From 6e76a7fbd85eba67c7c1b02142bdfcefbc5d8279 Mon Sep 17 00:00:00 2001 From: bruvzg <7645683+bruvzg@users.noreply.github.com> Date: Tue, 21 Feb 2023 10:45:49 +0200 Subject: [Linux] Process TTS callback on the main thread to avoid speech-dispatcher deadlock. --- platform/linuxbsd/tts_linux.cpp | 135 ++++++++++++++++++++++------------------ platform/linuxbsd/tts_linux.h | 8 ++- 2 files changed, 81 insertions(+), 62 deletions(-) (limited to 'platform/linuxbsd') diff --git a/platform/linuxbsd/tts_linux.cpp b/platform/linuxbsd/tts_linux.cpp index 04d7c5444f..6b8584ef6c 100644 --- a/platform/linuxbsd/tts_linux.cpp +++ b/platform/linuxbsd/tts_linux.cpp @@ -35,6 +35,11 @@ TTS_Linux *TTS_Linux::singleton = nullptr; +void TTS_Linux::_bind_methods() { + ClassDB::bind_method(D_METHOD("_speech_event", "msg_id", "client_id", "type"), &TTS_Linux::_speech_event); + ClassDB::bind_method(D_METHOD("_speech_index_mark", "msg_id", "client_id", "type", "index_mark"), &TTS_Linux::_speech_index_mark); +} + void TTS_Linux::speech_init_thread_func(void *p_userdata) { TTS_Linux *tts = (TTS_Linux *)p_userdata; if (tts) { @@ -76,76 +81,84 @@ void TTS_Linux::speech_init_thread_func(void *p_userdata) { void TTS_Linux::speech_event_index_mark(size_t p_msg_id, size_t p_client_id, SPDNotificationType p_type, char *p_index_mark) { TTS_Linux *tts = TTS_Linux::get_singleton(); - if (tts && tts->ids.has(p_msg_id)) { - MutexLock thread_safe_method(tts->_thread_safe_); - // Get word offset from the index mark injected to the text stream. - String mark = String::utf8(p_index_mark); - DisplayServer::get_singleton()->tts_post_utterance_event(DisplayServer::TTS_UTTERANCE_BOUNDARY, tts->ids[p_msg_id], mark.to_int()); + if (tts) { + tts->call_deferred(SNAME("_speech_index_mark"), p_msg_id, p_client_id, (int)p_type, String::utf8(p_index_mark)); + } +} + +void TTS_Linux::_speech_index_mark(size_t p_msg_id, size_t p_client_id, int p_type, const String &p_index_mark) { + _THREAD_SAFE_METHOD_ + + if (ids.has(p_msg_id)) { + DisplayServer::get_singleton()->tts_post_utterance_event(DisplayServer::TTS_UTTERANCE_BOUNDARY, ids[p_msg_id], p_index_mark.to_int()); } } void TTS_Linux::speech_event_callback(size_t p_msg_id, size_t p_client_id, SPDNotificationType p_type) { TTS_Linux *tts = TTS_Linux::get_singleton(); if (tts) { - MutexLock thread_safe_method(tts->_thread_safe_); - List &queue = tts->queue; - if (!tts->paused && tts->ids.has(p_msg_id)) { - if (p_type == SPD_EVENT_END) { - DisplayServer::get_singleton()->tts_post_utterance_event(DisplayServer::TTS_UTTERANCE_ENDED, tts->ids[p_msg_id]); - tts->ids.erase(p_msg_id); - tts->last_msg_id = -1; - tts->speaking = false; - } else if (p_type == SPD_EVENT_CANCEL) { - DisplayServer::get_singleton()->tts_post_utterance_event(DisplayServer::TTS_UTTERANCE_CANCELED, tts->ids[p_msg_id]); - tts->ids.erase(p_msg_id); - tts->last_msg_id = -1; - tts->speaking = false; - } + tts->call_deferred(SNAME("_speech_event"), p_msg_id, p_client_id, (int)p_type); + } +} + +void TTS_Linux::_speech_event(size_t p_msg_id, size_t p_client_id, int p_type) { + _THREAD_SAFE_METHOD_ + + if (!paused && ids.has(p_msg_id)) { + if ((SPDNotificationType)p_type == SPD_EVENT_END) { + DisplayServer::get_singleton()->tts_post_utterance_event(DisplayServer::TTS_UTTERANCE_ENDED, ids[p_msg_id]); + ids.erase(p_msg_id); + last_msg_id = -1; + speaking = false; + } else if ((SPDNotificationType)p_type == SPD_EVENT_CANCEL) { + DisplayServer::get_singleton()->tts_post_utterance_event(DisplayServer::TTS_UTTERANCE_CANCELED, ids[p_msg_id]); + ids.erase(p_msg_id); + last_msg_id = -1; + speaking = false; } - if (!tts->speaking && queue.size() > 0) { - DisplayServer::TTSUtterance &message = queue.front()->get(); - - // Inject index mark after each word. - String