Guo, Yejun [Wed, 7 Aug 2019 02:44:27 +0000 (10:44 +0800)]
FATE/dnn: let fate/dnn tests depend on ffmpeg static libraries
background:
DNN (deep neural network) is a sub module of libavfilter, and FATE/dnn
is unit test for the DNN module, one unit test for one dnn layer.
The unit tests are not based on the APIs exported by libavfilter,
they just directly call into the functions within DNN submodule.
There is an issue when run the following command:
build$ ../ffmpeg/configure --disable-static --enable-shared
make
make fate-dnn-layer-pad
And part of error message:
tests/dnn/dnn-layer-pad-test.o: In function `test_with_mode_symmetric':
/work/media/ffmpeg/build/src/tests/dnn/dnn-layer-pad-test.c:73: undefined reference to `dnn_execute_layer_pad'
The root cause is that function dnn_execute_layer_pad is a LOCAL symbol
in libavfilter.so, and so the linker could not find it when build dnn-layer-pad-test.
To check it, just run: readelf -s libavfilter/libavfilter.so | grep dnn
So, add dependency in fate/dnn Makefile with ffmpeg static libraries.
This is the same method used in fate/checkasm
Signed-off-by: Guo, Yejun <yejun.guo@intel.com> Signed-off-by: Pedro Arthur <bygrandao@gmail.com>
Guo, Yejun [Fri, 16 Aug 2019 02:36:56 +0000 (10:36 +0800)]
libavfilter/dnn/dnn_backend_tf: fix typo that variable uninitialized.
if it is initialized randomly, the tensorflow lib will report
error message such as:
Attempt to add output -7920 of depth_to_space4 not in range [0, 1) to node with type Identity
Signed-off-by: Guo, Yejun <yejun.guo@intel.com> Signed-off-by: Pedro Arthur <bygrandao@gmail.com>
avcodec/anm: Check input size for a frame with just a stop code
Fixes: Timeout (11sec -> 6sec) Fixes: 16344/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_ANM_fuzzer-5673032000995328 Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg Reviewed-by: Peter Ross <pross@xvid.org> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Marton Balint [Mon, 5 Aug 2019 19:50:42 +0000 (21:50 +0200)]
avformat/avio: add avio_print_string_array and avio_print
These functions can be used to print a variable number of strings consecutively
to the IO context. Unlike av_bprintf, no temporary buffer is necessary.
Limin Wang [Tue, 13 Aug 2019 01:39:47 +0000 (09:39 +0800)]
avfilter/f_select: yuv will use Y plane only for scenecut detect
At the moment scene change detection score uses all planes to detect scene
changes. In this regard this is similar how the frozen frames detection works.
However, in classic encoding scene change detection typically only uses the Y
plane.
We might get more resonable scores for scene change if we also use only
the Y plane for calculating the score if the pixel format is YUV. Although
this will require additional work once packed YUV formats are added,
because for the moment the generic scene sad score calculation has no way
to ignore some components in a packed format.
Signed-off-by: Limin Wang <lance.lmwang@gmail.com> Signed-off-by: Marton Balint <cus@passwd.hu>
matroska_reset_status (a function that is used during seeking (among
other things)) used an int for the return value of avio_seek which
returns an int64_t. Checking the return value then indicated an error
even though the seek was successfull for targets in the range of
2GB-4GB, 6GB-8GB, ... This error implied that the status hasn't been
reset and in particular, the old level was still considered to be in
force, so that ebml_parse returned errors because the newly parsed
elements were of course not contained in the previously active and still
wrongly considered active master element any more.
Addresses ticket #8084.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com> Signed-off-by: James Almer <jamrial@gmail.com>
convert_from_tensorflow.py: support conv2d with dilation
conv2d with dilation > 1 generates tens of nodes in graph, it is not
easy to parse each node one by one, so we do special tricks to parse
the conv2d layer.
Signed-off-by: Guo, Yejun <yejun.guo@intel.com> Signed-off-by: Pedro Arthur <bygrandao@gmail.com>
tomajsjiang [Wed, 7 Aug 2019 14:48:28 +0000 (22:48 +0800)]
lavf/utils: fix error like "offset 0x1f85: partial file"
fix error like "offset 0x1f85: partial file", the root cause is when
read the mp4 file from http, and the moov in the end of the mp4 file,
reconfig the buffer will drop some data.
