Re: [PATCH v3 2/2] worktree add: add --orphan flag

[Jacob Abel <jacobabel@nullpo.dev>]

On 22/11/15 11:09PM, Ævar Arnfjörð Bjarmason wrote:
>
> On Thu, Nov 10 2022, Jacob Abel wrote:
>
> So, on a second read-through...
>
> >  'git worktree add' [-f] [--detach] [--checkout] [--lock [--reason <string>]]
> > -		   [[-b | -B] <new-branch>] <path> [<commit-ish>]
> > +		   [[-b | -B | --orphan] <new-branch>] <path> [<commit-ish>]
>
> This synopsis is now at least partially wrong, and ....
>
> > +--orphan <new-branch>::
> > +	With `add`, create a new orphan branch named `<new-branch>` in the new
> > +	worktree. See `--orphan` in linkgit:git-switch[1] for details.
> > +
> >  --porcelain::
> > ....
> >  #define BUILTIN_WORKTREE_ADD_USAGE \
> >  	N_("git worktree add [-f] [--detach] [--checkout] [--lock [--reason <string>]]\n" \
> > -	   "                 [[-b | -B] <new-branch>] <path> [<commit-ish>]")
> > +	   "                 [[-b | -B | --orphan] <new-branch>] <path> [<commit-ish>]")
>
>
> ...here we say the same, but surely it's only:
>
> 	git worktree add --orphan new-branch /tmp/orphan
>
> And not e.g.:
>
> 	git worktree add --orphan new-branch /tmp/orphan origin/next
>
> Or whatever, but it's incompatible with <commit-ish>. I think this on
> top should fix it up:
>
> 	diff --git a/Documentation/git-worktree.txt b/Documentation/git-worktree.txt
> 	index 1310bfb564f..3afef985154 100644
> 	--- a/Documentation/git-worktree.txt
> 	+++ b/Documentation/git-worktree.txt
> 	@@ -10,7 +10,9 @@ SYNOPSIS
> 	 --------
> 	 [verse]
> 	 'git worktree add' [-f] [--detach] [--checkout] [--lock [--reason <string>]]
> 	-		   [[-b | -B | --orphan] <new-branch>] <path> [<commit-ish>]
> 	+		   [[-b | -B] <new-branch>] <path> [<commit-ish>]
> 	+'git worktree add' [-f] [--lock [--reason <string>]]
> 	+		   --orphan <new-branch> <path>
> 	 'git worktree list' [-v | --porcelain [-z]]
> 	 'git worktree lock' [--reason <string>] <worktree>
> 	 'git worktree move' <worktree> <new-path>
> 	diff --git a/builtin/worktree.c b/builtin/worktree.c
> 	index 71786b72f6b..2b811630b3a 100644
> 	--- a/builtin/worktree.c
> 	+++ b/builtin/worktree.c
> 	@@ -17,7 +17,10 @@
>
> 	 #define BUILTIN_WORKTREE_ADD_USAGE \
> 	 	N_("git worktree add [-f] [--detach] [--checkout] [--lock [--reason <string>]]\n" \
> 	-	   "                 [[-b | -B | --orphan] <new-branch>] <path> [<commit-ish>]")
> 	+	   "                 [[-b | -B] <new-branch>] <path> [<commit-ish>]"), \
> 	+	N_("git worktree add [-f] [--lock [--reason <string>]]\n" \
> 	+	   "                 --orphan <new-branch> <path>")
> 	+
> 	 #define BUILTIN_WORKTREE_LIST_USAGE \
> 	 	N_("git worktree list [-v | --porcelain [-z]]")
> 	 #define BUILTIN_WORKTREE_LOCK_USAGE \
> 	@@ -668,6 +671,9 @@ static int add(int ac, const char **av, const char *prefix)
> 	 	if (opts.orphan_branch && !opts.checkout)
> 	 		die(_("'%s' and '%s' cannot be used together"), "--orphan",
> 	 		    "--no-checkout");
> 	+	if (opts.orphan_branch && ac == 2)
> 	+		die(_("'%s' and '%s' cannot be used together"), "--orphan",
> 	+		    _("<commit-ish>"));
> 	 	if (lock_reason && !keep_locked)
> 	 		die(_("the option '%s' requires '%s'"), "--reason", "--lock");
> 	 	if (lock_reason)
> 	diff --git a/t/t2400-worktree-add.sh b/t/t2400-worktree-add.sh
> 	index 93c340f4aff..47461d02115 100755
> 	--- a/t/t2400-worktree-add.sh
> 	+++ b/t/t2400-worktree-add.sh
> 	@@ -326,6 +326,10 @@ test_expect_success '"add" --orphan/--no-checkout mutually exclusive' '
> 	 	test_must_fail git worktree add --orphan poodle --no-checkout bamboo
> 	 '
>
> 	+test_expect_success '"add" --orphan and <commit-ish> mutually exclusive' '
> 	+	test_must_fail git worktree add --orphan poodle bamboo main
> 	+'
> 	+
> 	 test_expect_success '"add" -B/--detach mutually exclusive' '
> 	 	test_must_fail git worktree add -B poodle --detach bamboo main
> 	 '

