When muxing, the AVStreams' side-data is typically set
by the caller before avformat_write_header();
it is not documented to be else. Yet the Matroska muxer
added an AVStereo3D side data if certain metadata
was present:
Since commit 4d686fb721
(adding support for AVStereo3D stream side-data),
the Matroska muxer checked certain stream tags that
contain Matroska's StereoMode and (if they are present)
converted this value into an AVStereo3D struct that
gets attached to the AVStream (reusing a function from
the demuxer). Afterwards the AVStereo3D side data struct
(whether it has just been added by the muxer or not) gets
parsed and converted back into a Matroska StereoMode.
Besides being an API violation this change broke
StereoMode values without a corresponding AVStereo3D
(namely the anaglyph ones).
This commit fixes this: A StereoMode given via tags
is now used-as-is; if no such tag exists and an AVStereo3D
side data exists, it is converted into the corresponding
StereoMode (if possible). This approach also fixes
handling of the anaglyph ones; the changes to the
matroska-stereo_mode are due to this.
The new STEREOMODE_STEREO3D_MAPPING has been put to
good use for this.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
It has undefined behaviour in case the value does not fit into an int.
Also stop allowing to override a stream level "alpha_mode" tag
by an AVFormatContext one and properly check that the stereo_mode
number given via a tag is actually in the range 0..14: Negative
values would have been treated as zero before this patch.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
When the marker writing code was merged from libav to FFmpeg
in dc62016c, it failed to take into account that the meaning of
cluster_pos had changed in bda5b662; in particular, the special
value for “I'm not currently working on a cluster” had changed
from 0 to -1. This makes the avio_write_marker() call never
be called. Update the if statement to fix it.
Fixes: Ticket9843
Signed-off-by: Steinar H. Gunderson <steinar+ffmpeg@gunderson.no>
Signed-off-by: Martin Storsjö <martin@martin.st>
Writing the duration SimpleTag is special: It's size is
reserved in advance via an EBML Void element (if seekable)
and this reserved space is overwritten when writing the trailer;
it does not use put_ebml_string().
The string to write is created via snprintf on a buffer
of size 20; this buffer is then written via put_ebml_binary()
with a size of 20.
EBML strings need not be zero-terminated; if not, they
are implicitly terminated by the element's length field.
snprintf() always zero-terminates the buffer, i.e.
the last byte can be discarded when using an EBML string.
This patch does this.
The FATE changes are as expected: One byte saved for every
track; the only exception is the matroska-qt-mode test:
An additional byte is saved because an additional byte
could be saved from the enclosing Tags length field.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
Do it only for video (the only thing for type for which HDR10+
makes sense).
This effectively reverts changes to several FATE ref-files
made in bda44f0f39.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
These two AVIOContexts currently coincide, but this is not
guaranteed to remain so (in fact, I have plans to write each
TrackEntry into its own AVIOContext).
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
It is only WebVTT which is special in WebM; hypothetical future
subtitle codecs in WebM will presumably use the ordinary code.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
The demuxer uses a extradata offset of 26, so we would need
to recreate the missing 26 bytes somehow in the muxer, but
we just don't. Remuxed files (like real/rv30.rm from the FATE-suite)
don't work due to missing extradata.
(The extradata offset also applies to RV40 and the extradata
is indeed lost upon remuxing, yet remuxing real/spygames-2MB.rmvb
works; our RV40 decoder does not use extradata at all.)
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
RV10 and RV20 are unsupported because creating the correct CodecPrivate
is unsupported (the demuxer uses a codecpriv_offset of 26, so one
would need to recreate the missing 26 bytes); COOK and SIPR are
unsupported, because Matroska uses a packetization mode that is
different from what FFmpeg uses in its packets (see
matroska_parse_rm_audio() in the demuxer).
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
bda44f0f39 added code that
potentially added another BlockMore master and BlockAdditional
data as well as BlockAddID number, yet it bumped the number
of EBML elements by four instead of only three.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
Matroska supports orthogonal transformations (both pure rotations
as well as reflections) via its 3D-projection elements, namely
ProjectionPoseYaw (for a horizontal reflection) as well as
ProjectionPoseRoll (for rotations). This commit adds support
for this.
