 512)), reason.data());
          }
        });
    const auto status = g_instance.WaitAny(future, 5000000000);
    if (status != wgpu::WaitStatus::Success) {
      Log.error("Failed to create {} adapter: {}", magic_enum::enum_name(backend),
                magic_enum::enum_name(status));
+      SDL_SetError("Graphics adapter request did not complete within its startup deadline");
      return false;
    }
    if (!g_adapter) {
@@ -738,11 +742,15 @@ bool initialize(AuroraBackend auroraBackend) {
                                    g_device = std::move(device);
                                  } else {
                                    Log.warn("Device request failed: {}", message);
+                                    const std::string_view reason{message};
+                                    SDL_SetError("Graphics device unavailable: %.*s",
+                                        static_cast<int>(std::min<size_t>(reason.size(), 512)), reason.data());
                                  }
                                });
    const auto status = g_instance.WaitAny(future, 5000000000);
    if (status != wgpu::WaitStatus::Success) {
      Log.error("Failed to create device: {}", magic_enum::enum_name(status));
+      SDL_SetError("Graphics device request did not complete within its startup deadline");
      return false;
    }
    if (!g_device) {
@@ -2,6 +2,7 @@
 #include <cstring>
 #include <ctime>
 #include <mutex>
+#include <limits>
 #include <string>
 #include <filesystem>
 #include <vector>
@@ -286,8 +287,33 @@ size_t load_from_cache(void const* key, size_t keySize, void* value, size_t valu
  if (ret == SQLITE_ROW) {
    // Hit
    const auto foundPtr = sqlite3_column_blob(load_stmt, 0);
-    foundSize = sqlite3_column_int64(load_stmt, 1);
-    const bool compressed = sqlite3_column_int(load_stmt, 2) != 0;
+    const auto declaredSize = sqlite3_column_int64(load_stmt, 1);
+    const auto storedSize = sqlite3_column_bytes(load_stmt, 0);
+    const auto compression = sqlite3_column_int(load_stmt, 2);
+    const bool compressed = compression == 1;
+    // Dawn asks for the size before allocating its destination. Validate here,
+    // not only during the copy: corrupt metadata must become a cache miss.
+    bool valid = declaredSize > 0 &&
+                 static_cast<uint64_t>(declaredSize) <= std::numeric_limits<size_t>::max() &&
+                 foundPtr != nullptr && storedSize > 0 && (compression == 0 || compression == 1);
+    if (valid && compressed) {
+#if defined(AURORA_CACHE_USE_ZSTD)
+      // Our writer uses ZSTD_compress, which records the original content size.
+      const auto frameSize = ZSTD_getFrameContentSize(foundPtr, static_cast<size_t>(storedSize));
+      valid = frameSize != ZSTD_CONTENTSIZE_ERROR && frameSize != ZSTD_CONTENTSIZE_UNKNOWN &&
+              frameSize == static_cast<uint64_t>(declaredSize);
+#else
+      valid = false;
+#endif
+    } else if (valid) {
+      valid = declaredSize == storedSize;
+    }
+    if (!valid) {
+      Log.error("Ignoring cache entry with inconsistent size or compression metadata");
+      check(sqlite3_reset(load_stmt));
+      return 0;
+    }
+    foundSize = static_cast<size_t>(declaredSize);
    if (value == nullptr) {
      g_hits.fetch_add(1, std::memory_order_relaxed);
    } else {
@@ -11,9 +11,6 @@
 #include <aurora/event.h>
 #include <aurora/gfx.h>
 #include <aurora/render_size_limits.hpp>
-#if defined(__APPLE__)
-#include <TargetConditionals.h>
-#endif
 #include <SDL3/SDL_error.h>
 #include <SDL3/SDL_events.h>
 #include <SDL3/SDL_keyboard.h>
@@ -411,21 +408,8 @@ bool create_window(AuroraBackend backend) {
    height = 480;
  }

-  Sint32 posX = g_config.hasWindowPosition ? g_config.windowPosX : SDL_WINDOWPOS_CENTERED;
-  Sint32 posY = g_config.hasWindowPosition ? g_config.windowPosY : SDL_WINDOWPOS_CENTERED;
-#if defined(__APPLE__) && TARGET_OS_IPHONE
-  // Without this the window falls back to the 1280x960 default on any device
-  // with no saved size.
-  if (const SDL_DisplayID primary = SDL_GetPrimaryDisplay(); primary != 0) {
-    SDL_Rect bounds{};
-    if (SDL_GetDisplayBounds(primary, &bounds) && bounds.w > 0 && bounds.h > 0) {
-      width = bounds.w;
-      height = bounds.h;
-    }
-    posX = SDL_WINDOWPOS_CENTERED_DISPLAY(primary);
-    posY = SDL_WINDOWPOS_CENTERED_DISPLAY(primary);
-  }
-#endif
+  const Sint32 posX = g_config.hasWindowPosition ? g_config.windowPosX : SDL_WINDOWPOS_CENTERED;
+  const Sint32 posY = g_config.hasWindowPosition ? g_config.windowPosY : SDL_WINDOWPOS_CENTERED;

  const auto props = SDL_CreateProperties();
  TRY(SDL_SetStringProperty(props, SDL_PROP_WINDOW_CREATE_TITLE_STRING, g_config.appName), "Failed to set {}: {}",
@@ -18,6 +18,7 @@ if (AURORA_ENABLE_GX)
    gx_fifo_test.cpp
    gx_test_stubs.cpp
    texture_bind_group_cache_key_test.cpp
+    renderer_regression_test.cpp
    ../lib/gfx/efb_ram_encoder.cpp
    # GX API implementations (encoders)
    ../lib/dolphin/gx/GXBump.cpp
@@ -66,6 +67,18 @@ if (AURORA_ENABLE_GX)
  )

  gtest_discover_tests(gx_fifo_tests)
+
+  option(AURORA_BUILD_GPU_TESTS "Build renderer pixel tests requiring a graphics device" OFF)
+  if (AURORA_BUILD_GPU_TESTS)
+    add_executable(gx_readback_tests gpu_readback_test.cpp)
+    target_compile_features(gx_readback_tests PRIVATE cxx_std_20)
+    target_include_directories(gx_readback_tests PRIVATE ../lib)
+    target_link_libraries(gx_readback_tests PRIVATE
+      aurora::core aurora::gx aurora::pad aurora::vi aurora::mtx aurora::si
+      dawn::dawncpp_headers absl::flat_hash_map absl::btree TracyClient)
+    add_test(NAME gx_readback_tests COMMAND gx_readback_tests "${CMAKE_CURRENT_BINARY_DIR}/readback-cache")
+    set_tests_properties(gx_readback_tests PROPERTIES TIMEOUT 90)
+  endif ()
 endif () # AURORA_ENABLE_GX

