Cancel speech recognition by timeout.

td/telegram/Td.cpp

@@ -3558,6 +3558,8 @@ void Td::clear() {
   LOG(DEBUG) << "TopDialogManager actor was cleared" << timer;
   updates_manager_actor_.reset();
   LOG(DEBUG) << "UpdatesManager actor was cleared" << timer;
+  voice_notes_manager_actor_.reset();
+  LOG(DEBUG) << "VoiceNotesManager actor was cleared" << timer;
   web_pages_manager_actor_.reset();
   LOG(DEBUG) << "WebPagesManager actor was cleared" << timer;
 }
@@ -3981,7 +3983,6 @@ void Td::init_managers() {
   documents_manager_ = make_unique<DocumentsManager>(this);
   video_notes_manager_ = make_unique<VideoNotesManager>(this);
   videos_manager_ = make_unique<VideosManager>(this);
-  voice_notes_manager_ = make_unique<VoiceNotesManager>(this);
 
   animations_manager_ = make_unique<AnimationsManager>(this, create_reference());
   animations_manager_actor_ = register_actor("AnimationsManager", animations_manager_.get());
@@ -4037,6 +4038,8 @@ void Td::init_managers() {
   updates_manager_ = make_unique<UpdatesManager>(this, create_reference());
   updates_manager_actor_ = register_actor("UpdatesManager", updates_manager_.get());
   G()->set_updates_manager(updates_manager_actor_.get());
+  voice_notes_manager_ = make_unique<VoiceNotesManager>(this, create_reference());
+  voice_notes_manager_actor_ = register_actor("VoiceNotesManager", voice_notes_manager_.get());
   web_pages_manager_ = make_unique<WebPagesManager>(this, create_reference());
   web_pages_manager_actor_ = register_actor("WebPagesManager", web_pages_manager_.get());
   G()->set_web_pages_manager(web_pages_manager_actor_.get());

td/telegram/Td.h

@@ -137,7 +137,6 @@ class Td final : public Actor {
   unique_ptr<DocumentsManager> documents_manager_;
   unique_ptr<VideoNotesManager> video_notes_manager_;
   unique_ptr<VideosManager> videos_manager_;
-  unique_ptr<VoiceNotesManager> voice_notes_manager_;
 
   unique_ptr<AnimationsManager> animations_manager_;
   ActorOwn<AnimationsManager> animations_manager_actor_;
@@ -185,6 +184,8 @@ class Td final : public Actor {
   ActorOwn<TopDialogManager> top_dialog_manager_actor_;
   unique_ptr<UpdatesManager> updates_manager_;
   ActorOwn<UpdatesManager> updates_manager_actor_;
+  unique_ptr<VoiceNotesManager> voice_notes_manager_;
+  ActorOwn<VoiceNotesManager> voice_notes_manager_actor_;
   unique_ptr<WebPagesManager> web_pages_manager_;
   ActorOwn<WebPagesManager> web_pages_manager_actor_;
 

td/telegram/VoiceNotesManager.cpp

@@ -94,7 +94,25 @@ class RateTranscribedAudioQuery final : public Td::ResultHandler {
   }
 };
 
-VoiceNotesManager::VoiceNotesManager(Td *td) : td_(td) {
+VoiceNotesManager::VoiceNotesManager(Td *td, ActorShared<> parent) : td_(td), parent_(std::move(parent)) {
+  voice_note_transcription_timeout_.set_callback(on_voice_note_transcription_timeout_callback);
+  voice_note_transcription_timeout_.set_callback_data(static_cast<void *>(this));
+}
+
+void VoiceNotesManager::tear_down() {
+  parent_.reset();
+}
+
+void VoiceNotesManager::on_voice_note_transcription_timeout_callback(void *voice_notes_manager_ptr,
+                                                                     int64 transcription_id) {
+  if (G()->close_flag()) {
+    return;
+  }
+
+  auto voice_notes_manager = static_cast<VoiceNotesManager *>(voice_notes_manager_ptr);
+  send_closure_later(voice_notes_manager->actor_id(voice_notes_manager),
+                     &VoiceNotesManager::on_pending_voice_note_transcription_failed, transcription_id,
+                     Status::Error(500, "Timeout expired"));
 }
 
