Closed fdagenais-cinesite closed 3 years ago
Thank you for you comments @andrewkaufman.
My changes were blindly based on the hash
methods, and as you pointed out a few of the plugs should have been removed from the hash instead of being added to the affect
method. In the latest commit, I changed both the affect
and hash
methods based on the plugs used by their respective compute
methods.
The
affects()
methods of theAtomsCrowdGenerator
andAtomsCrowdClothReader
nodes have been fixed by adding missing input plugs and a call to the base class inAtomsCrowdClothReader
. InAtomsCrowdGenerator
, most of the logic has been moved to the matchingBranchCreator::affectsBranch...()
method, since they are internally used byBranchCreator
to dirty its internal plugs.In Gaffer 0.58+, changes in the dirtying mechanism makes it important to have the behavior of
affects()
methods match the affects relationships of the hash/compute methods. Relationships not defined inaffects()
result in a failure to update the right outputs when an input plug is changed.