When a new time segment is received upstream is going to restart
with a new atom. Make the neededbytes and todrop variables
reflect that to avoid waiting too much or dropping the
initial bytes that contain the header.
The adapter might have data remaining from the previous segment,
push it all before clearing the adapter and starting a new segment.
It can accumulate data if it had pushed and got not-linked, returning
immediately without processing all the data. Before starting a new
segment this data should be handled.
Avoids accumulating all samples from a fragmented stream that could
lead to a 'index-too-big' error once it goes over 50MB of data. It
could reach that before 2h of playback so it doesn't take that long.
As upstream elements are providing data in time format they should
be the ones that have more information about the full media index
and should be able to seek if possible.
upstream_newsegment isn't really clear on what it means, it is set
to TRUE when the upstream element sends a segment in TIME format, so
rename it to be more clear about it.
It is important to know this because it means that upstream has
a notion of time and qtdemux is likely being driven by an upstream
element that is reading from a higher level abstraction than a file,
such as a DASH, MSS or DLNA element.
In fragmented streaming, multiple moov/moof will be parsed and their
previously stored samples array might leak when new values are parsed.
The parse_trak and callees won't free the previously stored values
before parsing the new ones.
In step-by-step, this is what happens:
1) initial moov is parsed, traks as well, streams are created. The
trak doesn't contain samples because they are in the moof's trun
boxes. n_samples is set to 0 while parsing the trak and the samples
array is still NULL.
2) moofs are parsed, and their trun boxes will increase n_samples and
create/extend the samples array
3) At some point a new moov might be sent (bitrate switching, for example)
and parsing the trak will overwrite n_samples with the values from
this trak. If the n_samples is set to 0 qtdemux will assume that
the samples array is NULL and will leak it when a new one is
created for the subsequent moofs.
This patch makes qtdemux properly free previous sample data before
creating new ones and adds an assert to catch future occurrences of
this issue when the code changes.
Most files don't contain the values for transposing the coordinates
back to the positive quadrant so qtdemux was ignoring the rotation
tag. To be able to properly handle those files qtdemux will also ignore
the transposing values to only detect the rotation using the values
abde from the transformation matrix:
[a b c]
[d e f]
[g h i]
https://bugzilla.gnome.org/show_bug.cgi?id=738681
The media start has nothing to do with the shift we have applied
but with the value of the first PTS. This is defined as:
Dt(0) = 0
Ct(0) = Dt(0) + CTTS(0)
So the media start is always the first CTTS.
https://bugzilla.gnome.org/show_bug.cgi?id=751361
Allows playing edts editted files with proper synchronization of
streams. This patch fixes the regression introduced by
bf95f93c01 that was added to fix
segment seeks handling.
Having the accumulated_base separated from the main segment.base
allows handling both segment seeks and edts editted files.
https://bugzilla.gnome.org/show_bug.cgi?id=751361
Buffers need not to start at running-time 0 so the last_dts needs
to be the value of the first buffer's dts as it is used to compute
the duration of the buffers. If it was left at 0 the first buffer
would have a larger duration when it shouldn't
https://bugzilla.gnome.org/show_bug.cgi?id=751361
Move the multiview caps calculations to the configure_stream()
function, so the rest of the video info is available, and
use the gst_video_multiview_guess_half_aspect() function to
determine if the half-aspect flag should be set on frame-packed
video.
The cslg atom provide information about the DTS shift. This is
needed in recent version of ctts atom where the offset can be
negative. When cslg is missing, we parse the CTTS table as proposed
in the spec to calculate these values.
In this implementation, we only need to know the shift. As GStreamer
cannot transport negative timestamps, we shift the timestamps forward
using that value and adapt the segment to compensate. This patch also
removes bogus offset of ctts_soffset, this offset shall be included
in the edit list.
https://bugzilla.gnome.org/show_bug.cgi?id=751103
We shift DTS forward to avoid negative timestamps which cannot be
represented with version 0 of the CTTS table. To stick with that
version (backward compatibility), the spec recommend using an
edit list entry to move back the presentation time to where it
should be.
https://bugzilla.gnome.org/show_bug.cgi?id=751242
In presence of a CTTS, the segment start/stop must be offset so
the segment start/stop include the PTS. This is needed since the
PTS cannot be negative in this format. This fixes issues where the
running time of the first buffer isn't at the start.
https://bugzilla.gnome.org/show_bug.cgi?id=740575
Remove a custom specialized version of gst_buffer_new_wrapped by
using gst_buffer_new_wrapped_full inside a macro to simplify
parameters and give it a more meaningful name.
