Fatih Uzunoğlu authored 9 months ago and Steve Lhomme committed 9 months ago

Windows 11 identifies as Windows 10 (NT 10),
so this comparison always yields false.

As a side note, native acrylic background
for Win32 windows only gained support with
the Windows 11 22H2 update, that's why
there is the check here.

Note that according to the general conventions
a transparent window must be frameless on
Windows. This is currently not the case,
because "frameless window don't support
areo snapping". So, it can be considered
as a known bug if native acrylic does not
work as intended. This grants a requirement
of change to use frameless window, and get
rid of the current workaround that simulates
frameless window in mainctx_win32.cpp. If
aero snapping does not work, it is Windows's
problem since we are using frameless windows
in every other platform.

We don't use the Cluster otherwise and can read at all.

Ref #28653

Changes for 1.4.2 'Road Runner':
--------------------------------

1.4.2 is a small release of dav1d, improving notably ARM, AVX-512 and PowerPC
 - AVX2 optimizations for 8-tap and new variants for 6-tap
 - AVX-512 optimizations for 8-tap and new variants for 6-tap
 - Improve entropy decoding on ARM64
 - New ARM64 optimizations for convolutions based on DotProd extension
 - New ARM64 optimizations for convolutions based on i8mm extension
 - New ARM64 optimizations for subpel and prep filters for i8mm

 This release includes enhancements and bug fixes.

  - Upgrading:
    This release is ABI compatible with the previous release.

  - Enhancement:
    Improved the detection of compiler support for AArch64 extensions,
    particularly SVE.

    Added vpx_codec_get_global_headers() support for VP9.

  - Bug fixes:

    Added buffer bounds checks to vpx_writer and vpx_write_bit_buffer.
    Fix to GetSegmentationData() crash in aq_mode=0 for RTC rate control.
    Fix to alloc for row_base_thresh_freq_fac.
    Free row mt memory before freeing cpi->tile_data.
    Fix to buffer alloc for vp9_bitstream_worker_data.
    Fix to VP8 race issue for multi-thread with pnsr_calc.
    Fix to uv width/height in vp9_scale_and_extend_frame_ssse3.
    Fix to integer division by zero and overflow in calc_pframe_target_size().
    Fix to integer overflow in vpx_img_alloc() & vpx_img_wrap()(CVE-2024-5197).
    Fix to UBSan error in vp9_rc_update_framerate().
    Fix to UBSan errors in vp8_new_framerate().
    Fix to integer overflow in vp8 encodeframe.c.
    Handle EINTR from sem_wait().

And without ragel.

- snap is left out because there's a new image coming

We cannot mix system tools and tools in contribs because qtpaths6 will only provide one of those.
So either we build all of them, or we build none of them.

For native compilation, native tools are built with the libraries.

Alexandre Janniaux authored 1 year ago and Steve Lhomme committed 9 months ago

The memory and file keystore are always used when running the test from
meson, so ensure they are built before.

Alexandre Janniaux authored 9 months ago and Steve Lhomme committed 9 months ago

Using relative path for plugin path leads to the test failing to launch
depending on where it is being launched from.

Alexandre Janniaux authored 9 months ago and Steve Lhomme committed 9 months ago

The variable is already quoted so there's no need to x-test them.

This reverts commit 4cd7d490.

This warning can occur in callbacks that never return. Some functions cannot
have a return statement because that would cause dead code warnings, while
they also cannot return void because the callback prototype is externally
mandated. Ideally the compiler would notice that the function does not
return, and sometimes it does, but evidently we cannot rely on that, so
we have to revert this warning back to a warning, not an error.

Fixes #28650.

Alexandre Janniaux authored 3 years ago and Steve Lhomme committed 9 months ago

The script explains how to generate the xcode project and will generate
a folder containing the project at the current working directory where
it has been used.

