ffmpeg.git
14 months agoavfilter/Makefile: add missing framesync dependency to bm3d & mix filters
Lou Logan [Wed, 6 Nov 2019 19:37:09 +0000 (10:37 -0900)]
avfilter/Makefile: add missing framesync dependency to bm3d & mix filters

Signed-off-by: Lou Logan <lou@lrcd.com>
14 months agoconfigure: select missing rdft for several filters
Lou Logan [Wed, 6 Nov 2019 19:35:28 +0000 (10:35 -0900)]
configure: select missing rdft for several filters

afir, sinc, superequalizer, surround

For afir fft is replaced with rdft as:
rdft_select="fft"

Signed-off-by: Lou Logan <lou@lrcd.com>
14 months agoavfilter/vf_dnn_processing: correct duplicate statement
leozhang [Fri, 8 Nov 2019 03:55:10 +0000 (11:55 +0800)]
avfilter/vf_dnn_processing: correct duplicate statement

Signed-off-by: leozhang <leozhang@qiyi.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
14 months agoavfilter/vf_dnn_processing: fix fate-source
Guo, Yejun [Fri, 8 Nov 2019 07:51:40 +0000 (15:51 +0800)]
avfilter/vf_dnn_processing: fix fate-source

Signed-off-by: Guo, Yejun <yejun.guo@intel.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
14 months agoavformat/mxfdec: cleanup on "essence prior to first PartitionPack"
Michael Niedermayer [Thu, 31 Oct 2019 12:30:52 +0000 (13:30 +0100)]
avformat/mxfdec: cleanup on "essence prior to first PartitionPack"

Fixes: memleak
Fixes: 18473/clusterfuzz-testcase-minimized-ffmpeg_DEMUXER_fuzzer-5738557074833408

Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
14 months agoavfilter/f_metadata: remove unneeded code
Limin Wang [Thu, 10 Oct 2019 06:03:54 +0000 (14:03 +0800)]
avfilter/f_metadata: remove unneeded code

Reviewed-by: Steven Liu <lq@onvideo.cn>
Signed-off-by: Limin Wang <lance.lmwang@gmail.com>
14 months agolavf/dashdec: drop unnecessary check before ff_format_io_close
Jun Zhao [Mon, 23 Sep 2019 16:30:17 +0000 (00:30 +0800)]
lavf/dashdec: drop unnecessary check before ff_format_io_close

ff_format_io_close will check the AVIOContext pointer pb, so drop
the unnecessary check before ff_format_io_close.

Signed-off-by: Jun Zhao <barryjzhao@tencent.com>
14 months agolavf/dashenc: enable probesize/max_analyze_duration setting in sub-demuxer
Jun Zhao [Sat, 2 Nov 2019 14:46:00 +0000 (22:46 +0800)]
lavf/dashenc: enable probesize/max_analyze_duration setting in sub-demuxer

Enable probesize/max_analyze_duration setting when open the sub-demuxer,
it's will be used to minimizing the initial delay.

Signed-off-by: Jun Zhao <barryjzhao@tencent.com>
14 months agolavf/hls: support probesize/max_analyze_duration when open sub-demuxer
Jun Zhao [Sat, 2 Nov 2019 11:48:28 +0000 (19:48 +0800)]
lavf/hls: support probesize/max_analyze_duration when open sub-demuxer

Add probesize/max_analyze_duration support when open the sub-demuxer,
it's will be used to minimizing the initial delay.

Signed-off-by: Jun Zhao <barryjzhao@tencent.com>
14 months agolavf/flvenc: Cosmetics: fix indentation
Jun Zhao [Sat, 2 Nov 2019 11:59:45 +0000 (19:59 +0800)]
lavf/flvenc: Cosmetics: fix indentation

fix indentation

Signed-off-by: Jun Zhao <barryjzhao@tencent.com>
14 months agolavf/mov: add log context dump in log message
Jun Zhao [Sun, 27 Oct 2019 02:01:30 +0000 (10:01 +0800)]
lavf/mov: add log context dump in log message

add log context dump in log message.

Reviewed-by: Michael Niedermayer <michael@niedermayer.cc>
Signed-off-by: Jun Zhao <barryjzhao@tencent.com>
14 months agolavf/hls: fix the log context setting in log message
Jun Zhao [Sun, 27 Oct 2019 02:00:24 +0000 (10:00 +0800)]
lavf/hls: fix the log context setting in log message

Fix the log context setting in log message

Signed-off-by: Jun Zhao <barryjzhao@tencent.com>
14 months agoavfilter/vf_dnn_processing: add a generic filter for image proccessing with dnn networks
Guo, Yejun [Thu, 31 Oct 2019 08:33:02 +0000 (16:33 +0800)]
avfilter/vf_dnn_processing: add a generic filter for image proccessing with dnn networks

This filter accepts all the dnn networks which do image processing.
Currently, frame with formats rgb24 and bgr24 are supported. Other
formats such as gray and YUV will be supported next. The dnn network
can accept data in float32 or uint8 format. And the dnn network can
change frame size.

