blender-staging.git
6 years agoAdded a brie instruction how to build simple memtest
Sergey Sharybin [Thu, 5 Sep 2013 16:32:44 +0000 (16:32 +0000)]
Added a brie instruction how to build simple memtest

6 years agoGhost test application could be compiled again
Sergey Sharybin [Thu, 5 Sep 2013 16:22:07 +0000 (16:22 +0000)]
Ghost test application could be compiled again

Was an IRC request to have Ghost application up
and running for investigation purposes.

6 years agoFix metaball f-curves not showing up in graph editor.
Brecht Van Lommel [Thu, 5 Sep 2013 16:15:01 +0000 (16:15 +0000)]
Fix metaball f-curves not showing up in graph editor.

6 years agoFix #36649 and #36650: animation of freestyle line thickness and linestyle did
Brecht Van Lommel [Thu, 5 Sep 2013 15:47:52 +0000 (15:47 +0000)]
Fix #36649 and #36650: animation of freestyle line thickness and linestyle did
not render correct in animations, due to missing updates.

6 years agoExpose theme color of shadow UVs (These include UVs that are drawn in
Antony Riakiotakis [Thu, 5 Sep 2013 15:09:10 +0000 (15:09 +0000)]
Expose theme color of shadow UVs (These include UVs that are drawn in
texture paint modes and as modified mesh uvs overlay) and UVs of other
objects.

6 years agoCtr-Alt-F radial control operator for texture painting, controls the
Antony Riakiotakis [Thu, 5 Sep 2013 14:02:59 +0000 (14:02 +0000)]
Ctr-Alt-F radial control operator for texture painting, controls the
rotation of the brush mask texture.

Unfortunately secondary path does not work here because we do not have a
permanent switch to choose between primary-secondary brush texture. Use
operator property instead.

6 years agoCode cleanup: use boolean instead of int for space_clip
Sergey Sharybin [Thu, 5 Sep 2013 13:37:53 +0000 (13:37 +0000)]
Code cleanup: use boolean instead of int for space_clip

6 years agoCode cleanup: get rid of crappy height cast to double
Sergey Sharybin [Thu, 5 Sep 2013 13:37:46 +0000 (13:37 +0000)]
Code cleanup: get rid of crappy height cast to double

6 years agoCode cleanup: use boolean instead of int for tracking API
Sergey Sharybin [Thu, 5 Sep 2013 13:37:36 +0000 (13:37 +0000)]
Code cleanup: use boolean instead of int for tracking API

6 years agoCompositor cleanup: Merge conversion operations into a single file (see also r59820).
Lukas Toenne [Thu, 5 Sep 2013 13:32:14 +0000 (13:32 +0000)]
Compositor cleanup: Merge conversion operations into a single file (see also r59820).

Most convert operations can share a common base class with a single socket reader (except channel separate/combine nodes).

6 years agoTidy up paint options, patch by Sebastian Koenig, with minor changes
Antony Riakiotakis [Thu, 5 Sep 2013 13:15:29 +0000 (13:15 +0000)]
Tidy up paint options, patch by Sebastian Koenig, with minor changes
(ommit texture paint changes, they made the interface more difficult to
discern).

Also, move stroke input samples to stroke panel

6 years agoBugfix #35920
Ton Roosendaal [Thu, 5 Sep 2013 13:03:03 +0000 (13:03 +0000)]
Bugfix #35920

Adding a new node in Node Editor failed for "High DPI" (Only Mac retina now).

- Py script for adding nodes was doing dpi magic, which it shouldn't. It has
  been replaced with a (temporary) API call to set the correct cursor location.
  (Thanks to Lukas T for helping here)

- The SpaceNode->cursor[2] property now is *only* storing the coordinate
  in "adding new node space". Use of this has been removed from the code where
  possible, with as only exception the code to draw noodles while adding them.

Special coder note: Nodes should respect the DPI value, and draw larger with
larger buttons if you increase this size. The hack here is that this can only
work nice if also the node positions are scaled accordingly.

A better fix could be to check on scaling the node view itself for it. That
then would also remove this Python API call that was added in this commit.
However, that again might fight with how buttons layout code works now...
needs some careful checking.

6 years agoRemoved unused variables used only in BLI_assert, this triggers compiler warning...
Lukas Toenne [Thu, 5 Sep 2013 12:10:49 +0000 (12:10 +0000)]
Removed unused variables used only in BLI_assert, this triggers compiler warning in release builds.

