diff options
Diffstat (limited to 'source')
-rw-r--r-- | source/blender/blenkernel/intern/anim_data.c | 17 |
1 files changed, 15 insertions, 2 deletions
diff --git a/source/blender/blenkernel/intern/anim_data.c b/source/blender/blenkernel/intern/anim_data.c index fdc23ac50af..c26fc6578f1 100644 --- a/source/blender/blenkernel/intern/anim_data.c +++ b/source/blender/blenkernel/intern/anim_data.c @@ -357,10 +357,23 @@ AnimData *BKE_animdata_copy(Main *bmain, AnimData *adt, const int flag) /* make a copy of action - at worst, user has to delete copies... */ if (do_action) { + /* Recursive copy of 'real' IDs is a bit hairy. Even if do not want to deal with usercount + * when copying ID's data itself, we still need to do so with sub-IDs, since those will not be + * handled by later 'update usercounts of used IDs' code as used e.g. at end of + * BKE_id_copy_ex(). + * So in case we do copy the ID and its sub-IDs in bmain, silence the 'no usercount' flag for + * the sub-IDs copying. + * Note: This is a bit weak, as usually when it comes to recursive ID copy. Should work for + * now, but we may have to revisit this at some point and add a proper extra flag to deal with + * that situation. Or refactor completely the way we handle such recursion, by flattening it + * e.g. */ + const int id_copy_flag = (flag & LIB_ID_CREATE_NO_MAIN) == 0 ? + flag & ~LIB_ID_CREATE_NO_USER_REFCOUNT : + flag; BLI_assert(bmain != NULL); BLI_assert(dadt->action == NULL || dadt->action != dadt->tmpact); - BKE_id_copy_ex(bmain, (ID *)dadt->action, (ID **)&dadt->action, flag); - BKE_id_copy_ex(bmain, (ID *)dadt->tmpact, (ID **)&dadt->tmpact, flag); + BKE_id_copy_ex(bmain, (ID *)dadt->action, (ID **)&dadt->action, id_copy_flag); + BKE_id_copy_ex(bmain, (ID *)dadt->tmpact, (ID **)&dadt->tmpact, id_copy_flag); } else if (do_id_user) { id_us_plus((ID *)dadt->action); |