These allow getting the absolute start timestamp of a fragment
without reading preceding timestamps. This fixes sync between
tracks if starting from fragments in different streams that don't
align exactly.
This also is a prerequisite for producing DASH content.
Signed-off-by: Martin Storsjö <martin@martin.st>
E-AC-3 samples should contain 6 audio blocks, so concatenate syncframes
in order to achieve this.
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
This introduces a new option to the mov demuxer: -use_mfra_for
(pts|dts). When it's given and moofs and a MFRA are present, the MFRA's
TFRAs are read for fragment start times.
Unfortunately some programs that produce fragmented mp4s use the TFRA
time field for dts and some for pts. There is no realistic way to detect
which is the case, hence the responsibility is punted onto the user.
This also means that no behavioural change is enabled by default - you
must pass either dts or pts for anything to happen.
Without this change, timestamps for some discontinuous fragmented mp4 are
wrong, and cause audio/video desync and are not usable for generating
HLS.
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
* commit 'e44ee1eb8db7393e9d43207c2e1812720e292e6d':
movenc: Simplify code by using an existing local pointer
Merged-by: Michael Niedermayer <michaelni@gmx.at>
* commit 'dad12ce452a9d69c0d9d53c375003947d5f1b02e':
movenc: Print a warning for an unhandled case of nonzero start dts with empty_moov
Conflicts:
libavformat/movenc.c
Merged-by: Michael Niedermayer <michaelni@gmx.at>
* commit 'dbb472cb2f2f799295a12b4922a6a8be2cccfdee':
movenc: Write edit lists for fragmented files as well, if necessary
Conflicts:
libavformat/movenc.c
The default for writing EDTS for fragmented mp4 is left at disabled
this can be overridden via command line with -use_editlist
but EDTS + fragments still does not fully work, which is why it is
left disabled by default
Merged-by: Michael Niedermayer <michaelni@gmx.at>
* commit '95ee4e2ce774e0339632d067161596bf3dadfc72':
movenc: Add some comments explaining subtle details in writing the edit lists
Merged-by: Michael Niedermayer <michaelni@gmx.at>
* commit '72f801619a1ae91969fee9a7d72519422433c998':
movenc: Adjust edit lists to trim out parts of tracks with negative pts
Conflicts:
libavformat/movenc.c
See: 66b45d8f7a
See: 14fd34d73b
Merged-by: Michael Niedermayer <michaelni@gmx.at>
* commit '8bef43388132b53f59a6e90add18900a3bb4cc60':
smoothstreamingenc: Simplify code by removing a redundant variable
Merged-by: Michael Niedermayer <michaelni@gmx.at>
Icecast uses HTTP 1.0 while Libav uses HTTP 1.1 and enables by
default chunked post.
Icecast actually forwards the HTTP chunk headers to the listener
as part of the media stream (without the chunk encoding HTTP headers)
causing the players to lose sync.
Disabling the option is enough to feed icecast properly.
Signed-off-by: Luca Barbato <lu_zero@gentoo.org>
This is necessary to get the right timestamp offset for content
that starts with dts != 0.
This currently only helps when writing fragmented files with a non-empty
moov atom. When writing an empty moov atom, we don't have any packets
yet, so we don't know the starting dts for the tracks.
Signed-off-by: Martin Storsjö <martin@martin.st>
This makes sure that audio preroll for e.g. AAC is signaled correctly.
Previously we only wrote the edit list correctly if we had negative
dts but started with pts == 0 (e.g. for video with B-frames).
Signed-off-by: Martin Storsjö <martin@martin.st>
In these cases, only drop dts. Because if we drop both we have no
timestamps at all for some files.
This improves playback of HLS streams from GoPro cameras.
Signed-off-by: Martin Storsjö <martin@martin.st>
Support only one independent substream right now, and only syncframes
containing 6 blocks.
Fixes part of ticket #3074
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
Trying to write to a stream id larger the the maximum requested is
a programming error, still there is no reason to leave a
reachable abort() in the codebase.
CC: libav-stable@libav.org
Fix basename computation code to allow just file names without any
directories in the path.
Signed-off-by: Vignesh Venkatasubramanian <vigneshv@google.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
In dahsmode Matroska is not writing the first Cluster for every
audio stream in the Cues element.
Signed-off-by: Frank Galligan <frankgalligan@gmail.com>
Reviewed-by: Vignesh Venkatasubramanian <vigneshv@google.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
syncframes in E-AC-3 can be combined to provide 6 audio blocks per
sample, thus requiring parsing for proper decoding.
