Unable to reproduce the issue on Win-7 and Win-10 using chrome reported version #60.0.3112.78 and latest canary #62.0.3167.0.

Following are the steps followed to reproduce the issue.
-----------
1. Navigated to URL: http://game.granbluefantasy.jp and played Granblue.
2. Enabled Sound.
3. Played and simultaneously performed other activities on desktop like typing in notepad, editing excel e.t.c
4. Did not observe any popping noises and stutters in it as found in attached gbf.mp3.

nihaomeeow@ - Could you please check the issue on latest canary #62.0.3167.0 by creating a new profile without any apps and extensions and please let us know if the issue still persist or not.

Thanks...!!

The problem persists in canary #62.0.3167.0.
To verify it's a chrome problem I tested chrome 57.0.2987.133 and 59.0.3071.86. No issues with these. So it must be something with chrome 60+

Thank you for providing more feedback. Adding requester "krajshree@chromium.org" to the cc list and removing "Needs-Feedback" label.

For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot

Tested this issue on Windows 10 with chrome #62.0.3171.0

Steps followed : 
1. Navigated to Test URL: http://game.granbluefantasy.jp/#tutorial/5 
2. Able to hear the Sound.

Observation:
-------------
Unable to hear the audio from test url when other application window is active

Attaching the screen-cast for reference.

nihaomeeow@ Could you please look into it and let us know your observations.

This was broken by Chrome 60. I reproduce this in 60.0.3112.90 (Official Build) (64-bit) (cohort: 60_90_win), it worked in 59. 60 also broke other things related to Granblue fantasy, see  issue #752791 . Users on reddit are also reporting visual lag: https://www.reddit.com/r/Granblue_en/comments/6rxqvk/suddently_lag_since_last_game_update/#dl8krsa

The most noticeable place for lag and sound issues is in combat. The sound issues appear to correlate with time periods where the game is rendering more complex graphics & playing more sounds at once - i.e. complex spell effects or boss special moves. This probably produces visible lag on computers slower than mine.

I'm playing with hardware acceleration disabled (because of other Chrome bugs that affect this game) so it's not video driver/video card related.

I've attached a video recording of the game where you can hear multiple audio glitches as described above. The background music & character voices produce popping, stuttering and slowdown.

Deleted: lag-and-audio-issues.mp4 9.0 MB

	lag-and-audio-issues.mp4 9.0 MB View Download

Here is a chrome://tracing recording that captures at least one instance of audio glitches, on one of the game's landing pages (no combat happening).

The glitching is (I'm pretty sure) in a suspicious time period of the trace where all the audio threads suddenly stop running as often: wasapi_render_thread, WebAudio Rendering Thread, AudioOutputDevice.

From looking at the trace, the audio glitching begins almost immediately after a 22ms-long GC inside CrRendererMain.

Deleted: trace_audio-glitch.json.gz 1.8 MB

trace_audio-glitch.json.gz 1.8 MB Download

Here's a devtools profile. The audio glitch starts around where the GC occurs and ends shortly before the end of the profile.

Deleted: Profile-20170806T233855.zip 1.6 MB

Profile-20170806T233855.zip 1.6 MB Download

> wasapi_render_thread, WebAudio Rendering Thread, AudioOutputDevice

Well, this does not sound like WebAudio's problem. WebAudio Rendering Thread is driven by AudioOutputDevice. So if it does not schedule the next render task, WebAudio Rendering Thread will not render anything. I looked at the tracing data
and the actual stuttering is starting from AudioDeviceThread.

Also it does not makes sense to me that the main thread GC can cause the audio thread to stop. olka@ might know why this can happen.

kg@ Could you suggest what "GC" stands for? Also could you please refer to time stamps on the graph and pids, to make communication more efficient? 
Could you record a glitch-free trace for comparison?

