January 29, 2015, 23:39 (GMT) |
Depsgraph: Compile fixes on mingw Seems that there were still too many headers removed, with sprintf and BLI_sprintf not available. |
January 29, 2015, 20:10 (GMT) |
Depsgraph: Removed a bit too much in the previous commit |
January 29, 2015, 15:48 (GMT) |
Depsgraph: headers cleanup Mainly avoid headers include from headers. |
January 29, 2015, 15:18 (GMT) |
Depsgraph: Fix compilation error after ogdf commit CMake should work now (ogdf stays disabled), should be no functional changes for scons. |
January 29, 2015, 13:44 (GMT) |
Depsgraph WIP: Ensure edges get added between nodes to show ownership hierarchies We can't really test that this looks right yet, since we don't have a layout algorithm working, but when this works, it will be necessary for identifying what an opnode is for. |
January 29, 2015, 13:11 (GMT) |
Depsgraph WIP: Stubbed in a few different layout engines for OGDF - still no dice |
January 29, 2015, 12:43 (GMT) |
Depsgraph WIP: Disable superfluous debug prints |
January 29, 2015, 12:32 (GMT) |
Depsgraph WIP: Removing unnecessary attributes for graph These were resulting in nodes being created with sizes of 0,0, as those flags expect that you manually set what you want. |
January 29, 2015, 12:28 (GMT) |
Depsgraph WIP: Basic support for OGDF export This introduces initial preliminary support for OGDF export, including the code to populate the graph, and all the supporting operator/rna stuff. Much of this has been copied + adapted from the code for the GraphViz exporting, though we are likely to need to adapt this later on. * Currently, there is no layout engine used, so the exported graph won't look like anything exists. But, at least all the data does get exported. * Debugging Note: Crashes related to the ogdf string class when assigning to attributes occur if you don't set the GraphAttributes flag corresponding to that attribute. |
January 29, 2015, 11:22 (GMT) |
Depsgraph WIP: Start of OGDF graph export Useful Notes: * The OGDF headers have to come before the Blender ones, or else the code will NOT compile under mingw64 (and potentially a few others). In particular, this results in many non-sensical errors in windef.h, windows.h, and winreg.h, potentially arising from redefined or already defined constants. |
January 29, 2015, 02:21 (GMT) |
Import v2012.07 of OGDF into extern libs As a quick test, this imports OGDF into the source tree (later it could be removed and added as a prebuilt static lib instead). The intention here is to use it for drawing the debug graphs for the depsgraph. Only scons is supported right now. |
January 29, 2015, 00:29 (GMT) |
Depsgraph: Compile fix for pchanmap |
January 28, 2015, 23:57 (GMT) |
Depsgraph: Tweaks to make graphviz-neato debug graphs more legible Dot will crash when graphs are generated using this option, but for neato graphs, at least we can see what the different items are this way. |
January 28, 2015, 15:56 (GMT) |
Depsgraph: Add debug print to bone done callback |
January 28, 2015, 15:06 (GMT) |
Depsgraph: Move root pchan map to own file |
January 28, 2015, 08:37 (GMT) |
Depsgraph: Fix non-initialized operation flags This was in fact root of the issue of victor rig i was having for quite some time now. |
January 28, 2015, 08:06 (GMT) |
Merge branch 'master' into depsgraph_refactor Conflicts: release/scripts/startup/bl_operators/anim.py |
January 26, 2015, 16:50 (GMT) |
Depsgraph: Add dirty check for whether bone was updated or not |
January 26, 2015, 15:02 (GMT) |
Depsgraph: Whole IK chain depends on the solver result This is the only way we can make sure all needed calculations are done to the time BKE_pose_bone_done() is called. |
January 26, 2015, 11:28 (GMT) |
Depsgraph: Use BLI_gset_add() instead of BLI_gset_insert() to avoid duplicate entries in set So it turns out that using BLI_gset_insert() will just cause the ghash-set implementation to blindly add duplicate entries if the key already exists in the hashtable - not exactly what we want, as that kindof violates the whole point of such data structures. |
|