 int32 VoiceNotesManager::get_voice_note_duration(FileId file_id) const {
@@ -302,6 +320,7 @@ void VoiceNotesManager::on_voice_note_transcribed(FileId file_id, string &&text,
     }
     bool is_inserted = pending_voice_note_transcription_queries_.emplace(transcription_id, file_id).second;
     CHECK(is_inserted);
+    voice_note_transcription_timeout_.set_timeout_in(transcription_id, TRANSCRIPTION_TIMEOUT);
   }
 }
 
@@ -335,6 +354,7 @@ void VoiceNotesManager::on_update_transcribed_audio(string &&text, int64 transcr
   }
   auto file_id = it->second;
   pending_voice_note_transcription_queries_.erase(it);
+  voice_note_transcription_timeout_.cancel_timeout(transcription_id);
 
   on_voice_note_transcribed(file_id, std::move(text), transcription_id, is_final);
 }
@@ -346,6 +366,8 @@ void VoiceNotesManager::on_pending_voice_note_transcription_failed(int64 transcr
   }
   auto file_id = it->second;
   pending_voice_note_transcription_queries_.erase(it);
+  voice_note_transcription_timeout_.cancel_timeout(transcription_id);
+
   on_voice_note_transcription_failed(file_id, std::move(error));
 }
 

td/telegram/VoiceNotesManager.h

@@ -13,7 +13,9 @@
 #include "td/telegram/telegram_api.h"
 #include "td/telegram/Version.h"
 
+#include "td/actor/actor.h"
 #include "td/actor/PromiseFuture.h"
+#include "td/actor/Timeout.h"
 
 #include "td/utils/common.h"
 #include "td/utils/FlatHashMap.h"
@@ -24,9 +26,9 @@ namespace td {
 
 class Td;
 
-class VoiceNotesManager {
+class VoiceNotesManager final : public Actor {
  public:
-  explicit VoiceNotesManager(Td *td);
+  VoiceNotesManager(Td *td, ActorShared<> parent);
 
   int32 get_voice_note_duration(FileId file_id) const;
 
@@ -66,6 +68,8 @@ class VoiceNotesManager {
   FileId parse_voice_note(ParserT &parser);
 
  private:
+  static constexpr int32 TRANSCRIPTION_TIMEOUT = 60;
+
   class VoiceNote {
    public:
     string mime_type;
@@ -78,6 +82,8 @@ class VoiceNotesManager {
     FileId file_id;
   };
 
+  static void on_voice_note_transcription_timeout_callback(void *voice_notes_manager_ptr, int64 transcription_id);
+
   VoiceNote *get_voice_note(FileId file_id);
 
   const VoiceNote *get_voice_note(FileId file_id) const;
@@ -88,11 +94,16 @@ class VoiceNotesManager {
 
   void on_voice_note_transcription_updated(FileId file_id);
 
+  void tear_down() final;
+
   Td *td_;
+  ActorShared<> parent_;
+
   FlatHashMap<FileId, unique_ptr<VoiceNote>, FileIdHash> voice_notes_;
 
   FlatHashMap<FileId, vector<Promise<Unit>>, FileIdHash> speech_recognition_queries_;
   FlatHashMap<int64, FileId> pending_voice_note_transcription_queries_;
+  MultiTimeout voice_note_transcription_timeout_{"VoiceNoteTranscriptionTimeout"};
 
   FlatHashMap<FileId, FlatHashSet<FullMessageId, FullMessageIdHash>, FileIdHash> voice_note_messages_;
   FlatHashMap<FullMessageId, FileId, FullMessageIdHash> message_voice_notes_;

test/online.cpp

@@ -424,7 +424,7 @@ class TestDownloadFile : public Task {
       unlink(file.local_->path_).ignore();
     }
 
-    size_t size = file.size_;
+    size_t size = narrow_cast<size_t>(file.size_);
     Random::Xorshift128plus rnd(123);
 
     size_t begin = 0;