From callback oddity perspective, I see that in the first 400 ms and between 7500-8800 ms WASAPIAudioOutputStream periodically decides to not call upper layers to render audio. It happens here [https://cs.chromium.org/chromium/src/media/audio/win/audio_low_latency_output_win.cc?type=cs&q=RenderAudioFromSource&l=476] and means platform audio buffer does not have enough space (= still has some audio to play). The code has been like this for years, so I'm not sure it's a source of the reported glitches, since it works in M59. (Adding Dale to comment)

So to me it looks like recent WebAudio changes are still the most probable root cause of the issue.

Could we add traces for WebAudio and make another recording of glitches?

I'd check the buffer size it's using; it's possible it's not a multiple of 10ms so we end up accumulating and thus skipping a read every now and then.

The main thread GC is just a coincidence, most likely. I merely wanted to call it out because I saw it in all the traces.

If it helps, the audio files this game plays are all 32khz audio encoded as mp3 (typically around 48kbps, it looks like). So it could be related to resampling?

It's very hard to get a trace without any glitches because any content playing sound tends to glitch periodically. 

Here's a sound that's looped in the background of a scene recently added to the game:
http://game-a5.granbluefantasy.jp/assets_en/sound/se/crowd_bgm_1.mp3
It loops fine without sounding glitchy in Firefox. If I just let the game sit and loop it in Chrome, it glitches periodically, though not as bad as some other recordings. When the scene begins playing character dialogues those glitch a couple times a minute with popping, slowdown, etc.

The glitches have obvious gaps when examined as a recording in Audacity or another editor. Recording & screenshot attached. This is just that BGM looping at a static dialogue scene, nothing is happening.

The samples within the glitches/gaps are not all zeroes, which is rather suspicious to me.

Deleted: gaps.png 8.7 KB

	gaps.png 8.7 KB View Download

Deleted: nonzero.png 27.9 KB

	nonzero.png 27.9 KB View Download

Deleted: glitch.wav 1.3 MB

glitch.wav 1.3 MB Download

I instrumented their game code and all the sound effects & music appear to be implemented using createBufferSource and decodeAudioData. The decode is producing 48khz buffers, so resampling wouldn't be involved (since the audio doesn't glitch on every loop).

If I open this isolated test case in a tab while granblue fantasy is running in another tab (idled on the home screen), and click Play, the audio glitches periodically while looping.

Deleted: test.html 64.4 KB

test.html 64.4 KB View Download

re #12:
It outputs at 480 byte frames, so per #14 it's 10 ms buffers. Also if it's due to rebuffering I would expect a stable pattern of skipped reads, but here they are in bursts.

kg@ could you please make one more trace with glitches (both audio and webaudio categories) and give some estimations around which time of the trace glitches occur?



hongchang@

re #13:

Thanks for following up on this with more observation.

> The samples within the glitches/gaps are not all zeroes, which is rather suspicious to me.

1. This is not the pure recording of WebAudio's output. Some numerical error can be added in some where in the chain. What kind of audio routing are you doing?

2. What worries me here is the frame size of glitches. They do not look like regular buffer sizes. (340~380 frames) If the WebAudio's new FIFO is the culprit, the glitch size must be multiple of 128.

Because this is a game with heavy user interaction, I think it might be related to this as well:
https://bugs.chromium.org/p/chromium/issues/detail?id=745898#c6

Also, examine stderr.  If the fifo is the culprit, there should be messages from the fifo about underruns.