It is only used to create temporary buffers to have its data copied.
Adds AC-3 muxing support. It is defined for mp4 and 3gp formats.
One extra feature that was added was the ability to add extension
atoms after set_caps as the AC-3 extension atom needs some data
that has to be extracted from the stream itself and is not
present on caps.
The MPEG-A format provides an extension to the ISO base media
file format to store stereoscopic content encoded with different
codecs like H.264 and MPEG-4:2. The stereo video media information(svmi)
atom declares the presence and storage method for the video.
Stereo video information for MPEG-A can also be supplied through
the 'stvi' atom (ref: ISO/IEC_14496-12, ISO/IEC_23000-11), which
is not implemented in this patch.
Also missing is support for stereo video encoded as separate video tracks
for now.
Based on a patch by Sreerenj Balachandran <sreerenj.balachandran@intel.com>
https://bugzilla.gnome.org/show_bug.cgi?id=611157
When performing seek, segment->start is being updated with desired_offset,
but in case of reverse playback segment->start should be 0 and
segment->stop should be updated with desired offset.
https://bugzilla.gnome.org/show_bug.cgi?id=750675
Only update the moov header into the caps if it's the finalised
moov at EOS time. Avoids posting a bogus moov at startup and
repeated updates in robust-recording mode
Implement a robust recording mode, where the output
file is always in a playable state, seeking and rewriting
the moov header at a configurable interval. Rewriting
moov is done using reserved space at the start of
the file, and a ping-pong strategy where the moov
is replaced atomically so it's never invalid.
Track when tags have actually changed, and don't write them into
the moov unless they've changed. Clear any existing tags when
re-writing them, so we can do progressive moov updating in robust
recording mode.
Write placeholder mdat as a free atom plus a 32-bit mdat
with '0' size, which means "rest of the file" in the spec.
Re-write it later to a full 64-bit extended size atom if needed.
Correctly update any edit lists each time the moov is recalculated,
updating existing table entries if they already exist instead of just
adding new ones.
qtdemux creates a samples array and gets the timestamps for buffers by
accumulating their durations. When doing reverse playback of fragments,
accumulating samples will lead to wrong timestamps as the timestamps
should go decreasing from fragment to fragment and the accumulation
will produce wrong results.
In this case, when receiving a discont for fragmented reverse playback,
the previous samples information should be flushed before new data
is processed.
The gst-launch script for example launch line to test qtdemux is
missing a queue before the decodebins, otherwise the gst-launch-1.0
command won't work.
https://bugzilla.gnome.org/show_bug.cgi?id=749054
When doing key uint seek, qtdemux calls gst_qtdemux_adjust_seek
to get proper offset. And then this offset is set to
segment.position and segment.time in gst_qtdemux_perform_seek but
segment.start is not updated.
After that, application sends segment query,
qtdemux sets start and stop to query using gst_segment_to_stream_time. Due
to the wrong value in segment.start, the stop position is smaller than
it should.
https://bugzilla.gnome.org/show_bug.cgi?id=746822
We always write the CTTS in qtmux. Ideally we only want to do that
for streams that need DTS, it should be present on the track information
rather than be decided based on each buffer
As qt uses durations, it doesn't matter, only the difference
between consecutive buffers is important. Also, collectpads
already replaces PTS/DTS with the running times for them.
Instead of checking various state variables around the muxer,
track the current muxing mode in a single 'mux_mode' enum.
Add some implementation notes about the different mux modes
gst_segment_do_seek() does that for us already, and doing it twice
will break non-flushing seeks in interesting ways. Leftover from 1.0
porting.
Also copy over segment offset and applied_rate, just in case.
When not in fast-start or fragmented mode, we need to be able
to rewrite the size of the mdat atom, or else the output just
won't be playable - the mdat placeholder with size == 0 will
cover the rest of the file, including any moov atom we write out.
https://bugzilla.gnome.org/show_bug.cgi?id=708808
Unlike many other seek flags, the KEY_UNIT seek
flag is not copied over into the GstSegment,
since it's only relevant for the seek itself,
so we need to pass it explicitly to the seek
handler here.
https://bugzilla.gnome.org/show_bug.cgi?id=745339
We need different symbol names, because these symbols are also present
in the fragmented plugin ... which will cause conflicts when doing
static linking
Using the sparse streams can make the push-based seeking return
too far in the stream. It also can lead to issues as the
sparse streams will be ignored when restarting playback and,
if the sparse stream is the one that has the earliest sample,
it will confuse qtdemux's offsets as one stream will have
an earlier offset than the demuxer's one which might lead to
early EOS.
