865091 - Videos don't play on YouTube

Status: RESOLVED FIXED

(Firefox OS Graveyard :: Gaia::Video, defect)

Description

[Simon Callan, Telefonica]

User Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10_8_2) AppleWebKit/537.31 (KHTML, like Gecko) Chrome/26.0.1410.65 Safari/537.31

Steps to reproduce:

Open YouTube app and try to play a video


Actual results:

You only see a blank screen - no video gets played


Expected results:

The video should play

Comment 2

[Jason Smith [:jsmith]]

Needs more information. Specifically:

1. What device is this reproducing on?
2. Which YouTube videos?

The general YouTube video functionality is already tested generally, so I'm wondering if this is video-specific (e.g. h264 requested) vs. device-specific.

Comment 3

[Simon Callan, Telefonica]

This is on ZTE Open.  Go to YouTube search for e.g. Cat Jumps Off Building FAIL by chendalson.
Or go to the home screen and try to play any of the videos. Some give a message "watch on your PC later" some simply give a black screen. I've not found a single video that plays yet..

Comment 4

[Simon Callan, Telefonica]

I've just tried testing on the LG device and the same videos work - but not on the ZTE Open

Comment 5

[Jason Smith [:jsmith]]

(In reply to Simon Callan, Telefonica from comment #4)
> I've just tried testing on the LG device and the same videos work - but not
> on the ZTE Open

Hmm...what I think you might be doing then is consuming a video that contains embedded h264 content in the web. 1.1 supports this, 1.01 does not.

The part that's weird is why you are not getting the vnd protocol served that takes you out to the video app.

Just to confirm - Are you getting sent to the video app here when you try to view a YouTube video?

Comment 6

[Simon Callan, Telefonica]

NO it stays inside the app - well as far as I can see..

The wierd thing now, is that all the videos pop up with an error page "YouTube won't play video" (on both the ZTE and LG and the Simulator)..

Comment 8

[Jason Smith [:jsmith]]

Confirmed by another person. Looks like a vnd protocol parsing regression.

Comment 9

[Jason Smith [:jsmith]]

See dupe for video of the problem.

Comment 10

[leo.bugzilla.gecko]

leo device has same problem, "YouTube won't play video".

Since request.open('GET', query) in youtube.js returns error, the streaming cannot be started.
By TCP dump, it received OK response from the server.

Comment 12

[Jason Smith [:jsmith]]

So we've got an automated test here for this:

https://github.com/mozilla/gaia-ui-tests/blob/master/gaiatest/tests/videoplayer/test_video_youtube.py

Zac - Is our automation failing right now due to this bug? If not, can you find out why?

I'm wondering if our current automated coverage would be able to catch this bug.

Comment 13

[Zac C (:zac)]

Jason, yes this has been failing on our automation for the last 18 hrs but we haven't enough people in that timezone to investigate/report it.

On Unagi/v1-train build we get the message that leo reports in comment #10.

Comment 14

[Diego Wilson [:diego]]

I tested on an earlier known good leo build. A week ago youtube videos were playing fine on this archived build. Now they all show the "YouTube won't play video". I don't think it's a regression per se. I think something changed in m.youtube.com that broke it on b2g.

Comment 15

[Jason Smith [:jsmith]]

(In reply to Diego Wilson [:diego] from comment #14)
> I tested on an earlier known good leo build. A week ago youtube videos were
> playing fine on this archived build. Now they all show the "YouTube won't
> play video". I don't think it's a regression per se. I think something
> changed in m.youtube.com that broke it on b2g.

Well, that's really bad then. I should have known this would eventually happen, given that we're relying on an unpublished API.

Chris - Can someone outreach to YouTube here?

Comment 16

[Jason Smith [:jsmith]]

Hmm...I tested this on Fx Android Nightly and it works fine. So it's probably something that changed on the site side in how we're interpreting the VND protocol then.

Given the bustage here though, we probably still need someone to establish with YouTube that we're consuming this API.

Comment 17

[Nick Davidson]

Attachment: Error message playing YouTube video

Issue still reproduses. Repro frequency:  100%, on 3 different devices
See video:  http://youtu.be/g6lT16ymo3E and screenshot for more info.

User is unable to watch any YouTube videos on Inari device going directly via browser to m.youtube.com or go via Everything.me section and selecting YouTube music videos. 

Repro Steps:
1) Updated to Inari Build ID: 20130424070201  
2) Open the browser and navigate to youtube.com 
3) Select a video to play

Actual:  User is getting  message “YouTube won’t play video:”  with blue “OK” button on a bottom

Expected:  The video plays with audio and no glitches is observed.

Repro frequency:  100%, on 3 different devices
See video:  http://youtu.be/g6lT16ymo3E and screenshot for more info.

