[PATCH/RFC 0/5] replay: support replaying 2-parent merges

[PATCH/RFC 0/5] replay: support replaying 2-parent merges

[Johannes Schindelin via GitGitGadget]

git history, the new history-rewriting builtin in v2.54, dies on any merge
in the rewrite path with replaying merge commits is not supported yet!. That
makes it not very useful for the workflows I actually have, where almost
every interesting branch contains at least one merge of a feature topic. The
natural fallback, git rebase --rebase-merges, is interactive and stops to
ask for re-resolution even when no re-resolution is needed.

This series lifts that limitation for the common 2-parent case. The
algorithm itself is not new: Elijah Newren wrote it down in his replay
design notes
[https://github.com/newren/git/blob/replay/replay-design-notes.txt] and
prototyped it in a 2022 work-in-progress sketch
[https://github.com/newren/git/commit/4c45e8955ef9bf7d01fd15d9106b3bdb8ea91b45].
What is new is wiring it into the replay_revisions() API that backs both git
replay and git history, plus three specific tweaks that make the trickier
cases work where the WIP sketch bailed out: identical conflict-marker labels
for the inner remerges of the original and the rewritten parents (so their
conflict-markered trees compare equal in the regions the user did not
touch), tolerating result.clean == 0 from those inner merges (their
well-defined conflict-markered trees are valid inputs to the outer 3-way
merge), and self-fallback for both merge parents combined with mapping the
rev-range boundary commits to the onto commit.

Octopus merges and revert-of-merge are surfaced as explicit errors at the
dispatch point. The split sub-command of git history continues to refuse
when its target is a merge: split semantics simply do not apply there. The
xdiff special mode for matching conflict-marker hunks across inner remerges,
the XDL_MERGE_FAVOR_BASE variant, and the modify/delete and binary-file
specials that the design notes flag as future work all remain future work.

While I was at it, git history reword had a pre-existing silent-success bug:
a positive return from replay_revisions() (which means "conflict, no updates
queued") was treated as success. Obviously this should never occur, as a
reword simply does not change any of the file contents, but bugs do happen.
The merge-replay work is complex enough to make that class of bugs more
likely, therefore I introduce error messages for those instances.

Johannes Schindelin (5):
  replay: support replaying 2-parent merges
  replay: short-circuit merge replay when parent and base trees are
    unchanged
  history.adoc: describe merge-replay support and its limits
  test-tool: add a "historian" subcommand for building merge fixtures
  t3454: cover merge-replay scenarios with the historian helper

 Documentation/git-history.adoc |  27 ++-
 Makefile                       |   1 +
 builtin/history.c              |  16 +-
 replay.c                       | 258 +++++++++++++++++++++++++--
 t/helper/meson.build           |   1 +
 t/helper/test-historian.c      | 189 ++++++++++++++++++++
 t/helper/test-tool.c           |   1 +
 t/helper/test-tool.h           |   1 +
 t/meson.build                  |   1 +
 t/t3451-history-reword.sh      |  21 ++-
 t/t3452-history-split.sh       |   6 +-
 t/t3454-history-merges.sh      | 308 +++++++++++++++++++++++++++++++++
 t/t3650-replay-basics.sh       |  46 ++++-
 13 files changed, 840 insertions(+), 36 deletions(-)
 create mode 100644 t/helper/test-historian.c
 create mode 100755 t/t3454-history-merges.sh


base-commit: 94f057755b7941b321fd11fec1b2e3ca5313a4e0
Published-As: https://github.com/gitgitgadget/git/releases/tag/pr-2106%2Fdscho%2Fsupport-merge-commits-in-git-history-v1
Fetch-It-Via: git fetch https://github.com/gitgitgadget/git pr-2106/dscho/support-merge-commits-in-git-history-v1
Pull-Request: https://github.com/gitgitgadget/git/pull/2106
-- 
gitgitgadget

[PATCH/RFC 1/5] replay: support replaying 2-parent merges

[Johannes Schindelin via GitGitGadget]

From: Johannes Schindelin <johannes.schindelin@gmx.de>

`git history` (introduced in v2.54) and the underlying `git replay`
infrastructure both refused to walk past any commit with more than
one parent, dying with "replaying merge commits is not supported
yet!". For real history-rewriting work this is a showstopper: the
natural fallback `git rebase --rebase-merges` is interactive and
stops to ask for re-resolution even when no re-resolution is needed.

Elijah Newren spelled out a way to lift this limitation in his
replay-design-notes [1] and prototyped it in a 2022
work-in-progress sketch [2]. The idea is that a merge commit M on
parents (P1, P2) records both an automatic merge of those parents
AND any manual layer the author put on top of that automatic merge
(textual conflict resolution and any semantic edit outside conflict
markers). Replaying M onto rewritten parents (P1', P2') must
preserve that manual layer, but the rewritten parents change the
automatic merge, so a simple cherry-pick is wrong: the manual layer
would be re-introduced on top of stale auto-merge text.

What works instead is a three-way merge of three trees the existing
infrastructure already knows how to compute. Let R be the recursive
auto-merge of (P1, P2), O be M's actual tree and N be the recursive
auto-merge of (P1', P2'). Then `git diff R O` is morally
`git show --remerge-diff M`: it captures exactly what the author
added on top of the automatic merge. A non-recursive 3-way merge
with R as the merge base, O as side 1 and N as side 2 layers that
manual contribution onto the freshly auto-merged rewritten parents
(N) and produces the replayed tree.

Implement `pick_merge_commit()` along those lines and dispatch to it
from `replay_revisions()` when the commit being replayed has exactly
two parents. Two specific points (learned the hard way) keep
non-trivial cases working where the WIP sketch [2] bailed out.
First, R and N use identical `merge_options.branch1` and `branch2`
labels ("ours"/"theirs"). When the original parents conflicted on a
region of a file, both R and N produce textually identical conflict
markers; the outer non-recursive merge then sees N == R in that
region and the user's manual resolution from O wins cleanly. Without
this, the conflict-marker text would differ between R and N (because
the inner merges would label the conflicts differently), and the
outer merge would itself be unclean even when the user did supply a
clean resolution. Second, an unclean inner merge
(`result.clean == 0`) is _not_ fatal: the tree merge-ort produces in
that case still has well-defined contents (with conflict markers in
the conflicted files) and is a valid input to the outer
non-recursive merge. Only a real error (`< 0`) propagates as
failure.

The replay propagates the textual diffs the user actually made in M;
it does _not_ extrapolate symbol-level intent. If rewriting the
parents pulls in genuinely new content (for example, a brand-new
caller of a function that the merge renamed), that new content stays
as the rewritten parents have it. Symbol-aware refactoring is out of
scope here, just as it is for plain rebase.

Octopus merges (more than two parents) and revert-of-merge are not
supported and are surfaced as explicit errors at the dispatch point.
The "split" sub-command of `git history` continues to refuse when
the targeted commit is itself a merge: split semantics do not apply
to merges. The pre-walk gate in `builtin/history.c` that previously
rejected any merge in the rewrite path now only rejects octopus
merges; rename it accordingly.

A small refactor in `create_commit()` makes the merge case possible:
the helper now takes a `struct commit_list *parents` rather than a
single parent pointer and takes ownership of the list. The single
existing caller in `pick_regular_commit()` builds and passes a
one-element list; the new `pick_merge_commit()` builds a two-element
list, with the order of the `from` and `merge` parents preserved.

Update the negative expectations in t3451, t3452 and t3650 that were
asserting the now-retired "not supported yet" message, replacing
them with positive coverage where it fits. Octopus rejection and
revert-of-merge rejection are covered by new positive tests in
t3650. A dedicated test script with merge-replay scenarios driven by
a new test-tool fixture builder will follow in a subsequent commit.

[1] https://github.com/newren/git/blob/replay/replay-design-notes.txt
[2] https://github.com/newren/git/commit/4c45e8955ef9bf7d01fd15d9106b3bdb8ea91b45

Helped-by: Elijah Newren <newren@gmail.com>
Assisted-by: Claude Opus 4.7
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
---
 builtin/history.c         |  16 ++-
 replay.c                  | 209 ++++++++++++++++++++++++++++++++++++--
 t/t3451-history-reword.sh |  21 ++--
 t/t3452-history-split.sh  |   6 +-
 t/t3650-replay-basics.sh  |  46 ++++++++-
 5 files changed, 269 insertions(+), 29 deletions(-)

diff --git a/builtin/history.c b/builtin/history.c
index 9526938085..00097b2226 100644
--- a/builtin/history.c
+++ b/builtin/history.c
@@ -195,15 +195,15 @@ static int parse_ref_action(const struct option *opt, const char *value, int uns
 	return 0;
 }
 
-static int revwalk_contains_merges(struct repository *repo,
-				   const struct strvec *revwalk_args)
+static int revwalk_contains_octopus_merges(struct repository *repo,
+					   const struct strvec *revwalk_args)
 {
 	struct strvec args = STRVEC_INIT;
 	struct rev_info revs;
 	int ret;
 
 	strvec_pushv(&args, revwalk_args->v);
-	strvec_push(&args, "--min-parents=2");
+	strvec_push(&args, "--min-parents=3");
 
 	repo_init_revisions(repo, &revs, NULL);
 
@@ -217,7 +217,7 @@ static int revwalk_contains_merges(struct repository *repo,
 	}
 
 	if (get_revision(&revs)) {
-		ret = error(_("replaying merge commits is not supported yet!"));
+		ret = error(_("replaying octopus merges is not supported"));
 		goto out;
 	}
 
@@ -289,7 +289,7 @@ static int setup_revwalk(struct repository *repo,
 		strvec_push(&args, "HEAD");
 	}
 
-	ret = revwalk_contains_merges(repo, &args);
+	ret = revwalk_contains_octopus_merges(repo, &args);
 	if (ret < 0)
 		goto out;
 
@@ -482,6 +482,9 @@ static int cmd_history_reword(int argc,
 	if (ret < 0) {
 		ret = error(_("failed replaying descendants"));
 		goto out;
+	} else if (ret) {
+		ret = error(_("conflict during replay; some descendants were not rewritten"));
+		goto out;
 	}
 
 	ret = 0;
@@ -721,6 +724,9 @@ static int cmd_history_split(int argc,
 	if (ret < 0) {
 		ret = error(_("failed replaying descendants"));
 		goto out;
+	} else if (ret) {
+		ret = error(_("conflict during replay; some descendants were not rewritten"));
+		goto out;
 	}
 
 	ret = 0;
diff --git a/replay.c b/replay.c
index f96f1f6551..3dbce095f9 100644
--- a/replay.c
+++ b/replay.c
@@ -1,6 +1,7 @@
 #define USE_THE_REPOSITORY_VARIABLE
 