[10348:1308:0810/173144.667:VERBOSE1:dhcp_proxy_script_adapter_fetcher_win.cc(262)] Error fetching PAC URL from DHCP: 2
[3092:10608:0810/173144.691:WARNING:pushpullfifo.cpp(154)] PushPullFIFO: underflow while pulling (underflowCount=4, availableFrames=96, requestedFrames=480, fifoLength=8192)
[10348:4308:0810/173144.722:VERBOSE1:dhcp_proxy_script_adapter_fetcher_win.cc(262)] Error fetching PAC URL from DHCP: 2
[3092:10608:0810/173144.851:WARNING:pushpullfifo.cpp(154)] PushPullFIFO: underflow while pulling (underflowCount=5, availableFrames=96, requestedFrames=480, fifoLength=8192)
[3092:10608:0810/173144.931:WARNING:pushpullfifo.cpp(154)] PushPullFIFO: underflow while pulling (underflowCount=6, availableFrames=96, requestedFrames=480, fifoLength=8192)
[3092:10608:0810/173145.090:WARNING:pushpullfifo.cpp(154)] PushPullFIFO: underflow while pulling (underflowCount=7, availableFrames=96, requestedFrames=480, fifoLength=8192)
[3092:10608:0810/173145.171:WARNING:pushpullfifo.cpp(154)] PushPullFIFO: underflow while pulling (underflowCount=8, availableFrames=96, requestedFrames=480, fifoLength=8192)
[3092:10608:0810/173145.250:WARNING:pushpullfifo.cpp(154)] PushPullFIFO: underflow while pulling (underflowCount=9, availableFrames=352, requestedFrames=480, fifoLength=8192)
[3092:10608:0810/173145.570:WARNING:pushpullfifo.cpp(154)] PushPullFIFO: underflow while pulling (underflowCount=10, availableFrames=96, requestedFrames=480, fifoLength=8192)
[3092:10608:0810/173145.651:WARNING:pushpullfifo.cpp(154)] PushPullFIFO: underflow while pulling (underflowCount=11, availableFrames=96, requestedFrames=480, fifoLength=8192)
[3092:10608:0810/173145.730:WARNING:pushpullfifo.cpp(154)] PushPullFIFO: underflow while pulling (underflowCount=12, availableFrames=96, requestedFrames=480, fifoLength=8192)
[3092:10608:0810/173145.810:WARNING:pushpullfifo.cpp(154)] PushPullFIFO: underflow while pulling (underflowCount=13, availableFrames=96, requestedFrames=480, fifoLength=8192)
[3092:10608:0810/173145.970:WARNING:pushpullfifo.cpp(154)] PushPullFIFO: underflow while pulling (underflowCount=14, availableFrames=352, requestedFrames=480, fifoLength=8192)
[3092:10608:0810/173146.051:WARNING:pushpullfifo.cpp(154)] PushPullFIFO: underflow while pulling (underflowCount=15, availableFrames=96, requestedFrames=480, fifoLength=8192)

Thanks.  So something is causing the audio thread to run out of time and underflow.

kg@

Thanks for the info. Can you tell me about the system that you're running the game? Hoe many CPU cores does it have? We have similar regression report from the system with the fewer CPU cores. (thus the audio rendering thread gets descheduled.)

kg@

Unfortunately none of my machines did not generate any glitch from the game. (linux, windows, osx)

re #21:
Windows 10 pro creators' update x64
Core i7-6700K 4ghz (4 cores + HT), 64gb RAM

re #23:

Thanks! If you're willing to try a tentative fix for the thread priority issue, you can try this patch: https://chromium-review.googlesource.com/c/614284

re #6:

Audio not playing while another window being active has always been a thing even prior to chrome v60. The game basically pauses playback for audio and animations somehow when it's not the active focus. There is an unofficial chrome extension which allows the audio to play when not in an active video so it's fixable, I'm not sure if it's intended or a bug however. I assume it's not related to this issue.

Thank you for providing more feedback. Adding requester "kkaluri@chromium.org" to the cc list and removing "Needs-Feedback" label.

For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot

re #24: Is there a way I can identify the audio thread(s) and manually boost them? Building Chrome from source would take me a while to set up, I don't have buildtools/etc on any of my machines right now.

re #6 and #26: Yeah the game intentionally mutes audio in the background. I use an extension to suppress that. Regardless, I hear glitches even while it's active, so that shouldn't prevent reproducing this (maybe it makes it harder?) My machine isn't under heavy load so I'm not sure why I see these glitches so reliably on a high-spec machine. The test case I provided earlier reproduces them too.

re #28: I've got some reports from other users that raising the priority greatly improves the glitch problem. However, I don't know how to manually raise the thread priority (not a process) on any platform.

