Diff
Modified: trunk/Source/WebCore/ChangeLog (272336 => 272337)
--- trunk/Source/WebCore/ChangeLog 2021-02-03 20:36:28 UTC (rev 272336)
+++ trunk/Source/WebCore/ChangeLog 2021-02-03 20:40:15 UTC (rev 272337)
@@ -1,3 +1,35 @@
+2021-02-03 Sihui Liu <[email protected]>
+
+ Create a SpeechRecognizer for each SpeechRecognitionRequest
+ https://bugs.webkit.org/show_bug.cgi?id=219699
+ <rdar://problem/72392097>
+
+ Reviewed by Youenn Fablet.
+
+ Make SpeechRecognizer take the ownerwhip of the SpeechRecognitionRequest it's about to handle, and manage its
+ state.
+
+ * Modules/speech/SpeechRecognizer.cpp:
+ (WebCore::SpeechRecognizer::SpeechRecognizer):
+ (WebCore::SpeechRecognizer::~SpeechRecognizer):
+ (WebCore::SpeechRecognizer::abort):
+ (WebCore::SpeechRecognizer::stop):
+ (WebCore::SpeechRecognizer::clientIdentifier const):
+ (WebCore::SpeechRecognizer::start):
+ (WebCore::SpeechRecognizer::startCapture):
+ (WebCore::SpeechRecognizer::stopCapture):
+ (WebCore::SpeechRecognizer::startRecognition):
+ (WebCore::SpeechRecognizer::abortRecognition):
+ (WebCore::SpeechRecognizer::stopRecognition):
+ (WebCore::SpeechRecognizer::reset): Deleted.
+ (WebCore::SpeechRecognizer::resetRecognition): Deleted.
+ * Modules/speech/SpeechRecognizer.h:
+ (WebCore::SpeechRecognizer::setInactive):
+ (WebCore::SpeechRecognizer::currentClientIdentifier const): Deleted.
+ * Modules/speech/cocoa/SpeechRecognizerCocoa.mm:
+ (WebCore::SpeechRecognizer::startRecognition):
+ (WebCore::SpeechRecognizer::resetRecognition): Deleted.
+
2021-02-03 Ryan Haddad <[email protected]>
Unreviewed, reverting r272300.
Modified: trunk/Source/WebCore/Modules/speech/SpeechRecognizer.cpp (272336 => 272337)
--- trunk/Source/WebCore/Modules/speech/SpeechRecognizer.cpp 2021-02-03 20:36:28 UTC (rev 272336)
+++ trunk/Source/WebCore/Modules/speech/SpeechRecognizer.cpp 2021-02-03 20:40:15 UTC (rev 272337)
@@ -26,6 +26,7 @@
#include "config.h"
#include "SpeechRecognizer.h"
+#include "SpeechRecognitionRequest.h"
#include "SpeechRecognitionUpdate.h"
#include <wtf/MediaTime.h>
@@ -35,24 +36,27 @@
namespace WebCore {
-SpeechRecognizer::SpeechRecognizer(DelegateCallback&& callback)
- : m_delegateCallback(WTFMove(callback))
+SpeechRecognizer::SpeechRecognizer(DelegateCallback&& delegateCallback, UniqueRef<SpeechRecognitionRequest>&& request)
+ : m_delegateCallback(WTFMove(delegateCallback))
+ , m_request(WTFMove(request))
{
}
-void SpeechRecognizer::reset()
+SpeechRecognizer::~SpeechRecognizer()
{
- if (!m_clientIdentifier)
+ if (m_state == State::Aborting || m_state == State::Stopping || m_state == State::Running)
+ m_delegateCallback(SpeechRecognitionUpdate::create(clientIdentifier(), SpeechRecognitionUpdateType::End));
+}
+
+void SpeechRecognizer::abort(Optional<SpeechRecognitionError>&& error)
+{
+ if (m_state == State::Aborting || m_state == State::Inactive)
return;
+ m_state = State::Aborting;
- stopCapture();
- resetRecognition();
- m_clientIdentifier = WTF::nullopt;
-}
+ if (error)
+ m_delegateCallback(SpeechRecognitionUpdate::createError(clientIdentifier(), *error));
-void SpeechRecognizer::abort()
-{
- ASSERT(m_clientIdentifier);
stopCapture();
abortRecognition();
}
@@ -59,25 +63,31 @@
void SpeechRecognizer::stop()
{
- ASSERT(m_clientIdentifier);
