Bug 216696

Created attachment 409130 [details] Patch

Comment on attachment 409130 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=409130&action=review > Source/WebCore/ChangeLog:8 > + Implement setSinkId and sinkId as per https://w3c.github.io/mediacapture-output/#htmlmediaelement-extensions, s/,/./ > LayoutTests/imported/w3c/ChangeLog:15 > + * web-platform-tests/audio-output/META.yml: Added. > + * web-platform-tests/audio-output/idlharness.https.window-expected.txt: Added. > + * web-platform-tests/audio-output/idlharness.https.window.html: Added. > + * web-platform-tests/audio-output/idlharness.https.window.js: Added. > + * web-platform-tests/audio-output/setSinkId.https-expected.txt: Added. > + * web-platform-tests/audio-output/setSinkId.https.html: Added. > + * web-platform-tests/audio-output/w3c-import.log: Added. It looks like you forgot to add these.

Created attachment 409146 [details] Patch

> It looks like you forgot to add these. The danger of git reset HEAD~1... Here they are!

Created attachment 409234 [details] Patch

Created attachment 409237 [details] Patch

Comment on attachment 409237 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=409237&action=review > Source/WebCore/html/HTMLMediaElement.cpp:2863 > + if (!document().processingUserGestureForMedia() && document().topDocument().settings().speakerSelectionRequiresUserGesture()) { What is the goal of accessing settings from topDocument here? > Source/WebCore/html/HTMLMediaElement.idl:92 > + [Conditional=MEDIA_STREAM, EnabledBySetting=ExposeSpeakers&PerElementSpeakerSelection] readonly attribute DOMString audioOutputDevice; > + [Conditional=MEDIA_STREAM, EnabledBySetting=ExposeSpeakers&PerElementSpeakerSelection, ImplementedAs=audioOutputHashedDeviceId] readonly attribute DOMString sinkId; > + [Conditional=MEDIA_STREAM, EnabledBySetting=ExposeSpeakers&PerElementSpeakerSelection] Promise<undefined> setAudioOutputDevice(DOMString deviceId); > + [Conditional=MEDIA_STREAM, EnabledBySetting=ExposeSpeakers&PerElementSpeakerSelection, ImplementedAs=setAudioOutputDevice] Promise<undefined> setSinkId(DOMString deviceId); From the spec you linked in the ChangeLog, I only saw sinkId and setSinkId. Where do audioOutputDevice and setAudioOutputDevice come from? Regardless, these can and should now be defined from their own IDL file using a partial interface (partial interfaces no longer require a separate implementation class with static functions, so the implementation can stay in HTMLMediaElement.h/cpp if that is the right place for it to live. > Source/WebCore/platform/graphics/avfoundation/objc/MediaPlayerPrivateAVFoundationObjC.mm:956 > +#if PLATFORM(MAC) Rather than using PLATFORM(MAC) here, I have been trying to encourage people to use more fine grained defines. Perhaps ENABLE(MEDIA_STREAM_OUTPUT_DEVICE_ID) or something like that? > Source/WebKit/Shared/WebPreferencesExperimental.yaml:146 > + category: experimental it is no longer needed (or desired) to set the category as it it implied by the file. > Source/WebKit/Shared/WebPreferencesExperimental.yaml:147 > + condition: ENABLE(WEB_RTC) Is this the right condition? In the rest of the patch, you have mostly been using ENABLE(MEDIA_STREAM).

