1192016 - False color patches in HTML5 MP4 video playback

Status: RESOLVED WORKSFORME

(Core :: Audio/Video: Playback, defect, P1)

Description

[aleth [:aleth]]

I've noticed lurid false colour patches in video playback recently.

At least on my machine, this is a good example (the patches I'm talking about are mainly pink and green): https://www.kickstarter.com/projects/514621648/tower-57

I've also come across it on youtube, so maybe this is related to bug 1185356.

Comment 1

[Chris Peterson [:cpeterson]]

Aleth, I can't reproduce the problem on my MacBook Pro running OS X 10.10. Can you share your about:support information about your hardware?

Do you see the problem if you flip the "media.hardware-video-decoding.enabled" pref to false?

Comment 2

[aleth [:aleth]]

(In reply to Chris Peterson [:cpeterson] from comment #1)

I've noticed the issue is not consistently reproducible, e.g. today (after a FX restart) for some reason the video in the above link plays back OK.

When it does happen, the false colour areas appear to creep in from the edges of the video.

> Aleth, I can't reproduce the problem on my MacBook Pro running OS X 10.10.
> Can you share your about:support information about your hardware?

This is a 2011 MBP running 10.10.4.

Application Basics
------------------

Name: Firefox
Version: 41.0a2
Build ID: 20150806004006
Update Channel: aurora
User Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.10; rv:41.0) Gecko/20100101 Firefox/41.0
Multiprocess Windows: 0/2 (default: false)

Graphics
--------

Asynchronous Pan/Zoom: none
Device ID: 0x0126
GPU Accelerated Windows: 2/2 OpenGL (OMTC)
Supports Hardware H264 Decoding: false
Vendor ID: 0x8086
WebGL Renderer: ATI Technologies Inc. -- AMD Radeon HD 6750M OpenGL Engine
windowLayerManagerRemote: true
AzureCanvasBackend: skia
AzureContentBackend: quartz
AzureFallbackCanvasBackend: none
AzureSkiaAccelerated: 0

Important Modified Preferences
------------------------------

media.gmp-gmpopenh264.lastUpdate: 1430781402
media.gmp-gmpopenh264.version: 1.4
media.gmp-manager.buildID: 20150806004006
media.gmp-manager.lastCheck: 1438866237
plugin.state.flash: 0

Comment 3

[avada]

(In reply to aleth [:aleth] from comment #0)
> I've noticed lurid false colour patches in video playback recently.
> 
> At least on my machine, this is a good example (the patches I'm talking
> about are mainly pink and green):
> https://www.kickstarter.com/projects/514621648/tower-57
> 
> I've also come across it on youtube, so maybe this is related to bug 1185356.

If I were you, I'd make a screenshot so others could see what's happening.

Comment 4

[aleth [:aleth]]

Attachment: Screen Shot

(In reply to avada from comment #3)
> If I were you, I'd make a screenshot so others could see what's happening.

OK, here's an example.

Comment 5

[Chris Peterson [:cpeterson]]

aleth, this bug looks a lot like a OS X video bug I reported (bug 1187103), which was just fixed in Nightly 42 (by bug 1187103). Can you still reproduce this bug in the latest Nightly 43 build? Was your screenshot from Nightly 43?

Comment 6

[aleth [:aleth]]

The screenshot was from Aurora 42. It also happened in 41.

I haven't seen it in Nightly 43 so far, which is good news. However, I am using Aurora 42 as my main browser, and despite the recent uplift I'm still experiencing the problem there.

Comment 7

[Chris Peterson [:cpeterson]]

(In reply to aleth [:aleth] from comment #6)
> I haven't seen it in Nightly 43 so far, which is good news. However, I am
> using Aurora 42 as my main browser, and despite the recent uplift I'm still
> experiencing the problem there.

That is interesting. That suggests the bug has been recently fixed in Nightly 43 or the bug is somehow hidden in the Nightly channel.

Are you using multi-process Firefox (aka "e10s")? It is enabled out-out in the Nightly channel, but available opt-in in the Aurora channel. Can you please test Nightly 43 and uncheck "Enable multi-process Nightly" in your Firefox Preferences? I recently saw an unrelated bug that was hidden in the Nightly channel because e10s is enabled.

Comment 8

[aleth [:aleth]]

> Are you using multi-process Firefox (aka "e10s")? 

I've seen it in 42 with e10s and in 41 with and without e10s, so it seems to be independent of that.

I'm not certain it's fixed in 43, as my profile on 43 is much lighter (testing only) so it's not an exact replication.

Comment 9

[aleth [:aleth]]

Unfortunately I've now also seen this in a current nightly.

Comment 10

[Anthony Jones (:ajones, :kentuckyfriedtakahe, :k17e)]

JYA - can you look into this one?

Comment 11

[Ryan VanderMeulen [:RyanVM]]

Does this still reproduce for you, aleth?

Comment 12

[aleth [:aleth]]

(In reply to Ryan VanderMeulen [:RyanVM UTC-4] from comment #11)
> Does this still reproduce for you, aleth?

I haven't seen this issue recently (since the updates to Fx 43, OSX 10.11). Resolving WFM for now, but I'll keep an eye open for it.

Comment 13

[Jean-Yves Avenard [:jya]]

We do not perform any conversions on mac anymore, it's all done by the GPU. so I don't believe this issue still occurs in 43 and later