Unable to reproduce the issue on Win-7 and Win-10 using chrome reported version #60.0.3112.78, latest stable #60.0.3112.101 and latest canary #62.0.3188.0.

Following are the steps followed to reproduce the issue.
-----------
1. Navigated to URL: http://game.granbluefantasy.jp and played Granblue.
2. Enabled Sound.
3. Played and simultaneously performed other activities on desktop like typing in notepad, editing excel e.t.c. Also opened another tab and opened other sites.
4. Did not observe any popping noises and stutters in it as found in attached gbf.mp3(comment #0), but the audio paused after switching to other tab or when performing any activity on desktop.

nihaomeeow@ - Could you please check the issue on latest stable #60.0.3112.101 by creating a new profile without any apps and extensions and please let us know if the issue still persist or not.
Removing Needs-Bisect label as of now as it is not reproducible. Please feel free to add the same if required.

Thanks...!!

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/f546d6dddbba639b593689b6e72d6c61fc1eb342

commit f546d6dddbba639b593689b6e72d6c61fc1eb342
Author: Hongchan Choi <hongchan@chromium.org>
Date: Mon Aug 21 17:32:30 2017

Raise the thread priority for WebAudio renderer

Since the introduction of stand-alone blink::WebThread for WebAudio 
rendering, several regression were reported. We speculate the root
cause of this is the lower thread priority of blink::WebThread
(normal priority) compared to the AudioDeviceThread
(realtime priority).

The analysis indicated that this regression gets worse when the
device has fewer CPU cores, resulting two threads (browser and main)
occupy the entire CPU load and making the WebThread descheduled from
the task scheduler.

This CL raises the thread priority for WebAudio renderer up to the
level of the browser main thread. (ThreadPriority::DISPLAY)

Blink Architecture discussion: https://groups.google.com/a/chromium.org/forum/?utm_medium=email&utm_source=footer#!msg/platform-architecture-dev/EnlQMTRwyrw/AeysuDuUAAAJ

Bug: 734539,  749344 
Change-Id: I5717e956228a1c4046e82e6f598d602ca5c82f35
Reviewed-on: https://chromium-review.googlesource.com/614284
Commit-Queue: Hongchan Choi <hongchan@chromium.org>
Reviewed-by: Kentaro Hara <haraken@chromium.org>
Reviewed-by: Kinuko Yasuda <kinuko@chromium.org>
Reviewed-by: Chris Palmer <palmer@chromium.org>
Reviewed-by: Raymond Toy <rtoy@chromium.org>
Cr-Commit-Position: refs/heads/master@{#495974}
[modify] https://crrev.com/f546d6dddbba639b593689b6e72d6c61fc1eb342/content/child/blink_platform_impl.cc
[modify] https://crrev.com/f546d6dddbba639b593689b6e72d6c61fc1eb342/content/child/blink_platform_impl.h
[modify] https://crrev.com/f546d6dddbba639b593689b6e72d6c61fc1eb342/third_party/WebKit/Source/platform/audio/AudioDestination.cpp
[modify] https://crrev.com/f546d6dddbba639b593689b6e72d6c61fc1eb342/third_party/WebKit/Source/platform/exported/Platform.cpp
[modify] https://crrev.com/f546d6dddbba639b593689b6e72d6c61fc1eb342/third_party/WebKit/public/platform/Platform.h

kg@

Could you try the app again with the latest Canary? (62.0.3196.0) The fix (reverting to the previous threading model) has landed there.

Problem still existed in  #60.0.3112.101 but latest Canary seems to have it fixed mostly for me as well. Good work.

Hi, this is still broken in 61.0.3163.79. Is the fix planned to be uplifted to release channel, or do I have to wait another 6+ weeks to be able to listen to sound in Chrome?

The fix (the workaround, actually) will be available in M62. It was incompatible with M61 branch, so we couldn't take the risk of breaking other parts. It's really unfortunate, but Canary has the fix and it is confirmed by other reporters as well.