git/list[1] front-page[2] threads[3] people[4] search[5] about
 

Re: [PATCH 3/3] commit: give correct advice for empty commit during a rebase

From
Johannes Schindelin <johannes.schindelin@gmx.de>
Date
Oct 25, 2019, 11:48 UTC
Message-ID
<nycvar.QRO.7.76.6.1910251341550.46@tvgsbejvaqbjf.bet>
In-Reply-To
<9d550cd9-08a2-eee5-7d7f-63678285accc@gmail.com>
Hi Phillip,
On Thu, 24 Oct 2019, Phillip Wood wrote:
Show 27 quoted lines
> On 23/10/2019 00:30, Johannes Schindelin via GitGitGadget wrote:
> > From: Johannes Schindelin <johannes.schindelin@gmx.de>
> >
> > In dcb500dc16c (cherry-pick/revert: advise using --skip, 2019-07-02),
> > `git commit` learned to suggest to run `git cherry-pick --skip` when
> > trying to cherry-pick an empty patch.
> >
> > However, it was overlooked that there are more conditions than just a
> > `git cherry-pick` when this advice is printed (which originally
> > suggested the neutral `git reset`): the same can happen during a rebase.
> >
> > Let's suggest the correct command, even during a rebase.
> >
> > While at it, we adjust more places in `builtin/commit.c` that
> > incorrectly assumed that the presence of a `CHERRY_PICK_HEAD` meant that
> > surely this must be a `cherry-pick` in progress.
> >
> > Note: we take pains to handle the situation when a user runs a `git
> > cherry-pick` _during_ a rebase. This is quite valid (e.g. in an `exec`
> > line in an interactive rebase). On the other hand, it is not possible to
> > run a rebase during a cherry-pick, meaning: if both `rebase-merge/` and
> > `sequencer/` exist, we still want to advise to use `git cherry-pick
> > --skip`.
>
> Thanks for working on this. It's unfortunate that rebase does not remove
> CHERRY_PICK_HEAD for empty commits as it does if the commit is not
> empty.

Oh, so that's what it is that bites me all the time? I frequently run interactive rebases and sometimes mess up authorship (taking authorship of patches that I did not, in fact, author). I guess what happens is that I `git commit` in the middle of a rebase that was interrupted by merge conflicts.

So I would actually like to see the _exact opposite_ of what you want to see: I want to keep `CHERRY_PICK_HEAD` even in the non-empty case.

Show 5 quoted lines
> I think this is because 'rebase --continue' will skip an empty commit
> so the user _has_ to run 'git commit' manually to keep it. If it had
> been designed so that 'rebase --continue' kept the empty commit and
> 'rebase --skip' skipped it then we would not have this problem but
> it's a bit late to worry about that now.
True.
Show 7 quoted lines
> I don't this patch can distinguish between an empty cherry-pick
> performed by the user while a rebase is in progress and an empty pick
> performed by rebase as both create CHERRY_PICK_HEAD while
> .git/rebase-merge exists. It seems to assume that CHERRY_PICK_HEAD was
> created by rebase and prints advise based on that which may or may not
> be the correct. I think we could distinguish the two by checking if
> CHERRY_PICK_HEAD matches .git/rebase-merge/stopped-sha or REBASE_HEAD.

I guess we could, but then, I would rather worry about that in the next cycle. In this cycle, I would rather fix the common case, which is that a `git rebase -i` fails and tells me to `git cherry-pick --skip` instead of `git rebase --skip`.

And even if I performed a `git cherry-pick` during a `git rebase` and the result would be an empty commit, I'd rather be told to `git rebase --skip` to continue...

