Re: [PATCH v3 2/8] merge-recursive: internal flag to avoid touching the worktree

[Junio C Hamano <gitster@pobox.com>]

Thomas Rast <tr@thomasrast.ch> writes:

> From: Thomas Rast <trast@inf.ethz.ch>
>
> o->call_depth has a double function: a nonzero call_depth means we
> want to construct virtual merge bases, but it also means we want to
> avoid touching the worktree.  Introduce a new flag o->no_worktree to
> trigger only the latter.
>
> Signed-off-by: Thomas Rast <trast@inf.ethz.ch>
> Signed-off-by: Junio C Hamano <gitster@pobox.com>
> ---

I notice that many hits from

   $ git grep -e '->call_depth' --and --not -e '->no_worktree'

are about how the progress is reported during recursive operations
or setting up ll_opts suitable for ancestor merges (both of which
are perfectly fine not to pay any attention to no_worktree), but
some others look iffy.  For example, function remove_file() decides
to update the in-core index only when call_depth is set (i.e. we are
doing a virtual parent) or clean (clean merge at the content level,
i.e. "both removed"), and decides to update the working tree only at
the top-level of the recursion and no_wd is passed.

 - As to "update_cache", if you do not update it while you are
   operating in the cache-only mode (aka ->no_worktree), I wonder
   where the result goes.  Shouldn't it be done for in-core merge as
   well?

 - As to "update_working_tree", there are few places where the
   function is called with no_wd that is not true, even when
   ->no_worktree is set.  Do you want to allow working tree to be
   modified in such a call?