 #include "git-compat-util.h"
+#include "commit-reach.h"
 #include "environment.h"
 #include "hex.h"
 #include "merge-ort.h"
@@ -77,15 +78,21 @@ static void generate_revert_message(struct strbuf *msg,
 	repo_unuse_commit_buffer(repo, commit, message);
 }
 
+/*
+ * Build a new commit with the given tree and parent list, copying author,
+ * extra headers and (for pick mode) the commit message from `based_on`.
+ *
+ * Takes ownership of `parents`: it will be freed before returning, even on
+ * error. Parent order is preserved as supplied by the caller.
+ */
 static struct commit *create_commit(struct repository *repo,
 				    struct tree *tree,
 				    struct commit *based_on,
-				    struct commit *parent,
+				    struct commit_list *parents,
 				    enum replay_mode mode)
 {
 	struct object_id ret;
 	struct object *obj = NULL;
-	struct commit_list *parents = NULL;
 	char *author = NULL;
 	char *sign_commit = NULL; /* FIXME: cli users might want to sign again */
 	struct commit_extra_header *extra = NULL;
@@ -96,7 +103,6 @@ static struct commit *create_commit(struct repository *repo,
 	const char *orig_message = NULL;
 	const char *exclude_gpgsig[] = { "gpgsig", "gpgsig-sha256", NULL };
 
-	commit_list_insert(parent, &parents);
 	extra = read_commit_extra_headers(based_on, exclude_gpgsig);
 	if (mode == REPLAY_MODE_REVERT) {
 		generate_revert_message(&msg, based_on, repo);
@@ -273,6 +279,7 @@ static struct commit *pick_regular_commit(struct repository *repo,
 {
 	struct commit *base, *replayed_base;
 	struct tree *pickme_tree, *base_tree, *replayed_base_tree;
+	struct commit_list *parents = NULL;
 
 	if (pickme->parents) {
 		base = pickme->parents->item;
@@ -327,7 +334,143 @@ static struct commit *pick_regular_commit(struct repository *repo,
 	if (oideq(&replayed_base_tree->object.oid, &result->tree->object.oid) &&
 	    !oideq(&pickme_tree->object.oid, &base_tree->object.oid))
 		return replayed_base;
-	return create_commit(repo, result->tree, pickme, replayed_base, mode);
+	commit_list_insert(replayed_base, &parents);
+	return create_commit(repo, result->tree, pickme, parents, mode);
+}
+
+/*
+ * Replay a 2-parent merge commit by composing three calls into merge-ort:
+ *
+ *   R = recursive merge of pickme's two original parents (auto-remerge of
+ *       the original merge, accepting any conflicts)
+ *   N = recursive merge of the (possibly rewritten) parents
+ *   O = pickme's tree (the user's actual merge, including any manual
+ *       resolutions)
+ *
+ * The picked tree comes from a non-recursive merge using R as the base,
+ * O as side1 and N as side2. `git diff R O` is morally `git show
+ * --remerge-diff $oldmerge`, so this layers the user's original manual
+ * resolution on top of the freshly auto-merged rewritten parents (see
+ * `replay-design-notes.txt` on the `replay` branch of newren/git).
+ *
+ * If the outer 3-way merge is unclean, propagate the conflict status to
+ * the caller via `result->clean = 0` and return NULL. The two inner
+ * merges (R and N) being unclean is _not_ fatal: the conflict-markered
+ * trees they produce are valid inputs to the outer merge, and using
+ * identical labels for both inner merges keeps the marker text
+ * byte-equal between R and N so the user's resolution recorded in O
+ * collapses the conflict cleanly there. Octopus merges (more than two
+ * parents) and revert-of-merge are rejected by the caller before this
+ * function is invoked.
+ */
+static struct commit *pick_merge_commit(struct repository *repo,
+					struct commit *pickme,
+					kh_oid_map_t *replayed_commits,
+					struct merge_options *merge_opt,
+					struct merge_result *result)
+{
+	struct commit *parent1, *parent2;
+	struct commit *replayed_par1, *replayed_par2;
+	struct tree *pickme_tree;
+	struct merge_options remerge_opt = { 0 };
+	struct merge_options new_merge_opt = { 0 };
+	struct merge_result remerge_res = { 0 };
+	struct merge_result new_merge_res = { 0 };
+	struct commit_list *parent_bases = NULL;
+	struct commit_list *replayed_bases = NULL;
+	struct commit_list *parents;
+	struct commit *picked = NULL;
+	char *ancestor_name = NULL;
+
+	parent1 = pickme->parents->item;
+	parent2 = pickme->parents->next->item;
+
+	/*
+	 * Map the merge's parents to their replayed counterparts. With the
+	 * boundary commits pre-seeded into `replayed_commits`, every parent
+	 * either has an explicit mapping (rewritten or boundary -> onto) or
+	 * sits outside the rewrite range entirely; the latter must stay at
+	 * the original parent commit, so use `parent` itself as the fallback
+	 * for both sides.
+	 */
+	replayed_par1 = mapped_commit(replayed_commits, parent1, parent1);
+	replayed_par2 = mapped_commit(replayed_commits, parent2, parent2);
+
+	/*
+	 * R: auto-remerge of the original parents.
+	 *
+	 * Use the same branch labels for the inner merges that compute R
+	 * and N so conflict markers (if any) are textually identical
+	 * between the two; the outer non-recursive merge can then collapse
+	 * the manual resolution from O against them.
+	 */
+	init_basic_merge_options(&remerge_opt, repo);
+	remerge_opt.show_rename_progress = 0;
+	remerge_opt.branch1 = "ours";
+	remerge_opt.branch2 = "theirs";
+	if (repo_get_merge_bases(repo, parent1, parent2, &parent_bases) < 0) {
+		result->clean = -1;
+		goto out;
+	}
+	merge_incore_recursive(&remerge_opt, parent_bases,
+			       parent1, parent2, &remerge_res);
+	parent_bases = NULL; /* consumed by merge_incore_recursive */
+	if (remerge_res.clean < 0) {
+		result->clean = remerge_res.clean;
+		goto out;
+	}
+
+	/* N: fresh merge of the (possibly rewritten) parents. */
+	init_basic_merge_options(&new_merge_opt, repo);
+	new_merge_opt.show_rename_progress = 0;
+	new_merge_opt.branch1 = "ours";
+	new_merge_opt.branch2 = "theirs";
+	if (repo_get_merge_bases(repo, replayed_par1, replayed_par2,
+				 &replayed_bases) < 0) {
+		result->clean = -1;
+		goto out;
+	}
+	merge_incore_recursive(&new_merge_opt, replayed_bases,
+			       replayed_par1, replayed_par2, &new_merge_res);
+	replayed_bases = NULL; /* consumed by merge_incore_recursive */
+	if (new_merge_res.clean < 0) {
+		result->clean = new_merge_res.clean;
+		goto out;
+	}
+
+	/*
+	 * Outer non-recursive merge: base=R, side1=O (pickme), side2=N.
+	 */
+	pickme_tree = repo_get_commit_tree(repo, pickme);
+	ancestor_name = xstrfmt("auto-remerge of %s",
+				oid_to_hex(&pickme->object.oid));
+	merge_opt->ancestor = ancestor_name;
+	merge_opt->branch1 = short_commit_name(repo, pickme);
+	merge_opt->branch2 = "merge of replayed parents";
+	merge_incore_nonrecursive(merge_opt,
+				  remerge_res.tree,
+				  pickme_tree,
+				  new_merge_res.tree,
+				  result);
+	merge_opt->ancestor = NULL;
+	merge_opt->branch1 = NULL;
+	merge_opt->branch2 = NULL;
+	if (!result->clean)
+		goto out;
+
+	parents = NULL;
+	commit_list_insert(replayed_par2, &parents);
+	commit_list_insert(replayed_par1, &parents);
+	picked = create_commit(repo, result->tree, pickme, parents,
+			       REPLAY_MODE_PICK);
+
+out:
+	free(ancestor_name);
+	free_commit_list(parent_bases);
+	free_commit_list(replayed_bases);
+	merge_finalize(&remerge_opt, &remerge_res);
+	merge_finalize(&new_merge_opt, &new_merge_res);
+	return picked;
 }
 
 void replay_result_release(struct replay_result *result)
@@ -407,17 +550,63 @@ int replay_revisions(struct rev_info *revs,
 	merge_opt.show_rename_progress = 0;
 	last_commit = onto;
 	replayed_commits = kh_init_oid_map();
+
+	/*
+	 * Seed the rewritten-commit map with each negative-side ("BOTTOM")
+	 * cmdline entry pointing at `onto`. This matters for merge replay:
+	 * a 2-parent merge whose first parent is the boundary (e.g. the
+	 * commit being reworded) must replay onto the rewritten boundary,
+	 * yet pick_merge_commit uses a self fallback so the second parent
+	 * (a side branch outside the rewrite range) is preserved as-is.
+	 * Pre-seeding the boundary disambiguates the two: in the map ->
+	 * rewritten, missing -> kept as-is.
+	 *
+	 * Only do this for the pick path; revert mode chains reverts
+	 * through last_commit and a pre-seeded boundary would short-circuit
+	 * that chain.
+	 */
+	if (mode == REPLAY_MODE_PICK) {
+		for (size_t i = 0; i < revs->cmdline.nr; i++) {
+			struct rev_cmdline_entry *e = &revs->cmdline.rev[i];
+			struct commit *boundary;
+			khint_t pos;
+			int hr;
+
+			if (!(e->flags & BOTTOM))
+				continue;
+			boundary = lookup_commit_reference_gently(revs->repo,
+								  &e->item->oid, 1);
+			if (!boundary)
+				continue;
+			pos = kh_put_oid_map(replayed_commits,
+					     boundary->object.oid, &hr);
+			if (hr != 0)
+				kh_value(replayed_commits, pos) = onto;
+		}
+	}
+
 	while ((commit = get_revision(revs))) {
 		const struct name_decoration *decoration;
 		khint_t pos;
 		int hr;
 
-		if (commit->parents && commit->parents->next)
-			die(_("replaying merge commits is not supported yet!"));
-
-		last_commit = pick_regular_commit(revs->repo, commit, replayed_commits,
-						  mode == REPLAY_MODE_REVERT ? last_commit : onto,
-						  &merge_opt, &result, mode);
+		if (commit->parents && commit->parents->next) {
+			if (commit->parents->next->next) {
+				ret = error(_("replaying octopus merges is not supported"));
+				goto out;
+			}
+			if (mode == REPLAY_MODE_REVERT) {
+				ret = error(_("reverting merge commits is not supported"));
+				goto out;
+			}
+			last_commit = pick_merge_commit(revs->repo, commit,
+							replayed_commits,
+							&merge_opt, &result);
+		} else {
+			last_commit = pick_regular_commit(revs->repo, commit, replayed_commits,
+							  mode == REPLAY_MODE_REVERT ? last_commit : onto,
+							  &merge_opt, &result, mode);
+		}
 		if (!last_commit)
 			break;
 
diff --git a/t/t3451-history-reword.sh b/t/t3451-history-reword.sh
index de7b357685..d103f866a2 100755
--- a/t/t3451-history-reword.sh
+++ b/t/t3451-history-reword.sh
@@ -201,12 +201,21 @@ test_expect_success 'can reword a merge commit' '
 		git switch - &&
 		git merge theirs &&
 
