Handle a DISPLAY_VIDEO_CAPTURE_THIS_TAB request in the browser process

This is the second step in implementing getCurrentBrowsingContextMedia behind
a runtime flag.

TL;DR: This is an API for capturing the current tab.

Explainer:
https://docs.google.com/document/d/1CIQH2ygvw7eTGO__Pcds_D46Gcn-iPAqESQqOsHHfkI

Design doc:
go/get-current-browsing-context-media

Intent-to-Prototype:
https://groups.google.com/u/3/a/chromium.org/g/blink-dev/c/NYVbRRBlABI/m/MJEzcyEUCQAJ

PR against spec:
w3c/mediacapture-screen-share#148

Next steps:
* Implement the confirmation-box.
* Implement unit-tests that rely on the confirmation-box.

Bug: 1136942
Change-Id: I8b25baa85565999ec44ed2f1b0bd1e19d6f148c4
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/2502628
Reviewed-by: Guido Urdaneta <[email protected]>
Reviewed-by: Yuri Wiitala <[email protected]>
Reviewed-by: Robert Kaplow <[email protected]>
Reviewed-by: Sergey Ulanov <[email protected]>
Commit-Queue: Sergey Ulanov <[email protected]>
Cr-Commit-Position: refs/heads/master@{#824534}

chrome/browser/media/capture_access_handler_base.cc

@@ -131,12 +131,15 @@ void CaptureAccessHandlerBase::UpdateMediaRequestState(
     blink::mojom::MediaStreamType stream_type,
     content::MediaRequestState state) {
   DCHECK_CURRENTLY_ON(BrowserThread::UI);
-  if ((stream_type !=
-       blink::mojom::MediaStreamType::GUM_DESKTOP_VIDEO_CAPTURE) &&
-      (stream_type != blink::mojom::MediaStreamType::GUM_TAB_VIDEO_CAPTURE) &&
-      (stream_type != blink::mojom::MediaStreamType::DISPLAY_VIDEO_CAPTURE) &&
-      (stream_type != blink::mojom::MediaStreamType::DISPLAY_AUDIO_CAPTURE)) {
-    return;
+  switch (stream_type) {
+    case blink::mojom::MediaStreamType::GUM_DESKTOP_VIDEO_CAPTURE:
+    case blink::mojom::MediaStreamType::GUM_TAB_VIDEO_CAPTURE:
+    case blink::mojom::MediaStreamType::DISPLAY_VIDEO_CAPTURE:
+    case blink::mojom::MediaStreamType::DISPLAY_AUDIO_CAPTURE:
+    case blink::mojom::MediaStreamType::DISPLAY_VIDEO_CAPTURE_THIS_TAB:
+      break;
+    default:
+      return;
   }
 
   if (state == content::MEDIA_REQUEST_STATE_DONE) {

chrome/browser/media/webrtc/display_media_access_handler.cc

@@ -73,7 +73,9 @@ bool DisplayMediaAccessHandler::SupportsStreamType(
     content::WebContents* web_contents,
     const blink::mojom::MediaStreamType stream_type,
     const extensions::Extension* extension) {
-  return stream_type == blink::mojom::MediaStreamType::DISPLAY_VIDEO_CAPTURE;
+  return stream_type == blink::mojom::MediaStreamType::DISPLAY_VIDEO_CAPTURE ||
+         stream_type ==
+             blink::mojom::MediaStreamType::DISPLAY_VIDEO_CAPTURE_THIS_TAB;
   // This class handles MEDIA_DISPLAY_AUDIO_CAPTURE as well, but only if it is
   // accompanied by MEDIA_DISPLAY_VIDEO_CAPTURE request as per spec.
   // https://w3c.github.io/mediacapture-screen-share/#mediadevices-additions
@@ -135,6 +137,16 @@ void DisplayMediaAccessHandler::HandleRequest(
   }
 #endif  // defined(OS_MAC)
 
+  if (request.video_type ==
+      blink::mojom::MediaStreamType::DISPLAY_VIDEO_CAPTURE_THIS_TAB) {
+    // TODO(crbug.com/1136942): Add support for capture-this-tab instead of
+    // returning an error
+    std::move(callback).Run(
+        blink::MediaStreamDevices(),
+        blink::mojom::MediaStreamRequestResult::NOT_SUPPORTED, nullptr);
+    return;
+  }
+
   std::unique_ptr<DesktopMediaPicker> picker = picker_factory_->CreatePicker();
   if (!picker) {
     std::move(callback).Run(
@@ -267,8 +279,7 @@ void DisplayMediaAccessHandler::OnPickerDialogResults(
           web_contents->GetLastCommittedURL(),
           url_formatter::SchemeDisplay::OMIT_CRYPTOGRAPHIC);
       ui = GetDevicesForDesktopCapture(
-          web_contents, &devices, media_id,
-          blink::mojom::MediaStreamType::DISPLAY_VIDEO_CAPTURE,
+          web_contents, &devices, media_id, pending_request.request.video_type,
           blink::mojom::MediaStreamType::DISPLAY_AUDIO_CAPTURE,
           media_id.audio_share, false /* disable_local_echo */,
           display_notification_, visible_url, visible_url);

