According to draft-pantos-http-live-streaming-07, 6.3.4,
the duration of the last media segment in the playlist
should be used as initial minimum reload delay.
Signed-off-by: Martin Storsjö <martin@martin.st>
With the current default PES packet size, and very small audio bitrates,
audio packet duration gets too long. For players, which wait for a whole
audio packet (or more) it takes a very long time to start playing sound.
For 24kbps audio, one PES packet is about 1 second long. On Motorola STBs,
we observe about 3 second delay before the playback starts with the
default setting.
Signed-off-by: Anton Khirnov <anton@khirnov.net>
Do not assume the audio packets being always smaller than
DEFAULT_PES_PAYLOAD_SIZE.
Signed-off-by: Jindřich Makovička <makovick@gmail.com>
Signed-off-by: Anton Khirnov <anton@khirnov.net>
The 'fiel' atoms can be found in H.264 tracks clobbering the extradata.
MJPEG supports non field based extradata, and this data should be
preserved when copying.
This fixes demuxing of file where the first packet is not audio. Such files
are generated by our idroq muxer. It also fixes demuxing of audio only
idroq files.
Compared to just overwriting the old extradata, this has the
advantage of letting the decoder know exactly when the
extradata changed (otherwise it is changed immediately when the
new extradata packet is demuxed, even if there's old queued packets
awaiting to be decoded). This makes it easier for decoders to
actually react to the change, so they won't have to inspect
the extradata for each packet to see if it might have changed.
This works when sequentially playing a file with sample rate
changes, but if seeking past a new extradata packet in the
file, it obviously doesn't work properly. That case doesn't
work in flash player either, so it's probably ok not to handle
it.
Signed-off-by: Martin Storsjö <martin@martin.st>
Support Main Profile at High 1440 Level in MXF container,
using essence coding label from SMPTE RDD 9, table 6.
Signed-off-by: Janne Grunau <janne-libav@jannau.net>
Audio header information might get scrambled and would not parse,
yet wsqva_read_packet would try to parse audio packets causing
segfaults such as floating point exception.
Fixes bugzilla #141.
Signed-off-by: Martin Storsjö <martin@martin.st>
This avoids a segfault if the probe function wasn't able to
determine the format.
The bug was found by Panagiotis H.M. Issaris.
Signed-off-by: Martin Storsjö <martin@martin.st>
v410 is a packed 10-bit 4:4:4 YCbCr format used in
QuickTime.
Signed-off-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>
Signed-off-by: Diego Biurrun <diego@biurrun.de>
It sets the supplied AVFormatContext pointer to NULL after freeing it,
which is safer and its name is consistent with other lavf functions.
Also deprecate av_close_input_file().
The existing functions defined in intfloat_readwrite.[ch] are
both slow and incorrect (infinities are not handled).
This introduces a new header with fast, inline conversion
functions using direct union punning assuming an IEEE-754
system, an assumption already made throughout the code.
The one use of Intel/Motorola extended 80-bit format is
replaced by simpler code sufficient under the present
constraints (positive normal values).
The old functions are marked deprecated and retained for
compatibility.
Signed-off-by: Mans Rullgard <mans@mansr.com>
If the sdp is generated before the rtp muxer is initialized
(e.g. as when called from the rtsp muxer), this has to be done,
otherwise the rtp muxer doesn't know that the input really is
in mp4 format.
Signed-off-by: Martin Storsjö <martin@martin.st>
If an annex b bitstream is muxed into mov, the actual written
sample is reformatted to mp4 syntax before writing.
Currently, the RTP hints that copy data from the normal video
track, where the payload data might be offset compared to the
original sample that the RTP hinting used (when 3 byte
annex b startcodes have been converted into 4 byte mp4 format
startcodes).
Signed-off-by: Martin Storsjö <martin@martin.st>
This implements reading the tag in the demuxer and adds support for writing it
in the muxer. Some example channel layout tables for muxing are included for
ac3, aac, and alac, but they are not utilized yet.
Use Sound Sample Description Version 2 for all MOV files.
Updated FATE references accordingly.
Note that ADPCM is treated as compressed audio in version 2.
704af3e29c broke publishing
of rtmp streams, at least publishing to Wowza servers.
This changes all invoke commands to use nb_invokes.