But if you feel strongly that this should be fixed differently, I'll gladly leave it to you ;-)

Ciao, Dscho

Show 123 quoted lines
>
> Best Wishes
>
>
> Phillip
>
> >
> > Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
> > ---
> >  builtin/commit.c       | 33 ++++++++++++++++++++++++---------
> >  t/t3403-rebase-skip.sh |  9 +++++++++
> >  2 files changed, 33 insertions(+), 9 deletions(-)
> >
> > diff --git a/builtin/commit.c b/builtin/commit.c
> > index e588bc6ad3..2beae13620 100644
> > --- a/builtin/commit.c
> > +++ b/builtin/commit.c
> > @@ -59,6 +59,9 @@ N_("The previous cherry-pick is now empty, possibly due to conflict resolution.\
> >  "    git commit --allow-empty\n"
> >  "\n");
> >
> > +static const char empty_rebase_advice[] =
> > +N_("Otherwise, please use 'git rebase --skip'\n");
> > +
> >  static const char empty_cherry_pick_advice_single[] =
> >  N_("Otherwise, please use 'git cherry-pick --skip'\n");
> >
> > @@ -122,7 +125,7 @@ static enum commit_msg_cleanup_mode cleanup_mode;
> >  static const char *cleanup_arg;
> >
> >  static enum commit_whence whence;
> > -static int sequencer_in_use;
> > +static int sequencer_in_use, rebase_in_progress;
> >  static int use_editor = 1, include_status = 1;
> >  static int have_option_m;
> >  static struct strbuf message = STRBUF_INIT;
> > @@ -183,6 +186,8 @@ static void determine_whence(struct wt_status *s)
> >  		whence = FROM_CHERRY_PICK;
> >  		if (file_exists(git_path_seq_dir()))
> >  			sequencer_in_use = 1;
> > +		if (file_exists(git_path_rebase_merge_dir()))
> > +			rebase_in_progress = 1;
> >  	}
> >  	else
> >  		whence = FROM_COMMIT;
> > @@ -453,8 +458,11 @@ static const char *prepare_index(int argc, const char **argv, const char *prefix
> >  	if (whence != FROM_COMMIT) {
> >  		if (whence == FROM_MERGE)
> >  			die(_("cannot do a partial commit during a merge."));
> > -		else if (whence == FROM_CHERRY_PICK)
> > +		else if (whence == FROM_CHERRY_PICK) {
> > +			if (rebase_in_progress && !sequencer_in_use)
> > +				die(_("cannot do a partial commit during a rebase."));
> >  			die(_("cannot do a partial commit during a cherry-pick."));
> > +		}
> >  	}
> >
> >  	if (list_paths(&partial, !current_head ? NULL : "HEAD", &pathspec))
> > @@ -950,10 +958,12 @@ static int prepare_to_commit(const char *index_file, const char *prefix,
> >  			fputs(_(empty_amend_advice), stderr);
> >  		else if (whence == FROM_CHERRY_PICK) {
> >  			fputs(_(empty_cherry_pick_advice), stderr);
> > -			if (!sequencer_in_use)
> > -				fputs(_(empty_cherry_pick_advice_single), stderr);
> > -			else
> > +			if (sequencer_in_use)
> >  				fputs(_(empty_cherry_pick_advice_multi), stderr);
> > +			else if (rebase_in_progress)
> > +				fputs(_(empty_rebase_advice), stderr);
> > +			else
> > +				fputs(_(empty_cherry_pick_advice_single), stderr);
> >  		}
> >  		return 0;
> >  	}
> > @@ -1156,8 +1166,11 @@ static int parse_and_validate_options(int argc, const char *argv[],
> >  	if (amend && whence != FROM_COMMIT) {
> >  		if (whence == FROM_MERGE)
> >  			die(_("You are in the middle of a merge -- cannot amend."));
> > -		else if (whence == FROM_CHERRY_PICK)
> > +		else if (whence == FROM_CHERRY_PICK) {
> > +			if (rebase_in_progress && !sequencer_in_use)
> > +				die(_("You are in the middle of a rebase -- cannot amend."));
> >  			die(_("You are in the middle of a cherry-pick -- cannot amend."));
