seems reasonable, but seems to indicate that we'd want a test for record_entry_contents (or a larger function, up to commit), checking that it gets the canonical size from the tree, not the convenient size. Even if this interface is renamed, there'd still be a chance commit was using the wrong one and we ought to test it.
John's patch
http:// bazaar. launchpad. net/~jameinel/ bzr/1.17- content- filtering- commit/ revision/ 4530
seems reasonable, but seems to indicate that we'd want a test for record_ entry_contents (or a larger function, up to commit), checking that it gets the canonical size from the tree, not the convenient size. Even if this interface is renamed, there'd still be a chance commit was using the wrong one and we ought to test it.