Blender Git Loki

Make mesh modifier tests not fail with optimized OSD lib

When building opensubdiv with more aggressive optimization flags
(-march=native -02) the output meshes would differ a bit from what we
expected in the current automated modifier test file.

The differences in vertex position is within the 1e-6 range, which I
would call is acceptable for floats. In addition to this, all the
modifier test that tests the subdiv modifier in particular pass without
any modifications. I've updated two tests in the modifier test file and
script to make it pass (listed below).

Updated following test categories:

1. Decimate test

Here there was a subdiv modifier applied before the actual decimate
modifier. Because the decimate modifier creates a queue of potential
vertices it can remove, it is highly sensitive to even small changes as
it drastically changes in which order the vertices are decimated in.

As this test should only be testing the decimate modifier, I pre-applied
the subdiv modifier in the test file.

2. RandomCubeModifier
For these tests I removed the subdiv modifier as well. As with decimate,
a small change in vertex position here can lead to quite different
results.

Reviewed By: Sergey, Bastien

Differential Revision: https://developer.blender.org/D9004

Fix T83876: blender crashes when baking particles + smoke sim

This was introduced in rBe5c0d4613a8943c712b57fb336997ecd78e6508e.

The issue is that the new fluid system does not use the pointcache
system for caching anymore, but still relies on pointcache for
other things. For example, it uses DEG_add_collision_relations
which internally creates relations with pointcache. Not sure if
there are other issues.

Ideally, this dependence should be resolved in one way or another
at some point, but that is out of scope for this fix.

Differential Revision: https://developer.blender.org/D9984

Reviewers: brecht

Fix T84260: NURBS edit mode lines not showing

When in edit mode, the edit lines for de-selected surfaces did not
show up.

The bug was caused by the is_gpencil bool which reused another flag.
Both grease pencil and nurbs surfaces use the edit_curve_handle shader.
A dedicated flag was added to make sure the is_gpencil bool is
set correctly.

Reviewed By: fclem

Maniphest Tasks: T84260

Differential Revision: https://developer.blender.org/D9985

Fix T84095: Eevee vextex color isn't working with hair

Regression introduced by {c766d9b9dc56}. When converting the vertex
buffer to a texture buffer the fetch mode wasn't checked and the short
was bitwise interpreted as a float. This change checks the fetch_mode
and select the correct texture buffer.

This could also be added to other places when needed. At this time it is
only added here to support vertex colors when used with hair particles.

Fix: Update normals when switching scene quality

Recent commits also updated normals for metaballs, curves and volumetric
objects. This change tags will tag to generate geometry for these new
types.

GPU: Add HQ normals workaround.

This change makes it possible for platforms to only support high quality
normal rendering. This is part of {T82856} where current AMD drivers
running on the polaris architecture does not support the low quality
setting due to a driver bug.

In a next commit the work around will be enabled.

GPU: Enable HQ normal work around for AMD Polaris

THe high quality normals work around is enabled for Polaris cards using
the official drivers. Since driver version 2.11.2 they fail to render
using low quality normals.

The detection of polaris cards is done by matching the opengl renderer.
The renderer strings have been extracted from various reports linked to
{T82856} but isn't complete as some reports are missing the exact
renderer as users don't always report via the help menu.

Fix mistake in rBef90a8e12caf in data transfer code.

Old mistake from 2.80 era when we got rid of DerivedMesh...

Not critical, but could be backported to 2.90LTS too.

Fix T84362: Crash when use Vertex Paint in subdivide strokes

Maniphest Tasks: T84362

Differential Revision: https://developer.blender.org/D9983

Fix T84345: Transforming the cursor fails with absolute grid-snap

Absolute grid snapping was using the pivot, which doesn't make sense
when transforming the cursor.

DrawManager: High quality normals for non meshes

This adds high quality normals for non meshes. These include

* Volumetric Object Wireframe
* Metaballs
* Extracted Curves
* Curves in edit mode

This is in preparation to fix a regression in recent AMD
drivers where the `GL_INT_2_10_10_10_REV` data type isn't
working in Polaris cards.

Fix T84106: attribute mix node worked incorrectly on float attributes

Fix imbuf.new & resize allowing zero & negative dimensions

Cleanup: remove unused optional argument to imbuf.new

While this didn't cause any problems, it was incorrect.

Fix T84191: remove python API limits for Image.scale() dimensions

Since the introduction in 2c23bb838979131667c922df1f01738e02e6ca16,
these were set to 10000 each.

This seems like an arbitrary limit (BKE_image_scale / IMB_scaleImBuf
don't have limits and I couldn't spot similar size restrictions in
image relating functions), now match what we do for creating images
(rna_Main_images_new), use INT_MAX.

Ref D9950

Cleanup: use const variables

Cleanup: correct array size in argument

Fix clang_array_check checking utility

Use python3 which is now supported,
only use CLANG_BIND_DIR & CLANG_LIB_DIR when they are set.

Also add immediate mode GPU API function calls.

Fix early exit check in BKE_appdir_app_template_has_userpref

Logical error in 84f21c170dda9e503de440c20bc2753002987901
This check was a no-op, replace this with a check for an empty string.

Fix crash reading files with unknown modifier ID's

Missing NULL check for the modifier type.