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
* commit '20a5956b8daeee4cb59d6fa00ec809b02c04d7f8':
dump: split audio and video probing on multiple lines
Conflicts:
libavcodec/utils.c
libavformat/dump.c
Merged-by: Michael Niedermayer <michaelni@gmx.at>
* commit '7ae9791b641d1183910b6e7faca23f7ae08f8579':
avformat: bump version after mime_type change
Conflicts:
libavformat/version.h
Merged-by: Michael Niedermayer <michaelni@gmx.at>
* commit 'b15b06ebf582ae81e47d236524c9ad6e10c8a0a7':
avformat: use const char* instead of uint8_t* for AVProbeData.mime_type
Conflicts:
libavformat/format.c
Merged-by: Michael Niedermayer <michaelni@gmx.at>
This makes the field consistent with AVInputFormat.mime_type and the
argument type of av_match_name.
Signed-off-by: Vittorio Giovara <vittorio.giovara@gmail.com>
* commit '28816050e47b6dba430a52e429d21a864cffda8e':
lavf: Set the stream time base hint properly for chained muxers
Conflicts:
libavformat/segment.c
Merged-by: Michael Niedermayer <michaelni@gmx.at>
By using ff_avc_write_annexb_extradata instead of the h264_mp4toannexb
BSF, the code for doing the conversion itself is kept much shorter,
there's less state to restore at the end, we don't risk leaving the
AVCodecContext in an inconsistent state if returning early due to
errors, etc.
Also add a missing free if the base64 encoding fails.
Signed-off-by: Martin Storsjö <martin@martin.st>
Fixes regression from Ticket3962
This basically favors the first global header while since 34751f8313
it was the last. If this heuristic turns out worse, do not hesitate to revert this and
reopen 3962 as a feature request for multiple STSD
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
Fixes ticket #3948
Based-on-patch-by: Michael Niedermayer <michaelni@gmx.at>
Signed-off-by: Alexander Strasser <eclipse7@gmx.net>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
Fixes use of uninitialized memory
Fixes: signal_sigsegv_504fb0_10_signal_sigsegv_504fb0_343_mewmew_ssa.avi
Found-by: Mateusz j00ru Jurczyk and Gynvael Coldwind
Fixes out of array read
Fixes: signal_sigsegv_844d59_10_signal_sigsegv_a17bb7_366_mpegts_mpeg2video_mp2_dvbsub_topfield.rec
Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
The -hls_allow_cache parameter enables explicitly setting the
EXT-X-ALLOW-CACHE tag in the manifest file. That tag indicates
whether the client MAY or MUST NOT cache downloaded media
segments for later replay.
Valid values are 1 (=YES) or 0 (=NO) and the EXT-X-ALLOW-CACHE
will not show in the manifest for other values (or if
-hls_allow_cache is not used.
Signed-off-by: Martin Storsjö <martin@martin.st>
Width, Height and Sample Rate should be in the AdaptationSet tag
only if all the contained representations have the same width,
height and sampling rate. Otherwise they should go into the
Representation tag. This patch adds this functionality and a fate
test for the same.
Signed-off-by: Vignesh Venkatasubramanian <vigneshv@google.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
Fix incorrect bandwidth computation in some cases. When the cue end
descriptor is null (i.e.) start_time_ns == -1, existing bandwidth
computed (if any) should be returned rather than returning 0.
Signed-off-by: Vignesh Venkatasubramanian <vigneshv@google.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
Fix an incorrect hard code in cues_end computation. Updating the fate
test reference files related to the fix as well. The earlier computation
was clearly wrong as the cues_end field was greater than the file size
itself in some cases.
Signed-off-by: Vignesh Venkatasubramanian <vigneshv@google.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
avio_flush() did nothing useful for read streams. Fix it to behave as
expected, and discard the currently read buffer properly.
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
If we setup AVIO interrupt callback and it will be returns 1 on socket
timeouts and we try to connect to non-existing streams on some servers
(like nginx-rtmp) we got FD leak.
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
When AVFMT_FLAG_NOBUFFER is set, the packets are not added to the
AVFormatContext packet list, so they need to be freed when they are
no longer needed.
Signed-off-by: Anton Khirnov <anton@khirnov.net>
The -hls_allow_cache parameter enables explicitly setting the
EXT-X-ALLOW-CACHE tag in the manifest file. That tag indicates
whether the client MAY or MUST NOT cache downloaded media
segments for later replay.
Valid values are 1 (=YES) or 0 (=NO) and the EXT-X-ALLOW-CACHE
will not show in the manifest for other values (or if
-hls_allow_cache is not used.
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
After this the order from the original file is stored through readorder
when doing ffmpeg -i input.ass -c copy output.mkv.
And now that the ASS muxer honors the ReadOrder, extracting the ass back
(without transcoding) restores the original order.
