Open mikibonacci opened 3 months ago
Interesting, indeed I don't know how it generates the hash. Maybe we can simply store an attribute in the constructor, but isn't it an issue of aiida-core that does not distinguish two different class entry-points?
This is indeed a bug in aiida-core
: https://github.com/aiidateam/aiida-core/issues/6320
It seems that the caching mechanism does not distinguish HubbardStructureData from standard StructureData. The very same simulation but changing the structure to be HubbardStructureData is considered identical to the original one, and so if caching is enabled this causes problems.
I guess there is just a check on the attributes of StructureData.
Pinning @bastonero and @mbercx: do you have comments on this? is it a known issue?