It might be useful to use Data::Dumper as text output for each comparable data structure, and do an actual unix-diff on each. This has the limitations of 1) treating both objects and code references as opaque, 2) relying on a unix command.