Blender Git Loki

Color picker fixes... (bug #1490)

Since color is stored RGB only in Blender, the conversion from and to HSV
causes values to be clipped all the time.
Solution found is adding a persistant hsv storage to the uiBut struct, and
have it working on that only while hsv button is open. Still, after usage
(leaving picker) the conversion will change values to rgb.

Also added; redraw event for editing buttons, to show correct changed
color while using nkey picker

Added to the top of the file: struct SpaceOops;

to stop a warning about SpaceOops defined inside of parameter list...

Kent

Kenneth Strom's patch to make Fractal Subdivide work in the -
direction as well as the positive.

Kent

Bug #1468

"Home" or "View All" in NLA window didn't set a good result in all cases.
Now it uses the Scene start/end frame, which isn't correct always, but at
least gives results.

NLA and Action drawing is total mess! Certainly nice job for cleanup for
next release.

Accidentally committed a line paste with declaration outside of { }...
Thanks jK!

This fix, bug #1342, makes blender consistant, but won't really solve
weakness in design.

Issue was: if you delete the 'mother metaball', which is an object with
name without a number, the entire metaball family isn't polygonized. Sofar
so good, where it not that rendering still worked, because it searches all
objects in the scene for a family member with lowest name.

Doing such a search within the main drawing loops will become a major
slowdown, so this better isn't done.

I don't mind sticking to (documented) consistant behaviour. Alternative is
a flag in object to denote it's a Mother Ball, and check for that instead.
This isn't nice to code in current Blender though... we don't have a signaling
system where to insert such tests reliably, making it very hackish.

Jiri; I hope you can agree with the solution! :) Just want to have tracker
cleaned up, refine it to real bugs.

Not a real fix, but report #1638 was a valid confusement...

When using Unified Render, the Gauss option cannot work, so the button then
disappears to indicate so.

Fix for #1617

Volumetric spothalo shadow clipped at z=0.0 instead of correct z= -1.0.
Just forgot that zbuffer for spot shadow uses entire signed integer range.

Bug fix #1341

Deleting all vertices in a Curve Path, crashed/corrupted a FollowPath
constraint using it.
Now it defaults to (0 0 0) for an empty path.

Three fixes;

- bug #1307; copy full scene didn't update links in constraints
(was an oldie bug!)
- also fixed same for Hooks, which even't didn't duplicate yet
- HOME in outliner (show hierarchy) now only does current scene

Fix for #1626

Pulldown menu "Apply deform" in (3d header) called 'make duplis real'.
Easy fix :)

minor fix, more a beauty glitch, bug 1612;
when adding first ipo curve position, editmode on the curve shows weird
handles, which solves when adding the next keys though...

Fix is that it sets all handles on zero size now. (is auto-handle, so no
harm done)

Fill and beautyfill now respect new edge/face selection flags.
Also flipped normals for default filled faces, they were pointing away
always.

Fix for #1569

Problem was dynamic linking of files with game logic. The links between
sensors/controllers/actuators got lost.
Never though i'd still understand the code, not too bad at all! :)

Fix for bug report #1566

Some 3d cards, IBM mostly, crash when glBitmap draws while in glPicking
mode (for select).
This was fixed for empty objects before, but armatures also draw bone
names... here the check for 'picking' was added

Cut'n'paste code, to fix bug 1500, resulted in crash when not rendering
extruded curves. thanks intrr for noticing! :)

Fix for bug #1562

There's an ancient code in Blender to denote a projected vertex coordinate
is invisble, it sets the x coord at 3200. :)
This wasn't updated while coding loopselect, nor edge select. Causing in
extreme zoomed in situations vertex selecting go wrong.

Also added; option "don't load GUI" in fileselector doesn't get saved in
files.

fix numerous memory leaks in the math types.
fixed bug: #1633 Memory leak in M_Mathutils_Vector

The math types ( matrix, vector, quad ) now make
copies of data passed to them rather than holding
a pointer to memory that cannot be freed, or that
may go away unexpectedly.

This also clarifies the problem of who is responsible
for freeing memory allocations.

Pre-checkin files are tagged mem_leak-1 in case this
breaks something.

Linestyle (dashed, using glStipple) seems not to work proper on
circle selecing. I can't find clues (thanks k-rich for help!) what it
exactly causes...seems like inverted draw + glStipple + draw circle
somehow confuses a driver... we can live without i guess?

This commit is to verify it now works ;)

Fix for bug #1500

Ray_transp on curve objects (filled polys) didnt work well, normals
were not pointing outside all nicely