This also changes the callbacks a bit to (a) avoid passing references
for already ref-counted types, and (b) don't pass PID since the
parent knows this and gives it to the child parser.
Issue #832
Change-Id: I7dd44436c8d1ad81d42a813d16f850175b85ad1a
This changes the default MP4 output to use TTML and adds a way to
choose which one is used. This is done with 'format=ttml+mp4' or
'format=vtt+mp4'.
This also fixes the boxes output in WebVTT in MP4.
Change-Id: Ieaa7fc44fbf4dc020a5bb70cfa3578ec10e088ce
This only supports TTML output; meaning the user can convert WebVTT
into TTML, but not the other way around. This will be useful for
DVB-sub subtitles that would be better supported within TTML.
This only adds text-based output; a follow-up will add MP4 support.
Change-Id: I0944b7df95d7765e55f203fc5e9a644f5c455dd8
- Use std::move to transfer ownership.
- Avoid libxml primitives outside XmlNode.
- Have attribute setting return errors.
- Mark bool returns with WARN_UNUSED_RESULTS and use return value.
- Use std::string over char*.
Change-Id: Ia00bb29c690025275e270f10e5c065722481c0c5
Alpine does not support python3 yet, but depot_tools enabled python3
by default recently.
Disable python3 for now.
Fixes#763.
Change-Id: I57cd414702e89cafbe1b8beee810f89760129d10
This adds a new path when parsing MPEG2-TS streams to ignore unsupported
streams. This allows extracting supported streams when some of the
streams are unsupported. For example, you can extract audio from a
file that has unsupported video.
Change-Id: I608fcb19d0a573bfd35e9272f60b0b69346ae11a
This adds more generic settings for regions and CSS styles. These are
global settings, so they go on the StreamInfo object.
Change-Id: Ibb76c060206152ccf8e9a067c09877226f67c927
Always set AUTOSELECT=YES for DVS tracks and also exclude it from
standard AUTOSELECT and DEFAULT configuration logic.
Fixes#857.
Change-Id: Ie179d02390ff09f6e8c0b54892c1d6d32a3c38d6
Now text cues are composed of nested fragments that can be individually
styled. This allows portions of the cue to be bold, etc. The
WebVTT parser doesn't parse the inputs, but the original tags are
preserved in WebVTT output. The WebVTT output will add tags if the
style elements are present in the cue object.
Change-Id: I6abba4175e376e4f753193f7d8cac63e958d3c89
We currently have a bug about non-deterministic output in the MPD
generator. This works around that bug by optionally doing everything
in a single thread. This allows us to run manifest comparisons without
making the major changes needed to add that feature.
Issue #177
Change-Id: I10e1084dac77841220161fbd2575cdcb5c13c00e
Now the Cue settings are a generic object that is parsed in WebVTT.
This will allow setting the settings in different parsers without having
to use WebVTT-specifics.
Change-Id: I36689bec725bd2e515af962b7174fc5977f96fa2
This sets the groundwork for more generic text cues by having a more
generic object for the settings and the body. This also changes the
TextSample to be immutable and accepts the fields in the constructor
instead of using setters.
Change-Id: I76b09ce8e8471a49e6bf447e8c187f867728a4bf
Now text-based WebVTT also uses the generic media pipeline. This
converts the WebVttTextOutputHandler to a WebVttMuxer to be more
consistent with the other muxer types.
This also allows choosing between single-segment text and multi-segment.
Before, we would generate both and use single-segment for DASH and
multi-segment for HLS; but now you can choose between either and either
are supported in both DASH and HLS.
Change-Id: I6f7edda09e01b5f40e819290d3fe6e88677018d9
Now the same pipeline for handling the audio/videos streams will handle
the segmented text streams too. This doesn't apply to the text output,
only to the MP4 variants. This also fixes a bug where we added the
X-TIMESTAMP-MAP tag even when there wasn't TS streams; this doesn't
otherwise change the behavior around that tag.
Change-Id: I03f7cea56efa42e96311c00841330629a14aa053
The test added in the previous CL was broken due to a rebase on another
change. This subtly changed some of the byte offsets that broke the
test. This wasn't caught since I didn't rebase and re-run the tests
before merging.
Change-Id: Id7e4c7688278eae37da1a14f1648263b4dda98cd
This changes it from an OriginHandler to a MediaParser and moves the
handling of it to the Demuxer. This will allow more generic handling
of text by giving it the same abstractions as video/audio handling.
Change-Id: Ibbde3c84d228ec8e83af1ed266ea97dbc9589c24
In addition to the MediaSample handling of the MediaParser, this now
adds callbacks for TextSample. This allows reading text streams from
the media files.
