This patch moves the various Is{Format}Enabled methods to
nsMediaDecoder. These methods are all implemented on top of prefs, and
they logically belong to the decoder class, because it needs to know
what media types it should be able to decode. This also helps reduce
the dependency on nsHTMLMediaElement.h.
This patch simply moves the MetadataTags typedef to nsMediaDecoder.h to
reduce the dependencies on nsHTMLMediaElements.h. The rest of the
changes are made to make this compile.
Remember parsed comments from the OpusTags header in
an array member variable. Add a method to generate
the nsHTMLMediaElement::MetadataTags hash table from
those entries, performing the same validatation we
do for Vorbis.
This feature is tested by adding the existing opus
test file to gMetadataTests.
IsValidVorbisTagName is moved to a static method
on nsOggCodecState so it can be shared among the
subclasses and easily called externally.
MetadataTags generation for Vorbis streams is moved
to the same method name on nsVorbisState to allow
sharing of the parsing and validation code as much
as possible, since both formats use the same scheme.
It's arguable whether contrustuction better belongs to
nsOggReader or to the nsCodecState subclasses, since in
theory the various multiplexed streams interact in
determining the canonical tag list, but it is the
per-codec streams which contain the metadata in Ogg.
Bug 748144 switched our opus decoding to use the
opus_multistream api so we could handle surround
audio files. The version of opus in our tree at
the time didn't include extern "C" {} protection
in the corresponding header, so we had to provide
our own.
With the update to the opus 1.0.0 source release
in bug 790381, this oversight is corrected and
we can remove our work-around.
Verify that tag names are ASCII within the valid character
ranges and that values are UTF-8, skipping comments where
these checks fail.
As written, the check in IsValidVorbisTagName() for the
separator character isn't necessary, because we've already
split on the first occurance. However, it's better to run
the extra check in case the function is reused elsewhere.
The Opus audio format supports multichannel (surround) audio, but our initial implementation used a simpler API which only supported mono and stereo output.
To handle these files gracefully, this patch uses the multichannel api and downmixes the output, if possible, to stereo, since we don't currently support surround sound playback.
nsOggReader::RangeEndTime() relied on CRC values from valid ogg
pages to keep track of already-searched data when looking for a
end timestamp. This caused quadratic behaviour searching in a
section of a file with no valid pages.
Instead, remember where we last looked, and backoff immediately
if we go more than the maximum length of a page into previously
scanned data. This avoids searching data we've eliminated as
containing valid Ogg pages and lets us search backward with O(N)
instead.
Implements a media.mozGetMetadata() method returning a new javascript object whose properties are key value pairs respresenting metadata tags from the media resource. This data is available after readystate enters METADATA_LOADED.
Currently this is only implemented for Ogg Vorbis streams.
Media format metadata is parsed out by the media decoders. In the nsCodecStateMachine::ReadMetadata subclasses we fill in an nsDataHashtable pointer using the format-specifc api.
The hash pointer is passed up to the media element as part of the MetadataLoaded event.
The hash is deleted if the load is aborted. The audio metadata is also reset to zero (as in the constructor), resolving a todo comment.
We trim the initial few samples out of the opus decoder,
to give the output time to converge, and to correct for
the encoding delay. Encoders store the delay in the preskip
field of the Ogg encapsulation header.
The previous code to do this was a hack based on the granulepos
values and could fail on some inputs. Instead, keep a count
of how many samples we want to trip, and remove packet data
until that value matches the preskip value from the header.
The value is set to the preskip value from the header when
the decoder is initialized. We also need to do this after
seek. To do this we add a specialized nsOggReader::ResetDecode
method which takes a boolean argument, set to true when
we are seeking to the start of the stream. In that case,
the method resets the skip count.
There is still an issue after general seeks. The spec recommends
trimming a full 80 ms (3840 frames) to allow the decoder to fully
settle from the previous state. It's tricky to do this inside
nsOpusState because it doesn't know where it is in the stream.
Also add some debug output to track the decode behaviour.