Created attachment 409263 [details] Patch

(In reply to Sam Weinig from comment #7) > Comment on attachment 409237 [details] > Patch > > View in context: > https://bugs.webkit.org/attachment.cgi?id=409237&action=review > > > Source/WebCore/html/HTMLMediaElement.cpp:2863 > > + if (!document().processingUserGestureForMedia() && document().topDocument().settings().speakerSelectionRequiresUserGesture()) { > > What is the goal of accessing settings from topDocument here? This is mirroring what is done for audio playing restriction. I changed it to use the current document for now. > > Source/WebCore/html/HTMLMediaElement.idl:92 > > + [Conditional=MEDIA_STREAM, EnabledBySetting=ExposeSpeakers&PerElementSpeakerSelection] readonly attribute DOMString audioOutputDevice; > > + [Conditional=MEDIA_STREAM, EnabledBySetting=ExposeSpeakers&PerElementSpeakerSelection, ImplementedAs=audioOutputHashedDeviceId] readonly attribute DOMString sinkId; > > + [Conditional=MEDIA_STREAM, EnabledBySetting=ExposeSpeakers&PerElementSpeakerSelection] Promise<undefined> setAudioOutputDevice(DOMString deviceId); > > + [Conditional=MEDIA_STREAM, EnabledBySetting=ExposeSpeakers&PerElementSpeakerSelection, ImplementedAs=setAudioOutputDevice] Promise<undefined> setSinkId(DOMString deviceId); > > From the spec you linked in the ChangeLog, I only saw sinkId and setSinkId. > Where do audioOutputDevice and setAudioOutputDevice come from? Regardless, > these can and should now be defined from their own IDL file using a partial > interface (partial interfaces no longer require a separate implementation > class with static functions, so the implementation can stay in > HTMLMediaElement.h/cpp if that is the right place for it to live. WG agreed to introduce navigator.mediaDevices.setAudioOutputDevice to set the audio output for a whole page. I filed an issue to rename SetSinkId to setAudioOutputDevice for consistency. I removed it for now and will introduce it later. > > Source/WebCore/platform/graphics/avfoundation/objc/MediaPlayerPrivateAVFoundationObjC.mm:956 > > +#if PLATFORM(MAC) > > Rather than using PLATFORM(MAC) here, I have been trying to encourage people > to use more fine grained defines. Perhaps > ENABLE(MEDIA_STREAM_OUTPUT_DEVICE_ID) or something like that? OK, let's use HAVE() instead.

Created attachment 409274 [details] Patch

Comment on attachment 409274 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=409274&action=review > Source/WTF/ChangeLog:8 > + > + * wtf/PlatformHave.h: Please fill in the changelog. > Source/WebKit/ChangeLog:14 > + Add internal flag to enable/disable user gesture requirement for setting audio output. > + > + * Shared/WebPreferencesExperimental.yaml: > + * UIProcess/API/C/WKPreferences.cpp: > + (WKPreferencesSetSpeakerSelectionRequiresUserGesture): > + (WKPreferencesGetSpeakerSelectionRequiresUserGesture): > + * UIProcess/API/C/WKPreferencesRefPrivate.h: This ChangeLog doesn't look accurate. You changed both WebPreferencesInternal.yaml and WebPreferencesExperimental.yaml. > Source/WebCore/html/HTMLMediaElement.idl:90 > + [Conditional=MEDIA_STREAM, EnabledBySetting=ExposeSpeakers&PerElementSpeakerSelection, ImplementedAs=audioOutputHashedDeviceId] readonly attribute DOMString sinkId; > + [Conditional=MEDIA_STREAM, EnabledBySetting=ExposeSpeakers&PerElementSpeakerSelection, ImplementedAs=setAudioOutputDevice] Promise<undefined> setSinkId(DOMString deviceId); Please move these out into a partial interface in their own IDL file. Also, why are they implemented under another name? Why not just call them sinkId/setSinkId in the implementation? > Source/WebKit/UIProcess/API/C/WKPreferencesRefPrivate.h:172 > +// Defaults to false. This currently seems to default to true. > Source/WebKit/UIProcess/API/C/WKPreferencesRefPrivate.h:174 > +WK_EXPORT void WKPreferencesSetSpeakerSelectionRequiresUserGesture(WKPreferencesRef preferencesRef, bool flag); > +WK_EXPORT bool WKPreferencesGetSpeakerSelectionRequiresUserGesture(WKPreferencesRef preferencesRef); Are these needed beyond use by WebKitTestRunner? If it's only going to be used by WebKitTestRunner, please use WKPreferencesSetInternalDebugFeatureForKey(preferences, false, WKStringCreateWithUTF8CString("SpeakerSelectionRequiresUserGesture"); instead of adding this new SPI (you can find others doing this around line 845 of TestController.cpp in WebKitTestRunner. > LayoutTests/ChangeLog:8 > + Skip new tests on WK1. Out of interest, what is WebKit2 specific about these?

Comment on attachment 409274 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=409274&action=review > Source/WebCore/ChangeLog:9 > + Introduce a setting to expose these methods and to enable/disable user gesture requirement. How can we test the user gesture requirement? It seems like that's something we wouldn't want to break.