Support for this in the demuxer has been added in
937bb6bbc1 and
the sample used in the matroska-dovi-write-config8 FATE-test
includes a displaymatrix indicating a rotation which is now
properly written and read, thereby providing coverage for
the relevant code in the muxer as well as the demuxer.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
This commit does for AVOutputFormat what commit
20f9727018 did for AVCodec:
It adds a new type FFOutputFormat, moves all the internals
of AVOutputFormat to it and adds a now reduced AVOutputFormat
as first member.
This does not affect/improve extensibility of both public
or private fields for muxers (it is still a mess due to lavd).
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
Signed-off-by: Anton Khirnov <anton@khirnov.net>
The general demuxing API uses parsers and decoders. Therefore
FFStream contains pointers to AVCodecContexts and
AVCodecParserContext and lavf/internal.h includes lavc/avcodec.h.
Yet actually only a few files files really use these; and it is best
when this number stays small. Therefore this commit uses opaque
structs in lavf/internal.h for these contexts and stops including
avcodec.h.
This also avoids including lavc/codec_desc.h implicitly. All other
headers are implicitly included as now (mostly through codec.h).
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
The field is not specific to Opus.
The mp2fixed encoder signals initial_padding and is used
by both the matroska-encoding-delay test as well as
the lavf-mkv tests which necessitated several FATE ref changes.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
Matroska generally requires timestamps to be nonnegative, but
there is an exception: Data that corresponds to encoder delay
and is not supposed to be output anyway can have a negative
timestamp. This is achieved by using the CodecDelay header
field: The demuxer has to subtract this value from the raw
(nonnegative) timestamps of the corresponding track.
Therefore the muxer has to add this value first to write
this raw timestamp.
Support for writing CodecDelay has been added in FFmpeg commit
d92b1b1bab and in Libav commit
a1aa37dd0b. The former simply
wrote the header field and did not apply any timestamp offsets,
leading to desynchronisation (if one uses multiple tracks).
The latter applied it at two places, but not at the one where
it actually matters, namely in mkv_write_block(), leading to
the same desynchronisation as with the former commit. It furthermore
used the wrong stream timebase to convert the delay to the
stream's timebase, as the conversion used the timebase from
before avpriv_set_pts_info().
When the latter was merged in 82e4f39883,
it was only done in a deactivated state that still did not
offset the timestamps when muxing due to "assertion failures
and av sync errors". a1aa37dd0b
made it definitely more likely to run into assertion failures
(namely if the relative block timestamp doesn't fit into an int16_t).
Yet all of the above issues have been fixed (in commits
962d631573,
5d3953a5dc and
4ebeab15b0. This commit therefore
enables applying CodecDelay, fixing ticket #7182.
There is just one slight regression from this: If one has input
with encoder delay where the first timestamp is negative, but
the pts of the part of the data that is actually intended to be
output is nonnegative, then the timestamps will currently by default
be shifted to make them nonnegative before they reach the muxer;
the muxer will then ensure that the shifted timestamps are retained.
Before this commit, the muxer did not ensure this; instead the
timestamps that the demuxer will output were shifted and
if the first timestamp of the actually intended output was zero
before shifting, then this unintentional shift just cancels
the shift performed before the packet reached the muxer.
(But notice that this only applies if all the tracks use the same
CodecDelay, or the relative sync between tracks will be impaired.)
This happens in the matroska-opus-remux and matroska-ogg-opus-remux
FATE tests. Future commits will forward the information that
the Matroska muxer has a limited capability to handle negative
timestamps so that the shifting in libavformat can take advantage
of it.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
Opus can be decoded to multiple samplerates (namely 48kHz, 24KHz,
16Khz, 12 KHz and 8Khz); libopus as well as our encoder wrapper
support these sample rates. The OpusHead contains a field for
this original samplerate. Yet the pre-skip (and the granule-position
in the Ogg-Opus mapping in general) are always in the 48KHz clock,
irrespective of the original sample rate.
Before commit c3c22bee63, our libopus
encoder was buggy: It did not account for the fact that the pre-skip
field is always according to a 48kHz clock and wrote a too small
value in case one uses the encoder with a sample rate other than 48kHz;
this discrepancy between CodecDelay and OpusHead led to Firefox
rejecting such streams.
In order to account for that, said commit made the muxer always use
48kHz instead of the actual sample rate to convert the initial_padding
(in samples in the stream's sample rate) to ns. This meant that both
fields are now off by the same factor, so Firefox was happy.
