Jeppe Øland <joland@xxxxxxxxx> writes: > I'm trying to use sparse checkouts in a large project, and one oddity > (bug?) came up. > > If I do a normal full clone of a repository, then do: > git worktree add ../sparse --no-checkout > cd ../sparse > git sparse-checkout set somepath > git checkout > Now indeed it is a sparse worktree, but the files from the root > directory are also there. I think this is working as intended. Without knowing about the non-tree entries' object names, you cannot create your next commit on top, as you need to be able to compute the tree object's contents. You'd update some files in "somepath", and that directory being populated you can compute the object name of the tree that represents the state of "somepath" after your update. But that object name must be mixed with other things in the same directory as "somepath" (i.e. in the root tree), so they must be known.
