Rémi Denis-Courmont authored 3 years ago and Hugo Beauzée-Luyssen committed 3 years ago

This keeps the C prefix for Chinese for consistency with the existing
AVS mapping. Another option would be to drop the C, and rename the rare
legacy video game coding also called AVS.

The AVI mapping is taken from libavformat's RIFF parser.

Alexandre Janniaux authored 3 years ago and Hugo Beauzée-Luyssen committed 3 years ago

module_list type in config_chain was raising the log:

    unhandled config var type (134)

Steve Lhomme authored 3 years ago and Rémi Denis-Courmont committed 3 years ago

It's usually 0 but just in case it isn't, that's how it should be allocated.

Steve Lhomme authored 3 years ago and Rémi Denis-Courmont committed 3 years ago

* MF_E_TRANSFORM_STREAM_CHANGE was added in 4.0
* MF_E_TRANSFORM_NEED_MORE_INPUT was added in 4.0
* MF_EVENT_FLAG_NO_WAIT was added in 4.0

Steve Lhomme authored 3 years ago and Rémi Denis-Courmont committed 3 years ago

There is only but that's the proper way to release it.

Steve Lhomme authored 3 years ago and Rémi Denis-Courmont committed 3 years ago

Otherwise we lose the beginning of the file.

Steve Lhomme authored 3 years ago and Rémi Denis-Courmont committed 3 years ago

- only set COBJMACROS before the MediaFoundation includes
- use initguid.h as in the rest of the code
- STDCALL is not used

Steve Lhomme authored 3 years ago and Rémi Denis-Courmont committed 3 years ago

We already link to it.

The API calls are supported since Vista and in UWP on Win10, except MFTEnumEx
which is available since Win7 and in UWP on Windows 10 1709/16299.

Steve Lhomme authored 3 years ago and Rémi Denis-Courmont committed 3 years ago

The MFT_ENUM_FLAG_TRANSCODE_ONLY is not needed in our case, we don't want
decoders tuned for that.

Steve Lhomme authored 3 years ago and Rémi Denis-Courmont committed 3 years ago

Not available on WinXP.

Steve Lhomme authored 3 years ago and Rémi Denis-Courmont committed 3 years ago

Asynchronous decoders need to send MFT_MESSAGE_NOTIFY_START_OF_STREAM afterwards.

Steve Lhomme authored 3 years ago and Rémi Denis-Courmont committed 3 years ago

No functional changes.

Steve Lhomme authored 3 years ago and Rémi Denis-Courmont committed 3 years ago

There can be more than one output available at a time.

Steve Lhomme authored 3 years ago and Rémi Denis-Courmont committed 3 years ago

And enqueue the audio/video output right after.

Steve Lhomme authored 3 years ago and Rémi Denis-Courmont committed 3 years ago

The buffer lock is done earlier so that audio/video copies have the proper
buffer to write to.

+ only call IMFSample_GetTotalLength when we are going to use the result (audio)

Steve Lhomme authored 3 years ago and Rémi Denis-Courmont committed 3 years ago

Like ComPtr::Reset() would do.

Steve Lhomme authored 3 years ago and Rémi Denis-Courmont committed 3 years ago

No need for else cases when the previous call returned or jump to an error.

Only push picture/audio buffers where they we actually created.

Steve Lhomme authored 3 years ago and Rémi Denis-Courmont committed 3 years ago

+ more logs

Steve Lhomme authored 3 years ago and Rémi Denis-Courmont committed 3 years ago

Push the PPS/SPS/VPS once

They were always zeroes.

No functional changes.

...rather than cfg.window_props. cfg.display was unused (always zeroes)
on vout_display_New() until this change.

This is consistent with vout_display_SetSize() affecting cfg.display
rather than cfg.window_props.

Only from vout_display_New() can (and should) the window size be used
rather than the ideal default display size. In the other call sites,
we actually want to compute the ideal size regardless of the actual
window size.

Take that case out of vout_display_GetDefaultDisplaySize() to enable
further refactoring. This is a non-functional change, except for the
added run-time warning.

This is no longer necessary.

In practice, this function is used to compensate mismatches between
the source format SAR and the display PAR and should only be based on
those parameters and the source format visible dimensions. Indeed, in
the two call sites of the function pass 0x0 as the display size.

In the vout_display_New() call site, the function is used to
work-around semi-broken window providers that do not provision a window
size, or only do so with a delay. Normally, it just copies the dimensions
from window_props.

Either way cfg->display dimensions are always zeroes.

Furthermore, the 3 hereto unused and hereby removed cases failed to
account for orientation correctly: from_source should have been false.

So far, the display was created without a requested size. Then its size
was immediately changed to that of the window.

With this change, the display is created directly with the correct size.
This matches the core video output behaviour.

As of the last rework, ownership of the original format has already been
passed to the video output by the point of enabling the window. It is
kept until the next format change, or the destruction of the video
output. So it shall not be freed if enabling the window fails.

Maxime Chapelet authored 3 years ago and Jean-Baptiste Kempf committed 3 years ago

the display can be created when the application state is active or inactive
inactive state actually happens when the application comes back from background and becomes active just after a very short amount of time
hence we should allow vout initialization only when application state is not background

Pierre Lamot authored 3 years ago and Hugo Beauzée-Luyssen committed 3 years ago

The MediaLib object can perform ML requests itself, this would register an event
called on object destruction, that would callback on itself but the MediaLib
would be in an invalid state (partially destroyed).

Anyway we shouldn't handle object destruction events after ML task is done or
canceled

fix: #26448