The EME draft spec changed this week to make MediaKeys.createSession() synchronous, and add an explicit MediaKeySession.generateRequest() method. This splits out the async part of the old MediaKeys.createSession() into its own function. Based on the minutes of the WG meeting, the change was well received, so we may as well make it sooner rather than later.

Nasty race condition, this used to be.

Comment on attachment 8488410 [details] [diff] [review] 1059043-split-createsession.patch Review of attachment 8488410 [details] [diff] [review]: ----------------------------------------------------------------- Looks good. ::: content/media/eme/MediaKeySession.h @@ +98,5 @@ > const nsString mKeySystem; > nsString mSessionId; > const SessionType mSessionType; > bool mIsClosed; > + bool mIsInitialized; Please make this mIsUninitialized and invert the true/false values, so that it matches the "uninitialized" flag in the spec. ::: content/media/eme/MediaKeys.cpp @@ +342,5 @@ > + NS_WARNING("Received activation for non-existent session!"); > + promise->MaybeReject(NS_ERROR_DOM_INVALID_ACCESS_ERR); > + return; > + } > + Assert that the session has a valid sessionId. It should be valid, but we'll be pretty stuffed if it isn't.

Comment on attachment 8488410 [details] [diff] [review] 1059043-split-createsession.patch Review of attachment 8488410 [details] [diff] [review]: ----------------------------------------------------------------- Forgot DOM peer review; yelled at by pre-push hook again.

Comment on attachment 8488410 [details] [diff] [review] 1059043-split-createsession.patch Ok, I was told that best of our best knowledge other implementations will change their MediaKeys implementation too.

Mass update firefox-status to track EME uplift.