browser(firefox): implement screencast (#6452)

browser_patches/firefox/BUILD_NUMBER

@@ -1,2 +1,2 @@
-1254
+1255
-Changed: lushnikov@chromium.org Thu 06 May 2021 11:02:00 AM PDT
+Changed: pavel.feldman@gmail.com Thu 06 May 2021 06:33:59 PM PDT

browser_patches/firefox/juggler/TargetRegistry.js

@@ -159,8 +159,8 @@ class TargetRegistry {
       target.updateColorSchemeOverride();
       if (!hasExplicitSize)
         target.updateViewportSize();
-      if (browserContext.screencastOptions)
+      if (browserContext.videoRecordingOptions)
-        target._startVideoRecording(browserContext.screencastOptions);
+        target._startVideoRecording(browserContext.videoRecordingOptions);
     };
 
     const onTabCloseListener = event => {
@@ -333,7 +333,8 @@ class PageTarget {
     this._url = 'about:blank';
     this._openerId = opener ? opener.id() : undefined;
     this._channel = SimpleChannel.createForMessageManager(`browser::page[${this._targetId}]`, this._linkedBrowser.messageManager);
-    this._screencastInfo = undefined;
+    this._videoRecordingInfo = undefined;
+    this._screencastRecordingInfo = undefined;
     this._dialogs = new Map();
 
     const navigationListener = {
@@ -352,6 +353,7 @@ class PageTarget {
     this._registry._browserBrowsingContextToTarget.set(this._linkedBrowser.browsingContext, this);
 
     this._registry.emit(TargetRegistry.Events.TargetCreated, this);
+    this._screencast = Cc['@mozilla.org/juggler/screencast;1'].getService(Ci.nsIScreencastService);
   }
 
   dialog(dialogId) {
@@ -497,43 +499,69 @@ class PageTarget {
     if (width < 10 || width > 10000 || height < 10 || height > 10000)
       throw new Error("Invalid size");
 
-    const screencast = Cc['@mozilla.org/juggler/screencast;1'].getService(Ci.nsIScreencastService);
     const docShell = this._gBrowser.ownerGlobal.docShell;
     // Exclude address bar and navigation control from the video.
     const rect = this.linkedBrowser().getBoundingClientRect();
     const devicePixelRatio = this._window.devicePixelRatio;
-    const videoSessionId = screencast.startVideoRecording(docShell, file, width, height, devicePixelRatio * rect.top);
+    const sessionId = this._screencast.startVideoRecording(docShell, true, file, width, height, 0, devicePixelRatio * rect.top);
-    this._screencastInfo = { videoSessionId, file };
+    this._videoRecordingInfo = { sessionId, file };
     this.emit(PageTarget.Events.ScreencastStarted);
   }
 
-  async _stopVideoRecording() {
+  _stopVideoRecording() {
-    if (!this._screencastInfo)
+    if (!this._videoRecordingInfo)
       throw new Error('No video recording in progress');
-    const screencastInfo = this._screencastInfo;
+    const videoRecordingInfo = this._videoRecordingInfo;
-    this._screencastInfo = undefined;
+    this._videoRecordingInfo = undefined;
-    const screencast = Cc['@mozilla.org/juggler/screencast;1'].getService(Ci.nsIScreencastService);
+    this._screencast.stopVideoRecording(videoRecordingInfo.sessionId);
-    const result = new Promise(resolve =>
-      Services.obs.addObserver(function onStopped(subject, topic, data) {
-        if (screencastInfo.videoSessionId != data)
-          return;
-
-        Services.obs.removeObserver(onStopped, 'juggler-screencast-stopped');
-        resolve();
-      }, 'juggler-screencast-stopped')
-    );
-    screencast.stopVideoRecording(screencastInfo.videoSessionId);
-    return result;
   }
 
-  screencastInfo() {
+  videoRecordingInfo() {
-    return this._screencastInfo;
+    return this._videoRecordingInfo;
+  }
+
+  async startScreencast({ width, height, quality }) {
+    // On Mac the window may not yet be visible when TargetCreated and its
+    // NSWindow.windowNumber may be -1, so we wait until the window is known
+    // to be initialized and visible.
+    await this.windowReady();
+    if (width < 10 || width > 10000 || height < 10 || height > 10000)
+      throw new Error("Invalid size");
+
+    const docShell = this._gBrowser.ownerGlobal.docShell;
+    // Exclude address bar and navigation control from the video.
+    const rect = this.linkedBrowser().getBoundingClientRect();
+    const devicePixelRatio = this._window.devicePixelRatio;
+    const screencastId = this._screencast.startVideoRecording(docShell, false, '', width, height, quality || 90, devicePixelRatio * rect.top);
+    const onFrame = (subject, topic, data) => {
+      this.emit(PageTarget.Events.ScreencastFrame, data);
+    };
+    Services.obs.addObserver(onFrame, 'juggler-screencast-frame');
+    this._screencastRecordingInfo = { screencastId, onFrame };
+    return { screencastId };
+  }
+
+  screencastFrameAck({ screencastId }) {
+    if (!this._screencastRecordingInfo || this._screencastRecordingInfo.screencastId !== screencastId)
+      return;
+    this._screencast.screencastFrameAck(screencastId);
+  }
+
+  stopScreencast() {
+    if (!this._screencastRecordingInfo)
+      throw new Error('No screencast in progress');
+    const screencastInfo = this._screencastRecordingInfo;
+    Services.obs.removeObserver(screencastInfo.onFrame, 'juggler-screencast-frame');
+    this._screencastRecordingInfo = undefined;
+    this._screencast.stopVideoRecording(screencastInfo.screencastId);
   }
 
