Brandon Williams <bmwill@xxxxxxxxxx> writes: > The main difference in v2 is that instead of piping through an index_state > struct into parse_pathspec, I ripped out the logic that needed to access the > index and either removed it completely if it wasn't needed anymore (stripping > submodule slash) or factored it out into its own function which can be called > after initializing a pathspec object (dying if a path descends into a > submodule). > > Brandon Williams (6): > pathspec: provide a more descriptive die message > submodule: add die_in_unpopulated_submodule function > pathspec: remove PATHSPEC_STRIP_SUBMODULE_SLASH_EXPENSIVE flag > ls-files: prevent prune_cache from overeagerly pruning submodules > pathspec: remove PATHSPEC_STRIP_SUBMODULE_SLASH_CHEAP > pathspec: convert find_pathspecs_matching_against_index to take an > index This was (perhaps surprisingly) quite a pleasant read, even compared to the other approach ;-).