Differential D10695
Skip undo steps when changing properties of the 3d cursor through the UI Authored by Germano Cavalcante (mano-wii) on Mar 11 2021, 5:12 PM. Tags None Subscribers None
Details Reinvestigating T57100, I realized that, even though the cursor values This is bad for obvious reasons: overhead, memory, useless Ctrl+Z... This patch proposes to untag the View3DCursor RNA struct as not having
Diff Detail
Event TimelineComment Actions I would not go that way, starting to not write undo steps even though we edit some undo-affected data is open gate for issues with modern undo system, which is much, much less forgiving to inconsistencies. Let's not open new potential vulnerabilities here, we already have enough to actually fix in undo code, that kind of micro-optimization is really not worth the risk currently. Comment Actions Understandable, we don't want to risk new gates for bugs. Not adding undo-steps would match the operator that changes the 3D cursor position when clicking in the 3D view, as it does not add undo-steps. This problem is similar to the operator that moves the 3D cursor with hold and drag (bpy.ops.transform.traslate(....)) as it creates problematic undo steps. Comment Actions @Germano Cavalcante (mano-wii) aaah... yeah I might have been a bit quick on this one, indeed since the cursor3d operator itself does not store undo steps, we can probably do the same for the RNA access to this data. note that your patch should be committed in two steps though, the fix in interface.c should go first, before making RNA 3Dcursor struct non-undoable. |