   dispose() {
     this._disposed = true;
-    if (this._screencastInfo)
+    if (this._videoRecordingInfo)
-      this._stopVideoRecording().catch(e => dump(`stopVideoRecording failed:\n${e}\n`));
+      this._stopVideoRecording();
+    if (this._screencastRecordingInfo)
+      this.stopScreencast();
     this._browserContext.pages.delete(this);
     this._registry._browserToTarget.delete(this._linkedBrowser);
     this._registry._browserBrowsingContextToTarget.delete(this._linkedBrowser.browsingContext);
@@ -551,6 +579,7 @@ class PageTarget {
 
 PageTarget.Events = {
   ScreencastStarted: Symbol('PageTarget.ScreencastStarted'),
+  ScreencastFrame: Symbol('PageTarget.ScreencastFrame'),
   Crashed: Symbol('PageTarget.Crashed'),
   DialogOpened: Symbol('PageTarget.DialogOpened'),
   DialogClosed: Symbol('PageTarget.DialogClosed'),
@@ -591,7 +620,7 @@ class BrowserContext {
     this.defaultUserAgent = null;
     this.touchOverride = false;
     this.colorScheme = 'none';
-    this.screencastOptions = undefined;
+    this.videoRecordingOptions = undefined;
     this.scriptsToEvaluateOnNewDocument = [];
     this.bindings = [];
     this.settings = {};
@@ -788,8 +817,8 @@ class BrowserContext {
     return result;
   }
 
-  async setScreencastOptions(options) {
+  async setVideoRecordingOptions(options) {
-    this.screencastOptions = options;
+    this.videoRecordingOptions = options;
     if (!options)
       return;
     const promises = [];

browser_patches/firefox/juggler/protocol/BrowserHandler.js

@@ -40,7 +40,7 @@ class BrowserHandler {
     ];
 
     const onScreencastStopped = (subject, topic, data) => {
-      this._session.emitEvent('Browser.screencastFinished', {screencastId: '' + data});
+      this._session.emitEvent('Browser.videoRecordingFinished', {screencastId: '' + data});
     };
     Services.obs.addObserver(onScreencastStopped, 'juggler-screencast-stopped');
     this._eventListeners.push(() => Services.obs.removeObserver(onScreencastStopped, 'juggler-screencast-stopped'));
@@ -204,8 +204,8 @@ class BrowserHandler {
     await this._targetRegistry.browserContextForId(browserContextId).setColorScheme(nullToUndefined(colorScheme));
   }
 
-  async ['Browser.setScreencastOptions']({browserContextId, dir, width, height, scale}) {
+  async ['Browser.setVideoRecordingOptions']({browserContextId, dir, width, height, scale}) {
-    await this._targetRegistry.browserContextForId(browserContextId).setScreencastOptions({dir, width, height, scale});
+    await this._targetRegistry.browserContextForId(browserContextId).setVideoRecordingOptions({dir, width, height, scale});
   }
 
   async ['Browser.setUserAgentOverride']({browserContextId, userAgent}) {

browser_patches/firefox/juggler/protocol/PageHandler.js

@@ -88,8 +88,8 @@ class PageHandler {
     // to be ignored by the protocol clients.
     this._isPageReady = false;
 
-    if (this._pageTarget.screencastInfo())
+    if (this._pageTarget.videoRecordingInfo())
-      this._onScreencastStarted();
+      this._onVideoRecordingStarted();
 
     this._eventListeners = [
       helper.on(this._pageTarget, PageTarget.Events.DialogOpened, this._onDialogOpened.bind(this)),
@@ -97,7 +97,8 @@ class PageHandler {
       helper.on(this._pageTarget, PageTarget.Events.Crashed, () => {
         this._session.emitEvent('Page.crashed', {});
       }),
-      helper.on(this._pageTarget, PageTarget.Events.ScreencastStarted, this._onScreencastStarted.bind(this)),
+      helper.on(this._pageTarget, PageTarget.Events.ScreencastStarted, this._onVideoRecordingStarted.bind(this)),
+      helper.on(this._pageTarget, PageTarget.Events.ScreencastFrame, this._onScreencastFrame.bind(this)),
       helper.on(this._pageNetwork, PageNetwork.Events.Request, this._handleNetworkEvent.bind(this, 'Network.requestWillBeSent')),
       helper.on(this._pageNetwork, PageNetwork.Events.Response, this._handleNetworkEvent.bind(this, 'Network.responseReceived')),
       helper.on(this._pageNetwork, PageNetwork.Events.RequestFinished, this._handleNetworkEvent.bind(this, 'Network.requestFinished')),
@@ -146,9 +147,13 @@ class PageHandler {
     helper.removeListeners(this._eventListeners);
   }
 
