Merge commit 'bf12a81cc67d62dd45c58e29fa0e9177331cc151'

* commit 'bf12a81cc67d62dd45c58e29fa0e9177331cc151':
  configure: Replace `pr` since it is not provided by busybox

Not merged as requested by Timothy Gu.

Merged-by: Hendrik Leppkes <h.leppkes@gmail.com>

Merge commit '5f3a081b42b84404a40a52c80ef7a354cf048c56'

* commit '5f3a081b42b84404a40a52c80ef7a354cf048c56':
  avi: Spin out the logic to position to the next non-interleaved stream

Merged-by: Hendrik Leppkes <h.leppkes@gmail.com>

swscale/arm: add ff_nv{12,21}_to_{argb,rgba,abgr,bgra}_neon_16

fate/api-{jpeg,png}-codec-param: add missing codec dependencies

fate/api-codec-param: fix codec context leak

avcodec/libdcadec: honor -err_detect option

Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/libdcadec: add some useful options

Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/libdcadec: implement logging callback

Don't print a warning when dcadec_context_filter() returns positive
warning code. Most relevant warnings are now output through the callback
function.

Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/libdcadec: fix request_channel_layout

Take request_channel_layout as a hint and don't force 2.0 downmix by
using both the 2CH and 6CH flags together.

Remove warnings about missing coefficients because they are no longer
relevant.

Honor AV_CH_LAYOUT_NATIVE and make it possible for native DTS channel
layout to be output.

Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

lavf/mxfdec: fix seeking before the first keyframe

Regression since 53f2ef2c4afb1d49a679dea9163cb0e4671f3117.
Fixes ticket #5017.

Signed-off-by: Marton Balint <cus@passwd.hu>

lavu/opencl: restore #if HAVE_THREADS

Fix regression since a8bb81a.

Spotted-by: RiCON

cosmetics: Fix weird indentations

lavc, lavu: use avutil/thread.h instead of redundant conditional includes

configure: fix vp9_d3d11va_hwaccel deps

Signed-off-by: James Almer <jamrial@gmail.com>

avutil/threadmessage: fix build without HAVE_THREADS, new attempt

avcodec/g723_1: fix license header

Merge commit 'cb49bb10ca7fcff2e382d9d989232b1a7f28e7da'

* commit 'cb49bb10ca7fcff2e382d9d989232b1a7f28e7da':
  build: Move -Wcast-qual to the extra_warnings

Merged-by: Hendrik Leppkes <h.leppkes@gmail.com>

Merge commit '0fc61c6ab6912a2f0c40fdd3f3c591bc2a33efd4'

* commit '0fc61c6ab6912a2f0c40fdd3f3c591bc2a33efd4':
  avi: Validate the stream-id for DV as well

Merged-by: Hendrik Leppkes <h.leppkes@gmail.com>

Merge commit 'd017ed878a45171f2f6c69fb9d76401c3c494110'

* commit 'd017ed878a45171f2f6c69fb9d76401c3c494110':
  avi: Use the correct data type

Merged-by: Hendrik Leppkes <h.leppkes@gmail.com>

Merge commit '7f57ea143c55ce5732ef7e31e4b75ae6c307af13'

* commit '7f57ea143c55ce5732ef7e31e4b75ae6c307af13':
  vsrc_color: Drop unneeded variable

Merged-by: Hendrik Leppkes <h.leppkes@gmail.com>

Merge commit 'ca5f386e75c592ce25b8184516fd0d580ccb31bb'

* commit 'ca5f386e75c592ce25b8184516fd0d580ccb31bb':
  lavf: G.723.1 muxer

Merged-by: Hendrik Leppkes <h.leppkes@gmail.com>

Merge commit 'f023d57d355ff3b917f1aad9b03db5c293ec4244'

* commit 'f023d57d355ff3b917f1aad9b03db5c293ec4244':
  lavc: G.723.1 encoder

Split existing FFmpeg G.723.1 encoder into a new file.