Yep, you are right. I applied the patch as part of this 2/2 patch and will
include it in v4. When it comes to attribution, is there a preferred way to
handle this?

>
> > -	if (ac < 2 && !new_branch && !opts.detach) {
> > +	/*
> > +	 * As the orphan cannot be created until the contents of branch
> > +	 * are staged, opts.orphan_branch should be treated as both a boolean
> > +	 * indicating that `--orphan` was selected and as the name of the new
> > +	 * orphan branch from this point on.
> > +	 */
>
> I've re-read this a couple of times, and I honestly still don't see what
> point is trying to drive home.
>
> So, "--orphan" is an OPT_STRING(), so it always has a value:
>
> 	$ ./git worktree add --orphan
> 	error: option `orphan' requires a value
>
> But we init it to NULL, and above we just used it as a boolean *and*
> below.
>
> I.e. this comment would seem to me to fit with code where the reader
> might be surprised that we're using "opts.orphan_branch" as a string
> from then on, but we're just copying that to "new_branch", then we
> always use "opts.orphan_branch" as a boolean for the rest of the
> function.
>
> I may be missing something, but I think this would probably be better
> just without this comment. E.g. we use "--track", "--lock-reason"
> etc. in similar ways, and those don't have a comment like that.
>

Originally the new orphan branch's name was passed into
`add_worktree(path, refname, opts)` via the `orphan_branch` field in `opts` and
the branch which was to be checked out first(to mimic `git checkout --orphan`)
was passed in via `refname`.

Now that the behavior was changed to use `git switch`, that
"checkout then make orphan" behavior was unneeded and `refname` also contains
the name of the orphan branch.

For `make_worktree_orphan(opts, child_env)` however since I used the same
function signature as `checkout_worktree(opts, child_env)`, `refname` wasn't
passed in and I used `opts->orphan_branch` to access the branch name from
that scope.

I can change `make_worktree_orphan(opts, child_env)` to
`make_worktree_orphan(ref, opts, child_env)` instead and then `orphan_branch`
would be able to be treated as a boolean like those other flags.

>
> > +	if (opts.orphan_branch) {
> > +		new_branch = opts.orphan_branch;
> > +	}
> > +
> > +	if (ac < 2 && !new_branch && !opts.detach && !opts.orphan_branch) {
>
> In general we shouldn't combine random "if"'s just because a a
> sufficiently smart compiler could discover a way to reduce work.
>
> But in this case these seem to be inherently connected, we always want
> the not-DWIM behavior with "orphan", no?
>
> So shouldn't this just be:
>
> 	if (opts.orphan_branch) {
> 		...
> 	} else if (ac < 2 && !new_branch && !opts.detach) {
> 		....
> 	}
>
> ?

Yes. I've updated that for v4.