Then commit f4bdeddc3c fixed the issue
in libopusenc; so the OpusHead is correct, but the CodecDelay is
still off*. This commit fixes this by effectively reverting
c3c22bee63.
*: Firefox seems to no longer abort when CodecDelay and OpusHead
are off.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
It is possible for the trailing padding to be zero, namely
e.g. if the AV_PKT_DATA_SKIP_SAMPLES side data is used
for leading padding. Matroska supports this (use a negative
DiscardPadding), but players do not; at least Firefox refuses
to play such a file. So for now only write DiscardPadding
if it is trailing padding and nonzero.
The fate-matroska-ogg-opus-remux was affected by this.
(I wish CodecDelay would not exist and DiscardPadding would
be used to instead trim the codec delay away (with the Block
timestamp corresponding to the time at which the actually
output audio is output).)
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
Regression since 67eea6cf02.
Affects only WebVTT when muxing WebM. (This is covered
by the webm-webvtt-remux FATE test which fails for several
FATE boxes on fate-ffmpeg.org.)
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
Up until now, only the first four bytes (the ones preceding
the OBU) were written because not enough space has been reserved
for the complete CodecPrivate. This commit changes this
by increasing the space reserved for the CodecPrivate (it is big
enough for every sane sequence header plus something extra);
the code falls back to writing four bytes in case the increased
space turns out to be insufficient.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
Up until now, updating extradata was very ad-hoc: The amount of
space reserved for extradata was not recorded when writing the
header; instead the AAC code simply presumed that it was enough.
This commit changes this by recording how much space is available.
This brings with it that the code for writing of and reserving space
for the CodecPrivate and for updating it diverges. They are therefore
split; this allows to put other common tasks like seeking to
right offset as well as writing padding (in case the new extradata did
not fill the whole reserved space) to this common function.
The code for filling up the reserved space is smarter than the code
it replaces; therefore it is no longer necessary to reserve more
than necessary just to be sure that one can add an EBML Void element
(whose minimum size is two) lateron. This is the reason for the change
to the aac-autobsf-adtstoasc test.
Reviewed-by: Michael Niedermayer <michael@niedermayer.cc>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
Instead pass extradata and extradata_size explicitly.
(It is not perfect, as ff_put_(wav|bmp)_header() still uses
the extradata embedded in codecpar, but this is not an issue
as long as their CodecPrivate isn't updated.)
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
This is in preparation for splitting writing and updating
extradata more thoroughly later.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
Because not all metadata is written as tags, the Matroska muxer
filters out the tags that are not written as tags.
Therefore the code first checks whether a Tag master element
needs to be opened for a given stream/chapter/attachment/global
metadata. If the answer turns out to be yes, it is checked again
whether a given AVDictionaryEntry is written as a tag.
This commit changes this: The Tag element is opened unconditionally
and in case it turns out that it was unneeded, it is discarded again.
This is possible because the Tag element is written into its own
dynamic buffer.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
This is possible by using a dynamic buffer to write them;
said dynamic buffer is (re)used and reset as appropriate.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
It is no longer converted since mkv_write_chapters() is called
before mkv_write_tags() which happens since commit
4ebfc13c33. Given the fact that
chapters can also be written late, mkv_write_chapters() has to
convert the metadata itself.
Fixes ticket #9812.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
Before this patch the muxer writes an invalid file
(namely one in which the Projection master is a child of
the Colour element) if the following conditions are met:
a) The stream contains AVMasteringDisplayMetadata without primaries
and luminance (i.e. useless AVMasteringDisplayMetadata).
b) The stream contains AV_PKT_DATA_SPHERICAL side data.
c) All the colour elements of the stream are equal to default
(i.e. unknown).
Fortunately these conditions are very unlikely to be met.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
Add a parameter to omit seq header when generating the av1C atom.
For now, this does not change any behavior. This will be used by a
follow-up patch to add AVIF support.
Signed-off-by: Vignesh Venkatasubramanian <vigneshv@google.com>
This avoids unnecessary rebuilds of most source files if only the
list of enabled components has changed, but not the other properties
of the build, set in config.h.
Signed-off-by: Martin Storsjö <martin@martin.st>
According to the documentation, the ISOBMFF 'equi' box must
be present for equirectangular projections.
Reviewed-by: Hendrik Leppkes <h.leppkes@gmail.com>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
