* normal offset not taken properly into account
* wrong usage of BVHTree (epsilon != radius) caused massive slowdowns in calculations, for example just opening test file took about 30 s on my machine, after fix about 0.5 s :)
crash when using too much stack memory, each recursive call for the sorting function would allocate 200 bytes in TransData per call (64bit os), now pass this as an argument.
this is because of fairly confusing logic, when an SRGB image has a float buffer added its always LINEAR, so we have to account for this elsewhere.
So the conversion from byte to float works like this:
* NONE -> NON-LINEAR
* SRGB/LINEAR -> LINEAR
...but none change the profile variable.
Submitted by Shane Ambler.
The original patches made an enum for action zone edges, changed positioning for minimised icons and repositioned minimised icon for operator properties panel.
I kept the enum idea, but further improved on the naming. Some switches used in place of if/else blocks and added some comments. See patch tracker for more comments.
* Fix var declaration in bpy_interface.c
* Remove forward declarations from py_capi_utils.h: they are unnecessary and break compiles (there were probably many warnings about this during compile with GCC).
Reported and patched by Torsten Rupp
When a time marker is added in the timeline window via keyboard (keycode M) or existing markers are selected/moved via
the mouse the marker resp. the selection state is not draw immediately in the timeline window region.
The patch adds notifiers where necessary to ensure updates.
This troubles were caused by "break" of ND_OBJECT case in outliner area listener,
so not all cases were handled.
Handle more data and actions in outline listener, but not refresh when it's
actually unneeded (there where problems with it without that "break" -- extra
refreshing could be made).
Patch by Shane Ambler
Reported by Anthony Edlin
From the patch details:
"
In response to bug #23078 the operator panel disappears when dragged above the top of the 3dview without showing the
azone icon to restore it.
This patch properly hides the operator panel if dragged near the top of the 3dview so that the azone icon is in place
"
Thanks!
Patch by Alexander Kuznetsov
Reported by Chidozie Oku
From patch description:
"
Handler is now released on every exit from File Selector. For example pressing ctrl-up and then changing editor type
to another also releases the handler.
When an area is changed from SPACE_FILE, ED_fileselect_exit is called for clean up. It takes function of freeing folder
list and files (before it was done in cancel or exec functions) because they must be released on every exit anyway.
op!=null means cancel or exec was not executed so a handler was not released. ED_fileselect_exit then releases the handler
without changing screens.
"
Thanks!
* Velocity for particles that were born at exactly integer frames was calculated wrong when they were born.
Note: If you had a raytrace acceleration related bug, please clear the pointcache for all particles, toggle a particle setting to reset pointcache and rebake to create a valid simulation.
from Dan Eicher (dna)
Basically just wraps distfactor_to_bone() and passes the correct head/tail depending on which bone type it's called
from.
note:
renamed envelope() --> evaluate_envelope()
This runs after changing a property and allows correcting incompatible options.
Returning True will redraw the UI.
Currently this is used for setting the write extension when saving files, so changing the image format also corrects the extension.
The same is accessible from python where its used when saving SVG/EPS/PNG files.
This fixes: [#23828] obj export problems, [#23760] Exporting OBJ and filetype ending
also fixed document submission operator.
Now the filename in the file selector is the one used for writing this means we remove the "Save Over" popup which could be overlooked too easily.
Instead display the filename field with red tint, and a note in the tooltip.
Added ND_OB_SELECT notifier to separate operator. Selected objects
aren't actually changing, but there is no existing ND which could be used
for outliner update.
- BKE_add_image_extension now sets the extension rather then appending. (no more image.jpg.tga)
- py/rna functions which have no return value now raise an error if a non-None value is returned.
- added back the red-alert flag so buttons can have a red highlight if somethings wrong.
color, made sample a modal operator now to solve this. It's an indirect
solution, but couldn't think of anything better, and it's useful to have
anyway.
Crash was caused by missed offscreen OpenGL buffer. Added checking around this stuff.
Also fixed crash of simple "Image from view operator".
Note: This commit fixes only crashing, you'll be still unable to use this tools.
- rna buttons with units set now use the units base value for snapping.
- bone head/tail radius could be set negative.
matt: removed a check in ui_is_but_unit() which made angle buttons return false, what was this for?
from Lorenzo Tozzi (oni_niubbo) with minor edits.
--- from the tracker
The present situation is this: due to bug#22274, during editing, UTF chars are stripped from buttons with a unit associated
(length, angles, etc.).