Merged-by: Hendrik Leppkes <h.leppkes@gmail.com>

Merge commit '165cc6fb9defcd79fd71c08167f3e8df26b058ff'

* commit '165cc6fb9defcd79fd71c08167f3e8df26b058ff':
  g723_1: Move sharable functions to a separate file

Merged-by: Hendrik Leppkes <h.leppkes@gmail.com>

Merge commit 'aac996cc01042194bf621d845bbe684549b5882e'

* commit 'aac996cc01042194bf621d845bbe684549b5882e':
  g723_1: Rename files to better reflect their purpose

Merged-by: Hendrik Leppkes <h.leppkes@gmail.com>

fate/api: add w32+os2 support for fate-api-threadmessage

Merge commit 'b74b88f30da2389f333a31815d8326d5576d3331'

* commit 'b74b88f30da2389f333a31815d8326d5576d3331':
  g723_1: Handle values at the ends of the table in lsp2lpc()

Merged-by: Hendrik Leppkes <h.leppkes@gmail.com>

Merge commit 'a0fa6d06b848f26b16ba12f0a9a4a85b93ab8022'

* commit 'a0fa6d06b848f26b16ba12f0a9a4a85b93ab8022':
  matroska: Warn when metadata references a non-existent element

Merged-by: Hendrik Leppkes <h.leppkes@gmail.com>

Merge commit 'f7986239f4dbec91c743c4c5eb0a2339bd325bf6'

* commit 'f7986239f4dbec91c743c4c5eb0a2339bd325bf6':
  dvenc: Validate the frame size before copying it

Merged-by: Hendrik Leppkes <h.leppkes@gmail.com>

fate/api: fix fate-api-threadmessage dependency

avutil/threadmessage: fix build without HAVE_THREADS

Merge commit '7139489c452ef8af6a745ec4e62056ee4ea4d6a8'

* commit '7139489c452ef8af6a745ec4e62056ee4ea4d6a8':
  pgssubdec: fix API compability layer

Merged-by: Hendrik Leppkes <h.leppkes@gmail.com>

Merge commit 'b0e8651a2a84553d08fbb2f7cb9697bd64fb1b55'

* commit 'b0e8651a2a84553d08fbb2f7cb9697bd64fb1b55':
  doc: Amend the MSYS2 Documentation

Merged-by: Hendrik Leppkes <h.leppkes@gmail.com>

Merge commit '4f979418c723652ad4e43115118c57a44bd46b52'

* commit '4f979418c723652ad4e43115118c57a44bd46b52':
  avfoundation: Simple capture

Not merged on request by Thilo Borgmann, original author of the
AVFoundation capture in FFmpeg.

Merged-by: Hendrik Leppkes <h.leppkes@gmail.com>

doc/filters: add more compand examples

Signed-off-by: Paul B Mahol <onemda@gmail.com>

avfilter/af_biquads: pass filter ctx to av_log calls

Signed-off-by: Paul B Mahol <onemda@gmail.com>

avutil/threadmessage: split the pthread condition in two

Fix a dead lock under certain conditions. Let's assume we have a queue of 1
message max, 2 senders, and 1 receiver.

Scenario (real record obtained with debug added):
    [...]
    SENDER #0: acquired lock
    SENDER #0: queue is full, wait
    SENDER #1: acquired lock
    SENDER #1: queue is full, wait
    RECEIVER: acquired lock
    RECEIVER: reading a msg from the queue
    RECEIVER: signal the cond
    RECEIVER: acquired lock
    RECEIVER: queue is empty, wait
    SENDER #0: writing a msg the queue
    SENDER #0: signal the cond
    SENDER #0: acquired lock
    SENDER #0: queue is full, wait
    SENDER #1: queue is full, wait

Translated:
 - initially the queue contains 1/1 message with 2 senders blocking on
   it, waiting to push another message.
 - Meanwhile the receiver is obtaining the lock, read the message,
   signal & release the lock. For some reason it is able to acquire the
   lock again before the signal wakes up one of the sender. Since it
   just emptied the queue, the reader waits for the queue to fill up
   again.
 - The signal finally reaches one of the sender, which writes a message
   and then signal the condition. Unfortunately, instead of waking up
   the reader, it actually wakes up the other worker (signal = notify
   the condition just for 1 waiter), who can't push another message in
   the queue because it's full.
 - Meanwhile, the receiver is still waiting. Deadlock.

This scenario can be triggered with for example:
    tests/api/api-threadmessage-test 1 2 100 100 1 1000 1000

One working solution is to make av_thread_message_queue_{send,recv}()
call pthread_cond_broadcast() instead of pthread_cond_signal() so both
senders and receivers are unlocked when work is done (be it reading or
writing).