> > Source/WebCore/html/HTMLMediaElement.idl:90 > > + [Conditional=MEDIA_STREAM, EnabledBySetting=ExposeSpeakers&PerElementSpeakerSelection, ImplementedAs=audioOutputHashedDeviceId] readonly attribute DOMString sinkId; > > + [Conditional=MEDIA_STREAM, EnabledBySetting=ExposeSpeakers&PerElementSpeakerSelection, ImplementedAs=setAudioOutputDevice] Promise<undefined> setSinkId(DOMString deviceId); > > Please move these out into a partial interface in their own IDL file. Also, > why are they implemented under another name? Why not just call them > sinkId/setSinkId in the implementation? sinkId/setSinkId is not very meaningful and we want to introduce navigator.mediaDevices.setAudioOutput which is the same but for a whole page. The idea would be to deprecate setSinkId in favour of setAudioOutput. > > LayoutTests/ChangeLog:8 > > + Skip new tests on WK1. > > Out of interest, what is WebKit2 specific about these? navigator.mediaDevices

Created attachment 409293 [details] Patch

Comment on attachment 409293 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=409293&action=review > Source/WebCore/platform/mediastream/mac/AudioMediaStreamTrackRendererUnit.cpp:42 > +#if PLATFORM(COCOA) > +#include "CoreAudioCaptureDevice.h" > +#include "CoreAudioCaptureDeviceManager.h" > +#endif Since this is a Mac only file, I don't think the platform check is needed here. > Source/WebCore/platform/mediastream/mac/AudioMediaStreamTrackRendererUnit.cpp:61 > +#if PLATFORM(MAC) Given the path of the file, I assume it is Mac only. Is this check really needed? > Source/WebCore/platform/mediastream/mac/AudioMediaStreamTrackRendererUnit.cpp:196 > +#if PLATFORM(MAC) Same here. > Source/WebCore/platform/mediastream/mac/AudioMediaStreamTrackRendererUnit.h:73 > +#if PLATFORM(MAC) And here. > Source/WebCore/testing/Internals.cpp:570 > + page.settings().setSpeakerSelectionRequiresUserGesture(false); It seems like this is duplicated here and in TestController.cpp. > Source/WebKit/UIProcess/API/C/WKPreferencesRefPrivate.h:174 > +// Defaults to true. > +WK_EXPORT void WKPreferencesSetSpeakerSelectionRequiresUserGesture(WKPreferencesRef preferencesRef, bool flag); > +WK_EXPORT bool WKPreferencesGetSpeakerSelectionRequiresUserGesture(WKPreferencesRef preferencesRef); Why is new SPI needed here?

(In reply to Sam Weinig from comment #15) > Comment on attachment 409293 [details] > Patch > > View in context: > https://bugs.webkit.org/attachment.cgi?id=409293&action=review > > > Source/WebCore/platform/mediastream/mac/AudioMediaStreamTrackRendererUnit.cpp:42 > > +#if PLATFORM(COCOA) > > +#include "CoreAudioCaptureDevice.h" > > +#include "CoreAudioCaptureDeviceManager.h" > > +#endif > > Since this is a Mac only file, I don't think the platform check is needed > here. Right. > > Source/WebCore/platform/mediastream/mac/AudioMediaStreamTrackRendererUnit.cpp:61 > > +#if PLATFORM(MAC) > > Given the path of the file, I assume it is Mac only. Is this check really > needed? It is needed. Source/WebCore/platform/mediastream/mac should really be renamed Source/WebCore/platform/mediastream/cocoa. > > Source/WebCore/testing/Internals.cpp:570 > > + page.settings().setSpeakerSelectionRequiresUserGesture(false); > > It seems like this is duplicated here and in TestController.cpp. If Internals sets it to true, it might stick for longer than expected. > > Source/WebKit/UIProcess/API/C/WKPreferencesRefPrivate.h:174 > > +// Defaults to true. > > +WK_EXPORT void WKPreferencesSetSpeakerSelectionRequiresUserGesture(WKPreferencesRef preferencesRef, bool flag); > > +WK_EXPORT bool WKPreferencesGetSpeakerSelectionRequiresUserGesture(WKPreferencesRef preferencesRef); > > Why is new SPI needed here? Oh right, it should be removed now that TestController does not use it anymore.

(In reply to youenn fablet from comment #16) > > > Source/WebCore/platform/mediastream/mac/AudioMediaStreamTrackRendererUnit.cpp:61 > > > +#if PLATFORM(MAC) > > > > Given the path of the file, I assume it is Mac only. Is this check really > > needed? > > It is needed. > Source/WebCore/platform/mediastream/mac should really be renamed > Source/WebCore/platform/mediastream/cocoa. I see. Very confusing. > > > > Source/WebCore/testing/Internals.cpp:570 > > > + page.settings().setSpeakerSelectionRequiresUserGesture(false); > > > > It seems like this is duplicated here and in TestController.cpp. > > If Internals sets it to true, it might stick for longer than expected. Really? TestController calls this in a function called resetPreferencesToConsistentValues() which seems to be called after every invocation.