6 years agoadd function for rotating linked list so at item is last.
Campbell Barton [Thu, 5 Sep 2013 11:40:38 +0000 (11:40 +0000)]
add function for rotating linked list so at item is last.

6 years agogrid-fill can now fill a single, closed edge loop.
Campbell Barton [Thu, 5 Sep 2013 11:07:49 +0000 (11:07 +0000)]
grid-fill can now fill a single, closed edge loop.
the active vertex is used when possible, else the corner vertex is detected.

this should replace the grid-fill-manager addon by Alexander Nedovizin.

6 years agoUpdate Ceres to 1.7.0 release
Sergey Sharybin [Thu, 5 Sep 2013 10:48:44 +0000 (10:48 +0000)]
Update Ceres to 1.7.0 release

For Blender this release is interesting because of:

- Covariance estimation (not used in Blender yet, but now we
  might use it for keyframe selection instead of havingown
  implementation).

- Significant performance improvements to loss function and
  dense linear solvers and automatic differentiation.

  Unfortunately, didn't notice speedup of tracking itself,
  but camera reconstruction now happens around 2 times faster
  on my laptop,

- Better inner iteration step acceptance and stopping.

6 years agoFix #36113, Translate's wrapping has 1 pixel gap in X and Y after scale node.
Lukas Toenne [Thu, 5 Sep 2013 10:45:21 +0000 (10:45 +0000)]
Fix #36113, Translate's wrapping has 1 pixel gap in X and Y after scale node.
The issue with wrapping is that it requires correct interpolation of the border pixels. Since interpolation is done at the far left end of the node tree in buffer/image/etc read operations, the wrapping
setting can not be used directly in those operations (otherwise in-line translate operations would cause conflicts). To make wrapping work correctly we need to add a buffer in front of the translate
operation, which can then be interpolated correctly based on wrapping. The WrapOperation becomes a variant of ReadBufferOperation, which uses its wrapping setting to determine the correct "extend" mode
for interpolation of the buffer.

6 years agoExtend mode option for MemoryBuffer reading in compositor. This will allow proper...
Lukas Toenne [Thu, 5 Sep 2013 10:45:19 +0000 (10:45 +0000)]
Extend mode option for MemoryBuffer reading in compositor. This will allow proper interpolation of pixel values when using wrapping in the Translate node. Implemented in inline functions, so won't cause
overhead if constant values are passed (as happens with most calls using the default argument).

6 years agoadd positive_mod() utility function.
Campbell Barton [Thu, 5 Sep 2013 10:12:00 +0000 (10:12 +0000)]
add positive_mod() utility function.

6 years agoMerge some operations into a single file
Sergey Sharybin [Thu, 5 Sep 2013 09:39:38 +0000 (09:39 +0000)]
Merge some operations into a single file

This merges all mix operations into a COM_MixBaseOperation
(naming could be better, but this way it corresponds to what's
going on with math operations.

Same was done with RenderLayers operations.

Overall this gives 20% of bf_compositor library compilation
time decrease. And it was rather annoying to have tens of
files with just a single-line constructors anyway.

TODO:
- All Convert operations could also be merged into a single file,
  but that would require adding some ConvertBaseOperation to
  reduce code duplication (ideally). Leaving it unchanged for now.

- Some operations' headers seems to be wrongly including MixOperation
  header, they need to include NodeOperation instead it seems.

  This is rather harmless, but would be nice to doublecheck on this
  eventually.

6 years agofix for error in bmesh api, selection setting needs to go via BM_elem_select_set...
Campbell Barton [Thu, 5 Sep 2013 08:38:06 +0000 (08:38 +0000)]
fix for error in bmesh api, selection setting needs to go via BM_elem_select_set else the total selected elements count isn't mantained.

6 years agoApply patch [#36601].
Nathan Letwory [Thu, 5 Sep 2013 06:58:27 +0000 (06:58 +0000)]
Apply patch [#36601].
Submitted by Saurabh Wankhade (sauraedron).

This patch adds camera shift and dof export and import in a Blender profile.