This second solution replaces the condition with two: one to notify the
senders, and one to notify the receivers. This prevents senders from
notifying other senders instead of a reader, and the other way around.
It also avoid broadcasting to everyone like the first solution, and is,
as a result in theory more optimized.

fate/api: test threadmessage

avutil/threadmessage: add av_thread_message_flush()

fate: add limited_input_seek tests

Signed-off-by: Simon Thelen <ffmpeg-dev@c-14.de>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

ffmpeg_dxva2: support vp9 hwaccel

avcodec: implement vp9 dxva2 hwaccel

vp9: add hwaccel hooks

vp9: fix pixel format changes with threading

cmdutils: use version accessor macros

Signed-off-by: Reynaldo H. Verdejo Pinochet <reynaldo@osg.samsung.com>

avfilter/showcqt: BASEFREQ and ENDFREQ cast to double

fix default basefreq/endfreq comparison
on platform that does not do comparison
in double type

found on zeranoe 32-bit build, where
default freq range is detected as non-default

Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

Remove Redundant Entry of MPEG2 Video Desc

Signed-off-by: Anshul Maheshwari <er.anshul.maheshwari@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

mjpegdec: consider chroma subsampling in size check

If the chroma components are subsampled, smaller buffers are allocated
for them. In that case the maximal block_offset for the chroma
components is not as large as for the luma component.

This fixes out of bounds writes causing segmentation faults or memory
corruption.

Reviewed-by: Michael Niedermayer <michael@niedermayer.cc>
Signed-off-by: Andreas Cadhalpun <Andreas.Cadhalpun@googlemail.com>

fate: change the CMP_TARGETs for the recent AAC encoder changes

The case of PNS was outdated and resulted in failures on some
kdfreebds systems.

Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>

aacenc_tns: tune and reduce artifacts

There are a couple of major changes here:

1. Start using TNS coefficient compression.
2. Start using 3 bits per coefficient maximum for short windows.
The bits we save from these 2 changes seem to make a nice impact on the
rest of the file/windows.

3. Remove special case gain checking for short windows.
4. Modify the coefficient loop to support up to 3 windows.
The additional restrictions on TNS were something that was no in the
specifications and furthermore restricting TNS to only low energy short
windows was done to compensate for bugs elsewhere in the code.

Overall, the improvements here reduce crackling artifacts heard in very
noisy tracks.

Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>

aacenc: move the TNS search and filtering before PNS

The original plan was to have TNS use data from the PNS search to better
tune itself to noise but this was never used nor necessary. This should
slightly boost the PNS accuracy if TNS was used.

Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>

avfilter/af_alimiter: add 3 more options

Signed-off-by: Paul B Mahol <onemda@gmail.com>

fate/concatdec: Use -bitexact

Fixes FATE failures on --enable-small builds.

Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

ffprobe: Do not print profile names in -bitexact

Instead, print "unknown" if it's unknown, or their numerical values if
they are known.

Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

ffplay: use hypot()

Reviewed-by: Marton Balint <cus@passwd.hu>
Signed-off-by: Ganesh Ajjanagadde <gajjanagadde@gmail.com>

avcodec/dvdsubdec: fix typo in dlog message

Likely accidental in 764900d6458a2f79166ff91df4f20ad39cd6acec.

Fixes: CID 1341578.
Reviewed-by: Michael Niedermayer <michael@niedermayer.cc>
Signed-off-by: Ganesh Ajjanagadde <gajjanagadde@gmail.com>

avformat/movenc-test: correct varargs usage

It is required to call va_end for each invocation of va_start within the
same function.

Fixes: CID 1341583.
Reviewed-by: Michael Niedermayer <michael@niedermayer.cc>
Signed-off-by: Ganesh Ajjanagadde <gajjanagadde@gmail.com>

avformat/rtsp: free opts dictionary on failure of getnameinfo

Fixes: CID 1341579.
Reviewed-by: Michael Niedermayer <michael@niedermayer.cc>
Signed-off-by: Ganesh Ajjanagadde <gajjanagadde@gmail.com>

avfilter/af_stereotools: fix logic fail

Signed-off-by: Paul B Mahol <onemda@gmail.com>

libavutil: add version component accessor macros

Pretty standard macros, these should help libav*
users avoid repeating ver.si.on parsing code,
which aids in compatibility-checking tasks like
identifying FFmpeg from Libav (_MICRO >= 100 check).
Something many are doing since we are not
intercompatible anymore.