 # DVD API tests
@@ -0,0 +1,363 @@
+// ROM-free integration probe. Links the actual maintained Aurora renderer.
+#include "gfx/common.hpp"
+#include "gfx/clear.hpp"
+#include "gfx/efb_ram_copy.hpp"
+#include "gfx/pipeline_cache.hpp"
+#include "gfx/texture.hpp"
+#include "gx/gx.hpp"
+#include "gx/fifo.hpp"
+#include "gx/command_processor.hpp"
+#include "gx/frame_interpolation.hpp"
+#include <dolphin/gx.h>
+#include <aurora/aurora.h>
+#include <array>
+#include <bit>
+#include <atomic>
+#include <chrono>
+#include <cstdio>
+#include <filesystem>
+#include <stdexcept>
+#include <thread>
+#include <vector>
+
+namespace {
+using namespace aurora;
+std::atomic<unsigned> errors{};
+std::atomic<unsigned> guestWrites{};
+// Keep destinations alive through shutdown, including any failing wait.
+std::array<uint8_t, 16 * 16 * 4 + 32> guarded;
+std::array<uint8_t, 16 * 16 * 4 + 32> guardedBake;
+void require(bool value, const char* message) {
+  if (!value) throw std::runtime_error(message);
+}
+void submit(bool final, bool download = false, bool async = false) {
+  auto encoder = webgpu::g_device.CreateCommandEncoder();
+  if (final) gfx::end_frame(encoder); else gfx::end_batch(encoder);
+  gfx::render(encoder);
+  if (download) gfx::efb_ram::encode_downloads(encoder);
+  if (async) gfx::efb_ram::encode_async_downloads(encoder);
+  auto commands = encoder.Finish();
+  webgpu::g_queue.Submit(1, &commands);
+  if (download) require(gfx::efb_ram::complete_downloads(), "EFB readback failed");
+  gfx::after_submit();
+  if (!final) require(gfx::resume_frame(), "Batch resume failed");
+}
+
+constexpr std::array<std::array<uint8_t, 4>, 4> colors{{
+    {255, 0, 0, 255}, {0, 255, 0, 255}, {0, 0, 255, 255}, {255, 255, 0, 255}}};
+using Pixels = std::vector<uint8_t>;
+Pixels expected(unsigned extent) {
+  Pixels bytes(extent * extent * 4);
+  // GX RGBA8: 4x4 tiles, sixteen A/R pairs followed by sixteen G/B pairs.
+  for (unsigned y = 0; y < extent; ++y) for (unsigned x = 0; x < extent; ++x) {
+    const auto color = colors[y * 4 / extent];
+    const auto tile = ((y / 4) * (extent / 4) + x / 4) * 64;
+    const auto pair = ((y % 4) * 4 + x % 4) * 2;
+    bytes[tile + pair] = color[3]; bytes[tile + pair + 1] = color[0];
+    bytes[tile + 32 + pair] = color[1]; bytes[tile + 33 + pair] = color[2];
+  }
+  return bytes;
+}
+
+Pixels run(unsigned splitEvery, bool async = false, bool offscreen = false, unsigned geometry = 0, bool capacityStress = false, bool interpolate = false, bool frameWorker = false, unsigned copyCase = 0) {
+  require(!async || !offscreen, "Combined probe mode is not supported");
+  guardedBake.fill(0xa5);
+  gx::g_gxState.clearColor = {0.f, 0.f, 0.f, 1.f};
+  require(frameWorker ? aurora_begin_frame() : gfx::begin_frame(), "Frame begin failed");
+  std::array<std::array<float, 3>, 4> positions{};
+  if (geometry) {
+    alignas(32) static std::array<uint8_t, 32768> fifo;
+    GXInit(fifo.data(), fifo.size());
+    gx::g_gxState.viewportPolicy = AURORA_VIEWPORT_NATIVE;
+    GXSetViewport(0.f, 0.f, 64.f, 64.f, 0.f, 1.f);
+    GXSetScissor(0, 0, 64, 64);
+    GXSetCullMode(GX_CULL_NONE);
+    GXSetZMode(false, GX_ALWAYS, false);
+    GXSetBlendMode(GX_BM_NONE, GX_BL_ONE, GX_BL_ZERO, GX_LO_COPY);
+    GXSetColorUpdate(true); GXSetAlphaUpdate(true);
+    GXSetNumTexGens(0); GXSetNumChans(1); GXSetNumTevStages(copyCase ? copyCase : 1);
+    for (unsigned stage = 1; stage < copyCase; ++stage) {
+      GXSetTevOrder(static_cast<GXTevStageID>(stage), GX_TEXCOORD_NULL, GX_TEXMAP_NULL, GX_COLOR0A0);
+      GXSetTevOp(static_cast<GXTevStageID>(stage), GX_PASSCLR);
+    }
+    GXSetTevOrder(GX_TEVSTAGE0, GX_TEXCOORD_NULL, GX_TEXMAP_NULL, GX_COLOR0A0);
+    GXSetTevOp(GX_TEVSTAGE0, GX_PASSCLR);
+    GXSetChanCtrl(GX_COLOR0A0, false, GX_SRC_REG, GX_SRC_VTX, GX_LIGHT_NULL, GX_DF_NONE, GX_AF_NONE);
+    const float projection[]{interpolate ? 0.f : 1.f, 1.f, 0.f, 1.f, 0.f, 0.f, -0.5f};
+    GXSetProjectionv(projection);
+    GXClearVtxDesc();
+    GXSetVtxDesc(GX_VA_POS, geometry == 2 ? GX_INDEX8 : GX_DIRECT);
+    GXSetVtxDesc(GX_VA_CLR0, GX_DIRECT);
+    if (geometry == 2) GXSetArray(GX_VA_POS, positions.data(), sizeof(positions), sizeof(positions[0]), true);
+    GXSetVtxAttrFmt(GX_VTXFMT0, GX_VA_POS, GX_POS_XYZ, GX_F32, 0);
+    GXSetVtxAttrFmt(GX_VTXFMT0, GX_VA_CLR0, GX_CLR_RGBA, GX_RGBA8, 0);
+    gx::fifo::drain();
+  }
+  const auto pipeline = gfx::pipeline_ref(gfx::clear::PipelineConfig{});
+  for (unsigned band = 0; band < 4; ++band) {
+    const auto c = colors[band];
+    if (geometry) {
+      const float top = 1.f - band * 0.5f;
+      const float bottom = top - 0.5f;
+      const float z = interpolate ? -1.f : 0.f;
+      positions = {{{-1.f, top, z}, {1.f, top, z}, {1.f, bottom, z}, {-1.f, bottom, z}}};
+      // Keep the same array address/format and change only its bytes between draws.
+      if (geometry == 2) GXInvalidateVtxCache();
+      if (geometry == 3) {
+        std::array<uint8_t, 64> raw{};
+        for (unsigned index = 0; index < positions.size(); ++index) {
+          for (unsigned axis = 0; axis < 3; ++axis) {
+            const auto bits = std::bit_cast<uint32_t>(positions[index][axis]);
+            for (unsigned byte = 0; byte < 4; ++byte)
+              raw[index * 16 + axis * 4 + byte] = bits >> (24 - byte * 8);
+          }
+          std::copy(c.begin(), c.end(), raw.begin() + index * 16 + 12);
+        }
+        require(gx::fifo::submit_raw_draw(GX_QUADS, GX_VTXFMT0, raw.data(), 4, raw.size()),
+                "Raw bridge rejected valid quad");
+      } else {
+      GXBegin(GX_QUADS, GX_VTXFMT0, 4);
+      for (unsigned index = 0; index < positions.size(); ++index) {
+        if (geometry == 2) GXPosition1x8(index);
+        else GXPosition3f32(positions[index][0], positions[index][1], positions[index][2]);
+        GXColor4u8(c[0], c[1], c[2], 255);
+      }
+      GXEnd();
+      }
+    } else {
+    gfx::push_draw_command(gfx::clear::DrawData{
+        .pipeline = pipeline,
+        .color = {c[0] / 255., c[1] / 255., c[2] / 255., 1.},
+        .depth = 0.5f,
+        .useScissor = true,
+        .scissor = {0, static_cast<int32_t>(band * 16), 64, 16}});
+    }
+    if (offscreen && band == 0) {
+      // Suspend a partially recorded EFB, bake an independently observable copy,
+      // then resume it before a possible capacity-boundary submission.
+      gfx::begin_offscreen(64, 64);
+      gfx::push_draw_command(gfx::clear::DrawData{
+          .pipeline = pipeline, .color = {1., 0., 1., 1.}, .depth = 0.25f});
+      if (capacityStress) for (unsigned draw = 0; draw < 24; ++draw) {
+        gfx::push_draw_command(gfx::clear::DrawData{
+            .pipeline = pipeline, .color = {1., 0., 1., 1.}, .depth = 0.25f,
+            .useScissor = true, .scissor = {0, 0, 4, 4}});
+      }
+      auto baked = gfx::new_render_texture(64, 64, GX_TF_RGBA8, "Aurora probe offscreen bake");
+      gfx::resolve_pass(baked, {0, 0, 64, 64}, false, false, false,
+                        {0.f, 0.f, 0.f, 1.f}, 1.f, GX_TF_RGBA8, nullptr, false,
+                        nullptr, false, 1.f, false, false, true);
+      gfx::efb_ram::schedule(guardedBake.data() + 16, 16, 16, GX_TF_RGBA8, baked);
+      gfx::end_offscreen();
+      if (capacityStress) {
+        require(gfx::efb_ram::prepare_downloads(), "Early bake readback preparation failed");
+        for (unsigned draw = 0; draw < 24; ++draw) {
+          gfx::push_draw_command(gfx::clear::DrawData{
+              .pipeline = pipeline, .color = {1., 0., 0., 1.}, .depth = 0.5f,
+              .useScissor = true, .scissor = {0, 0, 4, 4}});
+        }
+      }
+    }
+    if (splitEvery && band < 3 && (band + 1) % splitEvery == 0) submit(false);
+  }
+  auto texture = gfx::new_render_texture(64, 64, GX_TF_RGBA8, "Aurora probe persistent copy");
+  static std::array<uint8_t, 64 * 64 * 4> copyDestination;
+  if (copyCase) {
+    gx::fifo::drain();
+    GXSetTexCopySrc(0, 0, 64, 64);
+    GXSetTexCopyDst(64, 64, GX_TF_RGBA8, GX_FALSE);
+    GXCopyTex(copyDestination.data(), GX_FALSE);
+    texture = gx::g_gxState.copyTextures.at(copyDestination.data()).handle;
+  } else {
+  // A partial clear forces the real snapshot and clear-uniform paths after the copy.
+  gfx::resolve_pass(texture, {0, 0, 64, 64}, true, true, true, {0.f, 0.f, 0.f, 1.f},
+                    1.f, GX_TF_RGBA8, nullptr, false, nullptr, false, 1.f, false, false, true);
+  }
+  guarded.fill(0xa5);
+  const unsigned extent = async ? 4 : 16;
+  const unsigned bytes = extent * extent * 4;
+  gfx::efb_ram::schedule(guarded.data() + 16, extent, extent, GX_TF_RGBA8, texture);
+  const auto before = guestWrites.load(std::memory_order_acquire);
+  if (async) gfx::efb_ram::seal_async_downloads();
+  else require(gfx::efb_ram::prepare_downloads(), "Readback preparation failed");
+  if (frameWorker) {
+    require(!async && aurora_flush_efb_copies_to_ram(), "Worker-mode EFB readback failed");
+    aurora_end_frame();
+  } else submit(true, !async, async);
+  if (async) {
+    const auto deadline = std::chrono::steady_clock::now() + std::chrono::seconds(5);
+    while (guestWrites.load(std::memory_order_acquire) == before) {
+      webgpu::g_instance.ProcessEvents();
+      require(std::chrono::steady_clock::now() < deadline, "Async readback did not complete");
+      std::this_thread::sleep_for(std::chrono::milliseconds(1));
+    }
+  }
+  require(std::all_of(guarded.begin(), guarded.begin() + 16, [](auto b) { return b == 0xa5; }) &&
+          std::all_of(guarded.begin() + 16 + bytes, guarded.end(), [](auto b) { return b == 0xa5; }),
+          "Readback wrote outside its destination");
+  if (offscreen) {
+    require(std::all_of(guardedBake.begin(), guardedBake.begin() + 16, [](auto b) { return b == 0xa5; }) &&
+            std::all_of(guardedBake.end() - 16, guardedBake.end(), [](auto b) { return b == 0xa5; }),
+            "Offscreen readback wrote outside its destination");
+    for (unsigned tile = 0; tile < 16; ++tile) for (unsigned pair = 0; pair < 16; ++pair) {
+      const auto offset = 16 + tile * 64 + pair * 2;
+      require(guardedBake[offset] == 255 && guardedBake[offset + 1] == 255 &&
+              guardedBake[offset + 32] == 0 && guardedBake[offset + 33] == 255,
+              "Offscreen bake did not preserve expected magenta pixels");
+    }
+  }
+  Pixels pixels(guarded.begin() + 16, guarded.begin() + 16 + bytes);
+  return pixels;
+}
+} // namespace
+
+int main(int argc, char** argv) {
+  if (argc != 2) return 2;
+  std::filesystem::create_directories(argv[1]);
+  AuroraConfig config{};
+  config.appName = "Aurora readback regression tests";
+  config.userPath = argv[1];
+  config.cachePath = argv[1];
+  config.resourcesPath = argv[1];
+  config.desiredBackend = BACKEND_AUTO;
+  config.windowWidth = 64;
+  config.windowHeight = 64;
+  config.msaa = 1;
+  config.maxTextureAnisotropy = 1;
+  config.logLevel = LOG_INFO;
+  config.logCallback = [](AuroraLogLevel level, const char* module, const char* message, unsigned size) {
+    if (level >= LOG_ERROR) ++errors;
+    std::fprintf(stderr, "[%s] %.*s\n", module, static_cast<int>(size), message);
+  };
+  const auto initialized = aurora_initialize(1, argv, &config);
+  if (initialized.initializationStatus != AURORA_INITIALIZATION_SUCCESS) return 3;
+  aurora_set_skip_unready_pipelines(true);
+  aurora_set_guest_write_hooks(nullptr, [](const void*, size_t) {
+    guestWrites.fetch_add(1, std::memory_order_release);
+  });
+  try {
+    const auto prewarmQueued = gfx::queued_pipeline_count();
+    const auto prewarmDeadline = std::chrono::steady_clock::now() + std::chrono::seconds(60);
+    while (gfx::queued_pipeline_count() != 0) {
+      require(std::chrono::steady_clock::now() < prewarmDeadline, "Seeded pipeline prewarm did not finish");
+      webgpu::g_instance.ProcessEvents();
+      std::this_thread::sleep_for(std::chrono::milliseconds(1));
+    }
+    std::printf("Actual Aurora completed seeded startup queue (%u observed pending)\n", prewarmQueued);
+    const auto control = run(0);
+    for (unsigned band = 0; band < 4; ++band) {
+      const auto tile = band * 4 * 64;
+      std::fprintf(stderr, "band=%u ARGB=%u,%u,%u,%u\n", band, control[tile],
+                   control[tile + 1], control[tile + 32], control[tile + 33]);
+    }
+    require(control == expected(16), "Unsplit pixels differ from independently expected GX data");
+    for (unsigned iteration = 0; iteration < 9; ++iteration) {
+      const auto splitEvery = iteration % 3 + 1;
+      require(run(splitEvery) == control, "Split pixels differ from unsplit control");
+      std::printf("Actual Aurora split=%u iteration=%u matched native tiled readback\n", splitEvery, iteration);
+    }
+    for (unsigned iteration = 0; iteration < 9; ++iteration) {
+      require(run(iteration % 3 + 1, true) == expected(4), "Async pixels differ from expected GX data");
+      std::printf("Actual Aurora async iteration=%u matched native tiled readback\n", iteration);
+    }
+    for (unsigned splitEvery = 0; splitEvery < 4; ++splitEvery) {
+      require(run(splitEvery, false, true) == expected(16), "Offscreen interlude changed the suspended EFB");
+      std::printf("Actual Aurora offscreen split=%u preserved bake and suspended EFB\n", splitEvery);
+    }
+    for (unsigned splitEvery = 0; splitEvery < 4; ++splitEvery) {
+      require(run(splitEvery, false, false, true) == expected(16), "GX FIFO quad pixels differ from expected output");
+      std::printf("Actual Aurora GX FIFO split=%u preserved direct vertices, indices and uniforms\n", splitEvery);
+    }
+    for (unsigned splitEvery = 0; splitEvery < 4; ++splitEvery) {
+      require(run(splitEvery, false, false, 2) == expected(16), "Invalidated GX array pixels differ from expected output");
+      std::printf("Actual Aurora GX invalidation split=%u refreshed the same array address\n", splitEvery);
+    }
+    const gfx::StagingSizes physical{gfx::VertexBufferSize, gfx::UniformBufferSize,
+                                      gfx::IndexBufferSize, gfx::StorageBufferSize};
+    const auto uniformTail = gx::MaxUniformSize + 32 * gfx::staging_uniform_bytes(48);
+    for (unsigned buffer = 0; buffer < 4; ++buffer) {
+      auto limits = physical;
+      limits[buffer] = buffer == 0 ? 128 : buffer == 1 ? uniformTail + 512 :
+                       buffer == 2 ? 24 : 2 * gfx::staging_storage_bytes(48);
+      gfx::set_staging_capacity_limits_for_testing(limits);
+      const auto before = gfx::staging_split_count();
+      require(run(0, false, false, buffer == 1 ? 0 : buffer == 3 ? 2 : 1) == expected(16),
+              "Automatic capacity split changed pixels");
+      require(gfx::staging_split_count() > before, "Forced capacity did not split");
+      const auto highWater = gfx::staging_high_water();
+      for (unsigned i = 0; i < limits.size(); ++i)
+        require(highWater[i] <= limits[i], "Actual staging usage exceeded admission budget");
+      std::printf("Actual staging high-water V/U/I/S=%llu/%llu/%llu/%llu bytes\n",
+          static_cast<unsigned long long>(highWater[0]), static_cast<unsigned long long>(highWater[1]),
+          static_cast<unsigned long long>(highWater[2]), static_cast<unsigned long long>(highWater[3]));
+      std::printf("Actual Aurora automatic capacity buffer=%u splits=%llu matched pixels\n", buffer,
+                  static_cast<unsigned long long>(gfx::staging_split_count() - before));
+    }
+    auto limits = physical;
+    limits[0] = 128;
+    gfx::set_staging_capacity_limits_for_testing(limits);
+    require(run(0, false, false, 3) == expected(16), "Raw bridge capacity split changed pixels");
+    std::puts("Actual Aurora raw bridge capacity split preserved direct quad pixels");
+    limits = physical;
+    limits[1] = uniformTail + 768;
+    gfx::set_staging_capacity_limits_for_testing(limits);
+    const auto beforeBake = gfx::staging_split_count();
+    require(run(0, false, true, 0, true) == expected(16),
+            "Automatic offscreen split changed bake or suspended EFB");
+    require(gfx::staging_split_count() - beforeBake >= 9, "Offscreen test did not reuse all staging slots");
+    std::printf("Actual Aurora automatic offscreen/readback splits=%llu preserved all pixels\n",
+                static_cast<unsigned long long>(gfx::staging_split_count() - beforeBake));
+    gfx::set_staging_capacity_limits_for_testing(physical);
+    aurora_set_frame_interpolation_fps(120);
+    for (unsigned frame = 0; frame < 3; ++frame)
+      require(run(0, false, false, 2, false, true) == expected(16), "Perspective warmup changed pixels");
+    AuroraFrameInterpolationDiagnostics interpolation{};
+    gx::get_frame_interpolation_diagnostics(interpolation);
+    require(interpolation.matchable > 0 && interpolation.activeSamples > 0,
+            "Interpolation probe did not establish matching perspective draws");
+    limits = physical;
+    limits[3] = 2 * gfx::staging_storage_bytes(48);
+    gfx::set_staging_capacity_limits_for_testing(limits);
+    require(run(0, false, false, 2, false, true) == expected(16), "Interpolated split changed native pixels");
+    gx::get_frame_interpolation_diagnostics(interpolation);
+    require(!interpolation.replaySafe, "Split frame incorrectly retained interpolation replay");
+    aurora_set_frame_interpolation_fps(0);
+    std::puts("Actual Aurora matched perspective interpolation survived capacity split and disabled replay");
+    limits = physical;
+    limits[0] = 32; // One quad needs 64 bytes: typed rejection before any draw allocation.
+    gfx::set_staging_capacity_limits_for_testing(limits);
+    bool oversized = false;
+    try { run(0, false, false, 1); }
+    catch (const gfx::StagingCapacityError&) { oversized = true; }
+    require(oversized, "Oversized primitive was not rejected");
+    require(gfx::staging_usage() == gfx::StagingSizes{}, "Oversized primitive partially allocated");
+    gx::fifo::clear_buffer();
+    gfx::abort_frame();
+    gfx::set_staging_capacity_limits_for_testing(physical);
+    std::puts("Actual Aurora oversized primitive rejected before staging mutation");
+    require(run(0, false, false, 1) == expected(16), "Renderer failed after rejected primitive cleanup");
+    limits = physical;
+    limits[3] = 2 * gfx::staging_storage_bytes(48);
+    gfx::set_staging_capacity_limits_for_testing(limits);
+    aurora_set_frame_interpolation_fps(120);
+    for (unsigned frame = 0; frame < 16; ++frame)
+      require(run(0, false, false, 2, false, true, true) == expected(16),
+              "Frame-worker capacity split changed pixels");
+    // Grant preparation of the next frame before joining DONE, exactly as the
+    // real producer does; leave no worker waiting for a future begin_frame.
+    require(aurora_begin_frame(), "Final worker frame preparation failed");
+    aurora::wait_for_frame_worker();
+    gfx::abort_frame();
+    aurora_set_frame_interpolation_fps(0);
+    gfx::set_staging_capacity_limits_for_testing(physical);
+    std::puts("Actual Aurora frame worker completed 16 capacity-split perspective frames");
+    require(errors == 0, "Renderer reported an error");
+  } catch (const std::exception& error) {
+    std::fprintf(stderr, "FAIL: %s\n", error.what());
+    aurora_shutdown();
+    return 4;
+  }
+  aurora_shutdown();
+  if (errors != 0) return 4;
+  std::puts("Actual Aurora clear/resolve/snapshot/downsample/readback batches passed");
+}
@@ -462,8 +462,18 @@ TEST(FrameInterpolationContract, IndexedPaletteHistoryKeepsAbsoluteVertexSlots)
  std::array<uint8_t, uniformSize> changedSource{};
  aurora::gx::begin_frame_interpolation();
  const auto changedRanges = recordFrame(changedTopology, 91.0f, 9.0f, changedSource);
-  EXPECT_EQ(changedRanges[0].size, 0u);
+  // Staging may reserve a copy for sibling matching; the correctness contract
+  // is that an unmatched topology receives the current pose unchanged.
+  const auto expectedCurrent = changedSource;
  aurora::gx::finalize_frame_interpolation();
+  EXPECT_EQ(changedSource, expectedCurrent);
+  if (changedRanges[0].size != 0) {
+    // No replacement range also correctly selects the original current uniform.
+    ASSERT_EQ(changedRanges[0].size, uniformSize);
+    const auto& duplicated = aurora::gfx::testing::uniform_allocation(changedRanges[0].offset);
+    ASSERT_EQ(duplicated.size(), expectedCurrent.size());
+    EXPECT_EQ(std::memcmp(duplicated.data(), expectedCurrent.data(), expectedCurrent.size()), 0);
+  }

  aurora::gx::set_frame_interpolation_fps(0);
  aurora::gx::begin_frame_interpolation();
@@ -646,12 +656,34 @@ TEST(TevRegisterLivenessContract, PacksOneUniformWhenBothHalvesNeedInitialValue)
  auto config = baseline;
  config.tevStages[0].colorPass.a = GX_CC_C0;
  config.tevStages[0].alphaPass.a = GX_CA_A0;
+  config.tevStages[0].colorPass.b = GX_CC_KONST;
+  config.tevStages[0].kcSel = GX_TEV_KCSEL_K0;

-  const auto baselineInfo = aurora::gx::build_shader_info(baseline);
  const auto info = aurora::gx::build_shader_info(config);
  EXPECT_TRUE(info.loadsTevRegRgb.test(GX_TEVREG0));
  EXPECT_TRUE(info.loadsTevRegAlpha.test(GX_TEVREG0));
-  EXPECT_EQ(info.uniformSize, baselineInfo.uniformSize + sizeof(aurora::Vec4<float>));
+  // The final allocation is alignment-rounded, so adding one register need
+  // not increase it. Verify actual packing with a distinct following K color.
+  const auto savedReg = g_gxState.colorRegs[GX_TEVREG0];
+  const auto savedKColor = g_gxState.kcolors[GX_KCOLOR0];
+  g_gxState.colorRegs[GX_TEVREG0] = {11.f, 22.f, 33.f, 44.f};
+  g_gxState.kcolors[GX_KCOLOR0] = {55.f, 66.f, 77.f, 88.f};
+  EXPECT_TRUE(info.sampledKColors.test(GX_KCOLOR0));
+  aurora::gfx::testing::reset_uniform_allocations();
+  aurora::gx::build_uniform(info, 0, {}, {}, false);
+  const auto expectedReg = g_gxState.colorRegs[GX_TEVREG0];
+  const auto expectedKColor = g_gxState.kcolors[GX_KCOLOR0];
+  g_gxState.colorRegs[GX_TEVREG0] = savedReg;
+  g_gxState.kcolors[GX_KCOLOR0] = savedKColor;
+  const auto& bytes = aurora::gfx::testing::uniform_allocation(0);
+  const auto* reg = reinterpret_cast<const uint8_t*>(&expectedReg);
+  const auto found = std::search(bytes.begin(), bytes.end(), reg, reg + sizeof(aurora::Vec4<float>));
+  ASSERT_NE(found, bytes.end());
+  const size_t offset = static_cast<size_t>(found - bytes.begin());
+  ASSERT_LE(offset + 2 * sizeof(aurora::Vec4<float>), bytes.size());
+  EXPECT_EQ(std::memcmp(bytes.data() + offset + sizeof(aurora::Vec4<float>),
+                        &expectedKColor, sizeof(aurora::Vec4<float>)), 0);
+  aurora::gfx::testing::reset_uniform_allocations();
 }

 // BP registers (direct FIFO writes, no dirty state flush needed)
@@ -708,6 +740,52 @@ TEST_F(GXFifoTest, BlendMode_Logic) {
  EXPECT_EQ(g_gxState.blendOp, GX_LO_XOR);
 }