Change-Id: I6c00e286e98bc9aafe05b99cf2f7ce6f89d167a9
Instead of having the text readers reading from the file directly, they
now accept the data as a stream.
Change-Id: Id1b32c867a8058a68ae7aab5c568f77672a4401d
The Chromium sysroot has a broken C++11 implementation that doesn't
allow us to use <condition_variable> or <chrono>. So we need to use the
system sysroot to use all C++11 features.
Change-Id: I41bec7ebddf598e1bf0494ee8bb2844e67bc8acf
The KeySource now only handles fetching the keys and loading any PSSH
info from the license; it will not handle generating new PSSH info
based on the config.
This will allow the PSSH generation to access to the full
EncryptionConfig so we can add additional options to it.
Issue #756
Change-Id: Ia67387aa3d5ec0d723b7f5f21fc517f64c840393
If the user has the GPAC MP4 tools installed, this will use that to
provide a better, text-based diff of the MP4 files. This makes looking
at test failures much better than just "binary files differ".
Change-Id: Ifdca54c02226a9f0fe9ddcf9c6b28e960e568111
There were several enum types that all were used for key systems. This
combines them into one to make it more clear and only needing to update
one. This also uses a bit field to specify multiple key systems instead
of using a std::vector.
Change-Id: Ia88039835492a5bd47f449ba4b76187046deeec0
The resolution values are uninitialized when parsing AVC decoder
configuration record if it does not contain parameter sets (SPS/PPS),
thus they could contain undefined values.
Fixes#750.
Change-Id: I6d54ec9f8740acd45a67b4b7d69031e097fbacc1
Opening a named pipe can block until both ends are open, and we cannot
control when the other end will be open. Ideally, we would always
open files in a thread so that Packager can be used with piped inputs
from naive applications without a potential deadlock.
This change will defer opening WebVTT files until the parser Run()
method is called from a thread. This way, WebVTT files being sent in
from a pipe will never be able to block the main thread.
Previously, files were opened on the main thread before calling the
parser constructor, passing the open file to the constructor as an
argument. I also tried doing it in the parser's InitializeInternal()
method, but that is also called from the main thread.
Change-Id: I54cc68ed9d48a8dc697829119be84d4065b1ae1c
Under command line flag --mvex_before_trak.
This is needed to workaround Android MediaExtractor bug which
requires |mvex| to appear before |trak|.
Closes#711.
Change-Id: Id41d71af5c0016f59023dda6408bbf502e12ac55
Added --crypt_byte_block, --skip_byte_block to allow users to specify
protection pattern other than 1:9 for pattern based encryption scheme,
e.g. cbcs and cens.
Closes#710.
b/147307451
Change-Id: I9f64a7639170c737f138572689b28d17286325c7
Per https://dashif.org/docs/DASH-IF-IOP-v4.3.pdf 4.4.3.6 Signalling the last segment number in Period, there are three ways to signal the last segment number:
a. Use the lmsg signalling as defined in clause 4.4.3.5.
b. Use the Segment Timeline with @r value greater or equal to 0.
c. Add a Supplemental Descriptor with @schemeIdUri set to http://dashif.org/guidelines/last-segment-number with the @value set to the last segment number.
We do not support (a). This change adds support for (c) when Segment Timeline (b) is not used, i.e. when Representation has constant duration (could be approximate).
Under flag --dash_add_last_segment_number_when_needed (disabled by default).
AV1-ISOBMFF spec is updated recently to use start alignment instead of
end alignment for cbcs:
for the protected scheme cbcs:
- BytesOfProtectedData SHALL start on the first byte and end on the
last byte of the decode_tile structure (including any trailing bits).
- A subsample SHALL be created for each tile, even if its size is less
than 16 bytes.
- ...
cenc protection scheme is not affected.
Closes#698.
Change-Id: Ic83a478fb2602d830c30daf3206a1c2d2c238a08
Remove custom constructors, which is no longer needed with C++11. Use
C++ struct zero initialization instead.
Fixes#686.
Change-Id: I512da9f23a250e8b9ebf8bb8e0a39ad0f822d0d3
Added Dolby Vision backward compatible signalling, i.e. for Dolby Vision
profile 8, both base codec without Dolby Vision and HDR codec with Dolby
Vision are signalled.
This is achieved by using a new MuxerListener implementation
MultiCodecMuxerListener, which wraps multiple child MuxerListeners and
is able to delegate to the child MuxerListeners based on the codecs in
the stream.
Closes#341.