The following is a python script to halve the value of the first
channel of the pixel. It demos how to setup and execute dnn model
with python+tensorflow. It also generates .pb file which will be
used by ffmpeg.

import tensorflow as tf
import numpy as np
import imageio
in_img = imageio.imread('in.bmp')
in_img = in_img.astype(np.float32)/255.0
in_data = in_img[np.newaxis, :]
filter_data = np.array([0.5, 0, 0, 0, 1., 0, 0, 0, 1.]).reshape(1,1,3,3).astype(np.float32)
filter = tf.Variable(filter_data)
x = tf.placeholder(tf.float32, shape=[1, None, None, 3], name='dnn_in')
y = tf.nn.conv2d(x, filter, strides=[1, 1, 1, 1], padding='VALID', name='dnn_out')
sess=tf.Session()
sess.run(tf.global_variables_initializer())
output = sess.run(y, feed_dict={x: in_data})
graph_def = tf.graph_util.convert_variables_to_constants(sess, sess.graph_def, ['dnn_out'])
tf.train.write_graph(graph_def, '.', 'halve_first_channel.pb', as_text=False)
output = output * 255.0
output = output.astype(np.uint8)
imageio.imsave("out.bmp", np.squeeze(output))

To do the same thing with ffmpeg:
- generate halve_first_channel.pb with the above script
- generate halve_first_channel.model with tools/python/convert.py
- try with following commands
  ./ffmpeg -i input.jpg -vf dnn_processing=model=halve_first_channel.model:input=dnn_in:output=dnn_out:fmt=rgb24:dnn_backend=native -y out.native.png
  ./ffmpeg -i input.jpg -vf dnn_processing=model=halve_first_channel.pb:input=dnn_in:output=dnn_out:fmt=rgb24:dnn_backend=tensorflow -y out.tf.png

Signed-off-by: Guo, Yejun <yejun.guo@intel.com>
Signed-off-by: Pedro Arthur <bygrandao@gmail.com>
14 months agofftools/ffmpeg_opt: Fix mixed declarations and code
Andreas Rheinhardt [Wed, 6 Nov 2019 12:18:16 +0000 (13:18 +0100)]
fftools/ffmpeg_opt: Fix mixed declarations and code

Introduced in ed3c317d.

Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
14 months agoavcodec/extract_extradata_bsf: fix typo in comments
leozhang [Wed, 6 Nov 2019 07:00:59 +0000 (15:00 +0800)]
avcodec/extract_extradata_bsf: fix typo in comments

Signed-off-by: leozhang <leozhang@qiyi.com>
Reviewed-by: Gyan <ffmpeg@gyani.pro>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
14 months agoavdevice/v4l2: Remove av_assert0 when format not supported
Andriy Gelman [Tue, 5 Nov 2019 03:32:59 +0000 (22:32 -0500)]
avdevice/v4l2: Remove av_assert0 when format not supported

Before this commit an av_assert0 would fail if a v4l2 device did not
support a target format.
For example,
./ffmpeg -f v4l2 -codec:v h264 -i /dev/video0 -f mpegts -
would signal an abort if /dev/video0 did not support h264.

The new behaviour is to return an AVERROR(EINVAL) error code. An
av_assert0 has been added to verify this return.

Fixes #6629

Signed-off-by: Andriy Gelman <andriy.gelman@gmail.com>
Reviewed-by: Nicolas George <george@nsup.org>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
14 months agoavcodec/Makefile: add missing msmpeg4* dependencies to wmv1 encoder
Lou Logan [Tue, 5 Nov 2019 19:28:07 +0000 (10:28 -0900)]
avcodec/Makefile: add missing msmpeg4* dependencies to wmv1 encoder

Signed-off-by: Lou Logan <lou@lrcd.com>
14 months agoavcodec/Makefile: add missing h263data dependency to snow encoder
Lou Logan [Tue, 5 Nov 2019 19:21:10 +0000 (10:21 -0900)]
avcodec/Makefile: add missing h263data dependency to snow encoder

Signed-off-by: Lou Logan <lou@lrcd.com>
14 months agoavcodec/Makefile: add missing vorbis_data dependency to opus encoder
Lou Logan [Tue, 5 Nov 2019 19:17:01 +0000 (10:17 -0900)]
avcodec/Makefile: add missing vorbis_data dependency to opus encoder

Signed-off-by: Lou Logan <lou@lrcd.com>
14 months agoconfigure: select mdct for dca encoder
Lou Logan [Tue, 5 Nov 2019 19:01:55 +0000 (10:01 -0900)]
configure: select mdct for dca encoder

Signed-off-by: Lou Logan <lou@lrcd.com>
14 months agoavcodec/dvdec: correctly decode bottom mb row in 1080i field mode
Baptiste Coudurier [Tue, 18 Jun 2019 20:29:41 +0000 (13:29 -0700)]
avcodec/dvdec: correctly decode bottom mb row in 1080i field mode

14 months agoavcodec/dvdec: correctly set interlaced and tff
Baptiste Coudurier [Fri, 28 Jun 2019 17:33:05 +0000 (10:33 -0700)]
avcodec/dvdec: correctly set interlaced and tff

14 months agoavcodec/libvpxdec: reject video and alpha dimension mismatches
Marton Balint [Sun, 27 Oct 2019 17:10:36 +0000 (18:10 +0100)]
avcodec/libvpxdec: reject video and alpha dimension mismatches

Signed-off-by: Marton Balint <cus@passwd.hu>
Signed-off-by: James Zern <jzern@google.com>
14 months agoavcodec/libvpxdec: decode to custom framebuffers for vp9
Marton Balint [Sun, 27 Oct 2019 17:10:35 +0000 (18:10 +0100)]
avcodec/libvpxdec: decode to custom framebuffers for vp9

This avoids copying the full frame after decoding.