-		# It is not possible to replay merge commits embedded in the
-		# history (yet).
-		test_must_fail git -c core.editor=false history reword HEAD~ 2>err &&
-		test_grep "replaying merge commits is not supported yet" err &&
+		# Reword a non-merge commit whose descendants include the
+		# merge: replay carries the merge through.
+		reword_with_message HEAD~ <<-EOF &&
+		ours reworded
+		EOF
+		expect_graph <<-EOF &&
+		*   Merge tag ${SQ}theirs${SQ}
+		|\\
+		| * theirs
+		* | ours reworded
+		|/
+		* base
+		EOF
 
-		# But it is possible to reword a merge commit directly.
+		# And reword a merge commit directly.
 		reword_with_message HEAD <<-EOF &&
 		Reworded merge commit
 		EOF
@@ -214,7 +223,7 @@ test_expect_success 'can reword a merge commit' '
 		*   Reworded merge commit
 		|\
 		| * theirs
-		* | ours
+		* | ours reworded
 		|/
 		* base
 		EOF
diff --git a/t/t3452-history-split.sh b/t/t3452-history-split.sh
index 8ed0cebb50..ad6309f98b 100755
--- a/t/t3452-history-split.sh
+++ b/t/t3452-history-split.sh
@@ -36,7 +36,7 @@ expect_tree_entries () {
 	test_cmp expect actual
 }
 
-test_expect_success 'refuses to work with merge commits' '
+test_expect_success 'refuses to split a merge commit' '
 	test_when_finished "rm -rf repo" &&
 	git init repo &&
 	(
@@ -49,9 +49,7 @@ test_expect_success 'refuses to work with merge commits' '
 		git switch - &&
 		git merge theirs &&
 		test_must_fail git history split HEAD 2>err &&
-		test_grep "cannot split up merge commit" err &&
-		test_must_fail git history split HEAD~ 2>err &&
-		test_grep "replaying merge commits is not supported yet" err
+		test_grep "cannot split up merge commit" err
 	)
 '
 
diff --git a/t/t3650-replay-basics.sh b/t/t3650-replay-basics.sh
index 3353bc4a4d..368b1b0f9a 100755
--- a/t/t3650-replay-basics.sh
+++ b/t/t3650-replay-basics.sh
@@ -103,10 +103,48 @@ test_expect_success 'cannot advance target ... ordering would be ill-defined' '
 	test_cmp expect actual
 '
 
-test_expect_success 'replaying merge commits is not supported yet' '
-	echo "fatal: replaying merge commits is not supported yet!" >expect &&
-	test_must_fail git replay --advance=main main..topic-with-merge 2>actual &&
-	test_cmp expect actual
+test_expect_success 'using replay to rebase a 2-parent merge' '
+	# main..topic-with-merge contains a 2-parent merge (P) introduced
+	# via test_merge. Use --ref-action=print so this test does not
+	# mutate state for subsequent tests in this file.
+	git replay --ref-action=print --onto main main..topic-with-merge >result &&
+	test_line_count = 1 result &&
+
+	new_tip=$(cut -f 3 -d " " result) &&
+
+	# Result is still a 2-parent merge.
+	git cat-file -p $new_tip >cat &&
+	grep -c "^parent " cat >count &&
+	echo 2 >expect &&
+	test_cmp expect count &&
+
+	# Merge subject is preserved.
+	echo P >expect &&
+	git log -1 --format=%s $new_tip >actual &&
+	test_cmp expect actual &&
+
+	# The replayed merge sits on top of main: walking back via the
+	# first-parent chain reaches main.
+	git merge-base --is-ancestor main $new_tip
+'
+
+test_expect_success 'replaying an octopus merge is rejected' '
+	# Build an octopus side-branch so the rest of the test state stays
+	# untouched.
+	test_when_finished "git update-ref -d refs/heads/octopus-tip" &&
+	octopus_tip=$(git commit-tree -p topic4 -p topic1 -p topic3 \
+		-m "octopus" $(git rev-parse topic4^{tree})) &&
+	git update-ref refs/heads/octopus-tip "$octopus_tip" &&
+
+	test_must_fail git replay --ref-action=print --onto main \
+		topic4..octopus-tip 2>actual &&
+	test_grep "octopus merges" actual
+'
+
+test_expect_success 'reverting a merge commit is rejected' '
+	test_must_fail git replay --ref-action=print --revert=topic-with-merge \
+		topic4..topic-with-merge 2>actual &&
+	test_grep "reverting merge commits" actual
 '
 
 test_expect_success 'using replay to rebase two branches, one on top of other' '
-- 
gitgitgadget

[PATCH/RFC 2/5] replay: short-circuit merge replay when parent and base trees are unchanged

[Johannes Schindelin via GitGitGadget]

From: Johannes Schindelin <johannes.schindelin@gmx.de>

For the common `git history reword` case the rewrite changes only
commit messages, so every commit on the line being replayed has the
same tree as before. When such a rewrite reaches a 2-parent merge
whose rewritten parents AND merge bases all carry the same trees as
the originals, the inner auto-merge of the rewritten parents (N) is
tree-equal to the inner auto-merge of the original parents (R), and
the outer 3-way merge with R as the merge base, the original merge
tree as side 1 and N as side 2 yields the original tree as result.

Detect this in `pick_merge_commit()` before doing any merge work and
write the new merge commit directly with the original tree and the
rewritten parents. This saves two recursive merges and one
non-recursive merge per merge commit on the rewrite path, which
dominates the cost of `git history reword` across histories with
many merges.

The merge-base trees must be checked too, in order. Tree-same
parents over a tree-different base could still produce a different
auto-merge (a conflict region that did not exist before, or vice
versa), and the original resolution would be inappropriate to apply.

To avoid recomputing the merge bases when the fast path does not
apply, both pairs are computed up front and the slow path that
follows reuses them.

Assisted-by: Claude Opus 4.7
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
---
 replay.c | 67 ++++++++++++++++++++++++++++++++++++++++++++++++--------
 1 file changed, 58 insertions(+), 9 deletions(-)

diff --git a/replay.c b/replay.c
index 3dbce095f9..5dfdef1447 100644
--- a/replay.c
+++ b/replay.c
@@ -396,6 +396,64 @@ static struct commit *pick_merge_commit(struct repository *repo,
 	replayed_par1 = mapped_commit(replayed_commits, parent1, parent1);
 	replayed_par2 = mapped_commit(replayed_commits, parent2, parent2);
 
+	/*
+	 * Compute both pairs of merge bases up front. The fast path below
+	 * needs them for the tree-equality check, and the slow path that
+	 * follows reuses them to avoid recomputing.
+	 */
+	if (repo_get_merge_bases(repo, parent1, parent2, &parent_bases) < 0 ||
+	    repo_get_merge_bases(repo, replayed_par1, replayed_par2,
+				 &replayed_bases) < 0) {
+		result->clean = -1;
+		goto out;
+	}
+
+	/*
+	 * Fast path: when both rewritten parents carry the same trees as
+	 * the originals AND every merge base does too (in order), the
+	 * auto-merges R and N would be tree-equal (their inputs match
+	 * content-wise), so the outer 3-way merge trivially yields the
+	 * original merge's tree. Skip the inner merges and write the new
+	 * merge commit directly.
+	 *
+	 * This is the common case for `git history reword`, which only
+	 * changes commit messages and so leaves every tree on the line
+	 * being replayed unchanged. The merge-base trees must be checked
+	 * too: tree-same parents over a tree-different base could still
+	 * produce a different auto-merge (a conflict region that did not
+	 * exist before, or vice versa), and the original resolution would
+	 * be inappropriate.
+	 */
+	if (oideq(&repo_get_commit_tree(repo, parent1)->object.oid,
+		  &repo_get_commit_tree(repo, replayed_par1)->object.oid) &&
+	    oideq(&repo_get_commit_tree(repo, parent2)->object.oid,
+		  &repo_get_commit_tree(repo, replayed_par2)->object.oid)) {
+		struct commit_list *bo, *bn;
+		int bases_match = 1;
+
+		for (bo = parent_bases, bn = replayed_bases;
+		     bo && bn;
+		     bo = bo->next, bn = bn->next) {
+			if (!oideq(&repo_get_commit_tree(repo, bo->item)->object.oid,
+				   &repo_get_commit_tree(repo, bn->item)->object.oid)) {
+				bases_match = 0;
+				break;
+			}
+		}
+		if (bo || bn)
+			bases_match = 0;
+
+		if (bases_match) {
+			pickme_tree = repo_get_commit_tree(repo, pickme);
+			parents = NULL;
+			commit_list_insert(replayed_par2, &parents);
+			commit_list_insert(replayed_par1, &parents);
+			picked = create_commit(repo, pickme_tree, pickme,
+					       parents, REPLAY_MODE_PICK);
+			goto out;
+		}
+	}
+
 	/*
 	 * R: auto-remerge of the original parents.
 	 *
@@ -408,10 +466,6 @@ static struct commit *pick_merge_commit(struct repository *repo,
 	remerge_opt.show_rename_progress = 0;
 	remerge_opt.branch1 = "ours";
 	remerge_opt.branch2 = "theirs";
-	if (repo_get_merge_bases(repo, parent1, parent2, &parent_bases) < 0) {
-		result->clean = -1;
-		goto out;
-	}
 	merge_incore_recursive(&remerge_opt, parent_bases,
 			       parent1, parent2, &remerge_res);
 	parent_bases = NULL; /* consumed by merge_incore_recursive */
@@ -425,11 +479,6 @@ static struct commit *pick_merge_commit(struct repository *repo,
 	new_merge_opt.show_rename_progress = 0;
 	new_merge_opt.branch1 = "ours";
 	new_merge_opt.branch2 = "theirs";
-	if (repo_get_merge_bases(repo, replayed_par1, replayed_par2,
-				 &replayed_bases) < 0) {
-		result->clean = -1;
-		goto out;
-	}
 	merge_incore_recursive(&new_merge_opt, replayed_bases,
 			       replayed_par1, replayed_par2, &new_merge_res);
 	replayed_bases = NULL; /* consumed by merge_incore_recursive */
-- 
gitgitgadget

[PATCH/RFC 3/5] history.adoc: describe merge-replay support and its limits

[Johannes Schindelin via GitGitGadget]

From: Johannes Schindelin <johannes.schindelin@gmx.de>

Replace the blanket "does not (yet) work with histories that contain
merges" caveat now that 2-parent merges are supported via the R/O/N
algorithm. Spell out what works (the user's manual conflict
resolution and any semantic edits inside the merge are preserved
through the replay), what is intentionally out of scope (octopus
merges; symbol-level extrapolation when rewriting parents pulls in
genuinely new content), and what still requires interactive rebase
(merges that would actually conflict on replay).

