The existing code was RNA, so I thought it better to stick to that...

Hmm, thinking about this a bit more, it would probably best to change these nodes too, for consistency:

source/blender/nodes/shader/nodes/node_shader_normal_map.cc:
  26:     PointerRNA obptr = CTX_data_pointer_get(C, "active_object");

source/blender/nodes/shader/nodes/node_shader_tangent.cc:
  29:     PointerRNA obptr = CTX_data_pointer_get(C, "active_object");

source/blender/nodes/shader/nodes/node_shader_uvmap.cc:
  25:     PointerRNA obptr = CTX_data_pointer_get(C, "active_object");

source/blender/nodes/shader/nodes/node_shader_vertex_color.cc:
  21:   PointerRNA obptr = CTX_data_pointer_get(C, "active_object");

Added 3.5 tag, because it's something which makes the UV as attributes much more useful in geometry nodes.

In D17093#466543, @Brecht Van Lommel (brecht) wrote:

I don't think the active object is guaranteed to get evaluated by the depsgraph? I think it should fall back to the original object in that case.

I tried to trace through DEG_get_evaluated_id() and as far as I can see it just returns the original id if the depsgraph doesn't return an evaluated id. Since PointerRNA dataptr is always of the OB_MESH type DEG_get_evaluated_rna_pointer() will always use that function (DEG_get_evaluated_id() ).

So I'd think it is not needed?

I'll check it again extra thoroughly tomorrow when a bit more awake. But if it indeed works as described above, do you agree a fallback isn't needed?

Looked it over once again, and the fallback to the original id is already present in blender::deg_get_evaluated_id(...) so I don't think it's needed here.