https://bugzilla.gnome.org/show_bug.cgi?id=742661
Parse the 'sidx' atom and update the total duration according to the
parser result. The isoff parser code is imported from
gst-plugins-bad's dashdemux and a gst_isoff_sidx_parser_add_data()
function was factored out of the gst_isoff_sidx_parser_add_buffer()
function.
https://bugzilla.gnome.org/show_bug.cgi?id=743578
Keep global and stream tags separately and parse the udta node
that can be found under the trak atom. The udta will contain
stream specific tags and will be pushed as such
https://bugzilla.gnome.org/show_bug.cgi?id=692473
Tags received via events, when marked as stream tags, will
be stored on that stream's trak atom instead of being stored
in the main tags atom. This allows the resulting file to have
global and stream tags stored.
https://bugzilla.gnome.org/show_bug.cgi?id=692473
Refactor the functions that were bound to the 'moov' atom to
directly pass the desired 'udta' that should receive the tags.
This allows the tags to be written to 'udta' at the 'moov' or
the 'trak' level, creating tags that are for the container or
for a stream only.
https://bugzilla.gnome.org/show_bug.cgi?id=692473
For fragmented streams with extra data at the end of the mdat
qtdemux was not dropping those bytes and would try to use
that extra data as the beginning of a new atom, causing the
stream to fail.
https://bugzilla.gnome.org/show_bug.cgi?id=743407
Actually copy the codec data instead of copying nothing
and then bombing out because there's no data.
Fixes: gst-launch-1.0 audiotestsrc ! avenc_alac ! qtmux ! fakesink
https://bugzilla.gnome.org/show_bug.cgi?id=741783
When dealing with fragmented files, we will get more accurate duration
information via the mfra and moof atoms.
In order for playback to not stop at the initial duration (from the
moov atom), we need to check and update the various duration variables
when we find more information.
Fixes playback of fragmented files in pull mode
When seeking or finding the previous keyframe, do
comparisons against targets and segments using composition time
to correctly decide which sample times match.
Currently during header parsing, we scan through the entire file
and skip every moof+mdat chunk for fragmented mp4s, which makes
start-up incredibly slow. Instead, just stop at the first moof
chunk when have a moov, and start exposing the streams, so we
can go and start handling the moofs for real.
Empty segments in an edit list have a media_start time of -1,
as they don't actually play any media. Allow for that when
aligning to the reference stream in reverse play.
DTS delta is used to calculate sample duration. If buffer has missing DTS, we take either segment start or previous buffer end time, whichever is later.
This must only be done for non sparse streams, sparse streams can have gaps between buffers (which is handled later by adding extra empty buffer with duration that fills the gap)
https://bugzilla.gnome.org/show_bug.cgi?id=737095
The old default timescale of 1 millisecond produces irrational
numbers for a lot of framerate/audio-packet-duration multiples.
1/1800 is a nicer number, as it tends to produce better fractions
and therefore slightly higher accuracy overall
Change the way the output framerate is calculated
to ignore the first sample (which is sometimes truncated
in my testing) and use the new gst_video_guess_framerate()
function to recognise common standard framerates better.
Remove the code that was sorting the first 20 sample
durations and then ignoring the result.
This makes sense in DASH reverse playback, where the upstream dashdemux
will download DASH segments in reverse order, but push their buffers
forward to qtdemux and mark each segment start as DISCONT. This needs
to be forwarded downstream to the parser/decoder, otherwise it won't work.
https://bugzilla.gnome.org/show_bug.cgi?id=734443
When writing out a trak with an edit list, make sure the
overall file duration is also updated to reflect the
lengthening of the stream.
Add some more debug to qtdemux to warn about streams that
are longer than the file and get truncated.
Handle the transformation matrix cases where there are only simple rotations
(90, 180 or 270 degrees) and use a tag for those cases. This is a common scenario
when recording with mobile devices
https://bugzilla.gnome.org/show_bug.cgi?id=679522
They are very confusing for people, and more often than not
also just not very accurate. Seeing 'last reviewed: 2005' in
your docs is not very confidence-inspiring. Let's just remove
those comments.
It was used in the past in 0.10 when there was no explicit DTS
field in buffers, now we have it in 1.x series and we can
check it directly with GST_BUFFER_DTS_IS_VALID
Do not try to use subsequent buffer timestamps to calculate
sparse streams durations because the stream is sparse and
the buffers might not be 'time adjacent'. So rely on the
duration and give the option to the pad to provide
custom 'empty' buffers to represent the gaps in the
stream, this can vary on how the data is represented.