Assisted-by: Claude Opus 4.7
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
---
 Documentation/git-history.adoc | 27 ++++++++++++++++++++-------
 1 file changed, 20 insertions(+), 7 deletions(-)

diff --git a/Documentation/git-history.adoc b/Documentation/git-history.adoc
index 24dc907033..56328a7c59 100644
--- a/Documentation/git-history.adoc
+++ b/Documentation/git-history.adoc
@@ -40,13 +40,26 @@ at once.
 LIMITATIONS
 -----------
 
-This command does not (yet) work with histories that contain merges. You
-should use linkgit:git-rebase[1] with the `--rebase-merges` flag instead.
-
-Furthermore, the command does not support operations that can result in merge
-conflicts. This limitation is by design as history rewrites are not intended to
-be stateful operations. The limitation can be lifted once (if) Git learns about
-first-class conflicts.
+This command supports two-parent merge commits in the rewrite path:
+the auto-remerged tree of the original parents, the merge commit
+itself, and the auto-merged tree of the rewritten parents are
+combined so that the user's manual conflict resolution (textual or
+semantic) is preserved through the replay. Octopus merges (more than
+two parents) are not supported and are rejected with an error.
+
+The replay propagates the textual diffs the user actually made in
+the merge commit. It does _not_ extrapolate symbol-level intent: if
+rewriting the parents pulls in genuinely new content (for example, a
+new caller of a function that the merge renamed), that new content
+is _not_ rewritten by the replay and may need a follow-up edit.
+Symbol-aware refactoring is out of scope here, just as it is for
+plain rebase.
+
+The command does not support operations that can result in merge
+conflicts on the replayed merge itself. This limitation is by design
+as history rewrites are not intended to be stateful operations. Use
+linkgit:git-rebase[1] with the `--rebase-merges` flag when the
+rewrite is expected to require interactive conflict resolution.
 
 COMMANDS
 --------
-- 
gitgitgadget

[PATCH/RFC 4/5] test-tool: add a "historian" subcommand for building merge fixtures

[Johannes Schindelin via GitGitGadget]

From: Johannes Schindelin <johannes.schindelin@gmx.de>

The merge-replay tests added in a follow-up commit need a way to set
up specific topologies with full control over blob contents, parent
order, and per-side trees. Sequencing plumbing commands or driving
plain `git fast-import` from shell quickly becomes unreadable for
the kinds of scenarios that exercise non-trivial merge resolution
(textual conflicts, semantic edits outside the conflict region,
intentional limitations such as new content on one side).

Add a small `test-tool historian` subcommand that reads a tight,
shell-quoted, one-line-per-object DSL and feeds an equivalent stream
to a `git fast-import` child process. Each blob and commit is given
a logical name; the helper allocates fast-import marks on first use
and emits a lightweight tag for every commit so tests can refer to
the resulting object via `refs/tags/<name>`.

The DSL has just two directives:

  blob NAME LINE...
  commit NAME BRANCH SUBJECT [from=NAME] [merge=NAME]... [PATH=BLOB]...

A blob's content is the listed lines joined with `\n` (and a final
`\n`); a commit's tree is exactly the listed PATH=BLOB pairs (the
helper emits a `deleteall` so nothing leaks in from the implicit
parent). Token splitting is delegated to `split_cmdline()` so quoted
arguments work as in shell. Marks for parent references and file
contents go through the same `strintmap`-backed name resolver, which
keeps the helper itself trivially small: blob writing, tree
construction, commit creation and merge-base computation are all
handled by `git fast-import`.

Note that the DSL reserves the names `from` and `merge` (with a
trailing `=`) for parent specification; a tree path called `from` or
`merge` cannot be expressed via this helper. That is acceptable here
because every input is a tightly controlled test fixture and the
filenames are chosen by the test author.

The helper trusts its caller: malformed input results in a
fast-import error rather than a friendly diagnostic.

Wire the new subcommand into the Makefile and meson build, register
it in `t/helper/test-tool.{c,h}`.

Assisted-by: Claude Opus 4.7
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
---
 Makefile                  |   1 +
 t/helper/meson.build      |   1 +
 t/helper/test-historian.c | 189 ++++++++++++++++++++++++++++++++++++++
 t/helper/test-tool.c      |   1 +
 t/helper/test-tool.h      |   1 +
 5 files changed, 193 insertions(+)
 create mode 100644 t/helper/test-historian.c

diff --git a/Makefile b/Makefile
index cedc234173..b38678b484 100644
--- a/Makefile
+++ b/Makefile
@@ -832,6 +832,7 @@ TEST_BUILTINS_OBJS += test-hash-speed.o
 TEST_BUILTINS_OBJS += test-hash.o
 TEST_BUILTINS_OBJS += test-hashmap.o
 TEST_BUILTINS_OBJS += test-hexdump.o
+TEST_BUILTINS_OBJS += test-historian.o
 TEST_BUILTINS_OBJS += test-json-writer.o
 TEST_BUILTINS_OBJS += test-lazy-init-name-hash.o
 TEST_BUILTINS_OBJS += test-match-trees.o
diff --git a/t/helper/meson.build b/t/helper/meson.build
index 675e64c010..704edd1e1f 100644
--- a/t/helper/meson.build
+++ b/t/helper/meson.build
@@ -29,6 +29,7 @@ test_tool_sources = [
   'test-hash.c',
   'test-hashmap.c',
   'test-hexdump.c',
+  'test-historian.c',
   'test-json-writer.c',
   'test-lazy-init-name-hash.c',
   'test-match-trees.c',
diff --git a/t/helper/test-historian.c b/t/helper/test-historian.c
new file mode 100644
index 0000000000..2250d420c0
--- /dev/null
+++ b/t/helper/test-historian.c
@@ -0,0 +1,189 @@
+/*
+ * Build a small history out of a tiny declarative input. Used by tests
+ * that need specific merge topologies without long sequences of
+ * plumbing commands or fragile shell helpers.
+ *
+ * The historian reads stdin line by line and emits an equivalent
+ * stream to a `git fast-import` child process. It also allocates marks
+ * for named objects so tests can refer to commits and blobs by name.
+ *
+ * Input directives (one per line, shell-style quoting):
+ *
+ *	blob NAME LINE1 LINE2 ...
+ *	    Each LINE becomes a content line in the blob; lines are
+ *	    joined with '\n' and the blob ends with a final '\n'. With
+ *	    no LINEs, the blob is empty.
+ *
+ *	commit NAME BRANCH SUBJECT [from=PARENT] [merge=PARENT]... [PATH=BLOB]...
+ *	    Creates a commit on refs/heads/BRANCH using the listed
+ *	    file=blob mappings as the entire tree (no inheritance from
+ *	    parents). Up to one `from=` and any number of `merge=`
+ *	    parents may be given. `from=` defaults to the current branch
+ *	    tip; if BRANCH has no tip yet, the commit becomes a root.
+ *
+ * Each `commit NAME` directive also creates a lightweight tag
+ * `refs/tags/NAME` so tests can `git rev-parse NAME`.
+ *
+ * This helper trusts its caller; malformed input results in fast-import
+ * errors. That is fine because test scripts feed it tightly controlled
+ * input.
+ */
+
+#define USE_THE_REPOSITORY_VARIABLE
+
+#include "test-tool.h"
+#include "git-compat-util.h"
+#include "alias.h"
+#include "run-command.h"
+#include "setup.h"
+#include "strbuf.h"
+#include "strmap.h"
+#include "strvec.h"
+
+static int next_mark = 1;
+
+static int resolve_mark(struct strintmap *names, const char *name)
+{
+	int n = strintmap_get(names, name);
+	if (!n) {
+		n = next_mark++;
+		strintmap_set(names, name, n);
+	}
+	return n;
+}
+
+static void emit_data(FILE *out, const char *data, size_t len)
+{
+	fprintf(out, "data %"PRIuMAX"\n", (uintmax_t)len);
+	fwrite(data, 1, len, out);
+	fputc('\n', out);
+}
+
+static void emit_blob(FILE *out, struct strintmap *names,
+		      int argc, const char **argv)
+{
+	struct strbuf content = STRBUF_INIT;
+	int n = resolve_mark(names, argv[1]);
+	int i;
+
+	for (i = 2; i < argc; i++) {
+		strbuf_addstr(&content, argv[i]);
+		strbuf_addch(&content, '\n');
+	}
+
+	fprintf(out, "blob\nmark :%d\n", n);
+	emit_data(out, content.buf, content.len);
+	strbuf_release(&content);
+}
+
+static void emit_tag(FILE *out, const char *name, int mark)
+{
+	fprintf(out, "reset refs/tags/%s\nfrom :%d\n\n", name, mark);
+}
+
+static void emit_commit(FILE *out, struct strintmap *names,
+			int argc, const char **argv, int seq)
+{
+	int n = resolve_mark(names, argv[1]);
+	const char *branch = argv[2];
+	const char *subject = argv[3];
+	const char *rest;
+	int i;
+
+	fprintf(out, "commit refs/heads/%s\nmark :%d\n", branch, n);
+	fprintf(out, "author A <a@e> %d +0000\n", 1700000000 + seq);
+	fprintf(out, "committer A <a@e> %d +0000\n", 1700000000 + seq);
+	emit_data(out, subject, strlen(subject));
+
+	/*
+	 * fast-import requires `from` and `merge` to precede all file
+	 * operations; emit them first regardless of argv ordering.
+	 */
+	for (i = 4; i < argc; i++) {
+		if (skip_prefix(argv[i], "from=", &rest))
+			fprintf(out, "from :%d\n", resolve_mark(names, rest));
+		else if (skip_prefix(argv[i], "merge=", &rest))
+			fprintf(out, "merge :%d\n", resolve_mark(names, rest));
+	}
+
+	/*
+	 * The PATH=BLOB list is the entire tree; wipe whatever the
+	 * implicit parent contributed before re-applying it.
+	 */
+	fprintf(out, "deleteall\n");
+	for (i = 4; i < argc; i++) {
+		const char *eq;
+		size_t key_len;
+		char *path;
+
+		if (skip_prefix(argv[i], "from=", &rest) ||
+		    skip_prefix(argv[i], "merge=", &rest))
+			continue;
+		eq = strchr(argv[i], '=');
+		if (!eq)
+			die("bad commit spec '%s'", argv[i]);
+		key_len = eq - argv[i];
+		path = xmemdupz(argv[i], key_len);
+		fprintf(out, "M 100644 :%d %s\n",
+			resolve_mark(names, eq + 1), path);
+		free(path);
+	}
+
+	fputc('\n', out);
+	emit_tag(out, argv[1], n);
+}
+
+int cmd__historian(int argc, const char **argv UNUSED)
+{
+	struct child_process fi = CHILD_PROCESS_INIT;
+	struct strintmap names = STRINTMAP_INIT;
+	struct strbuf line = STRBUF_INIT;
+	int seq = 0;
+	int ret = 0;
+	FILE *fi_in;
+
+	if (argc != 1)
+		die("usage: test-tool historian <input");
+
+	setup_git_directory();
+
+	strvec_pushl(&fi.args, "fast-import", "--quiet", "--force", NULL);
+	fi.git_cmd = 1;
+	fi.in = -1;
+	fi.no_stdout = 1;
+	if (start_command(&fi))
+		die("failed to start git fast-import");
+	fi_in = xfdopen(fi.in, "w");
+
+	while (strbuf_getline_lf(&line, stdin) != EOF) {
+		const char **a = NULL;
+		int n;
+
+		strbuf_trim(&line);
+		if (!line.len || line.buf[0] == '#')
+			continue;
+
+		n = split_cmdline(line.buf, &a);
+		if (n < 0)
+			die("split_cmdline failed: %s",
+			    split_cmdline_strerror(n));
+
+		if (n >= 2 && !strcmp(a[0], "blob"))
+			emit_blob(fi_in, &names, n, a);
+		else if (n >= 4 && !strcmp(a[0], "commit"))
+			emit_commit(fi_in, &names, n, a, seq++);
+		else
+			die("unknown directive: %s", a[0]);
+
+		free(a);
+	}
+
+	if (fclose(fi_in))
+		die_errno("close fast-import stdin");
+	if (finish_command(&fi))
+		ret = 1;
+
+	strbuf_release(&line);
+	strintmap_clear(&names);
+	return ret;
+}
diff --git a/t/helper/test-tool.c b/t/helper/test-tool.c
index a7abc618b3..28bde98ce1 100644
--- a/t/helper/test-tool.c
+++ b/t/helper/test-tool.c
@@ -39,6 +39,7 @@ static struct test_cmd cmds[] = {
 	{ "hashmap", cmd__hashmap },
 	{ "hash-speed", cmd__hash_speed },
 	{ "hexdump", cmd__hexdump },
+	{ "historian", cmd__historian },
 	{ "json-writer", cmd__json_writer },
 	{ "lazy-init-name-hash", cmd__lazy_init_name_hash },
 	{ "match-trees", cmd__match_trees },
diff --git a/t/helper/test-tool.h b/t/helper/test-tool.h
index 7f150fa1eb..78cec8594a 100644
--- a/t/helper/test-tool.h
+++ b/t/helper/test-tool.h
@@ -32,6 +32,7 @@ int cmd__getcwd(int argc, const char **argv);
 int cmd__hashmap(int argc, const char **argv);
 int cmd__hash_speed(int argc, const char **argv);
 int cmd__hexdump(int argc, const char **argv);
+int cmd__historian(int argc, const char **argv);
 int cmd__json_writer(int argc, const char **argv);
 int cmd__lazy_init_name_hash(int argc, const char **argv);
 int cmd__match_trees(int argc, const char **argv);
-- 
gitgitgadget