+ if (m_state == State::Aborting || m_state == State::Inactive)
+ return;
+ m_state = State::Stopping;
+
stopCapture();
stopRecognition();
}
+SpeechRecognitionConnectionClientIdentifier SpeechRecognizer::clientIdentifier() const
+{
+ return m_request->clientIdentifier();
+}
+
#if ENABLE(MEDIA_STREAM)
-void SpeechRecognizer::start(SpeechRecognitionConnectionClientIdentifier clientIdentifier, Ref<RealtimeMediaSource>&& source, bool mockSpeechRecognitionEnabled, const String& localeIdentifier, bool continuous, bool interimResults, uint64_t maxAlternatives)
+void SpeechRecognizer::start(Ref<RealtimeMediaSource>&& source, bool mockSpeechRecognitionEnabled)
{
- ASSERT(!m_clientIdentifier);
- m_clientIdentifier = clientIdentifier;
- m_delegateCallback(SpeechRecognitionUpdate::create(*m_clientIdentifier, SpeechRecognitionUpdateType::Start));
-
- if (!startRecognition(mockSpeechRecognitionEnabled, clientIdentifier, localeIdentifier, continuous, interimResults, maxAlternatives)) {
- auto error = WebCore::SpeechRecognitionError { WebCore::SpeechRecognitionErrorType::ServiceNotAllowed, "Failed to start recognition"_s };
- m_delegateCallback(WebCore::SpeechRecognitionUpdate::createError(clientIdentifier, WTFMove(error)));
+ if (!startRecognition(mockSpeechRecognitionEnabled, clientIdentifier(), m_request->lang(), m_request->continuous(), m_request->interimResults(), m_request->maxAlternatives())) {
+ auto error = SpeechRecognitionError { SpeechRecognitionErrorType::ServiceNotAllowed, "Failed to start recognition"_s };
+ m_delegateCallback(SpeechRecognitionUpdate::createError(clientIdentifier(), WTFMove(error)));
return;
}
+ m_state = State::Running;
+ m_delegateCallback(SpeechRecognitionUpdate::create(clientIdentifier(), SpeechRecognitionUpdateType::Start));
startCapture(WTFMove(source));
}
@@ -88,15 +98,12 @@
weakThis->dataCaptured(time, data, description, sampleCount);
};
- auto stateUpdateCallback = [this, weakThis = makeWeakPtr(this)](const auto& update) {
- if (!weakThis)
- return;
-
- ASSERT(m_clientIdentifier && m_clientIdentifier.value() == update.clientIdentifier());
- m_delegateCallback(update);
+ auto stateUpdateCallback = [weakThis = makeWeakPtr(this)](const auto& update) {
+ if (weakThis)
+ weakThis->m_delegateCallback(update);
};
- m_source = makeUnique<SpeechRecognitionCaptureSource>(*m_clientIdentifier, WTFMove(dataCallback), WTFMove(stateUpdateCallback), WTFMove(source));
+ m_source = makeUnique<SpeechRecognitionCaptureSource>(clientIdentifier(), WTFMove(dataCallback), WTFMove(stateUpdateCallback), WTFMove(source));
}
#endif
@@ -107,7 +114,7 @@
return;
m_source = nullptr;
- m_delegateCallback(SpeechRecognitionUpdate::create(*m_clientIdentifier, SpeechRecognitionUpdateType::AudioEnd));
+ m_delegateCallback(SpeechRecognitionUpdate::create(clientIdentifier(), SpeechRecognitionUpdateType::AudioEnd));
}
#if !HAVE(SPEECHRECOGNIZER)
@@ -118,29 +125,19 @@
bool SpeechRecognizer::startRecognition(bool, SpeechRecognitionConnectionClientIdentifier, const String&, bool, bool, uint64_t)
{
- m_isRecognizing = true;
return true;
}
void SpeechRecognizer::abortRecognition()
{
- m_isRecognizing = false;
- m_delegateCallback(SpeechRecognitionUpdate::create(*m_clientIdentifier, SpeechRecognitionUpdateType::End));
