Signed-off-by: Rémi Denis-Courmont <remi@remlab.net>

No longer needed as the model now behaves correctly.
reverts 442bad11

- popup now created according to the selected items and models.
- fixes view/popup model abstraction.
- allows introducing new methods/models.

Never ends..

- Separate common methods into interface, allowing multiple
inheritance architecture (to use any implemented AbstractModel)
- Code factorization between PL/ML models (and PL/ML Items)
- Use proxy model for views and redirection to the PL/ML models.
Fixes filtering and also really make use of the current model (only the
PL Model was really used before, models were half switched).
- stop storing/serializing selection indexes
- ML Missing locks and bug fixes
- Move ML callback into event loop
- ML Items leaks fixes
- Code factorization
- Dead code removal

It might be better to convert from the proper character encoding to
UTF-8, but I do not know what proper is.

Signed-off-by: Rémi Denis-Courmont <remi@remlab.net>

tested with ticket 9022 samples to 2 channel and without channel
definition. Output files seemed to have correct channel amount in both
cases.

refs #9022

I am not taking a stand on whether pictures should or should not have
references when the pool is destroyed. It is hard to support for some
video outputs (and indeed not all of them do so correctly).
Nevertheless, it can happen, so the core might as well deal with this.

Feel free to undo this and add an assertion, if you fix all affected
code paths.

httplive: Choose initial stream to be the biggest one, avoiding audio-only as migration from that doesn't work in some cases

libmpgatofixed32: give mad-errors as msg_Err and give empty frames only if mad tells that it can't recover from that error

Signed-off-by: Rafaël Carré <funman@videolan.org>

Signed-off-by: Rafaël Carré <funman@videolan.org>

Signed-off-by: Rafaël Carré <funman@videolan.org>

Signed-off-by: Rafaël Carré <funman@videolan.org>