Hook SpVoice to fix narrator (#10)

windows/Sources/Hook/Hook.cpp

@@ -4,10 +4,17 @@
 #include <Detours.h>
 #include <string>
 #include <stdexcept>
+#include <sapi.h>
+#include <AtlBase.h>
+#include <AtlConv.h>
+#include <AtlCom.h>
 
 using namespace std::string_literals;
 #pragma clang diagnostic ignored "-Wmicrosoft-cast"
 
+#define VTABLE_INDEX_SPEAK 20
+#define VTABLE_INDEX_SKIP 23
+
 // Workaround for GetVolumeInformationW not working in a UWP application
 // This is by creating a handle to a file on the drive and then using GetVolumeInformationByHandleW with that handle
 static BOOL (WINAPI* TrueGetVolumeInformationW)(LPCWSTR, LPWSTR, DWORD, LPDWORD, LPDWORD, LPDWORD, LPWSTR, DWORD) = GetVolumeInformationW;
@@ -93,6 +100,39 @@ BOOL WINAPI SetCursorPosPatch(int x, int y) {
     return true;
 }
 
+HRESULT __stdcall SpeakPatch(ISpVoice* This, LPCWSTR pwcs, DWORD dwFlags, ULONG *pulStreamNumber) {
+    CW2A utf8(pwcs, CP_UTF8);
+    Runtime::speak(utf8.m_psz, dwFlags);
+    return S_OK;
+}
+
+HRESULT __stdcall SpeakSkipPatch(ISpVoice* This, LPCWSTR *pItemType, long lItems, ULONG *pulNumSkipped) {
+    Runtime::speakSkip();
+    return S_OK;
+}
+
+struct _ISpVoiceVTable {
+    void* speak;
+    void* skip;
+};
+
+_ISpVoiceVTable createISpVoiceVTable() {
+    CoInitializeEx(nullptr, 0);
+
+    CComPtr<ISpVoice> spVoice;
+    if (!SUCCEEDED(spVoice.CoCreateInstance(CLSID_SpVoice))) {
+        throw std::runtime_error("Failed to create ISpVoice instance");
+    }
+
+    auto vTable = *(void***)spVoice.p;
+    void* speak = vTable[VTABLE_INDEX_SPEAK];
+    void* skip = vTable[VTABLE_INDEX_SKIP];
+
+    return {speak, skip};
+}
+
+static _ISpVoiceVTable spVoiceVTable = createISpVoiceVTable();
+
 BOOL WINAPI DllMain(HINSTANCE hinst, DWORD dwReason, LPVOID reserved) {
     if (DetourIsHelperProcess()) {
         return true;
@@ -106,13 +146,17 @@ BOOL WINAPI DllMain(HINSTANCE hinst, DWORD dwReason, LPVOID reserved) {
         DetourAttach(&(PVOID&)TrueGetVolumeInformationW, GetVolumeInformationWPatch);
         DetourAttach(&(PVOID&)TrueClipCursor, ClipCursorPatch);
         DetourAttach(&(PVOID&)TrueSetCursorPos, SetCursorPosPatch);
+        DetourAttach(&(PVOID&)spVoiceVTable.speak, SpeakPatch);
+        DetourAttach(&(PVOID&)spVoiceVTable.skip, SpeakSkipPatch);
         DetourTransactionCommit();
     } else if (dwReason == DLL_PROCESS_DETACH) {
         DetourTransactionBegin();
         DetourUpdateThread(GetCurrentThread());
         DetourDetach(&(PVOID&)TrueGetVolumeInformationW, GetVolumeInformationWPatch);
         DetourDetach(&(PVOID&)TrueClipCursor, ClipCursorPatch);
         DetourDetach(&(PVOID&)TrueSetCursorPos, SetCursorPosPatch);
+        DetourDetach(&(PVOID&)spVoiceVTable.speak, SpeakPatch);
+        DetourDetach(&(PVOID&)spVoiceVTable.skip, SpeakSkipPatch);
         DetourTransactionCommit();
         Runtime::processDetach();
     }

windows/Sources/SandboxTest/SandboxTest.swift

@@ -10,6 +10,7 @@ class SandboxTest {
     "namedPipe": NamedPipeCommand(),
     "nameMax": NameMaxCommand(),
     "mouseMovements": MouseMovementsCommand(),
+    "speech": SpeechCommand(),
   ]
 
   static func main() throws {

windows/Sources/SandboxTest/SpeechCommand.swift

@@ -0,0 +1,36 @@
+import WinSDK
+import WindowsUtils
+import WinSDKExtras
+import CxxStdlib
+
+// Enable to run the full test, with a long message that will get cut off.
+let FULL_TEST = false
+
+// Use the win32 Speech API to convert text to speech
+class SpeechCommand: Command {
+  private static let SPF_ASYNC = DWORD(1)
+  private static let SPF_IS_NOT_XML = DWORD(1 << 4)
+
+  func execute(_ arguments: [String]) throws {
+    CoInitializeEx(nil, 0)
+    defer {
+      CoUninitialize()
+    }
+
+    var speak = SpeakApi()
+    let flags: DWORD = SpeechCommand.SPF_ASYNC | SpeechCommand.SPF_IS_NOT_XML
+    let message = FULL_TEST ? "Hello Fabric, Sandbox. This is a long message that will get cut off" : ""
+    let result = speak.Speak(std.string(message), flags)
+    guard result == S_OK else {
+       throw Win32Error("Failed to speak", result: result)
+    }
+
+    if FULL_TEST {
+      Sleep(2000)
+    }
+
+    speak.Skip()
+
+    print("Spoke")
+  }
+}

windows/Tests/IntergrationTests.swift

@@ -120,6 +120,14 @@ import WindowsUtils
       ["mouseMovements", "-Dsandbox.namedPipe=\(server.path)"], namedPipe: server)
     #expect(exitCode == 0)
   }
+
+  @Test func testSpeech() throws {
+    let server = try SandboxNamedPipeServer(
+      pipeName: "\\\\.\\pipe\\FabricSandbox" + randomString(length: 10))
+    let (exitCode, output) = try runIntergration(["speech"], namedPipe: server)
+    #expect(exitCode == 0)
+    #expect(output == "Spoke")
+  }
 }
 func runIntergration(
   _ args: [String], capabilities: [SidCapability] = [], filePermissions: [FilePermission] = [],
@@ -164,14 +172,19 @@ func runIntergration(
       accessPermissions: [.genericRead, .genericExecute])
   }
 
+  var commandLine = [testExecutable.path()] + args
+
   if let namedPipe = namedPipe {
     try grantNamedPipeAccess(
       pipe: namedPipe, appContainer: container, accessPermissions: [.genericRead, .genericWrite])
+    if namedPipe is SandboxNamedPipeServer {
+      commandLine.append("-Dsandbox.namedPipe=\(namedPipe.path)")
+    }
   }
 
   let outputConsumer = TestOutputConsumer()
   let process = SandboxedProcess(
-    application: testExecutable, commandLine: [testExecutable.path()] + args,
+    application: testExecutable, commandLine: commandLine,
     workingDirectory: workingDirectory, container: container, outputConsumer: outputConsumer)
   let exitCode = try process.run()
   return (exitCode, outputConsumer.trimmed())