handling callbacks that checked for double names... this was caused by
a bugfix I did a week ago for ESC on rename button in outliner.
Discovered in time by Andy, thanks!
Enabled Striding in NLA to also use FollowPath Constraint. Until now it
only worked when Object was parented to a Curve.
Interesting feature now is that it checks for all FollowPath Constraints,
and it chooses to stride over the first Path found with influence > 0.5.
Note: to make swithing to other paths work, map the time curve for these
paths to match the switch. This is not very NLA friendly... but let the
dudes in studio test it first. :)
Todo for later once; ensure proper blending...
the current frame was outside of strips. This gives potential errors with
playback that skips frames, or for a renderfarm. And, while editing the
poses got 'stuck' if you changed to different frames.
Now if no strips are found on 'current frame', always the first strip to
right is used, or the last strip to the left.
no shadow a little bit. Uses same threshold function as the "Bias" button.
(That latter still works, and will use an automatic bias value based on the
geometry, to prevent terminator errors in raytracing).
Anyhoo, with this manual bias you can get rid of terminator problems for
weird diffuse shaders now as well (like tangent or fresnel).
Committed a few more files than needed, that's just code cleanup.
For some ancient reason, the stride option only worked on a single strip,
and extended always. This made it nearly impossible to integrate it well
with other actions in NLA.
This commit changes it as follows;
- As any strip, the striding-strip also stops at the end of a strip
- This allows to put multiple different actions on a single path, and if
all of these have the striding option set, the actions will each do their
own individual stride.
- To match the different actions, a new "Action Offset" button was added
in the NLA Panel, which allows to internally cycle the action.
- Of course, blend-in and blend-out works nicely too.
Here's a quick AVI test with 2 actions. There's some slipping of the feet
between actions still, because of the stridebone blending. I might look
at that later, for now you can correct it with a simple Ipo on Armature too
http://www.blender.org/bf/0007_0151.avi
a the beztriple, so things like the hidden and handle select states were
set to random values. Added a beztriple.hide attribute so that the
hide attribute can be set/cleared from the BPy API.
Finally (after 10 years!) fixed the issue that was marked in a comment in
the top of the readfile.c... when using multiple library files at once,
the code that tries to map old to new pointers could fail when accidentally
pointers (in different files) had same old values.
Current code now uses proper lookup tables per open library file.
Last issues to solve;
- pointer tables still suffer the 'accidental pointer' error when linking
multiple levels of library deep, with recursive back linking to higher
levels... (I know is stupid, but should be solved!)
- The code for relative paths doesnt work for multiple levels, when the
files use different relative paths.
not use the proper call to increase user count && set the action to get
a tag "extern linked". This caused actions to get lost when indirectly
linked from library files.
So... now I understand file reading and libraries again, I couldn't leave
it to do a nice optimize step. This commit introduces bsearch() for finding
the elements in the pointer table, while reconstructing links. Tests reveil
a speedup of a factor 2-5 on complex scenes. And since undo uses same
system, quite nice in general!
(Note; speedup sometimes can't be noticed, when a lot of subsurfs or other
displaylists need to be generated.)
for Groups in official release still, which will give very bad results when
loading an Orange branch file with bf-blender, save over, and read back
in Orange branch.
Now, reading files with Groups in bf-blender will just completely ignore
them (not read nor save). That's destructive, but normal behaviour and
stable.
Phew... a whole day revision of library linking system... there was a bug
when objects got linked, then removed, and then linked via a group... this
was a weak part in the library system, when data is used both "direct"
(like local object is linked to scene) and "indirect" (when group from a
library has same object).
The crucial fixes are easier to find by reading the diff... this system
is just braincrunching complex! Just as note to self then;
- ID_ID oldmap-lookups are now protected
- removed adding oldmap lookups for already read data
For Matt & Basse: you can do a very good cleanup of old files by putting
the "rt" button on 127, and then save file. This will skip writing any
reference to library data in the file, with exception of group links.
Reload a file might give some link errors, but a 2nd save and load then
should result in a clean file.
- On file reading, a check is added to see if a Pose has channels to
missing bones. Poses are supposed to match an Armature completely.
(Thanks for crash file, Johnny!... but how did you do it!)
- Own collection: added depsgraph update on setting all layers (with the
ACCENTGRAVEKEY). This also could potentially crash with modifiers
- Renamed Xkey menu in NLA to more properly tell what it does. :)
- malformed nmeshes could crash Blender with a sigsegv. Related to old
behavior that accepted "faces" with one or two verts.
- removing unused var (store_edges) + doc update.
This is the case:
- Empty has Group duplicator
- Empty has NLA strips to animate the group
On linking the Empty to another group (with button in ObjectButtons), it
checks for the current strips in NLA, and tries to find the proper objects
in the new Group, based on name matches.
If not found, it sets the strip-objects to zero.
don't work consistant;
BLI_makestringcode()
This one works correct for paths that go up; it then makes a relative
path like : //../lib/file.blend
BLI_convertstringcode()
This one delivers different paths when it expands, the previous case
will just append the relative path, like /work/project/../lib/file.blend.
That string cannot be compared with the original... which happened to find
out if a library was read already.
Solution for now is that only the relative paths are compared, but it
won't work if multiple relative levels are used... like for libraries
that link recursive. To solve that, the entire 'relative' concept has to
be redesigned.
Image.New(name, w,h,depth)
Image.start- for animtex
Image.end - for animtex
Image.speed - for animtex
Image.packed - read only bool
See the pydocs for details.
data (local data links to lib data) and "indirect" (lib links to same lib
data).
A real solution should be found though... there's a conflict in the code
I don't fully grasp yet.
when it duplicated mesh data.
I'm not thrilled with how I implemented this code, but currently don't
know a better way. If someone more familiar with how blender duplicates
objects and converts things to meshes wants to have a look, it would be
appreciated.
Apparently python allows to create Key blocks without proper *from pointer
back to the owner of the Key. The Key unfortunately wasn't mean to be a
dynamic block linkable to any Mesh...
This patch sets the from pointer on any call to do_mesh_key, which is safe
to do anyway. I made a note to get rid of that convention once, but at
least now we got non-crashing blenders. :)
(In bugreport noted that we need to check Python code for it too!)
Adding Ipos to Sound strips in Sequencer didn't work anymore. Adding a new
curve with CTRL+click also wasn't smart enough to detect that only one
channel existed there... fixed that too.
Another ancient bug in unified render... the lamphalo wasn't filled in
correctly over sky, especially when the background had halos (or stars).
Only shows when using filters other than box though.
wire frame became very limited... the information of faces (vertex colors
and UV reside there) got lost.
Solved it nicely with creating a large index table, and use bsearch() to
get quickly the matching information.
Then I noticed the render code needed fixes too for wire, no proper UV's
were calculated over the wire edge.
The tool "Single User" -> "Object data" missed a tag to recalculate the
geometry. This caused meshes to disappear after entering editmode on the
former linked mesh.