Right now, the only sparse stream supported is tx3g subtitles.
Make sure empty segments are used and pushed with a gap event
to represent its data (or lack of it)
Each QtSegment is mapped into a GstSegment with the corresponding
media range. For empty QtSegments a gap event is pushed instead
of GstBuffers and it advances to the next QtSegment.
To make this work with seeks, need to keep track of the starting
'base' to make sure it remains consistently increasing when
pushing new segment events.
For example: if a seek makes qtdemux start from 5s, the first
segment will have a base=0. When the next segment is activated,
its base time will be QtSegment.time - qtdemux.segment_base so
that it doesn't include the first 5s that weren't played and
shouldn't be accounted on the running time
This purposedly will remove the fix made for
https://bugzilla.gnome.org/show_bug.cgi?id=700264, at this
point it was decided to respect the gaps, even if they cause
a delay on playback, because that's the way the file was crafted.
https://bugzilla.gnome.org/show_bug.cgi?id=345830
Mov spec says it uses a pascal style string, while isomedia uses
a null terminated one. Store the current atoms flavor into the HDLR
to be able to generate the correct output.
https://bugzilla.gnome.org/show_bug.cgi?id=705982
Make it clear what should be handled purely by mss mode:
1) Expose the streams on the first moof as there are no moov atoms
2) Properly cleanup streams on flushes
Add a note about the meaning of upstream_newsegment and mss_mode
for future reference.
Make all other special fragment handling shared for both dash
and mss streams.
In a fragmented scenario, qtdemux is operating in push mode
and it gets a fragmented buffer. While processing its data
downstream gets unlinked (or a input-selector changes its
active pad and returns not-linked). Qtdemux stops processing
this fragment and returns not-linked upstream, leaving the
remaining data in its adapter.
When it gets an EOS it should make sure that all the data it
had received is pushed before pushing EOS.
Some buffers can have multiple moov atoms inside and the strategy
of using the gst_adapter_prev_pts timestamp to get the base timestamp
for the media of the fragment would fail as it would reuse the same
base timestamp for all moofs in the buffer instead of accumulating
the durations for all of them.
Heres a better explanation of the issue:
qtdemux receives a buffer where PTS(buf) = X
buf -> moofA | moofB | moofC
The problem was that PTS(buf) was used as the base timestamp for
all 3 moofs, causing all buffers to be X based. In this case we want
only moofA to be X based as it is what the PTS on buf means, and the
other moofB and moofC just use the accumulated timestamp from the
previous moofs durations.
To solve this, this patch uses gst_adapter_prev_pts distance
result, this allows qtdemux to calculate if it should use the
resulting pts or just accumulate the samples as it can identify
if the moofs belong to the same upstream buffer or not.
https://bugzilla.gnome.org/show_bug.cgi?id=719783
In SmoothStreaming fragmented scenario, the timestamps are calculated
starting from the fragment buffer timestamp. When there is a not-linked
return from downstream, qtdemux will return upstream and will keep the
non-pushed data into its adapter.
On a new fragment buffer pushed to qtdemux, the new buffer timestamp
would overwrite the previous one that should be used on the still
to be pushed buffers. Because of this, this patch will also
update the fragment_start timestamp from the adapter last pts
to make sure the moof and timestamps are in sync and will result
in correct timestamps for all fragments.
In the scenario of "mdat | moov (with fragmented artifacts)" qtdemux
could read the moov again after the mdat because it was considering the
media as a fragmented one.
To avoid this loop this patch makes it store
the last processed moov_offset to avoid parsing it again.
And it also checks if there are any samples to play before
resturning to the mdat, so that it knows there is new data to be played.
https://bugzilla.gnome.org/show_bug.cgi?id=691570
When parsing a trak only free streams on failures if those streams
were created locally. They could have been created from a previous
fragment, in this case we they have valid info from the other fragment.
Including pads.
https://bugzilla.gnome.org/show_bug.cgi?id=691570
As for text subtitles and as suggested in #712643, throw
away the 2 byte terminator packets that some encoders insert.
This will make things better when remuxing and causes generation
of gap events.
Clean up the handling of mp4s streams. Use the generic esds
descriptor function to extract the palette, instead of hard coding
a wrong magic offset.
Add some more size safety checks when parsing ES descriptors, and
replace magic numbers with the descriptive constants that are already
defined.
Enhance dump output for stsd atoms.
Streams from both bug 712643 and historic bug 568278 now both work
correctly.
Fixes: #712643
Assume a file with atoms in the following order: moov, mdat, moof,
mdat, moof ...
