WebKit Bugzilla
New
Browse
Log In
×
Sign in with GitHub
or
Remember my login
Create Account
·
Forgot Password
Forgotten password account recovery
RESOLVED MOVED
Bug 229413
iOS / iPadOS 15 video playback performance regression after seek
https://bugs.webkit.org/show_bug.cgi?id=229413
Summary
iOS / iPadOS 15 video playback performance regression after seek
Dustin Kerstein
Reported
2021-08-23 11:24:24 PDT
In the latest iPadOS 15 Beta 6, when performing a manual seek while playing a video in Safari, the video will often stutter for several seconds before playing back normally. This behavior is not seen on iPadOS 14.x 1. Try playing this video in Safari on iPadOS 15 -
https://files.panomoments.com/bbb_sunflower_2160p_60fps_normal.mp4
2. Let it play for ~60 seconds - note correct playback 3. Manually seek to ~20 mark - note playback will stutter for several seconds before it starts to play correctly 4. If unable to replicate on first try, seek to several different points in the timeline (within the cached buffer) Please let me know if I can help with any further debug / testing.
Attachments
Add attachment
proposed patch, testcase, etc.
Dustin Kerstein
Comment 1
2021-08-23 11:32:03 PDT
Upon further testing, it appears this issue may be more of an iOS / iPadOS issue as it can be replicated in the native video player, not just Safari. Is it possible to loop in the Apple iOS/iPadOS folks?
Radar WebKit Bug Importer
Comment 2
2021-08-24 01:30:51 PDT
<
rdar://problem/82281799
>
Sam Sneddon [:gsnedders]
Comment 3
2021-08-24 01:33:21 PDT
(and closing this because per your investigation this isn't actually a WebKit bug)
Dustin Kerstein
Comment 4
2021-10-23 05:28:36 PDT
FYI, here are the related feedback assistant bugs. FB9688897 FB9666426 FB9554184 Some of these issues may be more in VideoToolbox than in WebKit.
Dustin Kerstein
Comment 5
2021-10-23 06:07:16 PDT
And a post on the Apple Developer Forums -
https://developer.apple.com/forums/thread/693002
Jer Noble
Comment 6
2021-10-25 16:16:52 PDT
One of the requirements for media playback on iOS and macOS is that the server responds to requests with a valid Range: header with a 206 response and a well formed Content-Range: response header. And this URL will give out a 200 response for a Range request... sometimes. Here's the output of curling that file with a 2-byte range twice in a row: ...> curl -I -r 0-1
https://files.panomoments.com/bbb_sunflower_2160p_60fps_normal.mp4
; curl -I -r 0-1
https://files.panomoments.com/bbb_sunflower_2160p_60fps_normal.mp4
HTTP/2 200 date: Mon, 25 Oct 2021 23:10:23 GMT content-type: video/mp4 content-length: 673223862 x-amz-id-2: XgpLClM0fxhLvc8mY6myO+G9hj66UnxBXcOVtvYSfwxLMG1fPvd4AyYynbYODF4Xvabx5acSc/E= x-amz-request-id: 6EYV634TXT20SEX4 last-modified: Fri, 01 May 2020 15:20:57 GMT etag: "68360a312337530961dd70c330199fda-40" cache-control: max-age=31536000 cf-cache-status: MISS expect-ct: max-age=604800, report-uri="
https://report-uri.cloudflare.com/cdn-cgi/beacon/expect-ct
" report-to: {"endpoints":[{"url":"
https:\/\/a.nel.cloudflare.com\/report\/v3?s=AvVUvLxBmK%2BkrYQj%2FSiAVLNbqev6lJ%2F1ayWyl8hZOHoXmIze2KGldWyDZZ2rfOfibVM%2FRYIhZkyivQ2L8V%2Bi2rbKWtKihzV%2Bjd6unDBXwqinVT7kgOWoEpPP5gzjqNNxr1Tk7tfEeeI%3D