-  _onScreencastStarted() {
+  _onVideoRecordingStarted() {
-    const info = this._pageTarget.screencastInfo();
+    const info = this._pageTarget.videoRecordingInfo();
-    this._session.emitEvent('Page.screencastStarted', { screencastId: info.videoSessionId, file: info.file });
+    this._session.emitEvent('Page.videoRecordingStarted', { screencastId: info.sessionId, file: info.file });
+  }
+
+  _onScreencastFrame(data) {
+    this._session.emitEvent('Page.screencastFrame', { data, deviceWidth: 0, deviceHeight: 0 });
   }
 
   _onPageReady(event) {
@@ -373,6 +378,18 @@ class PageHandler {
     return await this._contentPage.send('setInterceptFileChooserDialog', options);
   }
 
+  async ['Page.startScreencast'](options) {
+    return await this._pageTarget.startScreencast(options);
+  }
+
+  async ['Page.screencastFrameAck'](options) {
+    await this._pageTarget.screencastFrameAck(options);
+  }
+
+  async ['Page.stopScreencast'](options) {
+    await this._pageTarget.stopScreencast(options);
+  }
+
   async ['Page.sendMessageToWorker']({workerId, message}) {
     const worker = this._workers.get(workerId);
     if (!worker)

browser_patches/firefox/juggler/protocol/Protocol.js

@@ -226,7 +226,7 @@ const Browser = {
       canceled: t.Optional(t.Boolean),
       error: t.Optional(t.String),
     },
-    'screencastFinished': {
+    'videoRecordingFinished': {
       screencastId: t.String,
     },
   },
@@ -420,7 +420,7 @@ const Browser = {
         colorScheme: t.Nullable(t.Enum(['dark', 'light', 'no-preference'])),
       },
     },
-    'setScreencastOptions': {
+    'setVideoRecordingOptions': {
       params: {
         browserContextId: t.Optional(t.String),
         dir: t.String,
@@ -664,7 +664,7 @@ const Page = {
       workerId: t.String,
       message: t.String,
     },
-    'screencastStarted': {
+    'videoRecordingStarted': {
       screencastId: t.String,
       file: t.String,
     },
@@ -696,6 +696,11 @@ const Page = {
       opcode: t.Number,
       data: t.String,
     },
+    'screencastFrame': {
+      data: t.String,
+      deviceWidth: t.Number,
+      deviceHeight: t.Number,
+    },
   },
 
   methods: {
@@ -895,6 +900,23 @@ const Page = {
         message: t.String,
       },
     },
+    'startScreencast': {
+      params: {
+        width: t.Number,
+        height: t.Number,
+        quality: t.Number,
+      },
+      returns: {
+        screencastId: t.String,
+      },
+    },
+    'screencastFrameAck': {
+      params: {
+        screencastId: t.String,
+      },
+    },
+    'stopScreencast': {
+    },
   },
 };
 

browser_patches/firefox/juggler/screencast/HeadlessWindowCapturer.cpp

@@ -17,7 +17,7 @@ using namespace webrtc;
 
 namespace mozilla {
 
-rtc::scoped_refptr<webrtc::VideoCaptureModule> HeadlessWindowCapturer::Create(HeadlessWidget* headlessWindow) {
+rtc::scoped_refptr<webrtc::VideoCaptureModuleEx> HeadlessWindowCapturer::Create(HeadlessWidget* headlessWindow) {
   return new rtc::RefCountedObject<HeadlessWindowCapturer>(headlessWindow);
 }
 
@@ -41,6 +41,19 @@ void HeadlessWindowCapturer::DeRegisterCaptureDataCallback(rtc::VideoSinkInterfa
   }
 }
 
+void HeadlessWindowCapturer::RegisterRawFrameCallback(webrtc::RawFrameCallback* rawFrameCallback) {
+  rtc::CritScope lock2(&_callBackCs);
+  _rawFrameCallbacks.insert(rawFrameCallback);
+}
+
+void HeadlessWindowCapturer::DeRegisterRawFrameCallback(webrtc::RawFrameCallback* rawFrameCallback) {
+  rtc::CritScope lock2(&_callBackCs);
+  auto it = _rawFrameCallbacks.find(rawFrameCallback);
+  if (it != _rawFrameCallbacks.end()) {
+    _rawFrameCallbacks.erase(it);
+  }
+}
+
 void HeadlessWindowCapturer::NotifyFrameCaptured(const webrtc::VideoFrame& frame) {
   rtc::CritScope lock2(&_callBackCs);
   for (auto dataCallBack : _dataCallBacks)
@@ -63,10 +76,23 @@ int32_t HeadlessWindowCapturer::StartCapture(const VideoCaptureCapability& capab
     }
 
     if (dataSurface->GetFormat() != gfx::SurfaceFormat::B8G8R8A8) {
-      fprintf(stderr, "Uexpected snapshot surface format: %hhd\n", dataSurface->GetFormat());
+      fprintf(stderr, "Unexpected snapshot surface format: %hhd\n", dataSurface->GetFormat());
       return;
     }
 
