Blender Git Loki

Image and Draw wernt converting the paths of an image properly.
Getting teh size of an image with a relative path "//foobar.png" would always fail.

=== Sequencer ===

Bugfixes:
- Made the HDD-Audio-Code silence buffers in advance before doing a seek
operation. Stops noise on end of file or decoding errors.
- Corrected hanging of audio decoding loop bug on some MP3-files.
(you had to close blender and restart)
- Made preseeking code not seek before beginning of file.

seqaudio:
- made some "local globals" static.

Problems remaining: VBR-MP3-length isn't detected correctly so the
displayed sequence strip is too long...

Two issues:

* a misuse of the Mesh module meant that I kept creating new
meshes each time the script was run;

* the commas in the vertex color export looked weird.

Plumiferos bug report; duplicating strips with effects crashed.
Was in new code that handlerized callbacks... leaving function pointers
unassigned. Dubious code... I would NULL everything first before filling.

A fairly major revision of the VRML97 export script.
Changes include, but may not be limited to:

* Support for meshes with multiple materials/face images
* Mode modular design that more closely reflects Blender's
internal workings
* Position, scale, and orientations are now exclusively dealt
with in Transform nodes, making the math more unified and
way easier to understand.
* vertex colors either written when mesh has SHARED_COL
face property, or when mesh has vertex colors and
first material with VCOL_PAINT (a little crufty, but
maybe will try a better way later)
* Support for debugging output to the console by setting
the 'rt' button to 42 (for mild verbosity) or 43 (for more
verbosity)
* Potentially long lists like vertex coordinates, face indices,
etc. are now unindented (why potentially waste thousands of
tab characters?)
* All lines in the script are under 79 characters (as per the
python style guide).

A major revision will usually beget major bugs, so
please test, test, test to make sure I haven't broken your favorite
feature in this script! (I've done a number of tests loading
output into the Bitmanagement viewer and into 3DSMax.)

Stoped a list that didint need to be made.
~20% faster and memory saved by re-using the face and edge lists and their classes each pass rather then __init__'ing new ones.
Stupid mistake- was only collapsing edges with UV's

- fixed geometry init problem (stupid, forgot to call reset function)

Fixed some evil bugs in the poly reducer messing up UV's every now and then.
Added support for "Weighted Collapse" Before an edge could only collapse into its middle,
Now the edge collapses into a point bias'd by the 2 verts Concave/Convec "Pointyness" value as well as boundry weighting.
This works much better for boundry verts. - UV's Vcols and Weights are correctly interpolated into the new location.

Added a tool in the mesh menu for accessing the poly reduction tool.

Added Blender.Get('rt') so that the pythoner can access the value
of the rt button for setting debugging levels.

===Python API===
Change code for rigid body access to use attributes instead of methods; also
add documentation to epydoc.

#4201 the second
humm .. a bit more testing showed AVI codecs want it 2 byte aligned
found nothing in MSDN Library about that

Fix a buffer overflow in Blender.sys.splitext. It was checking for bounds
(although it did this one char too 'friendly'), but still happily copying
the string even if it was too long.

Pythoneerz, please check if this is the correct fix :)

Made import set the object layer to the current scene layer, applied the 3ds name to the object and the mesh.

Ghost can be any bound type in Bullet, so removed the edit restrictions

added 2 more Rigid Body Python access, should be enough to make a start for Collada 1.4 Physics Import/Export

Added first Rigid Body python access (mass). Basic others will follow. This helps Collada 1.4 Import/Export.

made some changes to vrml97_export, this script needs to be re-written, until then fixed some incorrect assumtions it made- was using the first world in the list for AMB color no matter what world was being used, was also using the image name rather then the file name for its image URL references.

Made it so when a text block is removed (using the cross button in the header), the previous text block becomes active, to be consistant with scene and screen delete operations. Also makes removing all the text's within a blend a lot less tedious - (Something I do for bug report Blend's)
change to drawtext was just to comment on what pop_space_text does, names not obvious.

*added new Object method Object.evaluatePose(frame)
- the new methods will advance an object's pose correctly to any frame

Example:
myobj = Blender.Object.Get('obert')
for x in range(10):
myobj.evaluatePose(x)
pose = myobj.getPose()
print 'Data at Frame %d' % x
for bone in pose.bones.values():
print bone.head, bone.tail
print bone.head, bone.tail

bug #4201 fixed
one from the stoneage

providing proper padding for non 4 byte aligned rows when passing data to windows codecs

NOTE however some popular codecs like divx refuse to work with image dimensions like that
-did put a nice printf in for that.

i'll leave this open on tracker since i expect surprises with one of the zillion codecs.
unless prooven to do no harm