+ m_delegateCallback(SpeechRecognitionUpdate::create(clientIdentifier(), SpeechRecognitionUpdateType::End));
}
void SpeechRecognizer::stopRecognition()
{
- m_isRecognizing = false;
- m_delegateCallback(SpeechRecognitionUpdate::create(*m_clientIdentifier, SpeechRecognitionUpdateType::End));
+ m_delegateCallback(SpeechRecognitionUpdate::create(clientIdentifier(), SpeechRecognitionUpdateType::End));
}
-void SpeechRecognizer::resetRecognition()
-{
- if (!m_isRecognizing)
- return;
- abortRecognition();
-}
-
#endif
} // namespace WebCore
Modified: trunk/Source/WebCore/Modules/speech/SpeechRecognizer.h (272336 => 272337)
--- trunk/Source/WebCore/Modules/speech/SpeechRecognizer.h 2021-02-03 20:36:28 UTC (rev 272336)
+++ trunk/Source/WebCore/Modules/speech/SpeechRecognizer.h 2021-02-03 20:40:15 UTC (rev 272337)
@@ -27,6 +27,8 @@
#include "SpeechRecognitionCaptureSource.h"
#include "SpeechRecognitionConnectionClientIdentifier.h"
+#include "SpeechRecognitionError.h"
+#include <wtf/UniqueRef.h>
#if HAVE(SPEECHRECOGNIZER)
#include <wtf/RetainPtr.h>
@@ -35,6 +37,7 @@
namespace WebCore {
+class SpeechRecognitionRequest;
class SpeechRecognitionUpdate;
class SpeechRecognizer : public CanMakeWeakPtr<SpeechRecognizer> {
@@ -41,21 +44,27 @@
WTF_MAKE_FAST_ALLOCATED;
public:
using DelegateCallback = Function<void(const SpeechRecognitionUpdate&)>;
- WEBCORE_EXPORT explicit SpeechRecognizer(DelegateCallback&&);
- WEBCORE_EXPORT ~SpeechRecognizer() = default;
+ WEBCORE_EXPORT explicit SpeechRecognizer(DelegateCallback&&, UniqueRef<SpeechRecognitionRequest>&&);
+ WEBCORE_EXPORT ~SpeechRecognizer();
#if ENABLE(MEDIA_STREAM)
- WEBCORE_EXPORT void start(SpeechRecognitionConnectionClientIdentifier, Ref<RealtimeMediaSource>&&, bool mockSpeechRecognitionEnabled, const String& localeIdentifier, bool continuous, bool interimResults, uint64_t maxAlternatives);
+ WEBCORE_EXPORT void start(Ref<RealtimeMediaSource>&&, bool mockSpeechRecognitionEnabled);
#endif
- WEBCORE_EXPORT void reset();
- WEBCORE_EXPORT void abort();
+ WEBCORE_EXPORT void abort(Optional<SpeechRecognitionError>&& = WTF::nullopt);
WEBCORE_EXPORT void stop();
- Optional<SpeechRecognitionConnectionClientIdentifier> currentClientIdentifier() const { return m_clientIdentifier; }
+ WEBCORE_EXPORT SpeechRecognitionConnectionClientIdentifier clientIdentifier() const;
SpeechRecognitionCaptureSource* source() { return m_source.get(); }
+ void setInactive() { m_state = State::Inactive; }
+
private:
- void stopInternal();
+ enum class State {
+ Inactive,
+ Running,
+ Stopping,
+ Aborting
+ };
#if ENABLE(MEDIA_STREAM)
void startCapture(Ref<RealtimeMediaSource>&&);
@@ -65,17 +74,14 @@
bool startRecognition(bool mockSpeechRecognitionEnabled, SpeechRecognitionConnectionClientIdentifier, const String& localeIdentifier, bool continuous, bool interimResults, uint64_t alternatives);
void abortRecognition();
void stopRecognition();
- void resetRecognition();
- Optional<SpeechRecognitionConnectionClientIdentifier> m_clientIdentifier;
DelegateCallback m_delegateCallback;
+ UniqueRef<SpeechRecognitionRequest> m_request;
std::unique_ptr<SpeechRecognitionCaptureSource> m_source;
+ State m_state { State::Inactive };
#if HAVE(SPEECHRECOGNIZER)
RetainPtr<WebSpeechRecognizerTask> m_task;
-#else
- // For testing.
