Blender Git Loki

GHOST/X11: enable EGL

This will replace GLX with EGL for X11. GLEW does not support GLX and EGL
at the same time. Most distributions build GLEW with GLX support, so we
have to use the externally provided GLEW and build with EGL support.
This effectively sets WITH_SYSTEM_GLEW to OFF for all Linux configurations.

Differential Revision: https://developer.blender.org/D12034

Fix T89976: Mirror Keys By Value performs wrong scale conversion

In the graph editor, Mirror Keys by Value would convert the value to mirror
over, to account for different units for linear & rotational properties.
The conversion was done in the different direction, though, resulting in
values that were too large by a factor of (180/pi)^2.

deps/win: Remove media foundation dep for ffmpeg

This caused a blender load error on windows N,
given we do not use these codecs they can safely
be disabled.

This will fix T90200 once the new libraries are
in SVN

Render: disable Z pass by default, leave only Combined

It was somewhat arbitrary to have this one pass enabled that adds a bit of
additional memory and render time overhead, even though it's not necessarily
more important than others.

Render: remove unused Blender Internal view layer settings

These should have been removed earlier but were forgotten.

Fix T90295: inconsistent render pass order between Cycles and Eevee

Cleanup: Consistent indent style for asset CMakeLists file

Was mixing 4 and 2 space indent in a single file.

Cleanup: Pass asset handle to asset iterator, rather than wrapped file

This iterator was introduced before `AssetHandle` existed, so it was
dealing with the file data directly. Now we want as little code as
possible to deal with the file data, all access should happen via the
`AssetHandle`.

Assets/UI: Sanity check argument for UILayout.template_asset_view()

Was already doing some sanity checks, but wasn't checking if the passed
property actually is a collection property, which is important.

Cleanup: Remove unnecessary code for asset view UI template

From what I can tell there is no reason anymore to do this. The design
has changed since this was added.

Assets/UI: Resolve major asset view UI template limitation

Before this, all asset view templates showing the same asset library
would show the same assets, even if they should show different ID types.
That was a major limitation since the design did forsee that this
template can be put anywhere in the UI to display various sub-sets of
assets.

Initially I did the ID type filtering close to the asset-list reading,
because I wanted to optimize reading so that we would only actually read
asset information from disk of the ID type to be shown. But this will be
quite complex and I'm not sure if I'll get to work on this anytime soon.
So this commit moves the filtering to the template display level solving
this limitation.

Note: This also adds the code to filter by tags, together with the ID
type. But it's not actually used anywhere yet.

install_deps: update OIDN to 1.4.1, and ISPC to 1.16.0.

Ref. T88438.

Install_deps: Add zstd dependency.

Ref T88438.

Install_deps: add flex dependency.

Was already installed on Debian-like and Fedore/Suse actually, now also
explicitely required on Arch and listed in docs.

Ref. T88438.

install_deps: Update llvm to 12.0 (with minimal now being 11.0).

Ref T88438.

install_deps: Update OIIO to 1.1.15.1.

Ref. T88438.

Cleanup: OSL buildlib patch: Remove `.rej` part.

This patch contained changes for an `.rej` rejection file generated by
failed patch apply... Definitly nothing to do here.

install_deps: Update OSL to 1.11.14.1.

This has been a huge pain to get working, for several reasons (new flags
needed, patching is now mandatory, etc.).

Further more, discovered that debian OIIO package is now silently
relying on OpenCV, without even proper handling of this dependency (at
least in the `-dev` package), so had to revert to force-build own OIIO
again on that distro for the time being.

Ref. T88438.

Fix 'BLI_task_parallel_mempool' keeping 'user_chunk' unchanged

When `BLI_task_parallel_mempool` does not use threading, the
`userdata_chunk` is allocated locally simulating a TLS.

However `func_reduce` is not called so the original chunk is ignored.

`task_parallel_iterator_no_threads` is another function that doesn't call
`func_reduce`. It also ignores `userdata_chunk_local` in the main iterator.

The solution in these cases is not to create a `userdata_chunk_local`.

This fixes T90131

Differential Revision: https://developer.blender.org/D12067

deps: Reduce llvm/clang footprint for windows

We shipped the whole bin folder for llvm/clang
while we only needed clang-format, by shipping
just the bits we need we save about 700 megabytes
off our svn lib download.