6 years agosimplify idprop reallocation with MEM_recallocN
Campbell Barton [Thu, 5 Sep 2013 04:26:08 +0000 (04:26 +0000)]
simplify idprop reallocation with MEM_recallocN

6 years agoBGE: Removing an unused rendertools argument from BL_ConvertActuators().
Mitchell Stokes [Thu, 5 Sep 2013 04:19:05 +0000 (04:19 +0000)]
BGE: Removing an unused rendertools argument from BL_ConvertActuators().

6 years agoFix broken GLSL shader after recent changes.
Brecht Van Lommel [Thu, 5 Sep 2013 00:37:10 +0000 (00:37 +0000)]
Fix broken GLSL shader after recent changes.

6 years agoFind missing files now keeps relative paths relative instead of turning them
Brecht Van Lommel [Thu, 5 Sep 2013 00:37:06 +0000 (00:37 +0000)]
Find missing files now keeps relative paths relative instead of turning them
into absolute paths.

6 years agoremove freebsd7 config. its fairly old and not maintained.
Campbell Barton [Wed, 4 Sep 2013 23:44:41 +0000 (23:44 +0000)]
remove freebsd7 config. its fairly old and not maintained.

6 years agofix [#36459] Official blenderplayer can not go fullscreen on Linux
Campbell Barton [Wed, 4 Sep 2013 23:43:25 +0000 (23:43 +0000)]
fix [#36459] Official blenderplayer can not go fullscreen on Linux
use modified patch from Alex Fraser (z0r)

6 years agoBGE: Removing the touch sensor and converting all touch sensors to collision sensors.
Mitchell Stokes [Wed, 4 Sep 2013 22:14:14 +0000 (22:14 +0000)]
BGE: Removing the touch sensor and converting all touch sensors to collision sensors.

6 years agoGhost_SDL: Fixing the DEF_CURSOR macro so it works in release builds.
Mitchell Stokes [Wed, 4 Sep 2013 21:57:14 +0000 (21:57 +0000)]
Ghost_SDL: Fixing the DEF_CURSOR macro so it works in release builds.

6 years agoremove error return value from BLI_bvhtree_insert, no callers were using.
Campbell Barton [Wed, 4 Sep 2013 20:33:50 +0000 (20:33 +0000)]
remove error return value from BLI_bvhtree_insert, no callers were using.
in the case of an error - assert, rather then fail silently since it wont be working as expected anyway.

6 years agouse strict flags for kdopbvh, also assert when invalid treetype is passed, rather...
Campbell Barton [Wed, 4 Sep 2013 20:03:33 +0000 (20:03 +0000)]
use strict flags for kdopbvh, also assert when invalid treetype is passed, rather then failing silently.

6 years agoCode cleanup / Cycles:
Thomas Dinges [Wed, 4 Sep 2013 16:24:58 +0000 (16:24 +0000)]
Code cleanup / Cycles:
* Removed unused member of the device_memory template.

6 years agoCode cleanup / Cycles:
Thomas Dinges [Wed, 4 Sep 2013 16:11:21 +0000 (16:11 +0000)]
Code cleanup / Cycles:
* Avoid special code, when Subsurface is enabled.
Ideally we should only use the function, and get rid of the extra duplicate, but this is slower on CUDA.

6 years agoCompile fix / Cycles:
Thomas Dinges [Wed, 4 Sep 2013 13:37:35 +0000 (13:37 +0000)]
Compile fix / Cycles:
* 'T' is now also used for the Subsurface closure.

6 years agoAdd new cursor types to data handling to account for Intuos 4 device (pen and eraser).
Nathan Letwory [Wed, 4 Sep 2013 12:10:02 +0000 (12:10 +0000)]
Add new cursor types to data handling to account for Intuos 4 device (pen and eraser).

Thanks to Ayden Polat for helping to find out the cursor type codes.

6 years agofix for building on msvc
Campbell Barton [Wed, 4 Sep 2013 06:50:15 +0000 (06:50 +0000)]
fix for building on msvc

6 years agobetter fix for r59794, register a gesture when the mouse goes outside the actionzone.
Campbell Barton [Wed, 4 Sep 2013 05:25:29 +0000 (05:25 +0000)]
better fix for r59794, register a gesture when the mouse goes outside the actionzone.

6 years agofix for window join action being interpreted as a split.
Campbell Barton [Wed, 4 Sep 2013 05:06:38 +0000 (05:06 +0000)]
fix for window join action being interpreted as a split.

