Prince Gupta authored 11 months ago and Jean-Baptiste Kempf committed 10 months ago

Control doesn't handle focusable children well

Prince Gupta authored 1 year ago and Jean-Baptiste Kempf committed 10 months ago

BaseModel has getCount

Olaf Hering authored 10 months ago and Jean-Baptiste Kempf committed 10 months ago

... to restore compatiblity with Python 3.6.

This was initroduced in commit 177d0499,
which broke compilation on openSUSE Leap 15, which uses Python 3.6 als
default Python interpreter.

Fixes #28636

Signed-off-by: Olaf Hering <olaf@aepfle.de>

Prince Gupta authored 10 months ago and Jean-Baptiste Kempf committed 10 months ago

fixes binding assignment,
using QObject/QGadget types in Qt.binding fails
see https://bugreports.qt.io/browse/QTBUG-125095

Fatih Uzunoğlu authored 11 months ago and Jean-Baptiste Kempf committed 10 months ago

Qt Quick Shapes is not used now, it is not
necessary to initialize its resources.

Fatih Uzunoğlu authored 11 months ago and Jean-Baptiste Kempf committed 10 months ago

At the moment, Qt 5 effects are used. This needs
to be reverted when MultiEffect is started being
used.

Fatih Uzunoğlu authored 11 months ago and Jean-Baptiste Kempf committed 10 months ago

For now Quick Shapes/Path is not used, we don't need
to build it.

Prince Gupta authored 11 months ago and Jean-Baptiste Kempf committed 10 months ago

implement SD_CAT_MYCOMPUTER based service discovery
module on Windows

allows to query volumes mounted on host computer
Ref #28597

Martin Storsjö authored 10 months ago and Jean-Baptiste Kempf committed 10 months ago

This backports
https://aomedia.googlesource.com/aom/+/5ccdc66ab6eb8eb300eda854fab4ff250b2c2f92
and
https://aomedia.googlesource.com/aom/+/fb21617c1f3ef49795597e006b68adfba6e54be0.

This makes libaom test compiling a more nontrivial function,
which according to AAPCS requires backing up and restoring SVE
registers.

On current versions of Clang when targeting Windows, compiling such
a function errors out. This added test makes libaom not try to
compile the SVE codepaths for this target.

See https://github.com/llvm/llvm-project/issues/80009 for reference
on toolchain support for SVE functions on Windows on aarch64.

visuve authored 11 months ago and Jean-Baptiste Kempf committed 10 months ago

Use system preferred RNG and skip the unnecessary initialization steps (in this
context, which is definitely not security related)
The vlc_rand_bytes function is now about 55x faster when called repeatedly in a loop
See https://learn.microsoft.com/en-us/windows/win32/api/bcrypt/nf-bcrypt-bcryptgenrandom

 for more details

Signed-off-by: Steve Lhomme <robux4@ycbcr.xyz>

Steve Lhomme authored 10 months ago and Jean-Baptiste Kempf committed 10 months ago

The reason we build tar is because it's missing xz support.

Rémi Denis-Courmont authored 11 months ago and Marvin Scholz committed 10 months ago

RFC9110 specifies that a client must handle a shorter response range
than requested in all circumstanges. Previously, RFC7233 only required
that behaviour for multipart ranges, which VLC did not use.

This matches the newer specification: VLC will try to resume from the
last received offset not only on unexpected error, but also on short
response.

Fixes #28627.

Especially when it's modified later.

Some dependencies are not applied properly otherwise.

Inverted test introduced in 9c80a115.

It requires the native qmlcachegen to build.

And we don't need the direct dependency on qsb.

We only need it as a native tool we call.

We don't call the cross compiled tools. fxc is pretending to be a native tool.

We don't need the qsb compiler to get the library dependencies of VLC.
In any case it should be the native one, not the cross-compiled one.

We never use that code directly. We use qsb but as native with manual detection.

Regression from 63f3e396

Thomas Guillem authored 10 months ago and Steve Lhomme committed 10 months ago

See previous commit.

Thomas Guillem authored 10 months ago and Steve Lhomme committed 10 months ago

Player timer states are not really states but events. It is possible to
receive a discontinuity while paused, in that case, the paused state was
overridden causing UI timers to continue while paused. Store each events
in different variables to fix this issue. Keeping the discontinuity
event in the internal player state is useless since
1ecb283c.

Fixes #28622

Thomas Guillem authored 1 year ago and Steve Lhomme committed 10 months ago

This module enables SpatialAudio when playing multi channels content.

This module has finer A/V sync delay:
 - 15-40ms on speaker/HDMI
 - -80ms with AirPlay
 - 80ms with bluetooth devices (was 250ms before).

Airplay playback is now more stable, the VLC delay (via
aout_TimingReport) is less than 100ms, so audio and video will play
almost instantaneously.

With audiounit_ios, the delay was 2 seconds making it harder to sync
audio and video (since input tracks had to be delayed too).

Note that even if the audio seems to be playing from the VLC side, the
airplay will be silent for the first 2seconds. This can be improved by
increasing the pts-delay (file-caching): 7seconds of pts-delay remove
this latency (from 2secs to 100ms), but this is not something that could
be enabled by default.

One other improvement is that the airplay device will stop immediately
when pausing/stopping/seeking.

This module has a priority of 100, less than auhal, that can handle
passthrough, but more than audiunit_ios. Therefore, this module will be
the default one on iOS/tvOS, but can be selected by the user on macOS
(or deselected on iOS/tvOS). This is temporary as it might be
interesting for macOS users to use this aout as it can handle Spatial
Audio.

Thomas Guillem authored 11 months ago and Steve Lhomme committed 10 months ago

No changes since the other aout module is macOS only (auhal) and won't
interfere.