nodejs-electron/disable-webspeech.patch

From d617766b236a93749ddbb50b75573dd35238ffc9 Mon Sep 17 00:00:00 2001
From: Allan Sandfeld Jensen <allan.jensen@theqtcompany.com>
Date: Fri, 29 Jan 2016 14:20:48 +0100
Subject: Fix building without WebSpeech

WebSpeech requires Google API keys, so we do not support it, and
disabling it also gets rid of libspeex and libFLAC.

Also splits speech recognition from speech synthesis, so we can
disable the former without disabling the latter.

Change-Id: I80a54ec4c6a13ea4c443289aa42369196c2e095a
Reviewed-by: Michal Klocek <michal.klocek@qt.io>
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
---
 chromium/content/browser/BUILD.gn                  | 56 ++++++++++++----------
 .../content/browser/browser_interface_binders.cc   | 10 ++++
 chromium/content/browser/browser_main_loop.cc      |  7 ++-
 chromium/content/browser/browser_main_loop.h       |  2 +
 chromium/content/child/runtime_features.cc         |  1 -
 chromium/media/BUILD.gn                            |  1 +
 chromium/media/media_options.gni                   |  2 +
 7 files changed, 52 insertions(+), 27 deletions(-)

--- src/content/browser/BUILD.gn.orig	2024-12-08 18:44:00.234689649 +0100
+++ src/content/browser/BUILD.gn	2024-12-18 12:20:19.753578327 +0100
@@ -2134,13 +2134,6 @@ source_set("browser") {
     "sms/user_consent_handler.h",
     "sms/webotp_service.cc",
     "sms/webotp_service.h",
-    "speech/speech_recognition_dispatcher_host.cc",
-    "speech/speech_recognition_dispatcher_host.h",
-    "speech/speech_recognition_manager_impl.cc",
-    "speech/speech_recognition_manager_impl.h",
-    "speech/speech_recognition_session.cc",
-    "speech/speech_recognition_session.h",
-    "speech/speech_recognizer.h",
     "speech/speech_synthesis_impl.cc",
     "speech/speech_synthesis_impl.h",
     "speech/tts_controller_impl.cc",
@@ -3340,13 +3333,6 @@ source_set("browser") {
       "service_worker/service_worker_usb_delegate_observer.cc",
       "service_worker/service_worker_usb_delegate_observer.h",
 
-      # Most speech code is non-Android.
-      "speech/network_speech_recognition_engine_impl.cc",
-      "speech/network_speech_recognition_engine_impl.h",
-      "speech/speech_recognition_engine.cc",
-      "speech/speech_recognition_engine.h",
-      "speech/speech_recognizer_impl.cc",
-      "speech/speech_recognizer_impl.h",
       "tracing/tracing_ui.cc",
       "tracing/tracing_ui.h",
 
@@ -3397,10 +3383,8 @@ source_set("browser") {
       "//components/soda:constants",
       "//components/soda:soda",
       "//components/soda:utils",
-      "//components/speech:speech",
       "//components/vector_icons",
       "//components/webauthn/json",
-      "//media/mojo/mojom:speech_recognition",
       "//ui/base",
     ]
   }
@@ -3471,6 +3455,37 @@ source_set("browser") {
     deps += [ "//ui/compositor" ]
   }
 
