On Mon, Jun 11, 2018 at 6:05 PM Duy Nguyen <pclouds@xxxxxxxxx> wrote: > > On Sat, Jun 9, 2018 at 9:58 PM Elijah Newren <newren@xxxxxxxxx> wrote: > > I read over the rest. Found a small grammatical error in a commit > > message. Found multiple places that still need conversion, from > > pushing up &the_index usages to callers of ll-merge.c and sha1-file.c > > instead of having them in those files, to mixes of _cache_ and _index_ > > functions as in apply.c and merge-recursive.c. However, Duy pointed > > out there was more work to do, > > Yes. This is just fyi, 40 patches later, ... Junio, just to be clear, I think I'll withdraw this 23-patch series (reviews are still welcome though). It does fix some potential bugs but it's not that critical. This will let me merge it with the other 40+ patches and reorganize better (after all this started out as a single rfc patch, I didn't realize what I got myself into) > i'm down to leaving the_index > in three files outside builtin/: merge-recursive, notes-merge.c and > transport.c. Even after the conversion we may need some more follow-up > patches because it now shows places where we should _not_ touch the > index at all, which may involve not simply passing NULL index_state to > some functions, but fixing them up to tolerate NULL index_state. So > it's going to be a few patch series until the_index is gone for good > [1]. > > [1] but like cheap horror movies, there's always a sequel: > the_repository is still spread in many places and hides dependencies > in the same way. We can't do anything about it though until struct > repository conversion is more or less complete. > -- > Duy -- Duy