Signed-off-by: Martin Storsjö <martin@martin.st>
malloc() is allowed to return NULL when zero is the argument. This
causes us to think malloc has failed and return AVERROR(ENOMEM). In
addition OS X malloc() returns an unfreeable non-NULL pointer for size
zero when alignment is greater than 16.
Pass the correct size in bits to mpeg4audio_get_config and add a flag
to disable parsing of the sync extension when the size is not known.
Latm with AudioMuxVersion 0 does not specify the size of the audio
specific config. Data after the audio specific config can be
misinterpreted as sync extension resulting in random and wrong configs.
Commit 035af99 made avconv always call an encoder when using the
null muxer. While useful for 2-pass encodes, it inadvertently
caused an extra memcpy of raw frames when decoding only.
This hack restores the old behaviour when only decoding while
allowing use of the null muxer with encoded streams as well.
Signed-off-by: Mans Rullgard <mans@mansr.com>
The media_type_mask is initialized via AVOptions for the
rtsp and sdp demuxers, but it isn't available as an option
for the rtp guessing demuxer (since it doesn't really make
sense there). Therefore, it must be manually initialized
instead, since a zero value means no media types at all
are accepted.
Signed-off-by: Martin Storsjö <martin@martin.st>
For the standardized 8 kHz sample rate, this works exactly the same.
For nonstandard sample rates, the different predefined G726
names (G726-16, G726-24, G726-32, G726-40) are interpreted as an
indication of the bits per coded sample, even though their
actual bitrates aren't what the name specifies.
This feels more sane than using free-form names for nonstandard
sample rate/bitrate combinations, e.g like G726-22, G726-33
for 11025 Hz.
Signed-off-by: Martin Storsjö <martin@martin.st>
This avoids overflow if frame_size is over 2147, since both
frame_size and AV_TIME_BASE are plain integers.
Signed-off-by: Martin Storsjö <martin@martin.st>
Not yet complete, for demuxing AAC the AAC header must be generated
manually.
Possibly the decoder could accept the header as extradata to simplify
this.
Signed-off-by: Anton Khirnov <anton@khirnov.net>
This simplifies the open functions by avoiding one function
call that needs error checking, reducing the amount of
extra bulk code.
Signed-off-by: Martin Storsjö <martin@martin.st>
This string will be passed to ff_http_auth_create_response
even if no proxy is used, resulting in reading uninitialized
memory. The other auth string is always initialized by
av_url_split.
Signed-off-by: Martin Storsjö <martin@martin.st>
This requires using a separate init function, since there
isn't necessarily any fmtp lines for this codec, so
parse_sdp_a_line won't be called. Incorporating it with the
alloc function wouldn't do either, since it is called before
the full rtpmap line is parsed (where the sample rate is
extracted).
Signed-off-by: Martin Storsjö <martin@martin.st>
Using an unsigned variable avoids problems with overflows.
There is further no need for a 64-bit intermediate here.
Signed-off-by: Mans Rullgard <mans@mansr.com>
This simplifies the decoder so it doesn't have to process an in-packet header
or handle arbitrary-sized packets. It also fixes decoding of files with large
headers.
AVFMT_NOTIMESTAMPS for crc, as it ignores the timestamps.
AVFMT_VARIABLE_FPS for framecrc, as it prints dts.
Many FATE changes, because avconv is no longer duplicating frames in
those tests.
Also added -vsync 0 for some tests to prevent avconv from dropping
frames until it can be fixed more properly.
AVFMT_NOTIMESTAMPS for md5, as it ignores the timestamps.
AVFMT_VARIABLE_FPS for framemd5, as it prints dts.
-vsync 0 for the vp8 test is needed because with vsync 2 the timestamp
guessing code gets confused by an altref frame that is never displayed
and drops a frame later.
Adding the thread count in frame level multithreading to has_b_frames
as an additional delay causes more problems than it solves.
For example inconsistent behaviour during timestamp calculation in
libavformat.
Thread count and frame level multithreading are both set by the user.
If the additional delay caused by frame level multithreading needs
to be considered in the calling code it has all information to take
it into account.
Should it become necessary to calculate a maximum delay inside
libavcodec it should be exported as its own field and not reusing
an existing field.