Signed-off-by: Marton Balint <cus@passwd.hu>
Signed-off-by: James Zern <jzern@google.com>
14 months agoavcodec/libvpxdec: pass decoder instances to vpx_init directly
Marton Balint [Sun, 27 Oct 2019 17:10:34 +0000 (18:10 +0100)]
avcodec/libvpxdec: pass decoder instances to vpx_init directly

If the alpha decoder init failed we presented the error message from the normal
decoder. This change should prevent such mistakes.

Signed-off-by: Marton Balint <cus@passwd.hu>
Signed-off-by: James Zern <jzern@google.com>
14 months agoavcodec/libvpxdec: make sure BlockAdditional side data size >= 8
Marton Balint [Sun, 27 Oct 2019 17:10:33 +0000 (18:10 +0100)]
avcodec/libvpxdec: make sure BlockAdditional side data size >= 8

Signed-off-by: Marton Balint <cus@passwd.hu>
Signed-off-by: James Zern <jzern@google.com>
14 months agoavcodec/libvpxenc: only allocate alpha U/V planes on size changes
Marton Balint [Sat, 2 Nov 2019 16:20:21 +0000 (17:20 +0100)]
avcodec/libvpxenc: only allocate alpha U/V planes on size changes

Signed-off-by: Marton Balint <cus@passwd.hu>
Signed-off-by: James Zern <jzern@google.com>
14 months agoavcodec/libvpxenc: fix alpha stride
Marton Balint [Sun, 27 Oct 2019 17:10:31 +0000 (18:10 +0100)]
avcodec/libvpxenc: fix alpha stride

Signed-off-by: Marton Balint <cus@passwd.hu>
Signed-off-by: James Zern <jzern@google.com>
14 months agoavformat/nutenc: Do not pass NULL to memcmp() in get_needed_flags()
Michael Niedermayer [Fri, 1 Nov 2019 09:02:29 +0000 (10:02 +0100)]
avformat/nutenc: Do not pass NULL to memcmp() in get_needed_flags()

This compared to the other suggestions is cleaner and easier to understand
keeping the condition in the if() simple.

This affects alot of fate tests.

See: [FFmpeg-devel] [PATCH 05/11] avformat/nutenc: Don't pass NULL to memcmp
See: [FFmpeg-devel] [PATCH]lavf/nutenc: Do not call memcmp() with NULL argument

Fixes: Ticket 7980

Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
14 months agodoc/bitstream_filters: Fix copy an paste typo
Limin Wang [Mon, 4 Nov 2019 11:09:59 +0000 (19:09 +0800)]
doc/bitstream_filters: Fix copy an paste typo

Signed-off-by: Limin Wang <lance.lmwang@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
14 months agoconfigure: select bswapdsp for ylc decoder
Lou Logan [Mon, 4 Nov 2019 23:25:31 +0000 (14:25 -0900)]
configure: select bswapdsp for ylc decoder

Signed-off-by: Lou Logan <lou@lrcd.com>
14 months agoconfigure: select fft for qdmc decoder
Lou Logan [Mon, 4 Nov 2019 23:23:41 +0000 (14:23 -0900)]
configure: select fft for qdmc decoder

Signed-off-by: Lou Logan <lou@lrcd.com>
14 months agoconfigure: select bswapdsp for mdec decoder
Lou Logan [Mon, 4 Nov 2019 23:20:10 +0000 (14:20 -0900)]
configure: select bswapdsp for mdec decoder

Signed-off-by: Lou Logan <lou@lrcd.com>
14 months agoconfigure: select bswapdsp for imm4 decoder
Lou Logan [Mon, 4 Nov 2019 23:14:41 +0000 (14:14 -0900)]
configure: select bswapdsp for imm4 decoder

Signed-off-by: Lou Logan <lou@lrcd.com>
14 months agoavcodec/Makefile: add missing ass dependency to ccaption decoder
Lou Logan [Mon, 4 Nov 2019 23:12:24 +0000 (14:12 -0900)]
avcodec/Makefile: add missing ass dependency to ccaption decoder

Signed-off-by: Lou Logan <lou@lrcd.com>
14 months agoconfigure: select audiodsp for acelp_kelvin decoder
Lou Logan [Mon, 4 Nov 2019 23:10:11 +0000 (14:10 -0900)]
configure: select audiodsp for acelp_kelvin decoder

Signed-off-by: Lou Logan <lou@lrcd.com>
14 months agoconfigure: select mdct for atrac3al decoder
Lou Logan [Mon, 4 Nov 2019 23:06:22 +0000 (14:06 -0900)]
configure: select mdct for atrac3al decoder

Signed-off-by: Lou Logan <lou@lrcd.com>
14 months agoconfigure: select mdct & sinewin for atrac3pal decoder
Lou Logan [Mon, 4 Nov 2019 23:01:45 +0000 (14:01 -0900)]
configure: select mdct & sinewin for atrac3pal decoder