+    webrtc::VideoCaptureCapability frameInfo;
+    frameInfo.width = dataSurface->GetSize().width;
+    frameInfo.height = dataSurface->GetSize().height;
+#if MOZ_LITTLE_ENDIAN()
+    frameInfo.videoType = VideoType::kARGB;
+#else
+    frameInfo.videoType = VideoType::kBGRA;
+#endif
+
+    for (auto rawFrameCallback : _rawFrameCallbacks) {
+      rawFrameCallback->OnRawFrame(dataSurface->GetData(), dataSurface->Stride(), frameInfo);
+    }
+
     int width = dataSurface->GetSize().width;
     int height = dataSurface->GetSize().height;
     rtc::scoped_refptr<I420Buffer> buffer = I420Buffer::Create(width, height);
@@ -87,7 +113,6 @@ int32_t HeadlessWindowCapturer::StartCapture(const VideoCaptureCapability& capab
         buffer->MutableDataU(), buffer->StrideU(),
         buffer->MutableDataV(), buffer->StrideV(),
         width, height);
-
     if (conversionResult != 0) {
       fprintf(stderr, "Failed to convert capture frame to I420: %d\n", conversionResult);
       return;

browser_patches/firefox/juggler/screencast/HeadlessWindowCapturer.h

@@ -10,6 +10,7 @@
 #include "media/base/videosinkinterface.h"
 #include "modules/video_capture/video_capture.h"
 #include "rtc_base/criticalsection.h"
+#include "video_engine/desktop_capture_impl.h"
 
 class nsIWidget;
 
@@ -19,9 +20,9 @@ namespace widget {
 class HeadlessWidget;
 }
 
-class HeadlessWindowCapturer : public webrtc::VideoCaptureModule {
+class HeadlessWindowCapturer : public webrtc::VideoCaptureModuleEx {
  public:
-  static rtc::scoped_refptr<webrtc::VideoCaptureModule> Create(mozilla::widget::HeadlessWidget*);
+  static rtc::scoped_refptr<webrtc::VideoCaptureModuleEx> Create(mozilla::widget::HeadlessWidget*);
 
   void RegisterCaptureDataCallback(
       rtc::VideoSinkInterface<webrtc::VideoFrame>* dataCallback) override;
@@ -29,6 +30,9 @@ class HeadlessWindowCapturer : public webrtc::VideoCaptureModule {
       rtc::VideoSinkInterface<webrtc::VideoFrame>* dataCallback) override;
   int32_t StopCaptureIfAllClientsClose() override;
 
+  void RegisterRawFrameCallback(webrtc::RawFrameCallback* rawFrameCallback) override;
+  void DeRegisterRawFrameCallback(webrtc::RawFrameCallback* rawFrameCallback) override;
+
   int32_t SetCaptureRotation(webrtc::VideoRotation) override { return -1; }
   bool SetApplyRotation(bool) override { return false; }
   bool GetApplyRotation() override { return true; }
@@ -54,6 +58,7 @@ class HeadlessWindowCapturer : public webrtc::VideoCaptureModule {
   mozilla::widget::HeadlessWidget* mWindow = nullptr;
   rtc::CriticalSection _callBackCs;
   std::set<rtc::VideoSinkInterface<webrtc::VideoFrame>*> _dataCallBacks;
+  std::set<webrtc::RawFrameCallback*> _rawFrameCallbacks;
 };
 
 }  // namespace mozilla

browser_patches/firefox/juggler/screencast/nsIScreencastService.idl

@@ -12,10 +12,12 @@ interface nsIDocShell;
 [scriptable, uuid(d8c4d9e0-9462-445e-9e43-68d3872ad1de)]
 interface nsIScreencastService : nsISupports
 {
-  AString startVideoRecording(in nsIDocShell docShell, in ACString fileName, in uint32_t width, in uint32_t height, in int32_t offset_top);
+  AString startVideoRecording(in nsIDocShell docShell, in boolean isVideo, in ACString fileName, in uint32_t width, in uint32_t height, in uint32_t quality, in int32_t offset_top);
 
   /**
    * Will emit 'juggler-screencast-stopped' when the video file is saved.
    */
   void stopVideoRecording(in AString sessionId);
+
+  void screencastFrameAck(in AString sessionId);
 };

browser_patches/firefox/juggler/screencast/nsScreencastService.cpp

@@ -7,6 +7,7 @@
 #include "ScreencastEncoder.h"
 #include "HeadlessWidget.h"
 #include "HeadlessWindowCapturer.h"
+#include "mozilla/Base64.h"
 #include "mozilla/ClearOnShutdown.h"
 #include "mozilla/PresShell.h"
 #include "mozilla/StaticPtr.h"
@@ -15,6 +16,7 @@
 #include "nsIRandomGenerator.h"
 #include "nsISupportsPrimitives.h"
 #include "nsThreadManager.h"
+#include "nsReadableUtils.h"
 #include "nsView.h"
 #include "nsViewManager.h"
 #include "webrtc/modules/desktop_capture/desktop_capturer.h"
@@ -24,6 +26,9 @@
 #include "webrtc/modules/video_capture/video_capture.h"
 #include "mozilla/widget/PlatformWidgetTypes.h"
 #include "video_engine/desktop_capture_impl.h"
+extern "C" {
+#include "jpeglib.h"
+}
 
 using namespace mozilla::widget;
 