Signed-off-by: Reynaldo H. Verdejo Pinochet <reynaldo@osg.samsung.com>

doc/encoders.texi: fix typo and clarify mpeg2_aac_low

Thanks to Brad Smith for finding the typo.

Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>

avcodec/hevc: Fix integer overflow of entry_point_offset

Fixes out of array read
Fixes: d41d8cd98f00b204e9800998ecf8427e/signal_sigsegv_321165b_7641_077dfcd8cbc80b1c0b470c8554cd6ffb.bit
Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

aacenc_ltp: use an AR filter for LTP encoding as well

Seems to work better. Information on why the decoder does this is
lacking.

Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>

aacenc: fix aac_pred option triggering an error

Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>

doc/encoders.texi: use words intead of numbers to describe option states

Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>

Changelog: append experimental flag removal to the AAC encoder entry

Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>

avcodec/dirac_parser: Check that there is a previous PU before accessing it

Fixes out of array read
Fixes: 99d142c47e6ba3510a74b872a1a2ae72/asan_heap-oob_11b36f4_3811_0f5c69e7609a88a580135678de1df844.dxa
Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/dirac_parser: Add basic validity checks for next_pu_offset and prev_pu_offset

Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/dirac_parser: Fix potential overflows in pointer checks

Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

doc/encoders.texi: remove forgotten mention of "experimental" from libvo-aacenc

Previous documentation commit overlooked that particular mention.

Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>

aacenc: remove the experimental flag

Thiss commit removes the experimental flag from the native AAC Encoder
and thus makes it the default.

After a lot of work, done by myself and Claudio Freire, the quality of
this encoder rivals and surpasses libfdk_aac in some situations. The
encoder had instability issues earlier which prevented it from having
its experimental flag removed, however the last commits done by Claudio
removed the last known source of instability and solved a lot of
problems which were previously observed. The issues were caused by the
various coding tools interfering with the scalefactor indices. Thus,
with these problems solved, it should now be possible to declare this
encoder as the default and recommend that the users should use it
instead of others provided by external libraries, as it is both faster
and has a subjectively higher quality with selected tracks.
The encoder has still yet to be fine tuned for every possible audio file
type like music or voice, so it is hoped that with the experimental flag
removed the users should be able to provide feedback and make the
encoder better than the alternatives for every type of audio and at
every bitrate.

Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>

doc/encoders.texi: update documentation for the native AAC encoder

Since the next commit removes the experimental flag from the encoder
it's better to update the documentation which has been around in its
current form for as long as the encoder itself.

Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>

aacenc: mark coders other than twoloop as experimental

ANMR has some interesting things coming up but is currently not in a
shape fit for non-experimental usage. Same with "FAST".

Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>

aacenc: mark the "faac"-like coder for removal

This coder produces a much lower quality audio than the rest, is much
slower and is unstable. Hasn't been updated for a very long time as
well, hence it is more appropriate to remove it since it also depends on
a big burden of a code (the encode_window_bands_info function which is
just as old, just as unstable and bad and in no way modifiable or
fixable).

Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>

avfilter/af_stereotools: check s->length size

Signed-off-by: Paul B Mahol <onemda@gmail.com>

avcodec/wmaprodec: Check bits per sample to be within the range not causing integer overflows

Fixes: 549d5aab1480d10f2a775ed90b0342f1/signal_sigabrt_7ffff6ae7cc9_5643_96bbb0cfe3e28be1dadfce1075016345.wma
Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/utils: Fix overflow in get_bit_rates computations

Fixes: 129ca3e28d73af7b1e24a9d4118e7a2d/signal_sigabrt_7ffff6ae7cc9_836_762b310fc3ef6087bd7771e5d8e90b9b.asf
Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/pcm: Fix overflow in bitrate computation

Fixes: 129ca3e28d73af7b1e24a9d4118e7a2d/signal_sigabrt_7ffff6ae7cc9_836_762b310fc3ef6087bd7771e5d8e90b9b.asf
Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/wmaprodec: Fix overflow of cutoff