Signed-off-by: Jun Zhao <barryjzhao@tencent.com> Signed-off-by: Zhongxing Jiang <tomajsjiang@tencent.com>
ff_write_chained essentially duplicated the functionality of
av_packet_rescale_ts. This has been changed.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com> Reviewed-by: Paul B Mahol <onemda@gmail.com> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
avformat/mux: Use const AVPacket * in compare functions
There is no reason for these functions to modify the given packets at
all.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com> Reviewed-by: Paul B Mahol <onemda@gmail.com> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
tools/target_dec_fuzzer: adjust pixel threshold for MSRLE, as it allows coding gigantic images on tiny input
Fixes: Timeout (12sec ->2sec) Fixes: 16125/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_MSRLE_fuzzer-5650846364205056 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>
fftools: Use right function signature and pointers
The option tables of the various fftools (in particular ffprobe) are
arrays of OptionDef; said type contains a union of a pointer to void and
a function pointer of type int (*)(void *, const char *, const char *)
as well as a size_t. Some entries (namely the common entry for writing a
report as well as several more of ffprobe's entries) used the pointer to
void to store a pointer to functions of type int (*)(const char *) or
type int (*)(const char *, const char *); nevertheless, when the functions
are actually called in write_option (in cmdutils.c), it is done via a
pointer of the first type.
There are two things wrong here:
1. Pointer to void can be converted to any pointer to incomplete or
object type and back; but they are nevertheless not completely generic
pointers: There is no provision in the C standard that guarantees their
convertibility with function pointers. C90 lacks a generic function
pointer, C99 made every function pointer a generic function pointer and
still disallows the convertibility with void *.
2. The signature of the called function differs from the signature
of the pointed-to type. This is undefined behaviour in C99 (given that
C90 lacks a way to convert function pointers at all, it doesn't say
anything about such a situation). It only works because none of the
functions this patch is about make any use of their parameters at all.
Therefore this commit changes the type of the relevant functions
to match the type used for the call and uses the union's function
pointer to store it. This is legal even in C90.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com> Reviewed-by: Paul B Mahol <onemda@gmail.com> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
avcodec/ffwavesynth: Fixes invalid shift with pink noise seeking
Fixes: left shift of negative value -961533698048 Fixes: 16242/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_FFWAVESYNTH_fuzzer-5738550670131200 Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg Reviewed-by: Nicolas George <george@nsup.org> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
avcodec/ffwavesynth: Fix integer overflow for some corner case values
Fixes: left shift of negative value -14671840 Fixes: 16000/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_FFWAVESYNTH_fuzzer-5145977817661440 Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Marton Balint [Sun, 4 Aug 2019 19:52:35 +0000 (21:52 +0200)]
avformat/mpegtsenc: fix PCR generation intervals
PCR generation was based on counting packets for both CBR and VBR streams.
Couting packets might have worked for CBR streams (when muxrate was specified)
but it only took into account the packets of a service (or the packets of the
PCR stream lately), so even that was problematic for multi program streams.
The new code works on actual timestamps for both CBR and VBR streams. For VBR
streams the behaviour of the old code is simulated by selecting a PCR interval
which is the highest multiple of the frame duration but still less than 100 ms.
It should be trivial to add support for setting the PCR interval for VBR
streams as well in a later patch.
The accuracy of PCR packets for CBR streams was greatly improved by preemtively
sending them at PCR intervals even if sending the payload of another stream
is in progress.
Marton Balint [Fri, 2 Aug 2019 07:46:51 +0000 (09:46 +0200)]
avformat/mpegtsenc: fix incorrect PCR selection with multiple programs
The MPEG-TS muxer had a serious bug related to the use of multiple programs:
in that case, the PCR pid selection was incomplete for all services except one.
This patch solves this problem and selects a stream to become PCR for each
service, preferably the video stream.
This patch also moves pcr calculation attributes to MpegTSWriteStream from
MpegTSService. PCR is a per-stream and not per-service thing, so it was
misleading to refer to it as something that is per-service.
Also remove *service from MpegTSWriteStream because a stream can belong to
multiple services so it was misleading to select one for each stream.
You can check the result with this example command:
[mpegts @ 0x37505c0] service 1 using PCR in pid=256
[mpegts @ 0x37505c0] service 2 using PCR in pid=257
[mpegts @ 0x37505c0] service 3 using PCR in pid=258
[mpegts @ 0x37505c0] service 4 using PCR in pid=256
Fixes ticket #8039.
v2: a video is stream is preferred if there are no programs, just like before
the patch.