The limit for dragging a gesture on an area corner was smaller then the area hot-spot,
so you could click on the right-most side of the bottom-left corner, drag left - pass the gesture threshold and still be in the same area. so a motion intended as a join would register as a split.
Happened more with high DPI values. fix by ensuring the drag limit is always higher then the hotspot.

6 years agofix for glitch splitting horizontally with a high dpi, could make areas smaller then...
Campbell Barton [Wed, 4 Sep 2013 03:52:25 +0000 (03:52 +0000)]
fix for glitch splitting horizontally with a high dpi, could make areas smaller then the header which pushed the original view out of the screen.

6 years agostyle cleanup
Campbell Barton [Wed, 4 Sep 2013 03:30:41 +0000 (03:30 +0000)]
style cleanup

6 years agotweaks to fly mode
Campbell Barton [Wed, 4 Sep 2013 01:49:20 +0000 (01:49 +0000)]
tweaks to fly mode
- rotating the view is faster.
- arrow keys work (was only wasd-rf before)
- when stationary wheel or +/- will set z direction (so mouse wheel always sets the fly axis back to z)

6 years agodebug option for modifier stack calculation USE_MODIFIER_VALIDATE,
Campbell Barton [Wed, 4 Sep 2013 01:29:34 +0000 (01:29 +0000)]
debug option for modifier stack calculation USE_MODIFIER_VALIDATE,
checks that every modifier gives a valid mesh (very slow) - so off by default even in debug mode.

6 years agoAdded better support for line styles in the Outliner.
Tamito Kajiyama [Wed, 4 Sep 2013 01:15:23 +0000 (01:15 +0000)]
Added better support for line styles in the Outliner.

6 years agoFix cycles CUDA/OpenCL build error after recent change.
Brecht Van Lommel [Wed, 4 Sep 2013 01:04:14 +0000 (01:04 +0000)]
Fix cycles CUDA/OpenCL build error after recent change.

6 years agoFix for redundant multiple animation data items of the same line style in DopeSheet.
Tamito Kajiyama [Wed, 4 Sep 2013 00:21:27 +0000 (00:21 +0000)]
Fix for redundant multiple animation data items of the same line style in DopeSheet.
This could happen when a line style is shared by multiple linesets within a scene.

6 years agoFix for missing filter button to visualize/hide animation data related to line styles.
Tamito Kajiyama [Tue, 3 Sep 2013 22:50:58 +0000 (22:50 +0000)]
Fix for missing filter button to visualize/hide animation data related to line styles.

6 years agoFix #36649: Drivers for Freestyle parameters in Render Layer panel don't appear in...
Tamito Kajiyama [Tue, 3 Sep 2013 22:49:30 +0000 (22:49 +0000)]
Fix #36649: Drivers for Freestyle parameters in Render Layer panel don't appear in Graph Editor.

The channels define code in ANIM_init_channel_typeinfo_data() was not synchronized with the
definition of channel types (eAnim_ChannelType) in ED_anim_api.h.

6 years agoFix a few issues reported by coverity scan.
Brecht Van Lommel [Tue, 3 Sep 2013 22:39:21 +0000 (22:39 +0000)]
Fix a few issues reported by coverity scan.

6 years agoCycles: add a sharpness input to the Cubic SSS falloff. When set to 1 this will
Brecht Van Lommel [Tue, 3 Sep 2013 22:39:17 +0000 (22:39 +0000)]
Cycles: add a sharpness input to the Cubic SSS falloff. When set to 1 this will
give a result more similar to the Compatible falloff option. The scale is x2
though to keep the perceived scatter radius roughly the same while changing the
sharpness. Difference with compatible will be mainly on non-flat geometry.

6 years agoreplace sqrt(dot()) with length functions.
Campbell Barton [Tue, 3 Sep 2013 22:32:03 +0000 (22:32 +0000)]
replace sqrt(dot()) with length functions.

6 years agosimplify length checks in DM_calc_auto_bump_scale
Campbell Barton [Tue, 3 Sep 2013 22:22:45 +0000 (22:22 +0000)]
simplify length checks in DM_calc_auto_bump_scale

6 years agofix for missing NULL check before calling BKE_sequencer_imbuf_to_sequencer_space().
Campbell Barton [Tue, 3 Sep 2013 21:54:17 +0000 (21:54 +0000)]
fix for missing NULL check before calling BKE_sequencer_imbuf_to_sequencer_space().

