Was using edge-slide & remove-doubles but this was error prone since remove doubles could fail in some cases or find doubles where it shouldn't (with very small scale objects).
This gives more predictable behavior when the edges of a loop wouldnt slide (in that case they would just drag over to one of the sides with no user control)
and multiple edge loops work better too. eg:
- http://www.graphicall.org/ftp/ideasman42/edge_loop_del_update.png
- ED_view3d_from_m4() got incorrect rotation from scaled cameras, was noticable with smoothview transitions.
- when you lock the camera to the view, any view edits would reset the cameras scale to 1.
- another problem with view locking if the camera was scaled and had a parent, the parent would be transformed incorrectly.
- fly mode was chaning object scale a little over time, now restore after applying scale so it never changes.
Only activate search-box items on mouse-release,
Otherwise this gives odd behavior when using the operator-search popup since some tools expect the mouse buttons not to be held when activating which includes operators that have their own popup menus.
In the case where the modifier stack didnt need deform-verts to calculate,
they would not be available for the preview either.
This fixes a bug caused by r57206 which set mirror to preview so the mirrored
weights would be displayed, but it only worked when there was an armature after it, see [#35545].
Hopefully it's more readable now. Took me a while to remmeber
all the stuff going on here while was looking into possibility
of implementing some feature here.
The tooltip seemed to hint that this tool is able to resolve all manner of
gimble-lock situations by untangling the curves (i.e. performing some kind of
equivalent-angles resolution, keeping in mind the nearest situations nearby).
However, this tool currently only performs corrections for the most basic case
when large jump+flip discontinuity artifacts appear in euler rotation curves as
a result of rotation values getting clipped to +/- 180 degrees, which arises
when these rotation curves are the result of baking some physics sim or so.
(Also, fixed an unrelated "replace-all" typo in a comment)
proposal was to add another special init hack for the viewer node->id, but rather would do it right and so moved all the special init hacks for constant ID backpointers (Scene for RenderLayer, Composite,
Defocus, FileOutput and MovieClip for MovieClip, MovieDistortion and Stabilization nodes). These are now part of the local init callbacks functions of the appropriate nodes, using the new initfunc_api
callback which takes a Context pointer, so they have access to Scene.
Couple of issues here:
- User shouldn't be able to run into dupligroup recursion.
It was checking already when setting a group for dupli.
Added check to operator which adds object to group.
- It's still possible files with recursion are hanging around,
so made simplify function robust to such kind of crap.
Gives approx 2x speedup on my laptop on such operations
as mesh subdivision in edit mode. Desktops with fancier
CPUs could benefit even more.
Thanks Campbell for review!
Issue was caused by missing X/Y displacement components
flip when flipping the normals (flipping the normals changes
the tangent space apparently and displacement vectors need
to be modified to correspond to new space).
Reported by Jonathan Williamson in IRC.
Now add_freestyle() in pipeline.c takes a second argument to enable/disable
stroke rendering. When stroke rendering is disabled, the function allocates
data structures but does not perform stroke rendering. The allocated data
structures (mostly left unpopulated with data elements) are intended to allow
for the Read Full Sample Layers (Shift-R) command in the compositor.
