1194662 - Videos on www.brightcove.com don't work in Firefox for Android and Firefox OS

Status: RESOLVED FIXED

(Web Compatibility :: Site Reports, defect, P1)

Description

[Marco Castelluccio [:marco]]

Looks like the Brightcove solution is used by a lot of media websites.

In Chrome the videos on their page play correctly.

In Firefox an error appears when you try to play a video ("This video is either unavailable or not supported in this browser" "Error Code: MEDIA_ERR_SRC_NOT_SUPPORTED" ...)

The error looks similar to the one in bug 1194658.

In Firefox OS there's a different error ("Could not download the video" "Error Code: PLAYER_ERR_TIMEOUT").

Comment 1

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

Does this happen in nightly?

Comment 2

[Ralph Giles (:rillian)]

CAn you give us a url which shows the problem to test against?

Comment 3

[Marco Castelluccio [:marco]]

(In reply to Ralph Giles (:rillian) from comment #2)
> CAn you give us a url which shows the problem to test against?

There was a video in the homepage, which is now gone.

I can now reproduce the problem with the second video at https://www.brightcove.com/en/solutions/media, in the section "Why the World's Most Innovative ...".


(In reply to Anthony Jones (:kentuckyfriedtakahe, :k17e) from comment #1)
> Does this happen in nightly?

I've tested with Firefox Beta on Android and a Nightly build of Firefox OS.

Comment 4

[Hallvord R. M. Steen [:hallvors]]

Hm... I've tested in Firefox Beta and Firefox Nightly and it seems the video is broken if there is an Android version number in the UA string. 

The amusing thing here is that one of the main reasons for adding an Android version number to the UA string was that a lot of sites using Brightcove's earlier video library version failed to play videos when we didn't have a version number. Now, today.. Brightcove's *new* video library fails if we *do* have an Android version number. Yay, we just can't win, can we?

Comment 5

[Hallvord R. M. Steen [:hallvors]]

So, this problem occurs when they try to use a M3U8 src for the VIDEO element, and the code that causes it is this one:

                var a, b = /^application\/(?:x-|vnd\.apple\.)mpegurl/i,
                    c = /^video\/mp4/i;
                vjs.Html5.patchCanPlayType = function() {
                    if (vjs.ANDROID_VERSION >= 4) {
                        if (!a) {
                            a = vjs.TEST_VID.constructor.prototype.canPlayType
                        }
                        vjs.TEST_VID.constructor.prototype.canPlayType = function(c) {
                            if (c && b.test(c)) {
                                return "maybe"
                            }
                            return a.call(this, c)
                        }
                    }

So for certain Android versions they make canPlayType() return "maybe" for M3U8.

As it seems the problem is trying to work around bugs in canPlayType() in some other implementations, it's not straightforward to fix it in a nice way. I think they may want to try to handle the error more gracefully and fall back to trying other formats (come to think of it, why not create a VIDEO element with several SOURCE children right away?). The minimal fix would be to teach the script that the hack is not to be applied in Firefox:

                    if (vjs.ANDROID_VERSION >= 4 && !vjs.IS_FIREFOX) {

(But then you might of course get the same problems in other non-Firefox non-M3U8 supporting browsers).

Comment 6

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

(In reply to Hallvord R. M. Steen [:hallvors] from comment #5)
> As it seems the problem is trying to work around bugs in canPlayType() in
> some other implementations, it's not straightforward to fix it in a nice
> way.

The best solution would be to catch the error we fire and fall back. Do you have any contacts in Brightcove?

Comment 7

[Hallvord R. M. Steen [:hallvors]]

We have contacts but so far we haven't actually gotten anyone to change anything.

Comment 8

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

(In reply to Hallvord R. M. Steen [:hallvors] from comment #7)
> We have contacts but so far we haven't actually gotten anyone to change
> anything.

Has any progress been made? Do we need to nag the right person in Brightcove?

Comment 9

[Hallvord R. M. Steen [:hallvors]]

Basically, this is only a problem in their old version. They don't want to maintain that version (even though the libs are usually loaded from their CDN), they would rather have their customers upgrade to the new one. From what I've seen the uptake is a bit slow..

Comment 10

[Hallvord R. M. Steen [:hallvors]]

Oops, sorry. I'm confusing myself. This bug is about the problem affecting the *new* version.. I'll dig up my contacts.

