Was happening when modifier stack detected that mesh is not deformed
and is not modified and attempted to share result across multiple
objects.
This was introduced in 2f77119.
Now functions which are supposed to return mesh owned by caller will
do so again. Shouldn't be a huge impact on memory print since the
data layers are referenced.
This also fixes commit rB1fd7b380f4cf8a0489b405de2819f228a4da5ea2 which
didn't do allocation for effect strips properly.
Reviewed By: brecht, campbellbarton
Differential Revision: https://developer.blender.org/D4970
Now a small threshold is used for mouse input,
avoiding delay when gizmos are activated on drag.
Tablet input threshold remains unchanged since
it's easier to make small movements when using a tablet.
A larger threshold for non-cursor input is now used (typically keyboard)
which improves usability when the "Pie Menu on Drag" key-map preference.
The idea is to share a mesh data-block as a result across all objects
which are sharing same original mesh and have no effective modifiers.
This mesh is owned by an original copy-on-written version of object data.
Tricky part is to make sure it is only initialized once, and currently a
silly mutex lock is used. In practice it only locks if the mesh is not
already there.
As an extra bonus, even viewport memory is also lower after this change.
Reviewers: brecht, mont29
Reviewed By: brecht, mont29
Differential Revision: https://developer.blender.org/D4954
This removes the dither patterns visible in wireframe mode.
This does decrease de depth perception but many users complained
about the visual noise it produces.
This commit extends dependency graph API with an argument which
denotes that all custom data layers are to be preserved. This
forces modifier stack re-evaluation with more inclusive mask.
Far from ideal, since this might fail in certain configurations
with indirectly used objects which might be missing layers needed
for the current object evaluation. But this is how it worked for
a long time, so should be good enough for until more sophisticated
solution is found.
In order to use this new behavior two things are to be passed:
- Pass keep_all_data_layers=True
- Pass a valid dependency graph.
The dependency graph is only needed if keep_all_data_layers=True
and is NOT to be passed if keep_all_data_layers=False.
If keep_all_data_layers=True the dependency graph MUST be passed.
Reviewers: mont29, brecht
Reviewed By: mont29
Maniphest Tasks: T64994, T64794
Differential Revision: https://developer.blender.org/D4940
It's closer to the default matcap now, but slightly less metallic and dark. The
reason to use studio lights as default is because the roughness and metallic
parameters of the material then have an effect, and because Texture color mode
does not work for matcaps.
Evaluation must never go to original objects and query them:
this is a huge violation of the entire idea of separating
state across viewports and render engines.
Allowed this to only happen for active dependency graph, where
we at least know order of dependency graph update and user
input.
That would break proxy behavior after a library reload.
The usual super-annoying loop-back pointers... At least that one is
easily detectable and can be fixed in-place.
Found while investigating T64764.
Masks were not really covered by Copy-on-Write due to mistake
in the dependency graph. After correcting that mistake a lot
of tools became broken, so majority of the patch is related
on making it so access to evaluated/tessellated masks is done.
When accessing evaluated mask state make sure access to an
evaluated dependency graph is done. This solves possible
access to NULL data on redo.
Fixes T64899: Re-doing new point addition causes crash
Reviewers: brecht
Reviewed By: brecht
Maniphest Tasks: T64899
Differential Revision: https://developer.blender.org/D4918
Those are two cases where keeping infamous backward `parents` pointers
of collections in sync is kind of impossible to do... So rebuilding
those relationships from scratch instead.
Fixes e.g. a crash when undoing, then reloading a library, and likely
many more weird ones like that.
Uncovered while investigating T64764.
It's not always possible to keep 'by hand' parent relationships valid in
collections hierarchy. Add functions to remake those
(re-using/factorizing code from `readfile.c` `lib_link_collection_data()`
function).
Can't stress again how painful it is to have those kind of backward
relationships in our data structures, those *always* end up being
serious issues to keep in sync... Should only be generated on the fly
when needed, period. :(
Note that this only fixes the core issue reported (caused by own dummy
mistake in rBd0df7fb3b94ea), investigating that report uncovered at leat
two more issues, including a crasher (when reloading after an undo)...
This will be needed in undo/redo case, since we do not re-read linked
IDs, their usercounts become total garbage (especially in 'used by local
ID' cases)...
We are not exposing RNA_ObjectBase in the 2.80 API.
Thus we can't have operators relying on it (e.g, CTX_data_visible_bases,
CTX_data_active_base, ...). Otherwise users won't be able to override
context for these operators.
This commit keep the CTX_data_.*bases() functions around so we don't
need to change the operators and potentially break things that late into
2.80. However as far as the Python scripters are concerned there is no
base to be overriden, ever.
That also simplify the guessing game addon developers have to play when
trying to override an operatori context. They still need to find whether an
operator requires editables, visibles, selected, ... objects. But at
least they don't need to find out whether the operators need base or
object.
The maximum particles per task of 256 was outdated and lead to too much thread
contention. Instead define a low fixed number of tasks per thread.
On a i7-7700HQ, creating 4 million particles went down from 31s to 4s.
Thanks to Oscar Abad, Sav Martin, Zebus3d, Sebastián Barschkis and Martin Felke
for testing and advice.
Differential Revision: https://developer.blender.org/D4910
When rendering viewport to an offscreen buffer the buffer was
constructed for non anti aliasing (0 samples). This made the objects
that are drawn by the `object_mode` including `wireframe` draw type
non-anti-aliased.
The offscreen buffers will be constructed based on the user setting for
viewport multisampling (`U.ogl_multisamples`). The same setting will
also be used when previewing scene strips in the sequencer. For now
this only improves wireframe drawing in the scene strips. To improve the
Anti aliasing in the scene strips we need to get finer control in the
draw manager. This will be part of a different patch I am preparing.
Please note that this patch also cleansup some unused code in the offscreen rendering (FSAA code was still existing, but never called)
Reviewed By: brecht
Maniphest Tasks: T64849
Differential Revision: https://developer.blender.org/D4907
This is old logic that no longer makes sense in the new depsgraph, and causes
issues when multiple threads try to modify the same bevel object.
Differential Revision: https://developer.blender.org/D4913
`BKE_collection_object_add_from()` would not check wether collections
were local or not... Trivial to fix.
Note that here I assume we do not use that function in some special
cases where we would like to edit linked datablocks. Think that is
reasonable stance, though.
* Replace Log view transform with Filmic Log.
* Remove Rec.709, DCI-P3 displays that were incomplete and outdated.
* Remove outdated RRT and Film transforms, replaced by Filmic.
* Remove camera responsive curves that don't work with HDR colors.
* Rename Default view transform to Standard.
We're breaking compatibility now for 2.80, so that we can add future
improvements on a clean config.
Part of the code was contributed by George Vogiatzis in D4782.
Differential Revision: https://developer.blender.org/D4900
