On Mon, Aug 27, 2018 at 7:32 PM Stefan Beller <sbel...@google.com> wrote: > > Besides some small conflicts on 'pu', like the previous part, it also > > breaks 'pu' because of API changes. The fix is trivial though, just > > prepend the_repository as the first argument for the broken function > > calls. > > This sounds like a problem. I said the same when sending the > object store lookup series, which ended via > 3a2a1dc1707 (Merge branch 'sb/object-store-lookup', 2018-08-02) > in master, but I do recall Junio being somewhat unhappy about it[1]. > > By just adding the argument to the function, it might merge easily > but not compile, which would have to be fixed up; and that object store > series seemed to touch a lot of functions that were also used in series > in-flight.
Which is why I state it clearly here so Junio could choose not to pick this series up (I'm totally ok with that, I could wait until the dust settles and send again). Another option is me adding a patch that renames diff_setup() to diff_setup2() or something that takes 'struct repository *', leaving diff_setup() as a light wrapper around diff_setup2(). I would need to wait until in-flight series are merged, then rename diff_setup2() back to diff_setup(). Not sure if it's worth doing. > > diff.c | 259 +++++++++++++++++++++++------------------ > > Ugh? That sounds like there is an interesting change coming. Yeah. sequencer.c also needs struct repository in lots of places (sha1-name.c comes second). But at least builtin/ code looks a lot like it's supposed to be when 'struct repository *' is introduced, we now have lots of function calls there that pass the_repository... -- Duy