@@ -33,9 +38,11 @@ NS_IMPL_ISUPPORTS(nsScreencastService, nsIScreencastService)
 
 namespace {
 
+const int kMaxFramesInFlight = 1;
+
 StaticRefPtr<nsScreencastService> gScreencastService;
 
-rtc::scoped_refptr<webrtc::VideoCaptureModule> CreateWindowCapturer(nsIWidget* widget) {
+rtc::scoped_refptr<webrtc::VideoCaptureModuleEx> CreateWindowCapturer(nsIWidget* widget) {
   if (gfxPlatform::IsHeadless()) {
     HeadlessWidget* headlessWidget = static_cast<HeadlessWidget*>(widget);
     return HeadlessWindowCapturer::Create(headlessWidget);
@@ -62,6 +69,18 @@ void NotifyScreencastStopped(const nsString& sessionId) {
   observerService->NotifyObservers(nullptr, "juggler-screencast-stopped", sessionId.get());
 }
 
+void NotifyScreencastFrame(const nsCString& frameData) {
+  nsString wideString;
+  CopyASCIItoUTF16(frameData, wideString);
+  nsCOMPtr<nsIObserverService> observerService = mozilla::services::GetObserverService();
+  if (!observerService) {
+    fprintf(stderr, "NotifyScreencastFrame error: no observer service\n");
+    return;
+  }
+
+  observerService->NotifyObservers(nullptr, "juggler-screencast-frame", wideString.get());
+}
+
 nsresult generateUid(nsString& uid) {
   nsresult rv = NS_OK;
   nsCOMPtr<nsIRandomGenerator> rg = do_GetService("@mozilla.org/security/random-generator;1", &rv);
@@ -80,11 +99,14 @@ nsresult generateUid(nsString& uid) {
 }
 }
 
-class nsScreencastService::Session : public rtc::VideoSinkInterface<webrtc::VideoFrame> {
+class nsScreencastService::Session : public rtc::VideoSinkInterface<webrtc::VideoFrame>,
+                                     public webrtc::RawFrameCallback {
  public:
-  Session(rtc::scoped_refptr<webrtc::VideoCaptureModule>&& capturer, RefPtr<ScreencastEncoder>&& encoder)
+  Session(rtc::scoped_refptr<webrtc::VideoCaptureModuleEx>&& capturer, RefPtr<ScreencastEncoder>&& encoder, gfx::IntMargin margin, uint32_t jpegQuality)
       : mCaptureModule(std::move(capturer))
-      , mEncoder(std::move(encoder)) {
+      , mEncoder(std::move(encoder))
+      , mJpegQuality(jpegQuality)
+      , mMargin(margin) {
   }
 
   bool Start() {
@@ -100,27 +122,113 @@ class nsScreencastService::Session : public rtc::VideoSinkInterface<webrtc::Vide
       return false;
     }
 
-    mCaptureModule->RegisterCaptureDataCallback(this);
+    if (mEncoder)
+      mCaptureModule->RegisterCaptureDataCallback(this);
+    else
+      mCaptureModule->RegisterRawFrameCallback(this);
     return true;
   }
 
   void Stop(std::function<void()>&& callback) {
-    mCaptureModule->DeRegisterCaptureDataCallback(this);
+    if (mEncoder)
+      mCaptureModule->DeRegisterCaptureDataCallback(this);
+    else
+      mCaptureModule->RegisterRawFrameCallback(this);
     int error = mCaptureModule->StopCapture();
     if (error) {
       fprintf(stderr, "StopCapture error %d\n", error);
     }
-    mEncoder->finish(std::move(callback));
+    if (mEncoder)
+      mEncoder->finish(std::move(callback));
+    else
+      callback();
+  }
+
+  void ScreencastFrameAck() {
+    rtc::CritScope lock(&mFramesInFlightCs);
+    --mFramesInFlight;
   }
 
   // These callbacks end up running on the VideoCapture thread.
   void OnFrame(const webrtc::VideoFrame& videoFrame) override {
+    if (!mEncoder)
+      return;
     mEncoder->encodeFrame(videoFrame);
   }
 
