avutil_version() currently performs several checks before
just returning the version. There is a static int that aims
to ensure that these tests are run only once. The reason is that
there used to be a slightly expensive check, but it has been removed
in 92e3a6fdac. Today running only
once is unnecessary and can be counterproductive: GCC 10 optimizes
all the actual checks away, but the checks_done variable and the code
setting it has been kept. Given that this check is inherently racy
(it uses non-atomic variables), it is best to just remove it.
Reviewed-by: Paul B Mahol <onemda@gmail.com>
Reviewed-by: James Almer <jamrial@gmail.com>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
(cherry picked from commit ff80090374)
Up until now, the scc demuxer not only read the line that it intends
to process, but also the next line, in order to be able to calculate
the duration of the current line. This approach leads to unnecessary
complexity and also to bugs: For the last line, the timing of the
next subtitle is not only logically indeterminate, but also
uninitialized and the same applies to the duration of the last packet
derived from it.* Worse yet, in case of e.g. an empty file, it is not
only the duration that is uninitialized, but the whole timing as well
as the line buffer itself.** The latter is used in av_strtok(), which
could lead to crashes. Furthermore, the current code always outputs
at least one packet, even for empty files.
This commit fixes all of this: It stops using two lines at a time;
instead only the current line is dealt with and in case there is
a packet after that, the duration of the last packet is fixed up
after having already parsed it; consequently the duration of the
last packet is left in its default state (meaning "unknown/up until
the next subtitle"). If no further line could be read, processing
is stopped; in particular, no packet is output for an empty file.
*: Due to stack reuse it seems to be zero quite often; for the same
reason Valgrind does not report any errors for a normal input file.
**: While ff_subtitles_read_line() claims to always zero-terminate
the buffer like snprintf(), it doesn't do so if it didn't read anything.
And even if it did, it would not necessarily help here: The current
code jumps over 12 bytes that it deems to have read even when it
hasn't.
Reviewed-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
(cherry picked from commit 60e12318bb)
It claims to always zero-terminate its buffer like snprintf(),
yet it does it not on EOF. Because of this the mcc demuxer
used uninitialized values when reading an empty input file.
Reviewed-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
(cherry picked from commit 8766361fc1)
Fix this by switching to av_dynarray_add_nofree() which is more
natural anyway because the entries of the array are pointers.
Reviewed-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
(cherry picked from commit 05c1f78a72)
av_image_copy() expects an array of four pointers according to its
declaration; although it currently only touches pointers that
are actually in use (depending upon the pixel format) this might
change at any time (as has already happened for the linesizes
in d7bc52bf45).
This fixes ticket #9264 as well as a warning from GCC 11.
Reviewed-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
(cherry picked from commit 9b17273c77)
When an flite filter instance is uninitialized and the refcount
of the corresponding voice_entry reaches zero, the voice is
unregistered, yet the voice_entry's pointer to the voice is not reset.
(Whereas some other pointers are needlessly reset.)
Because of this a new flite filter instance will believe said voice
to already be registered, leading to use-after-frees.
Fix this by resetting the right pointer instead of the wrong ones.
Reviewed-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
(cherry picked from commit 18ddb25c7a)
Could also happen if initializing flite failed* or if an unknown voice
has been selected or if registering the voice failed.
*: which it currently can't, because it is a no-op.
Reviewed-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
(cherry picked from commit 304cc03798)
This reverts commit 723c37d3b7.
Said commit was in preparation for auto-inserting the idet filter.
This has never happened; even if it did, the code is wrong, because
it segfaults if the filter instance doesn't have a name (having one
is not mandatory). Furthermore, it is documented for libavfilter to
not assign any semantics to the name, which this check violates.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
(cherry picked from commit f626a3d0e0)
h263_get_motion_length() forgot to take an absolute value;
as a consequence, a negative index was used to access an array.
This leads to potential crashes, but mostly it just accesses what
is to the left of ff_mvtab (unless one uses ASAN), thereby defeating
the purpose of the AV_CODEC_FLAG2_NO_OUTPUT because the sizes of
the returned packets differ from the sizes the encoder would actually
have produced.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
(cherry picked from commit 9207dc3b0d)
av_image_copy() expects an array of four pointers and linesizes
according to its declaration; it currently only pointers that are
actually in use (depending upon the pixel format), but this might
change at any time. It has already happened for the linesizes in
d7bc52bf45 and so increasing their
array fixes a stack-buffer overread.
This fixes a -Wstringop-overflow= and -Wstringop-overread warning
from GCC 11.2.
Reviewed-by: Linjie Fu <linjie.justin.fu@gmail.com>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
(cherry picked from commit b0b90100bf)
av_image_copy() expects an array of four pointers according to its
declaration; although it currently only touches pointers that
are actually in use (depending upon the pixel format) this might
change at any time (as has already happened for the linesizes
in d7bc52bf45).
This fixes a -Wstringop-overflow= warning with GCC 11.2.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
(cherry picked from commit b09ea67b40)
Due to this bush.aa (from the FATE suite) exported garbage metadata
with key "_040930".
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
(cherry picked from commit 0a76f8217e)
This fixes compilation errors in case nvenc is enabled
(e.g. autodected) with both nvenc-based encoders disabled
because nvenc uses ff_alloc_a53_sei(), yet only the nvenc-based
encoders require atsc_a53.
(This error does not manifest itself in case of static linking
(nothing pulls in nvenc.o), but it exists with shared builds.)
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
Signed-off-by: Timo Rothenpieler <timo@rothenpieler.org>
- quantizer delta and matrix level specific.
- support loop filter delta.
- support use superres.
Signed-off-by: Fei Wang <fei.w.wang@intel.com>
(cherry picked from commit 84c73102d9)
VAAPI needs 2 output surface for film grain frame. One used for
reference and the other used for applying film grain and pushing
to downstream.
Signed-off-by: Fei Wang <fei.w.wang@intel.com>
(cherry picked from commit 7871144cf8)
For film grain clip, vaapi_av1 decoder will cache additional 8
surfaces that will be used to store frames which apply film grain.
So increase the pool size by plus 8 to avoid leak of surface.
Signed-off-by: Fei Wang <fei.w.wang@intel.com>
(cherry picked from commit 53403158cc)
Defined in spec 5.9.24/5.9.25. Since function void
global_motion_params(AV1DecContext *s) already updates
gm type/params, the wminvalid parameter only need to get
the value from cur_frame.gm_invalid.
Signed-off-by: Tong Wu <tong1.wu@intel.com>
(cherry picked from commit 4e7a7d75e3)
Since order_hint_bits_minus_1 range is 0~7, cur_frame_hint can be
most 128. And similar return value for cbs_av1_get_relative_dist.
So if plus them and use int8_t for the result may lose its precision.
Signed-off-by: Fei Wang <fei.w.wang@intel.com>
(cherry picked from commit e7ff5722b1)
libdav1d 1.0.0 will be the first version supporting Dav1dSettings.n_threads == 0.
Signed-off-by: James Almer <jamrial@gmail.com>
(cherry picked from commit d873b5fffc)
The calling code does not handle failures and will fail with assertion failures later.
Seeking can always fail even when the position was previously read.
Fixes: Assertion failure
Fixes: 35253/clusterfuzz-testcase-minimized-ffmpeg_dem_MATROSKA_fuzzer-4693059982983168
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
(cherry picked from commit d115eec979)
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