- bool m_isRecognizing { false };
#endif
};
Modified: trunk/Source/WebCore/Modules/speech/cocoa/SpeechRecognizerCocoa.mm (272336 => 272337)
--- trunk/Source/WebCore/Modules/speech/cocoa/SpeechRecognizerCocoa.mm 2021-02-03 20:36:28 UTC (rev 272336)
+++ trunk/Source/WebCore/Modules/speech/cocoa/SpeechRecognizerCocoa.mm 2021-02-03 20:40:15 UTC (rev 272337)
@@ -45,11 +45,9 @@
bool SpeechRecognizer::startRecognition(bool mockSpeechRecognitionEnabled, SpeechRecognitionConnectionClientIdentifier identifier, const String& localeIdentifier, bool continuous, bool interimResults, uint64_t alternatives)
{
auto taskClass = mockSpeechRecognitionEnabled ? [WebSpeechRecognizerTaskMock class] : [WebSpeechRecognizerTask class];
- m_task = adoptNS([[taskClass alloc] initWithIdentifier:identifier locale:localeIdentifier doMultipleRecognitions:continuous reportInterimResults:interimResults maxAlternatives:alternatives delegateCallback:[this, weakThis = makeWeakPtr(this)](const WebCore::SpeechRecognitionUpdate& update) {
- if (!weakThis)
- return;
-
- m_delegateCallback(update);
+ m_task = adoptNS([[taskClass alloc] initWithIdentifier:identifier locale:localeIdentifier doMultipleRecognitions:continuous reportInterimResults:interimResults maxAlternatives:alternatives delegateCallback:[weakThis = makeWeakPtr(this)](const WebCore::SpeechRecognitionUpdate& update) {
+ if (weakThis)
+ weakThis->m_delegateCallback(update);
}]);
return !!m_task;
@@ -67,15 +65,6 @@
[m_task abort];
}
-void SpeechRecognizer::resetRecognition()
-{
- if (!m_task)
- return;
-
- auto task = std::exchange(m_task, nullptr);
- [task abort];
-}
-
} // namespace WebCore
#endif // HAVE(SPEECHRECOGNIZER)
Modified: trunk/Source/WebKit/ChangeLog (272336 => 272337)
--- trunk/Source/WebKit/ChangeLog 2021-02-03 20:36:28 UTC (rev 272336)
+++ trunk/Source/WebKit/ChangeLog 2021-02-03 20:40:15 UTC (rev 272337)
@@ -1,3 +1,23 @@
+2021-02-03 Sihui Liu <[email protected]>
+
+ Create a SpeechRecognizer for each SpeechRecognitionRequest
+ https://bugs.webkit.org/show_bug.cgi?id=219699
+ <rdar://problem/72392097>
+
+ Reviewed by Youenn Fablet.
+
+ We used the same SpeechRecognizer for multiple requests, so we had to reset SpeechRecognizer state between
+ requests, which makes things complicated and hard to debug (like nested reset calls). Now we have one dedicated
+ SpeechRecognizer for each request.
