This was a leftover of testing things in my fork, where I wasn't using
the master branch.

Marvin Scholz authored 5 years ago and Jean-Baptiste Kempf committed 5 years ago

Co-authored-by: Konstantin Pavlov <thresh@videolan.org>
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Marvin Scholz authored 5 years ago and Jean-Baptiste Kempf committed 5 years ago

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

Interrupt was already handled in waiting dialog functions, but it was done
after showing the dialog and before waiting for an user interaction. This could
lead to dialog being showed and immediately cancelled.

This commit check for the killed state before any dialog interactions, this
also fixes non waiting dialog functions shown while being killed.

If a module is interrupted while waiting in a vlc_*_i11e*() function, the i11e
function will return an error after restoring the interrupted state (from
vlc_interrupt_finish()). This error will be carefully handled by the module
that will close itself. If a second module is in the probe list, any call to a
vlc_*_i11e*() function will act as not interrupted. This cause any following
modules in the probe list to ignore this interrupted state.

To fix this issue, vlc_interrupt_finish() will now restore the interrupted
state to the killed state.

temp, the whole thing will be gone with adaptive rework

We already have a reference on the surface.

The other D3D11 picture_sys don't need to know about it.

We only need to keep it

+avoid a goto and remove duplicate variable set

Each derivative should have its own macro (like dxva2 does)

Each derivative should have its own macro (like d3d11va does)

Get the surface pointer directly using the va_surface.

Copy the structures, it's faster than pretending we create one from scratch.

Add a static_assert to make sure the structures we use are always compatible
with the main picture_context structure.

Get the surface pointer directly using the va_surface.

Copy the structures, it's faster than pretending we create one from scratch.

Add a static_assert to make sure the structures we use are always compatible
with the main picture_context structure.

As we have proper edit list support, we can allow late streams.
Should solve issues with user doing transcoding without tuning
sout-mux-caching

The free vout will be saved for later use when the decoder is destroyed. Until
then it is known internally to the decoder and can be stored for later use as
many times as starting the display thread/module will fail.

Rather than accessing vout->p each time.

The value of vout->p doesn't change during any of these calls.

Copy-paste, no functional changes.

Copy-paste, no functional changes.

The order of readdir() is completely filesystem dependent.

Slaves should be attached using the same order across multiple OSses/access
modules.

This fixes the test_libvlc_slaves on some system when the entry order from
readdir() is different.

This patch need to be backported to VLC 3.0.

Tested on iOS (where this variable can be different than 0).

Especially since the local sys will be freed after that.

After c0162d6d

It's not going to change during the lifetime of the converter module, unlike
the video context.

The callbacks are related to the decoder device, not the video context.

Instead of filling the picture context during get().

We can't just allocate the picture in the generic VA code and fill
the picture->context and picsys data as we need to handle the destructor
for each picture and release the resources when the picture is last
released.