GitLab

Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

On Windows if you allocate aligned memory you need to free it with an aligned
version of free.

This is similar to the old vlc_memalign() + vlc_free()
Signed-off-by: Thomas Guillem <thomas@gllm.fr>

free() is required by ISO C (hosted) and C++ anyway.

CID #1444701
Signed-off-by: Hugo Beauzée-Luyssen <hugo@beauzee.fr>

And set a picture->context as if it's coming from the decoder, otherwise
snapshots won't work.
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

For CPU formats mapped to GPU each texture of the array may be different.
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

This is the reverse of ReleasePictureSys
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

It may be used by other modules, even though it seems Windows is the only one
to provide a pitch for the main plane.
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

d3d11_surface is automatically added since D3D11 forces NV12 input for
DXGI_FORMAT_420_OPAQUE decoding.
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

We can't display it but we still need a proper texture format to display.
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Drivers crash during ID3D11VideoDevice::CreateVideoDecoderOutputView() if a
texture has more than 30 slices.

Fixes #18261
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

it doesn't matter what the format says
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

it's an old hack we don't need with picture_sys_t.
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

The level variable can be destroyed just after the filter chain is built.

This fixes indirect chroma conversion (i.e. when swscale cannot do the
conversion entirely).

Regression from ca546680.

Add CVPX_{NV12|UYVY|I420} to I420 and I420 to CVPX{NV12|UYVY|I420} filters.

cvpx buffers don't need the vout to be allocated.

This helper will be used by the videotoobox codec, the cvpx chroma and by the
cvpx opengl converter.

This new "chain" video filter will try to add a video converter before the
requested video filter.
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

used in I420 to NV12 conversion
Signed-off-by: Thomas Guillem <thomas@gllm.fr>

Signed-off-by: Thomas Guillem <thomas@gllm.fr>

Without this it's impossible to do some processing like splitting NV12 UV
planes, it requires double the resolution. In that case the code asserts on
assert(hstep > 0) in SSE_SplitPlanes().
Signed-off-by: Thomas Guillem <thomas@gllm.fr>

Signed-off-by: Thomas Guillem <thomas@gllm.fr>

This fixes a regression introduced by f9f0bbc1

This memory corruption was not visible since this function is generally used
to copy from GPU mapped buffers, that are often bigger than what you ask.

Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>