REGRESSION (iOS 15): HTMLAudioElement fails to load new audio when device is locked or safari is in background

[Rick Li]

Hi, we are currently attempting to get Safari to auto move on to the next audio track when a track ends. By executing JavaScript that replaces the src attribute with a new mp3, it worked fine with IOS 14.
However in IOS 15 Beta - 15.0 (19A5325f) if the screen is locked or safari is in the background, playback terminates at the end of the first track and does not progress to the second.  No error is raised.

To test, please use https://521dimensions.com/open-source/amplitudejs

[Radar WebKit Bug Importer]

<rdar://problem/82554226>

[Eric Carlson]

Created attachment 449683 [details]
Patch

[Eric Carlson]

Created attachment 449698 [details]
Patch

[Jer Noble]

Comment on attachment 449698 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=449698&action=review

> Source/WebCore/platform/audio/cocoa/MediaSessionManagerCocoa.mm:146
> +    if (!hasAudibleAudioOrVideoMediaType && m_previousHadAudibleAudioOrVideoMediaType) {
> +        if (!m_delayCategoryChangeTimer.isActive()) {
> +            m_delayCategoryChangeTimer.startOneShot(delayBeforeSettingCategoryNone);
> +            ALWAYS_LOG(LOGIDENTIFIER, "setting timer");
> +        }
> +        hasAudibleAudioOrVideoMediaType = true;
> +    }
> +    m_previousHadAudibleAudioOrVideoMediaType = hasAudibleAudioOrVideoMediaType;
> +

I wonder what will happen when a page is using WebAudio; we may go from Playback -> Ambient without hitting this timer.

> Source/WebKit/WebProcess/GPU/media/RemoteAudioSession.cpp:-148
> -#if PLATFORM(COCOA)
>      return m_category;
> -#else
> -    return AudioSession::CategoryType::None;
> -#endif

Is this necessary?

> LayoutTests/ChangeLog:12
> +        * media/audio-session-category-expected.txt: Added.
> +        * media/audio-session-category.html: Added.

Should these go into platform/ios/media or platform/mac/media?

[Jer Noble]

Comment on attachment 449698 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=449698&action=review

> Source/WebCore/platform/audio/cocoa/MediaSessionManagerCocoa.mm:133
> +                m_delayCategoryChangeTimer.stop();

I think this may be in the wrong place. It'll only get hit if there are no audible <video> or <audio> elements, but the session is "potentially audible".

> Source/WebCore/platform/audio/cocoa/MediaSessionManagerCocoa.mm:145
> +    if (!hasAudibleAudioOrVideoMediaType && m_previousHadAudibleAudioOrVideoMediaType) {
> +        if (!m_delayCategoryChangeTimer.isActive()) {
> +            m_delayCategoryChangeTimer.startOneShot(delayBeforeSettingCategoryNone);
> +            ALWAYS_LOG(LOGIDENTIFIER, "setting timer");
> +        }
> +        hasAudibleAudioOrVideoMediaType = true;
> +    }
> +    m_previousHadAudibleAudioOrVideoMediaType = hasAudibleAudioOrVideoMediaType;

I think it should move down here, in an "else" statement after the "if" (which may need to be restructured).  Maybe something like:

if (hasAudibleAudioOrVideoMediaType)
    m_delayCategoryChangeTimer.stop();
else if (m_previousHadAudibleAudioOrVideoMediaType) {
    ....
}

[Eric Carlson]

Created attachment 449735 [details]
Patch

[Eric Carlson]

Created attachment 449821 [details]
Patch for landing

[Eric Carlson]

Comment on attachment 449821 [details]
Patch for landing

Test failures are unrelated.

[EWS]

Committed r288466 (246352@main): <https://commits.webkit.org/246352@main>

All reviewed patches have been landed. Closing bug and clearing flags on attachment 449821 [details].

[Darin Adler]

Comment on attachment 449821 [details]
Patch for landing

View in context: https://bugs.webkit.org/attachment.cgi?id=449821&action=review

> Source/WebCore/platform/audio/cocoa/MediaSessionManagerCocoa.mm:97
> +    static const Seconds delayBeforeSettingCategoryNone = { 2_s };

constexpr auto delayBeforeSettingCategoryNone = 2_s;

[Eric Carlson]

Reopening to attach new patch.

[Eric Carlson]

Created attachment 449867 [details]
Address post-review feedback

[EWS]

Committed r288493 (246365@main): <https://commits.webkit.org/246365@main>

All reviewed patches have been landed. Closing bug and clearing flags on attachment 449867 [details].

[Eric Carlson]

Reopening to attach new patch.

[Eric Carlson]

Created attachment 458443 [details]
Patch

[Eric Carlson]

Comment on attachment 458443 [details]
Patch

Nothing to see here, I attached the patch to the wrong bug.