[PATCH/RFC 5/5] t3454: cover merge-replay scenarios with the historian helper

[Johannes Schindelin via GitGitGadget]

From: Johannes Schindelin <johannes.schindelin@gmx.de>

Add a dedicated test script for `git history reword` (and
`git replay` via the same code path) across 2-parent merges, using
the `test-tool historian` fixture builder so each scenario reads as
a small declarative recipe rather than a sequence of plumbing
commands.

The script exercises the cases that motivated the merge-replay
work:

  * a clean merge where each side touches unrelated files;

  * a non-trivial merge where the same line was changed on both
    sides and the user resolved by hand (textual manual resolution
    must be preserved through the replay);

  * a non-trivial merge where the user also touched a line outside
    any conflict region (a "semantic" edit must also be preserved
    through the replay);

  * an octopus merge in the rewrite path, which is rejected;

  * a function rename across the merge with a brand-new caller
    introduced by the rewritten parents. The pre-existing caller
    that the user manually renamed in the original merge must keep
    its rename, and the brand-new caller must _not_ be rewritten
    (calvin/hobbes naming chosen for legibility). This second part
    is the documented limitation: the replay propagates the textual
    diffs the user actually made, it does not extrapolate
    symbol-level intent. Symbol-aware refactoring is out of scope,
    just as it is for plain rebase.

The fixture builder lets each scenario sit in roughly a dozen lines
of historian directives plus the assertions, which keeps the test
file readable when more scenarios are added later.

Assisted-by: Claude Opus 4.7
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
---
 t/meson.build             |   1 +
 t/t3454-history-merges.sh | 308 ++++++++++++++++++++++++++++++++++++++
 2 files changed, 309 insertions(+)
 create mode 100755 t/t3454-history-merges.sh