Example: if the button displays '90°' and you click on it with LMB, the editing string will become '90'.
The problem arises if you use microns: '34µm' becomes '34' that blender interprets as 34 meters. So clicking on a button
and hitting enter won't confirm the previous value, but will change it (very badly also).
Of course nobody is using microns in blender, but the problem will arise when we will implement areas and option 'Separate
Units' will be enabled. The value '2m² 3cm²' will become '2m' during editing.
This patch solves the problem rewriting the string in a smarter way than just stripping the UTF chars: the unit is translated
from unit->name_short ('µm') to unit->name_alt ('um'). So clicking on '34µm' the editing string will become
'34um'.
--- end
note: rather then allowing empty strings in name_alt field I made it so if the unit system was the default one a NULL name_alt will just strip the string, since its the default its not needed.
- ignore MSVC warnings when FREE_WINDOWS is defined to quiet warnings.
- the CMake flags were not being set correctly making blender have weirdo colors (no -funsigned-char).
was missing a call to glLoadName(-1); so drawing commands after the bone were taken into account with the selection.
made some other minor changes that dont change functionality.
By Luca Bonavita (mindrones)
The patch renames and moves gl_round_box, gl_round_box_shade and gl_round_box_vertical_shade to UI_interface.h, so the extern usages are not needed anymore.
By Luca Bonavita (mindrones)
From detailed description: This patch doesnt change functionality, but uses the existing color math functions from math_color.c into
sequencer_draw.c.
- fix for minor inconsistency in armature selection, entering editmode and selecting a bone would move the manipulator because the selected bones, childs root wasnt selected on entering editmode.
- use copy_v3_v3 rather then VECCOPY in editarmature.c
The funny thing is: I only spotted this bug in March of this year. Almost one year after the original release. I think I don't parent objects to the camera often.
In terms of code I think that I can even think in a more elegant solution. I don't really need to rotate the camera, but simply to calculate its Modelview Matrix.
"""
m_rasterizer->SetViewMatrix(viewmat, cam->NodeGetWorldOrientation(), cam->NodeGetWorldPosition(), 1.0);
cam->SetModelviewMatrix(viewmat);
"""
The reason why I originally was rotating the camera was to make sure the frustum calculation was using the right camera frustum. For the frustum it takes the camera modelviewmatrix so the rotation really shouldn't be necessary. Leaving as it's for the time being.
* Note: the bug was never officially reported
- A property used in an 'active' poll for UI code in the NLA Editor was missed during the RNA renaming madness, leading to error prints in the console
- For matrix blending code, 'fsize' arrays were being declared wrong size (4 elements long instead of 3).
- use preset subclass with funcs for updating the keyconfig
- keyconfig filenames are used for the UI names as with presets (so separation anymore)
- keyconfig's are stored in the preset dir (scripts/cfg dir removed)
- only the active keyconfig script is loaded
- some bugfixes for saving keymaps
- user interactions no longer saves keyconfigs too, I think this needs to be re-worked.
developer note...
multiple keyconfigs at once are not really needed now that they are stored & accessed directly in python.
for now I left it alone but we could consider to remove this capability in the future.
The FullScreen state was never set. Therefore the window boundary was returning the wrong dimensions (it was assuming that it was not fullscreen).
* Note: blender.exe -W has a similar problem. We are working on that ...
Tested in Windows only. If someone can test in Linux and OSX please let me know if it's good there as well. To test it: ./blenderplayer -f myfile.blend
After last commit I realized that we have a different behavior when running blenderplayer or a runtime. Reason being is that when running blenderplayer the filename is mandatory. While in runtime mode it has no use for it (it can still be passed, but it makes no difference).
I also updated the -h (help) message to be multiplataform and to pass the filename argument only when in blenderplayer mode.
If someone wants to have fun it would be nice to port the blender new parsing code to the gameplayer.
* note: how come "noaudio" is used in the examples but it's not in the options list? *ouch*
I'm leaving as it's in case noaudio get implemented ...
Now reloading the user defaults also unloads/loads addons, resetting the state to the one set in the user preferences.
moved addon functions into bpy.utils
- bpy.utils.addon_enable(name, default_set=True)
- bpy.utils.addon_disable(name, default_set=True)
- bpy.utils.addon_reset_all(name, default_set=True)
the user preference operators now just wrap these.
argc always include the filename, therefore all the optional parameters should check for < argc instead of <=argc.