Comment 11

[Hallvord R. M. Steen [:hallvors]]

I've verified that we still have this problem - it also happens on https://brightcove.com/en/perform - and contacted Brightcove again. If we don't get through to the right person this time let's see if we have better contacts..

Comment 12

[Mike Taylor [:miketaylr]]

If we can solve this via a UA override (which requires having a stable domain for the player), I think we probably should.

Hallvord, do we know which domains the new and old players are served from?

Comment 13

[Hallvord R. M. Steen [:hallvors]]

While the JS is typically served from the same server (players.brightcove.net), the videos and other content might come from anywhere. It's not useful to spoof the CDN the scripts come from, but if we come across this on a higher profile site we should probably spoof.

Comment 14

[DS]

FTR I have the same issue on FF on a Galaxy Note 5, fully updated, at SMH.com.au and none of the videos play there though occasionally I do get sound, but the error message sits on ablack background.

Comment 15

[Mark Hammond [:markh] [:mhammond]]

(In reply to DS from comment #14)
> FTR I have the same issue on FF on a Galaxy Note 5, fully updated, at
> SMH.com.au and none of the videos play there though occasionally I do get
> sound, but the error message sits on ablack background.

Also on theage.com.au (which is the same company as smh.com.au) - according to Alexa these are the 55th and 30th top sites in Australia respectively, and 19th and 10th of .au addresses in that list - so these are certainly high-profile sites in this part of the world. Best I can tell these sites also use brightplayer - the video tag says:

<video ... data-fm-brightcove-player="videoOptions: ...}> and best I can tell the video loads from https://edge.api.brightcove.com/playback/v1/accounts/3910869709001/videos/ref:7379649

Comment 16

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

(In reply to Hallvord R. M. Steen [:hallvors] from comment #5)
> As it seems the problem is trying to work around bugs in canPlayType() in
> some other implementations, it's not straightforward to fix it in a nice
> way. I think they may want to try to handle the error more gracefully and
> fall back to trying other formats (come to think of it, why not create a
> VIDEO element with several SOURCE children right away?). The minimal fix
> would be to teach the script that the hack is not to be applied in Firefox:
> 
>                     if (vjs.ANDROID_VERSION >= 4 && !vjs.IS_FIREFOX) {
> 
> (But then you might of course get the same problems in other non-Firefox
> non-M3U8 supporting browsers).

Can you do me a favour and submit a pull request to video.js ?

https://github.com/videojs/video.js

Comment 17

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

I've done a pull request https://github.com/videojs/video.js/pull/3498

Comment 18

[Marco Castelluccio [:marco]]

I've created a new pull request (https://github.com/videojs/video.js/pull/3586) based off of Anthony's one with some test fixes (they couldn't merge Anthony one otherwise).

Comment 19

[Marco Castelluccio [:marco]]

https://github.com/videojs/video.js/pull/3586 has been merged.

Comment 20

[murrayww]

http://www.dailypress.com/tidewater-review/news/91765720-132.html

Black screen in Firefox 49.0.2

Fine in Chrome 54.0.2840.71 m

Do I need an extension/plugin ?

Thanks !

Comment 21

[Marco Castelluccio [:marco]]

(In reply to murrayww from comment #20)
> http://www.dailypress.com/tidewater-review/news/91765720-132.html
> 
> Black screen in Firefox 49.0.2
> 
> Fine in Chrome 54.0.2840.71 m
> 
> Do I need an extension/plugin ?
> 
> Thanks !

They're probably still using the old version of the Brightcove Player.
When they switch to the new version or if/when we ship HLS support, videos should work in Firefox too.

Comment 22

[Desigan Chinniah [:cyberdees] [:dees] [London - GMT]]

The following URL: http://www.dailypress.com/tidewater-review/news/91765720-132.html — is re-producible on desktop (together with :k17e) on a MacBook 12" Nightly 53.0a1

Comment 23

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

The upstream player is fixed. It will take some time for sites to update their players.

Comment 24

[johnk860]

Also cannot run brightcove player (Firefox 53.0.3 32 bit)  Ref:
http://ew.com/movies/2017/06/12/wonder-woman-impact-patty-jenkins-gal-gadot/?utm_campaign=entertainmentweekly&utm_source=facebook.com&utm_medium=social&xid=socialflow_facebook_entertainmentweekly

Comment 25

[Sergiu Logigan [:sergiu]]

Marking this issue as Resolved - Fixed.

We cannot reproduce the issue anymore on Google Pixel, Android 8.1.0 and Nightly 62.