July 29, 2016, 20:19 (GMT) |
Add GMock library which is responsive for dealing with mock objects |
July 29, 2016, 20:19 (GMT) |
GTests: Update to a newer Git version Required to get GMock working with GTest. |
July 29, 2016, 18:42 (GMT) |
Fix missing redraw of timeline when changing GPencil layer settings |
July 29, 2016, 16:49 (GMT) |
Cleanup space/tabs, Grrr. |
July 29, 2016, 16:47 (GMT) |
Fix a bunch of missing expand calls in object/particle physics area, was breaking linking of those. Initial report/patch by Alexander Gavrilov (@angavrilov), found more on the road. Nice demo of why we should use libquery ID looper in read code too - but that's for another day (also because read code needs to take care of some deprecated pointers sometimes)... |
July 29, 2016, 15:30 (GMT) |
Cycles: Fix wrong projection coordinates for points behind the camera Points behind camera will be flipped on projection, but that did not happen. This solves second issue reported at T48848. |
July 29, 2016, 15:30 (GMT) |
Cycles: Fix wrong check for object is behind camera after recent fixes |
July 29, 2016, 15:00 (GMT) |
Fix T48971: Append creates linked image textures if object has shape keys. Hating all those not-so-real ID types... Here there were two causes for the issue: 1) Linked shapekey ID was made local twice (once from mesh's make local, once by itself). Solved by not explicitely making shapekeys (nor any other non-linkable datatype) local. 2) Key->from 'back pointer' to its owner was messing 'still in used' detection of linked data after localization. Fixed with a hack for now, thinking correct solution might actually be to not consider this pointer at all in libquery ID looper, since it's nothing like and actual usage of mesh/lattice/curve. Again, shapekeys as ID is a joke, those should be mere struct, they have absolutely nothing to do in Main, period. :( Point 2) still demonstrates the need for better handling of IDs dependencies though, so far we only hit corner cases, but think there could also be valid cases generating those 'dependency cycles' between IDs (ID a using ID b which uses ID a), this will have to be addressed some day... |
July 29, 2016, 14:56 (GMT) |
WIP: correspondence update and drawing functions |
July 29, 2016, 14:35 (GMT) |
Only generate hair contacts for points that actually penetrate the mesh. |
July 29, 2016, 14:35 (GMT) |
Clear debug drawing when creating new contacts. |
July 29, 2016, 14:35 (GMT) |
Enable debug drawing of hair contacts. |
July 29, 2016, 14:35 (GMT) |
Simple contact detection method for hair editing, using the typical BVH closest point method. This method is just a placeholder, in the future it would be worthwhile to use Bullet collision detection with ghost objects for this. Also the code should eventually be abstracted a bit to make it agnostic to mesh formats, so it can be used for BMesh as well as DNA strand data or physics simulation. |
July 29, 2016, 14:35 (GMT) |
Hair edit tool settings for enabling and controlling deflection. Unused settings, deflection is not yet implemented. |
July 29, 2016, 14:35 (GMT) |
Add a cache structure for collision contact points. This should serve as a universal interface between possible collision detection systems (BVH, Bullet) and the various collision response systems. Using an new data structure means that the two sides of a collision handling feature (detection and response) can be treated independently, reducing code dependencies and complexity. The properties of contact points are modeled after btManifoldPoint from Bullet and may need to be adjusted in the future. |
July 29, 2016, 14:34 (GMT) |
Fix for frequent unnecessary shader recompilation during hair editing. The depsgraph component node for invalidating the strand shader was located in the GEOMETRY component of the object. This component by default gets tagged along with every other object part during editing. Only standardized way of avoiding this would be putting it in a different datablock, but the shader is very much tied to the particular object due to the displacement function from its node tree. For now the solution is a new component GEOMETRY_SHADER, which does not get tagged when the ID node is tagged. It still gets regular updates from node tree changes. This seems to be the least invasive way of avoiding unnecessary performance drops, and can be removed and replaced easily later on without disrupting other code work. |
July 29, 2016, 13:12 (GMT) |
Remove "debug build only" event printing. This makes no sense since user usually only has release build... And those noisy printings are behind debug flag anyway, so really no reason to only print them out in debug build. |
July 29, 2016, 13:06 (GMT) |
Cycles: Fix strict compilation warning |
July 29, 2016, 12:47 (GMT) |
Fix T48679: Bone transform only alters between translation and rotation There was some smart code in the transform which would alter between translation and rotation based on whether bone is connected or not and whether translation is locked or not. It could be handy to also fallback to scale if both rotation and translation are locked. |
July 29, 2016, 12:46 (GMT) |
Merge remote-tracking branch 'origin/master' into soc-2016-uv_tools |
|
|
|


Master Commits
MiikaHweb | 2003-2021