Fixes: 129ca3e28d73af7b1e24a9d4118e7a2d/signal_sigabrt_7ffff6ae7cc9_836_762b310fc3ef6087bd7771e5d8e90b9b.asf
Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avformat/smacker: fix integer overflow with pts_inc

Fixes: ce19e41f0ef1e52a23edc488faecdb58/asan_heap-oob_2504e97_4202_ffa0df1baed14022b9bfd4f8ac23d0cb.smk
Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avfilter/codecview: reindent after previous commit

avfilter/codecview: add QP support

Merge commit '9f57f134c19773d54269b6cb9ee455ff87c2e9e1'

* commit '9f57f134c19773d54269b6cb9ee455ff87c2e9e1':
  configure: ObjC support

Merged-by: Hendrik Leppkes <h.leppkes@gmail.com>

libavformat/mov.c: allow QuickTime metadata to come after traks

QuickTime metadata can come after trak data. Add indicator for which trak is being parsed (-1 if none) so that global metadata after the trak can be parsed.

Signed-off-by: Neil Birkbeck <neil.birkbeck@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

ffmpeg: When streamcopying, only add the input seek position when copying timestamps.

Using -ss as an input option shifts timestamps down by the seek, so it
doesn't have to be added to the recording time when checking whether to
stop.

Fixes #977

Signed-off-by: Simon Thelen <ffmpeg-dev@c-14.de>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

AAC encoder: ANMR, avoid empty search ranges

Avoid generating broken paths when some bands have an
empty search space.

aaccoder: prevent crash of anmr coder

If minq is negative, the range of sf_idx can be larger than
SCALE_MAX_DIFF allows, causing assertion failures later in
encode_scale_factors.

Reviewed-by: Claudio Freire <klaussfreire@gmail.com>
Signed-off-by: Andreas Cadhalpun <Andreas.Cadhalpun@googlemail.com>

swscale/utils: Fix for runtime error: left shift of negative value -1

Fixes: c106b36fa36db8ff8f3ed0c82be7bea2/asan_heap-oob_32699f0_6321_467b9a1d7e03d7cfd310b7e65dc53bcc.mov
Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/pgssubdec: Fix left shift of 255 by 24 places cannot be represented in type int

Fixes: b293a6479bb4b5286cff24d356bfd955/asan_generic_225c3c9_7819_cc526b657450c6cdef1371b526499626.mkv
Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

mips: add p5600 core optimizations

Signed-off-by: Vicente Olivert Riera <Vincent.Riera@imgtec.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

mips: remove --mips32r5 configure option

Having a configure option with the same name as a MIPS ISA is confusing,
so better to remove it. This option was being used to add some
optimizations to a specific core (p5600). We will add the optimizations
just when the p5600 core has been detected, in a later patch.

Signed-off-by: Vicente Olivert Riera <Vincent.Riera@imgtec.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avformat/flvdec: Fix left shift of 137 by 24 places cannot be represented in type int

Fixes: 3c857d4d90365731524716e6d051e43a/signal_sigsegv_7f4f59bcc29e_1386_20abd2c8e655cb9c75b24368e65fe3b1.flv
Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avformat/msf: Check channels when reading the header

Fixes integer overflow
Fixes: 0c2625f236ced104d402b4a03c0d65c7/asan_generic_274e1ce_5990_9314e7a67c26aecf011b178ade9f217c.avi
Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/jpeg2000dec: Check bpno in decode_cblk()

Fixes: undefined shift
Fixes: c409ef86f892335a0a164b5871174d5a/asan_heap-oob_1dff564_2159_162b7234616deab02b544410455eb07b.mov
Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avfilter: add audio emphasis filter

Signed-off-by: Paul B Mahol <onemda@gmail.com>

ffmpeg: enable echoing with command and debug modes

Allow seeing text when pressing 'c' or 'd'.

avcodec/mpegvideo: use constants for rc_strategy

swscale: use AV_OPT_TYPE_BOOL for {src,dst}_range options

swresample: use AV_OPT_TYPE_BOOL for linear_interp and cheby options

avformat: use AV_OPT_TYPE_BOOL in a bunch of places

avfilter: use AV_OPT_TYPE_BOOL in a bunch of places