+
+TEST_F(GXFifoTest, GenMode_FirstZeroWriteDecodesAndRepeatDeduplicates) {
+  reset_gx_state();
+  const auto before = g_gxState.pipelineStateGeneration;
+  decode_fifo(bp_cmd(0, 0));
+  EXPECT_EQ(g_gxState.numTevStages, 1u);
+  EXPECT_EQ(g_gxState.cullMode, GX_CULL_NONE);
+  EXPECT_EQ(g_gxState.numChans, 0u);
+  EXPECT_EQ(g_gxState.numTexGens, 0u);
+  EXPECT_EQ(g_gxState.numIndStages, 0u);
+  EXPECT_EQ(g_gxState.bpRegCache[0], 0u);
+  EXPECT_NE(g_gxState.pipelineStateGeneration, before);
+  const auto decoded = g_gxState.pipelineStateGeneration;
+  decode_fifo(bp_cmd(0, 0));
+  EXPECT_EQ(g_gxState.pipelineStateGeneration, decoded);
+}
+
+TEST_F(GXFifoTest, GenMode_FirstMaskedWritePreservesZeroResetBits) {
+  for (const u32 mask : {0u, 1u << 10}) {
+    reset_gx_state();
+    const auto before = g_gxState.pipelineStateGeneration;
+    decode_fifo(bp_cmd(0xFE, mask));
+    decode_fifo(bp_cmd(0, 0xFFFFFF));
+    EXPECT_EQ(g_gxState.bpRegCache[0], mask);
+    EXPECT_EQ(g_gxState.bpRegCache[0xFE], 0xFFFFFFu);
+    EXPECT_EQ(g_gxState.numTevStages, mask ? 2u : 1u);
+    EXPECT_EQ(g_gxState.cullMode, GX_CULL_NONE);
+    EXPECT_NE(g_gxState.pipelineStateGeneration, before);
+    decode_fifo(bp_cmd(0, 0));
+    EXPECT_EQ(g_gxState.numTevStages, 1u);
+    EXPECT_EQ(g_gxState.bpRegCache[0], 0u);
+  }
+}
+
+TEST_F(GXFifoTest, GenMode_ColdSingleStageApiSetupDecodes) {
+  reset_gx_state();
+  GXSetNumTevStages(1);
+  GXSetNumTexGens(0);
+  GXSetNumChans(0);
+  GXSetCullMode(GX_CULL_NONE);
+  const auto bytes = flush_and_capture();
+  decode_fifo(bytes);
+  EXPECT_EQ(g_gxState.numTevStages, 1u);
+  EXPECT_EQ(g_gxState.cullMode, GX_CULL_NONE);
+}
+
 TEST_F(GXFifoTest, BpMask_AppliesOnlyToNextWrite) {
  std::vector<u8> bytes;
  auto mask = bp_cmd(0xFE, 1u << 19);
@@ -2179,6 +2257,7 @@ TEST_F(GXFifoTest, DrawTopologyTemplatesPreserveExactGxIndexOrder) {
  const auto decodeAndReadIndices = [&](GXPrimitive primitive, u16 count) {
    std::vector<u8> fifo;
    append_test_draw(fifo, primitive, count);
+    aurora::gfx::testing::reset_vertex_push_record();
    decode_fifo(fifo);
    return aurora::gfx::testing::last_pushed_indices();
  };
@@ -2193,7 +2272,7 @@ TEST_F(GXFifoTest, DrawTopologyTemplatesPreserveExactGxIndexOrder) {
            (std::vector<u16>{0, 1, 2, 0, 2, 3, 0, 3, 4}));
  g_gxState.stateDirty = true;
  EXPECT_EQ(decodeAndReadIndices(GX_TRIANGLEFAN, 2),
-            (std::vector<u16>{0, 1}));
+            (std::vector<u16>{}));
  g_gxState.stateDirty = true;
  EXPECT_EQ(decodeAndReadIndices(GX_TRIANGLESTRIP, 6),
            (std::vector<u16>{0, 1, 2, 2, 1, 3, 2, 3, 4, 4, 3, 5}));
@@ -4166,7 +4245,9 @@ TEST_F(GXFifoTest, CopyTexClearTruePassesScratchRectAndUpdateMasksToResolve) {
  EXPECT_NEAR(resolve.clearColorValue.y(), 128.f / 255.f, 1.f / 255.f);
  EXPECT_NEAR(resolve.clearColorValue.z(), 192.f / 255.f, 1.f / 255.f);
  EXPECT_NEAR(resolve.clearColorValue.w(), 32.f / 255.f, 1.f / 255.f);
-  EXPECT_NEAR(resolve.clearDepthValue, 0x123456 / 16777216.f, 1.f / 16777216.f);
+  const float gxDepth = 0x123456 / 16777216.f;
+  EXPECT_NEAR(resolve.clearDepthValue, aurora::gx::UseReversedZ ? 1.f - gxDepth : gxDepth,
+              1.f / 16777216.f);
  EXPECT_EQ(resolve.resolveFormat, GX_TF_RGBA8);
  EXPECT_FALSE(resolve.halfScale);
  EXPECT_FALSE(resolve.forceOpaqueAlpha);
@@ -4186,7 +4267,7 @@ TEST_F(GXFifoTest, CopyTexColorFormatMarksResolvePersistent) {
  EXPECT_TRUE(records.front().persistentCopy);
 }

-TEST_F(GXFifoTest, RecurringColorCopyKeepsLaterResolveSkippable) {
+TEST_F(GXFifoTest, RecurringColorCopyPreservesEveryResolve) {
  std::array<u8, 152 * 114 * 4> image{};
  gxState().pixelFmt = GX_PF_RGBA6_Z24;

@@ -4200,7 +4281,7 @@ TEST_F(GXFifoTest, RecurringColorCopyKeepsLaterResolveSkippable) {
  const auto& records = aurora::gfx::testing::resolve_pass_records();
  ASSERT_EQ(records.size(), 2u);
  EXPECT_TRUE(records[0].persistentCopy);
-  EXPECT_FALSE(records[1].persistentCopy);
+  EXPECT_TRUE(records[1].persistentCopy);
 }

 TEST_F(GXFifoTest, ColorCopyAfterFrameGapRegainsPersistentProtection) {
@@ -4220,7 +4301,7 @@ TEST_F(GXFifoTest, ColorCopyAfterFrameGapRegainsPersistentProtection) {
  EXPECT_TRUE(records[1].persistentCopy);
 }

-TEST_F(GXFifoTest, CopyTexDepthFormatKeepsResolveSkippable) {
+TEST_F(GXFifoTest, CopyTexDepthFormatPreservesResolve) {
  std::array<u8, 4 * 4 * 4> image{};
  gxState().pixelFmt = GX_PF_RGBA6_Z24;

@@ -4230,7 +4311,7 @@ TEST_F(GXFifoTest, CopyTexDepthFormatKeepsResolveSkippable) {

  const auto& records = aurora::gfx::testing::resolve_pass_records();
  ASSERT_EQ(records.size(), 1u);
-  EXPECT_FALSE(records.front().persistentCopy);
+  EXPECT_TRUE(records.front().persistentCopy);
 }

 TEST_F(GXFifoTest, CopyDispResolveIsNotPersistent) {
@@ -299,6 +299,10 @@ std::pair<ByteBuffer, Range> copy_uniform(Range source) {
  return map_uniform(source.size);
 }
 uint32_t align_uniform(uint32_t value) { return (value + 255u) & ~255u; }
+uint64_t staging_uniform_bytes(uint64_t value) { return staging_padded(value, 256); }
+uint64_t staging_storage_bytes(uint64_t value) { return staging_padded(value, 256); }
+bool staging_has_space(const StagingSizes&) { return true; }
+void split_staging_batch() { throw StagingCapacityError("Unexpected split in FIFO unit test"); }

