Filters created with aout_filter_Create() already check the structure pointer
is not NULL.

Alexandre Janniaux authored 4 years ago and Steve Lhomme committed 4 years ago

Generate a simple operations structure for filters generating their own filter
callback via VIDEO_FILTER_WRAPPER().

The filter chain sets a mouse handler on video filters that don't have one,
just as before, by using a local version of the ops structure of the filter and
adding the local mouse callback.

Co-authored-by: Steve Lhomme <robux4@ycbcr.xyz>

Alexandre Janniaux authored 4 years ago and Steve Lhomme committed 4 years ago

Remove switch used for one case and use VLC_EGENERIC/VLC_SUCCESS when
suitable.

Alexandre Janniaux authored 4 years ago and Steve Lhomme committed 4 years ago

Avoid switch for one case and use VLC_EGENERIC/VLC_SUCCESS when
suitable.

And remove unneeded declaration of functions defined by VIDEO_FILTER_WRAPPER.

Just like in the video output, the user filters should adapt to the source,
better than they should adapt to the encoder. In particular, hardware support
will be favored in more cases.

Just like in the video output, the last converter chain is responsible to adapt
the decoded + prerendered + filtered pictures to the encoder requested input
format (see osys->converters display.c).

In the end we have the same filters as the video output: static, user, final
converter.

With the orientation changes done outside of the p_conv_nonstatic, this filter
cahin was always NULL (b_orient always false).

This is also how the display modules work, the orientation is usually done
while rendering the picture.

The p_final_conv_static filter is also the one that produces pictures on which
we render subpictures, which is when the reorientation is forced.

If the decoder and encoder alignment don't match, the last converter before the
encoder will take care of the alignment.

The last b_do_orient test is done without a local variable.

No functional changes.

No matter the value of b_do_orient, the filter chain will not be created. It
will only return a different error code.

If b_do_orient is true we returned VLC_EGENERIC a few lines above.
So b_do_orient is false, so we will return VLC_SUCCESS no matter what the
values of b_do_chroma and b_do_scale are.

- remove the enum differentiating each pass
- use the proper filter directly on each pass
- remove impossible test cases

Rather than a loop doing things differently in both passes.

No functional changes.

Add some more comments on the different filter chains.

Petri Hintukainen authored 4 years ago and Jean-Baptiste Kempf committed 4 years ago

Track language may change between playlists.

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

The new duplicated media should not share events of the orignal one.
Ie. parsing the original media should not trigger updates on the
duplicated media.

- Don't override the media set by the user from
  on_current_media_changed().

- Don't create a new media from an input_item. This will prevent to send
  media events with a media pointer that is unknown to the caller.

This callback is now always sending a media that is set with
libvlc_media_player_set_media().

Since the player is now fully asynchronous, a call to set_media() +
play() will be processed asynchronously if an older media is playing.
The user might want to know when its new media is actually playing.
Listening to the libvlc_MediaPlayerMediaChanged event is only way to do so.

Here is an example:

1. set_media(A) && play()
  -> get_media() -> A
2. set_media(B) && play()
  -> get_media() -> B
3. libvlc_MediaPlayerMediaChanged(A)
  -> get_media() -> B
4. libvlc_MediaPlayerMediaChanged(B)
  -> get_media() -> B

This will be used by libvlc_media to store the media instance into an
input_item. This will allow to get the media corresponding to an item
when propagating media events.

If the Cues we read in adaptive streaming are incorrect, then the adaptive
streaming cannot work at all with those streams.

It can only be loaded with the proper name, it is not automatically loaded.

It is necessary to know when opening the module because we start parsing the
Matroska data to find the mandatory headers and the Cues may be parsed during
that time.

Refs #21860.

When formatting a message to HTTP 1.1, chunked encoding is often
necessary. However if the input message structure is not tied to any
specific HTTP version, it cannot contain the chunked encoding, which
will only be added when formatting.