Buildbot has multiple cuda versions installed. Master selects the right
cuda version based on the architecture that is compiled. {D9179}.
When building BlenderLTS it defaults to CUDA 11.1 what isn't able to
compile `sm_30` architecture. This patch selects CUDA 10.1 as default.
Script create_msix_package.py will download the ZIP file
from the given URL. It will create the MSIX package
with the version number and publisher ID given.
Strongly recommended are the path to a valid PFX file, and the
password to use that PFX file. These are needed for signing
the resulting MSIX package. The signing step is optional though,
but the resulting MSIX package cannot be installed outside of the
Microsoft Store
Example
set VERSION=2.83.2.0
set URL=https://download.blender.org/release/Blender2.83/blender-2.83.2-windows64.zip
set PUBID=CN=PUBIDHERE
set PFX=X:\path\to\cert.pfx
set PFXPW=pwhere
python create_msix_package.py --version %VERSION% --url %URL% --publisher %PUBID% --pfx %PFX% --password %PFXPW%
Requirements:
* Python default from the Microsoft Store should do (3.8)
* requests can be installed with `pip install requests`
Note that for an LTS release that gets uploaded to its own LTS application release
in the store you need to specify the `--lts` switch on the command-line to the script.
Upon completion there will be a file with the
name blender-2.83.2.0-windows64.msix. In case PFX file and its password were
given on the command line MSIX package will also be signed for the Microsoft Store.
Related Wiki page: https://wiki.blender.org/wiki/Process/Release_On_Windows_Store
Reviewed By: jbakker
Maniphest Tasks: T77348, T79356
Differential Revision: https://developer.blender.org/D8310
On user level this fixes dead-lock of OpenCL render on Intel Iris GPUs.
Note that this patch does not include change in the logic which allows
or disallows OpenCL platforms to be used, that will happen after the
kernel fix is known to be fine for the currently officially supported
platforms.
The dead-lock was caused by wrong usage of memory barriers: as per the
OpenCL specification the barrier is to be executed by the entire work
group. This means, that the following code is invalid:
void foo() {
if (some_condition) {
return;
}
barrier(CLK_LOCAL_MEM_FENCE);
}
void bar() {
foo();
}
The Cycles code was mentioning this as an invalid code on CPU, while in
fact this is invalid as per specification. From the implementation side
this change removes the ifdefs around the CPU-only barrier logic, and
is implementing similar logic in the shader setup kernel.
Tested on NUC8i7HVK NUC.
The root cause of the dead-lock was identified by Max Dmitrichenko.
There is no measurable difference in performance of currently supported
OpenCL platforms.
Differential Revision: https://developer.blender.org/D9039
The example file in T81218 has a driver that maps a bone's X-location to
the number of BBone segments. This caused a dependency cycle, which
resulted in bad thread serialisation, which caused the crash.
This patch breaks the dependency cycle `BONE_LOCAL` →
`DRIVER(bones["Bone"].bbone_segments)` → `BONE_LOCAL`. The 'Driver Data'
relation now points to `BONE_SEGMENTS` when the driven property starts
with `bbone_`.
Differential Revision: https://developer.blender.org/D9122
The output layout was wrong and it's a mistery why it works on most
implementations since it's clearly a wrong usage.
Thanks @sebbas for helping narrowing down the issue.
Warp & weight vertex-group editing modifiers miscalculated vertex weight
inversion, the weights were multiplied before being subtracted from 1.
Ref D8241
Object in the 2.7x file is on no layer at all, while this is not really
expected nor considered valid, better to cope with the case gracefully
than crash.
To be backported to 2.90 and 2.83.
Was missing a dedicated entry for LMB select, this is to be consistent
with how it is done for meshes.
Maniphest Tasks: T80905
Differential Revision: https://developer.blender.org/D8935
On systems with 512kb stack this happened at around 13k points.
This happened at times with grease-pencil, although callers that
frequently use complex polygons should be using BLI_polyfill_calc_arena.
RNA diffing code was not dealing properly valid NULL PointerRNA (like
the empty texture slots of a ParticleSettings e.g., which were cause of
crash in that report).
To be backported to 2.90 and 2.83.
During undo/redo read code is expected to clear the `OB_MODE_EDIT`
bitflag of `Object.mode`, for some reasons.
This was not done anymore for re-used Objects, we need to add a special
handling case for that too.
Should be backported to 2.90 and 2.83 (will probably not be straight
forward for the latter).
Entering edit-mode after creating shape keys on a blank mesh would crash.
Regression in 9b9f84b317 which prevented initializing empty
shape keys when there is no shape key offset data available.
This prevents Blender from crashing when switching
to rendered view and the grid_data is NULL in the
lightcache (due to possible corruption), by switching
to the fallback lightcache. The fix extends the solution
for possible corruption in the cube_data as well.