Signed-off-by: Lou Logan <lou@lrcd.com>
14 months agoavformat/Makefile: add missing pcm dependency to sdx demuxer
Lou Logan [Fri, 1 Nov 2019 22:18:35 +0000 (14:18 -0800)]
avformat/Makefile: add missing pcm dependency to sdx demuxer

Signed-off-by: Lou Logan <lou@lrcd.com>
Signed-off-by: Carl Eugen Hoyos <ceffmpeg@gmail.com>
14 months agoavformat/Makefile: add missing pcm dependency to nsp demuxer
Lou Logan [Fri, 1 Nov 2019 22:14:44 +0000 (14:14 -0800)]
avformat/Makefile: add missing pcm dependency to nsp demuxer

Signed-off-by: Lou Logan <lou@lrcd.com>
Signed-off-by: Carl Eugen Hoyos <ceffmpeg@gmail.com>
14 months agoavformat/Makefile: add missing pcm dependency to hcom demuxer
Lou Logan [Fri, 1 Nov 2019 21:57:38 +0000 (13:57 -0800)]
avformat/Makefile: add missing pcm dependency to hcom demuxer

Signed-off-by: Lou Logan <lou@lrcd.com>
Signed-off-by: Carl Eugen Hoyos <ceffmpeg@gmail.com>
14 months agoconfigure: select riffdec for act demuxer
Lou Logan [Fri, 1 Nov 2019 21:50:33 +0000 (13:50 -0800)]
configure: select riffdec for act demuxer

Signed-off-by: Lou Logan <lou@lrcd.com>
Signed-off-by: Carl Eugen Hoyos <ceffmpeg@gmail.com>
14 months agoconfigure: select jpegtables for rtp muxer
Lou Logan [Fri, 1 Nov 2019 21:35:53 +0000 (13:35 -0800)]
configure: select jpegtables for rtp muxer

Signed-off-by: Lou Logan <lou@lrcd.com>
Signed-off-by: Carl Eugen Hoyos <ceffmpeg@gmail.com>
14 months agotools/enum_options: replace the deprecated API
Zhong Li [Sat, 2 Nov 2019 15:47:01 +0000 (23:47 +0800)]
tools/enum_options: replace the deprecated API

Signed-off-by: Zhong Li <zhongli_dev@126.com>
14 months agotools/probetest: replace the deprecated API
Zhong Li [Sat, 2 Nov 2019 15:47:00 +0000 (23:47 +0800)]
tools/probetest: replace the deprecated API

Signed-off-by: Zhong Li <zhongli_dev@126.com>
14 months agolavc/qsvenc: add the missing vp9 file
zhongli_dev@126.com [Mon, 4 Nov 2019 07:11:38 +0000 (07:11 +0000)]
lavc/qsvenc: add the missing vp9 file

It is missed in commit 3358380

Signed-off-by: Zhong Li <zhongli_dev@126.com>
14 months agolavf/asfdec: Simplify the check conditions
Jun Zhao [Sat, 15 Jun 2019 08:21:56 +0000 (16:21 +0800)]
lavf/asfdec: Simplify the check conditions

Simplify the check conditions

Signed-off-by: Jun Zhao <barryjzhao@tencent.com>
14 months agolavf/mpegtsenc: fix logic check error
Jun Zhao [Sat, 15 Jun 2019 08:19:18 +0000 (16:19 +0800)]
lavf/mpegtsenc: fix logic check error

fix the logic check error

fix #6751

Reviewed-by: Andriy Gelman <andriy.gelman@gmail.com>
Signed-off-by: Jun Zhao <barryjzhao@tencent.com>
14 months agoavformat/ftp: add AVOptions for authentication
Nicolas Frattaroli [Sat, 26 Oct 2019 22:37:32 +0000 (00:37 +0200)]
avformat/ftp: add AVOptions for authentication

This introduces two new AVOption options for the FTP protocol,
one named ftp-user to supply the username to be used for auth,
one named ftp-password to supply the password to be used for auth.

These are useful for when an API user does not wish to deal with
URL manipulation and percent encoding.

Setting them while also having credentials in the URL will use the
credentials from the URL. The rationale for this is that credentials
embedded in the URL are probably more specific to what the user is
trying to do than anything set by some API user.

Signed-off-by: Nicolas Frattaroli <ffmpeg@fratti.ch>
Signed-off-by: Marton Balint <cus@passwd.hu>
14 months agoavformat: Add max_probe_packets option
Andriy Gelman [Thu, 17 Oct 2019 14:49:20 +0000 (10:49 -0400)]
avformat: Add max_probe_packets option

Allows user to set maximum number of buffered packets when
probing a codec. It was a hard-coded parameter before this commit.

Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
14 months agolavc/qsvenc: enable vp9 encoder
Zhong Li [Sat, 26 Oct 2019 14:18:31 +0000 (22:18 +0800)]
lavc/qsvenc: enable vp9 encoder

1. must enable low_power mode since just VDENC can be supported by iHD
driver right now
2. Coding option1 and extra_data are not supported by MSDK
3. IVF header will be inserted in MSDK by default, but it is not needed
for FFmpeg, so disable it.