Comment on attachment 409293 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=409293&action=review > Source/WebCore/html/HTMLMediaElement.h:615 > + String audioOutputDeviceId() const final; > + String audioOutputDeviceIdOverride() const final { return m_audioOutputPersistentDeviceId; } Why are these `final`? Why is one implemented in here and one in the .cpp when they both just return a member variable?

(In reply to Eric Carlson from comment #18) > Comment on attachment 409293 [details] > Patch > > View in context: > https://bugs.webkit.org/attachment.cgi?id=409293&action=review > > > Source/WebCore/html/HTMLMediaElement.h:615 > > + String audioOutputDeviceId() const final; > > + String audioOutputDeviceIdOverride() const final { return m_audioOutputPersistentDeviceId; } > > Why are these `final`? These are MediaPlayerClient API. > Why is one implemented in here and one in the .cpp when they both just > return a member variable? audioOutputDeviceId() will later on either return the override or the page audioOutputDeviceId(), hence why it was in cpp in the main patch. I moved all of them to header file for now.

> > If Internals sets it to true, it might stick for longer than expected. > > Really? TestController calls this in a function called > resetPreferencesToConsistentValues() which seems to be called after every > invocation. Did some test and you are right it does not seem necessary so I'll remove it.

Created attachment 409457 [details] Patch for landing

Created attachment 409459 [details] Patch for landing

Committed r267472: <https://trac.webkit.org/changeset/267472> All reviewed patches have been landed. Closing bug and clearing flags on attachment 409459 [details].

<rdar://problem/69433680>

Sorry, only the validHandler needs to be immutable for us. If I make it mutable here's the error: ../../Source/WebKit/UIProcess/glib/UserMediaPermissionRequestManagerProxyGLib.cpp:70:17: error: no matching function for call to object of type 'const (lambda at ../../Source/WebKit/UIProcess/glib/UserMediaPermissionRequestManagerProxyGLib.cpp:54:29)' validHandler(std::move<WTF::CheckMoveParameter>(audioDevices), std::move<WTF::CheckMoveParameter>(videoDevices), std::move<WTF::CheckMoveParameter>(*deviceIdentifierHashSalt)); ^~~~~~~~~~~~ ../../Source/WebKit/UIProcess/glib/UserMediaPermissionRequestManagerProxyGLib.cpp:54:29: note: candidate function not viable: 'this' argument has type 'const (lambda at ../../Source/WebKit/UIProcess/glib/UserMediaPermissionRequestManagerProxyGLib.cpp:54:29)', but method is not marked const auto validHandler = [this, request](Vector<CaptureDevice>&& audioDevices, Vector<CaptureDevice>&& videoDevices, String&& deviceIdentifierHashSalt) mutable { ^ 1 error generated. Compiler killed by signal 1

oops wrong bug, sorry

On Safari Technology Release 115 it says: Added support for HTMLMediaElement.setSinkId Changeset 267472 in webkit shows all the changes made but I'm using Release 127 (Safari 15.0, WebKit 16612.1.18.11.3) and don't have the setSinkId functionality at all. Moreover, when using WKWebView I also don't have any "audiooutput" devices when enumerating devices. The goal is to use getUserMedia and toggle the outputs from earpiece to speaker on iOS. Any help will be much appreciated

(In reply to haimomesi from comment #27) > On Safari Technology Release 115 it says: > > Added support for HTMLMediaElement.setSinkId > > Changeset 267472 in webkit shows all the changes made but I'm using Release > 127 (Safari 15.0, WebKit 16612.1.18.11.3) and don't have the setSinkId > functionality at all. > > Moreover, when using WKWebView I also don't have any "audiooutput" devices > when enumerating devices. > > The goal is to use getUserMedia and toggle the outputs from earpiece to > speaker on iOS. > > Any help will be much appreciated Exposing audio output devices and setSinkId are still experimental features and both are still disabled by default. You should be able to enable these features for testing with Safari's Develop menu: Develop > Experimental Features > Allow per media element speaker device selection Develop > Experimental Features > Allow speaker device selection

Can we enabled this with WKWebView in some manner? Or maybe run a js script that will allow this behavior? Thank you for the quick reply

(In reply to haimomesi from comment #29) > Can we enabled this with WKWebView in some manner? > Or maybe run a js script that will allow this behavior? > No, WKWebView doesn't have API to enable experimental features.