 Vec2<uint32_t> get_render_target_size() noexcept { return s_renderTargetSize; }
 Vec2<uint32_t> get_frame_buffer_size() noexcept { return s_renderTargetSize; }
@@ -0,0 +1,174 @@
+#include "gx_test_common.hpp"
+#include "gfx/staging_map.hpp"
+#include "gx/pipeline.hpp"
+
+#include <thread>
+
+using aurora::gx::g_gxState;
+
+namespace {
+std::vector<u8> draw(GXPrimitive primitive, u16 count, GXVtxFmt format = GX_VTXFMT0) {
+  std::vector<u8> bytes{static_cast<u8>(primitive | format), static_cast<u8>(count >> 8),
+                        static_cast<u8>(count)};
+  bytes.resize(3 + count);
+  return bytes;
+}
+}
+
+TEST_F(GXFifoTest, MaximumQuadCountTerminatesWithoutOutOfRangeIndices) {
+  g_gxState.lastVtxFmt = GX_VTXFMT0;
+  g_gxState.lastVtxSize = 1;
+  for (const u16 count : {65532, 65533, 65534, 65535}) {
+    g_gxState.stateDirty = true;
+    decode_fifo(draw(GX_QUADS, count));
+    const auto& indices = aurora::gfx::testing::last_pushed_indices();
+    ASSERT_EQ(indices.size(), (count / 4) * 6 + (count % 4 == 3 ? 3 : 0));
+    for (const auto index : indices) ASSERT_LT(index, count);
+  }
+}
+
+TEST_F(GXFifoTest, IncompletePrimitivesNeverJoinAcrossDraws) {
+  g_gxState.lastVtxFmt = GX_VTXFMT0;
+  g_gxState.lastVtxSize = 1;
+  aurora::gfx::testing::use_draw_command_tracking(true);
+  decode_fifo(draw(GX_TRIANGLES, 4));
+  EXPECT_EQ(aurora::gfx::testing::last_pushed_indices(), (std::vector<u16>{0, 1, 2}));
+  decode_fifo(draw(GX_TRIANGLES, 5));
+  EXPECT_EQ(aurora::gfx::testing::last_pushed_indices(), (std::vector<u16>{4, 5, 6}));
+  const auto before = aurora::gfx::testing::last_pushed_indices();
+  decode_fifo(draw(GX_TRIANGLEFAN, 2));
+  EXPECT_EQ(aurora::gfx::testing::last_pushed_indices(), before);
+}
+
+TEST_F(GXFifoTest, MergeStopsBeforeSixteenBitIndexOverflow) {
+  g_gxState.lastVtxFmt = GX_VTXFMT0;
+  g_gxState.lastVtxSize = 1;
+  aurora::gfx::testing::use_draw_command_tracking(true);
+  decode_fifo(draw(GX_TRIANGLES, 65535));
+  decode_fifo(draw(GX_TRIANGLES, 3));
+  EXPECT_EQ(aurora::gfx::g_mergedDrawCallCount, 0u);
+  EXPECT_EQ(aurora::gfx::testing::last_pushed_indices(), (std::vector<u16>{0, 1, 2}));
+}
+
+TEST_F(GXFifoTest, VertexCacheInvalidationBreaksDrawMerging) {
+  g_gxState.lastVtxFmt = GX_VTXFMT0;
+  g_gxState.lastVtxSize = 1;
+  aurora::gfx::testing::use_draw_command_tracking(true);
+  decode_fifo(draw(GX_TRIANGLES, 3));
+  decode_fifo({GX_CMD_INVL_VC});
+  EXPECT_TRUE(g_gxState.stateDirty);
+  decode_fifo(draw(GX_TRIANGLES, 3));
+  EXPECT_EQ(aurora::gfx::g_mergedDrawCallCount, 0u);
+}
+
+TEST_F(GXFifoTest, EqualStrideVertexFormatChangeBreaksDrawMerging) {
+  aurora::gfx::testing::use_real_vertex_format_helpers(true);
+  g_gxState.vtxDesc[GX_VA_POS] = GX_DIRECT;
+  for (const auto format : {GX_VTXFMT0, GX_VTXFMT1}) {
+    g_gxState.vtxFmts[format].attrs[GX_VA_POS].cnt = GX_POS_XY;
+    g_gxState.vtxFmts[format].attrs[GX_VA_POS].type = GX_U8;
+  }
+  g_gxState.vtxFmts[GX_VTXFMT1].attrs[GX_VA_POS].frac = 1;
+  aurora::gfx::testing::use_draw_command_tracking(true);
+  for (const auto format : {GX_VTXFMT0, GX_VTXFMT1}) {
+    auto bytes = draw(GX_TRIANGLES, 3, format);
+    bytes.resize(9);
+    decode_fifo(bytes);
+  }
+  EXPECT_EQ(aurora::gfx::g_mergedDrawCallCount, 0u);
+}
+
+TEST_F(GXFifoTest, SingleExpandedPrimitiveCannotMergeWithTriangles) {
+  g_gxState.lastVtxFmt = GX_VTXFMT0;
+  g_gxState.lastVtxSize = 1;
+  aurora::gfx::testing::use_draw_command_tracking(true);
+  decode_fifo(draw(GX_POINTS, 1));
+  decode_fifo(draw(GX_TRIANGLES, 3));
+  EXPECT_EQ(aurora::gfx::g_mergedDrawCallCount, 0u);
+  EXPECT_EQ(aurora::gfx::testing::last_pushed_indices(), (std::vector<u16>{0, 1, 2}));
+}
+
+TEST(StagingMapping, RetiredCallbacksCannotPublishAnotherBuffersReadiness) {
+  using namespace aurora::gfx;
+  StagingMapState state;
+  const auto old = state.request();
+  EXPECT_EQ(state.request(), 0u);
+  state.reset();
+  const auto current = state.request();
+  EXPECT_FALSE(state.complete(old, BufferMapState::Mapped));
+  EXPECT_FALSE(state.complete(old, BufferMapState::Unmapped));
+  EXPECT_EQ(state.state(), BufferMapState::Mapping);
+  EXPECT_TRUE(state.complete(current, BufferMapState::Mapped));
+  EXPECT_FALSE(state.complete(current, BufferMapState::Unmapped));
+  EXPECT_EQ(state.state(), BufferMapState::Mapped);
+}
+
+TEST(StagingMapping, AsyncCompletionWakesWaiters) {
+  using namespace aurora::gfx;
+  StagingMapState state;
+  const auto generation = state.request();
+  std::thread callback([&] {
+    std::this_thread::sleep_for(std::chrono::milliseconds(10));
+    state.complete(generation, BufferMapState::Mapped);
+  });
+  const auto deadline = std::chrono::steady_clock::now() + std::chrono::seconds(2);
+  while (state.state() == BufferMapState::Mapping && std::chrono::steady_clock::now() < deadline)
+    state.wait_for_progress();
+  callback.join();
+  EXPECT_EQ(state.state(), BufferMapState::Mapped);
+}
+
+TEST(StagingCapacity, ReservesPaddingAndRejectsOverflow) {
+  using namespace aurora::gfx;
+  EXPECT_EQ(staging_padded(257, 256), 512u);
+  EXPECT_THROW(staging_padded(UINT64_MAX, 256), StagingCapacityError);
+  const StagingSizes used{0, 256, 0, 0}, demand{0, 256, 0, 0}, tail{0, 3840, 0, 0};
+  EXPECT_TRUE(staging_fits(used, demand, tail, {4, 4352, 4, 4}));
+  EXPECT_FALSE(staging_fits(used, demand, tail, {4, 4351, 4, 4}));
+  EXPECT_FALSE(staging_fits({UINT64_MAX, 0, 0, 0}, {1, 0, 0, 0}, {},
+                            {UINT64_MAX, UINT64_MAX, UINT64_MAX, UINT64_MAX}));
+}
+
+TEST(FrameInterpolationContract, IdenticalMeshesInDifferentViewportsDoNotShareHistory) {
+  using namespace aurora;
+  const auto savedViewport = gx::g_gxState.logicalViewport;
+  constexpr size_t positionOffset = sizeof(Mat4x4<float>);
+  constexpr size_t normalOffset = positionOffset + gx::MaxPnMtx * sizeof(Mat3x4<float>);
+  constexpr size_t uniformSize = normalOffset + gx::MaxPnMtx * sizeof(Mat3x4<float>);
+  const gx::FrameInterpolationDrawIdentity identity{0x1234, 0x5678, 0x9abc, 0xdef0};
+  const Mat4x4<float> projection{};
+  const auto record = [&](float x, std::array<uint8_t, uniformSize>& source) {
+    gx::g_gxState.pnMtx[0].pos = {{1.f, 0.f, 0.f, x}, {0.f, 1.f, 0.f, 0.f}, {0.f, 0.f, 1.f, 0.f}};
+    gx::g_gxState.pnMtx[0].nrm = {{1.f, 0.f, 0.f, 0.f}, {0.f, 1.f, 0.f, 0.f}, {0.f, 0.f, 1.f, 0.f}};
+    std::memcpy(source.data() + positionOffset, &gx::g_gxState.pnMtx[0].pos, sizeof(Mat3x4<float>));
+    std::memcpy(source.data() + normalOffset, &gx::g_gxState.pnMtx[0].nrm, sizeof(Mat3x4<float>));
+    return gx::record_interpolation_draw(identity, projection, 1, {
+        .sourceUniformData = source.data(), .uniformSize = source.size(), .projectionOffset = 0,
+        .positionOffset = positionOffset, .normalOffset = normalOffset, .currentMatrix = 0,
+        .indexedMatrices = true});
+  };
+  gx::set_frame_interpolation_fps(0);
+  gx::begin_frame_interpolation();
+  gx::set_frame_interpolation_fps(120);
+  gx::g_gxState.logicalViewport = {0.f, 0.f, 640.f, 240.f, 0.f, 1.f};
+  std::array<uint8_t, uniformSize> previous{};
+  gx::begin_frame_interpolation();
+  record(0.f, previous);
+  gx::finalize_frame_interpolation();
+  gfx::testing::reset_uniform_allocations();
+  gx::g_gxState.logicalViewport.top = 240.f;
+  std::array<uint8_t, uniformSize> current{};
+  gx::begin_frame_interpolation();
+  const auto ranges = record(20.f, current);
+  const auto expected = current;
+  gx::finalize_frame_interpolation();
+  EXPECT_EQ(current, expected);
+  if (ranges[0].size) {
+    const auto& duplicate = gfx::testing::uniform_allocation(ranges[0].offset);
+    ASSERT_EQ(duplicate.size(), expected.size());
+    EXPECT_EQ(std::memcmp(duplicate.data(), expected.data(), expected.size()), 0);
+  }
+  gx::g_gxState.logicalViewport = savedViewport;
+  gx::set_frame_interpolation_fps(0);
+  gx::begin_frame_interpolation();
+}
@@ -0,0 +1,349 @@
+# Building WiiCompiled and Retro Rewind on macOS
+
+This guide covers building **WiiCompiled** (base game) and **Retro Rewind** from source on macOS for Apple Silicon (`arm64`). Follow these instructions to compile the native executables directly.
+
+> [!NOTE]
+> If you only want to build the base game (**WiiCompiled**), look for sections marked **`(Skip if only building WiiCompiled)`** to bypass Retro Rewind and online payload steps.
+
+---
+
+## 1. Prerequisites
+
+### System Requirements
+- **Hardware**: Apple Silicon Mac (M1/M2/M3/M4)
+- **Operating System**: macOS 14 (Sonoma) or later
+- **Xcode Command Line Tools**:
+  ```bash
+  xcode-select --install
+  ```
+
+### Toolchain Dependencies
+Install the required tools using [Homebrew](https://brew.sh):
+```bash
+brew install cmake ninja
+brew install --cask dotnet-sdk@8
+```
+
+Verify that Clang, CMake, Ninja, and the .NET 8 runtime are available:
+```bash
+clang --version
+cmake --version
+ninja --version
+dotnet --list-runtimes   # Must list Microsoft.NETCore.App 8.x
+```
+
+---
+
+## 2. Required Game and Mod Assets
+
+Due to legal requirements, no proprietary Nintendo assets or code are included in this repository. You must provide your own legally dumped game files.
+
+1. **Mario Kart Wii PAL (`RMCP01`) Disc Image** *(Required)*:
+   - Supported formats: `.iso`, `.wbfs`, `.ciso`, `.rvz`, `.gcm`, `.gcz`.
+2. **nodtool** *(Required for disc extraction)*:
+   - Download the macOS Apple Silicon binary of [nodtool](https://github.com/encounter/nod/releases):
+     ```bash
+     curl -fsSL "https://github.com/encounter/nod/releases/download/v2.0.0-alpha.10/nodtool-macos-arm64" -o nodtool
+     chmod +x nodtool
+     ```
+3. **Retro Rewind Distribution** *(Skip if only building WiiCompiled)*:
+   - Download the [Retro Rewind](https://wiki.tockdom.com/wiki/Retro_Rewind) release package. You will need the `RetroRewind6` folder (which contains `Binaries/Code.pul`).
+4. **Retro-WFC Payload** *(Skip if only building WiiCompiled or building offline)*:
+   - Required for online multiplayer on Retro Rewind. Downloaded during setup from `https://rwfc.net/api/wfc/payload?g=RMCPD00`.
+
+---
+
+## 3. Step 1: Extract Disc Assets
+
+Extract your clean PAL `RMCP01` disc into the `Assets/` directory of the repository:
+
+```bash
+# Using nodtool directly into a temporary scratch directory
+mkdir -p /tmp/mkw-extract
+./nodtool extract /path/to/RMCP01.iso /tmp/mkw-extract
+
+# Copy extracted assets into the repository Assets directory
+rm -rf Assets/DATA/files Assets/DATA/sys
+mkdir -p Assets/DATA
+cp /tmp/mkw-extract/*/sys/main.dol Assets/main.dol
+cp /tmp/mkw-extract/*/files/rel/StaticR.rel Assets/StaticR.rel
+cp -R /tmp/mkw-extract/*/files Assets/DATA/files
+cp -R /tmp/mkw-extract/*/sys Assets/DATA/sys
+
+# Clean up temporary files
+rm -rf /tmp/mkw-extract
+```
+
+> [!TIP]
+> Alternatively, you can use the repository's helper script:
+> ```bash
+> Launcher/macos/extract-disc.command --game /path/to/RMCP01.iso --assets-dir Assets --nodtool ./nodtool
+> ```
+
+### Verify Extracted Asset Hashes
+Confirm that the extracted files match the expected clean PAL revision:
+```bash
+shasum -a 256 Assets/main.dol Assets/StaticR.rel
+```
+- `Assets/main.dol`: `80d18895b39c63bd80f457398bfcbb91b7d16ac116a41a88967e954080155b05`
+- `Assets/StaticR.rel`: `16d9d146112541fefea701ecb5bc1a496f9d50e4a752fbb5b6778e7c6399f67d`
+
+---
+
+## 4. Step 2: Build the Translator CLI
+
+Compile the static recompiler CLI:
+
+```bash
+dotnet build translator/src/Translator.Cli/Translator.Cli.csproj -c Release
+```
+
+Define a shell function to invoke the translator (ensuring paths with spaces are handled safely):
+```bash
+translator() {
+  dotnet "$(pwd)/translator/src/Translator.Cli/bin/Release/net8.0/Translator.Cli.dll" "$@"
+}
+```
+
+---
+
+## 5. Step 3: Translation
+
+### A. Translate Base Game Functions
+```bash
+mkdir -p generated/functions build/base
+
+translator translate-recursive 0x800060A4 \
+  --project projects/mkwii/recomp.yml \
+  --outdir generated/functions \
+  --output-metadata generated/base_translation_output.json \
+  --production-source-bundle generated/base_translation_sources.bin \
+  --no-function-files \
+  --prune-stale \
+  --threads $(sysctl -n hw.ncpu)
+```
+
+### B. Emit Base Manifest
+```bash
+translator emit-base-manifest \
+  --project projects/mkwii/recomp.yml \
+  --out build/base \
+  --functions-dir generated/functions \
+  --translation-output-metadata generated/base_translation_output.json \
+  --region P
+```
+
+---
+
+### C. Stage and Translate Retro Rewind *(Skip this step if you only want to build WiiCompiled)*
+
+1. Stage `Code.pul`:
+   ```bash
+   RETRO_DIR="/path/to/RetroRewind6"
+   mkdir -p PulsarPacks/completed/RetroRewind/RetroRewind6/Binaries
+   cp "$RETRO_DIR/Binaries/Code.pul" PulsarPacks/completed/RetroRewind/RetroRewind6/Binaries/Code.pul
+   ```
+
+2. **Retro-WFC Payload Setup (for Online Multiplayer)**:
+   Online play in Retro Rewind requires the shared Retro-WFC payload. Download and validate it:
+   ```bash
+   mkdir -p build/retro-wfc/binary
+   curl -fsSL --retry 3 "https://rwfc.net/api/wfc/payload?g=RMCPD00" \
+     -o build/retro-wfc/binary/payload.RMCPD00.bin
+
+   # Validate payload signature and integrity
+   translator validate-retro-wfc-payload --directory build/retro-wfc
+   ```
+
+3. Run Retro Rewind translation:
+   ```bash
+   mkdir -p build/mods/retro_rewind_full_cpp
+
+   translator translate-mod \
+     --project projects/mkwii/recomp.yml \
+     --profile retro-rewind \
+     --base-manifest build/base/mkwii_base_manifest.json \
+     --base-translation-output-metadata generated/base_translation_output.json \
+     --code-pul "$RETRO_DIR/Binaries/Code.pul" \
+     --mod-root "$RETRO_DIR" \
+     --mod-name "Retro Rewind" \
+     --region P \
+     --out build/mods/retro_rewind_full_cpp \
+     --prefer-cached-inputs \
+     --emit-cpp \
+     --threads $(sysctl -n hw.ncpu) \
+     --retro-wfc-payload build/retro-wfc/binary/payload.RMCPD00.bin
+   ```
+   > [!TIP]
+   > If you do not want online play or do not have an internet connection, replace `--retro-wfc-payload ...` with `--skip-retro-wfc`.
+
+---
+
+### D. Generate Data Initialization and Build Shards
+
+First, generate the embedded game data initializer:
+```bash
+translator generate-data-init --project projects/mkwii/recomp.yml
+```
+
+Next, generate the CMake build shards using **one** of the following options:
+
+#### Option 1: Base Game Only (WiiCompiled)
+```bash
+mkdir -p generated/build_shards
+translator emit-build-shards \
+  --project projects/mkwii/recomp.yml \
+  --base-metadata generated/base_translation_output.json \
+  --base-functions-dir generated/functions \
+  --native-source-dir runtime/src \
+  --out generated/build_shards
+```
+
+#### Option 2: Base Game + Retro Rewind
+```bash
+mkdir -p generated/build_shards
+translator emit-build-shards \
+  --project projects/mkwii/recomp.yml \
+  --base-metadata generated/base_translation_output.json \
+  --base-functions-dir generated/functions \
+  --native-source-dir runtime/src \
+  --out generated/build_shards \
+  --resolved-profile build/mods/retro_rewind_full_cpp/resolved_dispatch_profile.json \
+  --retro-cpp-dir build/mods/retro_rewind_full_cpp/cpp
+```
+
+---
+
+## 6. Step 4: Configure and Compile with CMake & Ninja
+
+Configure the native C++ build targeting Apple Silicon:
+
+```bash
+cmake -S runtime -B build-macos -G Ninja \
+  -DCMAKE_BUILD_TYPE=Release \
+  -DCMAKE_C_COMPILER=clang \
+  -DCMAKE_CXX_COMPILER=clang++ \
+  -DAURORA_SDL3_PROVIDER=vendor
+```
+
+Compile the desired target:
+
+```bash
+# To build WiiCompiled only:
+cmake --build build-macos --target WiiCompiled --parallel $(sysctl -n hw.ncpu)
+
+# OR to build both WiiCompiled and Retro Rewind:
+cmake --build build-macos --target WiiCompiled RetroRewind --parallel $(sysctl -n hw.ncpu)
+```
+
+Once compilation completes, the executables are ready in your build directory:
+- `build-macos/WiiCompiled`
+- `build-macos/RetroRewind` (if built)
+
+During the build, CMake automatically copies the required runtime assets into `build-macos/`:
+- `build-macos/dsp_coef.bin`
+- `build-macos/initial_pipeline_cache.db`
+- `build-macos/wii_bootstrap/`
+
+---
+
+## 7. Step 5: Running Executables from the Build Folder
+
+### Configure `Config.toml`
+The runtime reads configuration from `~/Library/Application Support/WiiCompiled/Config.toml`.
+
+Create the directory and configuration file:
+
+```bash
+mkdir -p "$HOME/Library/Application Support/WiiCompiled"
+```
+
+#### For Base Game Only (WiiCompiled):
+```toml
+# ~/Library/Application Support/WiiCompiled/Config.toml
+[video]
+widescreen = true
+resolution_multiplier = 1.0
+graphics_api = "metal"
+
+[paths]
+dvd_root = "/absolute/path/to/Wiicompiled/Assets/DATA"
+```
+
+#### For Base Game and Retro Rewind:
+```toml
+# ~/Library/Application Support/WiiCompiled/Config.toml
+[video]
+widescreen = true
+resolution_multiplier = 1.0
+graphics_api = "metal"
+
+[paths]
+dvd_root = "/absolute/path/to/Wiicompiled/Assets/DATA"
+retro_rewind_root = "/path/to/RetroRewind6"
+```
+
+> [!NOTE]
+> Ensure `dvd_root` points to the directory containing `files` and `sys/fst.bin`.
+
+### Launching the Game
+Run the compiled binaries directly from your terminal or by double clicking:
+
+```bash
+# Run base WiiCompiled
+./build-macos/WiiCompiled
+
+# Run Retro Rewind
+./build-macos/RetroRewind
+```
+
+
+
+Press **F10** in-game at any time to open the configuration bar (controls, resolution, display settings, audio).
+
+---
+
+## Quick Reference: Automated Helper Script
+
+The repository provides a script (`Launcher/local-build-macos.command`) that handles extraction, translation, and compilation in a single command.
+
+### Building Base Game Only:
+```bash
+Launcher/local-build-macos.command \
+  --profile base \
+  --output-dir build-macos/Products \
+  --game /path/to/RMCP01.iso \
+  --nodtool ./nodtool
+```
+
+### Building Both (with Online Retro-WFC Payload):
+```bash
+# 1. Download Retro-WFC payload into a staging directory:
+mkdir -p build/retro-wfc/binary
+curl -fsSL --retry 3 "https://rwfc.net/api/wfc/payload?g=RMCPD00" \
+  -o build/retro-wfc/binary/payload.RMCPD00.bin
+
+# 2. Run the automated build with the payload directory:
+Launcher/local-build-macos.command \
+  --profile both \
+  --output-dir build-macos/Products \
+  --base-output-dir build-macos/Products \
+  --game /path/to/RMCP01.iso \
+  --nodtool ./nodtool \
+  --retro-rewind-package-dir /path/to/RetroRewind6 \
+  --retro-wfc-offline-dir build/retro-wfc
+```
+
+### Building Both (Offline, Skipping Payload):
+```bash
+Launcher/local-build-macos.command \
+  --profile both \
+  --output-dir build-macos/Products \
+  --base-output-dir build-macos/Products \
+  --game /path/to/RMCP01.iso \
+  --nodtool ./nodtool \
+  --retro-rewind-package-dir /path/to/RetroRewind6 \
+  --skip-retro-wfc-payload
+```
+
+When finished, the compiled executables reside in `native-build-macos/` and the bundled `.app` packages are placed in `build-macos/Products/`.
@@ -58,7 +58,7 @@ profiles:
    module_link_base: 0x803992E0
    output: build/mods/retro_rewind_full_cpp
    enable_retro_wfc: true
-    retro_wfc_payload: http://nas.play.rwfc.net/payload?g=RMCPD00
+    retro_wfc_payload: https://rwfc.net/api/wfc/payload?g=RMCPD00
    retro_wfc_legacy_bootstrap_hook: 0x800ED6E8
    riivolution:
      xml: xml/RetroRewind6.xml
@@ -1,4 +1,4 @@
-cmake_minimum_required(VERSION 3.16)
+cmake_minimum_required(VERSION 3.16)
 project(mkw_recompiled)

 if(NOT CMAKE_CXX_COMPILER_ID MATCHES "^(Clang|AppleClang)$" OR NOT CMAKE_SIZEOF_VOID_P EQUAL 8)
@@ -7,13 +7,6 @@ endif()

 if(WIN32 AND MINGW AND CMAKE_SYSTEM_PROCESSOR MATCHES "^(AMD64|amd64|x86_64|X86_64)$")
    set(MKW_PLATFORM_WINDOWS TRUE)
-elseif(CMAKE_SYSTEM_NAME STREQUAL "iOS")
-    # Checked before macOS: APPLE is true for iOS too, so the macOS branch would
-    # otherwise claim an iOS build and select a backend that cannot compile
-    # there (<mach/mach_vm.h> is absent from the iOS SDK) or run there (/tmp is
-    # outside the sandbox).
-    set(MKW_PLATFORM_IOS TRUE)
-    set(MKW_PLATFORM_MACOS TRUE)   # shares the Apple arm64 substrate
 elseif(APPLE AND CMAKE_SYSTEM_PROCESSOR MATCHES "^(arm64|ARM64)$")
    # The first native macOS target is Apple Silicon. Intel and universal
    # binaries remain future compatibility work; do not silently claim them.
@@ -22,16 +15,13 @@ elseif(CMAKE_SYSTEM_NAME STREQUAL "Linux" AND CMAKE_SYSTEM_PROCESSOR MATCHES "^(
    set(MKW_PLATFORM_LINUX TRUE)
 else()
    message(FATAL_ERROR
-        "WiiCompiled supports 64-bit LLVM-MinGW Clang on Windows, native Linux "
-        "x86_64/aarch64, or Apple Clang on macOS arm64 or iOS arm64")
+        "WiiCompiled supports 64-bit LLVM-MinGW Clang on Windows, native Linux x86_64/aarch64, or Apple Clang on macOS arm64")
 endif()
 if(NOT CMAKE_BUILD_TYPE STREQUAL "Release")
    message(FATAL_ERROR "WiiCompiled only supports Release builds")
 endif()

 option(MKW_BUILD_PRODUCTS "Build translated WiiCompiled product targets" ON)
-set(MKW_IOS_BUNDLE_ID_PREFIX "it.tobyfox" CACHE STRING
-    "Reverse-DNS prefix for the iOS bundle identifier")

 # Preprocessor definitions that belong to this project's own code (the runtime,
 # the translated shards and the product glue) and to nothing else. They are
@@ -40,11 +30,6 @@ set(MKW_IOS_BUNDLE_ID_PREFIX "it.tobyfox" CACHE STRING
 # libraries and the products in cmake/PublicProducts.cmake) inherits them while
 # aurora-main and its third-party tree stay unaffected.
 set(MKW_PROJECT_COMPILE_DEFINITIONS NOMINMAX)
-if(MKW_PLATFORM_IOS)
-    # One project-wide answer to "is this iOS", so sources do not each pull in
-    # TargetConditionals.h and spell out the __APPLE__ && TARGET_OS_IPHONE dance.
-    list(APPEND MKW_PROJECT_COMPILE_DEFINITIONS MKW_PLATFORM_IOS=1)
-endif()

 set(CMAKE_CXX_STANDARD 17)

@@ -122,6 +107,43 @@ if(NOT MKW_NATIVE_PREBUILT_DIR)
    set_target_properties(mkw_cryptopp PROPERTIES UNITY_BUILD OFF)
 endif()

