February 27, 2015, 10:08 (GMT) |
Merge branch 'master' into depsgraph_refactor |
February 26, 2015, 16:17 (GMT) |
Depsgraph: Fix bad memory access in animation for proxy armatures |
February 26, 2015, 14:37 (GMT) |
Depsgraph: Code cleanup, wrong indentation |
February 26, 2015, 12:42 (GMT) |
Depsgraph: Armature object didn't reset recalc flags on object after full evaluation |
February 26, 2015, 09:26 (GMT) |
Merge branch 'master' into depsgraph_refactor |
February 25, 2015, 19:42 (GMT) |
Depsgraph: Code cleanup, comments |
February 25, 2015, 19:39 (GMT) |
Depsgraph: Fix update not happening correct in certain circumstances Because of the lack of real incremental updates it was possible that incorrect object matrix was used by the constraint stack. Now flush update will tag all operations within component for update, solving the issue with lack of incremental updates. It's not totally optimal implementation atm, but good enough for further tests and improvements. |
February 25, 2015, 17:41 (GMT) |
Depsgraph: Code cleanup |
February 25, 2015, 15:45 (GMT) |
Depsgraph: Correction to previous commit Update for the new frame happens a bit different and previous commit broke animation. |
February 25, 2015, 13:51 (GMT) |
Depsgraph: Make sure recalc is flushed before calling scene update handlers |
February 25, 2015, 13:20 (GMT) |
Depsgraph: Fix missing updates after recent visibility changes After some additional thoughts it's now not totally clear for me if we really want flush to happen on invisible objects or not. Will think about it in more details.. |
February 25, 2015, 10:13 (GMT) |
Depsgraph: Make sure changing DOF keeps valid depsgraph Not sure how it worked in master branch actually, maybe we need to do something like this there as well. |
February 25, 2015, 09:54 (GMT) |
Depsgraph: Correction for camera DOF relation For now it's unclear how to handle it on the camera datablock level, so moved parameters to the camera object level. This kinda of makes sense because for DOF you need to know object matrix and having DOF decoupled from it is kinda weird.. IN any case it's always easy to make it camera data relation and add relation between camera datablock and camera object. |
February 25, 2015, 09:44 (GMT) |
Depsgraph: Make error print a bit more verbose |
February 25, 2015, 08:58 (GMT) |
Merge branch 'master' into depsgraph_refactor |
February 23, 2015, 18:39 (GMT) |
Depsgraph: Fix for scene update might have been using wrong time |
February 23, 2015, 15:57 (GMT) |
Depsgraph: Rework the way how layer visibility are handled The main goals of this commit are: - Get rid of list of IDs which were tagged for update but are not visible - Make it possible to have proper time dependency re-tags of invisible objects This change is really quite the same as previous one: just uses ob->recalc flags to check whether object was updated or not. It might be a bit of an issue with materials, but they're always "visible", so should not cause any missing updates. In any case, we really need to get rid of all this temp sets and arrays, they might become quite huge in production files. |
February 23, 2015, 14:15 (GMT) |
Depsgraph: Changes in the way how tagging objects and relations for update are working together The idea is to get rid of rather obscure set which stored missing IDs for delayed update after the relations are up-to-date. New idea is to re-use the ID recalc flags and do needed re-scheduling during depsgraph build. This is combined together with layers visibility check so would not add extra slowdown or so. This change also makes it so ob->recalc flags are properly set and reset, which was discovered needed for the animation paths update and likely for other baking-like areas which does manual scene update. |
February 23, 2015, 13:22 (GMT) |
Depsgraph: Fix issue with missing materials update Now when materials are the parts of dependency graph we need to ensure relations are all correct when affecting material slots. |
February 23, 2015, 13:09 (GMT) |
Merge branch 'master' into depsgraph_refactor |
|