As described in id3v2.3.0.txt (4.2.1, TCON), multiple genres are
only possible as references to ID3v1 genres with an optional
refinement as a text. When downgrading multiple genres from
ID3v2.4.0, they are now converted to numbers when possible and
the first genre text without ID3v1 reference is added as a
refinement. The keywords RX and CR are supported too.
This incorporates [6ca536b5] (mp4 properties: handle the case when
mp4 file header has zero bitrate) from PR #899 with a more accurate
bitrate calculation and a unit test.
Since the resolve only use the filename, there is no reason not to probe
them, as a filename might be available through the IOStream interface.
When using a ByteVectorStream, the filename will be empty and user
defined resolvers won't be probed.
When a WAV file with float format without a fact chunk containing
a totalSamples value is encountered, the bitrate is not calculated.
However, since all samples have the same number of bits, the
number of samples can be calculated from the size of the data chunk
and number of channels and bits per sample, as it is done in the
PCM case.
When the picture block is large and comes before the comment block,
Windows will no longer display the tags. This can be fixed by
storing the comment block before the picture block instead of
appending it at the end.
This will allow editing the tags of WAV files which have data
appended at the end. The corresponding unit test checks that the
original contents are still available after editing the metadata
of such files.
Fixes to properly handle WavPack files with leading and
trailing non-audio blocks, non-standard sampling rates,
and DSD audio (introduced in WavPack 5).
This creates symetry with ID3v2::Tag::comment() in preferring frames with no description
when choosing which COMM frame should be updated. (Previously setComment() simply updated
the first COMM frame.)
Fixes#950
In SV7 these are a mix of signed and unsigned shorts; in SV8 they're all
signed. Storing them as an int is fine for signed or unsigned shorts as
it's wide enough to contain either of them.
Unfortunately there are no explicit tests for SV7 at the moment; that
would be ideal to add before the next release.
CC @carewolf
I'd imagined it being useful for calls to `strip()`, but it's not
actually used there since that's an OR-ed together set of flags
representing which tags to strip.