"}],"group":"cf-nel","max_age":604800} nel: {"success_fraction":0,"report_to":"cf-nel","max_age":604800} server: cloudflare cf-ray: 6a3f1e314a4f3610-LAX alt-svc: h3=":443"; ma=86400, h3-29=":443"; ma=86400, h3-28=":443"; ma=86400, h3-27=":443"; ma=86400 HTTP/2 206 date: Mon, 25 Oct 2021 23:10:23 GMT content-type: video/mp4 content-length: 2 x-amz-id-2: J66qJEh5ZFf0gYNYJToPmSPj8v+uLTO3Esb6MRT9NKvPnir/hsrGRieewKPTVlQ+Ll7Bihl/c5Y= x-amz-request-id: 6EYHMMJV2C0JWC0J last-modified: Fri, 01 May 2020 15:20:57 GMT etag: "68360a312337530961dd70c330199fda-40" content-range: bytes 0-1/673223862 cache-control: max-age=31536000 cf-cache-status: MISS expect-ct: max-age=604800, report-uri="
https://report-uri.cloudflare.com/cdn-cgi/beacon/expect-ct
" report-to: {"endpoints":[{"url":"
https:\/\/a.nel.cloudflare.com\/report\/v3?s=anXPlg4qkx%2FhEGnZs8fxWkwqLtETuYlkWwG%2FeSajCwwrYBE%2BP3RCPJAU%2F3zlKOXpAkCm7fFCarlTBsZhdljrmUo5qn4%2Fji6YMChkp8qzcB588sQ1pP6p4NORMLar88l%2FO1xqzY3NAAY%3D
"}],"group":"cf-nel","max_age":604800} nel: {"success_fraction":0,"report_to":"cf-nel","max_age":604800} server: cloudflare cf-ray: 6a3f1e350dae0575-LAX alt-svc: h3=":443"; ma=86400, h3-29=":443"; ma=86400, h3-28=":443"; ma=86400, h3-27=":443"; ma=86400 So the server is likely freaking out the media engine by responding to _some_ range requests with full files.
Dustin Kerstein
Comment 7
2021-10-25 17:17:48 PDT
(In reply to Jer Noble from
comment #6
)
> One of the requirements for media playback on iOS and macOS is that the > server responds to requests with a valid Range: header with a 206 response > and a well formed Content-Range: response header. And this URL will give > out a 200 response for a Range request... sometimes.
That indeed is an issue (related to Cloudflare) that I'm going to investigate, but I can replicate the same stuttering noted in the replication steps if I use the non-CDN URL: curl -I -r 0-1
https://s3.amazonaws.com/files.panomoments.com/bbb_sunflower_2160p_60fps_normal.mp4
This URL seems to provide 100% stable 206 responses. Are you able to replicate the stuttering using that URL?
Dustin Kerstein
Comment 8
2021-10-25 18:40:07 PDT
Actually, we may want to create a new bug ticket that is more focused on this new/worse behavior than this ticket was originally created for. Try going to this URL on Safari 15 (Desktop, iPadOS, iOS, etc.) -
https://s3.amazonaws.com/data.panomoments.com/processed/5849a51bc26d08000b62f9f8/58a5dbc8cd5f48000bdbbe1f/uhd_dashinit.mp4
curl -I -r 0-1
https://s3.amazonaws.com/data.panomoments.com/processed/5849a51bc26d08000b62f9f8/58a5dbc8cd5f48000bdbbe1f/uhd_dashinit.mp4
Returns 100% valid 206 responses, but something about the GOP (being dense in I-Frames / 100% I-Frames) causes Safari playback to fail (or take a very very long time to start). Are you able to replicate? If so, should we reopen or create a new ticket?
Dustin Kerstein
Comment 9
2021-10-25 18:45:48 PDT
The above behavior is potentially more accurately described by this older bug (which has gotten worse in recent Safari versions) -
https://bugs.webkit.org/show_bug.cgi?id=211295
Note
You need to
log in
before you can comment on or make changes to this bug.
Top of Page
Format For Printing
XML
Clone This Bug