+  if (enable_web_speech) {
+    sources += [
+      "speech/audio_buffer.cc",
+      "speech/audio_buffer.h",
+      "speech/audio_encoder.cc",
+      "speech/audio_encoder.h",
+      "speech/chunked_byte_buffer.cc",
+      "speech/chunked_byte_buffer.h",
+      "speech/endpointer/endpointer.cc",
+      "speech/endpointer/endpointer.h",
+      "speech/endpointer/energy_endpointer.cc",
+      "speech/endpointer/energy_endpointer.h",
+      "speech/endpointer/energy_endpointer_params.cc",
+      "speech/endpointer/energy_endpointer_params.h",
+      "speech/speech_recognition_engine.cc",
+      "speech/speech_recognition_engine.h",
+      "speech/speech_recognizer_impl.cc",
+      "speech/speech_recognizer_impl.h",
+      "speech/speech_recognition_dispatcher_host.cc",
+      "speech/speech_recognition_dispatcher_host.h",
+      "speech/speech_recognition_manager_impl.cc",
+      "speech/speech_recognition_manager_impl.h",
+      "speech/speech_recognizer.h",
+    ]
+    deps += [
+      "//components/speech",
+      "//content/browser/speech/proto",
+      "//third_party/flac",
+    ]
+  }
+
   if ((is_linux || is_chromeos) && use_dbus) {
     deps += [ "//dbus" ]
   }
--- src/content/browser/browser_interface_binders.cc.orig	2025-07-19 11:32:34.193979827 +0200
+++ src/content/browser/browser_interface_binders.cc	2025-07-20 14:51:05.789212478 +0200
@@ -62,7 +62,9 @@
 #include "content/browser/renderer_host/render_process_host_impl.h"
 #include "content/browser/service_worker/service_worker_host.h"
 #include "content/browser/shared_storage/shared_storage_worklet_host.h"
+#if BUILDFLAG(ENABLE_WEB_SPEECH)
 #include "content/browser/speech/speech_recognition_dispatcher_host.h"
+#endif
 #include "content/browser/storage_access/storage_access_handle.h"
 #include "content/browser/tracing/trace_report/trace_report.mojom.h"
 #include "content/browser/tracing/trace_report/trace_report_internals_ui.h"
@@ -101,7 +103,9 @@
 #include "media/mojo/mojom/media_metrics_provider.mojom.h"
 #include "media/mojo/mojom/media_player.mojom.h"
 #include "media/mojo/mojom/remoting.mojom.h"
+#if BUILDFLAG(ENABLE_WEB_SPEECH)
 #include "media/mojo/mojom/speech_recognizer.mojom.h"
+#endif
 #include "media/mojo/mojom/video_decode_perf_history.mojom.h"
 #include "media/mojo/mojom/video_encoder_metrics_provider.mojom.h"
 #include "media/mojo/mojom/webrtc_video_perf.mojom.h"
@@ -209,7 +213,9 @@
 #else  // BUILDFLAG(IS_ANDROID)
 #include "content/browser/direct_sockets/direct_sockets_service_impl.h"
 #include "media/mojo/mojom/renderer_extensions.mojom.h"
+#if BUILDFLAG(ENABLE_WEB_SPEECH)
 #include "media/mojo/mojom/speech_recognition.mojom.h"  // nogncheck
+#endif
 #include "third_party/blink/public/mojom/hid/hid.mojom.h"
 #include "third_party/blink/public/mojom/installedapp/installed_app_provider.mojom.h"
 #endif  // BUILDFLAG(IS_ANDROID)
@@ -865,11 +871,13 @@ void PopulateFrameBinders(RenderFrameHos
   map->Add<blink::mojom::SharedWorkerConnector>(
       base::BindRepeating(&BindSharedWorkerConnector, base::Unretained(host)));
 
+#if BUILDFLAG(ENABLE_WEB_SPEECH)
   map->Add<media::mojom::SpeechRecognizer>(
       base::BindRepeating(&SpeechRecognitionDispatcherHost::Create,
                           host->GetProcess()->GetDeprecatedID(),
                           host->GetRoutingID()),
       GetIOThreadTaskRunner({}));
+#endif
 
   map->Add<blink::mojom::SpeechSynthesis>(base::BindRepeating(
       &RenderFrameHostImpl::GetSpeechSynthesis, base::Unretained(host)));
@@ -1194,6 +1202,7 @@ void PopulateBinderMapWithContext(
 #if !BUILDFLAG(IS_ANDROID)
   map->Add<blink::mojom::DirectSocketsService>(
       base::BindRepeating(&DirectSocketsServiceImpl::CreateForFrame));
+#if BUILDFLAG(ENABLE_WEB_SPEECH)
   map->Add<media::mojom::SpeechRecognitionContext>(base::BindRepeating(
       &EmptyBinderForFrame<media::mojom::SpeechRecognitionContext>));
   map->Add<media::mojom::SpeechRecognitionClientBrowserInterface>(
@@ -1205,6 +1214,7 @@ void PopulateBinderMapWithContext(
   map->Add<media::mojom::MediaPlayerObserverClient>(base::BindRepeating(
       &EmptyBinderForFrame<media::mojom::MediaPlayerObserverClient>));
 #endif
+#endif
 #if BUILDFLAG(ENABLE_UNHANDLED_TAP)
   map->Add<blink::mojom::UnhandledTapNotifier>(base::BindRepeating(
       &EmptyBinderForFrame<blink::mojom::UnhandledTapNotifier>));
--- src/content/browser/browser_main_loop.cc.orig	2025-07-19 11:32:34.193979827 +0200
+++ src/content/browser/browser_main_loop.cc	2025-07-20 14:51:05.789212478 +0200
@@ -91,7 +91,9 @@
 #include "content/browser/screenlock_monitor/screenlock_monitor_device_source.h"
 #include "content/browser/service_host/utility_process_host.h"
 #include "content/browser/sms/sms_provider.h"
+#if BUILDFLAG(ENABLE_WEB_SPEECH)
 #include "content/browser/speech/speech_recognition_manager_impl.h"
+#endif
 #include "content/browser/speech/tts_controller_impl.h"
 #include "content/browser/startup_data_impl.h"
 #include "content/browser/startup_task_runner.h"
@@ -1162,12 +1164,14 @@ void BrowserMainLoop::ShutdownThreadsAnd
     midi_service_->Shutdown();
   }
 
+#if BUILDFLAG(ENABLE_WEB_SPEECH)
   {
     TRACE_EVENT0("shutdown",
                  "BrowserMainLoop::Subsystem:SpeechRecognitionManager");
     io_thread_->task_runner()->DeleteSoon(
         FROM_HERE, speech_recognition_manager_.release());
   }
+#endif
 
