Differential D9237
Add a callback to `IDTypeInfo` to allow preservation of some data accross memfile undos Authored by Bastien Montagne (mont29) on Oct 16 2020, 5:23 PM.
Details This is essentially adding that new callback, and using it only for already existing Scene's 3DCursor. Note that the place where this is called has been moved again, after all have been lib-linked, such that those callbacks may also work on ID pointers.
Diff Detail
Event TimelineComment Actions This will be a big usability improvement. Though I'm not sure this patch is enough to handle all cases:
In both cases it's probably reasonable for the pointers to get lost, but we should avoid crashing.
Comment Actions For this to work reliably we will need a way to visit and validate pointers from/to brushes on each undo step. A down side for clearing the pointers is it means undoing (back to the initial state), then redoing can leave you with cleared ID pointers.
Comment Actions We could undo/redo the datablock pointers, but leave all other fields in brushes and tool settings unchanged? This could be implemented by cleaing invalid datablock pointers automatically, and (manually?) implementing a function to copy over datablock pointers from the undo-restored brush or tool settings where it makes sense. At least for objects, materials, images, and scenes this is probably good. Palette datablock should be handled like brushes, and not be part of scene data. Textures are the more complicated case, where they can be either scene data or purely brush/UI data, not sure there is a great solution in the current design. Probably you also want to copy them over. Comment Actions In fact, I wonder if it's that important to keep brushes themselves out of undo? Most of the issues users are complaining about are actually related to tool settings, not brushes themselves? Now that I think about it a bit more, am somewhat wary about removing more ID types from undo process tbh. So indeed maybe have a util function to only swap (i.e. keep) regular data from old datablocks into new ones accross undo would be a much cleaner and future-proof solution (would even add it as a callback to IDTypeInfo structure itself. That would also make that 'hack' a bit more obvious/easy to find in the code, instead of hiding it in a specific function in readfile.c. Comment Actions Agree syncing non-pointer data seems a lot safer than bypassing non-screen ID's from undo. The down-sides with this are it could be a tedious to add the callbacks that sync data, but that's a lot less hassle than properly dealing with brushes/textures/palettes/images being added/removed. | ||||||||||||||||