text; - String language; - SPDVoice **voices = spd_list_synthesis_voices(tts->synth); - if (voices != nullptr) { - SPDVoice **voices_ptr = voices; - while (*voices_ptr != nullptr) { - if (String::utf8((*voices_ptr)->name) == message.voice) { - language = String::utf8((*voices_ptr)->language); - break; - } - voices_ptr++; - } - free_spd_voices(voices); - } - PackedInt32Array breaks = TS->string_get_word_breaks(message.text, language); - for (int i = 0; i < breaks.size(); i += 2) { - const int start = breaks[i]; - const int end = breaks[i + 1]; - text += message.text.substr(start, end - start + 1); - text += ""; - } + } + if (!speaking && queue.size() > 0) { + DisplayServer::TTSUtterance &message = queue.front()->get(); - spd_set_synthesis_voice(tts->synth, message.voice.utf8().get_data()); - spd_set_volume(tts->synth, message.volume * 2 - 100); - spd_set_voice_pitch(tts->synth, (message.pitch - 1) * 100); - float rate = 0; - if (message.rate > 1.f) { - rate = log10(MIN(message.rate, 2.5f)) / log10(2.5f) * 100; - } else if (message.rate < 1.f) { - rate = log10(MAX(message.rate, 0.5f)) / log10(0.5f) * -100; + // Inject index mark after each word. + String text; + String language; + SPDVoice **voices = spd_list_synthesis_voices(synth); + if (voices != nullptr) { + SPDVoice **voices_ptr = voices; + while (*voices_ptr != nullptr) { + if (String::utf8((*voices_ptr)->name) == message.voice) { + language = String::utf8((*voices_ptr)->language); + break; + } + voices_ptr++; } - spd_set_voice_rate(tts->synth, rate); - spd_set_data_mode(tts->synth, SPD_DATA_SSML); - tts->last_msg_id = spd_say(tts->synth, SPD_TEXT, text.utf8().get_data()); - tts->ids[tts->last_msg_id] = message.id; - DisplayServer::get_singleton()->tts_post_utterance_event(DisplayServer::TTS_UTTERANCE_STARTED, message.id); - - queue.pop_front(); - tts->speaking = true; + free_spd_voices(voices); } + PackedInt32Array breaks = TS->string_get_word_breaks(message.text, language); + for (int i = 0; i < breaks.size(); i += 2) { + const int start = breaks[i]; + const int end = breaks[i + 1]; + text += message.text.substr(start, end - start + 1); + text += ""; + } + spd_set_synthesis_voice(synth, message.voice.utf8().get_data()); + spd_set_volume(synth, message.volume * 2 - 100); + spd_set_voice_pitch(synth, (message.pitch - 1) * 100); + float rate = 0; + if (message.rate > 1.f) { + rate = log10(MIN(message.rate, 2.5f)) / log10(2.5f) * 100; + } else if (message.rate < 1.f) { + rate = log10(MAX(message.rate, 0.5f)) / log10(0.5f) * -100; + } + spd_set_voice_rate(synth, rate); + spd_set_data_mode(synth, SPD_DATA_SSML); + last_msg_id = spd_say(synth, SPD_TEXT, text.utf8().get_data()); + ids[last_msg_id] = message.id; + DisplayServer::get_singleton()->tts_post_utterance_event(DisplayServer::TTS_UTTERANCE_STARTED, message.id); + + queue.pop_front(); + speaking = true; } } @@ -204,7 +217,7 @@ void TTS_Linux::speak(const String &p_text, const String &p_voice, int p_volume, if (is_paused()) { resume(); } else { - speech_event_callback(0, 0, SPD_EVENT_BEGIN); + _speech_event(0, 0, (int)SPD_EVENT_BEGIN); } } diff --git a/platform/linuxbsd/tts_linux.h b/platform/linuxbsd/tts_linux.h index 3fe7b659d0..3fcbe3207b 100644 --- a/platform/linuxbsd/tts_linux.h +++ b/platform/linuxbsd/tts_linux.h @@ -45,7 +45,8 @@ #include #endif -class TTS_Linux { +class TTS_Linux : public Object { + GDCLASS(TTS_Linux, Object); _THREAD_SAFE_CLASS_ List queue; @@ -63,6 +64,11 @@ class TTS_Linux { static TTS_Linux *singleton; +protected: + static void _bind_methods(); + void _speech_event(size_t p_msg_id, size_t p_client_id, int p_type); + void _speech_index_mark(size_t p_msg_id, size_t p_client_id, int p_type, const String &p_index_mark); + public: static TTS_Linux *get_singleton(); -- cgit v1.2.3