"brian m. carlson" <sand...@crustytoothpaste.net> writes: > This is the tenth in a series of patches to convert from unsigned char > [20] to struct object_id. This series mostly involves changes to the > refs code. After these changes, there are almost no references to > unsigned char in the main refs code. > > The series has not been rebased on master since the last submission, but > I can do so if that's more convenient.
Thanks for working on this. I took a look at the conflicting parts between what happened on 'master' since than and this series. Two things stood out: - many calls to resolve_refdup() started passing NULL when the caller does not care the object name (they just want to know what ref the symref points at, if any). - the implementation of ref peeling has been updated quite a bit. In addition, builtin/pack-objects.c will lose want_object_in_pack() and a new function ignore_object() will be introduced to replace its role in a topic in flight. Having said all that, I think I managed to come up with a clean conflict resolutions to all integration branches 'master', 'next', 'jch' and 'pu' (that is, I pretend that the topic is mergeable to these branches immediately, make a trial merge and test the result before teaching the resolution to my rerere database, even though the trial merges to more stable branches than the branch the topic will start cooking in would be immediately discarded); at least the results of these all passed the tests, so this round that is based on a tad older version of 'master' looks alright.