6 years agomove timeit macros into their own include, since they are only used for testing and...
Campbell Barton [Tue, 3 Sep 2013 21:22:43 +0000 (21:22 +0000)]
move timeit macros into their own include, since they are only used for testing and unrelated to PIL_time.h typical use.

6 years agoFix #36449: switching between render slots during render would leave some tiles
Brecht Van Lommel [Tue, 3 Sep 2013 20:59:24 +0000 (20:59 +0000)]
Fix #36449: switching between render slots during render would leave some tiles
black until the end of the render.

6 years agoRelated to #36532: show message in modifier when cloth point cache read fails,
Brecht Van Lommel [Tue, 3 Sep 2013 19:58:09 +0000 (19:58 +0000)]
Related to #36532: show message in modifier when cloth point cache read fails,
for example due to a subsurf modifier preceding a cloth modifier.

6 years agoFix #36194, Halo material on Layer 2 produces flares on layer 1. Halo flares were...
Lukas Toenne [Tue, 3 Sep 2013 18:59:52 +0000 (18:59 +0000)]
Fix #36194, Halo material on Layer 2 produces flares on layer 1. Halo flares were simply rendered on the first available halo-enabled layer. Now use the flare layers correctly.

6 years agoFix compile error with strict compile flags.
Brecht Van Lommel [Tue, 3 Sep 2013 18:01:31 +0000 (18:01 +0000)]
Fix compile error with strict compile flags.

6 years agoFix #36189: link group, delete instanced object, link group again would not
Brecht Van Lommel [Tue, 3 Sep 2013 17:24:07 +0000 (17:24 +0000)]
Fix #36189: link group, delete instanced object, link group again would not
create an instance. Now it creates the instance even if the group was already
linked into the .blend file.

6 years agoRelated to #36382: for linked object point caches, show a message that disk
Brecht Van Lommel [Tue, 3 Sep 2013 16:28:53 +0000 (16:28 +0000)]
Related to #36382: for linked object point caches, show a message that disk
cache must be used for baking.

6 years agoFix #36359: sequencer frame indicator not updating during animation render.
Brecht Van Lommel [Tue, 3 Sep 2013 16:02:55 +0000 (16:02 +0000)]
Fix #36359: sequencer frame indicator not updating during animation render.

6 years agoFixed crash on Visual Studio
Alexander Kuznetsov [Tue, 3 Sep 2013 15:30:07 +0000 (15:30 +0000)]
Fixed crash on Visual Studio

6 years agoFix a possible bug in knife detected by coverity.
Howard Trickey [Tue, 3 Sep 2013 11:37:47 +0000 (11:37 +0000)]
Fix a possible bug in knife detected by coverity.

A selection copy may have been passed a null pointer.
Also quieted a couple other complaints by coverity.

6 years agoFix #36587: Tracking markers fail to track near the left and right edge of a movie...
Sergey Sharybin [Tue, 3 Sep 2013 10:39:55 +0000 (10:39 +0000)]
Fix #36587: Tracking markers fail to track near the left and right edge of a movie clip.

Track margin checks needed some tweaks to deal better with the fact
that normalized values for the same pixel values might be different
across X and Y axis.

Also, non-centered patters are expected to be handling better now.

6 years agoFix #36643: Clip Editor: undesired Markers created "automatically" while I track...
Sergey Sharybin [Tue, 3 Sep 2013 09:18:08 +0000 (09:18 +0000)]
Fix #36643: Clip Editor: undesired Markers created "automatically" while I track my marker

Was a typo from recent commint from my own.

6 years agoColor managed color didn't work properly for float sequencer frames.
Sergey Sharybin [Tue, 3 Sep 2013 09:18:04 +0000 (09:18 +0000)]
Color managed color didn't work properly for float sequencer frames.

Like, it seems it never worked actually.

6 years agoBugfix: When deleting all keyframes from F-Curves, don't delete the F-Curve if
Joshua Leung [Tue, 3 Sep 2013 06:46:17 +0000 (06:46 +0000)]
Bugfix: When deleting all keyframes from F-Curves, don't delete the F-Curve if
it has a driver

6 years agodisplay angle in axis/angle buttons as degrees as with euler rotations.
Campbell Barton [Tue, 3 Sep 2013 06:27:03 +0000 (06:27 +0000)]
display angle in axis/angle buttons as degrees as with euler rotations.