Signed-off-by: Zhong Li <zhongli_dev@126.com>
14 months agolavc/qsvenc: Fix bitrate_limit to allow AVC encode in limited bitrate
Linjie Fu [Tue, 15 Oct 2019 07:56:23 +0000 (15:56 +0800)]
lavc/qsvenc: Fix bitrate_limit to allow AVC encode in limited bitrate

MFXVideoENCODE_Query calls CheckVideoParamQueryLike in MSDK and
will determine whether to set param.mfx.TargetKbps to the allowed
minTargetKbps according to the bitrate_limit in extco2 buffer.

Thus q->param.ExtParam must be set before MFXVideoENCODE_Query in case
minTargetKbps is written to TargetKbps by default.

1080P AVC encoding with option "-bitrate_limit 0 -b:v 100k":
Before patch:
            902 kbps
After patch:
            156 kbps
Signed-off-by: Linjie Fu <linjie.fu@intel.com>
14 months agoavcodec/v4l2_m2m: Fix indentation
Andriy Gelman [Sun, 27 Oct 2019 04:19:47 +0000 (00:19 -0400)]
avcodec/v4l2_m2m: Fix indentation

Signed-off-by: Aman Gupta <aman@tmm1.net>
14 months agoavcodec/v4l2_buffers: Fix infinite loop
Andriy Gelman [Sun, 27 Oct 2019 04:19:46 +0000 (00:19 -0400)]
avcodec/v4l2_buffers: Fix infinite loop

This part of the code counts the number of planes returned by the v4l2
device for each queried capture/output buffer.
When testing the GPU h264 encoder on Nvidia's Jetson Nano, this caused an
infinite loop because avbuf->buf.length included some empty buffers (i.e.
where avbuf->buf.m.planes[i].length = 0), meaning that the counter was
never incremented and break was never reached.
This is fixed in the commit by using a well defined iteration range.

Signed-off-by: Aman Gupta <aman@tmm1.net>
14 months agotools/target_dec_fuzzer: Adjust threshold for G2M
Michael Niedermayer [Wed, 9 Oct 2019 08:09:22 +0000 (10:09 +0200)]
tools/target_dec_fuzzer: Adjust threshold for G2M

G2M allows large images from small input and also reallocates
multiple buffers on any resolution change.

Fixes: Timeout (22sec -> 5sec)
Fixes: 18022/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_G2M_fuzzer-5089192530411520

Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
14 months agoavcodec/aptx: Check the number of channels
Michael Niedermayer [Wed, 9 Oct 2019 07:53:26 +0000 (09:53 +0200)]
avcodec/aptx: Check the number of channels

Fixes: store to null pointer of type 'uint32_t' (aka 'unsigned int')
Fixes: 18021/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_APTX_HD_fuzzer-5761738313564160

Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
14 months agoavcodec/rv10: Check input for minimal frame size
Michael Niedermayer [Wed, 9 Oct 2019 07:08:38 +0000 (09:08 +0200)]
avcodec/rv10: Check input for minimal frame size

Fixes: Timeout (18sec -> 4sec)
Fixes: 18012/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_RV20_fuzzer-5767486145822720

Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
14 months agoavcodec/aacdec_template: Check samplerate
Michael Niedermayer [Tue, 8 Oct 2019 20:52:56 +0000 (22:52 +0200)]
avcodec/aacdec_template: Check samplerate

Fixes: signed integer overflow: 2 * 1881153568 cannot be represented in type 'int'
Fixes: 17996/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_AAC_FIXED_fuzzer-5687126468853760

Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
14 months agoavcodec/truemotion2: Fix several integer overflows in tm2_low_res_block()
Michael Niedermayer [Sat, 26 Oct 2019 19:39:41 +0000 (21:39 +0200)]
avcodec/truemotion2: Fix several integer overflows in tm2_low_res_block()

Fixes: signed integer overflow: 1077952576 + 1355863565 cannot be represented in type 'int'
Fixes: 16196/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_TRUEMOTION2_fuzzer-5679842317565952

Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
14 months agoavcodec/truemotion2: Fix several integer overflows with *Yo, *Uo, *Vo
Michael Niedermayer [Sat, 26 Oct 2019 19:34:44 +0000 (21:34 +0200)]
avcodec/truemotion2: Fix several integer overflows with *Yo, *Uo, *Vo

Fixes: signed integer overflow: 538976288 - -2080374792 cannot be represented in type 'int'
Fixes: 16196/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_TRUEMOTION2_fuzzer-5144044274974720

Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
14 months agoavcodec/utils: Check block_align
Michael Niedermayer [Thu, 31 Oct 2019 08:31:22 +0000 (09:31 +0100)]
avcodec/utils: Check block_align

Fixes: out of array access
Fixes: 18432/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_WMAV2_fuzzer-5675574936207360
Fixes: 18326/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_WMAV2_fuzzer-5071752362721280
Fixes: 18384/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_WMAV1_fuzzer-5769439500304384

Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Reviewed-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
14 months agoavfilter/vf_lut3d: simplify code
Limin Wang [Wed, 30 Oct 2019 13:20:55 +0000 (21:20 +0800)]
avfilter/vf_lut3d: simplify code