Marton Balint [Tue, 13 Aug 2019 21:05:44 +0000 (23:05 +0200)]
ffplay: properly detect all window size changes
SDL_WINDOWEVENT_SIZE_CHANGED should be used instead of SDL_WINDOWEVENT_RESIZED
because SDL_WINDOWEVENT_RESIZED is only emitted if the resize happened due to
an external event.
98e419cb added SIMD for the convolution filter for x64 systems. As
usual, it used a check of the form
if (ARCH_X86_64)
ff_convolution_init_x86(s);
and thereby relied on the compiler eliminating this pseudo-runtime check
at compiletime for non x64 systems (for which ff_convolution_init_x86
isn't defined) to compile. But vf_convolution.c contains more than one
filter and if the convolution filter is disabled, but one of the other
filters (prewitt, sobel, roberts) is enabled, the build will fail on x64,
because ff_convolution_init_x86 isn't defined in this case.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
Lars Kiesow [Mon, 12 Aug 2019 13:58:14 +0000 (15:58 +0200)]
libavfilter/vf_scale: Ensure scaled video is divisible by n
This patch adds a new option to the scale filter which ensures that the
output resolution is divisible by the given integer when used together
with `force_original_aspect_ratio`. This works similar to using `-n` in
the `w` and `h` options.
This option respects the value set for `force_original_aspect_ratio`,
increasing or decreasing the resolution accordingly.
The use case for this is to set a fixed target resolution using `w` and
`h`, to use the `force_original_aspect_ratio` option to make sure that
the video always fits in the defined bounding box regardless of aspect
ratio, but to also make sure that the calculated output resolution is
divisible by n so in can be encoded with certain encoders/options if
that is required.
Signed-off-by: Lars Kiesow <lkiesow@uos.de> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
avcodec/vp56: Consider the alpha start as end of the prior header
Fixes: Timeout (23sec -> 71ms) Fixes: 15661/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_VP6A_fuzzer-6257865947348992 Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg Reviewed-by: Peter Ross <pross@xvid.org> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
gxw [Wed, 7 Aug 2019 09:52:00 +0000 (17:52 +0800)]
avutil/mips: refine msa macros CLIP_*.
Changing details as following:
1. Remove the local variable 'out_m' in 'CLIP_SH' and store the result in
source vector.
2. Refine the implementation of macro 'CLIP_SH_0_255' and 'CLIP_SW_0_255'.
Performance of VP8 decoding has speed up about 1.1%(from 7.03x to 7.11x).
Performance of H264 decoding has speed up about 0.5%(from 4.35x to 4.37x).
Performance of Theora decoding has speed up about 0.7%(from 5.79x to 5.83x).
3. Remove redundant macro 'CLIP_SH/Wn_0_255_MAX_SATU' and use 'CLIP_SH/Wn_0_255'
instead, because there are no difference in the effect of this two macros.
Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Fixes: signed integer overflow: 1 + 2147483647 cannot be represented in type 'int' Fixes: 16041/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_HEVC_fuzzer-5685680656613376 Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg Reviewed-by: James Almer <jamrial@gmail.com> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
libavformat/subfile: Fix SEEK_CUR and SEEK_END seeking
Up until now, when performing a SEEK_END seek, the subfile protocol
ignored the desired position (relative to EOF) and used the current
absolute offset in the input file instead.
And when performing a SEEK_CUR seek, the current position has been
ignored.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
Chip Kerchner [Tue, 13 Aug 2019 00:21:24 +0000 (02:21 +0200)]
lsws/ppc/yuv2rgb_altivec: Replace vec_lvsl/vec_perm with vec_xl
gcc 6.x and 7.x generate wrong code for little endian machines
for the vec_lvsl/vec_perm instruction combos in some cases.
The bug was fixed in version 8.x
If these instructions are replaced with vec_xl, the problem goes
away for all versions of the compilers.
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg Reviewed-by: Tomas Härdin <tjoppen@acc.umu.se> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Carl Eugen Hoyos [Sun, 30 Jun 2019 23:45:36 +0000 (01:45 +0200)]
lavc/zmbvenc: Do not left-shift negative values.
