1175220 - Corrupted webm video

Status: RESOLVED FIXED

(Core :: Audio/Video, defect)

Description

[Fanolian]

User Agent: Mozilla/5.0 (Windows NT 6.3; WOW64; rv:41.0) Gecko/20100101 Firefox/41.0
Build ID: 20150616030201

Steps to reproduce:

Load https://www.microsoft.com/microsoft-hololens/Assets/img/anim/home/HeroRoom_hydrate_md.webm
or https://www.microsoft.com/microsoft-hololens/Assets/img/anim/home/HeroRoom_hydrate_lg.webm

The videos can be found on https://www.microsoft.com/microsoft-hololens/en-us.
Please visit the site with a viewport min-width 961px as it serves .mp4 version, which plays fine, for <961px.


Actual results:

Video is corrupted.

From mozregression:
Last good revision: b8effe1ac173
First bad revision: 9a60bf065716
Pushlog: https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=b8effe1ac173&tochange=9a60bf065716

From about:support

Graphics
Adapter Description	NVIDIA GeForce GTX 760
Adapter Description (GPU #2)	Intel(R) HD Graphics 4600
Adapter Drivers	nvd3dumx,nvwgf2umx,nvwgf2umx nvd3dum,nvwgf2um,nvwgf2um
Adapter Drivers (GPU #2)	igdumdim64 igd10iumd64 igd10iumd64 igdumdim32 igd10iumd32 igd10iumd32
Adapter RAM	4095
Adapter RAM (GPU #2)	Unknown
Asynchronous Pan/Zoom	wheel input enabled
ClearType Parameters	D [ Gamma: 2200 Pixel Structure: R ClearType Level: 100 Enhanced Contrast: 100 ] D [ Gamma: 2400 Pixel Structure: R ClearType Level: 100 Enhanced Contrast: 100 ]
Device ID	0x1187
Device ID (GPU #2)	0x0412
Direct2D Enabled	true
DirectWrite Enabled	true (6.3.9600.17795)
Driver Date	5-27-2015
Driver Date (GPU #2)	3-16-2015
Driver Version	9.18.13.5306
Driver Version (GPU #2)	10.18.14.4170
GPU #2 Active	false
GPU Accelerated Windows	1/1 Direct3D 11 (OMTC)
Subsys ID	282b1462
Subsys ID (GPU #2)	0000000c
Supports Hardware H264 Decoding	true
Vendor ID	0x10de
Vendor ID (GPU #2)	0x8086
WebGL Renderer	Google Inc. -- ANGLE (NVIDIA GeForce GTX 760 Direct3D11 vs_5_0 ps_5_0)
windowLayerManagerRemote	true
AzureCanvasBackend	direct2d 1.1
AzureContentBackend	direct2d 1.1
AzureFallbackCanvasBackend	cairo
AzureSkiaAccelerated	0

(Please advise a better bug title, thanks.)

Comment 1

[Virtual_ManPL [:Virtual] 🇵🇱 - (please needinfo? me - so I will see your comment/message/reply/question/etc.)]

[Tracking Requested - why for this release]: Regression

Comment 2

[Ralph Giles (:rillian)]

Thanks for the report. Fascinating. Is it possible to get a screenshot of the corruption to get a better idea where the failure is?

The file crashes firefox on my windows vm(!) so I can't reproduce directly.
https://crash-stats.mozilla.com/report/index/24e514f9-16df-4092-b713-6720a2150617

David, does my crash look like another alignment problem to you?

Comment 3

[(Away)]

No, it's a "vmovdqu" instruction and the "u" means it accepts unaligned data.

You got an EXCEPTION_ILLEGAL_INSTRUCTION which at first glance would mean your CPU (or VM, or OS) doesn't support AVX2. Presumably we'd check for support before running the AVX2 version of that function?

Comment 4

[Fanolian]

Attachment: corrupted video screenshot

(In reply to Ralph Giles (:rillian) from comment #2)
> Thanks for the report. Fascinating. Is it possible to get a screenshot of
> the corruption to get a better idea where the failure is?

Attached is a screenshot of the corruption.
Here is a video capture of the corruption: https://gfycat.com/SilentInsecureChital

Comment 5

[Ralph Giles (:rillian)]

(In reply to David Major [:dmajor] from comment #3)
> No, it's a "vmovdqu" instruction and the "u" means it accepts unaligned data.
> 
> You got an EXCEPTION_ILLEGAL_INSTRUCTION which at first glance would mean
> your CPU (or VM, or OS) doesn't support AVX2. Presumably we'd check for
> support before running the AVX2 version of that function?

We're supposed to, yes. Thanks for the analysis!

Comment 6

[jen strickland :jenstrickland]

Tracked for 40 and 41, due to regression (unaffected in 39), and from what I read above, to ensure "we check for support before running the AVX2 version of that function".

Comment 7

[Fanolian]

I cannot reproduce it on 2015-06-20 Nightly build. Perhaps it is fixed by bug 1151175.
Since bug 1151175 has not been uplifted to 40, I do not close this bug for now.

Comment 8

[Ralph Giles (:rillian)]

Thanks for testing with the newer nightly, Fanolian. I was wondering if the the libvpx update would fix it. Other candidate is the vpx hardware acceleration, but I believe that's only available with newer intel graphics.

Comment 9

[Ralph Giles (:rillian)]

Now that bug 1151175 has been uplifted to 40, we should be able to close this.