Signed-off-by: Limin Wang <lance.lmwang@gmail.com>
Reviewed-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
14 months agoswscale/swscale_unscaled: fix gbrap10be md5 different on big endian system
Limin Wang [Wed, 30 Oct 2019 13:20:52 +0000 (21:20 +0800)]
swscale/swscale_unscaled: fix gbrap10be md5 different on big endian system

You can reproduce it by below command:
./ffmpeg -f lavfi -i "testsrc=duration=1:rate=30" -vf format=gbrap10 -vcodec rawvideo \
    -pix_fmt gbrap10le -flags +bitexact -sws_flags +accurate_rnd+bitexact -fflags +bitexact  \
    -frames:v 1 -f nut md5:

little-endian:
f91e2edd8098276579c1929e5e160416
big-endian:
ba4d011dbbdc78ccbf6cc7d698630929

Signed-off-by: Limin Wang <lance.lmwang@gmail.com>
Reviewed-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
14 months agolibavformat/flacenc: reject too big picture blocks
Mattias Wadman [Wed, 30 Oct 2019 13:01:28 +0000 (14:01 +0100)]
libavformat/flacenc: reject too big picture blocks

A too big picture will case the muxer to write a truncated block size (uint24)
causing the output file to be corrupt.

How to reproduce:

Write a file with truncated block size:
ffmpeg -y -f lavfi -i sine -f lavfi -i color=red:size=2400x2400 -map 0:a:0 -map 1:v:0 -c:v:0 bmp -disposition:1 attached_pic -t 1 test.flac

Try to decode:
ffmpeg -i test.flac test.wav

Signed-off-by: Mattias Wadman <mattias.wadman@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
14 months agoavcodec/libtwolame: fix mono default bitrate
James Cowgill [Fri, 1 Nov 2019 08:51:07 +0000 (08:51 +0000)]
avcodec/libtwolame: fix mono default bitrate

As of libtwolame 0.4.0, 384 kbps is not accepted as a valid bitrate
for encoding mono audio and the maximum bitrate is now halved to 192
kbps to comply with the MP2 standard. Example error:

    twolame_init_params(): 384kbps is an invalid bitrate for mono encoding.

Adjust the default bitrate calculation to take this into account.

Signed-off-by: James Cowgill <jcowgill@debian.org>
14 months agodoc/filters: fix overlay_opencl document typo
Steven Liu [Fri, 1 Nov 2019 08:42:36 +0000 (16:42 +0800)]
doc/filters: fix overlay_opencl document typo

Reported-by: Yabo Wei <weiyabo@kuaishou.com>
Signed-off-by: Steven Liu <lq@chinaffmpeg.org>
14 months agoavfilter/vf_median: add radiusV option
Paul B Mahol [Wed, 30 Oct 2019 09:14:23 +0000 (10:14 +0100)]
avfilter/vf_median: add radiusV option

14 months agoavfilter/af_afade: start crossfading only when first stream reached end
Paul B Mahol [Thu, 31 Oct 2019 09:22:05 +0000 (10:22 +0100)]
avfilter/af_afade: start crossfading only when first stream reached end

14 months agoavformat/hlsenc: fix the av_dirname path isn't include separator in the end of string
Limin Wang [Wed, 23 Oct 2019 15:55:15 +0000 (23:55 +0800)]
avformat/hlsenc: fix the av_dirname path isn't include separator in the end of string

Reviewed-by: Liu Steven <lq@chinaffmpeg.org>
Signed-off-by: Limin Wang <lance.lmwang@gmail.com>
14 months agoconfigure: fix standalone compilation of mxf muxer
Lou Logan [Wed, 30 Oct 2019 00:11:46 +0000 (16:11 -0800)]
configure: fix standalone compilation of mxf muxer

Signed-off-by: Lou Logan <lou@lrcd.com>
Signed-off-by: Paul B Mahol <onemda@gmail.com>
14 months agoavformat/apngenc: Add deinit function
Andreas Rheinhardt [Wed, 23 Oct 2019 12:59:41 +0000 (14:59 +0200)]
avformat/apngenc: Add deinit function

Prevents memleaks when the trailer is never written (e.g. when there was
a write error when writing the header).

Fixes ticket #8347.

Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
Signed-off-by: James Almer <jamrial@gmail.com>
14 months agoavfilter/af_afade: check for eof after crossfade later
Paul B Mahol [Wed, 30 Oct 2019 18:07:19 +0000 (19:07 +0100)]
avfilter/af_afade: check for eof after crossfade later

Fixes memleaks and #8346

14 months agoavcodec/mips: msa optimizations for vc1dsp
gxw [Mon, 21 Oct 2019 07:56:47 +0000 (15:56 +0800)]
avcodec/mips: msa optimizations for vc1dsp

Performance of WMV3 decoding has speed up from 3.66x to 5.23x tested on 3A4000.

Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
14 months agoavfilter/f_sidedata: fix Wtautological-constant-out-of-range-compare
Zhao Zhili [Tue, 15 Oct 2019 17:17:54 +0000 (01:17 +0800)]
avfilter/f_sidedata: fix Wtautological-constant-out-of-range-compare

Reviewed-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
14 months agoFATE: add a test for freeezedetect
Limin Wang [Mon, 21 Oct 2019 10:12:13 +0000 (18:12 +0800)]
FATE: add a test for freeezedetect

Signed-off-by: Limin Wang <lance.lmwang@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
14 months agodoc/filters: move drawgraph to video filters chapter
Paul B Mahol [Wed, 30 Oct 2019 16:55:44 +0000 (17:55 +0100)]
doc/filters: move drawgraph to video filters chapter

14 months agodoc/filters: move agraphmonitor to multimedia filters chapter
Paul B Mahol [Wed, 30 Oct 2019 16:47:05 +0000 (17:47 +0100)]
doc/filters: move agraphmonitor to multimedia filters chapter

14 months agoavfilter/vf_sr: correct flags since the filter changes frame w/h
Guo, Yejun [Mon, 28 Oct 2019 13:11:02 +0000 (21:11 +0800)]
avfilter/vf_sr: correct flags since the filter changes frame w/h

If filter changes frame w/h, AVFILTER_FLAG_SUPPORT_TIMELINE_GENERIC
cannot be supported.

Signed-off-by: Guo, Yejun <yejun.guo@intel.com>
Signed-off-by: Pedro Arthur <bygrandao@gmail.com>
14 months agoavfilter/dnn: add a new interface to query dnn model's input info
Guo, Yejun [Mon, 21 Oct 2019 12:38:17 +0000 (20:38 +0800)]
avfilter/dnn: add a new interface to query dnn model's input info

to support dnn networks more general, we need to know the input info
of the dnn model.

background:
The data type of dnn model's input could be float32, uint8 or fp16, etc.
And the w/h of input image could be fixed or variable.

Signed-off-by: Guo, Yejun <yejun.guo@intel.com>
Signed-off-by: Pedro Arthur <bygrandao@gmail.com>
14 months agoavfilter/dnn: get the data type of network output from dnn execution result
Guo, Yejun [Mon, 21 Oct 2019 12:38:10 +0000 (20:38 +0800)]
avfilter/dnn: get the data type of network output from dnn execution result

so,  we can make a filter more general to accept different network
models, by adding a data type convertion after getting data from network.

After we add dt field into struct DNNData, it becomes the same as
DNNInputData, so merge them with one struct: DNNData.

Signed-off-by: Guo, Yejun <yejun.guo@intel.com>
Signed-off-by: Pedro Arthur <bygrandao@gmail.com>
14 months agodnn: add tf.nn.conv2d support for native model
Guo, Yejun [Mon, 21 Oct 2019 12:38:03 +0000 (20:38 +0800)]
dnn: add tf.nn.conv2d support for native model

Unlike other tf.*.conv2d layers, tf.nn.conv2d does not create many
nodes (within a scope) in the graph, it just acts like other layers.
tf.nn.conv2d only creates one node in the graph, and no internal
nodes such as 'kernel' are created.

The format of native model file is also changed, a flag named
has_bias is added, so change the version number.

Signed-off-by: Guo, Yejun <yejun.guo@intel.com>
Signed-off-by: Pedro Arthur <bygrandao@gmail.com>
14 months agodoc: fix typo in muxers documentation
Alfred E. Heggestad [Wed, 30 Oct 2019 09:48:44 +0000 (10:48 +0100)]
doc: fix typo in muxers documentation

14 months agoavformat/hls: Set AVFMT_TS_DISCONT flag on HLS input format
Philip Langdale [Sun, 27 Oct 2019 18:39:08 +0000 (11:39 -0700)]
avformat/hls: Set AVFMT_TS_DISCONT flag on HLS input format

There have been many reports over the years about problems when
taking an HLS stream as input to `ffmpeg` where there are timestamp
discontinuities present. This is explicitly supported in the
HLS spec (EXT-X-DISCONTINUITY) and often used for ad injection.

Various fixes and work-arounds have been proposed over the years,
but one step that seems obvious, even if it's not a complete fix,
is to mark the HLS input format as supporting discontinuities. This
will prevent timestamp fixup logic in ffmpeg.c kicking in that ends
up mangling the timestamps unnecessarily.

I've tested this out with an example provided by Joe Koberg early
last year, and it is sufficient to allow `ffmpeg` to download and
mux the stream correctly. Joe had briefly suggested that other
situations can still be handled incorrectly, but this seems like
a strict improvement.

Joe's example:

https://s3.amazonaws.com/playon-test-videos/discont_test_new/discont_test.m3u8

Reviewed-by: Steven Liu <lq@onvideo.cn>
Reviewed-by: Dennis Mungai <dmngaie@gmail.com>
14 months agovp4: prevent unaligned memory access in loop filter
Peter Ross [Thu, 20 Jun 2019 21:52:31 +0000 (07:52 +1000)]
vp4: prevent unaligned memory access in loop filter

VP4 applies a loop filter during motion compensation, causing the block offset
will often by unaligned. This produces a bus error on some platforms, namely
ARMv7 NEON.

This patch adds a unaligned version of the loop filter function pointer
to VP3DSPContext.