Based on a patch by Michael Niedermayer.
Signed-off-by: Janne Grunau <janne-libav@jannau.net>
Note: FCPublish/FCUnpublish are adobe server specific and not described
in the rtmp specification. Some servers might not cope with them at
all.
Signed-off-by: Luca Barbato <lu_zero@gentoo.org>
The Apple HTTP Live Streaming demuxer's implementation of
seeking searches for the MPEG TS segment which contains the
requested timestamp. In its current implementation it assumes
that the first segment will start from 0.
But, MPEG TS streams do not necessarily start with timestamp
(near) 0, causing seeking to fail for those streams.
This also occurs when using live streaming of HTTP Live Streams.
In this case sliding playlists may be used, which means that in
that case only the last x encoded segments are stored, the earlier
segments get deleted from disk and removed from the playlist.
Because of this, when starting playback of a stream in the middle
of such a broadcast, the initial segment fetched after parsing
the m3u8 playlist will not start from timestamp (near) 0, causing
(the admittedly limited live) seeking to fail.
This patch changes this demuxers seeking implementation to use
the initial DTS as an offset for searching the segments containing
the requested timestamp.
Signed-off-by: Martin Storsjö <martin@martin.st>
The tls protocol handles connections via proxies internally.
With TLS/SSL, the peer verification requires that the client
speaks directly with the server, since the proxy doesn't have
the remote server's private key.
Signed-off-by: Martin Storsjö <martin@martin.st>
This opens a plain TCP connection through the proxy via the
CONNECT HTTP method. Normally, this is allowed for connections
on port 443, but can in general be used to allow connections
to any port (depending on proxy configuration), and could thus
be used to tunnel any TCP connection via a HTTP proxy.
Signed-off-by: Martin Storsjö <martin@martin.st>
RTCP timestamps are only necessary to synchronize time between
multiple streams. For a single stream, the RTP packet timestamp
provides more reliable timing. As a result, single-stream RTP
sessions should now have accurate and monotonic PTS.
Signed-off-by: Martin Storsjö <martin@martin.st>
TLSv1 is compatible with SSLv3, so this doesn't change much
in terms of compatibility. By explicitly using TLSv1, OpenSSL
sends the server name indication (SNI) header, which we
already set using SSL_set_tlsext_host_name (earlier, this
didn't have any effect).
SNI allows servers to serve SSL content for different host
names with separate certificates on one single port (vhosts).
Signed-off-by: Martin Storsjö <martin@martin.st>
This makes the function accept the format of creation_time
as output by demuxers (e.g. the mov demuxer), making the
creation timestamp stay intact if transcoding.
Signed-off-by: Martin Storsjö <martin@martin.st>
This function is used in muxers for parsing the 'creation_time'
metadata key, for converting it to a time value.
This makes it match the behaviour of the exported 'creation_time'
metadata from demuxers, where it is in UTC, too.
Signed-off-by: Martin Storsjö <martin@martin.st>
Converting to double before the multiplication rather than after
avoids an integer overflow in some cases.
Signed-off-by: Mans Rullgard <mans@mansr.com>
The interrupt callback has to be passed in during opening (setting it
after opening isn't enough), since a blocking open couldn't be
interrupted otherwise.
Options are passed down to procotols and also need to be available
during open() in most cases.
Signed-off-by: Anton Khirnov <anton@khirnov.net>
This is a better io interrupt callback function, which has an
opaque parameter, which is given to the interrupt callback.
This allows callers to precisely cancel IO for one single
AVFormatContext, without interrupt other ones in the same
process.
Note, it's not needed in AVIOContext, at the moment.
Signed-off-by: Anton Khirnov <anton@khirnov.net>
Add a decoder for the VBLE Lossless Codec, which
still has a cult following. Used to be popular
several years ago on doom9.
Signed-off-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>
Signed-off-by: Martin Storsjö <martin@martin.st>
This is needed because the twinvq decoder cannot rely on bit_rate to be set.
The API documentation says that bit_rate is set by libavcodec, not by the
user.
Tested with both Basic and Digest authentication, and tested with
both proxy authentication and authentication for the requested
resource at the same time.
Signed-off-by: Martin Storsjö <martin@martin.st>
The error was hidden before, to avoid showing an error on the
first request where no auth has been provided, when the server
indicates which authentication method to use.