+  // These callbacks end up running on the VideoCapture thread.
+  void OnRawFrame(uint8_t* videoFrame, size_t videoFrameStride, const webrtc::VideoCaptureCapability& frameInfo) override {
+    if (!mJpegQuality)
+      return;
+    {
+      rtc::CritScope lock(&mFramesInFlightCs);
+      if (mFramesInFlight >= kMaxFramesInFlight)
+        return;
+      ++mFramesInFlight;
+    }
+
+    jpeg_compress_struct info;
+    jpeg_error_mgr error;
+    info.err = jpeg_std_error(&error);
+    jpeg_create_compress(&info);
+
+    unsigned char* bufferPtr = nullptr;
+    unsigned long bufferSize;
+    jpeg_mem_dest(&info, &bufferPtr, &bufferSize);
+
+    info.image_width = frameInfo.width - mMargin.LeftRight();
+    info.image_height = frameInfo.height - mMargin.TopBottom();
+
+#if MOZ_LITTLE_ENDIAN()
+    if (frameInfo.videoType == webrtc::VideoType::kARGB)
+      info.in_color_space = JCS_EXT_BGRA;
+    if (frameInfo.videoType == webrtc::VideoType::kBGRA)
+      info.in_color_space = JCS_EXT_ARGB;
+#else
+    if (frameInfo.videoType == webrtc::VideoType::kARGB)
+      info.in_color_space = JCS_EXT_ARGB;
+    if (frameInfo.videoType == webrtc::VideoType::kBGRA)
+      info.in_color_space = JCS_EXT_BGRA;
+#endif
+
+    // # of color components in input image
+    info.input_components = 4;
+
+    jpeg_set_defaults(&info);
+    jpeg_set_quality(&info, mJpegQuality, true);
+
+    jpeg_start_compress(&info, true);
+    while (info.next_scanline < info.image_height) {
+      JSAMPROW row = videoFrame + (mMargin.top + info.next_scanline) * videoFrameStride + 4 * mMargin.left;
+      if (jpeg_write_scanlines(&info, &row, 1) != 1) {
+        fprintf(stderr, "JPEG library failed to encode line\n");
+        break;
+      }
+    }
+
+    jpeg_finish_compress(&info);
+    jpeg_destroy_compress(&info);
+
+    nsCString base64;
+    nsresult rv = mozilla::Base64Encode(reinterpret_cast<char *>(bufferPtr), bufferSize, base64);
+    if (NS_WARN_IF(NS_FAILED(rv)))
+      return;
+
+    NS_DispatchToMainThread(NS_NewRunnableFunction(
+        "NotifyScreencastFrame", [base64]() -> void {
+          NotifyScreencastFrame(base64);
+        }));
+
+    free(bufferPtr);
+  }
+
  private:
-  rtc::scoped_refptr<webrtc::VideoCaptureModule> mCaptureModule;
+  rtc::scoped_refptr<webrtc::VideoCaptureModuleEx> mCaptureModule;
   RefPtr<ScreencastEncoder> mEncoder;
+  uint32_t mJpegQuality;
+  rtc::CriticalSection mFramesInFlightCs;
+  uint32_t mFramesInFlight = 0;
+  gfx::IntMargin mMargin;
 };
 
 
@@ -140,7 +248,7 @@ nsScreencastService::nsScreencastService() = default;
 nsScreencastService::~nsScreencastService() {
 }
 
-nsresult nsScreencastService::StartVideoRecording(nsIDocShell* aDocShell, const nsACString& aFileName, uint32_t width, uint32_t height, int32_t offsetTop, nsAString& sessionId) {
+nsresult nsScreencastService::StartVideoRecording(nsIDocShell* aDocShell, bool isVideo, const nsACString& aVideoFileName, uint32_t width, uint32_t height, uint32_t quality, int32_t offsetTop, nsAString& sessionId) {
   MOZ_RELEASE_ASSERT(NS_IsMainThread(), "Screencast service must be started on the Main thread.");
 
   PresShell* presShell = aDocShell->GetPresShell();
@@ -154,7 +262,7 @@ nsresult nsScreencastService::StartVideoRecording(nsIDocShell* aDocShell, const
     return NS_ERROR_UNEXPECTED;
   nsIWidget* widget = view->GetWidget();
 
-  rtc::scoped_refptr<webrtc::VideoCaptureModule> capturer = CreateWindowCapturer(widget);
+  rtc::scoped_refptr<webrtc::VideoCaptureModuleEx> capturer = CreateWindowCapturer(widget);
   if (!capturer)
     return NS_ERROR_FAILURE;
 
@@ -167,22 +275,24 @@ nsresult nsScreencastService::StartVideoRecording(nsIDocShell* aDocShell, const
   margin.top += offsetTop;
 
   nsCString error;
-  RefPtr<ScreencastEncoder> encoder = ScreencastEncoder::create(error, PromiseFlatCString(aFileName), width, height, margin);
+  RefPtr<ScreencastEncoder> encoder;
-  if (!encoder) {
+  if (isVideo) {
-    fprintf(stderr, "Failed to create ScreencastEncoder: %s\n", error.get());
+    encoder = ScreencastEncoder::create(error, PromiseFlatCString(aVideoFileName), width, height, margin);
-    return NS_ERROR_FAILURE;
+    if (!encoder) {
+      fprintf(stderr, "Failed to create ScreencastEncoder: %s\n", error.get());
+      return NS_ERROR_FAILURE;
+    }
   }
 
-  auto session = std::make_unique<Session>(std::move(capturer), std::move(encoder));
-  if (!session->Start())
-    return NS_ERROR_FAILURE;
-
   nsString uid;
   nsresult rv = generateUid(uid);
   NS_ENSURE_SUCCESS(rv, rv);
-
   sessionId = uid;
-  mIdToSession.emplace(uid, std::move(session));
+
+  auto session = std::make_unique<Session>(std::move(capturer), std::move(encoder), margin, isVideo ? 0 : quality);
+  if (!session->Start())
+    return NS_ERROR_FAILURE;
+  mIdToSession.emplace(sessionId, std::move(session));
   return NS_OK;
 }
 
