Nils,

 

We are still not able to fix the below e-mail trial that describes issues in 
getting information to be able to do testing and monitoring.

.

Any new development in the past months?

 

Can you help us how to fix the flowing we have customers using your browser, 
but we can't help them with the testing and monitoring.

 

Below the e-mail trail we had. No change in our in the ability to fix the 
issue. 

 

Thanks for helping!

 

Benjamin

 

 

---------- Forwarded message ---------
From: Muly Oved <m...@testrtc.com>
Date: Mon, Jul 25, 2016 at 9:45 AM
Subject: Re: Firefox
To: Nils Ohlmeier <nohlme...@mozilla.com>
Cc: Tsahi Levent-Levi <tsa...@testrtc.com>, Andrey Korin <a.ko...@outlook.com>, 
Puneet Goyal <puneet.pra...@gmail.com>

 

Hi

 

Managed to bring up everything back with Firefox 47.0.1

Run the test from this site 
https://webrtc.github.io/samples/src/content/peerconnection/states/

 

I believe this test isolate the problem to be related to media and not network 
as all test is on the same web page

1.      See attached test report, notice fps is most of the time < 1fps
2.      Attached channels stats, they are rather low but also for the out rtp.
3.      Attached Server stats, can see CPU is not too busy, machine spec is 
bare-metal 4 cores (8 Threads), 8GB. 100 Mbps.
4.      The test itself is run inside docker Ubuntu 16.04 based on this images 
https://github.com/SeleniumHQ/docker-selenium

Will love to get any feedback on this.

 

Muly

 

 

On Thu, Jul 21, 2016 at 11:07 PM Nils Ohlmeier <nohlme...@mozilla.com> wrote:

Hi Muly,

 

I asked our automation guys and they confirmed that 47.0 has a problem with 
Selenium. You should upgrade to 47.0.1, as it should be fixed in that version.

If it is still not working in 47.0.1 for you please let me know.

 

Best

  Nils Ohlmeier

 

On Jul 21, 2016, at 04:11, Muly Oved <m...@testrtc.com> wrote:

 

Nils

 

Thanks for the answer, started to retest but right now we have a problem with 
Firefox 47.0 not working with selenium, trying to solve this and will update.

 

Muly

 

On Tue, Jul 19, 2016 at 1:43 AM Nils Ohlmeier <nohlme...@mozilla.com> wrote:

Hi Muly,

 

I apologize for the very late reply.

 

I don’t have an obvious answer to the problem you are describing. For Linux you 
could try to set the user pref 'layers.acceleration.force-enabled’ to true. 
That setting requires a restart of the browser. What it does is that it 
bypasses quite a bit of graphics conversion on Linux if the graphic card is 
supported.

 

But the other question is why are the bit rates so much lower?

Are these rates bidirectional? What is sitting on the other side of the 
connection - always the same identical machine and OS, or a different OS?

 

I think I would like to see a tcpdump/wireshark trace from one of these calls 
to check if we can see lots of RTCP traffic indicating frame los.

 

Best regards

  Nils

 

On May 31, 2016, at 05:53, Muly Oved <m...@testrtc.com> wrote:

 

Nils

 

Trust you are doing good,

 

We done a lot of work to integrate our system with Firefox but we seem to have 
a major performance issue which we are not sure why. 

 

 


Service Setup

OS

FF 

Audio Bit Rate

Video Bit Rate

Video FPS


Browserstack

OS X Mavericks (v 10.9)

46

50

930

30

        
Win 7

46

49

940

30

        
Win XP

46

50

950

30

        
Win 10 

46 

49

900

26

                                                


Docker

Linux

46

4

450

1


Baremetal

8core

Linux

46

16

592

9


Azure VM

Windows 2012

46

8

352

5

 

Some background about the setup

1.      We open AppRTC webrtc page
2.      Add hidden video element to the page and load a VGA webm video from CDN
3.      we capture stream from video using HTMLMediaElement.mozCaptureStream() 
and use this stream as fake input for getUserMedia
4.      As far as we see from the SDP the codec used is VP8  

On our main setup Linux on Docker, the video get play in 1fps, 

running without docker Linux or Window strong cloud machines give us 9/5 fps

 

Not sure why but same on BrowserStack OSX and Windows machine show much better 
results but still cannot run HD size video and reach our Docker+Chrome setup 
which has no problem to run HD.

