Bug 1285848 - Part 1: Implement Rice Delta Decoding. r=francois. MozReview-Commit-ID: 16T47WuwwYb

Bug 1304389 - Don't write session file if content hasn't changed; r?Yoric (THIS COMMIT CURRENTLY FAILS BECAUSE Components IS NOT AVAILABLE IN THE PROMISEWORKER. I'M NOT SURE HOW TO FIX IT.) I saw a HN post about Firefox chewing through I/O writing recovery.js every few seconds. I figured I'd take a stab in the dark at fixing it. This commit adds logic to the sessionstore worker to no-op writes if written content hasn't changed. It does so by maintaining a hash of the last written data. If the hash of data we're writing is identical to a previous write, we skip the write. The code doesn't achieve absolute minimum write I/O. For example, we could hash the file at read time to avoid the initial write. This optimization can be performed later, if wanted. I didn't do it because I was striving to keep the change as small as possible. For this change to have the desired effect to reduce I/O, an idle Firefox's session state must not change. If there are things like timestamps derived from Date.now() sneaking into the session data, this would undermine this optimization. This change also assumes that it is acceptable to not update the mtime of the session file. I don't see any code looking at the mtime, so I think this will be fine. MozReview-Commit-ID: 7FSkgmUdO8T

Bug 1304389 - Indent code performing session write; r?Yoric This will make the next change easier to read. MozReview-Commit-ID: 3RjrwEDK2zF

Bug 1304568: Fix some bad casts. r?drno MozReview-Commit-ID: XJhCywlTpg

Bug 1304983 - Improve test_safe_browsing_initial_download.py to wait for all safebrowsing files being downloaded. MozReview-Commit-ID: IQC9VObLSgi

Bug 1206232 - Add temporary permission states to SitePermissions.jsm. MozReview-Commit-ID: AyXyK5tMMRm

Bug 1275877 - Use ContentTask.spawn to prevent intermittent test failure;r=jlongster MozReview-Commit-ID: 72kyZBHtEma

Bug 1287058 - Supports SafeBrowsing v4 partial update. r=gcp MozReview-Commit-ID: A9ZNpD435Ab

Bug 1304729 - Followup to fix careless mistake. r?botond MozReview-Commit-ID: 5gsVh4JgroW

Bug 1304729 - When transitioning from a pinch to a pan, make sure we start the pan with the correct coordinates. r=botond MozReview-Commit-ID: CC4gPT2jubR

Bug 1304729 - Ensure we schedule a composite if we change the scroll position during a pinch gesture with no zoom change. r=botond MozReview-Commit-ID: AF2aoJZ1ssG

Bug 1301675 - Assert that a MediaStreamTrackSource is not stopped more than once. r?jib MozReview-Commit-ID: Dmp9TnIhmax

Bug 1301675 - Uncomment assertion for when removed live track was not found. r?jib It needs an extra guard for when we're an audio element and a live video track was removed. That's ok and we should ignore it. MozReview-Commit-ID: FVw3lDKd4oU

Bug 1301675 - Fix track cloning test. r?jib It relied on track.stop() not working for MediaStreamAudioDestinationNode tracks. MozReview-Commit-ID: 1E53JpzAiMs

Bug 1301675 - Rename BasicUnstoppableTrackSource to BasicTrackSource. r?jib They are not unstoppable any longer. We just don't forward Stop() to the real source. MozReview-Commit-ID: FdFccMsD3eb

Bug 1301675 - Received tracks from RTCPeerConnection are stoppable. r?jib MozReview-Commit-ID: L8ag0BEWL39

Bug 1301675 - Test that a track from MediaStreamAudioDestinationNode can be stopped. r?padenot MozReview-Commit-ID: 5wViWJQmI2H

Bug 1301675 - Implement AudioDestinationTrackSource. r?padenot So stop()ing the output track can clean up the destination node. MozReview-Commit-ID: CmMXRCpOeFY

Bug 1301675 - Clarify why we don't need to do anything on DecoderCaptureTrackSource::Stop(). r?jib MozReview-Commit-ID: H1zaO1FKWvK

Bug 1301675 - Implement StreamCaptureTrackSource::Stop. r?jib MozReview-Commit-ID: 2qwNba6bA80