MSE: Report UseCounter and RAPPOR for two problematic MSE usages
This change adds UseCounter (for publicly visible % of Page Visits
statistics) and RAPPOR reporting to help us gauge the frequency of these
types of MSE API usages encountered in Chrome usage. Such data may
assist subsequent work to clarify the spec, improve our implementation,
and potentially deprecate support.
"KeyframeTimeGreaterThanDependant" reporting: (Bug 739931)
If nonkeyframe's PTS precedes the PTS of the keyframe necessary to
decode it, this (in ISO-BMFF terminology) is not SAP Type 1 (though
other bytestreams might also encounter this GOP structure, too.) Our
buffering mechanism should gracefully handle this type of GOP, but the
MSE spec is unclear for how to precisely report resulting buffered
ranges and handle buffering/playback of overlaps of these types of
streams.
"MuxedSequenceMode" reporting: (Bug 737757)
If a multitrack SourceBuffer is used in 'sequence' AppendMode, the spec
leads to frequently surprising and undesirable results, usually due to
automatic timestampOffset updates based on one track after a
discontinuity are applied to all tracks.
At least bug 739931 may impact current PTS/DTS compliance work tracked
by bug 718641.
BUG=739931,737757,718641
Change-Id: I4fabb4ae0b389c5ce2eecb361d4b67c6d4874b04
Reviewed-on: https://chromium-review.googlesource.com/567558
Reviewed-by: Rick Byers <rbyers@chromium.org>
Reviewed-by: Chrome Cunningham <chcunningham@chromium.org>
Reviewed-by: Alexei Svitkine (slow) <asvitkine@chromium.org>
Commit-Queue: Matthew Wolenetz <wolenetz@chromium.org>
Cr-Commit-Position: refs/heads/master@{#486512}22 files changed