* commit 'c463dfc7e49929a9891884312b23b27d14729c51':
rtpdec_hevc: Drop a duplicated, nonstandard entry
Merged-by: Michael Niedermayer <michaelni@gmx.at>
The RFC spec draft only specifies the "H265" name - there is no
specification saying how to interpret "HEVC" (if such a packet
format is specified it could be an entirely different format).
Since this is a very new standard (still a draft), there is little
need for compatibility with existing, broken implementations. Therefore
remove the extra alias, to avoid the risk of encouraging incorrect
usage.
Intentionally keeping the ff_hevc_dynamic_handler name for the
handler, to use "hevc" consistently as name for the codec instead
of "h265" within the library internals as long as there only is one
single variant in actual use.
Signed-off-by: Martin Storsjö <martin@martin.st>
Considering the palette is located at the end of extradata may be flawed
when the extradata contains the palette followed by the BottomUp field.
When the BottomUp field is present, exclude it from the palette.
Fixes part of ticket #1304
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
Before this commit, the code was muxing up to the 2nd line after
"[Events]" (assuming it to be the "Format:" line). The remaining are
generally "Comment:" directives which can stay in that place. mkvextract
behaves that way so it seems there is no reason for that extra
complexity.
Consider a file created with something like:
cat file1.mp3 file2.mp3 > result.mp3
Then if file2.mp3 has gapless information, result.mp3 would stop playing
something in the middle. This happens because the gapless info directs
the decoder to discard all samples after a certain position. To make
matters worse, the gapless info of file2.mp3 will be used when playing
the file1.mp3 part, because the gapless info is located at the end of
the file.
While handling concatenated gapless files correctly would be insane and
a lot of effort (especially without scanning the whole file on opening),
it's easy to prevent at least early EOF. Playback will happen to work,
even if it's slightly broken.
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
The code already had skipping of initial padding, but discarding
trailing frame padding was missing.
This is somewhat questionable, because it will make the decoder discard
any data after the declared file size in the LAME header. But note that
skipping full frames at the end of the stream is required. Encoders
actually create such files.
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
As suggested, this just adds _DEFAULT_SOURCE while preserving
_SVID_SOURCE for whatever old systems still need it.
Signed-off-by: Daniel Bomar <dbdaniel42@gmail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
* commit '2bb2c2bd75e5f4b28a945511cda77e0a1a44c758':
rtpenc_chain: Pass the initial time_base hint on to the chained muxer
See: [FFmpeg-devel] [PATCH] avformat/rtpenc_chain: Set timebase
Merged-by: Michael Niedermayer <michaelni@gmx.at>
In practice this hint is ignored - the rtp muxer always overwrites
the stream time base without taking the hint into account. But as
a general practice this is the correct way to pass a time base hint
on to a chained muxer.
This avoids warnings about using the codec time base as hint
being deprecated.
Signed-off-by: Martin Storsjö <martin@martin.st>
The size variable is (correctly) unsigned, but is passed to several functions
which take signed parameters, such as avio_read, sometimes after having
numbers added to it. So ensure that size remains within the bounds that
these functions can handle.
CC: libav-stable@libav.org
Signed-off-by: Diego Biurrun <diego@biurrun.de>
It appears this breaks build with MSVC
until someone who has MSVC setup has time to investigate and
workaround/fix this, its better to revert so that build is not broken
Thats even more so as the original commit only fixed a hypothetical issue
This reverts commit e587a428d7.
This adds a new option -hls_flags single_file that creates one .ts file
for HLS and adds byteranges to the .m3u8 file, instead of creating one
.ts file for each segment.
This is helpful at least for storing large number of videos, as the
number of files per video is drastically reduced and copying and storing
those files takes less requests and inodes.
This is based on work by Nicolas Martyanoff, discussed on ffmpeg-devel
in July 2014. That patch seems abandoned by the author, and contained
unrelated changes. This patch tries to add the minimum amount of code to
support the byterange playlists.
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
Fixes Ticket3901
the seek test error codes change due to a change in the failure path,
this could be avoided by changing the respective error codes to EINVAL
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
bmp pipe needs the bmp parser which is not bug free and should thus not be favored
over the bmp image2 demuxer
that also means this change could be reverted in case bmp pipe is improved so it
handles all single bmp images correctly
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
Previously, the returned error codes were intentionally ignored
(see fadd3a6821), to avoid aborting if the directory already
existed. If the mkdir actually failed, this was caught when
opening files within the directory fails anyway.
By handling the error code here (but explicitly ignoring EEXIST),
the error messages and return codes in these cases are more
appropriate and less confusing.
Signed-off-by: Martin Storsjö <martin@martin.st>