Now the error is printed if an authentication method was used,
but failed.
Signed-off-by: Martin Storsjö <martin@martin.st>
The chunksize internal variable has two different uses - for
reading, it's the amount of data left of the current chunk
(or -1 if the server doesn't send data in chunked mode), where
it's only an internal state variable. For writing, it's used
to decide whether to enable chunked encoding (by default), by
using the value 0, or disable chunked encoding (value -1).
This, while consistent, doesn't make much sense to expose
as an AVOption. This splits the usage of the internal variable
into two variables, chunksize which is used for reading (as
before), and chunked_post which is the user-settable option,
with the values 0 and 1, where 1 is default.
Signed-off-by: Martin Storsjö <martin@martin.st>
It is found in some 8svx files (e.g. ones created by SoX).
Currently the decoder reuses the 8svx functions because we already have
handling of a single large planar packet for the compressed 8svx codecs.
The return value ret isn't an error code that can be passed
to ERR_error_string().
This makes the error messages printed actually contain useful
information.
Signed-off-by: Martin Storsjö <martin@martin.st>
The caller expects the seekhead struct to be freed when calling
matroska_write_seekhead. Currently, the structure is leaked if the
seek fails.
Signed-off-by: Martin Storsjö <martin@martin.st>
This is to make developers aware of the fact that they will
start using the new init function at some point.
Signed-off-by: Martin Storsjö <martin@martin.st>
It might make sense not to make the function completely mandatory
immediately at the next bump, which might be quite soon after
the function was introduced.
Signed-off-by: Martin Storsjö <martin@martin.st>
This makes rdt work again, which has been broken since
603b8bc2a1. This commit made
opening a demuxer without a file (or in this case, with a filename
which can't be opened) fail, unless the demuxer actually declared
AVFMT_NOFILE.
Signed-off-by: Martin Storsjö <martin@martin.st>
Since 1.0.0, this function is deprecated. A new function,
CRYPTO_THREADID_set_callback is available, but if not set at all,
it uses the address of errno as thread id, which should be
sufficient for most systems.
On windows, it never was necessary to use this function even
before 1.0.0, it used the right win32 API function for this
by default.
Signed-off-by: Martin Storsjö <martin@martin.st>
All current usages of it are incompatible with localization.
For example strcasecmp("i", "I") != 0 is possible, but would
break many of the places where it is used.
Instead use our own implementations that always treat the data
as ASCII.
Signed-off-by: Martin Storsjö <martin@martin.st>
Earlier, sc->samples_per_frame was used for setting the frame size,
but all files don't have that set properly. The frame size is a
known constant for these codecs.
If frame_size isn't set, the mov/3gp muxer refuses to mux it.
This fixes stream copy of audio from
https://roundup.libav.org/file1248/Video_With_AMR-NB_Audio.3gp
to another 3gp file (roundup issue 2468).
Signed-off-by: Martin Storsjö <martin@martin.st>
These packets are valid packets, and consist of 1 byte (which
contains the mode bits).
This had been analyzed and reported by Igor Levin, igor d levin comverse com.
Signed-off-by: Martin Storsjö <martin@martin.st>
Note, this protocol doesn't yet check verify the server
certificate against a local database of trusted CA root
certificates.
Signed-off-by: Martin Storsjö <martin@martin.st>
Streams from RTSP or SDP that do not match an allowed type will
be skipped entirely, which allows video-only or audio-only
streaming from servers that provide both.
Signed-off-by: Martin Storsjö <martin@martin.st>
This requires some workarounds in the WAV muxer and demuxer. We need to write
the correct bits_per_coded_sample and block_align in the muxer. In the
demuxer, we cannot rely on the bits_per_coded_sample value, so we use the bit
rate and sample rate to determine the value.
This avoids having the decoder rely on AVCodecContext.bit_rate, which is not
required to be set by the user for decoding according to our API.
bits_per_coded_sample should be 8.
block_align is calculated incorrectly, but it is not needed anyway.
packet pts should be calculated in samples.
packet duration can be set.
This fixes false positives of has_codec_delay_been_guessed() for
streams where not every input picture generates an output picture,
such as interlaced H264.