+# TLS for non-Windows guest network HLE (runtime/src/hle/net/network_ssl.cpp) - the Windows path
+# uses Schannel (a Windows-only OS API), which has no equivalent on Linux/Android, so this project
+# needs its own TLS library there. mbed TLS was chosen over OpenSSL specifically because it cross-
+# compiles cleanly for Android with nothing beyond a plain C toolchain (no perl/asm build-script
+# dependency the way OpenSSL's build has), matching how this project already prefers toolchain-
+# simple libraries (see Crypto++ above, similarly stripped of ASM/SIMD for portability).
+# Fetched at build time from a pinned upstream release tarball with a checked SHA-256, the same way
+# aurora-main's own dependencies (SDL, zlib, etc.) are pulled in - not committed as a vendored
+# source tree, so the repository ships the compiled dependency rather than ~280 tracked upstream
+# files. Bump MKW_MBEDTLS_VERSION/MKW_MBEDTLS_SHA256 together when updating; the hash comes from
+# upstream's own signed `mbedtls-<version>-sha256sum.txt` release asset.
+#
+# The alias exists on every platform so the link lines in cmake/PublicProducts.cmake stay
+# platform-independent, but it is only populated where network_ssl.cpp actually compiles the mbed
+# TLS path (`#ifndef _WIN32`). Windows keeps Schannel and must not fetch anything: its builds run
+# with FETCHCONTENT_FULLY_DISCONNECTED=ON against the offline dependency set prepared by
+# Launcher/Prepare-Dependencies.ps1, so an unconditional fetch would fail a clean configure there
+# and would also add a dependency Windows never links.
+add_library(mkw_mbedtls INTERFACE)
+add_library(mkw::mbedtls ALIAS mkw_mbedtls)
+if(NOT MKW_PLATFORM_WINDOWS)
+    include(FetchContent)
+    set(MKW_MBEDTLS_VERSION "3.6.7")
+    set(MKW_MBEDTLS_SHA256 "a7e8bcbec0e6f761b4af24f25677626b35f762f68eef79c08677a363212d11f6")
+    FetchContent_Declare(mkw_mbedtls_upstream
+        URL "https://github.com/Mbed-TLS/mbedtls/releases/download/mbedtls-${MKW_MBEDTLS_VERSION}/mbedtls-${MKW_MBEDTLS_VERSION}.tar.bz2"
+        URL_HASH SHA256=${MKW_MBEDTLS_SHA256})
+    # Subproject mode already defaults ENABLE_TESTING off and skips codegen (GEN_FILES), but
+    # ENABLE_PROGRAMS defaults on and installation/package-config isn't wanted for a linked-in copy.
+    set(ENABLE_PROGRAMS OFF CACHE BOOL "" FORCE)
+    set(ENABLE_TESTING OFF CACHE BOOL "" FORCE)
+    set(MBEDTLS_FATAL_WARNINGS OFF CACHE BOOL "" FORCE)
+    set(DISABLE_PACKAGE_CONFIG_AND_INSTALL ON CACHE BOOL "" FORCE)
+    FetchContent_MakeAvailable(mkw_mbedtls_upstream)
+    target_link_libraries(mkw_mbedtls INTERFACE MbedTLS::mbedtls MbedTLS::mbedx509 MbedTLS::mbedcrypto)
+endif()
+
 set(MKW_TRANSLATED_COMPILE_JOBS 0 CACHE STRING
    "Cap on concurrently compiling translated shard TUs via a Ninja job pool (0 = uncapped). \
 Scheduling only - never affects output bytes, so it is deliberately outside the canonical flag fingerprint.")
@@ -151,16 +173,6 @@ else()
    if(NOT EXISTS ${MKW_AURORA_DIR}/CMakeLists.txt)
        message(FATAL_ERROR "Requested aurora-main but ${MKW_AURORA_DIR} is missing")
    endif()
-    # aurora picks shared third-party libraries whenever BUILD_SHARED_LIBS is
-    # merely undefined, and its extern/ tree calls unset(BUILD_SHARED_LIBS CACHE)
-    # while working around xxhash, which deletes any -D the user passed. Defining
-    # it here, before aurora is added, is what actually sticks across
-    # reconfigures. Apple builds ship a self-contained bundle, so a dylib
-    # resolved from the build tree or from Homebrew cannot be allowed to leak
-    # into the link.
-    if(APPLE)
-        set(BUILD_SHARED_LIBS OFF)
-    endif()
    set(DAWN_ENABLE_D3D11 OFF CACHE BOOL "" FORCE)
    if(MKW_PLATFORM_WINDOWS)
        set(DAWN_ENABLE_D3D12 ON CACHE BOOL "" FORCE)
@@ -274,18 +286,8 @@ if(MKW_PLATFORM_MACOS)
    # mkw_co_init/mkw_co_switch at link time.
    enable_language(ASM)
    list(APPEND SOURCES "${CMAKE_CURRENT_LIST_DIR}/src/platform/macos/co_switch.S")
-    set_source_files_properties("${CMAKE_CURRENT_LIST_DIR}/src/platform/macos/co_switch.S"
-        PROPERTIES LANGUAGE ASM SKIP_UNITY_BUILD_INCLUSION ON SKIP_PRECOMPILE_HEADERS ON)
-    # iOS cannot use the macOS backend: <mach/mach_vm.h> is absent from that SDK
-    # and /tmp is outside the sandbox.
-    if(MKW_PLATFORM_IOS)
-        list(REMOVE_ITEM SOURCES "${CMAKE_CURRENT_LIST_DIR}/src/guest_flat_memory_macos.cpp")
-    else()
-        list(REMOVE_ITEM SOURCES "${CMAKE_CURRENT_LIST_DIR}/src/guest_flat_memory_ios.cpp")
-    endif()
 else()
    list(REMOVE_ITEM SOURCES "${CMAKE_CURRENT_LIST_DIR}/src/guest_flat_memory_macos.cpp")
-    list(REMOVE_ITEM SOURCES "${CMAKE_CURRENT_LIST_DIR}/src/guest_flat_memory_ios.cpp")
 endif()
 set(MKW_PLATFORM_SOURCE "${CMAKE_CURRENT_LIST_DIR}/src/platform/host_platform.cpp")
 set(MKW_BASE_PRODUCT_SOURCE "${CMAKE_CURRENT_LIST_DIR}/src/product/base_product.cpp")
@@ -312,25 +314,31 @@ target_link_libraries(mkw_platform_paths_tests PRIVATE mkw_platform)
 target_compile_features(mkw_platform_paths_tests PRIVATE cxx_std_17)
 add_test(NAME mkw_platform_paths_tests COMMAND mkw_platform_paths_tests)

-add_executable(mkw_touch_auto_accelerate_tests
-    "${CMAKE_CURRENT_LIST_DIR}/tests/touch_auto_accelerate_tests.cpp"
-    "${CMAKE_CURRENT_LIST_DIR}/src/hle/input/touch_pad.cpp")
-target_include_directories(mkw_touch_auto_accelerate_tests PRIVATE
-    "${CMAKE_CURRENT_LIST_DIR}/include"
-    "${CMAKE_CURRENT_LIST_DIR}/src"
-    "${CMAKE_CURRENT_LIST_DIR}/third_party/toml11/single_include"
-    "${CMAKE_CURRENT_LIST_DIR}/third_party/toml11"
-    "${MKW_AURORA_DIR}/include"
-    "${CMAKE_BINARY_DIR}/_deps/imgui-src"
-    "${CMAKE_BINARY_DIR}/_deps/sdl-src/include")
-if(TARGET mkw_native_prebuilt)
-    target_include_directories(mkw_touch_auto_accelerate_tests PRIVATE
-        $<TARGET_PROPERTY:mkw_native_prebuilt,INTERFACE_INCLUDE_DIRECTORIES>)
-endif()
-target_compile_features(mkw_touch_auto_accelerate_tests PRIVATE cxx_std_20)
-target_compile_definitions(mkw_touch_auto_accelerate_tests PRIVATE SDL_MAIN_HANDLED TARGET_PC)
-target_link_libraries(mkw_touch_auto_accelerate_tests PRIVATE mkw_platform)
-add_test(NAME mkw_touch_auto_accelerate_tests COMMAND mkw_touch_auto_accelerate_tests)
+add_executable(mkw_nand_save_tests "${CMAKE_CURRENT_LIST_DIR}/tests/nand_save_tests.cpp")
+target_include_directories(mkw_nand_save_tests PRIVATE "${CMAKE_CURRENT_LIST_DIR}/include")
+target_compile_features(mkw_nand_save_tests PRIVATE cxx_std_17)
+add_test(NAME mkw_nand_save_tests COMMAND mkw_nand_save_tests)
+
+add_executable(mkw_nand_settings_tests "${CMAKE_CURRENT_LIST_DIR}/tests/nand_settings_tests.cpp")
+find_package(Threads REQUIRED)
+target_link_libraries(mkw_nand_settings_tests PRIVATE Threads::Threads)
+target_include_directories(mkw_nand_settings_tests PRIVATE "${CMAKE_CURRENT_LIST_DIR}/include")
+target_compile_features(mkw_nand_settings_tests PRIVATE cxx_std_17)
+add_test(NAME mkw_nand_settings_tests COMMAND mkw_nand_settings_tests)
+
+add_executable(mkw_sc_serial_tests "${CMAKE_CURRENT_LIST_DIR}/tests/sc_serial_tests.cpp")
+target_include_directories(mkw_sc_serial_tests PRIVATE "${CMAKE_CURRENT_LIST_DIR}/include")
+target_compile_features(mkw_sc_serial_tests PRIVATE cxx_std_17)
+add_test(NAME mkw_sc_serial_tests COMMAND mkw_sc_serial_tests)
+
+# The input expression engine is self-contained, so it can be exercised without
+# linking the runtime or SDL.
+add_executable(mkw_input_expr_tests
+    "${CMAKE_CURRENT_LIST_DIR}/tests/test_expr.cpp"
+    "${CMAKE_CURRENT_LIST_DIR}/src/input_expr.cpp")
+target_include_directories(mkw_input_expr_tests PRIVATE "${CMAKE_CURRENT_LIST_DIR}/include")
+target_compile_features(mkw_input_expr_tests PRIVATE cxx_std_17)
+add_test(NAME mkw_input_expr_tests COMMAND mkw_input_expr_tests)

 # HostContext deliberately keeps the platform-specific context primitive out
 # of fiber_manager.cpp. Exercise the Linux libco handoff directly so future
@@ -347,11 +355,9 @@ if(MKW_PLATFORM_LINUX)
    add_test(NAME mkw_linux_host_context_tests COMMAND mkw_linux_host_context_tests)
 endif()

-if(MKW_PLATFORM_MACOS AND NOT MKW_PLATFORM_IOS)
+if(MKW_PLATFORM_MACOS)
    # Exercise the Apple Silicon context ABI and the public host-memory
-    # contracts separately from translated products. Host-only: the flat memory
-    # test pulls in guest_flat_memory_macos.cpp, whose <mach/mach_vm.h> does not
-    # exist in the iOS SDK, and none of these can run on a device anyway.
+    # contracts separately from translated products.
    enable_language(ASM)
    add_executable(mkw_macos_context_abi_tests
        "${CMAKE_CURRENT_LIST_DIR}/tests/macos_context_abi_tests.cpp"
@@ -421,7 +427,7 @@ else()
        target_compile_definitions(mkw_macos_native_compile PRIVATE SDL_MAIN_HANDLED TARGET_PC)
        target_link_libraries(mkw_macos_native_compile PRIVATE
            aurora::gx aurora::pad aurora::si aurora::vi aurora::mtx
-            PNG::PNG mkw::pugixml mkw::toml11 mkw::cryptopp)
+            mkw::pugixml mkw::toml11 mkw::cryptopp)
        set_target_properties(mkw_macos_native_compile PROPERTIES UNITY_BUILD OFF)
    endif()
    add_custom_target(mkw_platform_paths_check DEPENDS mkw_platform)
@@ -1,8 +0,0 @@
-GameCube Button Icons and Controls - Zacksly
-Licensed under CC BY 3.0 - https://zacksly.itch.io
-
-The PNG files in this directory are taken unmodified from the "Buttons Outline /
-White" set of that pack, at 256px. They are drawn tinted and partly transparent
-at runtime, which is a rendering choice and not an edit to the artwork.
-
-Full licence: http://creativecommons.org/licenses/by/3.0/
@@ -77,18 +77,11 @@ add_library(mkw_runtime_common OBJECT ${SOURCES})
 mkw_configure_object_target(mkw_runtime_common)
 target_compile_features(mkw_runtime_common PRIVATE cxx_std_20)
 target_compile_definitions(mkw_runtime_common PRIVATE
+    SDL_MAIN_HANDLED
    _DISABLE_STRING_ANNOTATION _DISABLE_VECTOR_ANNOTATION)
-if(NOT MKW_PLATFORM_IOS)
-    # iOS is the one target where SDL must own main(): its entry point is what
-    # drives UIApplicationMain. Defining this makes <SDL3/SDL_main.h> a no-op,
-    # so the app links a bare main() that UIKit never calls.
-    target_compile_definitions(mkw_runtime_common PRIVATE SDL_MAIN_HANDLED)
-endif()
 target_link_libraries(mkw_runtime_common PRIVATE
    aurora::gx aurora::pad aurora::si aurora::vi aurora::mtx)
-# The touch overlay decodes its button artwork at startup.
-target_link_libraries(mkw_runtime_common PRIVATE PNG::PNG)
-target_link_libraries(mkw_runtime_common PRIVATE mkw_platform mkw::pugixml mkw::toml11 mkw::cryptopp)
+target_link_libraries(mkw_runtime_common PRIVATE mkw_platform mkw::pugixml mkw::toml11 mkw::cryptopp mkw::mbedtls)
 if(MKW_PLATFORM_WINDOWS)
    target_link_libraries(mkw_runtime_common PRIVATE shell32 windowsapp)
 elseif(MKW_PLATFORM_LINUX)
@@ -199,17 +192,14 @@ function(mkw_configure_product target)
        "${MKW_RUNTIME_SOURCE_DIR}/.."
        "${MKW_RUNTIME_SOURCE_DIR}/../aurora-main/include")
    target_compile_definitions(${target} PRIVATE
-        _DISABLE_STRING_ANNOTATION _DISABLE_VECTOR_ANNOTATION TARGET_PC)
-    if(NOT MKW_PLATFORM_IOS)
-        target_compile_definitions(${target} PRIVATE SDL_MAIN_HANDLED)
-    endif()
+        SDL_MAIN_HANDLED _DISABLE_STRING_ANNOTATION _DISABLE_VECTOR_ANNOTATION TARGET_PC)
    target_compile_features(${target} PRIVATE cxx_std_20)
    mkw_apply_common_compile_options(${target})
    # The dispatch-table and registration shards compile inside the product target itself and
    # include the same fat translated headers; bound them by the same pool.
    mkw_bound_translated_compiles(${target})
    target_link_libraries(${target} PRIVATE
-        mkw_platform mkw_base_shared mkw::pugixml mkw::toml11 mkw::cryptopp)
+        mkw_platform mkw_base_shared mkw::pugixml mkw::toml11 mkw::cryptopp mkw::mbedtls)

    target_link_libraries(${target} PRIVATE
        aurora::gx aurora::pad aurora::si aurora::vi aurora::mtx)
@@ -231,26 +221,6 @@ function(mkw_configure_product target)
            $<TARGET_FILE:sqlite3> $<TARGET_FILE_DIR:${target}>)
    endif()

-    if(MKW_PLATFORM_IOS AND NOT CMAKE_HOST_APPLE)
-        set(shim "${MKW_RUNTIME_SOURCE_DIR}/src/platform/ios/compiler_rt_shim.c")
-        target_sources(${target} PRIVATE "${shim}")
-        set_source_files_properties("${shim}" PROPERTIES SKIP_PRECOMPILE_HEADERS ON SKIP_UNITY_BUILD_INCLUSION ON)
-    endif()
-
-    if(MKW_PLATFORM_IOS)
-        # CMake bundles iOS targets with its own default Info.plist, whose
-        # CFBundleIdentifier is empty and which carries none of the iOS keys, so
-        # iOS refuses to install the result. Supply a real one.
-        set_target_properties(${target} PROPERTIES
-            MACOSX_BUNDLE TRUE
-            MACOSX_BUNDLE_INFO_PLIST "${CMAKE_CURRENT_LIST_DIR}/ios/Info.plist.in"
-            MACOSX_BUNDLE_EXECUTABLE_NAME "${target}"
-            MACOSX_BUNDLE_BUNDLE_NAME "${target}"
-            MACOSX_BUNDLE_GUI_IDENTIFIER "${MKW_IOS_BUNDLE_ID_PREFIX}.wiicompiled"
-            MACOSX_BUNDLE_BUNDLE_VERSION "1"
-            MACOSX_BUNDLE_SHORT_VERSION_STRING "1.0")
-    endif()
-
    if(MKW_PLATFORM_WINDOWS)
        target_link_libraries(${target} PRIVATE
            dbghelp user32 winmm ws2_32 iphlpapi secur32 crypt32 windowsapp)
@@ -293,19 +263,6 @@ function(mkw_configure_product target)
    add_custom_command(TARGET ${target} POST_BUILD COMMAND ${CMAKE_COMMAND} -E copy_directory
        "${MKW_WII_BOOTSTRAP_SOURCE_DIR}" "$<TARGET_FILE_DIR:${target}>/wii_bootstrap")

