- convertblender.c, remove assignments to unused vars.
- readfile.c, fix 2 possible crashes. null pointers were being checked for then used later without checking.
- space_graph.c, use switch statement for automatic color assignment rather then a float array.
This commit partially fixes the problems with Shapekeys from older
files, as seen from the Regression suite (relative.blend and
dolphin.blend in particular).
In older files, keyblock->slidermax was never truly set to 1.0 even
though the UI may have shown such a value (which was bizzarely being
sourced from somewhere else). Hence, after loading the files in 2.5,
the shapekeys wouldn't animate, as the value would get clamped between
0 and 0.
To fix this, I've added a version patch which corrects these
situations in old files, and I've adjusted the slider-RNA code so that
it is not possible to set up such clamping anymore.
TODO:
The fixes detailed here only make it possible for these files to work
again in 2.5. However, I haven't been able to find a way to get the
files to actually work in 2.5 without manually changing the active
shapekey (per object) after loading the files with these patches
applied. Possibly it's just some depsgraph magic needed, unless
there's still some other evil voodoo in the shapekey code
* Argh my bad, sorry about this!
* Now only the actual data array is saved to avoid constant re-allocations, but no relations to active data are kept.
* Also reverted Ton's quick fix for the crash as it's not needed anymore.
Serious *bad* crash in undo introduced by commit Janne dec 21st.
Time window now stores some kind of cache for fluids/cloth, but
it's pointing (in SpaceTime) to data inside Objects. (Not ID).
That's really not allowed... this commit fixes crashes but the
cache code really needs to be redesigned. I'm also afraid this
crash is going to frustrate everyone using physics...
Child-of constraint issue: on adding, it wasn't checking owner
correctly for Bones, resulting in a constraint working in wrong
space; it looked as if transform was applied double when moving
the object. Only adding via Py API went wrong btw.
Also found a silly check for drawing constraints, which caused
constraint initialization to happen for every object on every
redraw!
Implementation note: con->flag CONSTRAINT_SPACEONCE was only used
for child-of constraints in Bones, so I've patched it on file
reading to always set the flag. Marked with XXX, so it can be
removed one day. Now at least things get corrected well for
imported armatures.
Makes adding new flags give ambiguous results and also makes it less easy to tell whats intended.
In some places it looks like OB_RECALC_TIME should be left out too.
* New option to "Regrow hair" for each frame.
* This was perhaps more a feature request, but there was a similar useful feature called "animated hair" in particles at some point.
* The previous behavior for hair growing was inconsistent to say the least, so this is a nice option to have.
This allows manual (point by point) animation of their control verts,
although many other settings cannot really be animated with any
visible effects yet. Interestingly, lattices also had IPO block
pointers, though they were never really used (AFAIK).
Todo:
- Animation Editor support has yet to be added. I've got a few other
things to add to, so will group those changes together.
- Object actuator rotation was being scaled by: (1/0.02)*2*PI/360 == 0.872, since revision 2.
- Remove scaling and use do_versions to adjust existing files.
- skip fixing file paths on undo.
- simplify bpath alloc and free functions, also pass Main structure so as not to rely on G.main, (needed for file load).
only tags the ID and does the actual flush/update delayed, before the next
redraw. For objects the update was already delayed, just flushing wasn't
yet.
This should help performance in python and animation editors, by making
calls to RNA property update quicker. Still need to add calls in a few
places where this was previously avoided due to bad performance.
Issue with Library linking: if you already have libraries linked in,
and you attempt to link more data, using internally already linked
stuff (like a group with group links) the linkage would fail. It did
resolve on save-load though.
There's a long commented history in this code explaining it. I found
out it was because of manual step-by-step linking versus normal file
reads. In the first case, the "lib pointer map" was missing references.
Particle cache reading: crash when loading .blend on a different
endian system, code was dumping arrays in .blend without DNA.
General warning for devs: avoid generic write_data and dynamic
arrays in DNA.
* Separate caches were causing quite a lot of problems both in principle and practice.
* For example it doesn't really make sense to have different frame ranges for normal and high resolution smoke, but this was fully possible before.
* Also to fully bake the smoke you had to do a "Bake All Dynamics", which completely defeats the whole point of the feature!
* As a result of this change the smoke cache usage is much much simpler and less error prone.
* This is quite a big change, but hopefully there should be less rather than more problems as a result :)
Some other related changes:
* Changing the cache name now works for disk caches properly too, it now just renames the cache files so should be faster too!
* Smoke is now always forced to disk cache with step 1 on file load as there were some strange cases where smoke was trying to use memory cache.
* Disabled smoke debug prints from console.
* Disabled changing smoke parameters when smoke is baked.
Note to users: The unfortunate side effect of this is that old high resolution simulations have to be baked again, but in return you get much better and more logical functionality. Sorry none the less!
- Reading 2.4x files could cause print "missing region type".
Appeared the 2.50 do_versions patch differed from default region
definitions
- Sometimes editors showed wrong button for type browsing.
Was because variable wasn't correctly reset on saves.
Also use const char in many other parts of blenders code.
Currently this gives warnings for setting operator id, label and description since these are an exception and allocated beforehand.
- incorrect range check broke ZYX euler rotations, use MIN/MAX constants so this doesn't happen again.
- BGE Armature PyAPI also wasn't using correct min/max with rotation modes.
- clamp on file read rather then when calling the rotation functions, so developers don't use invalid args without realizing it.
- added assert() checks for debug builds so invalid axis constants don't slip through.
Ton will work on moving the File menu out of the Info space before release.
notes.
- reply Operator isn't working anymore.
- UI for reports is commented out so its not mixed with the file menu.
Issue: in user preferences window, using file selecting caused the
the userpref window to be saved, and not closing.
Reason: design error (by me) in using screen->full tag for denoting
a temporarily screen (like file window). Fixed by using a new
screen->temp variable for it.
System remained unstable though, noticed another issue with freeing
temp screens in wrong places. Seems nice stable now! Will check on
the wiki for relarted issues now.
- use sizeof() in more places.
- fixed some off by 1 bugs copying strings. setting curve font family for instance was 1 char too short.
- replace strncpy and strcpy with BLI_strncpy
Now the active strip doesn't just get cleared on fileload, but is relinked properly.
I had originally intended that files shouldn't be able to be saved with NLA data still in Tweakmode, but this turns out to be a bit more troublesome to get working as that would make undo keep popping out of this mode too.
Also reverting 32743 (bugfix for 24418), which was a hack around this.
fix included in report from Martijn Berger (mberger)
made some small changes.
- use ints rather then unsigned long for printing, values are not likely to be very large.
- CMake remove strict flags from collada build dir since I had warnings in the collada headers.
- added xml2 to collada libraries else I couldnt get collada building.
the brush system matches the brush mode with the object mode, but this doesn't work for 2D image view paint.
since the poll() function doesnt have access to the context, for now just check if no paint modes are active, default to texture paint.
