4 years ago
4 years ago


(Reporter: rillian, Assigned: rillian)


(Blocks 1 bug)

42 Branch
Dependency tree / graph

Firefox Tracking Flags

(firefox42 affected)



(2 attachments, 1 obsolete attachment)

Per consensus at yesterday's meeting, we want to reduce scope by disabling MSE webm playback in Firefox 42. It's not yet demonstrably stable, and we're running out of time for the 42 aurora branch. This feature can ride the train from 43 nightly.
WebM hasn't never enabled for 42. The code looks the same as it did when we turned it off in bug 1097436 last November.

So this bug is about making sure it stays off while we uplift other MSE changes from 43, where it is enabled for linux-only on 43. See bug 1097436 and bug 1190970.
Posted patch Disable MSE WebM in 42 (obsolete) — Splinter Review
Here's a patch on top of the uplift stack from bug 1197083 to turn MSE WebM back off for linux.

I think the expedient thing to do is to land the full stack of MSE changes on Aurora 42, including WebM-specific ones, to reduce variance. That makes it easier to backport further fixes, and to change our mind later.
Attachment #8656305 - Flags: review?(ajones)
Attachment #8656305 - Flags: review?(ajones) → review+
I don't believe this is the right approach as you'll be able to create a mediasource object but unable to attach any sourcebuffer object to it.
The easiest is simply not to uplift the commit where mediasource was made available on all platforms.
Blocks: 1197083
(In reply to [PTO Until 02/Sep] Jean-Yves Avenard [:jya] from comment #4)
> I don't believe this is the right approach as you'll be able to create a
> mediasource object but unable to attach any sourcebuffer object to it.

I don't understand your objection. Is mediasource not properly pref-protected? How is this different from what's in 43?
Flags: needinfo?(jyavenard)
Comment on attachment 8656305 [details] [diff] [review]
Disable MSE WebM in 42

Approval Request Comment
[Feature/regressing bug #]: MSE
[User impact if declined]: Linux users will see possibly-unstable MSE WebM from YouTube instead of more-stable but possibly-unsupported mp4 or flash.
[Describe test coverage new/current, TreeHerder]: Green on try; we haven't tested this configuration since 41.
[Risks and why]: Risk is moderate. I don't understand jya's comment, and this is a configuration we haven't tested on 43. However we want to try turning this off, and better to get data earlier in the cycle.
[String/UUID change made/needed]: None.
Attachment #8656305 - Flags: approval-mozilla-aurora?
Comment on attachment 8656305 [details] [diff] [review]
Disable MSE WebM in 42

ok, let's disable it then.
Attachment #8656305 - Flags: approval-mozilla-aurora? → approval-mozilla-aurora+
Not sure what could be confusing.
Mediasource *was* pref protected, and on top of that we had a pref to enable MP4 and a pref to enable webm.

So on Linux, you couldn't use mediasource at all. Attempting to create a mediasource object on Linux would have thrown an error. This has always been the case until now.

In 43 however, the pref to enable mediasource was removed and on Linux and other Unix OS where h264/AAC wasn't available webm was enabled by default and the other systems MP4 was enabled by default.

That change ( was uplifted.

What you've done in the patch however is simply to disable webm once again while allowing mediasource.

So now, on Linux, you'll be able to create a mediasource object but you won't be able to do anything with it. You won't be able to create a source buffer.

IMHO this is bad. Bug 1190970 should have simply not be uplifted and now should be reverted so we don't end up with a half baked unusable mediasource on Linux.
Flags: needinfo?(jyavenard)
Ok, I see what you mean. Thanks. Here's an updated version which also turns on mediasource on Linux.
Attachment #8656305 - Attachment is obsolete: true
Attachment #8657379 - Flags: review?(ajones)
Attachment #8657379 - Flags: feedback?(jyavenard)
Comment on attachment 8657379 [details] [diff] [review]
Disable MSE WebM in 42 v2

Review of attachment 8657379 [details] [diff] [review]:

You will want to also revert the 2nd part of bug 1190970, otherwise you will fail some mochitests.
Attachment #8657379 - Flags: feedback?(jyavenard) → feedback+
Attachment #8657379 - Flags: review?(ajones) → review+
Thanks. I think we don't want to revert bug 1190970 but only the 'linux: false' clauses. Otherwise the test will start failing when 42 goes to beta.
Attachment #8658317 - Flags: review?(jyavenard)
Comment on attachment 8658317 [details] [diff] [review]
Update interface mochitest

Attachment #8658317 - Flags: review?(jyavenard) → review+
Carrying forward a=sylvestre on the disable patch.
Ralph, how is that any different to reverting bug 1190970? Looks the same to me has:

    1.11  // IMPORTANT: Do not change this list without review from a DOM peer!
    1.12 -    {name: "MediaSource", linux: false, release: false},
    1.13 +    "MediaSource", has:

    1.12 -    "MediaSource",
    1.13 +    {name: "MediaSource", linux: false},

Your original patch removed the linux and release guards at the same time.
You need to log in before you can comment on or make changes to this bug.