+
+ * UIProcess/SpeechRecognitionServer.cpp:
+ (WebKit::SpeechRecognitionServer::requestPermissionForRequest):
+ (WebKit::SpeechRecognitionServer::handleRequest):
+ (WebKit::SpeechRecognitionServer::stop):
+ (WebKit::SpeechRecognitionServer::abort):
+ (WebKit::SpeechRecognitionServer::invalidate):
+ * UIProcess/SpeechRecognitionServer.h:
+
2021-02-03 Tim Horton <[email protected]>
REGRESSION (r270392): Launch time regression under IOSurface::maximumSize()
Modified: trunk/Source/WebKit/UIProcess/SpeechRecognitionServer.cpp (272336 => 272337)
--- trunk/Source/WebKit/UIProcess/SpeechRecognitionServer.cpp 2021-02-03 20:36:28 UTC (rev 272336)
+++ trunk/Source/WebKit/UIProcess/SpeechRecognitionServer.cpp 2021-02-03 20:40:15 UTC (rev 272337)
@@ -71,64 +71,46 @@
return;
auto identifier = weakRequest->clientIdentifier();
+ auto request = m_requests.take(identifier);
if (error) {
- m_requests.remove(identifier);
sendUpdate(identifier, WebCore::SpeechRecognitionUpdateType::Error, WTFMove(error));
return;
}
- handleRequest(*weakRequest);
+ ASSERT(request);
+ handleRequest(makeUniqueRefFromNonNullUniquePtr(WTFMove(request)));
});
}
-void SpeechRecognitionServer::handleRequest(WebCore::SpeechRecognitionRequest& request)
+void SpeechRecognitionServer::handleRequest(UniqueRef<WebCore::SpeechRecognitionRequest>&& request)
{
- if (!m_recognizer) {
- m_recognizer = makeUnique<WebCore::SpeechRecognizer>([this, weakThis = makeWeakPtr(this)](auto& update) {
- if (!weakThis)
- return;
+ if (m_recognizer)
+ m_recognizer->abort(WebCore::SpeechRecognitionError { WebCore::SpeechRecognitionErrorType::Aborted, "Another request is started"_s });
- auto clientIdentifier = update.clientIdentifier();
- if (!m_requests.contains(clientIdentifier))
- return;
+ auto clientIdentifier = request->clientIdentifier();
+ m_recognizer = makeUnique<WebCore::SpeechRecognizer>([this, weakThis = makeWeakPtr(this)](auto& update) {
+ if (!weakThis)
+ return;
- sendUpdate(update);
+ sendUpdate(update);
- auto type = update.type();
- if (type != WebCore::SpeechRecognitionUpdateType::Error && type != WebCore::SpeechRecognitionUpdateType::End)
- return;
+ if (update.type() == WebCore::SpeechRecognitionUpdateType::Error)
+ m_recognizer->abort();
+ else if (update.type() == WebCore::SpeechRecognitionUpdateType::End)
+ m_recognizer->setInactive();
+ }, WTFMove(request));
- if (m_isResetting)
- return;
- m_isResetting = true;
-
- m_recognizer->reset();
- m_requests.remove(clientIdentifier);
- m_isResetting = false;
- });
- }
-
- if (auto currentClientIdentifier = m_recognizer->currentClientIdentifier()) {
- auto error = WebCore::SpeechRecognitionError { WebCore::SpeechRecognitionErrorType::Aborted, "Another request is started"_s };
- sendUpdate(*currentClientIdentifier, WebCore::SpeechRecognitionUpdateType::Error, error);
- m_recognizer->reset();
- }
-
- auto clientIdentifier = request.clientIdentifier();
#if ENABLE(MEDIA_STREAM)
auto sourceOrError = m_realtimeMediaSourceCreateFunction();
if (!sourceOrError) {
- m_requests.remove(clientIdentifier);
sendUpdate(WebCore::SpeechRecognitionUpdate::createError(clientIdentifier, WebCore::SpeechRecognitionError { WebCore::SpeechRecognitionErrorType::AudioCapture, sourceOrError.errorMessage }));
return;
}
WebProcessProxy::muteCaptureInPagesExcept(m_identifier);
-
bool mockDeviceCapturesEnabled = m_checkIfMockSpeechRecognitionEnabled();
