1175006 - [EME] Reset state so we re-try creating a trial GMPVideoDecoder whenever Firefox updates

Status: RESOLVED FIXED

(Firefox :: General, defect)

Description

[Chris Pearce [:cpearce (Not reading bugmail)]]

Currently the first time Gecko tries to use EME it creates a 
"trial GMPVideoDecoder". If that fails, Gecko forevermore reports to content that EME doesn't work.

It could be the case that updating Firefox fixes the GMPVideoDecoderTrialCreator, and currently there's no way for us to recover from this other than a GMP update. So I think we should reset the prefs that GMPVideoDecoderTrialCreator sets whenever Firefox updates, so that we'll get another chance to make EME work.

Comment 1

[Chris Pearce [:cpearce (Not reading bugmail)]]

Attachment: Patch

Reset the gmp-trial-create pref when Firefox updates, so that if something's changed to make EME work or not work, we'll re-try instead of assuming the old result still holds.

Without this, clients that have tried to use EME and failed because it's broken with e10s at the moment will forever assume they can't use EME.

Comment 2

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/9c45565cdccc

Comment 3

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/9c45565cdccc

Comment 4

[Chris Pearce [:cpearce (Not reading bugmail)]]

Comment on attachment 8622843 [details] [diff] [review]
Patch

Approval Request Comment
[Feature/regressing bug #]: EME/bug 1169129

[User impact if declined]: This patch makes Firefox reset the state added in bug 1169129 every time it updates. Recall in bug 1169129 I'm adding code that tests that Adobe EME can decode video/audio before reporting to JS that Adobe EME works. If that code is wrong and we falsely report that we can't use Adobe EME when we actually can, then the user is stuck being unable to use EME forever. However with this patch, the next time we update Firefox, we'll try again, getting us out of the bad state.

[Describe test coverage new/current, TreeHerder]: Lots of EME mochitests and for the feature being touched here, though nothing for this specific case.

[Risks and why]: Low; if this goes wrong, at worst we won't cache the result of the "can Adobe EME decode?" check, meaning we'll just re-do it, which is very low cost.

[String/UUID change made/needed]: None.

Comment 5

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/releases/mozilla-aurora/rev/4ab88b90a70e