> > +		}
> >  	}
> >  	if (fixup_message && squash_message)
> >  		die(_("Options --squash and --fixup cannot be used together"));
> > @@ -1628,9 +1641,11 @@ int cmd_commit(int argc, const char **argv, const char *prefix)
> >  			reduce_heads_replace(&parents);
> >  	} else {
> >  		if (!reflog_msg)
> > -			reflog_msg = (whence == FROM_CHERRY_PICK)
> > -					? "commit (cherry-pick)"
> > -					: "commit";
> > +			reflog_msg = (whence != FROM_CHERRY_PICK)
> > +					? "commit"
> > +					: rebase_in_progress && !sequencer_in_use
> > +					? "commit (rebase)"
> > +					: "commit (cherry-pick)";
> >  		commit_list_insert(current_head, &parents);
> >  	}
> >
> > diff --git a/t/t3403-rebase-skip.sh b/t/t3403-rebase-skip.sh
> > index 1f5122b632..77b03ac49f 100755
> > --- a/t/t3403-rebase-skip.sh
> > +++ b/t/t3403-rebase-skip.sh
> > @@ -76,4 +76,13 @@ test_expect_success 'moved back to branch correctly' '
> >
> >  test_debug 'gitk --all & sleep 1'
> >
> > +test_expect_success 'correct advice upon empty commit' '
> > +	git checkout -b rebase-skip &&
> > +	test_commit a1 &&
> > +	test_tick &&
> > +	git commit --amend -m amended --no-edit &&
> > +	test_must_fail git rebase -m --onto a1 HEAD^ 2>err &&
> > +	test_i18ngrep "git rebase --skip" err
> > +'
> > +
> >  test_done
> >
>
>
Previous: Phillip WoodNext: Phillip Wood
Message 7 of 31 in “commit: fix advice for empty commits during rebases”
  1. 0/3 commit: fix advice for empty commits during rebasesJohannes Schindelin via GitGitGadget, Oct 22, 2019
  2. 2/3 sequencer: export the function to get the path of `.git/rebase-merge/`Johannes Schindelin via GitGitGadget, Oct 22, 2019
  3. 1/3 cherry-pick: add test for `--skip` advice in `git commit`Johannes Schindelin via GitGitGadget, Oct 22, 2019
  4. 3/3 commit: give correct advice for empty commit during a rebaseJohannes Schindelin via GitGitGadget, Oct 22, 2019
  5. Junio C HamanoOct 23, 2019
  6. Phillip WoodOct 24, 2019
  7. Johannes SchindelinOct 25, 2019
  8. Phillip WoodOct 25, 2019
  9. Junio C HamanoNov 8, 2019
  10. Johannes SchindelinNov 8, 2019
  11. Phillip WoodNov 11, 2019
  12. Junio C HamanoOct 23, 2019
  13. Johannes SchindelinOct 25, 2019
  14. Junio C HamanoOct 29, 2019
  15. Johannes SchindelinOct 29, 2019
  16. 0/9 commit: fix advice for empty commits during rebasesPhillip Wood, Dec 6, 2019
  17. 2/9 cherry-pick: add test for `--skip` advice in `git commit`Phillip Wood, Dec 6, 2019
  18. 4/9 sequencer: write CHERRY_PICK_HEAD for reword and editPhillip Wood, Dec 6, 2019
  19. 3/9 cherry-pick: check commit error messagesPhillip Wood, Dec 6, 2019
  20. 1/9 t3404: use test_cmp_revPhillip Wood, Dec 6, 2019
  21. Junio C HamanoDec 6, 2019
  22. 6/9 commit: encapsulate determine_whence() for sequencerPhillip Wood, Dec 6, 2019
  23. Junio C HamanoDec 6, 2019
  24. Phillip WoodDec 18, 2019
  25. 5/9 commit: use enum value for multiple cherry-picksPhillip Wood, Dec 6, 2019
  26. Junio C HamanoDec 6, 2019
  27. 9/9 [RFC] rebase -i: leave CHERRY_PICK_HEAD when there are conflictsPhillip Wood, Dec 6, 2019
  28. Phillip WoodDec 18, 2019
  29. 7/9 commit: give correct advice for empty commit during a rebasePhillip Wood, Dec 6, 2019
  30. 8/9 [RFC] rebase: fix advice when a fixup creates an empty commitPhillip Wood, Dec 6, 2019
  31. Elijah NewrenFeb 26, 2020

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.