Ensure that all threads on a multi-core system are used.
The issue was that BLI_task module was trying to be smart and
used heuristic to find optimal number of iterations per thread.
This heuristic assumes that tasks are light-weight, which is
not a case for subdivision surface.
On a higher subdivision level with a file from T70826 the
evaluation time goes down from 0.25 to 0.17 seconds per modifier
evaluation.
When D6189 is finalized we can being some extra performance
improvement.
Usage of spinlock during heavy IO gave reduced performance
see D6267 for details.
Reviewed By: sergey
Differential Revision: https://developer.blender.org/D6267
We have no dedicated image context tab, so for now making sure we don't
end up passing its ID as the pinned one.
If we ever get one, we then will need a different solution here, like
changing the ID that owns the data to be the object, instead of the
empty image datablock.
Differential Revision: https://developer.blender.org/D6273
rBc6cbcf83d015 caused to early out e.g when not all faces were selected
(but surrounding faces were, so implicitly all vertices were selected).
Now take (mixed also) selection mode into account.
Maniphest Tasks: T71554
Differential Revision: https://developer.blender.org/D6254
The CANCELLED return value from an operator is intended for
signaling that the operator aborted execution without changing
anything, and an Undo stack entry should not be created.
When a Python operator crashes, it is not safe to assume that
it did nothing, so it should interpret it as FINISHED instead.
Otherwise, the undo system behavior after an operator failure
feels broken.
Differential Revision: https://developer.blender.org/D6241
Rename and separate Layers and Materials Specials menu from other buttons for better consistency
Reviewed By: antoniov
Differential Revision: https://developer.blender.org/D6271
Was caused by "wrong" EOL characters used in the patch: the file is
actuallyu saved using CRLF EOL style.
The patch was using CRLF as well for until recent change in the C
runtime.
This commit adds a new command line argument --debug-ghost and
makes it so X11 errors happening during context initialization
are only printed when this new flag is sued.
There is no need to flood users with errors when their GPU is
not supporting latest OpenGL version. Or, at a very minimum,
the error must be more meaning full.
Differential Revision: https://developer.blender.org/D6057
In order to recover from a transient Focus Out - Focus In disruption
in the middle of a shortcut, which can be caused by certain window
managers, Blender has code that checks which modifier keys are pressed
after Focus In and restores the modifier state based on that.
If one of the Ctrl, Shift, Alt, Super keys is not mapped anywhere
in the active keyboard layout, XKeysymToKeycode returns the invalid
zero keycode, and reading the key state produces garbage, which can
cause an invalid modifier state. Check the return value to avoid this.
2.79 also did this [select the new instances] which was useful.
2.79 also kept the instancer selected [this patch deselects]
Reviewed By: mont29
Maniphest Tasks: T68191
Differential Revision: https://developer.blender.org/D6233
Some types were documented in bpy.types aren't accessible there.
For now, disable documenting types from add-ons and some types from
bl_operators, bl_ui... since these are mostly for internal use.
`WS_CHILD` is a different kind of child window that what we define as
child window. See http://forums.codeguru.com/showthread.php?491604.
Setting this style flag seems to mess things up a bit in our
configuration. The name bar is actually being overlapped by the Windows
task bar then. Not totally sure why this happens, but I think it's
because windows with the `WS_CHILD` style are positioned relative to the
parent, not the desktop (screen without taskbar). So it uses the full
space available when maximized, which isn't clipped by the taskbar
anymore.
This only frees brush_rng and random_tex_array when they were actually
previously allocated.
In a unit test (see D6246) I want to be able to partially start Blender
so that I can load a blend file. To prevent memory leaks, I also want to
be able to release memory, which currently requires calling
`BKE_blender_free()`. This unconditionally calls `RE_texture_rng_exit()`
and `BKE_brush_system_exit()`, which now crash on freeing `NULL`. This
patch fixes that.
Allocation (`BKE_brush_system_init()`) and freeing
(`BKE_brush_system_exit()`) are done asymmetrically. The allocation
functions are called from `main()` in the creator module, but the
freeing is done by `BKE_blender_free()` the Window Manager. Ideally we
symmetrise this and initialise Blender from outside the window manager
(so that the initialisation can be done without WM and Python too), but
for now I'm happy when things don't crash.
Reviewed by: sergey via pair programming
The render view window was never closed actually, just moved behind the
main window. It's properly closed now.
It should also behave more like expected when there already is a
temporary window open (e.g. Preferences).
This patch includes a modifiers that developed for NPR rendering.
- MultiStroke modifier that generates multiple strokes around the original ones.
Differential Revision: https://developer.blender.org/D5795
The script requires Python 3.7 as a very minimum, and CentOS is
only 3.6.
On macOC there was an access to a None object, due to missing
implementation of code signer on this platform.