@@ -201,4 +311,13 @@ nsresult nsScreencastService::StopVideoRecording(const nsAString& aSessionId) {
   return NS_OK;
 }
 
+nsresult nsScreencastService::ScreencastFrameAck(const nsAString& aSessionId) {
+  nsString sessionId(aSessionId);
+  auto it = mIdToSession.find(sessionId);
+  if (it == mIdToSession.end())
+    return NS_ERROR_INVALID_ARG;
+  it->second->ScreencastFrameAck();
+  return NS_OK;
+}
+
 }  // namespace mozilla

browser_patches/firefox/patches/bootstrap.diff

@@ -1143,12 +1143,15 @@ index 2459a4c3a464becb5b668af6571940a275239848..f7c1f7c0f6da0749a502cdc40d2b6eda
    /** Synthesize a touch event. The event types supported are:
     *    touchstart, touchend, touchmove, and touchcancel
 diff --git a/dom/media/systemservices/video_engine/desktop_capture_impl.cc b/dom/media/systemservices/video_engine/desktop_capture_impl.cc
-index 7bc92fe4408c2878c9d7c8bdb97a7c257258ee31..8885feebedf53c0748cef19d80ce5aa23adc900b 100644
+index 7bc92fe4408c2878c9d7c8bdb97a7c257258ee31..b20480c3c0ca96097e61d37f44e127d45ab4648a 100644
 --- a/dom/media/systemservices/video_engine/desktop_capture_impl.cc
 +++ b/dom/media/systemservices/video_engine/desktop_capture_impl.cc
-@@ -125,8 +125,9 @@ int32_t ScreenDeviceInfoImpl::GetOrientation(const char* deviceUniqueIdUTF8,
+@@ -123,10 +123,11 @@ int32_t ScreenDeviceInfoImpl::GetOrientation(const char* deviceUniqueIdUTF8,
+   return 0;
+ }
  
- VideoCaptureModule* DesktopCaptureImpl::Create(const int32_t id,
+-VideoCaptureModule* DesktopCaptureImpl::Create(const int32_t id,
++VideoCaptureModuleEx* DesktopCaptureImpl::Create(const int32_t id,
                                                 const char* uniqueId,
 -                                               const CaptureDeviceType type) {
 -  return new rtc::RefCountedObject<DesktopCaptureImpl>(id, uniqueId, type);
@@ -1199,21 +1202,93 @@ index 7bc92fe4408c2878c9d7c8bdb97a7c257258ee31..8885feebedf53c0748cef19d80ce5aa2
        time_event_(EventWrapper::Create()),
  #if defined(_WIN32)
        capturer_thread_(
+@@ -437,6 +444,19 @@ void DesktopCaptureImpl::DeRegisterCaptureDataCallback(
+   }
+ }
+ 
++void DesktopCaptureImpl::RegisterRawFrameCallback(RawFrameCallback* rawFrameCallback) {
++  rtc::CritScope lock(&_apiCs);
++  _rawFrameCallbacks.insert(rawFrameCallback);
++}
++
++void DesktopCaptureImpl::DeRegisterRawFrameCallback(RawFrameCallback* rawFrameCallback) {
++  rtc::CritScope lock(&_apiCs);
++  auto it = _rawFrameCallbacks.find(rawFrameCallback);
++  if (it != _rawFrameCallbacks.end()) {
++    _rawFrameCallbacks.erase(it);
++  }
++}
++
+ int32_t DesktopCaptureImpl::StopCaptureIfAllClientsClose() {
+   if (_dataCallBacks.empty()) {
+     return StopCapture();
+@@ -644,6 +664,12 @@ void DesktopCaptureImpl::OnCaptureResult(DesktopCapturer::Result result,
+   frameInfo.height = frame->size().height();
+   frameInfo.videoType = VideoType::kARGB;
+ 
++  size_t videoFrameStride =
++      frameInfo.width * DesktopFrame::kBytesPerPixel;
++  for (auto rawFrameCallback : _rawFrameCallbacks) {
++    rawFrameCallback->OnRawFrame(videoFrame, videoFrameStride, frameInfo);
++  }
++
+   size_t videoFrameLength =
+       frameInfo.width * frameInfo.height * DesktopFrame::kBytesPerPixel;
+   IncomingFrame(videoFrame, videoFrameLength, frameInfo);
 diff --git a/dom/media/systemservices/video_engine/desktop_capture_impl.h b/dom/media/systemservices/video_engine/desktop_capture_impl.h
-index 137281596e66793b8525ca2a5d156b4bd171da65..28978483e54e6424b6144d3a5509583df5885fdb 100644
+index 137281596e66793b8525ca2a5d156b4bd171da65..0e64c0b810b306e4a2509e4ad0d9596034d338fa 100644
 --- a/dom/media/systemservices/video_engine/desktop_capture_impl.h
 +++ b/dom/media/systemservices/video_engine/desktop_capture_impl.h
-@@ -160,7 +160,8 @@ class DesktopCaptureImpl : public DesktopCapturer::Callback,
+@@ -42,6 +42,21 @@ namespace webrtc {
+ 
+ class VideoCaptureEncodeInterface;
+ 
++class RawFrameCallback {
++ public:
++  virtual ~RawFrameCallback() {}
++
++  virtual void OnRawFrame(uint8_t* videoFrame, size_t videoFrameLength, const VideoCaptureCapability& frameInfo) = 0;
++};
++
++class VideoCaptureModuleEx : public VideoCaptureModule {
++ public:
++  virtual ~VideoCaptureModuleEx() {}
++
++  virtual void RegisterRawFrameCallback(RawFrameCallback* rawFrameCallback) = 0;
++  virtual void DeRegisterRawFrameCallback(RawFrameCallback* rawFrameCallback) = 0;
++};
++
+ // simulate deviceInfo interface for video engine, bridge screen/application and
+ // real screen/application device info
+ 
+@@ -154,13 +169,14 @@ class BrowserDeviceInfoImpl : public VideoCaptureModule::DeviceInfo {
+ // As with video, DesktopCaptureImpl is a proxy for screen sharing
+ // and follows the video pipeline design
+ class DesktopCaptureImpl : public DesktopCapturer::Callback,
+-                           public VideoCaptureModule,
++                           public VideoCaptureModuleEx,
+                            public VideoCaptureExternal {
+  public:
    /* Create a screen capture modules object
     */
-   static VideoCaptureModule* Create(const int32_t id, const char* uniqueId,
+-  static VideoCaptureModule* Create(const int32_t id, const char* uniqueId,
 -                                    const CaptureDeviceType type);
++  static VideoCaptureModuleEx* Create(const int32_t id, const char* uniqueId,
 +                                    const CaptureDeviceType type,
 +                                    bool captureCursor = true);
    static VideoCaptureModule::DeviceInfo* CreateDeviceInfo(
        const int32_t id, const CaptureDeviceType type);
  