Reported-by: Mike Melanson <mike@multimedia.cx>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
14 months agoavcodec/wmalosslessdec: Fix some integer anomalies
Michael Niedermayer [Sun, 27 Oct 2019 23:12:59 +0000 (00:12 +0100)]
avcodec/wmalosslessdec: Fix some integer anomalies

Fixes: left shift of negative value -341180
Fixes: 18401/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_WMALOSSLESS_fuzzer-5686380134400000

Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Reviewed-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
14 months agoavcodec/adpcm: Fix invalid shifts in ADPCM DTK
Michael Niedermayer [Sun, 27 Oct 2019 22:38:47 +0000 (23:38 +0100)]
avcodec/adpcm: Fix invalid shifts in ADPCM DTK

Fixes: left shift of negative value -1
Fixes: 18397/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_ADPCM_DTK_fuzzer-5675653487132672

Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Reviewed-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
14 months agoavcodec/apedec: Only clear the needed buffer space, instead of all
Michael Niedermayer [Sun, 27 Oct 2019 22:26:52 +0000 (23:26 +0100)]
avcodec/apedec: Only clear the needed buffer space, instead of all

Fixes: Timeout (15sec -> 0.4sec)
Fixes: 18396/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_APE_fuzzer-5730080487112704

Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Reviewed-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
14 months agoavfilter/asrc_anoisesrc: change color variable to int
Limin Wang [Sat, 26 Oct 2019 13:23:24 +0000 (21:23 +0800)]
avfilter/asrc_anoisesrc: change color variable to int

Or it'll cause invalid color and s->filter is NULL.
Please reproduce it with below command on big endian system:
$ ./ffmpeg -f lavfi -i "anoisesrc=d=60:c=1:r=48000" -f s16le -c:a pcm_s16le  -f
null  -
Segmentation fault (core dumped)

Signed-off-by: Limin Wang <lance.lmwang@gmail.com>
Reviewed-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
14 months agoavcodec/libvorbisdec: Fix insufficient input checks leading to out of array reads
Michael Niedermayer [Sat, 26 Oct 2019 21:58:02 +0000 (23:58 +0200)]
avcodec/libvorbisdec: Fix insufficient input checks leading to out of array reads

Fixes: 16144/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_LIBVORBIS_fuzzer-5638618940440576
Fixes: out of array read

Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
14 months agoavcodec/g723_1dec: fix invalid shift with negative sid_gain
Michael Niedermayer [Sat, 26 Oct 2019 19:56:03 +0000 (21:56 +0200)]
avcodec/g723_1dec: fix invalid shift with negative sid_gain

Fixes: left shift of negative value -1
Fixes: 18395/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_G723_1_fuzzer-5710313034350592

Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Reviewed-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
14 months agoavformat/dss: Remove superfluous headers
Andreas Rheinhardt [Sat, 26 Oct 2019 04:59:57 +0000 (06:59 +0200)]
avformat/dss: Remove superfluous headers

Both attributes.h and bswap.h have been included from the very beginning
of this muxer without there being any reason to do so.

Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
14 months agoavcodec/tiff: check the black level denominator
James Almer [Sat, 26 Oct 2019 15:01:16 +0000 (12:01 -0300)]
avcodec/tiff: check the black level denominator

Fixes ticket #8327.

Reviewed-by: Michael Niedermayer <michael@niedermayer.cc>
Signed-off-by: James Almer <jamrial@gmail.com>
14 months agoavcodec/nvenc: turn feature check failures into warnings
hydra3333 [Sun, 27 Oct 2019 02:30:54 +0000 (19:30 -0700)]
avcodec/nvenc: turn feature check failures into warnings

Signed-off-by: Timo Rothenpieler <timo@rothenpieler.org>
14 months agoavfilter/vf_vfrdet: also report average delta
Paul B Mahol [Tue, 29 Oct 2019 12:05:31 +0000 (13:05 +0100)]
avfilter/vf_vfrdet: also report average delta

14 months agoavfilter/vf_vfrdet: fix reporting max delta
Paul B Mahol [Tue, 29 Oct 2019 11:57:58 +0000 (12:57 +0100)]
avfilter/vf_vfrdet: fix reporting max delta

If only first delta was big it was previously discarded.

14 months agoavfilter: add median filter
Paul B Mahol [Tue, 15 Jan 2019 10:32:49 +0000 (11:32 +0100)]
avfilter: add median filter

14 months agoavformat/wc3movie: fix memleak when read content size not equvipe input size
Steven Liu [Thu, 10 Oct 2019 02:07:56 +0000 (10:07 +0800)]
avformat/wc3movie: fix memleak when read content size not equvipe input size

Signed-off-by: Steven Liu <lq@chinaffmpeg.org>
14 months agoavformat/swfdec: fix memleak when inflateInit failed
Steven Liu [Thu, 10 Oct 2019 02:07:55 +0000 (10:07 +0800)]
avformat/swfdec: fix memleak when inflateInit failed

Signed-off-by: Steven Liu <lq@chinaffmpeg.org>
14 months agoavformat/smacker: fix memleak when avformat_new_stream failed
Steven Liu [Thu, 10 Oct 2019 02:07:54 +0000 (10:07 +0800)]
avformat/smacker: fix memleak when avformat_new_stream failed

Signed-off-by: Steven Liu <lq@chinaffmpeg.org>
14 months agoavformat/rtmpproto: fix memleak when open rtmp failed
Steven Liu [Thu, 10 Oct 2019 02:07:53 +0000 (10:07 +0800)]
avformat/rtmpproto: fix memleak when open rtmp failed

Signed-off-by: Steven Liu <lq@chinaffmpeg.org>