6 years agoreorder BLI_strict_flags.h include so its not conflicting with stdio.h on apple.
Campbell Barton [Tue, 3 Sep 2013 04:39:12 +0000 (04:39 +0000)]
reorder BLI_strict_flags.h include so its not conflicting with stdio.h on apple.

6 years agoremove WINDOWS checks in cmake, this variable isn't defined on windows and checks...
Campbell Barton [Tue, 3 Sep 2013 03:04:44 +0000 (03:04 +0000)]
remove WINDOWS checks in cmake, this variable isn't defined on windows and checks weren't needed.
also remove redundant dot_v3v3 in shrinkwrap constraint.

6 years agoTweaks to New Driver creation behaviour
Joshua Leung [Tue, 3 Sep 2013 00:28:23 +0000 (00:28 +0000)]
Tweaks to New Driver creation behaviour

* 'Show Debug' now enabled for all newly created drivers. For most users, it is
useful to be able to see this to help figure out what's going on

* Removed failed experiment of creating new drivers with Generator FModifiers. I
had hoped that this would make it easier to create drivers that doubled or
halved the input values, but that has proved to not be the case, and instead
made harder for most users to set things up (as they'd have to remove these
first).

Now, when adding drivers from the UI, these get created with two keyframes (at
(0,0) and (1,1) for a 1-1 mapping), which can be easily tweaked normally.

However, for backwards compatability of scripts (notably rigify, and perhaps
some others out there), when creating drivers from scripts, they will still get
created with Generator FModifiers for now. We can review this situation again
for 2.7, but for now it seems ok.

6 years agoGraph Editor: Preserve active curve when using AKEY to toggle selection status
Joshua Leung [Mon, 2 Sep 2013 23:32:21 +0000 (23:32 +0000)]
Graph Editor: Preserve active curve when using AKEY to toggle selection status
of keyframe verts

Previously, every time you toggled the selection of all keyframes (using AKEY),
the active curve would get deselected and deactivated. However, this was a pain
when trying to tweak the shape of a particular curve, as doing this would cause
that curve to either fade into the background or into the jumble of other
curves.

6 years agoApply Scale for Empties
Joshua Leung [Mon, 2 Sep 2013 22:57:22 +0000 (22:57 +0000)]
Apply Scale for Empties

It is now possible to use "Apply Scale" for Empties. While Empties don't exactly
have any Object data attached to them which can be used for supporting "true"
apply scale (i.e. with non-uniform scaling), they do have a drawsize value which
controls how large the empties are drawn (before scaling). This works by taking
the scale factor on the most-scaled axis, and combines this with the existing
empty drawsize to maintain the correct dimensions on that axis at least.

Core Assumptions:
1) Most scaled empties have uniform scaling anyways (i.e. most empties used for
bone shapes)
2) On balance, preserving non-uniform scaling of empties after apply scale is
not as important as not being able to do it at all

6 years agoadd checks for bad args to RNA_def_property_ui_range & RNA_def_property_range and...
Campbell Barton [Mon, 2 Sep 2013 22:28:18 +0000 (22:28 +0000)]
add checks for bad args to RNA_def_property_ui_range & RNA_def_property_range and fix one instance where (min > max).
also remove redundant float/double conversion in ui_get_but_step_unit()

6 years agoallow negative offset for floor constraint.
Campbell Barton [Mon, 2 Sep 2013 22:24:44 +0000 (22:24 +0000)]
allow negative offset for floor constraint.
(since this isnt a projection, it just clamps on an axis this can be supported).

6 years agoAdd poll message when poll fails because of window missing from the context.
Campbell Barton [Mon, 2 Sep 2013 19:28:44 +0000 (19:28 +0000)]
Add poll message when poll fails because of window missing from the context.
This means calling invoke operators from python in states that dont support will give a meaningful error message.

also reduce context lookups when polling.

6 years agoAck... Followup to r59743: in fact, bitflags groups are not always wanted/needed...
Bastien Montagne [Mon, 2 Sep 2013 18:33:06 +0000 (18:33 +0000)]
Ack... Followup to r59743: in fact, bitflags groups are not always wanted/needed, thanks to Campell for notifying me about this!

So now, their generation is controlled by a flag, else previous "simple values" group ids are generated (one per poly region, no need here to reduce the number of used IDs!).