Change-Id: I1967bb1ed503087cdd011c364e5fb5647d516ca4
HLS uses the EXT-X-MEDIA-SEQUENCE tag at the start of a live playlist in
order to specify the first segment sequence number. This is because any
live playlist have a limited number of segments, and they also keep
updating with new segments while removing old ones. When a player refreshes
the playlist, this information is important for keeping track of segments
positions.
When the packager starts, it naturally starts this count from zero. However,
there are many situations where the packager may be restarted, without this
meaning starting this value from zero (but continuing a previous sequence).
The most common situations are problems in the encoder feeding the packager.
With those cases in mind, this parameter allows to set the initial
EXT-X-MEDIA-SEQUENCE value. This way, it's possible to continue the sequence
number from previous packager run.
Closes#691.
Call CRYPTO_library_init to properly initialize crypto engine, which
enables AES-NI (Hardware AES) if it is supported by CPU.
Also added a performance benchmark test.
Closes#198.
Change-Id: I962a2da588d2f4f6cbe00c83ecc9a832db0e6042
Also roll_boringssl.py script is also adjusted to work with latest
boringssl.
Disable ASM compilation on Windows x64 which fails to compile. It also
means there are no HW AES on Windows x64.
Issue #198.
Change-Id: Ib7e8ff506f014c8c733f1882eeeddbe34fa28511
For MP4 files, mp4 parser tries to open and read the same file again
to handle trailing MOOV box.
Open FIFO again may result in packager hang.
Update the code to only attempt to load MOOV for local regular files,
as FIFO and remote files are unseekable anyway.
Fixes#664.
Change-Id: Ib286d2876d202cd5a248ffe70b13589b3cc74bc9
Although FRAME-RATE is optional per HLS specification, but per HLS
Authoring Specification for Apple Devices (https://apple.co/30n90DC):
Each EXT-X-STREAM-INF tag MUST have a FRAME-RATE attribute.
Also, iOS and TVos refuse to play the content if VIDEO-RANGE is
present but FRAME-RATE is missing.
Issue #632, #634.
Change-Id: Ica282f023a7e8538f7b506094e2286840cf5c193
Calculate FRAME-RATE from sample duration as in the DASH solution.
Right now calculated frame-rate may not be accurate in some scenarios,
so we avoid setting the attribute in HLS unless it is more than 30
fps. We will set it unconditionally once it is fixed.
Fixes#634.
Change-Id: I87b6e9a047d959ae88dd4dcb2b4786527ba5c9fc
- Parse and extract transfer_characteristics from H264/H265 VUI
parameters.
- Set VIDEO-RANGE attribute in HLS according to HLS specification:
https://tools.ietf.org/html/draft-pantos-hls-rfc8216bis-02#section-4.4.4.2
- Also added an end to end test.
Fixes#632.
Change-Id: Iadf557d967b42ade321fb0b152e8e7b64fe9ff3e
- Add relevant FOURCCs for Dolby Vision.
- Parse DOVIDecoderConfigurationRecord (dvcC, dvvC) to generate
Dolby Vision codec string.
- Propagate Dolby Vision configs (dvcC, dvvC, hvcE) from Demuxer
to Muxer.
- Add a Dolby Vision end to end test.
Support for backward compatibility signaling in DASH and HLS will be
added in a later CL.
Issue #341
Change-Id: If1385df5f48e04b59cb7661130bea48e26b453bf
Latest version of FFmpeg encodes non standard channel layout, e.g. 5.1(side), in AAC using PCE.
This is now supported with the below changes:
- Allow channel_configuration in ADTS header to be 0, as the cctual channel layout is specified
in PCE.
- Add GetFrameSizeWithoutParsing to determine the frame size before actually parsing the frame.
- Skip and resume later if not the whole frame is available.
- Also ensure that the next frame starts with a proper sync word.
Fixes#598.
Store and use the segment start time in the original time scale to
delete old segments, instead of storing it in seconds, as the segment
file name is generated using the time with the time scale.
Fixes#625.
Change-Id: I7524d597b1ffc081dd1399d6fb3ea47c13502881
- Parses parameter set NAL units in the samples.
- Calculate pixel width and height from track width and height.
Fixes#621, #627.
Change-Id: Ic1e120dccbd220b01168f7bf4effeaa43f95b055
We used to use the target segment duration provided by the user.
Unfortunately that does not work for iframe only playlist which
the target duration is effectively equal to the GOP duration.
In the new approach, we estimate the target duration from the initial
few blocks (10 blocks right now).
Fixes#610.
Change-Id: Ie8bf943e157149ca7ed3b9382fe0a1088d0774e2
Add dash_accessibilities stream descriptor, which is a semi-colon
separated list of accessibility_scheme_id_uri=value. It is optional.