chrome/browser/media/webrtc/media_capture_devices_dispatcher.cc

@@ -34,6 +34,7 @@
 #include "extensions/buildflags/buildflags.h"
 #include "extensions/common/constants.h"
 #include "media/base/media_switches.h"
+#include "third_party/blink/public/common/features.h"
 #include "third_party/blink/public/mojom/mediastream/media_stream.mojom-shared.h"
 
 #if defined(OS_ANDROID)
@@ -163,6 +164,19 @@ void MediaCaptureDevicesDispatcher::ProcessMediaAccessRequest(
   }
 #endif
 
+  // Kill switch for getCurrentBrowsingContextMedia() on browser side to prevent
+  // renderer from bypassing blink side checks.
+  if (request.video_type ==
+      blink::mojom::MediaStreamType::DISPLAY_VIDEO_CAPTURE_THIS_TAB) {
+    if (!base::FeatureList::IsEnabled(
+            blink::features::kRTCGetCurrentBrowsingContextMedia)) {
+      std::move(callback).Run(
+          blink::MediaStreamDevices(),
+          blink::mojom::MediaStreamRequestResult::NOT_SUPPORTED, nullptr);
+      return;
+    }
+  }
+
   for (const auto& handler : media_access_handlers_) {
     if (handler->SupportsStreamType(web_contents, request.video_type,
                                     extension) ||

chrome/browser/media/webrtc/media_stream_capture_indicator.cc

@@ -111,16 +111,13 @@ ObserverMethod GetObserverMethodToCall(const blink::MediaStreamDevice& device) {
     case blink::mojom::MediaStreamType::GUM_DESKTOP_AUDIO_CAPTURE:
     case blink::mojom::MediaStreamType::DISPLAY_VIDEO_CAPTURE:
     case blink::mojom::MediaStreamType::DISPLAY_AUDIO_CAPTURE:
+    case blink::mojom::MediaStreamType::DISPLAY_VIDEO_CAPTURE_THIS_TAB:
       return IsDeviceCapturingDisplay(device)
                  ? &MediaStreamCaptureIndicator::Observer::
                        OnIsCapturingDisplayChanged
                  : &MediaStreamCaptureIndicator::Observer::
                        OnIsCapturingWindowChanged;
 
-    case blink::mojom::MediaStreamType::DISPLAY_VIDEO_CAPTURE_THIS_TAB:
-      // TODO(crbug.com/1136942): Finish wiring getCurrentBrowsingContextMedia.
-      NOTIMPLEMENTED();
-      FALLTHROUGH;
     case blink::mojom::MediaStreamType::NO_SERVICE:
     case blink::mojom::MediaStreamType::NUM_MEDIA_TYPES:
       NOTREACHED();
@@ -324,13 +321,10 @@ int& MediaStreamCaptureIndicator::WebContentsDeviceUsage::GetStreamCount(
     case blink::mojom::MediaStreamType::GUM_DESKTOP_AUDIO_CAPTURE:
     case blink::mojom::MediaStreamType::DISPLAY_VIDEO_CAPTURE:
     case blink::mojom::MediaStreamType::DISPLAY_AUDIO_CAPTURE:
+    case blink::mojom::MediaStreamType::DISPLAY_VIDEO_CAPTURE_THIS_TAB:
       return IsDeviceCapturingDisplay(device) ? display_stream_count_
                                               : window_stream_count_;
 
-    case blink::mojom::MediaStreamType::DISPLAY_VIDEO_CAPTURE_THIS_TAB:
-      // TODO(crbug.com/1136942): Finish wiring getCurrentBrowsingContextMedia.
-      NOTIMPLEMENTED();
-      FALLTHROUGH;
     case blink::mojom::MediaStreamType::NO_SERVICE:
     case blink::mojom::MediaStreamType::NUM_MEDIA_TYPES:
       NOTREACHED();

content/browser/renderer_host/media/in_process_video_capture_device_launcher.cc

@@ -149,8 +149,8 @@ void InProcessVideoCaptureDeviceLauncher::LaunchDeviceAsync(
 #endif  // !defined(OS_ANDROID)
 