E.g. That was causing "blenderplayer -f 800 600 myfile.blend" to play in a wrong Bpp (since it was trying to parse the filename to fullScreenBpp).
Bug introduced on rev. 16448 (in 2.49) and merged to 2.5 in rev. 19323
This happened/was most noticeable when clicking on a bone of an armature in posemode but not active (i.e. some object other than the armature is active when trying to select the bone).
* multiply the diffuse and specular colors by their respective intensities
* only output <transparency> if MA_TRANSP flag is on
* only output <reflective> and <reflectivity> if MA_RAYMIRROR flag is on
- option only available to cmake, scons and make have this enabled always.
- without this clang/llvm can compile blender
- this was the second biggest internal lib, 192mb -> 172mb for all blenders libs (with debug flags), so gives some speedup to linking.
this was the one module left to be renamed.
We didn't get to any conclusion in the mailing list, but Campbell suggested texture and I think its a good name. We can change any time before next Beta, but for now I think it's handy to have something.
The original proposal is from May:
http://lists.blender.org/pipermail/bf-committers/2010-May/027587.html
The reason this didnt work is all sibclasses of pythons type() or PyType_Type in C, have to have their size set to sizeof(PyHeapTypeObject) rather then sizeof(PyTypeObject) as you might expect.
This is strange since its not a heap-class (defined in pythons runtime), but a static C type, so Im not sure about this, and cant find any documentation but it seems to work ok.
Nurb's material number used to be clamped to [0;-1] for case object hasn't got
materials. Forced to keep material number on zero if object hasn't got any materials.
- operator properties are now converted into python property() class members which bypass the operator 'properties' member.
self.properties.mysetting
... can now be written as ...
self.mysetting
- fix for error reloading rigify
* Now cloth reads cache using subframes when rendering.
* Cloth cache also was reset every time on the start frame which kind of defeats the point of the caching.
eg:
bpy.types.Scene.myprop = BoolProperty()
note, this uses an ugly python metaclass, this should be replaced with a C implimentation which is included but commented out, causing crashes in pythons GC which gives no hint as to where the bug comes from.
Fixed bug #23657, "Modifiers dosen't work when you select diffrent mesh for object"
Multires modifier now adds empty mdisps if they're missing, rather than displaying a warning
Switching an object's mesh will now check for a multires modifier; if found the modifier's total number of levels are reset to match the mesh's mdisps
Switching the mesh also forces a multires update so that sculpted changes aren't lost
- rna internal deferred properties now store the functions as PyObjects rather then C function pointers
- Property functions now allow the first non keyword argument to be a class.
By Lukas Tönne
In the console RNA paths can use double quotes (" ") or single quotes (' ') to select from a collection, an
ID property or a vector/color component. The RNA_path_resolve function however only accepts double quotes. This patch
adds the ability to use single quotes in areas other than the console too.
PS. Note the very nice patch ID :)
Buttons for editing RNA paths/array index for F-Curves that aren't working are now actually functional. This means that when invalid paths are present, they can be manually fixed up.
Setting the classes __dict__ member directly didnt update the types slots (python internal type stuff used with subclassing), so class.bl_rna was returning an incorrect value.
Set the value using typical python setattr command.
from Alexander Kuznetsov (alexk)
--- copied from the tracker
Every image inside Blender is in linear color space and gets converted to SRGB upon saving.
Level node analyzed the linear image, which was not the one user saw because other output nodes converted image to
sRGB.
This fix analyzes the image that user see (converting it to correct color space).
Here is difference:
http://www.pasteall.org/pic/show.php?id=5559
First histogram (before the fix) tells that image is underexposed, which is not the case.
cant test if this fix solves the problem matt is having but it at least fixes an error caused by classes created in exec(),
when the properties pointers were copied the hash key still referred to the python object which could be freed. in most cases this wouldnt happen (would be kept in bytecode) but with exec() the property string is freed immediately.
from Mathew Burrack (mburrack)
...also applied a fix so this flag wont be written but including this patch for older startup.blend's.
--- from the tracker
I've created a patch for the texture path corruption bug, 23337. Basically, G_FILE_RELATIVE_REMAP was improperly getting
saved out to the startup.blend file, causing issues when the autosave timer went off. The proper fix is to mask out
that flag so it doesn't get written out to .blend files itself, but since that doesn't fix any pre-existing startup.blend
files, I just mask it out when startup.blend is read in instead.
I've tested it locally and so far, it seems to fix all the issues I've had with texture image paths getting corrupted.
