Zhao Zhili authored 5 years ago and Thomas Guillem committed 5 years ago

strerror_r can be GNU-specific on Android.

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

Zhao Zhili authored 5 years ago and Thomas Guillem committed 5 years ago

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

getmntent_r is not necessarily available when the mntent.h header is present
(cf.  Android).

It's part of vd that is already passed to the module loader.

We keep VLC_VIDEO_CONTEXT_NONE to 0 to make sure some context are not created
accidentally with a type value of 0.

Johan Gunnarsson authored 5 years ago and Thomas Guillem committed 5 years ago

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

Qt requires container elements to be copyable, https://bugreports.qt.io/browse/QTBUG-54685
but class with const members can't be copy-assignable. Remove the const values
so that we can copy SoutChain.

This patch fix the compilation with Clang 9.

This is needed so that the built contribs are actually packaged.

We try to get the minimum amount that is guaranteed to work. We don't want to
fail in the middle of playback because we actually did not have enough surfaces.

It gives 1 when it should be 8. Other codecs seem to be OK.

We don't use the display pool (vout_GetPicture) anymore.

The frame rate is a field per second. To get a duration we divide seconds by
the number of fields. This what the vlc_tick_from_samples() call does. But we
divide by 0 without this fix.

We don't want to use the hardware decoder for now.

This is similar to what is done in the display module.

This is similar to what is done in the display module.

Fill the proper i_width/i_height corresponding to the padded decoder surfaces.

We don't need to pass the lavc pixel format anymore

To avoid confusion with the output format of the VA.

It's already done before in the decoder helper.

This way other decoder owner (thumbnail, transcode, etc) benefit from the same
fix in a consistent manner.

Now we clean the value directly in the dec->fmt_out whereas before it was
cleaned at the vout level and the decoder didn't know about it.

It should be done before the decoder allocates it's picture pool. This is the
case as the default pool is created during the format_update, which called in
the same function later.

Other decoder owners are not impacted as they don't provide a default pool and
just rely on a direct picture_NewFromFormat() call.

This way other decoder owner (thumbnail, transcode, etc) benefit from the same
fix in a consistent manner.

Now we clean the value directly in the dec->fmt_out whereas before it was
cleaned at the vout level and the decoder didn't know about it.

In the end if the decoder didn't write it before it would not have written it
later anyway. And if it really wants to, it can still do it later and issue an
update_format.

There's no reason to use something else than what the decoder provides.

It's already done in the decoder helper.

Now we always set p_owner->video to the updated format even if dimensions/SAR
didn't change.

This way other decoder owner (thumbnail, transcode, etc) benefit from the same
fix in a consistent manner.

Now we clean the value directly in the dec->fmt_out whereas before it was
cleaned at the vout level and the decoder didn't know about it.

In the end if the decoder didn't write it before it would not have written it
later anyway. And if it really wants to, it can still do it later and issue an
update_format.

It's already done in the decoder helper.

This way other decoder owner (thumbnail, transcode, etc) benefit from the same
fix in a consistent manner.

Now we clean the value directly in the dec->fmt_out whereas before it was
cleaned at the vout level and the decoder didn't know about it.

In the end if the decoder didn't write it before it would not have written it
later anyway. And if it really wants to, it can still do it later and issue an
update_format.

Many of them do trivial things that can be factorized in decoder_helpers.

It's redundant when FixDisplayFormat is called from the update_format callback.

When called from CreateVoutIfNeeded in the get_device callback, the chroma of
the source format doesn't change the decoder device creation. The value is
likely not even set during that call anyway.

It's already done in the decoder helper.

It's already done in the decoder helper.

It's already done in the decoder helper.

It's already done in the decoder helper.

It's already done in the decoder helper.

All decoder owners do this fixup.