-    # Touch control artwork. Only the touch build reads these, but they are copied
-    # everywhere the other assets are so the bundle layout stays uniform.
-    set(MKW_TOUCH_ASSET_DIR "${MKW_RUNTIME_SOURCE_DIR}/assets/touch")
-    if(NOT EXISTS "${MKW_TOUCH_ASSET_DIR}/a.png")
-        message(FATAL_ERROR "Missing touch control artwork: ${MKW_TOUCH_ASSET_DIR}")
-    endif()
-    # Cleared first: copy_directory merges, so a removed icon would otherwise
-    # linger in the bundle and get shipped.
-    add_custom_command(TARGET ${target} POST_BUILD
-        COMMAND ${CMAKE_COMMAND} -E rm -rf "$<TARGET_FILE_DIR:${target}>/touch"
-        COMMAND ${CMAKE_COMMAND} -E copy_directory
-            "${MKW_TOUCH_ASSET_DIR}" "$<TARGET_FILE_DIR:${target}>/touch")
-
    set(MKW_DSP_COEFFICIENT_ROM "${MKW_RUNTIME_SOURCE_DIR}/assets/dsp/dsp_coef.bin")
    if(NOT EXISTS "${MKW_DSP_COEFFICIENT_ROM}")
        message(FATAL_ERROR "Missing Wii DSP coefficient ROM: ${MKW_DSP_COEFFICIENT_ROM}")
@@ -330,24 +287,20 @@ function(mkw_configure_product target)
        "${MKW_INITIAL_PIPELINE_CACHE}"
        "$<TARGET_FILE_DIR:${target}>/initial_pipeline_cache.db")

-    if(MKW_PLATFORM_IOS)
-        # An .ipa is a zip with the bundle under Payload/. Left unsigned: the
-        # sideloaders people actually install with (AltStore, SideStore) sign on
-        # the device with the user's own Apple ID, and a signature applied here
-        # would only be replaced. WiiCompiled.entitlements beside this file names
-        # the one entitlement the runtime needs, for whatever does the signing.
-        add_custom_command(TARGET ${target} POST_BUILD
-            COMMAND ${CMAKE_COMMAND} -E rm -rf "$<TARGET_FILE_DIR:${target}>/../ipa"
-            COMMAND ${CMAKE_COMMAND} -E make_directory "$<TARGET_FILE_DIR:${target}>/../ipa/Payload"
-            COMMAND ${CMAKE_COMMAND} -E copy_directory
-                "$<TARGET_BUNDLE_DIR:${target}>"
-                "$<TARGET_FILE_DIR:${target}>/../ipa/Payload/${target}.app"
-            COMMAND ${CMAKE_COMMAND} -E chdir "$<TARGET_FILE_DIR:${target}>/../ipa"
-                ${CMAKE_COMMAND} -E tar cf "${CMAKE_BINARY_DIR}/${target}-unsigned.ipa" --format=zip Payload
-            COMMAND ${CMAKE_COMMAND} -E rm -rf "$<TARGET_FILE_DIR:${target}>/../ipa"
-            COMMENT "Packaging ${target}-unsigned.ipa")
+    # Non-Windows TLS (runtime/src/hle/net/network_ssl.cpp's mbed TLS path) needs a trusted root
+    # CA bundle to verify server certificates against - Windows gets this for free from the OS via
+    # Schannel, mbed TLS does not ship one itself. Not SHA256-pinned like the DSP ROM above: unlike
+    # a fixed hardware ROM, this bundle is expected to be refreshed periodically as CAs rotate.
+    # Windows gets its trust store from Schannel, so only the platforms that actually build the
+    # mbed TLS path need the bundle beside the executable.
+    if(NOT MKW_PLATFORM_WINDOWS)
+        set(MKW_CA_CERTIFICATE_BUNDLE "${MKW_RUNTIME_SOURCE_DIR}/assets/certs/cacert.pem")
+        if(NOT EXISTS "${MKW_CA_CERTIFICATE_BUNDLE}")
+            message(FATAL_ERROR "Missing TLS root CA bundle: ${MKW_CA_CERTIFICATE_BUNDLE}")
+        endif()
+        add_custom_command(TARGET ${target} POST_BUILD COMMAND ${CMAKE_COMMAND} -E copy_if_different
+            "${MKW_CA_CERTIFICATE_BUNDLE}" "$<TARGET_FILE_DIR:${target}>/cacert.pem")
    endif()
-
 endfunction()

 add_executable(WiiCompiled "${MKW_BASE_PRODUCT_SOURCE}" ${MKW_BASE_REGISTRATION_SOURCES})
@@ -381,11 +334,6 @@ endif()
 # tuning rather than leaving target-specific performance on the table.
 if(CMAKE_SYSTEM_PROCESSOR MATCHES "^(AMD64|amd64|x86_64|X86_64)$")
    set(MKW_BASELINE_ARCH_FLAG -march=x86-64-v3)
-elseif(MKW_PLATFORM_IOS)
-    # iOS 17 runs on the A12 and later, and the build host is never the device;
-    # -mcpu=native there would tune for whatever Mac (or Linux box) did the
-    # compile, and upstream clang rejects it when cross-compiling.
-    set(MKW_BASELINE_ARCH_FLAG -mcpu=apple-a12)
 elseif(CMAKE_SYSTEM_PROCESSOR MATCHES "^(aarch64|arm64|ARM64)$")
    set(MKW_BASELINE_ARCH_FLAG -mcpu=native)
 else()
@@ -1,71 +0,0 @@
-# Cross-compile the iOS arm64 products from a Linux host.
-#
-# Needs upstream clang and lld (Debian 13: clang-19 lld-19 llvm-19) and an
-# iPhoneOS SDK, either copied out of Xcode or sparse-cloned from
-# github.com/xybp888/iOS-SDKs. Nothing from Theos or xtool. See README.md,
-# "iOS from Linux".
-#
-#   cmake -S runtime -B build-ios -G Ninja -DCMAKE_BUILD_TYPE=Release \
-#       -DCMAKE_TOOLCHAIN_FILE=cmake/ios-linux-toolchain.cmake \
-#       -DCMAKE_DISABLE_FIND_PACKAGE_absl=TRUE -DAURORA_DAWN_PROVIDER=package
-
-set(IOS_SDK "/opt/iPhoneOS.sdk" CACHE PATH "iPhoneOS SDK copied from Xcode")
-set(MKW_IOS_LLVM_BIN "/usr/lib/llvm-19/bin" CACHE PATH "Directory holding clang, ld64.lld and llvm-ar")
-set(MKW_IOS_CLANG_RT "" CACHE FILEPATH
-    "Optional: Xcode's libclang_rt.ios.a. Left empty, src/platform/ios/compiler_rt_shim.c covers what the runtime needs")
-list(APPEND CMAKE_TRY_COMPILE_PLATFORM_VARIABLES IOS_SDK MKW_IOS_LLVM_BIN MKW_IOS_CLANG_RT)
-
-set(CMAKE_SYSTEM_NAME iOS)
-set(CMAKE_SYSTEM_PROCESSOR arm64)
-set(CMAKE_OSX_ARCHITECTURES arm64 CACHE STRING "")
-set(CMAKE_OSX_DEPLOYMENT_TARGET 17.0 CACHE STRING "")
-set(CMAKE_OSX_SYSROOT "${IOS_SDK}" CACHE PATH "")
-set(CMAKE_SYSROOT "${IOS_SDK}")
-
-set(x "${CMAKE_HOST_EXECUTABLE_SUFFIX}")
-set(CMAKE_C_COMPILER "${MKW_IOS_LLVM_BIN}/clang${x}")
-set(CMAKE_CXX_COMPILER "${MKW_IOS_LLVM_BIN}/clang++${x}")
-set(CMAKE_OBJC_COMPILER "${MKW_IOS_LLVM_BIN}/clang${x}")
-set(CMAKE_OBJCXX_COMPILER "${MKW_IOS_LLVM_BIN}/clang++${x}")
-set(CMAKE_ASM_COMPILER "${MKW_IOS_LLVM_BIN}/clang${x}")
-set(CMAKE_AR "${MKW_IOS_LLVM_BIN}/llvm-ar${x}" CACHE FILEPATH "")
-set(CMAKE_RANLIB "${MKW_IOS_LLVM_BIN}/llvm-ranlib${x}" CACHE FILEPATH "")
-set(CMAKE_STRIP "${MKW_IOS_LLVM_BIN}/llvm-strip${x}" CACHE FILEPATH "")
-# Not every LLVM distribution ships these (llvm-mingw does not); nothing in
-# this build needs them, but CMake's Darwin support likes to know where they are.
-foreach(tool INSTALL_NAME_TOOL:install-name-tool LIPO:lipo OTOOL:otool)
-    string(REPLACE ":" ";" tool "${tool}")
-    list(GET tool 0 var)
-    list(GET tool 1 exe)
-    if(EXISTS "${MKW_IOS_LLVM_BIN}/llvm-${exe}${x}")
-        set(CMAKE_${var} "${MKW_IOS_LLVM_BIN}/llvm-${exe}${x}" CACHE FILEPATH "")
-    endif()
-endforeach()
-set(CMAKE_LINKER "${MKW_IOS_LLVM_BIN}/ld64.lld${x}" CACHE FILEPATH "")
-
-foreach(lang C CXX OBJC OBJCXX ASM)
-    set(CMAKE_${lang}_COMPILER_TARGET arm64-apple-ios17.0)
-endforeach()
-
-# Apple's clang searches the SDK's SubFrameworks implicitly and UIKit's own
-# headers import from there.
-foreach(lang C CXX OBJC OBJCXX)
-    set(CMAKE_${lang}_FLAGS_INIT "-iframework ${IOS_SDK}/System/Library/SubFrameworks")
-endforeach()
-# The SDK carries its own libc++ headers and Apple's clang uses those. Upstream
-# clang prefers a libc++ shipped beside itself when there is one (llvm-mingw
-# does), which pairs the wrong C++ library with Darwin's C headers.
-foreach(lang CXX OBJCXX)
-    string(APPEND CMAKE_${lang}_FLAGS_INIT " -nostdinc++ -isystem ${IOS_SDK}/usr/include/c++/v1")
-endforeach()
-
-set(CMAKE_EXE_LINKER_FLAGS_INIT "-fuse-ld=lld ${MKW_IOS_CLANG_RT}")
-set(CMAKE_SHARED_LINKER_FLAGS_INIT "-fuse-ld=lld ${MKW_IOS_CLANG_RT}")
-set(CMAKE_MODULE_LINKER_FLAGS_INIT "-fuse-ld=lld ${MKW_IOS_CLANG_RT}")
-string(STRIP "${CMAKE_EXE_LINKER_FLAGS_INIT}" CMAKE_EXE_LINKER_FLAGS_INIT)
-
-set(CMAKE_FIND_ROOT_PATH "${IOS_SDK}")
-set(CMAKE_FIND_ROOT_PATH_MODE_PROGRAM NEVER)
-set(CMAKE_FIND_ROOT_PATH_MODE_LIBRARY ONLY)
-set(CMAKE_FIND_ROOT_PATH_MODE_INCLUDE ONLY)
-set(CMAKE_FIND_ROOT_PATH_MODE_PACKAGE BOTH)
@@ -1,26 +0,0 @@
-<?xml version="1.0" encoding="UTF-8"?>
-<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
-<plist version="1.0">
-<dict>
-  <key>CFBundleExecutable</key><string>${MACOSX_BUNDLE_EXECUTABLE_NAME}</string>
-  <key>CFBundleIdentifier</key><string>${MACOSX_BUNDLE_GUI_IDENTIFIER}</string>
-  <key>CFBundleName</key><string>${MACOSX_BUNDLE_BUNDLE_NAME}</string>
-  <key>CFBundleDisplayName</key><string>${MACOSX_BUNDLE_BUNDLE_NAME}</string>
-  <key>CFBundleVersion</key><string>${MACOSX_BUNDLE_BUNDLE_VERSION}</string>
-  <key>CFBundleShortVersionString</key><string>${MACOSX_BUNDLE_SHORT_VERSION_STRING}</string>
-  <key>CFBundlePackageType</key><string>APPL</string>
-  <key>LSRequiresIPhoneOS</key><true/>
-  <key>MinimumOSVersion</key><string>${CMAKE_OSX_DEPLOYMENT_TARGET}</string>
-  <key>UIDeviceFamily</key><array><integer>1</integer><integer>2</integer></array>
-  <key>UILaunchScreen</key><dict/>
-  <key>UIRequiresFullScreen</key><true/>
-  <key>CADisableMinimumFrameDuration</key><true/>
-  <key>UIFileSharingEnabled</key><true/>
-  <key>LSSupportsOpeningDocumentsInPlace</key><true/>
-  <key>UISupportedInterfaceOrientations</key>
-  <array>
-    <string>UIInterfaceOrientationLandscapeLeft</string>
-    <string>UIInterfaceOrientationLandscapeRight</string>
-  </array>
-</dict>
-</plist>
@@ -1,11 +0,0 @@
-<?xml version="1.0" encoding="UTF-8"?>
-<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
-<plist version="1.0">
-<dict>
-  <key>application-identifier</key><string>CHANGEME.com.example.wiicompiled</string>
-  <key>com.apple.developer.team-identifier</key><string>CHANGEME</string>
-  <key>get-task-allow</key><true/>
-  <key>com.apple.developer.kernel.increased-memory-limit</key><true/>
-  <key>com.apple.developer.kernel.extended-virtual-addressing</key><true/>
-</dict>
-</plist>
@@ -58,6 +58,7 @@ inline void Flush(bool force = false) {
 // still active.  A window close is an intentional successful exit, so end the
 // process directly and do not run the crash/atexit paths.
 [[noreturn]] inline void ExitForAuroraWindowClose() noexcept {
+    settings_overlay::ReleaseControllers();
    WindowPlacementPersistence::Flush(true);
 #if defined(_WIN32)
    ::ExitProcess(0);
@@ -1,38 +1,28 @@
 #pragma once

-#include "runtime_config.h"
+#include "nand_path.h"
+#include "nand_settings.h"

-#include <algorithm>
 #include <array>
-#include <cctype>
 #include <cstddef>
 #include <cstdint>
-#include <filesystem>
-#include <fstream>
 #include <iomanip>
-#include <optional>
-#include <random>
 #include <sstream>
 #include <string>
-#include <string_view>
 #include <utility>

 namespace RuntimeConsoleIdentity {

 struct Identity {
    std::string serial;
+    std::string productCode;
+    std::string area;
+    std::string gameRegion;
    std::array<uint8_t, 6> mac;
 };

-inline bool IsValidSerial(const std::string& serial) {
-    return serial.size() == 9 &&
-           serial != "000000000" &&
-           std::all_of(serial.begin(), serial.end(),
-                       [](unsigned char value) { return std::isdigit(value) != 0; });
-}
-
 inline Identity FromSerial(std::string serial) {
-    // Keep Nintendo's Wii OUI. The suffix is derived from the persisted serial
+    // Keep Nintendo's Wii OUI. The suffix is derived from the NAND serial
    // so every API exposes one coherent, stable virtual-console identity.
    uint32_t hash = 2166136261u;
    for (const unsigned char value : serial) {
@@ -46,6 +36,7 @@ inline Identity FromSerial(std::string serial) {

    return {
        std::move(serial),
+        {}, {}, {},
        {
            0x00,
            0x09,
@@ -57,83 +48,23 @@ inline Identity FromSerial(std::string serial) {
    };
 }

-inline std::optional<std::string> ReadSerial(const std::filesystem::path& path) {
-    std::ifstream input(path);
-    std::string line;
-    if (!input || !std::getline(input, line)) {
-        return std::nullopt;
+inline Identity LoadFromNand() {
+    const auto root = RuntimeNandPath::DiscoverNandRootPath();
+    const auto settings = RuntimeNandSettings::Read(root);
+    if (!settings || !RuntimeNandSettings::HasIdentity(*settings)) {
+        RuntimeNandPath::FailNandRoot(
+            "NAND setting.txt is missing or has invalid console identity fields (SERNO, CODE, AREA, GAME)",
+            root / "title/00000001/00000002/data/setting.txt");
    }
-    constexpr std::string_view prefix = "serial=";
-    if (line.rfind(prefix, 0) != 0) {
-        return std::nullopt;
-    }
-    std::string serial = line.substr(prefix.size());
-    if (!IsValidSerial(serial)) {
-        return std::nullopt;
-    }
-    return serial;
-}
-
-inline bool WriteSerial(const std::filesystem::path& path, const std::string& serial) {
-    std::error_code ec;
-    std::filesystem::create_directories(path.parent_path(), ec);
-    if (ec) {
-        return false;
-    }
-
-    std::filesystem::path temporary = path;
-    temporary += ".tmp";
-    {
-        std::ofstream output(temporary, std::ios::trunc);
-        if (!output) {
-            return false;
-        }
-        output << "serial=" << serial << '\n';
-        output.close();
-        if (!output) {
-            return false;
-        }
-    }
-
-    std::filesystem::rename(temporary, path, ec);
-    if (!ec) {
-        return true;
-    }
-    std::filesystem::remove(temporary, ec);
-    return false;
-}
-
-inline std::string GenerateSerial() {
-    std::random_device entropy;
-    std::seed_seq seed{
-        entropy(),
-        entropy(),
-        entropy(),
-        entropy(),
-    };
-    std::mt19937 generator(seed);
-    std::uniform_int_distribution<uint32_t> distribution(100000000u, 999999999u);
-    return std::to_string(distribution(generator));
-}
-
-inline Identity LoadOrCreate(const std::filesystem::path& path) {
-    if (const auto serial = ReadSerial(path)) {
-        return FromSerial(*serial);
-    }
-
-    const std::string generated = GenerateSerial();
-    if (WriteSerial(path, generated)) {
-        return FromSerial(generated);
-    }
-
-    // Remain operational in a read-only environment. This fallback matches
-    // Dolphin's deterministic serial while keeping the same valid identity shape.
-    return FromSerial("123456789");
+    Identity identity = FromSerial(settings->at("SERNO"));
+    identity.productCode = settings->at("CODE");
+    identity.area = settings->at("AREA");
+    identity.gameRegion = settings->at("GAME");
+    return identity;
 }

 inline const Identity& Current() {
-    static const Identity identity =
-        LoadOrCreate(RuntimeConfigFile::ApplicationDataDirectory() / "ConsoleIdentity.txt");
+    static const Identity identity = LoadFromNand();
    return identity;
 }

