Blender Git Loki

* silence compiler about warnings for C++ files a bit more.
- from what I can see now, the larger part of warnings is now about conversions "possible loss of data" (ie. double to float, etc).

soc-2008-mxcurioni: merged changes to revision 17246

speedup occlusion by removing octree raytracing method and use the bucket's face list.

* doc update

This is coverity issue CID: 456
fixes a buffer overrun issue.

Kent

corrections to the C_WARN and CC_WARN variables for scons config. I didn't actually changed what the flags were, just the format, but Platform Maintainers, please check.

paint onto the final derived mesh rather then the original mesh data, so painting onto a modified mesh (subsurf, mirror etc) works as expected.

basic projection painting - should work without crashing and not be too slow.

does not use any opengl for calculations (2.5 friendly!) and free from screen pixel artifacts. raytracing is used for occlusion.
Details
* Uses a 2D screenspace bucket grid that store intersecting faces and a list of UV pixels to optimize lookups between the brush rectangle UV pixels
* Buckets and faces are initialized when a brush first touches them.
* on initializing all the faces used have their UV pixels are converted into screenspace and copied into the buckets for painting (if the pixel is not occluded by a ray cast).

still a lot to do - blend modes, float buffer, image wrapping, image texture filtering.

Add feature requests here
http://wiki.blender.org/index.php/Wahooney_re/Paint_Branch_Proposals

* Added a new turbulence type: Time. It's not entirely well tested, but so far working ok. It's smoother looking than 'velocity' but may need more in depth investigation.

Fix compile flags in linux config to fit Nathan's changes.

Warning to others: Those flags are lists now, be careful and update other configs if needed.

* Minimum SCons version is now 1.0.0
- Code has been changed to reflect this (ie. deprecated functions are not anymore used)
* clean up the C and C++ compiler flags mess.
- in the environment construction of BlenderLib all the compile flag governing options have been split in the *C*, *CC* and *CXX* containing equivalents.
C is for C compiler only flags. CC is for C and C++ compiler flags and CXX is for C++ compiler only flags.
All the platform default config files need to be double checked and fixed wherever it looks necessary. Either DIY, or send me a note with needed changes.
- a start for the BlenderLib parameter list has been made - all the SConscripts need to be checked and modified to hand in flags properly.
* A theeth request: make -jN settable in the config file.
- I give you BF_NUMJOBS, which is set to 1 by default. In your user-config.py, set BF_NUMJOBS=4 to have 4 parallel jobs handled. Yay.

Lots of code cleanup and some new features.

Clean and split editbone duplication methods for easy reuse.

Copy and retarget selected bones to stroke conversion method.

This is the start of the infamous templating system.

A simple demo video is available: http://blenderartists.org/~theeth/bf/etch-a-ton/retarget.ogv and http://vimeo.com/2111535 (for those who dislike ogg/theora)

* Build aborts when giving options on command-line when WITH_BF_DOCS=True
- make sure epydoc generation doesn't get a fit over options given on scons command-line -> don't use arguments from command-line.

Build fix (C90): Declaration after statement

Fix for dependency graph cycle print, regular "Parent" relation was
incorrectly printed as "Curve Parent".

Bugfix for [#17879] Speed vectors/velocity data not working on ALL fluids.

* Added a new, slightly experimental force field type: 'Spin'.
It works similarly to Vortex, but it's a more controllable - it
doesn't send the particles off accelerating into space, but
keeps them spinning around the Z axis of the force field object.

This is safe in the branch, but Jahka if you have any feedback,
I'd be curious to hear :)

moving stuff home. first draft for template retargetting

merge trunk 17213:17221

EditVert hash *is* used elsewhere in the code, so just to be safe, use a scratch array instead.

This is actually much safer than juggling values in the tmp union all the time.