Add dash_roles stream descriptor, which is a semi-colon separated
list of strings. It is optional.
Closes#565.
Change-Id: Idb1c20bb410fdd016db07e11fe507c102a3dd8ea
And also dvhe with dvh1.
Apple's specification does not recommend video formats with the
parameter sets stored in the samples.
It also fails mediastreamvalidator checks and some Apple devices /
platforms refused to play.
See https://apple.co/30n90DC 1.10.
Replaced with the corresponding formats with the parameter sets stored
in the sample descriptions instead.
Fixes#587.
Change-Id: Ic5d3f6fde115b1d09d1dcac32cef5fe0ad246aa0
We have logics in bandwidth calculation to ignore segments that is
smaller than half of target duration. The logic does not have any
effect right now as the target duration in mpd/hls params is always
zero.
This change will set target duration in mpd/hls params, thus it can fix
part of issue #581 as the last segment which is less than half of
target duration.
Issue #581.
Fixes#498.
Change-Id: Ieb2dbf4da9fc72a7b9de802cda4294f1954d29b4
Segment start,end time was used previously, which could result in
problems if there are discontinuity in the streams. E.g. if the
stream has timestamp, 10000, 10001, 10002 and then next segment
comes in with timestamp 1. With the previous logic, all the segments
would remain in the time shift buffer until after 10000 segments
even with a small time shift buffer depth of 10.
This could also happen when timestamp wraps around, which could
happen during long time of live streaming.
This change will also be useful to support multi-period live DASH.
Fixes#563.
Change-Id: Ie078d76c6e4af13ade9ad46191c8e3529069ed4d
This is needed for Dolby Vision as Dolby Vision uses NAL unit type 62
to carry private information.
Issue #341.
Change-Id: Ib3f06d8f36ef3d89c33162dd74c012a0c3a805e5
- Define BaseDescriptor and generic read / write operations.
- Define descriptors: ESDescriptor, DecoderConfigDescriptor,
DecoderSpecificInfoDescriptor, SLConfigDescriptor.
DecoderSpecificInfoDescriptor and all other descriptors can now
handle arbitrary length size, not limiting to 64 byte for
DecoderSpecificInfoDescriptor, which was placed to limit
ESDescriptor length size to one byte.
- Now DecoderConfigDescriptor is able to handle reading and writing
of all fields including buffer_size_db, which was not handled
earlier.
Fixes#536.
Change-Id: Ia8a775f8bf6e90e3343a85f0e643bc44cd017c7a
Addresses binary/str conversion problems like
'can't use a string pattern on a bytes-like object'.
With this change, packager_test.py will work with any version of python
after python2.5.
Fixes#553.
Change-Id: I02066942e4bcdddc1db1daf761abab4cc46169a9
With the update, the script needs to be launched from
out/{Debug,Release} directory.
Also updated docs to make it clear how to use the tool.
Fixes#538.
Change-Id: I37b71afc33e3b73ff5232e43d79c52ffa5859d57
Add crypto_period_seconds to Widevine key request
When using key rotation with Widevine DRM, a key server has to know
the duration of crypto period to relate generated keys to the media
playback time. This helps the server to provide relevant keys to
a client during license request.
Closes#544.
The file deletion could fail if it is still held by the web server.
This CL adds retry logic to retry the deletion.
Closes#533.
Change-Id: Ib42ed8db11f32897a29e3b58c1be89d05f5dd034
VLC seems to generate access units with extra AUDs. In #526, the below
sequence is seen:
AUD | SPS | PPS | SPS | PPS | AUD | SEI | SEI | SEI | IDR_SLICE
Previously we exit early when seeing AUD, which results in delayed
processing of the access unit.
The behavior is changed to continue processing the following NAL units
to workaround the content issue.
Closes#526.
Change-Id: I80f571c0711c6db1337eb393fce36fae5432b6c5
Generate an audio only master playlist if there are no videos and
subtitles.
We do not support mixing audio only EXT-X-STREAM-INF with video
EXT-X-STREAM-INF right now.
Fixes#461.
Change-Id: I999b335ad7abbe183ffcb0f5d471948977c2772f
The new directory permission is set to the permission of the last
existing directory in the file path. If none of the directory exists,
it is default to 0755.
Previously we use base::CreateDirectoryAndGetError(), which always sets
the new directory permission to 0700, which is not what we want.
Fixes#499.
Change-Id: Iee9429d5e63ada9588f74ff20d3cce28a1a6437b
It allows users to override the default language for text tracks.