 

Any hint will be highly appreciated.

 

Muly

 

 

 

 

 

 

 

 

 

 

On Fri, Mar 4, 2016 at 11:02 PM Nils Ohlmeier <nohlme...@mozilla.com> wrote:

Hi guys,

 

I filled this bug for you https://bugzilla.mozilla.org/show_bug.cgi?id=1253720 
and referenced your jsfiddle in it.

I was not able to CC you with your testrc.com <http://testrc.com/>  email 
addresses, as they don’t appear to be bugzilla accounts. Feel free to CC 
yourself with whatever bugzilla account you have.

If you want to file any bug in the future yourself just go to 
https://bugzilla.mozilla.org/, click ‘New’ at the top, choose ‘Core’ from the 
list of Components and then chose one of the WebRTC:* components (or just 
‘WebRTC’ if you are not sure which component fits best).

 

Also feel free to comment in the bug to indicate your dependency on this bug. 
That always helps with prioritizing bugs properly on our end.

 

Best

  Nils

 

On Mar 1, 2016, at 00:55, Tsahi Levent-Levi <tsa...@testrtc.com> wrote:

 

Nils,

 

Can you please direct us to where do we file such bugs?

 

Regards,

Tsahi

 

 

On Tue, Mar 1, 2016 at 10:46 AM, Nils Ohlmeier <nohlme...@mozilla.com> wrote:

Hi Muly,

 

I’m not an expert on the media rendering side. But from what I’m seeing in your 
jsfiddle (awesome, thanks for that), I would recommend to file a bug (I can 
also file it for you if that helps). Quite likely that you hit an unknown bug 
here.

 

Best

  Nils Ohlmeier

 

On Feb 25, 2016, at 02:11, Muly Oved <m...@testrtc.com> wrote:

 

Niles, 

 

Need some advice on this, 

 

We where busy integrate media recorder and other things and we managed to push 
local video into the webrtc session on FireFox to simulate a camera when 
testing.

 

We now facing a problem where things stop working when the local video get 
"seek" event (need this to play the video in loops) 

 

Here jsfiddle that demonstrate the problem on FireFox

 

https://jsfiddle.net/411h3y27/10/

 

Will appreciate any input on this.

Muly

 

 

On Wed, Oct 28, 2015 at 10:48 PM Tsahi Levent-Levi <tsa...@testrtc.com> wrote:

Thanks Nils. 

We'll see which option works best for us. 

Regards, 
Tsahi 

בתאריך 28 באוק׳ 2015 22:25,‏ "Nils Ohlmeier" <nohlme...@mozilla.com> כתב:

 

Hi,

 

Sorry I actually asked around, but then forgot to reply to you guys.

 

So there are two things you could use in Firefox:

 

1) On the ‘about:webrtc' page there is a button ‘Start AEC Logging’ which we 
believe behaves similar to Google’s audio recording feature.

2) You could start looking into the media recorder API.

 

Both alternatives probably need some work to get them to where you need them, 
but hopefully they are at least good starting points.

 

Best

  Nils Ohlmeier

 

On Oct 28, 2015, at 03:06, Tsahi Levent-Levi <tsa...@testrtc.com> wrote:

 

Nils,

 

Ca you please help us with this? We're in the process of adding Firefox 
support, but this one is blocking us.

 

Regards,

Tsahi

 

 

On Tue, Oct 20, 2015 at 8:18 AM, Muly Oved <m...@testrtc.com> wrote:

Hi

 

Thanks for the links, that did help and it seem to work fine also a bit fragile 
as it depend on the script loading order and sometimes the client code can 
capture the getUserMedia() before the test code.

 

Anyway, have another question, we need to record the incoming audio (maybe 
later also video) for audio quality purpose.

 

In Chrome we use the webrtc-internals page to request audio recording.

 

Is there a recommended way to achieve this in Firefox?

 

Thanks

Muly

 

 

 

 

On Tue, Sep 15, 2015 at 11:11 PM Nils Ohlmeier <nohlme...@mozilla.com> wrote:

Hi Muli,

 

please to meet you.