diff --git a/t/meson.build b/t/meson.build
index 7528e5cda5..25b0119d43 100644
--- a/t/meson.build
+++ b/t/meson.build
@@ -397,6 +397,7 @@ integration_tests = [
   't3450-history.sh',
   't3451-history-reword.sh',
   't3452-history-split.sh',
+  't3454-history-merges.sh',
   't3500-cherry.sh',
   't3501-revert-cherry-pick.sh',
   't3502-cherry-pick-merge.sh',
diff --git a/t/t3454-history-merges.sh b/t/t3454-history-merges.sh
new file mode 100755
index 0000000000..2eb3c947eb
--- /dev/null
+++ b/t/t3454-history-merges.sh
@@ -0,0 +1,308 @@
+#!/bin/sh
+
+test_description='git history reword across merge commits
+
+Exercises the merge-replay path in `git history reword` using the
+`test-tool historian` test fixture builder so each scenario is
+described in a small declarative input rather than a sprawling
+sequence of plumbing commands. The interesting cases are:
+
+  * a clean merge with each side touching unrelated files;
+  * a non-trivial merge whose conflicting line was resolved by hand
+    (textually) and whose resolution must be preserved through the
+    replay;
+  * a non-trivial merge with a manual *semantic* edit (an additional
+    change outside the conflict region) that must also be preserved.
+'
+
+GIT_TEST_DEFAULT_INITIAL_BRANCH_NAME=main
+export GIT_TEST_DEFAULT_INITIAL_BRANCH_NAME
+
+. ./test-lib.sh
+
+# Replace the commit's message via a fake editor and run reword.
+reword_to () {
+	new_msg="$1"
+	target="$2"
+	write_script fake-editor.sh <<-EOF &&
+	echo "$new_msg" >"\$1"
+	EOF
+	test_set_editor "$(pwd)/fake-editor.sh" &&
+	git history reword "$target" &&
+	rm fake-editor.sh
+}
+
+build_clean_merge () {
+	test-tool historian <<-\EOF
+	# Setup:
+	#       A (a) --- C (a, h) ----+--- M (a, g, h)
+	#        \                    /
+	#         +-- B (a, g) ------+
+	#
+	# Topic touches `g` only; main touches `h` only. The auto-merge
+	# at M is clean.
+	blob a "shared content"
+	blob g guarded
+	blob h host
+	commit A main "A" a=a
+	commit B topic "B (introduces g)" from=A a=a g=g
+	commit C main "C (introduces h)" a=a h=h
+	commit M main "Merge topic" merge=B a=a g=g h=h
+	EOF
+}
+
+test_expect_success 'clean merge: both sides touch unrelated files' '
+	test_when_finished "rm -rf repo" &&
+	git init repo &&
+	(
+		cd repo &&
+		build_clean_merge &&
+
+		reword_to "AA" A &&
+
+		# The merge is still a 2-parent merge with the same subject
+		# and tree (clean replay leaves content unchanged).
+		test_cmp_rev HEAD^{tree} M^{tree} &&
+
+		echo "Merge topic" >expect-subject &&
+		git log -1 --format=%s HEAD >subject &&
+		test_cmp expect-subject subject &&
+
+		git rev-list --merges HEAD~..HEAD >merges &&
+		test_line_count = 1 merges
+	)
+'
+
+build_textual_resolution () {
+	test-tool historian <<-\EOF
+	# Both sides change the same line of `a`; the user resolved with
+	# their own combined text, recorded directly as the merge tree.
+	blob a_v1 line1 line2 line3
+	blob a_main line1 line2-main line3
+	blob a_topic line1 line2-topic line3
+	blob a_resolution line1 line2-merged-by-hand line3
+	commit A main "A" a=a_v1
+	commit B topic "B (line2 on topic)" from=A a=a_topic
+	commit C main "C (line2 on main)" a=a_main
+	commit M main "Merge topic" merge=B a=a_resolution
+	EOF
+}
+
+test_expect_success 'non-trivial merge: textual manual resolution is preserved' '
+	test_when_finished "rm -rf repo" &&
+	git init repo &&
+	(
+		cd repo &&
+		build_textual_resolution &&
+
+		reword_to "AA" A &&
+
+		git show HEAD:a >after &&
+		test_write_lines line1 line2-merged-by-hand line3 >expect &&
+		test_cmp expect after
+	)
+'
+
+build_semantic_edit () {
+	test-tool historian <<-\EOF
+	# Topic and main conflict on line2 of `a`. The user's resolution
+	# at M not only picks combined text on line2 but ALSO touches
+	# line5 (a "semantic" edit outside any conflict region) -- this
+	# kind of edit is invisible to a naive pick-one-side strategy and
+	# must be preserved by replay.
+	blob a_v1 line1 line2 line3 line4 line5
+	blob a_main line1 line2-main line3 line4 line5
+	blob a_topic line1 line2-topic line3 line4 line5
+	blob a_resolution line1 line2-merged line3 line4 line5-touched
+	commit A main "A" a=a_v1
+	commit B topic "B (line2 on topic)" from=A a=a_topic
+	commit C main "C (line2 on main)" a=a_main
+	commit M main "Merge topic" merge=B a=a_resolution
+	EOF
+}
+
+test_expect_success 'non-trivial merge: semantic edit outside conflict region is preserved' '
+	test_when_finished "rm -rf repo" &&
+	git init repo &&
+	(
+		cd repo &&
+		build_semantic_edit &&
+
+		reword_to "AA" A &&
+
+		git show HEAD:a >after &&
+		test_write_lines line1 line2-merged line3 line4 line5-touched \
+			>expect &&
+		test_cmp expect after
+	)
+'
+
+build_octopus () {
+	test-tool historian <<-\EOF
+	blob a "x"
+	commit A main "A" a=a
+	commit B b1 "B" from=A a=a
+	commit C b2 "C" from=A a=a
+	commit D b3 "D" from=A a=a
+	commit O main "octopus" merge=B merge=C merge=D a=a
+	EOF
+}
+
+test_expect_success 'octopus merge in the rewrite path is rejected' '
+	test_when_finished "rm -rf repo" &&
+	git init repo &&
+	(
+		cd repo &&
+		build_octopus &&
+
+		test_must_fail git -c core.editor=true history reword \
+			--dry-run A 2>err &&
+		test_grep "octopus" err
+	)
+'
+
+build_with_boundary_other_than_onto () {
+	test-tool historian <<-\EOF
+	# Setup an "evil merge" topology where the rewrite range crosses
+	# a 2-parent merge whose first parent sits outside that range:
+	#
+	#   side -- O (a=v0)
+	#            \
+	#             M (parent1=O, parent2=R, a=v0, s=top)
+	#            /
+	#   A (a=v0) -- R (a=v0) -- T (a=v0, s=top)
+	#       |
+	#       reword target
+	#
+	# The walk for `history reword A` excludes A and its ancestors,
+	# so O sits outside the rewrite range and is not the boundary
+	# either. Replaying M correctly requires that first parent to
+	# remain at O (preserve, not replant).
+	blob v0 line1 line2 line3
+	blob top "marker"
+	commit X side "X" v0=v0
+	commit O side "O" v0=v0
+	commit A main "A" from=X v0=v0
+	commit R main "R" v0=v0
+	commit M main "Merge side into main" from=O merge=R v0=v0 s=top
+	commit T main "T" v0=v0 s=top
+	EOF
+}
+
+# A descendant merge whose first parent sits outside the rewrite
+# range is a topology that any reasonable replay of merges has to
+# handle correctly: the first parent must be preserved verbatim,
+# while the in-range second parent is rewritten. Without that, the
+# replayed merge would silently graft itself onto a different
+# ancestry than the author chose, which is far worse than a loud
+# failure.
+test_expect_success 'merge whose first parent sits outside the rewrite range keeps that parent' '
+	test_when_finished "rm -rf repo" &&
+	git init repo &&
+	(
+		cd repo &&
+		build_with_boundary_other_than_onto &&
+
+		reword_to "AA" A &&
+
+		# The replayed M (now HEAD~) is still a 2-parent merge.
+		# Its first parent is the original O (preserved, outside
+		# the rewrite range), its second parent is the rewritten
+		# R. T was rebased on top of M, so HEAD = T.
+		git rev-list --parents -1 HEAD~ >parents &&
+		new_p1=$(awk "{print \$2}" parents) &&
+		new_p2=$(awk "{print \$3}" parents) &&
+
+		# First parent is preserved verbatim.
+		test_cmp_rev O $new_p1 &&
+
+		# Second parent is the rewritten R: a fresh commit whose
+		# subject is still "R" but whose OID differs from the
+		# original (because its parent A is now reworded).
+		echo R >expect &&
+		git log -1 --format=%s $new_p2 >actual &&
+		test_cmp expect actual &&
+		! test_cmp_rev R $new_p2 &&
+
+		# T was rebased on top of the new M, and its tree still
+		# contains the s=top marker introduced in the original M.
+		echo "marker" >expect &&
+		git show HEAD:s >actual &&
+		test_cmp expect actual
+	)
+'
+
+build_function_rename () {
+	test-tool historian <<-\EOF
+	# Topic renames harry() -> hermione() (defs.h plus caller1). main
+	# adds caller2 calling harry(); the original merge M manually
+	# renames caller2 to hermione(). The "newer" base on a side branch
+	# contains caller2 AND a brand-new caller3 calling harry();
+	# replaying onto `newer` therefore introduces caller3 into the
+	# merged tree.
+	blob defs_harry "void harry(void);"
+	blob defs_hermione "void hermione(void);"
+	blob harry_call "harry();"
+	blob hermione_call "hermione();"
+	commit A main "A" defs.h=defs_harry caller1=harry_call
+	commit B topic "B (rename)" from=A defs.h=defs_hermione caller1=hermione_call
+	commit C main "C (caller2 calls harry)" defs.h=defs_harry caller1=harry_call caller2=harry_call
+	commit M main "Merge topic" merge=B defs.h=defs_hermione caller1=hermione_call caller2=hermione_call
+	commit NEW newer "newer base with caller3" from=A defs.h=defs_harry caller1=harry_call caller2=harry_call caller3=harry_call
+	EOF
+}
+
+# This case checks two things at once. First, the manual semantic
+# edit in M (renaming caller2) must be preserved when we replay onto
+# a different base; that is the case `git history` and `git replay`
+# need to handle correctly, even though nothing in the conflict
+# markers tells us about it. Second, a file that only enters the
+# tree via the rewritten parents (caller3, present on the `newer`
+# base) is _not_ renamed by the replay. The replay propagates the
+# textual diffs the user actually made in M; it does _not_ infer
+# the user's symbol-level intent ("rename every caller of harry").
+# This is a known and intentional limitation. Symbol-aware
+# refactoring is out of scope here, just as it is for plain rebase.
+test_expect_success 'preserves manual rename of pre-existing caller; does not extrapolate to new files' '
+	test_when_finished "rm -rf repo" &&
+	git init repo &&
+	(
+		cd repo &&
+		build_function_rename &&
+
+		# Replay (C, B, M) onto the newer base. A `main..M` style
+		# range across two unrelated branches is awkward; spin up a
+		# temp branch and use --advance.
+		git branch tmp main &&
+		git replay --ref-action=print --onto NEW A..tmp >result &&
+		new_tip=$(cut -f 3 -d " " result) &&
+
+		# defs.h and caller1 came from B (clean cherry-pick of the
+		# rename commit) and must reflect the rename.
+		echo "void hermione(void);" >expect &&
+		git show $new_tip:defs.h >actual &&
+		test_cmp expect actual &&
+
+		echo "hermione();" >expect &&
+		git show $new_tip:caller1 >actual &&
+		test_cmp expect actual &&
+
+		# caller2 existed in the original M; its manual rename to
+		# hermione() is the semantic edit the replay must preserve.
+		echo "hermione();" >expect &&
+		git show $new_tip:caller2 >actual &&
+		test_cmp expect actual &&
+
+		# caller3 only exists on the newer base, so it was brought
+		# in by N (the auto-merge of the rewritten parents). The
+		# replay has no way to know the user intended to rename
+		# every caller; caller3 keeps harry(). The resulting tree
+		# is therefore _not_ symbol-correct and needs a follow-up
+		# edit. This is the documented limitation.
+		echo "harry();" >expect &&
+		git show $new_tip:caller3 >actual &&
+		test_cmp expect actual
+	)
+'
+
+test_done
-- 
gitgitgadget

Re: [PATCH/RFC 0/5] replay: support replaying 2-parent merges

[D. Ben Knoble]

Hi Dscho,

On Wed, May 6, 2026 at 6:44 PM Johannes Schindelin via GitGitGadget
<gitgitgadget@gmail.com> wrote:
>
> git history, the new history-rewriting builtin in v2.54, dies on any merge
> in the rewrite path with replaying merge commits is not supported yet!. That
> makes it not very useful for the workflows I actually have, where almost
> every interesting branch contains at least one merge of a feature topic. The
> natural fallback, git rebase --rebase-merges, is interactive and stops to
> ask for re-resolution even when no re-resolution is needed.
>
> This series lifts that limitation for the common 2-parent case. The
> algorithm itself is not new: Elijah Newren wrote it down in his replay
> design notes
> [https://github.com/newren/git/blob/replay/replay-design-notes.txt] and
> prototyped it in a 2022 work-in-progress sketch
> [https://github.com/newren/git/commit/4c45e8955ef9bf7d01fd15d9106b3bdb8ea91b45].
> What is new is wiring it into the replay_revisions() API that backs both git
> replay and git history, plus three specific tweaks that make the trickier
> cases work where the WIP sketch bailed out: identical conflict-marker labels
> for the inner remerges of the original and the rewritten parents (so their
> conflict-markered trees compare equal in the regions the user did not
> touch), tolerating result.clean == 0 from those inner merges (their
> well-defined conflict-markered trees are valid inputs to the outer 3-way
> merge), and self-fallback for both merge parents combined with mapping the
> rev-range boundary commits to the onto commit.
>
> Octopus merges and revert-of-merge are surfaced as explicit errors at the
> dispatch point. The split sub-command of git history continues to refuse
> when its target is a merge: split semantics simply do not apply there. The
> xdiff special mode for matching conflict-marker hunks across inner remerges,
> the XDL_MERGE_FAVOR_BASE variant, and the modify/delete and binary-file
> specials that the design notes flag as future work all remain future work.
>
> While I was at it, git history reword had a pre-existing silent-success bug:
> a positive return from replay_revisions() (which means "conflict, no updates
> queued") was treated as success. Obviously this should never occur, as a
> reword simply does not change any of the file contents, but bugs do happen.
> The merge-replay work is complex enough to make that class of bugs more
> likely, therefore I introduce error messages for those instances.

Fixing this bug sounded interesting; I had a hard time spotting it
while skimming the first 2 patches.

Did I just miss it? Is it worth splitting that fix out to a separate patch?

Best,
-- 
D. Ben Knoble

Re: [PATCH/RFC 0/5] replay: support replaying 2-parent merges

[Johannes Schindelin]

[-- Attachment #1: Type: text/plain, Size: 2343 bytes --]

Hi Ben,

On Thu, 7 May 2026, D. Ben Knoble wrote:

> On Wed, May 6, 2026 at 6:44 PM Johannes Schindelin via GitGitGadget
> <gitgitgadget@gmail.com> wrote:
> >
> > [...]
> >
> > While I was at it, git history reword had a pre-existing
> > silent-success bug: a positive return from replay_revisions() (which
> > means "conflict, no updates queued") was treated as success. Obviously
> > this should never occur, as a reword simply does not change any of the
> > file contents, but bugs do happen. The merge-replay work is complex
> > enough to make that class of bugs more likely, therefore I introduce
> > error messages for those instances.
> 
> Fixing this bug sounded interesting; I had a hard time spotting it
> while skimming the first 2 patches.