If not specified, --default_language applies to both audio and text
tracks.
Issue #430.
Change-Id: I86a9baba2072be27b6661fa7b65a8bc8b6adb3cc
Add hls_characteristics stream descriptor, which is a colon or semi-colon
separated list of strings. It is optional.
Fixes#430.
Change-Id: Ifcf79316e68768ff065891933de565cd0ff32ec4
https://tools.ietf.org/html/rfc8216#section-4.1
The peak segment bit rate of a Media Playlist is the largest bit rate
of any contiguous set of segments whose total duration is between 0.5
and 1.5 times the target duration.
Fixes#498.
Change-Id: I1f28972b9cc5977735e47906bdcd88ba3942db5a
kFrameSizeCodeTable rows are ordered by 32kHz, 44.1kHz and 48kHz,
which is the reverse of fscod (48kHz, 44.1kHz and 32kHz).
Also updated unittests.
Fixes#487.
Change-Id: Icb0afb8bb895afde0028eee05b403bc85bf7b538
Dump the first 512 bytes of the buffer if container is unknown. This
can help us find out what the actual container is; and fix the
container detection bug if there is.
Closes#505.
Change-Id: I4a8fe5954d0419ef2ccbb9067ec2e9ffe1da417e
Created SubsampleGenerator to generate subsamples.
This is part of the EncryptionHandler clean up to make it more modular
and testable.
Change-Id: I6f4076b057027c72335beb3cbf1965341eb18031
Created EncryptorFactory to set up the encryptors.
This is part of the EncryptionHandler clean up to make it more modular
and testable.
Change-Id: I839bcd8a84fa873396360d67afb540fef1345673
Only Marlin Adaptive Streaming Specification – Simple Profile is
supported.
Two additional updates:
- Remove FairPlay ContentProtection element from DASH mpd as FairPlay
does not define a signaling in DASH.
- Updated end to end test to include all DRMs we support.
Closes#381.
Change-Id: Id12269b471ea34983b782cbd92f687332292ef59
Implemented according to bitstream spec at
https://aomediacodec.github.io/av1-spec/.
It will be used to do AV1 subsample encryption later.
Issue #453.
Change-Id: I84d8a2a780d95f2c9f430ee598838b97474cc0af
This was introduced earlier to indicate FairPlay protection system. But
in fact, it is sufficient to just use the system id for the indication.
- Also updated various parts of the pipeline to support empty PSSH.
- Added an additional FairPlay end to end test using fMP4.
Change-Id: Ica48b7b5235e9a2b5a7f722bcd0fc1ef2073ac13
Note that TTML in ISO-BMFF is not supported yet.
Also updated packager_test.py:
- Added a test using TTML passthrough.
- Computed output extension from input extension unless output_format
is specified.
Fixes#478.
Change-Id: Ia917fc4ed3c326782791ed67601fba02ea28b11d
The time for the previous segment was used when generating the segment
name. This resulted in the first segment being overwritten and
mismatching manifest and media files. It led to playback problems.
Issue #472.
Change-Id: Ia8130ce261585e1a2ede83b26de3e32508de087f
The VP9 level is computed when the container is missing a codec config
or if the level is missing from the codec config.
This fixes VP9 in ISO-BMFF files generated by FFmpeg v4.0.2 or earlier
which does not have level set in the codec config.
Fixes#469.
Change-Id: I685bfd48be16ee6b2209da1c3173f7d6bb02b36a
Implemented per AV1 Codec ISO Media File Format Binding at
https://aomediacodec.github.io/av1-isobmff/
And AOM AV1 codec mapping in Matroska/WebM at
https://github.com/Matroska-Org/matroska-specification/blob/av1-mappin/codec/av1.md
Note that AV1 specific boxes are not supported in this CL, i.e.
AV1 Forward Key Frame sample group entry 'av1f', AV1 Multi-Frame
sample group entry 'av1m' etc are not supported. These boxes are optional.
We will add support later if they are useful to the clients / players.
Encryption is not supported yet.
Issue #453.
Change-Id: I630432d0a9bf82d263ffaf40e57f67fc65eee902
We used to skip all the remaining bytes if the size is less than
|crypt_byte_block| blocks.
This interpretation is incorrect. We should only leave the last partial
16-byte block un-encrypted.
Change-Id: I4f09600efa52e2fdf4d0b661dfc418dcb675f9f6
Renamed "FakeMediaHandler" to "CachingMediaHandler" so that it
is easier to differentiate from "FakeInputMediaHandler". It will
be used as the single method to verify media handler output.
This is part of an effort to find a cleaner way to test media handlers.
