logical parts, and include files altered to denote internal and external
functions.
include/editmesh.h: internal calls for editmesh_xxx.c files
include/BIF_editmesh.h: external calls for these files
src/editmesh.c: basic alloc/lists and in/out editmode, undo, separate
src/editmesh_lib.c: basic utility calls for all editmesh_xxx.c (no UI)
src/editmesh_add.c: add prim, add duplicate, add vertex/edge/face (UI)
src/editmesh_mods.c: selecting, transforming (UI)
src/editmesh_loop.c: loop tools like knife, loop select, loop subdiv (UI)
src/editmesh_tools.c: other tools (extrude, spin, etc) (UI)
And a new file:
src/meshtools.c: tools for Mesh outside of editmode (normals, draw flags)
Certain actuators (Add Object/Replace Mesh/Scene/Track To) used to always require a target, or they would not influence the scene. If the actuator target is always set from Python, this could be annoying.
- EditVlak -> EditFace
- variables called 'evl' -> 'efa'
- functions with 'vlak' in it now have 'face'
Just thought was nice starter for editmesh recode...
detected as others, LR_COMMANDKEY
Unline previous commit, Apple key doesnt map to control anymore, but is
handled separate where needed. Now only for undo (Apple-Z)
- Made unified API for undo calls, to be found in space.c
BIF_undo_push(char *str)
BIF_undo(void)
BIF_redo(void)
These calls will do all undo levels, including editmode and vpaint.
The transition is work in progress, because mesh undo needs recode.
- New global hotkey CTR+Z for undo
Note: 'shaded draw mode' still is SHIFT+Z, the old CTRL+Z was to recalc
the lighting in shaded mode, which already became much more interactive,
like during/after any transform().
Recalc hotkey now is SHIFT+ALT+Z
CTRL+<any modifier>+Z is redo.
- For OSX users; the Apple-key ("Command") now maps to CTRL as well. This
disables the one-mouse-button hack for rightmouse btw, will be fixed in
next commit. At least we can use Apple-Z :)
- Old Ukey for undo is still there, as a training period... my preference is
to restore Ukey to "reload original data" as in past, and only use new
CTRL+Z for undo.
- Added undo_push() for all of editobject.c and editview.c. Meaning we can
start using/testing global undo in the 3d window. Please dont comment on
missing parts for now, first I want someone to volunteer to tackle all of
that.
- Since the global undo has a full 'file' in memory, it can save extremely
fast on exit to <temp dir>/quit.blend. That's default now when global undo
is enabled. It prints "Saved session recovery to ..." in console then.
- In file menu, a new option is added "Recover Last Session". Note that this
reads the undo-save, which is without UI.
- With such nice new features we then can also kill the disputed
Cancel/Confirm menu on Q-KEY.
- Added fix which initializes seam/normal theme color on saved themes.
They showed black now.... (Note: that's in usiblender.c!)
that a displist has to be recalculated, you only have to free it.
In the init_render_mesh() not all situations were checked OK.
Thanks Basse for finding it!
Wrong usage of B_NOP code in button events (xxxx | B_NOP), which is useles
since it is defined as -1.
Reminder for all: use a B_NOP if you want a button event not to be passed
on to the event queues.
1) Curve deform
http://www.blender3d.org/cms/Curve_Deform.392.0.html
Works simple as expected, but keep track of the rotation axis
in F7 buttons (Track X Y Z)
Only Mesh deform supported now.
Code changes:
- centralized deformation calls in curve_modifiers() mesh_modifiers()
etcetera. Here also other effects can be added like wave. Now the
evaluation order is fixed, but should become optional. It also doesnt
use the Displist anymore as deform-input. That latter part is unfinished
yet.
This code also is used for Hooks and will be needed for softbody
- made convention stricter that displists are being checked on in
drawobject(), this to prevent routines to make new displists recursively
(like armature does). Now a freedisplist() is sufficient to signal that
a new displaylist should be made.
2) Object Hooks
http://www.blender3d.org/cms/Object_Hooks.391.0.html
Support for Hooks is added to Mesh, Lattice, Curve and Surface objects.
For Armatures this would require some more work & research.
Main goal for this feature is to provide quick & simple access to the
underlying geometry in Objects on Object level, supporting hierarchies and
Ipos etc.
1) append in current active layer (off by default)
2) append at 3d cursor (off by default). When more than one object is selected this will first calculate the bounding box of the centers of the objects, and then use the distance of the bounding box center to the 3d cursor to transform all objects.
Some implicit cast warnings and more importantly, changed the parameter type to int for functions passing the particles random number seed around. No need to use floats there, it's an int anyway.
Effect.c is now warning free (on MSVC 6.0 at least)
The algorithm used before to split up the mesh before unwrapping was based on
vertices, resulting in leakage in some setups. Now it's edge based, allowing to
select only two edges of a triangle, and not automatically selecting the third.
Read today's meeting minutes for a description.
Bjornmose: We'll need some example and screenshots for the dev pages of blender3d.org. If you can do some, neat, if not, I'll do them.
Matt, Emilie and Chris: Please review UI addition. Pixel alignement and all that fun stuff has not be overly looked for (though I did some cleaning by adding alignment blocks).
- file-to-memory save
- incremental difference steps (compression)
everthing has been tightly coded to use minimum of memcpy or allocs. In
fact this system works with a single full buffer (=file) in memory, and undosteps as differences from it.
Speed gain is factor 4-8 faster. I've added it in CTRL+ALT+T timer menu for
a test. Please note the gain is especially in the undo-storing, not in
retrieving undo.
Also new: file read option to skip UI read (file menu). This now also is
default for the undo system.
Removed a block of a bunch of unused vars...
Added a return NULL to the end of a funtion that is suppose to
return something and could fall through.
editconstraint.c
Added a newline to the end of the file to get rid of a stupid warning.
Kent
They were not needed were for functions that were in the same .c file
higher up and I think its sort of a non standard way of doing things.
(using prototypes would probably be more inline with the other code but
as I mentioned they are not needed and were causing warnings in gcc)
Kent
LEVEL_2_CPP_WARNINGS += -Wstrict-prototypes
LEVEL_2_CPP_WARNINGS += -Wmissing-prototypes
Every version of g++ i've used has stated these two were not for g++
but for c or objectivec.
Kent
- transform (grab, rot, scale, etc)
- all button commands, including menus
I sacrificed for now the UKEY in the 3d window for it. Shift+U does a redo.
(Only in 3d window)
What this system does is saving files in the temp directory (user pref).
The filenames are cycled around (32 in total now).
This commit will follow shortly with a userpref for it, not to frustrate
people who want to work in normal fashion with blender.
I cleaned up the code a little did a couple of these:
if (blah > stuff - wah) blah = stuff - wah;
changed to....
tmp = stuff - wah;
if (blah > tmp) blah = tmp;
and combined multiple if statements
Kent
separate it...
1) Curve/Surface editmode undo
Uses same syntax as mesh undo, so simple to integrate. Edit-curve data is
also quite simpler, so no need for any hack at all.
It re-uses the undo system from next point, which is nice short & clean
local code
2) Framework for global undo
The undo calls themselves are commented out. In a next commit I want to
enable it for a couple of main features, for further feedback.
The speed goes surprisingly well, especially with this new version that
'pushes' undo after a command, ensuring interactivity isnt frustrated
3) framework for texture based icons in Blender
Not activated code, but tested here. Part of 2.3 UI project.
btw: Johnny Matthews will assist in (and complete) the undo project
I added ifdef's around a var declaration since the code using
it had them... Also initalized two vars that were questionable if they
could reach a state where they weren't assigned.
Kent
