This commit makes BKE_image_acquire_ibuf referencing result, which means once
some area requested for image buffer, it'll be guaranteed this buffer wouldn't
be freed by image signal.
To de-reference buffer BKE_image_release_ibuf should now always be used.
To make referencing working correct we can not rely on result of
image_get_ibuf_threadsafe called outside from thread lock. This is so because
we need to guarantee getting image buffer from list of loaded buffers and it's
referencing happens atomic. Without lock here it is possible that between call
of image_get_ibuf_threadsafe and referencing the buffer IMA_SIGNAL_FREE would
be called. Image signal handling too is blocking now to prevent such a
situation.
Threads are locking by spinlock, which are faster than mutexes. There were some
slowdown reports in the past about render slowdown when using OSX on Xeon CPU.
It shouldn't happen with spin locks, but more tests on different hardware would
be really welcome. So far can not see speed regressions on own computers.
This commit also removes BKE_image_get_ibuf, because it was not so intuitive
when get_ibuf and acquire_ibuf should be used.
Thanks to Ton and Brecht for discussion/review :)
These timing data can then be used during conversion to Curve objects, to create a path animation (i.e. an Evaluation Time F-Curve) exactly reproducing the drawing movements.
Aside from this "main feature", the patch brings several fixes/enhancements:
* Stroke smoothing/simplifying will no more move the start/end points of a stroke (this was rather annoying sometimes!).
* Also optimized smoothing code (even though not really noticeable on a modern computer, it now uses less memory and runs faster).
* When converting to curve, you now have the following new possibilities:
** Normalize the weight values (currently, they will get "stroke width * 0.1", i.e. would range by default from 0.0 to 0.3...).
** Scale the radius values to your liking (again, currently they are set from stroke width times 0.1)!
** Link all strokes into a single curve, using zero-radius sections (this is mandatory to use the dynamic feature!).
Here is a small demo video: http://youtu.be/VwWEXrnQAFI
Will update user manual later today.
This commit allows you to set the RGB <-> XYZ axis colours used for things like
the mini axis indicator, grid axis indicators, manipulators, transform
constraint indicators, F-Curves (when using XYZ to RGB colouring option), and
perhaps something else I've missed. Previously, these places all used hardcoded
defines (220 * i/j/k), but the readability of these colours was often quite
poor, especially when used with certain themes.
The settings for these colours can be found under the "User Interface" section
of the themes (i.e. same set of colours is used across editors). I could have
made these per editor, but since it's unlikely that these will need to be too
different across editors in practice (+ being easier to version patch), they are
stored under the UI section.
Adding new image texture to Meshes didn't initialize UVs to 0-1 default.
This makes initial display of textures on meshes not work.
This fixes my favorite demo case: Open Blender, drop image from desktop on cube.
PATCH: [#32989] Activate backup files filter in File Browser
Contributed by Georg Kronthaler, many thanks!
(I just moved the icon to a different place reserved for file browser icons)
* enables the filtering of backup files in the file browser
* adds a 'filter backup files'-icon to the filter buttons
* adds new icons for backup files in list and thumbnail view
* enables file preview for the backup files
Patch by Yasuhiro Fujii, thanks!
Original issue was that in vases viewport's lens are different from default
value switching between perspective and orthographic projections will change
viewplane a lot, which is disorienting and annoying.
Issue was caused by buffer shadows were binding buffer after offscreen buffers
was bind which lead to some unpredictable results.
Made it so ED_view3d_draw_offscreen wouldn't bind any buffers and for proper
shadows ED_view3d_draw_offscreen_init should be manually be called before
drawing to an offscreen.
This should also make open gl render with AA enabled a bit faster.
Also fixed missing sequencer cache invalidation when open gl render type is
changing.
Material and Rendered modes are still a TODO for sequencer.
- update face normals when triangulating.
- avoid divide by zero when interpolating customdata on a zero length edge.
- replace zero float comparisons with fabsf() < FLT_EPSILON to avoid numeric error.
also renamed BLI_heap_empty() --> BLI_heap_is_empty() so its obviously readonly function.
Made it so meshes, curves, surfaces and metaballs are scaling to a grid cell size,
which makes them behave consistently now.
There're still issues to be resolved still:
- Lattice is not scaled to grid cell size yet, it uses slightly different add
function which makes scaling a bit tricky and hacky. Would prefer to do a
bit bigger refactor here, so it's a TODO for now.
- Cameras, speakers and other helpers are not scaling. They don't have data
on which scale could be applied and perhaps it should be some kind of draw
scale. Also would consider it's a TODO for now.
- move object_iterators.c --> view3d_iterators. (ED_object.h had to include ED_view3d.h which isn't so nice)
- move projection functions from view3d_view.c --> view3d_project.c (view3d_view was becoming a mishmash of utility functions and operators).
- some some cmake includes as system-includes.
Documentation & Test blend files:
------------------
http://wiki.blender.org/index.php/User:MiikaH/GSoC-2012-Smoke-Simulator-Improvements
Credits:
------------------
Miika Hamalainen (MiikaH): Student / Main programmer
Daniel Genrich (Genscher): Mentor / Programmer of merged patches from Smoke2 branch
Google: For Google Summer of Code 2012
* Get rid of ED_object_add_generic_invoke() and all invoke callbacks using it, it was doing nothing exec() callbacks would not do. In fact, its only action (setting part of common add ops properties, like loc, layers, etc.) was needed too by direct exec call, so it was done twice in case of using invoke()!
* Replace custom invoke code for metaballs by WM_menu_invoke helper (as already used by lamps).
* Add a new OBJECT_OT_empty_add op, to allow direct addition of empties of a given drawtype.
* And some general code cleanup (like trailing spaces, empty lines, ...).
Did quite a bunch of tests/verifications, but obviously could not tackle all possible scenarios... Anyway, if any, bugs should arize quite quickly (but I don’t expect any! :p ).
This is apart of a code cleanup to make ED_view3d_project_short/ED_view3d_project_int/ED_view3d_project_float interchangeable. Currently they work very differently in a way thats quite confusing (and cause of bugs in blender that remain uncorrected) - fixes coming.
There are also cases where ED_view3d_project_short is used, then the values are converted from shorts into int's after because ED_view3d_project_int() behaves differently, will unify behavior of these functions after this commit.
- rather then clip/noclip versions, pass flags (for bound-box clip, window clip).
- rather then store the invalid clip-value, return success (or error value clip_near, clip_bb, clip_win, overflow).
- remove local copies of project functions from drawobject.c: view3d_project_short_clip, view3d_project_short_noclip, view3d_project_short_clip_persmat.
add functions:
- ED_view3d_project_short_global() global space projection
- ED_view3d_project_short_object() object space projection.
- ED_view3d_project_short_ex() take perspective matrix and local space option as args.
- ED_view3d_project_base() - special function to set the Object 'Base' screen coords (sx, sy), since this is a common enough operation.
A correct fix for this bug likely involves changuing how operators are called in the event loop but such changes better not be made just before the release.
add an option to apply the transform of an object when re-parenting.
the reason for this is you dont always want to apply transform because...
- your object may be driven/animated and the transformation will be lost.
- with negative scales and rotation it can end up giving odd/not-useful results.
fix pose-group-sort and pose-group-moving being disabled for pinned poses.
also fix for own missing NULL check for pose mask clear which would crash when run without an active object
The way it works now, navigation keys (up/down, number, a-z) - won't apply to a submenu if there are no active buttons in that menu, instead the parent menu gets the event.