Change-Id: Ie391fe23ee3389d13b012ae3093744a453106c19
Negative duration is not allowed, so set the duration of that sample to
an arbitrary small value in case it is needed to decode future samples.
Issue #451.
Change-Id: I9250d71d163f769ea2657d56e108b6dbd583de67
Keep bundled binutil scripts but not downloading actual binaries by default.
Automatic downloading of binutils has been causing problems for some users:
see #164, #412, #440.
Using bundled binutils helps reduce linking time, but packager codebase is
relatively small, so the gain is not significant.
Local testing shows that the full build time increases from 1m42s to 1m50s, i.e.
an increase of about 8s, or ~8%.
User can still enable the usage of bundled binutils by running
'python src/packager/third_party/binutils/download.py' and set
'linux_use_bundled_binutils' and 'linux_use_bundled_gold' to 1 in GYP_DEFINES.
Change-Id: I7ecae2333861f41054f957d0a6081dae948167bd
Note that STYLE and REGION are not supported in mp4 container due to
spec limitation as 14496-30:2014 does not specify a way to signal
styles/regions inside mp4.
Closes#344.
Change-Id: I05c14df916f7b2c7ca4364ee9407e0eda4dc7a3f
- Also fixed compilations in Alpine Linux and other flavors of Linux.
- Added container versions in docker files to always use a verified
version.
Closes#164.
Change-Id: I949a8709e4d70c49129c9c2e8608dd78193d964c
Configurable under --enable_legacy_widevine_hls_signaling, off by
default.
When it is enabled, do not fail if provider or content_id are missing,
but log a warning instead.
Bug: 112268769.
Change-Id: I2531aa7474d2818700b90fa0679b49891bb935ef
--io_block_size determines the buffer size for each read/write in
ThreadedIO.
A big io_block_size does not help much as long as --io_cache_size
is big enough.
Instead, it may cause problems. For example, Linux pipes block until
reading io_block_size number of bytes, which causes a large delay
when reading audio streams.
Change-Id: I5bdd3d61388579c7e8647cdab3152135a40a752b
In some ISO-BMFF files, there is an initial non-zero composition offset,
but there is no EditList present.
This is against ISO-BMFF spec recommentation [1] and we believe in most
cases it is just missing the EditList.
[1] 14496-12:2015 8.6.6.1
It is recommended that such an edit be used to establish a presentation
time of 0 for the first presented sample, when composition offsets are
used.
Issue: #112.
Change-Id: I178d5ec9d8c294c9f70aac4f4dd6254c824e2255
Previously it is possible that the same file path is generated when
the function is called consecutively in the same thread.
The problem can be re-produced in Windows. Does not seem to be
re-producible in Linux and Mac.
Fixes#448, #449.
Change-Id: Ia0163492e3494eba00f56f4d356aa1010b9660cc
- Also make it explicit that MemoryFile does not support opening an
already open file. An error will be returned when trying to open an
already open file. Previously the code may crash with memory problems.
- Also updated packager_test to use different test directories for
different tests.
Fixes Appveyor crash due to memory corruption: #449.
Appveyor may still fail but will contain a meaningful error logging.
Change-Id: Ibc9346ef7f301e416a4a09f120bca56504c939d8
This also ensures that it does not violate std::sort() requirement on
strict ordering, which is enforced in gcc/g++ though not in clang.
std::sort() strict ordering requirement:
std::sort() need a comparator that return true iff the first argument is
strictly lower than the second one. That is: must return false when they
are equal.
Change-Id: I781cf4ed4125fcad212eba5430a264f3a3d71c16
This is recommended by Apple and other content providers:
https://apple.co/2uTTAc4
- Set default --clear_lead to 12 seconds, i.e. about 2 segments.
- Set default --fragment_duration to 0 so fragment (sub-segment) is not
generated by default.
Change-Id: Ie6ec3ab6f0ce973547989c663a92b637a2fdc47c
It is possible to see a negative timestamp if the input mp4 file
contains EditList.
Also increase the queue size for EncryptionKey to be able to handle
smaller crypto_period_duration during testing.
Change-Id: I68278bf482d6662e771d9b80f4d5409605065aac
Configurable with --transport_stream_offset_ms.
This is needed to compensate for possible negative timestamps in
inputs, which could happen on ISO-BMFF with EditLists.
Issue #112.
Change-Id: I0fce8766c9df2911b9bb859c1e54052a8ed2abfb
In some ISO-BMFF files, there is an initial non-zero composition offset,
but there is no EditList present.
This is against ISO-BMFF spec recommentation [1] and we believe in most
cases it is just missing the EditList.