Will update obj exporter too.

6 years agowarning cleanup: correct some odd returns and quiet strict flag warnings on some...
Campbell Barton [Mon, 2 Sep 2013 17:59:04 +0000 (17:59 +0000)]
warning cleanup: correct some odd returns and quiet strict flag warnings on some systems.

6 years agoMake strict compiler happy for release builds.
Sergey Sharybin [Mon, 2 Sep 2013 17:24:56 +0000 (17:24 +0000)]
Make strict compiler happy for release builds.

6 years agoFix [#36636] Incorrect assignment of Smoothing Groups on export
Bastien Montagne [Mon, 2 Sep 2013 17:14:41 +0000 (17:14 +0000)]
Fix [#36636] Incorrect assignment of Smoothing Groups on export

In fact, smooth groups are supposed to be bitflags, not simply integer values (would be far too much simple!). This adds quite a bit of work, as with only 32 values, we can't just assign one to each group. Somewhat related to the "Four colors theorem"! ;)

Here we simply use the first available bit for current smooth group (i.e. first bit not used by any of the already defined contiguous groups).

6 years agoFix #36628, Muting bump node alters material visibility.
Lukas Toenne [Mon, 2 Sep 2013 17:08:03 +0000 (17:08 +0000)]
Fix #36628, Muting bump node alters material visibility.
The internal connections used for muted nodes are a "best guess" only. The algorithm would connect linked input/output sockets first, regardless of sensible connections. Replaced this by a more strict
connection method now which only uses matching types, so that Normal outputs won't get input from values, etc.

This is still far from perfect though. Eventually it may be necessary to implement dedicated methods for certain types of nodes to respect different semantics - using the plain socket type is just not
enough information. Also the value used for connections from muted nodes with no matching input will currently be the hidden(!) socket input values, this can lead to unexpected results. Needs further
thought ...

6 years agoRelated to #36632: cycles textured draw mode now shows the image from an image
Brecht Van Lommel [Mon, 2 Sep 2013 16:05:51 +0000 (16:05 +0000)]
Related to #36632: cycles textured draw mode now shows the image from an image
texture node even if it's not active, in case there is no active texture node.

6 years agoCycles:
Thomas Dinges [Mon, 2 Sep 2013 15:43:28 +0000 (15:43 +0000)]
Cycles:
* Document the members of the Ray struct.

6 years agoFix #36639, textures not reloaded after fixing paths.
Lukas Toenne [Mon, 2 Sep 2013 14:23:42 +0000 (14:23 +0000)]
Fix #36639, textures not reloaded after fixing paths.
The BKE_bpath_traverse_id method updates path variables of image data blocks, but did not cause the image to be reloaded (or rather, tagged so it gets reloaded on next use). This requires also updating
all the ImageUser "ok" tags, which happens in some odd places in other cases where reload happens (e.g. the reload operator and the image path button handler function in image template).

6 years agoFix #36640: node editor header with world nodes not drawing correct after recent...
Brecht Van Lommel [Mon, 2 Sep 2013 14:14:02 +0000 (14:14 +0000)]
Fix #36640: node editor header with world nodes not drawing correct after recent bugfix.

6 years agoFix #36124: VSE - Input Color option does not work for video files
Sergey Sharybin [Mon, 2 Sep 2013 13:54:12 +0000 (13:54 +0000)]
Fix #36124: VSE - Input Color option does not work for video files

Byte images and movies will now fully follow input color space.
Before this non-sRGB input colorspace for byte images and movies
behave really doggy (results in preview and final render were
totally different).

To prevent data loss, if byte image is set not stored in sequencer's
space it'll be internally converted to float buffer.

In theory some setups might be rendering a bit different now, but
new behavior is totally expected and someone used non-sRGB input
space for byte images/movies had Convert Float enabled anyway.

6 years agoFix for the BKE_image_walk_all_users function: This included a check for movie/sequen...
Lukas Toenne [Mon, 2 Sep 2013 13:45:57 +0000 (13:45 +0000)]
Fix for the BKE_image_walk_all_users function: This included a check for movie/sequence image types meant for the original use to set IMA_NEED_FRAME_RECALC flag. But the function name suggests it should be generic, so better move this filter into the callback.

