On 03/06 01:02, Junio C Hamano wrote: > I'd expect that when that day with no scripted parts of "git > submodule" remains comes, the main entry point functions in > builtin/submodule--helper.c (like module_list(), update_clone(), > module_set_branch(), etc.) will become helper functions that live in > submodule-lib.c and would be called from builtin/submodule.c. And > the conversion would rip out calls to parse_options() in each of > these functions that would migrate to submodule-lib.c > > Side note: instead of adding submodule-lib.c, you could add them > directly to submodule.c if they are small enough. I am however > modeling after how the "diff" family was converted to C; the > diff-lib.c layer is "library-ish helpers that get pre-parsed > command line arguments and performs a single unit of work" that > utilizes service routines at the lower layer that are in diff.c > and submodule-lib.c and submodule.c will be in a similar kind of > relationship. There does exist a `submodule.c` outside of `builtin/` which has various helper functions. Will that require renaming to `submodule-lib.c`? BTW `set-branch` is a subcommand of `git submodule` so do we have to put it into `submodule-lib.c` if there were to be one? What is the motivation behind modelling it on the diff-family?