Fix T79737
Reviewed By: fclem
Differential Revision: https://developer.blender.org/D8835
The state of the passes was not set back after the first sample.
This is a simpler fix than rBecfbc5fb55d5 which needs the new draw manager
features from 2.90.
custom split normals data
Clearing custom split normals would get rid of the CD_CUSTOMLOOPNORMAL
layer - but editing data `lnor_spacearr` would be kept.
Adding a CD_CUSTOMLOOPNORMAL layer (if none exists yet) should be done
in `edbm_average_normals_exec` / `BKE_editmesh_lnorspace_update` /
`BM_lnorspace_update` / `BM_lnorspacearr_store`. The thing is that if
the editing data `lnor_spacearr` would still be valid after `Clear
Custom Split Normals Data`, blender would happily call
`BM_lnorspace_rebuild` instead. Doing that without a CD_CUSTOMLOOPNORMAL
layer is asking for trouble.
Now clear lnor_spacearr on `Clear Custom Split Normals Data` as well.
Thx @mont29 for feedback here.
Maniphest Tasks: T80159
Differential Revision: https://developer.blender.org/D8730
When Blender is started in fullscreen mode from the command line,
or if the fullscreen state is saved in the startup file, all temporary windows
will also open in fullscreen mode. When closing the fullscreen File Browser,
Blender would either crash or parent window becomes black.
This does not happen if the Blender switches to full screen manually.
`NSWindowCollectionBehaviorFullScreenPrimary` should be set for windows that
can enter full-screen mode. Otherwise macOS will turn the wrong window into
full-screen.
Similar fix: rB4b39de677d20
Differential Revision: https://developer.blender.org/D8708
Reviewed by: Julian Eisel
If a different object was active, clicking on a linked armature's pose
in the Outliner would enter Pose Mode for it.
This would actually cause a failed assert, but in release builds the
armature would just enter pose mode.
Steps to reproduce were:
* Link in armature object
* Activate a different object
* In the Outliner, un-collapse the armature object
* Activate Pose Mode by clicking on its pose there
Problem is again with the embedded data, we want to make those local
together with their owner ID, but sometimes we are actually dealing with
copies here, which are inheritently already local.
Code did not considered that possibility before, leading to access to a
NULL `lib` pointer.
This should also be back-ported to 2.83 LTS release.
Maniphest Tasks: T80104
Differential Revision: https://developer.blender.org/D8731
Issue was with our dear posebones again... when applying overrides we
keep the same address/pointer for the IDs themselves, (which avoids us
the need to remap their usages), but their inner data is often
re-allocated.
Therefore, we need once again to go over armature objects and invalidate
their posebone pointers.
This should also be back-ported to Blender LTS 2.83.
Maniphest Tasks: T80078
Differential Revision: https://developer.blender.org/D8734
Checks to preserve the active spline on duplication
required an active vertex too.
Now having no active vertex doesn't prevent duplicate
from keeping the spline active.
Reviewed by: @mano-wii
Ref D8729
Steam just released a SteamVR update with OpenXR Developer Preview
support:
https://steamcommunity.com/games/250820/announcements/detail/2396425843528787270.
Once SteamVR is set up for OpenXR (see link above), it works with
Blender "out of the box", thanks to OpenXR!
We have to apply the sRGB transform workaround for SteamVR though,
otherwise it renders way too dark. Done in the next commit.
Note that AMD users may still only see a pink screen, because the
OpenGL-DirectX compatibility fails. I will check on a fix again.
For SteamVR on Linux we may have to wait for until it supports OpenGL
rendering for OpenXR. Alternatively, we *could* add initial Vulkan
support at Ghost level and use Vulkan<->OpenGL interoperability
extensions, Monado uses these as well.
This skips the subdivision operation if the mesh has no loops, avoiding
the crash.
Reviewed By: sergey
Maniphest Tasks: T80039
Differential Revision: https://developer.blender.org/D8696
Use invert_m4_m4_safe_ortho when joining objects so zero scaled axis
doesn't cause all points to be scaled to zero.
Instead geometry is left un-scaled on degenerate axes.
Report a warning in this case since users may want to adjust the
active objects scale.
Unlike invert_m4_m4_safe, this calculates zeroed axes.
Useful when we need to use the inverse of an objects matrix,
keeping the valid axis, only filling in the zeroed ones.
This prevents UV layer mix up in
MeshBatchCache.cd_used/cd_needed/cd_used_over_time which depends on the
extraction method.
One object's mesh can be accessed with MR_EXTRACT_MESH, another object
that uses the same mesh can use MR_EXTRACT_BMESH based on
(Object.mode & OB_MODE_EDIT), this causes a problem as the edit-mesh
and the mesh aren't always in sync, the custom data layers wont
necessarily match up, causing T77359.
Reviewed by @jbakker, @brecht
Ref D8645