-@@ -192,7 +193,7 @@ class DesktopCaptureImpl : public DesktopCapturer::Callback,
+@@ -170,6 +186,8 @@ class DesktopCaptureImpl : public DesktopCapturer::Callback,
+   void DeRegisterCaptureDataCallback(
+       rtc::VideoSinkInterface<VideoFrame>* dataCallback) override;
+   int32_t StopCaptureIfAllClientsClose() override;
++  void RegisterRawFrameCallback(RawFrameCallback* rawFrameCallback) override;
++  void DeRegisterRawFrameCallback(RawFrameCallback* rawFrameCallback) override;
+ 
+   int32_t SetCaptureRotation(VideoRotation rotation) override;
+   bool SetApplyRotation(bool enable) override;
+@@ -192,7 +210,7 @@ class DesktopCaptureImpl : public DesktopCapturer::Callback,
  
   protected:
    DesktopCaptureImpl(const int32_t id, const char* uniqueId,
@@ -1222,7 +1297,15 @@ index 137281596e66793b8525ca2a5d156b4bd171da65..28978483e54e6424b6144d3a5509583d
    virtual ~DesktopCaptureImpl();
    int32_t DeliverCapturedFrame(webrtc::VideoFrame& captureFrame,
                                 int64_t capture_time);
-@@ -240,6 +241,7 @@ class DesktopCaptureImpl : public DesktopCapturer::Callback,
+@@ -215,6 +233,7 @@ class DesktopCaptureImpl : public DesktopCapturer::Callback,
+   rtc::CriticalSection _apiCs;
+ 
+   std::set<rtc::VideoSinkInterface<VideoFrame>*> _dataCallBacks;
++  std::set<RawFrameCallback*> _rawFrameCallbacks;
+ 
+   int64_t _incomingFrameTimesNanos
+       [kFrameRateCountHistorySize];  // timestamp for local captured frames
+@@ -240,6 +259,7 @@ class DesktopCaptureImpl : public DesktopCapturer::Callback,
    void process();
  
   private:
@@ -1584,6 +1667,18 @@ index 77b4c4ea3581e3b66b0b40dae33c807b2d5aefd8..84af4461b9e946122527ac974dc30da5
    void updateTimeZone();
  
    void internalResyncICUDefaultTimeZone();
+diff --git a/media/libjpeg/jconfig.h b/media/libjpeg/jconfig.h
+index f2723e654098ff27542e1eb16a536c11ad0af617..b0b480551ff7d895dfdeb5a9800874858929c8ba 100644
+--- a/media/libjpeg/jconfig.h
++++ b/media/libjpeg/jconfig.h
+@@ -17,6 +17,7 @@
+ /* #undef D_ARITH_CODING_SUPPORTED */
+ 
+ /* Support in-memory source/destination managers */
++#define MEM_SRCDST_SUPPORTED 1
+ /* #undef MEM_SRCDST_SUPPORTED */
+ 
+ /* Use accelerated SIMD routines. */
 diff --git a/netwerk/base/nsINetworkInterceptController.idl b/netwerk/base/nsINetworkInterceptController.idl
 index 64a4a71b03b28872f376aac8eee12805bebd1bd8..f6fa7d731f3b0c7c4fcb26babad3fc2cdb29aec1 100644
 --- a/netwerk/base/nsINetworkInterceptController.idl