It's this part:

@@ -482,6 +482,9 @@ static int cmd_history_reword(int argc,
 	if (ret < 0) {
 		ret = error(_("failed replaying descendants"));
 		goto out;
+	} else if (ret) {
+		ret = error(_("conflict during replay; some descendants were not rewritten"));
+		goto out;
 	}
 
 	ret = 0;
@@ -721,6 +724,9 @@ static int cmd_history_split(int argc,
 	if (ret < 0) {
 		ret = error(_("failed replaying descendants"));
 		goto out;
+	} else if (ret) {
+		ret = error(_("conflict during replay; some descendants were not rewritten"));
+		goto out;
 	}
 
 	ret = 0;

> Did I just miss it? Is it worth splitting that fix out to a separate patch?

Well, you _could_ argue that they were not bugs at all: a `git history
reword` isn't supposed to be able to result in merge conflicts, nor is
`git history split` because they leave the respective commits tree-same
(in the `split` case, the second commit).

I could see the point were anybody to suggest using `BUG()` instead of
`error()` here, but erred on the "nicer to the user" side.

The only way this _might_ be triggered before this patch series is most
likely by playing games with replace objects. Or maybe you cannot trigger
it at all.

With the changes in this here patch series, I wasn't so certain that I had
covered all the edge cases (an early iteration of the quick short-cut in
patch 2/5 keyed only on the parent commits' trees, and forgot to verify
the merge _bases_' trees, for example). That's why I think it matters more
now than it did before.

Ciao,
Johannes

Re: [PATCH/RFC 0/5] replay: support replaying 2-parent merges

[Ben Knoble]

> 
> Le 7 mai 2026 à 11:06, Johannes Schindelin <johannes.schindelin@gmx.de> a écrit :
> 
> Hi Ben,
> 
>> On Thu, 7 May 2026, D. Ben Knoble wrote:
>> 
>>> On Wed, May 6, 2026 at 6:44 PM Johannes Schindelin via GitGitGadget
>>> <gitgitgadget@gmail.com> wrote:
>>> 
>>> [...]
>>> 
>>> While I was at it, git history reword had a pre-existing
>>> silent-success bug: a positive return from replay_revisions() (which
>>> means "conflict, no updates queued") was treated as success. Obviously
>>> this should never occur, as a reword simply does not change any of the
>>> file contents, but bugs do happen. The merge-replay work is complex
>>> enough to make that class of bugs more likely, therefore I introduce
>>> error messages for those instances.
>> 
>> Fixing this bug sounded interesting; I had a hard time spotting it
>> while skimming the first 2 patches.
> 
> It's this part:
> 
> @@ -482,6 +482,9 @@ static int cmd_history_reword(int argc,
>    if (ret < 0) {
>        ret = error(_("failed replaying descendants"));
>        goto out;
> +    } else if (ret) {
> +        ret = error(_("conflict during replay; some descendants were not rewritten"));
> +        goto out;
>    }
> 
>    ret = 0;
> @@ -721,6 +724,9 @@ static int cmd_history_split(int argc,
>    if (ret < 0) {
>        ret = error(_("failed replaying descendants"));
>        goto out;
> +    } else if (ret) {
> +        ret = error(_("conflict during replay; some descendants were not rewritten"));
> +        goto out;
>    }
> 
>    ret = 0;

Thanks, super helpful.

(Perhaps later) if we can say _which_ descendants weren’t rewritten, that might be good.

>> Did I just miss it? Is it worth splitting that fix out to a separate patch?
> 
> Well, you _could_ argue that they were not bugs at all: a `git history
> reword` isn't supposed to be able to result in merge conflicts, nor is
> `git history split` because they leave the respective commits tree-same
> (in the `split` case, the second commit).

I seem to recall Patrick making a similar argument, but don’t let me put words in anyone’s mouth. 

> I could see the point were anybody to suggest using `BUG()` instead of
> `error()` here, but erred on the "nicer to the user" side.
> 
> The only way this _might_ be triggered before this patch series is most
> likely by playing games with replace objects. Or maybe you cannot trigger
> it at all.
> 
> With the changes in this here patch series, I wasn't so certain that I had
> covered all the edge cases (an early iteration of the quick short-cut in
> patch 2/5 keyed only on the parent commits' trees, and forgot to verify
> the merge _bases_' trees, for example). That's why I think it matters more
> now than it did before.
> 
> Ciao,
> Johannes

Makes sense, thanks.

Re: [PATCH/RFC 1/5] replay: support replaying 2-parent merges

[Phillip Wood]

Hi Johannes

On 06/05/2026 23:43, Johannes Schindelin via GitGitGadget wrote:
> 
> Elijah Newren spelled out a way to lift this limitation in his
> replay-design-notes [1] and prototyped it in a 2022
> work-in-progress sketch [2]. The idea is that a merge commit M on
> parents (P1, P2) records both an automatic merge of those parents
> AND any manual layer the author put on top of that automatic merge
> (textual conflict resolution and any semantic edit outside conflict
> markers). Replaying M onto rewritten parents (P1', P2') must
> preserve that manual layer, but the rewritten parents change the
> automatic merge, so a simple cherry-pick is wrong: the manual layer
> would be re-introduced on top of stale auto-merge text.
> 
> What works instead is a three-way merge of three trees the existing
> infrastructure already knows how to compute. Let R be the recursive
> auto-merge of (P1, P2), O be M's actual tree and N be the recursive
> auto-merge of (P1', P2'). Then `git diff R O` is morally
> `git show --remerge-diff M`: it captures exactly what the author
> added on top of the automatic merge. A non-recursive 3-way merge
> with R as the merge base, O as side 1 and N as side 2 layers that
> manual contribution onto the freshly auto-merged rewritten parents
> (N) and produces the replayed tree.

So we cherry-pick the difference between the user's conflict resolution 
O and the auto-merge M of the original parents onto the auto-merge N of 
the replayed parents. If we have a topology that looks like

         |
         A
        /|\
       / B \
      E  |  D
         C /
         |/
         O

then running

     git replay --onto E --ancestry-path B..O

will replay C and O onto E. If the changes in E and D conflict but those 
conflicts do not overlap with the conflicts in M that were resolved to 
create O then the replayed version of O will contain conflict markers 
from the conflicting changes in E and D. Because the previous conflict 
resolution applies to N without conflicts we do not recognize that there 
are still conflicts in N that need to be resolved.

Having realized this I went to look at Elijah's notes and they recognize 
this possibility and suggest extending the xdiff merge code to detect 
when N has conflicts that do not correspond to the conflicts in M. That 
sounds like quite a lot of work. I've not put much effort into coming up 
with a counterexample but think that because "git replay" and "git 
history" do not yet allow the commits in the merged branches to be 
edited we may be able to safely use the implementation proposed in this 
series if both merge parents have been rebased (or we might want all the 
merge bases of the new merge to be a descendants of "--onto"). In the 
example above if both the parents were rebased onto E then any new 
conflicts would happen when picking D rather than when recreating the merge.

