* commit '620f88a0b94a651c6cc912b1fb32d308762d59b5':
vaapi_h264: Add missing return value check
h264_metadata: Fix clearing SEI payload in error case
cbs_h2645: Fix memory leak on when reading SEI fails
cbs: Add a missing return value check
Philip Langdale [Tue, 14 Nov 2017 05:11:27 +0000 (21:11 -0800)]
avcodec: Don't assume separate u and v planes in ff_alloc_picture
alloc_frame_buffer in ff_alloc_picture asserts that the linesize
of planes 1 and 2 are the same. If the pixfmt has a single uv
plane, like NV12, this won't be true.
So, let's only do this check if there are more than 2 planes.
We never hit this with previous hw formats because they don't set
linesize to meaningful values, but the cuda hw format sets the
values based on the underlying data layout.
Dale Curtis [Sat, 4 Nov 2017 00:10:55 +0000 (17:10 -0700)]
Avoid corrupting diagnostic state with _Pragma changes.
The macros for ICC and MSVC correctly push and pop the diagnostic
state of the compiler when disabling deprecation warnings. The
ones for clang/gcc should do the same. Without this, if a blanket
deprecation warning is applied to the code base it'll be flipped
back on incorrectly with FF_ENABLE_DEPRECATION_WARNINGS.
Signed-off-by: Dale Curtis <dalecurtis@chromium.org> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Currently if you use the multiple_requests=1 option and try to
receive a chunked-encoded response, http_buf_read() will hang forever.
After this patch, EOF is emulated once a 0-byte final chunk is
received by setting a new flag. This flag is reset in ff_http_do_new_request(),
which is used to make additional requests on the open socket.
Reviewed-by: Ronald S. Bultje <rsbultje@gmail.com> Signed-off-by: Aman Gupta <aman@tmm1.net>
Aman Gupta [Fri, 10 Nov 2017 20:34:41 +0000 (12:34 -0800)]
avcodec/videotoolbox: pass through hevc param changes to the decoder
This includes the SEI_PREFIX/SEI_SUFFIX NALUs, which can contain
updates like HEVC_SEI_TYPE_ACTIVE_PARAMETER_SETS. Previously, hevc
samples with this SEI present would not playback correctly.
See for example https://github.com/lhc70000/iina/issues/1123
Aman Gupta [Fri, 10 Nov 2017 20:25:39 +0000 (12:25 -0800)]
avcodec/videotoolbox: use decode_params to propagate H264 PPS changes and restart on SPS changes
This fixes decoding of H264 video samples with SPS and PPS changes.
See for example https://s3.amazonaws.com/tmm1/videotoolbox/spschange.ts,
which previously stalled the decoder and failed to produce any new frames
after the SPS change.
Also see https://s3.amazonaws.com/tmm1/videotoolbox/ppschange.ts, which
uses multiple PPS and would previously cause VT decode failures.
If the VideoToolbox session needs to be restarted, and
videotoolbox_start() fails for some reason (for instance, if the video
is interlaced and the decoder is running on iOS), avcodec will return
AVERROR_EXTERNAL. This can be used by the API user to switch to another
decoder.
Aman Gupta [Fri, 10 Nov 2017 20:18:35 +0000 (12:18 -0800)]
avcodec/h264: implement new decode_params callback for PPS/SPS
This callback will be used by the VideoToolbox H264 hwaccel so that it
can receive SPS and PPS NALUs. VideoToolbox requires PPS changes to be
fed into the decoder session, and for the session to be recreated when
the SPS changes.
Aman Gupta [Mon, 25 Sep 2017 23:45:53 +0000 (16:45 -0700)]
avcodec/h264, videotoolbox: return AVERROR_INVALIDDATA when no frames are produced
The only reason videotoolbox wouldn't produce frames is if the data fed
to it was invalid, so returning AVERROR_INVALIDDATA makes sense here.
Further, it means AVERROR_EXTERNAL can be used in further commits to signal
fatal VideoToolbox errors, letting the user know that they need to fallback to
another decoder.
Signed-off-by: Aman Gupta <aman@tmm1.net> Reviewed-by: Michael Niedermayer <michael@niedermayer.cc>
Aman Gupta [Mon, 13 Nov 2017 21:02:44 +0000 (15:02 -0600)]
This fixes a deadlock while reading a chunked https response, if
multiple_requests=1 is also set. Without an EOF to signal the end of
the last chunk, tls_read gets stuck forever trying to read more data
than is available. This occurs with the http protocol reproducibly,
because http.c always reads 4kb at a time, and the last chunk of an
http response is often much smaller.
After this commit, tls_read always returns any buffered plaintext
first before attempting to read more encrypted data off the
underlying tcp socket.
avcodec/aacpsdsp_template: Fix integer overflows in ps_decorrelate_c()
Fixes: runtime error: signed integer overflow: 1939661764 - -454942263 cannot be represented in type 'int' Fixes: 3191/clusterfuzz-testcase-minimized-5688798451073024 Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Fixes: runtime error: left shift of negative value -801112064 Fixes: 3492/clusterfuzz-testcase-minimized-5784775283441664 Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
avcodec/mdct_*: Fix integer overflow in addition in RESCALE()
Fixes: runtime error: signed integer overflow: 1219998458 - -1469874012 cannot be represented in type 'int' Fixes: 3443/clusterfuzz-testcase-minimized-5369987105554432 Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
avcodec/snowdec: Fix integer overflow in header parsing
Fixes: 3984/clusterfuzz-testcase-minimized-5265759929368576 Fixes: runtime error: signed integer overflow: -1085585801 + -1094995529 cannot be represented in type 'int' Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Thomas Köppe [Sun, 12 Nov 2017 16:24:37 +0000 (16:24 +0000)]
Fix missing used attribute for inline assembly variables
Variables used in inline assembly need to be marked with attribute((used)).
Static constants already were, via the define of DECLARE_ASM_CONST.
But DECLARE_ALIGNED does not add this attribute, and some of the variables
defined with it are const only used in inline assembly, and therefore
appeared dead. This change adds a macro DECLARE_ASM_ALIGNED that marks
variables as used.
This change makes FFMPEG work with Clang's ThinLTO.
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Mark Thompson [Tue, 24 Oct 2017 23:15:55 +0000 (00:15 +0100)]
vaapi_h264: Add workaround for bad SEI in old Intel drivers
With pre-2.0 Intel drivers in CBR mode, if an explicit SEI message with
the old (now deprecated) type is not included, the driver generates and
inserts some timing SEI which is almost certainly invlaid. Before 7a4fac5e91789b73e07bd4ad20493cfde028df76 we always inserted our own SEI
so this would not be visible, but since then it has been possible to
disable that. We would also like to avoid using the deprecated type,
and using the new type, while working in old drivers, does not suppress
the spurious message like the old type does.
Therefore, suppress the bad SEI insertion by providing a zero-length
buffer with the old type, which the driver can insert harmlessly.
projects / ffmpeg / log