Re: [PATCH] worktree: avoid dead-code in conditional

From: Eric Sunshine <sunshine@sunshineco.com>
Date: Jun 24, 2020, 23:00 UTC
Message-ID: <CAPig+cRdRyWAjU7QADPt39rDgJ4njDtN_pwo8JMVD=YX09a-Vg@mail.gmail.com>
In-Reply-To: <xmqqsgek5fg3.fsf@gitster.c.googlers.com>

On Wed, Jun 24, 2020 at 5:20 PM Junio C Hamano <gitster@pobox.com> wrote:

Show 13 quoted lines

> Eric Sunshine <sunshine@sunshineco.com> writes:
> > get_worktrees() retrieves a list of all worktrees associated with a
> > repository, including the main worktree. The location of the main
> > worktree is determined by get_main_worktree() which needs to handle
> > three distinct cases for the main worktree after absolute-path
> > conversion:
> >
> >   * <bare-repository>/.
> >   * <main-worktree>/.git/. (when $CWD is .git)
> >   * <main-worktree>/.git (when $CWD is any worktree)
>
> It is unclear from the above but I would assume that you are talking
> about the returned path from get_git_common_dir().

No, I am talking about the result of strbuf_add_absolute_path() applied to get_git_common_dir(). I intentionally explicitly stated "after absolute-path conversion" in the commit message to be clear about this, but perhaps I should have mentioned strbuf_add_absolute_path() by name to avoid confusion.

> I can certainly understand why there needs two distinct cases
> (i.e.. bare vs non-bare), but why is this codepath (or any caller of
> get_git_common_dir()) forced to care about the two cases?

Callers of get_git_common_dir() aren't forced to care. But after applying strbuf_add_absolute_path() we are forced to care. This is the result of get_git_common_dir() for the three cases:

    .     (within bare repo)
    .git  (within any worktree)
    .     (within .git)

After applying strbuf_add_absolute_path(), we get:

    /whatever/proj.git/.   (within bare proj)
    /whatever/proj/.git    (within any worktree)
    /whatever/proj/.git/.  (within .git)

These need to be normalized, respectively, to:

    /whatever/proj.git  (within bare proj)
    /whatever/proj      (within any worktree)
    /whatever/proj      (within .git)

Show 8 quoted lines

> I wonder if the right "fix" to this instance, at the same time
> preventing similar breakages in the future, is rather make sure
> get_git_common_dir() not to return the redundant path with ".git/."
> suffixed? For that matter, I do not know why the bare case must
> need "/." suffix. There seem to be about a dozen callers of the
> function, but don't some of them share a similar issue?
>
> Puzzled.

Your puzzlement may arise from the misunderstanding regarding get_git_common_dir() vs. strbuf_add_absolute_path()?

It might be nice to refine strbuf_add_absolute_path() to return more aesthetically pleasing results (i.e. not add trailing "." when unneeded), but such a change is outside the scope of this small patch and may have repercussions which could be a time sink (and I haven't the time to devote to it).

Show 19 quoted lines

> Let's look at the other two grep hits from worktree.c
>
>     strbuf_reset(&path);
>     strbuf_addf(&path, "%s/worktrees/%s/HEAD", get_git_common_dir(), id);
>
>     worktree = xcalloc(1, sizeof(*worktree));
>     worktree->path = strbuf_detach(&worktree_path, NULL);
>     worktree->id = xstrdup(id);
>     add_head_info(worktree);
>
> done:
>     strbuf_release(&path);
>     strbuf_release(&worktree_path);
>     return worktree;
> }
>
> This looks somewhat bogus. "sturct strbuf path" is populated, but
> is released without ever getting used, isn't it? Am I grossly
> misreading the code?

I think you're reading it just fine; that code is indeed dead. This appears to be fallout from 5193490442 (worktree: add a function to get worktree details, 2015-10-08), which unfortunately (I think) got little or no review[1].

This could indeed use a "dead-code removal" patch of its own, though it's unrelated to the current patch.

FOOTNOTES

[1] I had reviewed the very first version of the patch series which
    introduced these changes, but then effectively went offline for
    many, many months, thus was unable to review any of the re-rolls.