Web lists-archives.com

Re: [PATCH] rebase -r: always reword merge -c




Hi Dscho

On 29/04/2019 17:14, Johannes Schindelin wrote:
Hi Phillip,

On Fri, 26 Apr 2019, Phillip Wood wrote:

From: Phillip Wood <phillip.wood@xxxxxxxxxxxxx>

If a merge can be fast-forwarded then make sure that we still edit the
commit message if the user specifies -c. The implementation follows the
same pattern that is used for ordinary rewords that are fast-forwarded.

Signed-off-by: Phillip Wood <phillip.wood@xxxxxxxxxxxxx>
---

OMG I was bitten twice by this very bug in the past week, and planned on
looking into it next week. Thanks for beating me to it.

Two comments:

diff --git a/sequencer.c b/sequencer.c
index 0db410d590..ff8565e7a8 100644
--- a/sequencer.c
+++ b/sequencer.c
@@ -3248,6 +3248,10 @@ static int do_merge(struct repository *r,
  		rollback_lock_file(&lock);
  		ret = fast_forward_to(r, &commit->object.oid,
  				      &head_commit->object.oid, 0, opts);
+		if (flags & TODO_EDIT_MERGE_MSG) {
+			run_commit_flags |= AMEND_MSG;
+			goto fast_forward_edit;
+		}
  		goto leave_merge;
  	}

@@ -3351,6 +3355,7 @@ static int do_merge(struct repository *r,
  		 * value (a negative one would indicate that the `merge`
  		 * command needs to be rescheduled).
  		 */
+	fast_forward_edit:

It is *slightly* awkward that this is an `else` arm of an `if (ret)`, but
I do not necessarily think that it would be better to move the label
before the `if` than what you did; Your version comes out more readable,
still.

I did wonder about adding braces but I'm not sure that makes it any clearer. I agree having the label before the `if (ret)` would be less clear as the reader has to then think what ret will be in that case to work out what will happen.

  		ret = !!run_git_commit(r, git_path_merge_msg(r), opts,
  				       run_commit_flags);

diff --git a/t/t3430-rebase-merges.sh b/t/t3430-rebase-merges.sh
index 4c69255ee6..3d484a3c72 100755
--- a/t/t3430-rebase-merges.sh
+++ b/t/t3430-rebase-merges.sh
@@ -164,6 +164,16 @@ test_expect_success 'failed `merge <branch>` does not crash' '
  	grep "^Merge branch ${SQ}G${SQ}$" .git/rebase-merge/message
  '

+test_expect_success 'fast-forward merge -c still rewords' '
+	git checkout -b fast-forward-merge-c H &&
+	set_fake_editor &&

set_fake_editor affects global state AFAIR (setting and exporting
`EDITOR`), therefore this would need to be run in a subshell, i.e.
enclosed in parentheses.

The other test files are not very consistent about that. I'll re-roll. Note that I do not export any FAKE_* variables, so later tests should not be affected even if the fake editor runs.

Best Wishes

Phillip
+	FAKE_COMMIT_MESSAGE=edited GIT_SEQUENCE_EDITOR="echo merge -c H G >" \
+		git rebase -ir @^ &&
+	echo edited >expected &&
+	git log --pretty=format:%B -1 >actual &&
+	test_cmp expected actual
+'
+

The rest looks good, thank you!
Dscho

  test_expect_success 'with a branch tip that was cherry-picked already' '
  	git checkout -b already-upstream master &&
  	base="$(git rev-parse --verify HEAD)" &&
--
2.21.0