[1] 14496-12:2015 8.6.6.1
It is recommended that such an edit be used to establish a presentation
time of 0 for the first presented sample, when composition offsets are
used.
Issue: #112.
Fixes: b/110782437.
Change-Id: I23d33810ce536b09a1e22a2644828d824c1314f5
- EditLists in input files are parsed and applied to sample timestamps.
- An EditList will be inserted in the ISO-BMFF output if
- There is an offset between the initial presentation timestamp (pts)
and decoding timestamp (dts). Chrome, as of M67, still uses dts in
buffered range API [1], which creates various problems when buffered
range by pts does not align with buffered range by dts. There is
another bug in Chrome that applies EditList to pts only [2]. This
means that we can insert an EditList to align pts range and dts range.
- MediaSamples have negative timestamps (e.g. for Audio Priming).
You may notice the below change on some contents:
- Some media duration is reduced by one or two frames. This is because
EditList in the input file was ignored in the previous code, so video
streams start with a zero dts and a non-zero pts; the smaller of dts
and pts was used as the starting timestamp (related to the earlier
workaround for Chrome's dts bug), so the calculated duration was
actually a bit larger than the actual duration. Now with EditList
applied, the initial pts is reduced to zero, so the media duration is
also reduced to reflect the actual and correct media duration.
It may also result in negative timestamps in TS/HLS Packed Audio, which
will be addressed in a follow up CL.
Fixes#112.
Partially address b/110782437.
[1] https://crbug.com/718641, fixed but behind MseBufferByPts.
[2] https://crbug.com/354518. Chrome is planning to enable the fix for
[1] before addressing this bug, so we are safe.
Change-Id: I59317740ad3807ca66fa74b3a18fdf7f32c96aeb
DASH live profile with static MPD was incorrectly handled in the same
way as live profile with dynamic MPD, i.e. by assuming the synchronized
Representations, which is incorrect.
It can be easily re-produced in multiple period outputs, i.e. with Ad
cues. It may happen with regular contents as well, due to race
condition, though with a much lower chance of occurring.
Fixes#435.
Bug: 111359775.
Change-Id: I7de087f5dd8602b4c4e35cb697d589fa3699e8a5
They will result in empty DASH representations, which is not spec
compliant.
For text, if the cue is before the max end time, it will still be
dispatched as the text samples intercepted by the cue can be split into
two at the cue point.
Bug: 111359775.
Change-Id: I55c2025c4e9d64c88e6a685c0cf3024a0cc4a6d8
MPD cannot be validated right now since we do not generate deterministic
mpd with multiple inputs due to thread racing.
Exclude mpd comparison in this case.
Change-Id: I03b68b4969a39e20927c8a9b1475f72493682696
I.e. enable --generate_dash_if_iop_compliant_mpd in mpd_generator by
default.
It was enabled in the main packager binary in v1.6.1.
Change-Id: Icfb19758c52c107e8ab17d3fb581923fa291cc0a
WebVTT cues without payload may not carry meaningful information, but it
is allowed by WebVTT specification [1]. It could also be useful
sometimes, e.g. to signal the time progression in live case.
Fixes#433.
[1] https://www.w3.org/TR/webvtt1/#types-of-webvtt-cue-payload
Change-Id: I9e31f4a3789cbdafb7667b64f4019834190ecfc0
- Support decryption verification in _CheckTestResults
- Allow diff_files in _CheckTestResults
- Update all functional tests to use _CheckTestResults
Change-Id: I3f9c02f35808eba787becf9b1e5c1ce9238f943e
Instead, caclulating average bandwidth by dividing the sum of the
sizes of every segment by the sum of the durations of every segment.
This aligns with the requirement in HLS spec:
https://tools.ietf.org/html/draft-pantos-http-live-streaming-23 4.1.
BandwidthEstimator is also simplified to handle all blocks only.
Fixes#361
Change-Id: I89e7d415a841f4d4048f199de8dae7ffa250467b
There are non-ASCII characters in comment section of some source
code, which is causing compilation problems on some systems with
a codepage that cannot represent these characters.
Fixes#419.
Change-Id: I20d68a201263d515290ee440b52c5354b739099a
Before we had an assert that would catch if a sample had an
invalid times, however input may have bad times.
We did have a message if we saw a sample with a duration equal to
zero. This expands that check to check if the time is valid in
general and will ignore any sample that is not valid.
Fixes#425
Change-Id: I9774bfbdbd401f3016d2c345665b9973d1889db7
This flag was designed for two purpose:
- Grouping fragments into subsegments, achieving three level hierarchy:
segment < subsegment < fragment.
