Re: [PATCH] rebase --abort: cleanup refs/rewritten

["SZEDER Gábor" <szeder.dev@gmail.com>]

On Fri, Apr 26, 2019 at 11:32:12AM +0100, Phillip Wood wrote:
> From: Phillip Wood <phillip.wood@dunelm.org.uk>
> 
> When `rebase -r` finishes it removes any refs under refs/rewritten
> that it has created. However if the rebase is aborted these refs are
> not removed. This can cause problems for future rebases. For example I
> recently wanted to merge a updated version of a topic branch into an
> integration branch so ran `rebase -ir` and removed the picks and label
> for the topic branch from the todo list so that
>     merge -C <old-merge> topic
> would pick up the new version of topic. Unfortunately
> refs/rewritten/topic already existed from a previous rebase that had
> been aborted so the rebase just used the old topic, not the new one.
> 
> Signed-off-by: Phillip Wood <phillip.wood@dunelm.org.uk>
> ---
> 
> Notes:
>     This is based on pw/rebase-i-internal, it would be nicer to base it on
>     maint but there are function name clashes adding sequencer.h to rebase.c
>     an maint. Those clashes are fixed in pw/rebase-i-internal
> 
>  builtin/rebase.c         | 13 ++++++++++---
>  t/t3430-rebase-merges.sh |  8 ++++++++
>  2 files changed, 18 insertions(+), 3 deletions(-)
> 
> diff --git a/builtin/rebase.c b/builtin/rebase.c
> index 82bd50a1b4..e2e49c8239 100644
> --- a/builtin/rebase.c
> +++ b/builtin/rebase.c
> @@ -761,9 +761,16 @@ static int finish_rebase(struct rebase_options *opts)
>  	 * user should see them.
>  	 */
>  	run_command_v_opt(argv_gc_auto, RUN_GIT_CMD);
> -	strbuf_addstr(&dir, opts->state_dir);
> -	remove_dir_recursively(&dir, 0);
> -	strbuf_release(&dir);
> +	if (opts->type == REBASE_INTERACTIVE) {
> +		struct replay_opts replay = REPLAY_OPTS_INIT;

This patch and the topic 'pw/rebase-abort-clean-rewritten' can't be
compiled on its own, because it starts using 'struct replay_opts'
here, which is defined in 'sequencer.h', but 'builtin/rebase.c'
doesn't include that header yet.  (Though 'pu' already builds fine,
because commit 0609b741a4 (rebase -i: combine rebase--interactive.c
with rebase.c, 2019-04-17) in the parallel topic
'pw/rebase-i-internal' adds the necessary #include.)

So, to keep future bisects from potentially tipping over the compiler
error, this patch should either #include "sequencer.h", or be applied
on top of 'pw/rebase-i-internal'.

> +
> +		replay.action = REPLAY_INTERACTIVE_REBASE;
> +		sequencer_remove_state(&replay);
> +	} else {
> +		strbuf_addstr(&dir, opts->state_dir);
> +		remove_dir_recursively(&dir, 0);
> +		strbuf_release(&dir);
> +	}
>  
>  	return 0;
>  }
> diff --git a/t/t3430-rebase-merges.sh b/t/t3430-rebase-merges.sh
> index 4c69255ee6..6ebebf7098 100755
> --- a/t/t3430-rebase-merges.sh
> +++ b/t/t3430-rebase-merges.sh
> @@ -224,6 +224,14 @@ test_expect_success 'refs/rewritten/* is worktree-local' '
>  	test_cmp_rev HEAD "$(cat wt/b)"
>  '
>  
> +test_expect_success '--abort cleans up refs/rewritten' '
> +	git checkout -b abort-cleans-refs-rewritten H &&
> +	GIT_SEQUENCE_EDITOR="echo break >>" git rebase -ir @^ &&
> +	git rev-parse --verify refs/rewritten/onto &&
> +	git rebase --abort &&
> +	test_must_fail git rev-parse --verify refs/rewritten/onto
> +'
> +
>  test_expect_success 'post-rewrite hook and fixups work for merges' '
>  	git checkout -b post-rewrite &&
>  	test_commit same1 &&
> -- 
> 2.21.0
>