Environmental  Variables:
Inari Build ID: 20130425070201
Kernel Date: Feb 21
Gecko: http://hg.mozilla.org/releases/mozilla-b2g18_v1_0_1/rev/5d454864b668
Gaia: 052253176f3ad495a1086e47a28fffbf07613388

Comment 18

[Ben Francis [:benfrancis]]

A community member mentioned on IRC yesterday that they'd noticed a change on YouTube's server within the last week. They were using the same mechanism we use to launch YouTube videos in the app they were developing.

Comment 19

[sajirdalas]

(In reply to Ben Francis [:benfrancis] from comment #18)
> A community member mentioned on IRC yesterday that they'd noticed a change
> on YouTube's server within the last week. They were using the same mechanism
> we use to launch YouTube videos in the app they were developing.

I was using an external library to expose the raw stream URL for the video, using a similar method of what youtube.js uses on the gaia video app. Someone on github noted that youtube changed / dropped the CORS headers for the http://www.youtube.com/get_video_info endpoint. 

Sure enough while reproducing this bug youtube.js throws a query error for http://www.youtube.com/get_video_info

Comment 20

[Jason Smith [:jsmith]]

Good analysis in comment 19 - that confirms some of my theories I had on the problem here. Sounds like what we should do here is:

1. Adjust our youtube.js to take into account youtube's changes
2. Get help from Business Development (given this is a preinstalled app for v1.01 ship) to lock down a deal about our usage of the VND protocol API with YouTube. Going to redirect the needinfo request to Tom now as a result.

Comment 21

[Lawrence Mandel [:lmandel] (use needinfo)]

If we can't lock down #2, I suggested a mitigation strategy in bug 843334 that involves setting up a Mozilla owned server side service and piping the YouTube vnd content through it. This should allow us to respond to changes in YouTube content without the need to update B2G. The solution isn't free but may be preferable to shipping without the ability to respond to future YouTube changes.

Comment 22

[Thomas Elin [:relaas]]

I'm happy to go to Youtube but need a clear direction of next steps and how to fix this issue. Harald can act as my proxy until we have something to share with them.

Comment 23

[[:fabrice] Fabrice Desré]

Attachment: patch

Use system XHR to bypass CORS limitation from Youtube...

Comment 24

[sajirdalas]

(In reply to Fabrice Desré [:fabrice] from comment #23)
> Created attachment 742623 [details] [diff] [review]
> patch
> 
> Use system XHR to bypass CORS limitation from Youtube...

That seems to have fixed youtube.js... It's returning a source correctly now

Comment 25

[Dale Harvey (:daleharvey)]

Comment on attachment 742623 [details] [diff] [review]
patch

Stolen the review, tested this and works great

Comment 26

[Dale Harvey (:daleharvey)]

Merged in: https://github.com/mozilla-b2g/gaia/commit/8757b15adc8a9e3cfeed4283fc7802964af9c35e

Comment 27

[John Ford [:jhford] CET/CEST Berlin Time]

Uplifted 8757b15adc8a9e3cfeed4283fc7802964af9c35e to:
v1-train: cb6c6e9e25c313dd7fd339d83528c0b7a940f630
v1.0.1: 09046df332cd217cf4fad55c48ce9ce3f14ee5a4

Comment 29

[Jason Smith [:jsmith]]

Our automation is passing this test on a 5/1 build.

Comment 32

[bov]

Tested on Inari. Commercial RIL: 094. Build Id: 20130506003326
It is possible to play videos from YouTube APP, and  from the site www.youtube.com

Comment 33

[Jason Smith [:jsmith]]

Works for me is inappropriate to use here - we've confirmed a fix here. You want to set the status flag in this case to verified.

Comment 34

[Chris Lord [:cwiiis]]

Seems this fix doesn't work on master (gaia master and gecko master) - has anyone else tried?

Comment 35

[Jason Smith [:jsmith]]

(In reply to Chris Lord [:cwiiis] from comment #34)
> Seems this fix doesn't work on master (gaia master and gecko master) - has
> anyone else tried?

Probably not. No one really tests on master these days unless you are doing gaia development.

Comment 36

[Diego Wilson [:diego]]

On master before this patch I used to see an "unable to play video" message. With the patch I can't play the video but I don't see the message either. I think the issue affecting all branches was fixed, but it uncovered some further problem in master.

Comment 37

[John Hammink]

This problem was at least not appearing on  the 5/13 partner build which we were checking against for IOT certification in Chile.

Comment 38

[John Hammink]

Apparently, it reappears on 5/17 build however in IOT.  Reopening.

Comment 39

[Jason Smith [:jsmith]]

Please do not reopen bugs. There are new youtube bugs filed right now that followed the fix here. It's already landed at this point too.

If you are seeing this bug again, then file a new bug for the specific device. I'm a bit surprised though - I tested this on a 5/17 1.01 inari build yesterday and this was working fine. So double check the build ID of the device you are testing on.