- Indicate whether to generate 'sidx' box in media segments (when the
value is set to a negative number).
There are no practical use case for the first purpose. Removing it to
simplify the code and reduce the confusion.
Introduce another flag --generate_sidx_in_media_segments for the second
purpose.
Change-Id: I4be7cd42662fb324c1158b978e05768ee49dd048
It was implemented to workaround Chromium's DTS
https://crbug.com/398130, but the workaround does not really work in
all situations.
Remove it now as we already have another workaround available.
Change-Id: I291f559d78120fb743a6679b7d927e5bbc5b6b4e
DTS was used in ChunkingHandler. As a result, SegmentInfo contained
timestamp in DTS. MP4Muxer has a logic to change SegmentInfo to use
PTS but not in other muxers.
Benefits of using PTS in ChunkingHandler:
- De-dup the redundant logic in MP4Muxer
- Ensure consistent behavior in different output containers
- Consistent with other timestamps, e.g. Ad Cue timestamps
Issue #413
Change-Id: Ib671badf144e0c0866d60f4ff0ac0cbbdd33817e
We always assumed that text stream started at zero in the text chunker.
This meant that if the text stream started later than zero (like in a
live stream) we would generate a lot of empty segments.
Instead the text chunker will assume that segments should be adjusted
to align with multiple of segment duration. The text chunker will assume
that an earlier component in the pipeline (i.e. text padder) will ensure
that the first sample has the time that we want to start segments at
(even it needs to add an empty sample).
Issue #416
Change-Id: Ie45844354d6e9448787cae896841b5ab31721ed6
Previously, the text padder media handler would assume that text always
started at time zero. This would work for VOD but would result with a
large pad at the start of LIVE content.
To avoid this, the text padder will use a bias to test whether or not it
thinks the content starts at zero. Right now the bias is set to be 10
minutes, but will later be configurable with a command line flag.
10 minutes was used as LIVE content will have much larger values and VOD
content should have much lower values.
Issue: #416
Change-Id: I07af15a577392fb030e36f052085cd4e667700e8
This is a more faithful implementation of more_rbsp_data().
There could be trailing null bytes in NAL units. This isn't valid per
H264 specification, but the referenced bug includes a sample where the
PPS in the avcC record includes a trailing null byte.
Workaround the problem so packager does not fail.
A similar problem is workarounded in Chrome:
https://codereview.chromium.org/1107593004Closes#418
Change-Id: I28cb8a9371945dc094f766c3e559d7a66859b451
This allows UDP receive buffer size to be increased to avoid potential
packet loss result from receive buffer overrun.
Another related flag is --io_cache_size. buffer_size in UDP options
defines the UDP buffer size of the underlying system while
io_cache_size defines the size of the internal circular buffer managed
by Shaka Packager.
Closes#411
Change-Id: I57c843a88f13da546417dadc2a78df2bee0a00f3
Under VLOG(1).
IO Cache circular buffer full could result in UDP packet loss with UDP
inputs.
Related to #390 and #411.
Change-Id: Ia5636a70827978b5c2f71f6495256d758766632f
When we built the cue alignment media handler, we did not remove
the old solution.
This change removes all the code for the old solution.
Change-Id: I851b284c449c7d25aaabc2f55df5579ba7b5aad1
Originally the create and dispatch of stream data in the cue
alignment tests were separate to improve readability. The idea was
to separate building each stream and dispatching the streams muxed
together would make it easier to understand the streams.
However, coming back to the code, this did not hold up as it took just as
long to understand the code. So merging the two together allows for the
code to be more concise.
Change-Id: I77f6bb67e7fb0ca7af0553c3baf07082770da74c
In the media handler there is an example of what a one-to-many media
handler would be. It used the trick play handler as an example, but that
handler is now a one-to-one media handler. Changed the example to use the
replicator media handler.
Change-Id: I26908c6e27ea4a697a19c0fa0179c60842a449d2
Made ChainHandlers to MediaHandler::Chain so that it can be used in our
test code as well. After all it is a pretty helpful function.
Change-Id: I8d83ee184052cd9fa9b37f2741c96f3223d5ab48
Jacob Trimble
KongQun Yang
koln67
akata01
sr90
Weiguo Shao
Joe Foraci
Zajcev Evgeny
Alen Vrečko
yokoyama10
Sergio Garcia Murillo
Tim Lansen
Joey Parrish
Abhay Sundaram
Tomohiro IKEDA
Daniel Cantarín
Vladimir Kazakov
Phillip Baxter
srebrnyp
Denis Akhmetzyanov
Aaron Vaage
Yohann Connell