   TtsControllerImpl::GetInstance()->Shutdown();
 
@@ -1405,13 +1409,14 @@ void BrowserMainLoop::PostCreateThreadsI
     media_stream_manager_ =
         std::make_unique<MediaStreamManager>(audio_system_.get());
   }
-
+#if BUILDFLAG(ENABLE_WEB_SPEECH)
   {
     TRACE_EVENT0("startup",
                  "BrowserMainLoop::PostCreateThreads:InitSpeechRecognition");
     speech_recognition_manager_.reset(new SpeechRecognitionManagerImpl(
         audio_system_.get(), media_stream_manager_.get()));
   }
+#endif
 
   {
     TRACE_EVENT0("startup",
diff --git a/chromium/content/browser/browser_main_loop.h b/chromium/content/browser/browser_main_loop.h
index a8c8c6b5d16..7d92e0a701e 100644
--- a/content/browser/browser_main_loop.h
+++ b/content/browser/browser_main_loop.h
@@ -367,8 +367,10 @@ class CONTENT_EXPORT BrowserMainLoop {
 
   std::unique_ptr<midi::MidiService> midi_service_;
 
+#if defined(ENABLE_WEB_SPEECH)
   // Must be deleted on the IO thread.
   std::unique_ptr<SpeechRecognitionManagerImpl> speech_recognition_manager_;
+#endif
 
 #if BUILDFLAG(IS_WIN)
   std::unique_ptr<media::SystemMessageWindowWin> system_message_window_;
diff --git a/chromium/content/child/runtime_features.cc b/chromium/content/child/runtime_features.cc
index 3a2e713ccc3..6af5afde20e 100644
--- a/content/child/runtime_features.cc
+++ b/content/child/runtime_features.cc
@@ -465,7 +465,6 @@ void SetRuntimeFeaturesFromCommandLine(const base::CommandLine& command_line) {
       {wrf::EnablePushMessaging, switches::kDisableNotifications, false},
       {wrf::EnableScriptedSpeechRecognition, switches::kDisableSpeechAPI,
        false},
-      {wrf::EnableScriptedSpeechSynthesis, switches::kDisableSpeechAPI, false},
       {wrf::EnableScriptedSpeechSynthesis, switches::kDisableSpeechSynthesisAPI,
        false},
       {wrf::EnableSharedWorker, switches::kDisableSharedWorkers, false},
--- src/media/BUILD.gn.orig	2025-07-19 11:32:35.633979930 +0200
+++ src/media/BUILD.gn	2025-07-20 14:51:07.213212462 +0200
@@ -61,6 +61,7 @@ buildflag_header("media_buildflags") {
     "ENABLE_MSE_MPEG2TS_STREAM_PARSER=$enable_mse_mpeg2ts_stream_parser",
     "PLATFORM_HAS_OPTIONAL_HEVC_DECODE_SUPPORT=$platform_has_optional_hevc_decode_support",
     "PLATFORM_HAS_OPTIONAL_HEVC_ENCODE_SUPPORT=$platform_has_optional_hevc_encode_support",
+    "ENABLE_WEB_SPEECH=$enable_web_speech",
     "USE_ARC_PROTECTED_MEDIA=$use_arc_protected_media",
     "USE_CHROMEOS_PROTECTED_AV1=$use_chromeos_protected_av1",
     "USE_CHROMEOS_PROTECTED_MEDIA=$use_chromeos_protected_media",
diff --git a/chromium/media/media_options.gni b/chromium/media/media_options.gni
index 5667cac61f3..48ea95a99a6 100644
--- a/media/media_options.gni
+++ b/media/media_options.gni
@@ -66,6 +66,8 @@ declare_args() {
   # kAllowClearDolbyVisionInMseWhenPlatformEncryptedDvEnabled.
   enable_platform_encrypted_dolby_vision = proprietary_codecs && is_win
 
+  enable_web_speech = true
+
   # Enable logging override, e.g. enable DVLOGs through level 2 at build time.
   # On Cast devices, these are logged as INFO.
   # When enabled on Fuchsia, these are logged as VLOGs.
-- 
cgit v1.2.1


--- src/components/speech/BUILD.gn.orig	2024-12-08 18:34:57.021340049 +0100
+++ src/components/speech/BUILD.gn	2024-12-18 12:18:14.187490208 +0100
@@ -6,12 +6,6 @@ import("//build/config/features.gni")
 
 source_set("speech") {
   sources = [
-    "audio_buffer.cc",
-    "audio_buffer.h",
-    "audio_encoder.cc",
-    "audio_encoder.h",
-    "chunked_byte_buffer.cc",
-    "chunked_byte_buffer.h",
     "downstream_loader.cc",
     "downstream_loader.h",
     "downstream_loader_client.h",
@@ -27,7 +21,6 @@ source_set("speech") {
     "//mojo/public/cpp/system",
     "//services/network/public/cpp",
     "//services/network/public/mojom",
-    "//third_party/flac",
   ]
 
   if (!is_android) {