Closed Bug 1068055 Opened 6 years ago Closed 3 years ago

[Gingerbread] VP9/WEBM video artifacts, only plays on half screen

Categories

(Core :: Audio/Video: Playback, defect)

33 Branch
All
Android
defect
Not set
normal

Tracking

()

RESOLVED INCOMPLETE
Tracking Status
firefox32 --- wontfix
firefox33 + wontfix
firefox34 + wontfix
firefox35 + wontfix
firefox36 + wontfix
firefox37 --- wontfix
relnote-firefox --- 33+
fennec + ---

People

(Reporter: u421692, Assigned: milan)

References

Details

(Keywords: regression)

Attachments

(1 file)

Attached image screenshot
Environment:
Device: HTC Desire HD(Android 2.3.5)
Build: Firefox 33 Beta 6

Steps to reproduce:
1. http://people.mozilla.org/~atrain/mobile/tests/media.html
2. Play first video

Expected result:
Video plays without any issues

Actual result:
VP9/WEBM has video artifacts, only plays on half screen(see attached screenshot)
Component: Graphics, Panning and Zooming → Video/Audio
Product: Firefox for Android → Core
Version: Trunk → 33 Branch
Mihai is this a regression?
tracking-fennec: --- → ?
Flags: needinfo?(mihai.g.pop)
Can you also capture OMX related output in console and attach it here? I also searched through history and see a similar bug 993432 filed against the HTC Desire HD
Logs: https://pastebin.mozilla.org/6526848

2014-05-22 - good build
2014-05-23 - bad build
http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=b40296602083&tochange=e9b2b72f4e6c

Can only guess that this regressed from Bug 908503 - Update demuxer in MP4Reader to be usable for <video>
Flags: needinfo?(mihai.g.pop)
Keywords: regression
Blocks: 908503
[Tracking Requested - why for this release]: Regression in Fx32
tracking-fennec: ? → 32+
Anthony, does that ring a bell? Thanks
Flags: needinfo?(ajones)
That commit imports some android code which appears to be interfering with VP9 playback on Android itself. The C++ code is in a separate namespace but the C code is not. I've spent some time working on it but haven't found a sensible solution yet.
Not me this time. Looks like the culprit is:

78bc7d04bac9 Andreas Gal — Bug 1012407 - Part 6: Eliminate VBO use and DrawQuads and manually draw no-repeat quads (r=Bas)

http://hg.mozilla.org/mozilla-central/rev/78bc7d04bac9
Blocks: 1012407
No longer blocks: 908503
Milan - this looks like a graphics issue introduced in Andreas' patch. Can you find someone to look into it?
Flags: needinfo?(ajones) → needinfo?(milan)
Will take care of it - Anthony, since you can reproduce this - any assertions when running a debug build?
Flags: needinfo?(ajones)
(In reply to Anthony Jones (:kentuckyfriedtakahe, :k17e) from comment #7)
> Not me this time. Looks like the culprit is:
> 
> 78bc7d04bac9 Andreas Gal — Bug 1012407 - Part 6: Eliminate VBO use and
> DrawQuads and manually draw no-repeat quads (r=Bas)
> 
> http://hg.mozilla.org/mozilla-central/rev/78bc7d04bac9

Even after the "fix for part 6" in https://hg.mozilla.org/mozilla-central/rev/5b6ae4c0a754 ?
(In reply to Milan Sreckovic [:milan] from comment #9)
> Will take care of it - Anthony, since you can reproduce this - any
> assertions when running a debug build?

I'll check that. You should be able to repro it on any Gingerbread phone, or at least I could on the first one I tried.

I checked nightly, beta and a lot of builds in the regression range. I'll try that specific commit.
Flags: needinfo?(ajones)
It is still broken on 5b6ae4c0a754.
"Any Gingerbread phone" isn't easy to find these days.  I found Samsung GT-S5670L, with 2.3.4 on it, but Nightly runs fine, shows that video without scaling it.  Also no luck reproducing on Samsung S, with 2.3.6 on it with the Nightly.
Flags: needinfo?(milan)
Trying to get the data remotely - can somebody that can reproduce this try setting preference layout.css.devPixelsPerPx to 1 and see if it affects this bug?
I imagine we're tickling a driver bug - the phone Mihai listed is Adreno GPU, the two that I tried with are PowerVR and VideoCore.  Anthony, what's the device you're testing with?
I have a Nexus S on 2.3.4 and am unable to reproduce this.
(In reply to Milan Sreckovic [:milan] from comment #14)
> Trying to get the data remotely - can somebody that can reproduce this try
> setting preference layout.css.devPixelsPerPx to 1 and see if it affects this
> bug?

I was able to reproduce this issue on a HTC Desire S (Android 2.3.3), setting preference layout.css.devPixelsPerPx to 1 did not make any difference.
(In reply to Mihai Pop from comment #17)
> I was able to reproduce this issue on a HTC Desire S (Android 2.3.3)...

Makes sense, also Adreno.
It does not seem we are going to have a patch for 33. wontfix then!
(In reply to Sylvestre Ledru [:sylvestre] from comment #19)
> It does not seem we are going to have a patch for 33. wontfix then!

Right, probably not - Anthony is going to bring along the phone next week, so we can take a look at it.
Added to the 33 release notes (known issues)
Milan - Were you able to debug with Anthony this week?
Flags: needinfo?(milan)
I have the phone in my hand now, but no, haven't had a chance to actually look at it.  Next week is a short week, but should be able to take care of it.
Milan - As you're currently working on this one, I'm assigning to you. (Like to know who's tracking beta bugs.)
Assignee: nobody → milan
Duplicate of this bug: 1086542
We've been focused on some higher priority issues and so are going to have to ship this again in 34. We can try for a fix again in 35.
Flags: needinfo?(milan)
Re-ping Milan for this one - we've shipped a couple of releases already with this regression, is this on your radar still and can we expect any resolution in the next week to get onto beta?
Flags: needinfo?(milan)
We're out of time for 35, but tracking once more for 36 to get an answer to comment 27 and see if we'll get any traction here.
Yes, let's try one more cycle.
Flags: needinfo?(milan)
tracking-fennec: 32+ → +
We have this bug since 32 and still no activity, I don't see the point of tracking it for much longer.
Component: Audio/Video → Audio/Video: Playback
Gingerbread is no longer supported.
Status: NEW → RESOLVED
Closed: 3 years ago
Resolution: --- → INCOMPLETE
You need to log in before you can comment on or make changes to this bug.