Age | Commit message (Collapse) | Author |
|
filling it for verts/edges/faces on every redraw.
this may introduce bugs which I didn't catch, but they are very easy to identify in a debug build which has asserts to ensure the arrays are valid before use.
in my own test drawing ~98,304 quads - this gave an overall ~16% drawing speedup.
|
|
|
|
- BM_mesh_elem_toolflags_ensure / bmo_flag_layer_alloc / bmo_flag_layer_free / bmo_flag_layer_clear
- BM_mesh_select_flush
- EDBM_index_arrays_init
notes:
- mostly use openmp `sections` to split operations on vert/edge/face since this is a fairly minor change.
- split tool flag pool in 3, this means we can allocate exact sizes needed and iterate on them in threads without alloc'ing.
|
|
BLI_array_alloca() for smaller arrays.
|
|
layer.
Reported and patch by MiikaH
|
|
give a corrected value, instead pass a fallback so callers don't allow zero by accident.
|
|
|
|
name length.
|
|
which uses stack memory always and doesn't need to be freed explicitly.
|
|
|
|
When camera is the pivot of 3d window, and you go to camera view, moving out of
view with MMB drag causes zooming to stop working. Zooms depend on view3d "dist"
value, which then became zero.
This fix just makes dist "1.0" then, arbitrary but keeps things at least work.
(Tried restoring to previous 'dist', but this fails in cases too)
|
|
range must be left to the default (0..1 and -1..1).
|
|
|
|
|
|
|
|
comments.
|
|
|
|
|
|
|
|
|
|
PROP_POINTER had PROP_UNSIGNED!). Harmless, but stupid ;)
|
|
Gives approx 10% speedup in my own micro-benchmark looking up operators.
|
|
|
|
properly.
|
|
are not actually used here.
|
|
|
|
|
|
edges/verts outside the clip area but the geometry would still get cut
|
|
mouse line would reset to 0/0/0
a few areas that use ED_view3d_win_to_segment_clip() didnt take into account the case where the segment was filly clipped, some callers even needed the segment not to be clipped.
- added ED_view3d_win_to_segment()
- ED_view3d_win_to_segment_clip() now returns FALSE if the segment is totally clipped, but the start/ends of the line are not zero'd as they were before.
|
|
clipping against the object space bounds.
more fixes needed here but at least correct whats there.
also ensure clipping is initialized from the object before handling modal operator events.
|
|
|
|
the mouse. - snap, angle snap, midpoint snap.
|
|
text stepping when ctrl is held.
This adds back the problem that double-clicking on a single char wont select it. Double click selection may need its own logic.
|
|
Regression from r34115.
|
|
duplifaces like blender internal.
|
|
Already since like 2.61 - when Array draw was added - drawing only shadow
in bottom of menus with bad looking corners.
|
|
sockets' operator, ctrl+h), newly enabled render passes will not show up in the Render Layers node. The SOCK_HIDDEN flag cancels out the SOCK_UNAVAIL flag in that case. Disable the SOCK_HIDDEN flag as well when showing new passes to avoid confusion.
|
|
adding meshes were scaling the user input values so the distance on the button didnt relate to the scale of the object added.
Now use an invoke function that scales unset default values.
|
|
|
|
from Tobias Johansson (mutze)
with some minor edits (don't treat '!' as an operator - python centric)
|
|
(would use MEM_mallocN in one case but malloc in another). better use blenders MEM_mallocN for both.
|
|
bevel modifier was making zero area faces & edges that made scanfill fail (since it no longer removes doubles when filling ngons)
|
|
|
|
Really was caused by a previous bevel making
a two-edged face, which caused other faces to
be dropped when copying a bmesh.
The quadstrip code needed to be more careful
to avoid creating two-edge faces.
|
|
as render engine. Still missing is colors and texture slots, but that's too
tricky to fix this close to release.
|
|
IK-Solver crashes Blender. No constraint is created for target in that case, just needed to add a check.
|
|
which only have one texture slot.
also quiet float/double warning.
|
|
This was a regression in svn rev52718 caused by the fact that we can not
free packet fun until we've finished all manipulation with decoded frame
since frame and packet could share same pointers.
For now restored old behavior of next_packet which seems to be well
tested and better not do bigger refactoring here so close to release.
Memory leak fixed by that revision was fixed by calling av_free_packet
just before avcodec_decode_video2 in cases we're at the end of file.
Tested with valgrind and could not see any memory leaks in ffmpeg
area.
|
|
was incorrect assert which didnt consider having no faces.
|
|
in the viewport.
|