@@ -0,0 +1,170 @@
+#pragma once
+
+// The single vocabulary shared by everything that has to turn a Config.toml
+// controller name into a real button: the F10 settings bar, the macro engine,
+// and the startup mapping pass. Keeping one table here means a name that the
+// settings bar offers is always a name the config parser accepts, and vice
+// versa; the two used to drift because each side carried its own copy.
+
+#include <algorithm>
+#include <array>
+#include <cstdint>
+#include <string>
+#include <string_view>
+
+#include <SDL3/SDL_gamepad.h>
+#include <dolphin/pad.h>
+
+namespace ControllerNames {
+
+// A GameCube button as the game sees it, with the Config.toml key that selects
+// it. Order matches RuntimeConfigFile::kControllerButtonKeys.
+struct GameCubeButtonItem {
+    const char* configKey;
+    const char* label;
+    PADButton padButton;
+};
+
+inline constexpr std::array<GameCubeButtonItem, PAD_BUTTON_COUNT> kGameCubeButtons = {{
+    {"a", "A", PAD_BUTTON_A},
+    {"b", "B", PAD_BUTTON_B},
+    {"x", "X", PAD_BUTTON_X},
+    {"y", "Y", PAD_BUTTON_Y},
+    {"start", "Start", PAD_BUTTON_START},
+    {"z", "Z", PAD_TRIGGER_Z},
+    {"l", "L", PAD_TRIGGER_L},
+    {"r", "R", PAD_TRIGGER_R},
+    {"up", "D-pad Up", PAD_BUTTON_UP},
+    {"down", "D-pad Down", PAD_BUTTON_DOWN},
+    {"left", "D-pad Left", PAD_BUTTON_LEFT},
+    {"right", "D-pad Right", PAD_BUTTON_RIGHT},
+}};
+
+// A physical button on the host pad. Names are positional (south/east/...)
+// rather than Xbox-labelled so one config reads the same on any hardware.
+struct NativeButtonItem {
+    const char* configName;
+    const char* label;
+    uint32_t nativeButton;
+};
+
+inline constexpr auto kNativeButtons = std::to_array<NativeButtonItem>({
+    {"disabled", "Unmapped", PAD_NATIVE_BUTTON_DISABLED},
+    {"left_trigger", "Left trigger (LT / L2)", PADEncodeAxisButton(SDL_GAMEPAD_AXIS_LEFT_TRIGGER, false)},
+    {"right_trigger", "Right trigger (RT / R2)", PADEncodeAxisButton(SDL_GAMEPAD_AXIS_RIGHT_TRIGGER, false)},
+    {"left_stick_left", "Left stick left", PADEncodeAxisButton(SDL_GAMEPAD_AXIS_LEFTX, true)},
+    {"left_stick_right", "Left stick right", PADEncodeAxisButton(SDL_GAMEPAD_AXIS_LEFTX, false)},
+    {"left_stick_up", "Left stick up", PADEncodeAxisButton(SDL_GAMEPAD_AXIS_LEFTY, true)},
+    {"left_stick_down", "Left stick down", PADEncodeAxisButton(SDL_GAMEPAD_AXIS_LEFTY, false)},
+    {"right_stick_left", "Right stick left", PADEncodeAxisButton(SDL_GAMEPAD_AXIS_RIGHTX, true)},
+    {"right_stick_right", "Right stick right", PADEncodeAxisButton(SDL_GAMEPAD_AXIS_RIGHTX, false)},
+    {"right_stick_up", "Right stick up", PADEncodeAxisButton(SDL_GAMEPAD_AXIS_RIGHTY, true)},
+    {"right_stick_down", "Right stick down", PADEncodeAxisButton(SDL_GAMEPAD_AXIS_RIGHTY, false)},
+    {"unmapped", "Default", PAD_NATIVE_BUTTON_INVALID},
+    {"south", "South (A / Cross)", SDL_GAMEPAD_BUTTON_SOUTH},
+    {"east", "East (B / Circle)", SDL_GAMEPAD_BUTTON_EAST},
+    {"west", "West (X / Square)", SDL_GAMEPAD_BUTTON_WEST},
+    {"north", "North (Y / Triangle)", SDL_GAMEPAD_BUTTON_NORTH},
+    {"back", "Back / Select / Create", SDL_GAMEPAD_BUTTON_BACK},
+    {"guide", "Guide / Home / PS", SDL_GAMEPAD_BUTTON_GUIDE},
+    {"start", "Start / Options", SDL_GAMEPAD_BUTTON_START},
+    {"left_stick", "Left stick click (L3)", SDL_GAMEPAD_BUTTON_LEFT_STICK},
+    {"right_stick", "Right stick click (R3)", SDL_GAMEPAD_BUTTON_RIGHT_STICK},
+    {"left_shoulder", "Left bumper (LB / L1)", SDL_GAMEPAD_BUTTON_LEFT_SHOULDER},
+    {"right_shoulder", "Right bumper (RB / R1)", SDL_GAMEPAD_BUTTON_RIGHT_SHOULDER},
+    {"dpad_up", "D-pad Up", SDL_GAMEPAD_BUTTON_DPAD_UP},
+    {"dpad_down", "D-pad Down", SDL_GAMEPAD_BUTTON_DPAD_DOWN},
+    {"dpad_left", "D-pad Left", SDL_GAMEPAD_BUTTON_DPAD_LEFT},
+    {"dpad_right", "D-pad Right", SDL_GAMEPAD_BUTTON_DPAD_RIGHT},
+    {"misc1", "Misc 1 / Share / Mic", SDL_GAMEPAD_BUTTON_MISC1},
+    {"right_paddle1", "Right paddle 1", SDL_GAMEPAD_BUTTON_RIGHT_PADDLE1},
+    {"left_paddle1", "Left paddle 1", SDL_GAMEPAD_BUTTON_LEFT_PADDLE1},
+    {"right_paddle2", "Right paddle 2", SDL_GAMEPAD_BUTTON_RIGHT_PADDLE2},
+    {"left_paddle2", "Left paddle 2", SDL_GAMEPAD_BUTTON_LEFT_PADDLE2},
+    {"touchpad", "Touchpad click", SDL_GAMEPAD_BUTTON_TOUCHPAD},
+    {"misc2", "Misc 2", SDL_GAMEPAD_BUTTON_MISC2},
+    {"misc3", "Misc 3 / GC L click", SDL_GAMEPAD_BUTTON_MISC3},
+    {"misc4", "Misc 4 / GC R click", SDL_GAMEPAD_BUTTON_MISC4},
+    {"misc5", "Misc 5", SDL_GAMEPAD_BUTTON_MISC5},
+    {"misc6", "Misc 6", SDL_GAMEPAD_BUTTON_MISC6},
+});
+
+inline std::string TrimToken(std::string_view token) {
+    const size_t begin = token.find_first_not_of(" \t");
+    if (begin == std::string_view::npos) {
+        return {};
+    }
+    const size_t end = token.find_last_not_of(" \t");
+    return std::string(token.substr(begin, end - begin + 1));
+}
+
+inline const NativeButtonItem* FindNativeButton(std::string_view configName) {
+    const std::string name = TrimToken(configName.substr(0, configName.find('@')));
+    const auto it = std::find_if(kNativeButtons.begin(), kNativeButtons.end(),
+                                 [&](const NativeButtonItem& item) { return name == item.configName; });
+    return it == kNativeButtons.end() ? nullptr : &*it;
+}
+
+// Falls back to the "unmapped" entry so callers always have a label to draw.
+inline const NativeButtonItem& NativeButtonForValue(uint32_t nativeButton) {
+    const auto it = std::find_if(kNativeButtons.begin(), kNativeButtons.end(),
+                                 [&](const NativeButtonItem& item) { return PADAxisButtonIdentity(nativeButton) == PADAxisButtonIdentity(item.nativeButton); });
+    return it == kNativeButtons.end() ? *FindNativeButton("unmapped") : *it;
+}
+
+inline const GameCubeButtonItem* FindGameCubeButton(std::string_view configKey) {
+    const std::string key = TrimToken(configKey);
+    const auto it = std::find_if(kGameCubeButtons.begin(), kGameCubeButtons.end(),
+                                 [&](const GameCubeButtonItem& item) { return key == item.configKey; });
+    return it == kGameCubeButtons.end() ? nullptr : &*it;
+}
+
+// "up" or "up,a" -> the OR of those GC button bits. Unknown names are skipped so
+// a typo costs one button instead of the whole macro.
+inline uint16_t GameCubeMaskFromKeys(std::string_view keys) {
+    uint16_t mask = 0;
+    size_t begin = 0;
+    while (begin <= keys.size()) {
+        const size_t comma = keys.find(',', begin);
+        const std::string_view token =
+            keys.substr(begin, comma == std::string_view::npos ? std::string_view::npos : comma - begin);
+        if (const GameCubeButtonItem* item = FindGameCubeButton(token)) {
+            mask |= static_cast<uint16_t>(item->padButton);
+        }
+        if (comma == std::string_view::npos) {
+            break;
+        }
+        begin = comma + 1;
+    }
+    return mask;
+}
+
+inline std::string GameCubeKeysFromMask(uint16_t mask) {
+    std::string keys;
+    for (const auto& item : kGameCubeButtons) {
+        if ((mask & static_cast<uint16_t>(item.padButton)) == 0) {
+            continue;
+        }
+        if (!keys.empty()) {
+            keys += ',';
+        }
+        keys += item.configKey;
+    }
+    return keys;
+}
+
+inline std::string GameCubeLabelsFromMask(uint16_t mask) {
+    std::string labels;
+    for (const auto& item : kGameCubeButtons) {
+        if ((mask & static_cast<uint16_t>(item.padButton)) == 0) {
+            continue;
+        }
+        if (!labels.empty()) {
+            labels += " + ";
+        }
+        labels += item.label;
+    }
+    return labels.empty() ? std::string("None") : labels;
+}
+
+} // namespace ControllerNames
@@ -15,17 +15,6 @@ namespace GuestFlat {
 // of a global.
 inline constexpr uint64_t kGuestSpaceSize = 0x1'0000'0000ull;
 inline constexpr size_t kGuestPageSize = 0x1000;
-#ifdef MKW_PLATFORM_IOS
-// No fixed base works on every device: probing an iPhone 17 Pro and an iPad
-// Pro M5 gave disjoint sets of free 4 GiB windows (448 GiB only, versus 12-48
-// GiB), and the extended-virtual-addressing entitlement changes neither. So the
-// base is whatever the kernel hands out at Initialize().
-//
-// Note that memory_access.h routes every flat access to the checked path on
-// Apple, so this has no reader yet. It is groundwork, not a hot path.
-extern uint8_t* g_flatGuestBase;
-#define MKW_FLAT_GUEST_BASE (GuestFlat::g_flatGuestBase)
-#else
 #if defined(__x86_64__)
 // 16 TiB: clear of the Windows ASan shadow (32 TiB) and of the usual image/heap
 // placement.
@@ -51,10 +40,7 @@ inline constexpr uintptr_t kFixedFlatGuestBase = 0x0000'0010'0000'0000ull;
 #error "guest_flat_memory.h has no fixed flat guest base chosen for this architecture"
 #endif

-// Fixed base so the emitted access is `[reg + imm64-in-register]` with no load
-// of a global.
 #define MKW_FLAT_GUEST_BASE (reinterpret_cast<uint8_t*>(GuestFlat::kFixedFlatGuestBase))
-#endif

 enum class Backing {
    Owned,
@@ -0,0 +1,67 @@
+#pragma once
+
+// Per-port expression bindings for the GameCube controls, plus import of a
+// Dolphin GCPadNew.ini.
+
+#include <array>
+#include <cstdint>
+#include <string>
+
+#include <dolphin/pad.h>
+
+namespace InputBindings {
+
+// The controls an expression can drive, in Dolphin's own naming so an
+// imported config maps across without translation.
+struct ControlInfo {
+    const char* dolphinName;
+    const char* label;
+    uint16_t padButton;   // 0 for the analog-only controls below
+    int analog;           // 0 none, 1 trigger L, 2 trigger R
+};
+
+inline constexpr std::array<ControlInfo, 14> kControls = {{
+    {"Buttons/A", "A", PAD_BUTTON_A, 0},
+    {"Buttons/B", "B", PAD_BUTTON_B, 0},
+    {"Buttons/X", "X", PAD_BUTTON_X, 0},
+    {"Buttons/Y", "Y", PAD_BUTTON_Y, 0},
+    {"Buttons/Z", "Z", PAD_TRIGGER_Z, 0},
+    {"Buttons/Start", "Start", PAD_BUTTON_START, 0},
+    {"D-Pad/Up", "D-pad Up", PAD_BUTTON_UP, 0},
+    {"D-Pad/Down", "D-pad Down", PAD_BUTTON_DOWN, 0},
+    {"D-Pad/Left", "D-pad Left", PAD_BUTTON_LEFT, 0},
+    {"D-Pad/Right", "D-pad Right", PAD_BUTTON_RIGHT, 0},
+    {"Triggers/L", "L", PAD_TRIGGER_L, 1},
+    {"Triggers/R", "R", PAD_TRIGGER_R, 2},
+    {"Triggers/L-Analog", "L analog", 0, 1},
+    {"Triggers/R-Analog", "R analog", 0, 2},
+}};
+
+void Reload() noexcept;
+
+// The pad library has PADBlockInput but no matching query, so the settings
+// overlay reports its own state here.
+void SetInputBlocked(bool blocked) noexcept;
+bool InputBlocked() noexcept;
+
+// Mix expression output into a freshly read status set. Call once per guest
+// PADRead, after every other input source has been merged.
+void Apply(PADStatus* statuses) noexcept;
+
+std::string GetExpression(uint32_t port, size_t control) noexcept;
+// Returns false and fills error if the text does not parse; the binding is
+// left unchanged in that case.
+bool SetExpression(uint32_t port, size_t control, const std::string& text, std::string& error) noexcept;
+
+// True while the control's expression is above the press threshold.
+bool IsActive(uint32_t port, size_t control) noexcept;
+
+// The default Dolphin config location on Windows, then next to the executable.
+std::string DefaultDolphinConfigPath() noexcept;
+
+// Imports [GCPad<padIndex>] into the given port. Returns the number of controls
+// imported, or -1 on failure with error filled.
+int ImportDolphinConfig(const std::string& path, int padIndex, uint32_t port,
+                        std::string& summary, std::string& error) noexcept;
+
+} // namespace InputBindings
@@ -0,0 +1,53 @@
+#pragma once
+
+// Dolphin-compatible input expressions.
+//
+// Values are doubles in Dolphin's ControlState style; a control counts as
+// pressed above kConditionThreshold. Timing matches Dolphin: wall-clock
+// seconds on a steady clock, so an expression copied from GCPadNew.ini
+// behaves the same here as it does there.
+
+#include <filesystem>
+#include <functional>
+#include <memory>
+#include <string>
+#include <vector>
+
+namespace InputExpr {
+
+inline constexpr double kConditionThreshold = 0.5;
+
+// Resolves a backtick-quoted input name to its current value.
+using InputSource = std::function<double(const std::string&)>;
+
+struct Node;
+
+class Expression {
+public:
+    Expression();
+    ~Expression();
+    Expression(Expression&&) noexcept;
+    Expression& operator=(Expression&&) noexcept;
+
+    // Returns false and fills error on a syntax problem.
+    static bool Parse(const std::string& text, Expression& out, std::string& error);
+
+    bool Empty() const { return m_root == nullptr; }
+    double Evaluate(const InputSource& source) const;
+
+    // Input names the expression references, for diagnostics.
+    std::vector<std::string> ReferencedInputs() const;
+
+private:
+    std::unique_ptr<Node> m_root;
+};
+
+// Parses a Dolphin GCPadNew.ini and returns the expression text for each
+// control of the requested pad, keyed by Dolphin's own control names
+// ("Buttons/A", "D-Pad/Up", "Triggers/L", ...). Returns false if the file
+// cannot be read or the section is missing.
+bool ReadDolphinConfig(const std::filesystem::path& path, int padIndex,
+                       std::vector<std::pair<std::string, std::string>>& controls,
+                       std::string& deviceName, std::string& error);
+
+} // namespace InputExpr
@@ -1,6 +1,7 @@
 #pragma once

 #include "runtime_config.h"
+#include "nand_settings.h"
 #include "runtime_log.h"
 #include "system_bridge.h"

@@ -163,7 +164,7 @@ inline std::filesystem::path CreateManagedNandRoot() {
    return root;
 }

-inline std::filesystem::path DiscoverNandRootPath() {
+inline std::filesystem::path ResolveNandRootPath() {
    const std::string configPath = RuntimeConfigFile::NandRoot();
    if (!configPath.empty()) {
        const auto path = ResolveConfiguredPath(configPath);
@@ -179,4 +180,16 @@ inline std::filesystem::path DiscoverNandRootPath() {
    return CreateManagedNandRoot();
 }

