GitLab

Most blocks contains aligned samples.

Causes out of sync, frame dropping.

This is a fixup of 48a64287. Also destroy the mpg123_handle with mpg123_delete()
when flushing.

The memcpy used will of course make p_ssa_style and p_default_style
refer to the same text_style_t, resulting in a use-after-free during
clean-up (as well as a leak due to the original p_ssa_style->p_style
being lost).
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Moreover, this error case does not seem to constitute a critical decoding
error.

corrupted block is implicitely a discontinuity
fixup of ce6db786

Fixes playback of e.g. vp90-2-05-resize.ivf in conformance collection.
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

We don't have to force output chroma on any platforms (let the system decide
for the best chroma).

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

currently we only support media time or abs generated
from demux

And require 0.38 to activate VAAPI for everyone (VP9&HEVC)

Close #18011

CID #1402754

CID #1402784

unwanted style propagation on non teletext

When reusing the context we already keep a reference via ID3D11DeviceContext_GetDevice()
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

This way we can have a custom I420 with a texture for each plane

This replaces the pf_get_cc callback for video decoders. Packetizers should
still use pf_get_cc.

This replaces the decoder_RequestReload() function. The module implementation
that returns this status should not have modified or released the input block
passed in pf_decoder parameter.

This replaces the decoder_t->b_error boolean.

For now, only decoders that were setting b_error return VLCDEC_ECRITICAL, but a
lot more decoders should use this value since critical errors are often ignored.

Use only one callback for every decoder types:

int (*pf_decode)(decoder_t *, block_t *p_block);

There is now only one way to send output frames/blocks from a decoder module:
using decoder_QueueVideo(), decoder_QueueAudio() and decoder_QueueSub()
functions.

This fixes transcoding not receiving any output when a decoder used
decoder_Queue*() function.

The pf_packetize callback is kept unchanged. A packetizer shouldn't be
asynchronous at all (and this simplify the locking for decoder core).

The pf_decode callback returns, for now, only one value: SUCCESS. This will
allow a module to send more status.

No need to extract into a secondary one to send to the vout and so a
CopySubresourceRegion() for every frame decoded.

Rework the mutex locking as in this new case we don't do anything in
Extract() anymore.
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

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

Ref #17971

This adds explicit casts where needed

Shared by all code using the opaque format.
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Remove the videotoolbox-zero-copy variable since direct rendering is always
enabled.

Map the CVPX pixel format to a VLC fourcc, this buffer will be renderered by
the OpenGL vout with the cvpx converter (mapping an IOSurface to an OpenGL
texture).

On Macos, the destination pixel format shouldn't set touched at all to let the
decoder output frames in its native format (and don't do any color conversion).
This need to be tested on iOS (hence the #ifdef).