Blender Git Loki

Merge branch 'id-ensure-unique-memory-address' into undo-experiments

Merge branch 'readfile-rework-refcount-handling' into id-ensure-unique-memory-address

Cleanup: remove timing debug calls.

We now know that extra time spent here is neglectable.

Fix crash on undo case.

We cannot use libquery too early in undo case, because of all the black
magic we perform over our data-blocks there...

Merge branch 'id-ensure-unique-memory-address' into undo-experiments

Merge branch 'readfile-rework-refcount-handling' into id-ensure-unique-memory-address

Conflicts:
source/blender/blenkernel/intern/main.c

Merge branch 'libquery-handle-ui-pointers' into readfile-rework-refcount-handling

Merge branch 'master' into libquery-handle-ui-pointers

Conflicts:
source/blender/blenkernel/intern/lib_query.c

Merge branch 'libquery-handle-ui-pointers' into readfile-rework-refcount-handling

Merge branch 'master' into libquery-handle-ui-pointers

Merge branch 'master' into libquery-handle-ui-pointers

Merge branch 'undo-experiments' into undo-experiments-swap-reread-datablocks

Fix previous commit.

undoexp: Update for new 'ID memaddress unique' feature.

Merge branch 'undo-experiments' into undo-experiments-swap-reread-datablocks

undoexp: Update for 'unique ID memaddress' change.

Merge branch 'id-ensure-unique-memory-address' into undo-experiments

Initial rough idea of 'unique ID address' change.

Goal of this code is to ensure we only ever use a given memory address
once, for all IDs generated in a bmain.

This is crucial if we want to be able to use ID pointers instead of ID
names in undo speedup work, when re-using existing IDs.

undoexp: remove second libmap for re-used existing datablock addresses.

These should not be needed anymore, since they were used by re-used
data-blocks, which are not lib-linked anymore.

undoexp: do not liblink re-used datablocks anymore.

Since all existing IDs are supposed to use their same memory address,
unchanged data-blocks should not have any changed pointer to other IDs,
so no need to liblink them at all.

Note that am not 100% sure that is fully valid currently, as we use ID
names to find old IDs, which is not 100% safe and secure solution...