+inline std::filesystem::path DiscoverNandRootPath() {
+    static const auto root = [] {
+        const auto resolved = ResolveNandRootPath();
+        std::string error;
+        if (!RuntimeNandSettings::Ensure(resolved, error)) {
+            FailNandRoot(error.c_str(), RuntimeNandSettings::FilePath(resolved));
+        }
+        return resolved;
+    }();
+    return root;
+}
+
 } // namespace RuntimeNandPath
@@ -0,0 +1,59 @@
+#pragma once
+
+#include <filesystem>
+#include <fstream>
+#include <istream>
+
+namespace RuntimeNandSave {
+
+enum class Contents { Missing, Blank, Nonzero, Error };
+enum class ReadAction { Proceed, Missing, Error, RecoveryNeeded };
+
+// A failed read is not evidence that a save is blank. Check badbit before EOF:
+// an I/O failure may set both, whereas a successful short final read sets EOF.
+inline Contents InspectStream(std::istream& input) {
+    if (!input) return Contents::Error;
+    char block[4096];
+    for (;;) {
+        input.read(block, sizeof(block));
+        if (input.bad() || (input.fail() && !input.eof())) return Contents::Error;
+        for (std::streamsize i = 0; i < input.gcount(); ++i) {
+            if (block[i] != 0) return Contents::Nonzero;
+        }
+        if (input.eof()) return Contents::Blank;
+    }
+}
+
+inline Contents InspectFile(const std::filesystem::path& path) {
+    std::error_code ec;
+    const auto status = std::filesystem::symlink_status(path, ec);
+    if (ec && ec != std::errc::no_such_file_or_directory) return Contents::Error;
+    if (!std::filesystem::exists(status)) return Contents::Missing;
+    if (!std::filesystem::is_regular_file(path, ec) || ec) return Contents::Error;
+    std::ifstream input(path, std::ios::binary);
+    return InspectStream(input);
+}
+
+// Probe only read-only opens of the actual save and its exact write shadow.
+// No probe writes, removes, or repairs data, and backups are not save aliases.
+inline ReadAction CheckRead(const std::filesystem::path& path, int mode) {
+    const auto name = path.filename();
+    const bool isMain = name == "rksys.dat";
+    if (mode != 1 || (!isMain && name != "rksys.dat.nandsafe.tmp")) return ReadAction::Proceed;
+    const auto contents = InspectFile(path);
+    if (contents == Contents::Error) return ReadAction::Error;
+    if (contents == Contents::Nonzero) return ReadAction::Proceed;
+    if (isMain) {
+        auto shadow = path;
+        shadow += ".nandsafe.tmp";
+        const auto shadowContents = InspectFile(shadow);
+        if (shadowContents == Contents::Error) return ReadAction::Error;
+        // The next write normally discards an old shadow. Preserve a possible
+        // recovery source when there is no usable original, without promoting
+        // an uncommitted (and potentially incomplete) shadow to the real save.
+        if (shadowContents == Contents::Nonzero) return ReadAction::RecoveryNeeded;
+    }
+    return contents == Contents::Blank ? ReadAction::Missing : ReadAction::Proceed;
+}
+
+} // namespace RuntimeNandSave
@@ -0,0 +1,220 @@
+#pragma once
+
+#include <array>
+#include <atomic>
+#include <chrono>
+#include <ctime>
+#include <cstdint>
+#include <filesystem>
+#include <fstream>
+#include <map>
+#include <optional>
+#include <string>
+#include <utility>
+
+#ifdef _WIN32
+#include <windows.h>
+#else
+#include <unistd.h>
+#endif
+
+namespace RuntimeNandSettings {
+
+using Settings = std::map<std::string, std::string>;
+
+inline std::filesystem::path FilePath(const std::filesystem::path& root) {
+    return root / "title/00000001/00000002/data/setting.txt";
+}
+
+// Wii setting.txt is a 256-byte buffer encrypted with a rotating XOR key.
+inline std::optional<Settings> Read(const std::filesystem::path& nandRoot) {
+    std::ifstream input(FilePath(nandRoot), std::ios::binary);
+    std::array<uint8_t, 256> bytes{};
+    if (!input.read(reinterpret_cast<char*>(bytes.data()), bytes.size())) {
+        return std::nullopt;
+    }
+    uint32_t key = 0x73B5DBFAu;
+    std::string decoded;
+    for (const uint8_t byte : bytes) {
+        const char value = static_cast<char>(byte ^ static_cast<uint8_t>(key));
+        key = (key << 1) | (key >> 31);
+        if (value == '\0') {
+            break;
+        }
+        if (value != '\r') {
+            decoded += value;
+        }
+    }
+    Settings settings;
+    for (size_t start = 0; start < decoded.size();) {
+        const size_t end = decoded.find('\n', start);
+        const std::string line = decoded.substr(start, end - start);
+        const size_t equals = line.find('=');
+        if (equals != std::string::npos && equals != 0) {
+            settings.emplace(line.substr(0, equals), line.substr(equals + 1));
+        }
+        if (end == std::string::npos) {
+            break;
+        }
+        start = end + 1;
+    }
+    return settings;
+}
+
+inline bool HasIdentity(const Settings& settings) {
+    const auto serial = settings.find("SERNO");
+    if (serial == settings.end() || serial->second.empty() || serial->second.size() > 9 ||
+        serial->second.find_first_not_of("0123456789") != std::string::npos ||
+        serial->second.find_first_not_of('0') == std::string::npos) {
+        return false;
+    }
+    for (const auto& field : {std::pair{"CODE", 5u}, {"AREA", 3u}, {"GAME", 2u}}) {
+        const auto value = settings.find(field.first);
+        if (value == settings.end() || value->second.empty() ||
+            value->second.size() > field.second) {
+            return false;
+        }
+    }
+    return true;
+}
+
+// Dolphin's normal (non-deterministic) first-boot algorithm. It is independent
+// of the ES device ID. Matching another NAND requires that NAND's saved serial.
+inline std::string GenerateSerial(std::time_t now) {
+    if (now < 0) {
+        return {};
+    }
+    const auto digits = std::to_string(now % 1000000000);
+    return std::string(9 - digits.size(), '0') + digits;
+}
+
+// This recompilation targets the European disc. These are Dolphin's PAL boot
+// defaults; an existing setting.txt always takes precedence, in every region.
+inline std::optional<std::array<uint8_t, 256>> EncodeNew(const std::string& serial) {
+    const Settings identity{{"SERNO", serial}, {"CODE", "LEH"}, {"AREA", "EUR"}, {"GAME", "EU"}};
+    if (!HasIdentity(identity)) {
+        return std::nullopt;
+    }
+    std::array<uint8_t, 256> bytes{};
+    size_t position = 0;
+    uint32_t key = 0x73B5DBFAu;
+    const auto writeByte = [&](char value) {
+        bytes[position++] = static_cast<uint8_t>(value) ^ static_cast<uint8_t>(key);
+        key = (key << 1) | (key >> 31);
+    };
+    for (const std::string& line : {std::string("AREA=EUR\r\n"), std::string("MODEL=RVL-001(EUR)\r\n"),
+             std::string("DVD=0\r\n"), std::string("MPCH=0x7FFE\r\n"), std::string("CODE=LEH\r\n"),
+             "SERNO=" + serial + "\r\n", std::string("VIDEO=PAL\r\n"), std::string("GAME=EU\r\n")}) {
+        for (;;) {
+            if (position + line.size() > bytes.size()) {
+                return std::nullopt;
+            }
+            const auto start = position;
+            const auto savedKey = key;
+            bool hasNull = false;
+            for (const char value : line) {
+                writeByte(value);
+                hasNull |= bytes[position - 1] == 0;
+            }
+            if (!hasNull) {
+                break;
+            }
+            // Nintendo stops at an encoded NUL. Dolphin inserts an extra LF
+            // before this line and retries with the shifted encryption key.
+            position = start;
+            key = savedKey;
+            writeByte('\n');
+        }
+    }
+    return bytes; // The unused tail stays raw zero, as in Dolphin.
+}
+
+// Atomically claim our own scratch directory. A collision belongs to another
+// launch (or a previous crashed launch); leave it untouched and try another name.
+inline std::optional<std::filesystem::path> CreateScratchDirectory(
+    const std::filesystem::path& parent, const std::string& token, std::error_code& ec) {
+    for (unsigned attempt = 0; attempt < 128; ++attempt) {
+        const auto candidate = parent / (".setting-init-" + token + "-" + std::to_string(attempt));
+        ec.clear();
+        if (std::filesystem::create_directory(candidate, ec)) return candidate;
+        if (ec && ec != std::errc::file_exists) return std::nullopt;
+    }
+    ec = std::make_error_code(std::errc::file_exists);
+    return std::nullopt;
+}
+
+// Never replace an existing file, including an unreadable or damaged one.
+// Publish a complete file atomically so simultaneous launches use one identity.
+inline bool Ensure(const std::filesystem::path& root, std::string& error,
+                   std::time_t now = std::time(nullptr)) {
+    const auto path = FilePath(root);
+    std::error_code ec;
+    const auto status = std::filesystem::symlink_status(path, ec);
+    if (ec && ec != std::errc::no_such_file_or_directory) {
+        error = "Cannot inspect NAND setting.txt: " + ec.message();
+        return false;
+    }
+    if (std::filesystem::exists(status)) {
+        const auto existing = Read(root);
+        if (existing && HasIdentity(*existing)) {
+            return true;
+        }
+        error = "Existing NAND setting.txt is unreadable or invalid; restore it from this console's backup";
+        return false;
+    }
+
+    const auto bytes = EncodeNew(GenerateSerial(now));
+    if (!bytes) {
+        error = "Cannot initialize NAND settings: invalid system clock";
+        return false;
+    }
+    ec.clear();
+    std::filesystem::create_directories(path.parent_path(), ec);
+    if (ec) {
+        error = "Cannot create NAND settings directory: " + ec.message();
+        return false;
+    }
+    static std::atomic<unsigned> sequence{0};
+#ifdef _WIN32
+    const auto processId = GetCurrentProcessId();
+#else
+    const auto processId = getpid();
+#endif
+    const auto scratch = CreateScratchDirectory(path.parent_path(),
+        std::to_string(processId) + "-" + std::to_string(
+            std::chrono::steady_clock::now().time_since_epoch().count()) + "-" +
+            std::to_string(sequence++), ec);
+    if (!scratch) {
+        error = "Cannot create temporary NAND settings directory: " + ec.message();
+        return false;
+    }
+    const auto temporary = *scratch / "setting.txt";
+    bool written = false;
+    {
+        std::ofstream output(temporary, std::ios::binary);
+        output.write(reinterpret_cast<const char*>(bytes->data()), bytes->size());
+        output.close();
+        written = static_cast<bool>(output);
+    }
+    bool published = false;
+    if (written) {
+#ifdef _WIN32
+        published = MoveFileExW(temporary.c_str(), path.c_str(), MOVEFILE_WRITE_THROUGH) != 0;
+#else
+        published = ::link(temporary.c_str(), path.c_str()) == 0;
+#endif
+    }
+    std::filesystem::remove(temporary, ec);
+    std::filesystem::remove(*scratch, ec);
+    // A competing launcher may have published its settings first. Always read
+    // the winner from NAND rather than using our unpersisted candidate serial.
+    const auto persisted = Read(root);
+    if (persisted && HasIdentity(*persisted)) {
+        return true;
+    }
+    error = published ? "Cannot read newly initialized NAND setting.txt" :
+                        "Cannot persist NAND setting.txt; check NAND directory permissions";
+    return false;
+}
+
+} // namespace RuntimeNandSettings
@@ -11,6 +11,7 @@
 #include <iomanip>
 #include <iostream>
 #include <limits>
+#include <map>
 #include <optional>
 #include <sstream>
 #include <string>
@@ -89,8 +90,9 @@ struct RuntimeUserConfig {
    // comma-separated SDL-style physical button names ("south", or
    // "dpad_up,left_shoulder") as values; pressing either bound button counts.
    std::array<std::optional<std::string>, 12> controllerButtons;
-    // On-screen mobile touch controls: auto-accelerate locks A after a 1-second hold.
-    std::optional<bool> touchAutoAccelerate;
+    std::optional<bool> rumbleEnabled;
+    std::optional<int32_t> muteHotkey;
+    std::map<std::string, std::string> controllerExpressions;
 };

 namespace RuntimeConfigFile {
@@ -276,19 +278,7 @@ inline std::filesystem::path ApplicationDataDirectory() {
    return std::filesystem::current_path() / kApplicationDirectoryName;
 }

-inline std::filesystem::path& ConfigPathOverride() {
-    static std::filesystem::path path;
-    return path;
-}
-
-inline void SetConfigPathOverride(std::filesystem::path path) {
-    ConfigPathOverride() = std::move(path);
-}
-
 inline std::filesystem::path ResolveConfigPath() {
-    if (!ConfigPathOverride().empty()) {
-        return ConfigPathOverride();
-    }
    return ApplicationDataDirectory() / kConfigFileName;
 }

@@ -335,9 +325,6 @@ inline void EnsureConfigFile() {
              "# guest can observe it. Set to false to mix inline on the guest\n"
              "# thread exactly as the runtime did before.\n"
              "mix_worker = true\n\n"
-              "[controller]\n"
-              "# On-screen touch controls auto-accelerate latch (hold A for 1 second to lock)\n"
-              "touch_auto_accelerate = true\n\n"
              "[network]\n"
              "enabled = true\n\n"
              "[discord]\n"
@@ -424,6 +411,20 @@ inline RuntimeUserConfig ParseConfigDocument(const toml::value& document) {
            FindConfigValue<std::string>(document, "controller", buttonKeys[index]);
    }

+    config.rumbleEnabled = FindConfigValue<bool>(document, "controller", "rumble");
+    if (auto value = FindConfigInt(document, "audio", "mute_key")) {
+        config.muteHotkey = *value;
+    }
+
+    if (const auto* section = document.contains("controller") ? &document.at("controller") : nullptr;
+        section != nullptr && section->is_table()) {
+        for (const auto& [key, value] : section->as_table()) {
+            if (key.rfind("expr_", 0) == 0 && value.is_string()) {
+                config.controllerExpressions[key] = value.as_string();
+            }
+        }
+    }
+
    config.widescreen = FindConfigValue<bool>(document, "video", "widescreen");
    config.windowPosX = FindConfigInt(document, "video", "window_x");
    config.windowPosY = FindConfigInt(document, "video", "window_y");
@@ -484,8 +485,6 @@ inline RuntimeUserConfig ParseConfigDocument(const toml::value& document) {
    config.wiiAccelOffsetY = FindConfigValue<double>(document, "controller", "wii_accel_offset_y");
    config.wiiAccelOffsetZ = FindConfigValue<double>(document, "controller", "wii_accel_offset_z");
    config.wiiAccelTrace = FindConfigValue<bool>(document, "controller", "wii_accel_trace");
-    config.touchAutoAccelerate =
-        FindConfigValue<bool>(document, "controller", "touch_auto_accelerate");
    config.networkEnabled = FindConfigValue<bool>(document, "network", "enabled");
    config.discordPresenceEnabled = FindConfigValue<bool>(document, "discord", "enabled");
    config.discordClientId = FindConfigValue<std::string>(document, "discord", "client_id");
@@ -696,6 +695,34 @@ inline bool SetControllerButton(size_t index, std::string value) {
    return WriteSetting("controller", kControllerButtonKeys[index], FormatString(value));
 }

+inline std::string ControllerExpression(const std::string& key) {
+    const auto it = Get().controllerExpressions.find(key);
+    return it == Get().controllerExpressions.end() ? std::string() : it->second;
+}
+
+inline bool SetControllerExpression(const std::string& key, const std::string& value) {
+    Mutable().controllerExpressions[key] = value;
+    return WriteSetting("controller", key, FormatString(value));
+}
+
+inline bool RumbleEnabled(bool fallback = true) {
+    return Get().rumbleEnabled.value_or(fallback);
+}
+
+inline bool SetRumbleEnabled(bool value) {
+    Mutable().rumbleEnabled = value;
+    return WriteSetting("controller", "rumble", value ? "true" : "false");
+}
+
+inline int32_t MuteHotkey(int32_t fallback) {
+    return Get().muteHotkey.value_or(fallback);
+}
+
+inline bool SetMuteHotkey(int32_t value) {
+    Mutable().muteHotkey = value;
+    return WriteSetting("audio", "mute_key", std::to_string(value));
+}
+
 inline bool SetAudioVolume(float value) {
    value = std::clamp(value, 0.0f, 1.0f);
    Mutable().audioVolume = value;
@@ -868,16 +895,6 @@ inline bool SetWiiAccelOffset(const std::array<double, 3>& offset) {
    return ok;
 }

-// On-screen mobile touch controls: auto-accelerate locks A after a 1-second hold.
-inline bool TouchAutoAccelerate(bool fallback = true) {
-    return Get().touchAutoAccelerate.value_or(fallback);
-}
-
-inline bool SetTouchAutoAccelerate(bool value) {
-    Mutable().touchAutoAccelerate = value;
-    return WriteSetting("controller", "touch_auto_accelerate", value ? "true" : "false");
-}
-
 // Target frame rate for frame interpolation, or 0 to disable it.
 inline uint32_t FrameInterpolationFps(uint32_t fallback = 0) {
    return Get().frameInterpolationFps.value_or(fallback);
@@ -1030,9 +1047,6 @@ inline void LogLoadedConfig() {
            if (config.retroRewindRoot) {
                std::cout << " retro_rewind_root=" << *config.retroRewindRoot;
            }
-            if (config.touchAutoAccelerate) {
-                std::cout << " touch_auto_accelerate=" << (*config.touchAutoAccelerate ? "true" : "false");
-            }
        }
        std::cout << std::endl;
        return true;
--- a/Show More
+++ b/Show More