Blender Git Loki

Tile Branch:
* converted the MovieClip node to tiles.
* added color profile conversion operation

Channel Matte Node Tile
node from trunk converted to tile.

Fixed use of deprecated bNodeStack values from sockets. These have been replaced by the default_value pointer, which points to one of the bNodeSocketValue*** structs, depending on socket type. The bNodeStack ns member in bNodeSocket might still work occasionally since it is restored in writefile.c for forward compatibility (happens on undo push too).

Fixed registration functions for last section of compositor nodes in registerCompositNodes (dilate/erode 2, box mask, ellipse mask, bokeh image, bokeh blur, switch and opencl test). The register functions need a tree type argument now instead of a plain ListBase.

adding double edge mask to the list of to be converted nodes + blenderplayer stubs update

svn merge ^/trunk/blender -r43091:43824(HEAD)

Conflicted file manually update:
===============================
readfile.c

Note 1:
=======
The Double Edge Mask is not in tile since it's not a simple convert.
The author of this node is aware of it and should have it working there eventually.
(I'm going to add it in COM_Converter.cpp in a next commit)

Note 2:
=======
In COM_compositor.cpp we are passing only "Compositing" instead of what we have in trunk "Compositing 1 Gamma"
If we decide to go with trunk then we need to incorporate the 43283 and 43287 commits.

Something like:
COM_compositor.cpp:51
- editingtree->stats_draw(editingtree->sdh, (char*)"Compositing");
+ char str[128];
+ BLI_snprintf(str, sizeof(str), "Compositing %d %s", curnode, node->name)
+ editingtree->stats_draw(editingtree->sdh, str);

Normalize Node - Tile
The implementation is simple. Fancy alternatives to gather the minimum and
maximum through recursive loop in the tiles proved inefficient (Jeroen Bakker's
tests)

I'm also using the generic NodeTwoFloats to store the temporary values calculated in the loop.
And different from Blender trunk I'm not checking to see if the value is between the accepted Z range
(e.g. (*val <= BLENDER_ZMAX)). I can't see why that would be necessary. To avoid overflow? If so this be handled (aka clamped) outside the nodes in a generic way.

Credits to Jeroen Bakker for the solution in using mutex (copy pasted from his ToneMap code) and review

displace node cleanup (using clamp) + simple displace glitch fix + not using protected variables

Tile branch
* Added the Glare streaks operation
* Added a clear method to the MemoryBuffer

Jeroen Bakker
- At Mind -

TileBranch
* removed a python null-value when doing the next steps:
1. start blender
2. go to the compositor view
3. press 'N'

issue was that no node_tree is available

Displacement Node Simple version - Tile

Displace Node - Tile
--------------------
I had to make it into a Complex node to allow the readEWA to be used.
(the other alternative is to expose readEWA to all node types when dx and dy are
passed as parameter to read).

This node also uses the readEWA. It's the last one that needs that (from trunk nodes at least)

Thanks for Jeroen Bakker for the help in the determineDependingAreaOfInterest. Using +1 as max for the uv was producing precision problems resulting in a strange artifact (a vertical and a horizontal line from the origin (0,0))

TileBranch
* resolution propagation patch

Resolution were not always propagated between read/write buffers. This will be important when implementing Buffer caches

Jeroen Bakker
- At Mind -

TileBranch
* Removed storage of tiles to disc. in order to support buffer caching
* Cleaned up spaces/tabs

Jeroen

fix compile: doubleedgemask define will only be available when trunk get merged in tile again

including the non implemented nodes in the COM_Convert.cpp to make it easy to see what's missing
(thanks bash tools :)

cleanup - removing unecessary #includes
(COM_InputSocket.h and COM_OutputSocket.h - as suggested by Jeroen Bakker)

Color Matte Node - tiles
thanks Jeroen Bakker for review + small corrections

Tile - Chroma Matte Node
note - shadow adjust (c->t3) and lift (c->fsize) are no longer used in the algorith (since somewhere in 2.5 dev).
The properties are still in ChromaNode and have a RNA property.

This should be address in trunk, not in Tile I think.
If those are parameters needed for a more robust algorith we bring it to Tile later.

Separate Combine YCbCr nodes + fix for YUV
------------------------------------------
YUV was missing in COM_Converter.cpp

Sep/Comb YCC could even inherit each other (given the setMode function is th same). But I think it's fine to duplicate that. Thoughts?

Also, I'm not using m_privateprops C++ like naming. As we are not doing this in the other nodes.

reviewed/approved by Jeroen Bakker.