Maxime Chapelet authored 3 years ago and Steve Lhomme committed 9 months ago

We have to generate dSYM bundle from the script because Xcode might
already have generated them before binary lipo/copy execution.

Alexandre Janniaux authored 4 years ago and Steve Lhomme committed 9 months ago

Add an XCodeGen project spec to generate an XCode project which will be
using autoconf and automake to build, as well as exposing the generated
artifacts to other XCode projects.

This will be generating the build artifacts inside the product build
directory from XCode, so `Clean build folder` will work as expected in
the editor, and use dynamic plugins to copy them towards the final
application for now.

This project allows testing VLC directly through the iosvlc.m test
driver, without building VLCKit nor VLC for iOS or another test
application, by directly using XCode to run it on the targetted
platform.

Exposing such project also allows debug symbols and debugguer with
breakpoint support to work automatically, even when using the artifacts
in another project.

The xcode.sh script is added as a wrapper script to handle the
specificities of XCode when calling external buildsystem, in particular
the multi-arch building for archiving application and packages.

The copy_plugins.sh is added as a build phase script so that application
can install the plugins matching the architecture being built. It
currently probably fails with cases where some plugins are not built for
every architecture of a platform since I haven't tested that yet. It is
also quite "dumb" and re-do everything from scratch for now.

Alexandre Janniaux authored 3 years ago and Steve Lhomme committed 9 months ago

Ensure the flag definitions are stored in the config.status file.

Alexandre Janniaux authored 3 years ago and Steve Lhomme committed 9 months ago

When building from XCode, the build folder won't be located into a
sub-directory so remove the restriction.

Alexandre Janniaux authored 3 years ago and Steve Lhomme committed 9 months ago

The contribs were built from contrib/contrib-xxx but installed into
build/contrib/xxx-yy so removing the build folder led to the contrib
being unused since every .target files from contrib/contrib-xxx would
still exist.

Alexandre Janniaux authored 3 years ago and Steve Lhomme committed 9 months ago

Using absolute path enables the script from being called from a
different folder than a subfolder of the source directory in the
future.

Alexandre Janniaux authored 9 months ago and Steve Lhomme committed 9 months ago

The build system is generating plugins and installing them at the PREFIX
location, but it is also often used by extra build scripts using it.
Those build scripts were trying to list the plugins by listing the files
in the PREFIX folder, but it was often inadequately using stall
plugins.

By generating an up-to-date list of plugins by the build system, we can
check when the plugin list have changed and which plugins should be
copied to the final application.

Alexandre Janniaux authored 9 months ago and Steve Lhomme committed 9 months ago

The simulators are constrained to specific iOS and tvOS minimal version
and not respecting the constraint would lead to incorrect build and
linking failure.

Not sure how this should really be handled yet though, maybe the
build.sh script should get some parameter for that, or define the
targets depending on the SDK version.

Alexandre Janniaux authored 3 years ago and Steve Lhomme committed 9 months ago

When run from an environment not supporting colors, we should remove
them instead of printing the color descriptor anyway.

Apparently on ArchLinux qmake6 is in the PATH, but not qtpaths6.

Khalid Masum authored 9 months ago and Steve Lhomme committed 9 months ago

Previously:
- $PATH was set incorrectly to wasm-emscripten/bin
  which is supposed to be wasm-emscripten/build/bin

- $PATH was updated after making the tools

Hence,
 - The tools would be rebuilt every time build script is run.

 - The setting of PATH variable would do nothing anyway even
   if it was correct, as correct PATH is going to be set by the
   configure script later anyway.

This patch will repurpose the line to avoid rebuilding of already
built tools.

This fixes the M4 parsing in VSCode at no cost for VLC.

Alexandre Janniaux authored 10 months ago and Steve Lhomme committed 9 months ago

When the trace is not matching one of the expected field, continue was
called before the iterator was incremented, leading to infinitely
calling continue.

Co-authored-by: Thomas Guillem <thomas@gllm.fr>