 

On Sep 13, 2015, at 06:41, Muly Oved <m...@testrtc.com> wrote:

 

Hi

 

Glad to get in touch, I have a ready made questions...

 

Some background:

1.      We run chrome browser inside docker using Selenium
2.      For Chrome we collect the statistic using chrome://webrtc-internals/ 
but we now move to use getStats so will be able to collect also for Firefox

1.      On script startup we override RTCPeerConnection()
2.      When channel open we start call getStats() very 1 second

#1 Wonder if you think override RTCPeerConnection() is valid approach or there 
maybe problem with this?

 

No, as long as you pass through all the API calls to the actual 
mozRTCPeerConnection, you should be able to do that. In fact I think Firefox 
Hello uses a similar approach of wrapping a lot of the API calls if not the 
whole PeerConnection.

 

#2 For Chrome we use those flag to simulate audio and video

 

disable-user-media-security

use-fake-ui-for-media-stream

use-fake-device-for-media-stream

use-file-for-fake-audio-capture=/media/audio.wav

use-file-for-fake-video-capture=/media/video.y4m

 

In Firefox we have build in fake device for audio and video. You can simply 
activate them by adding ‘fake: true’ to your gUM constraints. So for example 
the typical audio + video constraint would look like this: {audio: true, video: 
true, fake: true}

 

Alternatively you can also activate the fake device without using the 
constraints, but instead setting a user preference in about:config.

https://dxr.mozilla.org/mozilla-central/source/dom/media/tests/mochitest/head.js?offset=600#244

The pref itself as you can see in the example above is called 
'media.navigator.streams.fake'

 

Based on this

 

https://hg.mozilla.org/mozilla-central/rev/cb7eb7fcf67b

 

I would ignore that code if I were you. What that change does is that on 
certain Linux test machines we have actual fake video drivers installed. As 
Firefox build in fake devices take a different code path that a regular 
external video device, this ensures we get testing overage for external device.

 

Seem there is some way to inject faked audio and video stream to Firefox

 

Is it possible to inject any audio file and video file to FF or this is limited 
to some build-in generated stream in FF?

 

 

Here is a test case which loads a video from a file and streams that through 
the PeerConnection instead of the gUM video:

https://dxr.mozilla.org/mozilla-central/source/dom/media/tests/mochitest/test_peerConnection_capturedVideo.html?offset=10200

 

 

Any example how to use it from Selenium?

 

Sorry we don’t use Selenium to test our platform parts. The Firefox Hello guys 
use marionette, which uses the same web-driver technology as Selenium AFAIK. 
But they mostly cover the Firefox Hello UI pieces, not the underlying 
audio/video pieces.

 

 

I hope this helps. Let me know if you have any further questions.

 

Best

  Nils Ohlmeier

 

Muly

 

 

 

Question is also asked here:

 

How can I pass a fake media stream to firefox

http://stackoverflow.com/questions/31859459/how-can-i-pass-a-fake-media-stream-to-
 
<http://stackoverflow.com/questions/31859459/how-can-i-pass-a-fake-media-stream-to-firefox-from-command-line>
 firefox-from-command-line

 

How can a fake media stream and media device be passed to firefox via command 
line/selenium

https://groups.google.com/forum/#!msg/selenium-users/ZANuzTA2VYQ/KqMqDEaij1IJ

 

https://groups.google.com/forum/#!msg/discuss-webrtc/87xwPYD6sl4/52h-vvGAjSQJ

 

 

 

On Sun, Sep 13, 2015 at 4:21 PM Tsahi Levent-Levi <tsa...@testrtc.com> wrote:

Hi Nils,

 

It was great meeting with you in person and have you speak at the Kranky Geek 
event.

 

As we discussed, I am sending out this intro :-)

 

Nils - Muly is our lead developer. He is currently looking into adding our 
testRTC service support Firefox. He has some integration questions to ask...

 

Muly - Nils is the person to talk to. He said you can harass you as much as you 
need.

 

Please keep me CC'd on these emails if possible.

 

Regards,

Tsahi

 

 

 

 

 

 

 

 

_______________________________________________
dev-media mailing list
dev-media@lists.mozilla.org
https://lists.mozilla.org/listinfo/dev-media

Reply via email to