- m_recognizer->start(clientIdentifier, sourceOrError.source(), mockDeviceCapturesEnabled, request.lang(), request.continuous(), request.interimResults(), request.maxAlternatives());
+ m_recognizer->start(sourceOrError.source(), mockDeviceCapturesEnabled);
#else
- m_requests.remove(clientIdentifier);
sendUpdate(clientIdentifier, WebCore::SpeechRecognitionUpdateType::Error, WebCore::SpeechRecognitionError { WebCore::SpeechRecognitionErrorType::AudioCapture, "Audio capture is not implemented"_s });
#endif
}
@@ -136,34 +118,33 @@
void SpeechRecognitionServer::stop(WebCore::SpeechRecognitionConnectionClientIdentifier clientIdentifier)
{
MESSAGE_CHECK(clientIdentifier);
- if (m_recognizer && m_recognizer->currentClientIdentifier() == clientIdentifier) {
- m_recognizer->stop();
+
+ if (m_requests.remove(clientIdentifier)) {
+ sendUpdate(clientIdentifier, WebCore::SpeechRecognitionUpdateType::End);
return;
}
- if (m_requests.remove(clientIdentifier))
- sendUpdate(clientIdentifier, WebCore::SpeechRecognitionUpdateType::End);
+ if (m_recognizer && m_recognizer->clientIdentifier() == clientIdentifier)
+ m_recognizer->stop();
}
void SpeechRecognitionServer::abort(WebCore::SpeechRecognitionConnectionClientIdentifier clientIdentifier)
{
MESSAGE_CHECK(clientIdentifier);
- if (m_recognizer && m_recognizer->currentClientIdentifier() == clientIdentifier) {
- m_recognizer->abort();
+ if (m_requests.remove(clientIdentifier)) {
+ sendUpdate(clientIdentifier, WebCore::SpeechRecognitionUpdateType::End);
return;
}
- if (m_requests.remove(clientIdentifier))
- sendUpdate(clientIdentifier, WebCore::SpeechRecognitionUpdateType::End);
+ if (m_recognizer && m_recognizer->clientIdentifier() == clientIdentifier)
+ m_recognizer->abort();
}
void SpeechRecognitionServer::invalidate(WebCore::SpeechRecognitionConnectionClientIdentifier clientIdentifier)
{
MESSAGE_CHECK(clientIdentifier);
- if (m_requests.remove(clientIdentifier)) {
- if (m_recognizer && m_recognizer->currentClientIdentifier() == clientIdentifier)
- m_recognizer->abort();
- }
+ if (m_recognizer && m_recognizer->clientIdentifier() == clientIdentifier)
+ m_recognizer->abort();
}
void SpeechRecognitionServer::sendUpdate(WebCore::SpeechRecognitionConnectionClientIdentifier clientIdentifier, WebCore::SpeechRecognitionUpdateType type, Optional<WebCore::SpeechRecognitionError> error, Optional<Vector<WebCore::SpeechRecognitionResultData>> result)
@@ -178,6 +159,7 @@
void SpeechRecognitionServer::sendUpdate(const WebCore::SpeechRecognitionUpdate& update)
{
+ WTFLogAlways("[%p]SpeechRecognitionServer::sendUpdate update.type[%d], update.clientIdentifier[%llu]", this, update.type(), update.clientIdentifier().toUInt64());
send(Messages::WebSpeechRecognitionConnection::DidReceiveUpdate(update));
}
Modified: trunk/Source/WebKit/UIProcess/SpeechRecognitionServer.h (272336 => 272337)
--- trunk/Source/WebKit/UIProcess/SpeechRecognitionServer.h 2021-02-03 20:36:28 UTC (rev 272336)
+++ trunk/Source/WebKit/UIProcess/SpeechRecognitionServer.h 2021-02-03 20:40:15 UTC (rev 272337)
@@ -67,7 +67,7 @@
private:
void requestPermissionForRequest(WebCore::SpeechRecognitionRequest&);
- void handleRequest(WebCore::SpeechRecognitionRequest&);
+ void handleRequest(UniqueRef<WebCore::SpeechRecognitionRequest>&&);
void sendUpdate(WebCore::SpeechRecognitionConnectionClientIdentifier, WebCore::SpeechRecognitionUpdateType, Optional<WebCore::SpeechRecognitionError> = WTF::nullopt, Optional<Vector<WebCore::SpeechRecognitionResultData>> = WTF::nullopt);
void sendUpdate(const WebCore::SpeechRecognitionUpdate&);