Fixes the following ubsan errors with the sample from ticket #7980:
libavcodec/zmbvenc.c:243:29: runtime error: left shift of negative value -4
libavcodec/zmbvenc.c:244:28: runtime error: left shift of negative value -2
This may break some valid tiff files, it appears the specification does not require
the offsets to be increasing. They increase in the 2 test files i have though except
the last offset which is 0 (an end marker) and for which a special case is added to
avoid asking for a sample for that end marker.
See: [FFmpeg-devel] [PATCH 2/2] avcodec/tiff: Detect infinite retry loop
for an alternative implementation
Fixes: Timeout (Infinite -> Finite) Fixes: 15706/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_TIFF_fuzzer-5114674904825856
This variant was requested by paul on IRC 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>
Fixes: signed integer overflow: 2082471995 * 36 cannot be represented in type 'int' Fixes: 16025/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_DDS_fuzzer-5136663778426880 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>
Fixes: left shift of 1 by 31 places cannot be represented in type 'int' Fixes: 15817/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_MPC8_fuzzer-5636626409062400 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>
avcodec/alsdec: Fix integer overflows of raw_samples in decode_var_block_data()
This also makes the code consistent with the existing similar MUL64()
in decode_var_block_data()
Fixes: signed integer overflow: -7277630735906765035 + -3272193951413647896 cannot be represented in type 'long' Fixes: 16015/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_ALS_fuzzer-5666552818434048 Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
avcodec/alsdec: Fix integer overflow of raw_samples in decode_blocks()
Fixes: signed integer overflow: 2147483424 - -1772303236 cannot be represented in type 'int' Fixes: 15708/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_ALS_fuzzer-5067890362941440 Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
avcodec/vc1_block: Fix invalid shifts in vc1_decode_i_blocks()
Fixes: left shift of negative value -9 Fixes: 15299/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_MSS2_fuzzer-5660922678345728 Fixes: 15557/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_VC1IMAGE_fuzzer-5673351911047168 Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
avcodec/vc1_block: fix invalid shift in vc1_decode_p_mb()
Fixes: left shift of negative value -5 Fixes: 15294/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_VC1_fuzzer-5733921754447872 Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
avcodec/aacdec_template: fix integer overflow in imdct_and_windowing()
Fixes: signed integer overflow: 2147483645 + 4 cannot be represented in type 'int' Fixes: 15418/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_AAC_FIXED_fuzzer-5685269069561856 Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
avformat/rtpdec_mpeg4: Fix integer parameters size check in SDP fmtp line
=== PROBLEM ===
I was trying to record h264 + aac streams from an RTSP server to mp4 file. using this command line:
ffmpeg -v verbose -y -i "rtsp://<ip>/my_resources" -codec copy -bsf:a aac_adtstoasc test.mp4
FFmpeg then fail to record audio and output this logs:
[rtsp @ 0xcda1f0] The profile-level-id field size is invalid (40)
[rtsp @ 0xcda1f0] Error parsing AU headers
...
[rtsp @ 0xcda1f0] Could not find codec parameters for stream 1 (Audio: aac, 48000 Hz, 1 channels): unspecified sample format
In SDP provided by my RTSP server I had this fmtp line:
a=fmtp:98 streamType=5; profile-level-id=40; mode=AAC-hbr; config=1188; sizeLength=13; indexLength=3; indexDeltaLength=3;
In FFmpeg code, I found a check introduced by commit 24130234cd9dd733116d17b724ea4c8e12ce097a. It disallows values greater than 32 for fmtp line parameters.
RFC-4566 (SDP: Session Description Protocol) do not give any limit of size on interger parameters given in an fmtp line.
However, In RFC-6416 (RTP Payload Format for MPEG-4 Audio/Visual Streams) give examples of "profile-level-id" values for AAC, up to 55.
=== FIX ===
As each parameter may have its own min and max values
I propose to introduce a range for each parameter.
For this patch I used RFC-3640 and ISO/IEC 14496-1 as reference for validity ranges.
This patch fix my problem and I now can record my RTSP AAC stream to mp4.
It has passed the full fate tests suite sucessfully.
Signed-off-by: Olivier Maignial <olivier.maignial@smile.fr> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Jun Zhao [Mon, 5 Aug 2019 04:53:55 +0000 (12:53 +0800)]
examples/encode_video: only add sequence end code for mpeg1/2 video
Only add sequence end code for mpeg1/mpeg2 video, or else use the encoder
libx264 or libx265 in this sample, decoding the output file will get
unknow NALU type error.
projects / ffmpeg / log