I haven't figured out how to properly test the remap-on-save option in the save as dialog, though, so I don't know if
I accidentally broke that or not (although I don't see how I could have).
* Committing new logo, same design as old one, just larger size, by Alexander Kuznetsov
. Thank you!
Note: Could only test on Vista. But it should also work on XP machines.
Note 2 (to jesterKing, elubie: Vista uses the winblender.ico also for the blend files, it doesn't use the winblenderfile.ico. But I guess Windows XP use that still?
If not we could remove that, also blenderplayer, is same as blendericon.
- move: material.add_texture(tex, coords, mapto) --> material.texture_slots.add()
- added material.texture_slots.create(index), material.texture_slots.clear(index)
- texture slot functions also work for lamp and world now.
Other minor changes
- allow rna functions to set FUNC_NO_SELF and FUNC_USE_SELF_ID at once.
- [#23317] Changed some operators' RNA to accept lengths, a modification I made to this patch made it not work as intended, removed this edit so unit buttons appier in the UI for certain operators.
- Sphinx doc gen, 2 columns rather then 3, didnt quite fit in some cases.
- use 3 column lists for inherited props, funcs and for references to save on vertical space.
- use the blender version string for the upload path and PDF name.
* When removing a modifier that effects the depsgraph the DAG_scene_sort has to be done after the update call so that dynamic systems can clear the collision surface properly.
* Note that this fix might slightly change the simulation results of some files that use the stickiness value, but lowering the value should fix these issues.
- set_frame() --> frame_set()
- set_context_pointer() --> context_pointer_set()
material adding works for curves and metaballs, new function to remove materials.
materials.link() didnt well fit how this is used elsewhere
- order matters
- it can be linked more than once.
- remove(material), isnt that useful since you need to manage indicies.
... use list style functions instead. materials.append(mat) / materials.pop(index)
Reported and patched by Mariusz Maximus
I overlooked this one while fixing up for win64 compiling, because I had already manually created this directory.
* Although not strictly a bug it is the expected behavior and won't mess anything else up.
* Note: the lattice is applied to the actual hair keys instead of the calculated strands so the applied result will differ a bit from the original.
* The actual problem was that the total amount of particles was rendered at all, since only the displayed percentage was calculated correctly.
* New behavior is that before baking (baking is always done for full % of particles) the display % is used for rendering too for dynamic particles.
* Also added a warning below the display % slider to inform about the situation.
Reported by Reiner Prokein
After aligning active camera toggling between last view and camera worked wrong in such a way that the view previous to the one to which was aligned was rotated to. In some cases rotation to last view seemed incomplete and ortho was used. Thanks to M.G. Kishalmi (lmg) for showing through a Skype screen share.
This is now fixed by copying the current view info before active camera is aligned.
http://www.vrchannel.de/blender/cylinder_rename.png
Mesh Tube > Mesh Cylinder
NURBS Tube > NURBS Cylinder
Metaball Cylinder > Metaball Capsule
I know that naming is something not everyone agrees on, but these terms look geometrically correct.
- ID properties now suopport non utf-8 strings for their values but not their keys.
- moved utility functions into py_capi_utils.c from bpy_utils and bpy_rna.
- import/export paths have to be printed with repr() or %r, so non utf-8 chars are escaped.
When creating project files the multi-configuration generator is used. CMAKE_BUILD_TYPE is then not set, so always release libs were copied.
For now copy both release and debug.
from Alexander Kuznetsov (alexk)
bugfix for [#23553] F2 on filebrowser = bug?
from the tracker
--- snip
File Explorer redraws weirdly on second press of F2 if non-default view or file types were selected previously.
This patch prohibits second call of file selector in the same window.
The bug goes much deeper. If file selector is never closed properly (cancel or select), it never gets released. (at
least the handler). If you press F2 or Ctrl-F3 ten times and than "Back to Previous" and repeat all of this
several times Blender will freeze. Also after calling file selector at least two times, on cancellation Blender will
return to full area independently to what state it was before.
---
include small unrelated change to quiet unpack prints when fonts are not found.
- function renames, move WM functions into collections wm.add_keymap() --> wm.keymaps.new()
note: new is used for named items in a collection, which return the result.
- Action.get_frame_range() is now a readonly property 'frame_range', floats rather then ints.
- OBJ import/export now work with non utf8 paths. (all exporters and importers need changes like this)
- strip non utf8 chars from new ID blocks (also applies to renaming)
- set the file rename button to allow non-utf8 chars.