6 years agoFix crash when adjusting plane track after re-tracking point tracks
Sergey Sharybin [Mon, 2 Sep 2013 12:02:46 +0000 (12:02 +0000)]
Fix crash when adjusting plane track after re-tracking point tracks

Issue was caused by tacks map merge re-allocating the tracks and this
didn't update plane tracks.

Ideally tracks_map_merge shall not re-allocate tracks, but for now
just update plane tracks. Shouldn't be too much slow anyway and could
always be tweaked without affecting any artists.

6 years agoSome function didn't match their declaration
Sergey Sharybin [Mon, 2 Sep 2013 11:19:21 +0000 (11:19 +0000)]
Some function didn't match their declaration

6 years agoRemoved some unused old function declarations for loop node register, awkward that...
Lukas Toenne [Mon, 2 Sep 2013 10:34:27 +0000 (10:34 +0000)]
Removed some unused old function declarations for loop node register, awkward that this is still there ...

6 years agofix [#36412] Pivot of active element switches to last selected face after duplicating...
Campbell Barton [Mon, 2 Sep 2013 04:39:48 +0000 (04:39 +0000)]
fix [#36412] Pivot of active element switches to last selected face after duplicating vertices.

6 years agobmesh: internal api change, remove BMOElemMapping, its not needed since the values...
Campbell Barton [Mon, 2 Sep 2013 03:13:51 +0000 (03:13 +0000)]
bmesh: internal api change, remove BMOElemMapping, its not needed since the values can be stored in the ghash directly.
saves 24 bytes per element (per vertex/edge/face on a 64bit system), for bmesh operators.

6 years agoadd hash iterator functions to access the pointer to the value.
Campbell Barton [Mon, 2 Sep 2013 02:58:53 +0000 (02:58 +0000)]
add hash iterator functions to access the pointer to the value.

6 years agotext editor cursor motion (left/right arrows) with selected text typically jumps...
Campbell Barton [Mon, 2 Sep 2013 00:47:27 +0000 (00:47 +0000)]
text editor cursor motion (left/right arrows) with selected text typically jumps to either side of the selection previously the cursor would move and loose the selection too.
text button fields already did this.

6 years agoadd attributes to ghash and edgehash functions.
Campbell Barton [Sun, 1 Sep 2013 22:47:44 +0000 (22:47 +0000)]
add attributes to ghash and edgehash functions.

6 years agofix odd (intentional) behavior with vertex parent,
Campbell Barton [Sun, 1 Sep 2013 22:38:41 +0000 (22:38 +0000)]
fix odd (intentional) behavior with vertex parent,
curve children of a triangle vertex parent would only display their relationship line to the first vertex. (confusing)

also added OB_TYPE_SUPPORT_PARVERT macro.

6 years agosupport for vertex parenting in object mode for object types which support it (mesh...
Campbell Barton [Sun, 1 Sep 2013 22:01:21 +0000 (22:01 +0000)]
support for vertex parenting in object mode for object types which support it (mesh, lattice, curve, surface)
previously this had to be done one by one.

both single and triagle vertex parents can be made, selected based on distance to the verts.

Developer notes:
- looks like this was old TODO, enums existed but weren't used.
- only meshes currently support using.
- added BKE_object_as_kdtree(), may come in handy for similar cases.

6 years agoCleanup!
Bastien Montagne [Sun, 1 Sep 2013 21:49:35 +0000 (21:49 +0000)]
Cleanup!

Also fixed a nasty (but inofensive for now) stuff, we had both MOD_UVPROJECT_MAX and MOD_UVPROJECT_MAXPROJECTORS (the former used in RNA code, the later in modifier code)! Kept MOD_UVPROJECT_MAXPROJECTORS.

6 years agouse strict flags for kdtree, and replace ints with unsigned ints where possible.
Campbell Barton [Sun, 1 Sep 2013 20:17:56 +0000 (20:17 +0000)]
use strict flags for kdtree, and replace ints with unsigned ints where possible.
also replace callocs with mallocs since zeroing memory can be avoided.

6 years agoFix [#36479] Particle instance Modifier random position snaps objects to 0 position
Bastien Montagne [Sun, 1 Sep 2013 15:31:09 +0000 (15:31 +0000)]
Fix [#36479] Particle instance Modifier random position snaps objects to 0 position

Most work (tracking bug) done by Anthony Edlin, many thanks!

Great example of why having duplicate code is never a good idea... ;)