     case blink::mojom::MediaStreamType::GUM_DESKTOP_VIDEO_CAPTURE:
-      FALLTHROUGH;
-    case blink::mojom::MediaStreamType::DISPLAY_VIDEO_CAPTURE: {
+    case blink::mojom::MediaStreamType::DISPLAY_VIDEO_CAPTURE:
+    case blink::mojom::MediaStreamType::DISPLAY_VIDEO_CAPTURE_THIS_TAB: {
       const DesktopMediaID desktop_id = DesktopMediaID::Parse(device_id);
       if (desktop_id.is_null()) {
         DLOG(ERROR) << "Desktop media ID is null";

content/browser/renderer_host/media/media_stream_manager.cc

@@ -242,6 +242,8 @@ const char* StreamTypeToString(blink::mojom::MediaStreamType type) {
       return "DISPLAY_AUDIO_CAPTURE";
     case blink::mojom::MediaStreamType::DISPLAY_VIDEO_CAPTURE:
       return "DISPLAY_VIDEO_CAPTURE";
+    case blink::mojom::MediaStreamType::DISPLAY_VIDEO_CAPTURE_THIS_TAB:
+      return "DISPLAY_VIDEO_CAPTURE_THIS_TAB";
     case blink::mojom::MediaStreamType::NO_SERVICE:
       return "NO_SERVICE";
     case blink::mojom::MediaStreamType::NUM_MEDIA_TYPES:
@@ -1465,7 +1467,8 @@ void MediaStreamManager::SetUpRequest(const std::string& label) {
   request->SetVideoType(request->controls.video.stream_type);
 
   const bool is_display_capture =
-      request->video_type() == MediaStreamType::DISPLAY_VIDEO_CAPTURE;
+      request->video_type() == MediaStreamType::DISPLAY_VIDEO_CAPTURE ||
+      request->video_type() == MediaStreamType::DISPLAY_VIDEO_CAPTURE_THIS_TAB;
   if (is_display_capture && !SetUpDisplayCaptureRequest(request)) {
     FinalizeRequestFailed(label, request,
                           MediaStreamRequestResult::SCREEN_CAPTURE_FAILURE);
@@ -1510,7 +1513,9 @@ void MediaStreamManager::SetUpRequest(const std::string& label) {
 
 bool MediaStreamManager::SetUpDisplayCaptureRequest(DeviceRequest* request) {
   DCHECK_CURRENTLY_ON(BrowserThread::IO);
-  DCHECK(request->video_type() == MediaStreamType::DISPLAY_VIDEO_CAPTURE);
+  DCHECK(request->video_type() == MediaStreamType::DISPLAY_VIDEO_CAPTURE ||
+         request->video_type() ==
+             MediaStreamType::DISPLAY_VIDEO_CAPTURE_THIS_TAB);
 
   // getDisplayMedia function does not permit the use of constraints for
   // selection of a source, see

content/browser/renderer_host/media/video_capture_controller.cc

@@ -69,6 +69,11 @@ void LogVideoFrameDrop(media::VideoCaptureFrameDropReason reason,
       UMA_HISTOGRAM_ENUMERATION("Media.VideoCapture.FrameDrop.DisplayCapture",
                                 reason, kEnumCount);
       break;
+    case blink::mojom::MediaStreamType::DISPLAY_VIDEO_CAPTURE_THIS_TAB:
+      UMA_HISTOGRAM_ENUMERATION(
+          "Media.VideoCapture.FrameDrop.DisplayCaptureCurrentTab", reason,
+          kEnumCount);
+      break;
     default:
       // Do nothing
       return;
@@ -103,6 +108,11 @@ void LogMaxConsecutiveVideoFrameDropCountExceeded(
           "Media.VideoCapture.MaxFrameDropExceeded.DisplayCapture", reason,
           kEnumCount);
       break;
+    case blink::mojom::MediaStreamType::DISPLAY_VIDEO_CAPTURE_THIS_TAB:
+      UMA_HISTOGRAM_ENUMERATION(
+          "Media.VideoCapture.MaxFrameDropExceeded.DisplayCaptureCurrentTab",
+          reason, kEnumCount);
+      break;
     default:
       // Do nothing
       return;

tools/metrics/histograms/histograms_xml/histogram_suffixes_list.xml

@@ -19223,6 +19223,7 @@ reviews. Googlers can read more about this at go/gwsq-gerrit.
 <histogram_suffixes name="VideoCaptureFrameDrop" separator=".">
   <suffix name="DeviceCapture" label=""/>
   <suffix name="DisplayCapture" label=""/>
+  <suffix name="DisplayCaptureCurrentTab" label=""/>
   <suffix name="GumDesktopCapture" label=""/>
   <suffix name="GumTabCapture" label=""/>
   <affected-histogram name="Media.VideoCapture.FrameDrop"/>