When decoding a bitstream with weighted-bipred enabled,
the results on ARM and x86 platforms may differ.
The reason for the inconsistency is that the value of
STRIDE_ALIGN differs between platforms. And STRIDE_ALIGN
is set to the buffer stride of temporary buffers for U
and V components in mc_part_weighted.
If the buffer stride is 32 or 64 (as on x86 platforms),
the U and V pixels can be interleaved row by row without
overlapping, resulting in correct output.
However, on ARM platforms where the stride is 16,
the V component did overwrite part of the U component's pixels,
leading to incorrect predicted pixels.
The bug can be reproduced by the following bitstream.
https://trac.ffmpeg.org/attachment/ticket/11357/inter_weighted_bipred2.264
Fixes: ticket 11357
Commit-msg-mostly-by: Bin Peng <pengbin@visionular.com>
Reviewed-by: Bin Peng <pengbin@visionular.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
This differs from libvorbis by
stddev: 2.44 PSNR: 88.58 MAXDIFF: 41 bytes: 834304/ 834304
for the file from the ticket
Fixes: Ticket11427
Regression since: dc89cf804a
This is a similar solution to what james proposed earlier in
[FFmpeg-devel] [PATCH] avcodec/vorbisdec: don't abort on EOD when decoding residuals
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
The current vvc_probe function checks for SPS, PPS, and IRAP NAL units but does not enforce their detection order. This can lead to false positives when PPS or IRAP appear before SPS, causing malformed or non-compliant bitstreams to be incorrectly identified as valid VVC streams.
This commit modifies the logic to use boolean flags and enforces the correct detection sequence: SPS must be detected before PPS, and PPS before IRAP. This ensures more accurate probing and prevents invalid bitstreams from being misidentified.
For streams with wrong ordering a lower score is returned
This change addresses issues reported in:
- https://trac.ffmpeg.org/ticket/11496
- https://trac.ffmpeg.org/ticket/11087
Signed-off-by: xiongweixiao <xiongweixiaoxxw@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Elapsed time is much less important than the actual progress, and I am guessing
several ffmpeg CLI frontends use the existing status line for showing progress,
so putting a new field in the beginning is less optimal anyway.
Related to ticket #11582.
Signed-off-by: Marton Balint <cus@passwd.hu>
Codec probing was primarily added to the wav demuxer to support DTS-in-wav
files, but DTS probing functions return AVPROBE_SCORE_EXTENSION+1, so we can be
a bit more strict with the required score.
This fixes MP3 misdetections for some wav files.
Fixes ticket #11581.
Signed-off-by: Marton Balint <cus@passwd.hu>
Videos of "StarWars - Making Magic" consist of 640x480 codec3 frames
which establish a background, and a 320x240 codec48 video put on top
at random left/top offsets.
To support this, a new default buffer "fbuf", which holds the final
image to be presented, is added, since codec37/47/48 need their 2/3 buffers
to be private to themselves. The decoded result is then copied to the fbuf,
honoring the left/top offsets if required.
Signed-off-by: Manuel Lauss <manuel.lauss@gmail.com>
Change the size detection a bit to recognize common video sizes,
as the FOBJ codecs>=37 cannot always be trusted, since they can
be embedded in a larger frame.
Signed-off-by: Manuel Lauss <manuel.lauss@gmail.com>
Some videos of "StarWars - Making Magic" have this subcompression
type: data just consists of the 16 byte codec48 header; the DOS player
and the c48 decoder in the "Mysteries of the Sith" game engine ignore it.
Signed-off-by: Manuel Lauss <manuel.lauss@gmail.com>
liboapv will seemingly encode correct 4:4:4 output, but report profile_idc 33, which
is specifically the profile value for 4:2:2 10bit.
Signed-off-by: James Almer <jamrial@gmail.com>
In the call to vkGetPhysicalDeviceImageFormatProperties2(), we were
previously requesting the properties of the first fallback format (e.g.
VK_FORMAT_R8_UNORM for VK_FORMAT_G8_B8R8_2PLANE_420_UNORM) instead of
the actual format in use.
We don’t do anything with it afterwards, but there is no reason to keep
querying the wrong format.
Fix chroma_location being cleared by setrange and setfield filters.
This was forgotten in 201f1cba15.
Signed-off-by: Tobias Rapp <t.rapp@noa-archive.com>
It was possible for the buffer pointers for the last tile to go over the
end of the unit buffer leading to a read overflow during decode of the
macroblock layer. Check all tile component sizes to prevent this case
and also catch related tile size mismatch errors earlier.
Halt tile component decoding at the first entropy error (this will be a
desync and is not recoverable). If any tile components contain errors
then discard the frame unless the output-corrupt flag is set.
Also fixes CID 1646764, which is the error case where the tile component
is too large for get_bits to handle.
No reason to build the exact same table once per decoding thread.
Reviewed-by: Mark Thompson <sw@jkqxz.net>
Signed-off-by: James Almer <jamrial@gmail.com>
Abort as soon as we're done reading the slice header instead of running extra checks
that assume slice data may follow.
Signed-off-by: James Almer <jamrial@gmail.com>
Prevents printing bogus errors about the value being 0, when in fact we
overread the available slice buffer.
Signed-off-by: James Almer <jamrial@gmail.com>
Since GCC 10 and llvm.org Clang 11, -fno-common is the default.
However Apple's Xcode Clang hasn't followed suit yet, and still
defaults to -fcommon.
Compiling with -fcommon causes uninitialized global variables to
be treated as "common" (which allows multiple object files to have
similar definitions).
Common variables seem to have the issue that their intended alignment
isn't signaled, so the linker assumes that they may need alignment
according to their full size.
With large global tables, this can lead to linker warnings like
this, with Xcode 16.3:
ld: warning: reducing alignment of section __DATA,__common from 0x8000 to 0x4000 because it exceeds segment maximum alignment
This can be reproduced with a small snippet like this:
char table[16385];
int main(int argc, char* argv[]) { return 0; }
Compiling with -fno-common avoids this issue and warning, and
matches the default behaviour of other compilers. (Compiling with
-fno-common also avoids the risk of accidentally accepting
duplicate definitions of global variables, as long as they are
uninitialized.)
Signed-off-by: Martin Storsjö <martin@martin.st>