The first moov usually doesn't contain any sample entries atoms (or
they are all set to 0 length), because the real samples are signaled
at the moofs. In push mode, qtdemux parses the moov and then finds the mdat,
but then it has 0 entries and assumes it is EOS.
This patch makes it continue parsing in case it is a fragmented file so that
it might find the moofs and play the media.
https://bugzilla.gnome.org/show_bug.cgi?id=710623
In push mode, when qtdemux can't use a seek to skip the mdat buffer it has
to buffer it for later use.
The issue is that after parsing the next moov/moof, there might be some
trailing bytes from the next atom in the file. This data was being discarded
along with the already parsed moov/moof and playback would fail to continue
after the contents of this moov/moof are played.
This is particularly bad on fragmented files that have the mdat before the
corresponding moof. So you'd get:
mdat|moof|mdat|moof ...
When a moof was received, it usually came with some extra bytes that would
belong to the next mdat (because upstream doesn't care about atoms alignment).
So those bytes were being discarded and playback would fail.
This patch makes qtdemux store those extra bytes to reuse them later after the
mdat is emptied.
https://bugzilla.gnome.org/show_bug.cgi?id=710623
Doing so would be a regression over 1.0 and breaks the unit test.
However the result will be most likely unusable, so let's post
a warning message on the bus.
Use g_date_time seconds manipulation to allow to cover the quicktime
spec for creation_time. It uses seconds since 1904.
Both paths could be done using the generic approach of seconds since
1904 with GDateTime handling, but the first path using seconds from
1970 should be more commonly found and avoids a few objects creation and
ref/unref, so keep it there for performance.
Additionally, the code for handling seconds since 1970 changed from >
to >= because having 0 seconds since 1970 is also a valid case for that
path to handle.
https://bugzilla.gnome.org/show_bug.cgi?id=707975
The streamable property only make sense for fragmented formats.
For regular MP4, when downstream is not seekable we can't rewrite
the headers, so qtmux can only work with fast-start=TRUE, where
the headers are written finishing the file.
For fragmented MP4, when streamable is not seekable and the streamable
property is FALSE, we must enforce streamable=TRUE warning the user
about this change
https://bugzilla.gnome.org/show_bug.cgi?id=707242
The most common use case for fragmented MP4 (Dash and Smooth Streaming)
is producing streamable content (even for VOD). streamable=FALSE would only
be used to generate fragmented MP4 with and index of MOOF's that could
be reproduced without a playlist/manifest
https://bugzilla.gnome.org/show_bug.cgi?id=707242
Check for GST_SEEK_TYPE_NONE for stop poistion and only update
the stop time if it is requested. Otherwise just maintain whatever
was stored at the segment
https://bugzilla.gnome.org/show_bug.cgi?id=707530
When the segment has a defined stop position, qtdemux should check
when streams reach this position and mark those as EOS. When all
streams are EOS it will return GST_FLOW_EOS to upstream to allow
the pipeline to finish instead of continuously consume buffers
from upstream that are not useful for the segment.
https://bugzilla.gnome.org/show_bug.cgi?id=707530
When handling seeks in push mode, qtdemux converts the seek to bytes
and pushes upstream. It needs to keep track of the seek and the
subsequent segment to be able to map them back to the requested
seek time and properly preserve the segment stop of the seek.
This is done by using the start offset in bytes of the seek,
that should be the same of the segment from upstream. And this
is also backwards compatible with what qtdemux already was using.
https://bugzilla.gnome.org/show_bug.cgi?id=707530
Amendment 2 of ISO/IEC 14496-15 (AVC file format) is defining a new
structure for fragmented MP4 called "avc3". The principal difference
between AVC1 and AVC3 is the location of the codec initialisation
data (e.g. SPS, PPS). In AVC1 this data is placed in the initial
MOOV box (moov.trak.mdia.minf.stbl.stsd.avc1) but in AVC3 this data
goes in the first sample of every fragment (i.e. the first sample in
each mdat box). The principal reason for avc3 is to make it easier
for client implementations, because it removes the requirement to
insert the SPS+PPS in to the decoder pipeline every time there is a
representation change.
This commit adds support for the "avc3" atom, which is almost identical
to the "avc1" atom, except it does not contain any SPS or PPS data.
https://bugzilla.gnome.org/show_bug.cgi?id=702004
This should make decoders able to precisely push buffers until the stop
time in case they need the next keyframe to do it.
Also, according to gst_segment_clip, it should only push a buffer that
the starting ts is strictly smaller than the segment stop, so we change
the min < comparison for <=
