Hi Dscho

On 10/10/2019 22:31, Johannes Schindelin wrote:

Hi Phillip,

On Thu, 10 Oct 2019, Phillip Wood via GitGitGadget wrote:

quoted

From: Phillip Wood <redacted>

Prior to commit 356ee4659b ("sequencer: try to commit without forking
'git commit'", 2017-11-24) the sequencer would always run the
post-commit hook after each pick or revert as it forked `git commit` to
create the commit. The conversion to committing without forking `git
commit` omitted to call the post-commit hook after creating the commit.

Signed-off-by: Phillip Wood <redacted>

Makes sense.

quoted

---
  builtin/commit.c              |  2 +-
  sequencer.c                   |  5 +++++
  sequencer.h                   |  1 +
  t/t3404-rebase-interactive.sh | 17 +++++++++++++++++
  4 files changed, 24 insertions(+), 1 deletion(-)

diff --git a/builtin/commit.c b/builtin/commit.c
index d898a57f5d..adb8c89c60 100644
--- a/builtin/commit.c
+++ b/builtin/commit.c

@@ -1653,7 +1653,7 @@ int cmd_commit(int argc, const char **argv, const char *prefix)

  	repo_rerere(the_repository, 0);
  	run_command_v_opt(argv_gc_auto, RUN_GIT_CMD);
-	run_commit_hook(use_editor, get_index_file(), "post-commit", NULL);
+	run_post_commit_hook(use_editor, get_index_file());

Does it really make sense to abstract the hook name away? It adds a lot
of churn for just two callers...

I'll drop the new function in the reroll

quoted

  	if (amend && !no_post_rewrite) {
  		commit_post_rewrite(the_repository, current_head, &oid);
  	}

diff --git a/sequencer.c b/sequencer.c
index 3ce578c40b..b4947f6969 100644
--- a/sequencer.c
+++ b/sequencer.c

@@ -1173,6 +1173,10 @@ static int run_prepare_commit_msg_hook(struct repository *r,
  	return ret;
  }

+void run_post_commit_hook(int editor_is_used, const char *index_file) {
+	run_commit_hook(editor_is_used, index_file, "post-commit", NULL);
+}
+

If we must have a separate `run_post_commit_hook()`, then it should be
an `inline` function, defined in the header. Or even a macro to begin
with.

quoted

  static const char implicit_ident_advice_noconfig[] =
  N_("Your name and email address were configured automatically based\n"
  "on your username and hostname. Please check that they are accurate.\n"

@@ -1427,6 +1431,7 @@ static int try_to_commit(struct repository *r,
  		goto out;
  	}

+	run_post_commit_hook(0, r->index_file);

So this is the _actual_ change of this patch.

quoted

  	if (flags & AMEND_MSG)
  		commit_post_rewrite(r, current_head, oid);

diff --git a/sequencer.h b/sequencer.h
index b0419d6ddb..e3e73c5635 100644
--- a/sequencer.h
+++ b/sequencer.h

@@ -203,4 +203,5 @@ int sequencer_get_last_command(struct repository* r,
  			       enum replay_action *action);
  LAST_ARG_MUST_BE_NULL
  int run_commit_hook(int editor_is_used, const char *index_file, const char *name, ...);
+void run_post_commit_hook(int editor_is_used, const char *index_file);
  #endif /* SEQUENCER_H */

diff --git a/t/t3404-rebase-interactive.sh b/t/t3404-rebase-interactive.sh
index d2f1d5bd23..d9217235b6 100755
--- a/t/t3404-rebase-interactive.sh
+++ b/t/t3404-rebase-interactive.sh

@@ -1467,4 +1467,21 @@ test_expect_success 'valid author header when author contains single quote' '
  	test_cmp expected actual
  '

+test_expect_success 'post-commit hook is called' '
+	test_when_finished "rm -f .git/hooks/post-commit commits" &&
+	mkdir -p .git/hooks &&
+	write_script .git/hooks/post-commit <<-\EOS &&
+	git rev-parse HEAD >>commits

Should `commits` be initialized before this script is written, e.g.
using

	>commits &&

Good point, especially if it is renamed to actual as Junio suggests

quoted

+	EOS
+	set_fake_editor &&

The `set_fake_editor` function sets a global environment variable, and
therefore needs to be run in a subshell. Therefore, this line (as well
as the next one) need to be enclosed in `( ... )`.

There are ~80 instances of 
set_fake_editor/test_set_editor/set_cat_todo_editor in that file that 
are not in subshells. I've converted them in a preparatory patch (that 
was fun), removing about 20 that can now safely rely on EDITOR=: 
(hopefully that will ameliorate the performance hit of ~60 extra 
subshells a little)

quoted

+	FAKE_LINES="edit 4 1 reword 2 fixup 3" git rebase -i A E &&
+	echo x>file3 &&

We usually leave no space after the `>`, but we _do_ leave a space
_before_ the `>`.

quoted

+	git add file3 &&
+	FAKE_COMMIT_MESSAGE=edited git rebase --continue &&
+	# rev-list does not support -g --reverse
+	git rev-list --no-walk=unsorted HEAD@{5} HEAD@{4} HEAD@{3} HEAD@{2} \
+		HEAD@{1} HEAD >expected &&

Wouldn't this be better as:

	git rev-parse HEAD@{5} HEAD@{4} HEAD@{3} HEAD@{2} HEAD@{1} HEAD \
		>expect &&

Good point

quoted

+	test_cmp expected commits

We usually use the name `expect` instead of `expected` in the test
suite.

OK

Thanks for looking at this series

Phillip

Thanks,
Dscho

quoted

+'
+
  test_done
--
gitgitgadget