Thanks

Phillip

> Implement `pick_merge_commit()` along those lines and dispatch to it
> from `replay_revisions()` when the commit being replayed has exactly
> two parents. Two specific points (learned the hard way) keep
> non-trivial cases working where the WIP sketch [2] bailed out.
> First, R and N use identical `merge_options.branch1` and `branch2`
> labels ("ours"/"theirs"). When the original parents conflicted on a
> region of a file, both R and N produce textually identical conflict
> markers; the outer non-recursive merge then sees N == R in that
> region and the user's manual resolution from O wins cleanly. Without
> this, the conflict-marker text would differ between R and N (because
> the inner merges would label the conflicts differently), and the
> outer merge would itself be unclean even when the user did supply a
> clean resolution. Second, an unclean inner merge
> (`result.clean == 0`) is _not_ fatal: the tree merge-ort produces in
> that case still has well-defined contents (with conflict markers in
> the conflicted files) and is a valid input to the outer
> non-recursive merge. Only a real error (`< 0`) propagates as
> failure.
> 
> The replay propagates the textual diffs the user actually made in M;
> it does _not_ extrapolate symbol-level intent. If rewriting the
> parents pulls in genuinely new content (for example, a brand-new
> caller of a function that the merge renamed), that new content stays
> as the rewritten parents have it. Symbol-aware refactoring is out of
> scope here, just as it is for plain rebase.
> 
> Octopus merges (more than two parents) and revert-of-merge are not
> supported and are surfaced as explicit errors at the dispatch point.
> The "split" sub-command of `git history` continues to refuse when
> the targeted commit is itself a merge: split semantics do not apply
> to merges. The pre-walk gate in `builtin/history.c` that previously
> rejected any merge in the rewrite path now only rejects octopus
> merges; rename it accordingly.
> 
> A small refactor in `create_commit()` makes the merge case possible:
> the helper now takes a `struct commit_list *parents` rather than a
> single parent pointer and takes ownership of the list. The single
> existing caller in `pick_regular_commit()` builds and passes a
> one-element list; the new `pick_merge_commit()` builds a two-element
> list, with the order of the `from` and `merge` parents preserved.
> 
> Update the negative expectations in t3451, t3452 and t3650 that were
> asserting the now-retired "not supported yet" message, replacing
> them with positive coverage where it fits. Octopus rejection and
> revert-of-merge rejection are covered by new positive tests in
> t3650. A dedicated test script with merge-replay scenarios driven by
> a new test-tool fixture builder will follow in a subsequent commit.
> 
> [1] https://github.com/newren/git/blob/replay/replay-design-notes.txt
> [2] https://github.com/newren/git/commit/4c45e8955ef9bf7d01fd15d9106b3bdb8ea91b45
> 
> Helped-by: Elijah Newren <newren@gmail.com>
> Assisted-by: Claude Opus 4.7
> Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
> ---
>   builtin/history.c         |  16 ++-
>   replay.c                  | 209 ++++++++++++++++++++++++++++++++++++--
>   t/t3451-history-reword.sh |  21 ++--
>   t/t3452-history-split.sh  |   6 +-
>   t/t3650-replay-basics.sh  |  46 ++++++++-
>   5 files changed, 269 insertions(+), 29 deletions(-)
> 
> diff --git a/builtin/history.c b/builtin/history.c
> index 9526938085..00097b2226 100644
> --- a/builtin/history.c
> +++ b/builtin/history.c
> @@ -195,15 +195,15 @@ static int parse_ref_action(const struct option *opt, const char *value, int uns
>   	return 0;
>   }
>   
> -static int revwalk_contains_merges(struct repository *repo,
> -				   const struct strvec *revwalk_args)
> +static int revwalk_contains_octopus_merges(struct repository *repo,
> +					   const struct strvec *revwalk_args)
>   {
>   	struct strvec args = STRVEC_INIT;
>   	struct rev_info revs;
>   	int ret;
>   
>   	strvec_pushv(&args, revwalk_args->v);
> -	strvec_push(&args, "--min-parents=2");
> +	strvec_push(&args, "--min-parents=3");
>   
>   	repo_init_revisions(repo, &revs, NULL);
>   
> @@ -217,7 +217,7 @@ static int revwalk_contains_merges(struct repository *repo,
>   	}
>   
>   	if (get_revision(&revs)) {
> -		ret = error(_("replaying merge commits is not supported yet!"));
> +		ret = error(_("replaying octopus merges is not supported"));
>   		goto out;
>   	}
>   
> @@ -289,7 +289,7 @@ static int setup_revwalk(struct repository *repo,
>   		strvec_push(&args, "HEAD");
>   	}
>   
> -	ret = revwalk_contains_merges(repo, &args);
> +	ret = revwalk_contains_octopus_merges(repo, &args);
>   	if (ret < 0)
>   		goto out;
>   
> @@ -482,6 +482,9 @@ static int cmd_history_reword(int argc,
>   	if (ret < 0) {
>   		ret = error(_("failed replaying descendants"));
>   		goto out;
> +	} else if (ret) {
> +		ret = error(_("conflict during replay; some descendants were not rewritten"));
> +		goto out;
>   	}
>   
>   	ret = 0;
> @@ -721,6 +724,9 @@ static int cmd_history_split(int argc,
>   	if (ret < 0) {
>   		ret = error(_("failed replaying descendants"));
>   		goto out;
> +	} else if (ret) {
> +		ret = error(_("conflict during replay; some descendants were not rewritten"));
> +		goto out;
>   	}
>   
>   	ret = 0;
> diff --git a/replay.c b/replay.c
> index f96f1f6551..3dbce095f9 100644
> --- a/replay.c
> +++ b/replay.c
> @@ -1,6 +1,7 @@
>   #define USE_THE_REPOSITORY_VARIABLE
>   
>   #include "git-compat-util.h"
> +#include "commit-reach.h"
>   #include "environment.h"
>   #include "hex.h"
>   #include "merge-ort.h"
> @@ -77,15 +78,21 @@ static void generate_revert_message(struct strbuf *msg,
>   	repo_unuse_commit_buffer(repo, commit, message);
>   }
>   
> +/*
> + * Build a new commit with the given tree and parent list, copying author,
> + * extra headers and (for pick mode) the commit message from `based_on`.
> + *
> + * Takes ownership of `parents`: it will be freed before returning, even on
> + * error. Parent order is preserved as supplied by the caller.
> + */
>   static struct commit *create_commit(struct repository *repo,
>   				    struct tree *tree,
>   				    struct commit *based_on,
> -				    struct commit *parent,
> +				    struct commit_list *parents,
>   				    enum replay_mode mode)
>   {
>   	struct object_id ret;
>   	struct object *obj = NULL;
> -	struct commit_list *parents = NULL;
>   	char *author = NULL;
>   	char *sign_commit = NULL; /* FIXME: cli users might want to sign again */
>   	struct commit_extra_header *extra = NULL;
> @@ -96,7 +103,6 @@ static struct commit *create_commit(struct repository *repo,
>   	const char *orig_message = NULL;
>   	const char *exclude_gpgsig[] = { "gpgsig", "gpgsig-sha256", NULL };
>   
> -	commit_list_insert(parent, &parents);
>   	extra = read_commit_extra_headers(based_on, exclude_gpgsig);
>   	if (mode == REPLAY_MODE_REVERT) {
>   		generate_revert_message(&msg, based_on, repo);
> @@ -273,6 +279,7 @@ static struct commit *pick_regular_commit(struct repository *repo,
>   {
>   	struct commit *base, *replayed_base;
>   	struct tree *pickme_tree, *base_tree, *replayed_base_tree;
> +	struct commit_list *parents = NULL;
>   
>   	if (pickme->parents) {
>   		base = pickme->parents->item;
> @@ -327,7 +334,143 @@ static struct commit *pick_regular_commit(struct repository *repo,
>   	if (oideq(&replayed_base_tree->object.oid, &result->tree->object.oid) &&
>   	    !oideq(&pickme_tree->object.oid, &base_tree->object.oid))
>   		return replayed_base;
> -	return create_commit(repo, result->tree, pickme, replayed_base, mode);
> +	commit_list_insert(replayed_base, &parents);
> +	return create_commit(repo, result->tree, pickme, parents, mode);
> +}
> +
> +/*
> + * Replay a 2-parent merge commit by composing three calls into merge-ort:
> + *
> + *   R = recursive merge of pickme's two original parents (auto-remerge of
> + *       the original merge, accepting any conflicts)
> + *   N = recursive merge of the (possibly rewritten) parents
> + *   O = pickme's tree (the user's actual merge, including any manual
> + *       resolutions)
> + *
> + * The picked tree comes from a non-recursive merge using R as the base,
> + * O as side1 and N as side2. `git diff R O` is morally `git show
> + * --remerge-diff $oldmerge`, so this layers the user's original manual
> + * resolution on top of the freshly auto-merged rewritten parents (see
> + * `replay-design-notes.txt` on the `replay` branch of newren/git).
> + *
> + * If the outer 3-way merge is unclean, propagate the conflict status to
> + * the caller via `result->clean = 0` and return NULL. The two inner
> + * merges (R and N) being unclean is _not_ fatal: the conflict-markered
> + * trees they produce are valid inputs to the outer merge, and using
> + * identical labels for both inner merges keeps the marker text
> + * byte-equal between R and N so the user's resolution recorded in O
> + * collapses the conflict cleanly there. Octopus merges (more than two
> + * parents) and revert-of-merge are rejected by the caller before this
> + * function is invoked.
> + */
> +static struct commit *pick_merge_commit(struct repository *repo,
> +					struct commit *pickme,
> +					kh_oid_map_t *replayed_commits,
> +					struct merge_options *merge_opt,
> +					struct merge_result *result)
> +{
> +	struct commit *parent1, *parent2;
> +	struct commit *replayed_par1, *replayed_par2;
> +	struct tree *pickme_tree;
> +	struct merge_options remerge_opt = { 0 };
> +	struct merge_options new_merge_opt = { 0 };
> +	struct merge_result remerge_res = { 0 };
> +	struct merge_result new_merge_res = { 0 };
> +	struct commit_list *parent_bases = NULL;
> +	struct commit_list *replayed_bases = NULL;
> +	struct commit_list *parents;
> +	struct commit *picked = NULL;
> +	char *ancestor_name = NULL;
> +
> +	parent1 = pickme->parents->item;
> +	parent2 = pickme->parents->next->item;
> +
> +	/*
> +	 * Map the merge's parents to their replayed counterparts. With the
> +	 * boundary commits pre-seeded into `replayed_commits`, every parent
> +	 * either has an explicit mapping (rewritten or boundary -> onto) or
> +	 * sits outside the rewrite range entirely; the latter must stay at
> +	 * the original parent commit, so use `parent` itself as the fallback
> +	 * for both sides.
> +	 */
> +	replayed_par1 = mapped_commit(replayed_commits, parent1, parent1);
> +	replayed_par2 = mapped_commit(replayed_commits, parent2, parent2);
> +
> +	/*
> +	 * R: auto-remerge of the original parents.
> +	 *
> +	 * Use the same branch labels for the inner merges that compute R
> +	 * and N so conflict markers (if any) are textually identical
> +	 * between the two; the outer non-recursive merge can then collapse
> +	 * the manual resolution from O against them.
> +	 */
> +	init_basic_merge_options(&remerge_opt, repo);
> +	remerge_opt.show_rename_progress = 0;
> +	remerge_opt.branch1 = "ours";
> +	remerge_opt.branch2 = "theirs";
> +	if (repo_get_merge_bases(repo, parent1, parent2, &parent_bases) < 0) {
> +		result->clean = -1;
> +		goto out;
> +	}
> +	merge_incore_recursive(&remerge_opt, parent_bases,
> +			       parent1, parent2, &remerge_res);
> +	parent_bases = NULL; /* consumed by merge_incore_recursive */
> +	if (remerge_res.clean < 0) {
> +		result->clean = remerge_res.clean;
> +		goto out;
> +	}
> +
> +	/* N: fresh merge of the (possibly rewritten) parents. */
> +	init_basic_merge_options(&new_merge_opt, repo);
> +	new_merge_opt.show_rename_progress = 0;
> +	new_merge_opt.branch1 = "ours";
> +	new_merge_opt.branch2 = "theirs";
> +	if (repo_get_merge_bases(repo, replayed_par1, replayed_par2,
> +				 &replayed_bases) < 0) {
> +		result->clean = -1;
> +		goto out;
> +	}
> +	merge_incore_recursive(&new_merge_opt, replayed_bases,
> +			       replayed_par1, replayed_par2, &new_merge_res);
> +	replayed_bases = NULL; /* consumed by merge_incore_recursive */
> +	if (new_merge_res.clean < 0) {
> +		result->clean = new_merge_res.clean;
> +		goto out;
> +	}
> +
> +	/*
> +	 * Outer non-recursive merge: base=R, side1=O (pickme), side2=N.
> +	 */
> +	pickme_tree = repo_get_commit_tree(repo, pickme);
> +	ancestor_name = xstrfmt("auto-remerge of %s",
> +				oid_to_hex(&pickme->object.oid));
> +	merge_opt->ancestor = ancestor_name;
> +	merge_opt->branch1 = short_commit_name(repo, pickme);
> +	merge_opt->branch2 = "merge of replayed parents";
> +	merge_incore_nonrecursive(merge_opt,
> +				  remerge_res.tree,
> +				  pickme_tree,
> +				  new_merge_res.tree,
> +				  result);
> +	merge_opt->ancestor = NULL;
> +	merge_opt->branch1 = NULL;
> +	merge_opt->branch2 = NULL;
> +	if (!result->clean)
> +		goto out;
> +
> +	parents = NULL;
> +	commit_list_insert(replayed_par2, &parents);
> +	commit_list_insert(replayed_par1, &parents);
> +	picked = create_commit(repo, result->tree, pickme, parents,
> +			       REPLAY_MODE_PICK);
> +
> +out:
> +	free(ancestor_name);
> +	free_commit_list(parent_bases);
> +	free_commit_list(replayed_bases);
> +	merge_finalize(&remerge_opt, &remerge_res);
> +	merge_finalize(&new_merge_opt, &new_merge_res);
> +	return picked;
>   }
>   
>   void replay_result_release(struct replay_result *result)
> @@ -407,17 +550,63 @@ int replay_revisions(struct rev_info *revs,
>   	merge_opt.show_rename_progress = 0;
>   	last_commit = onto;
>   	replayed_commits = kh_init_oid_map();
> +
> +	/*
> +	 * Seed the rewritten-commit map with each negative-side ("BOTTOM")
> +	 * cmdline entry pointing at `onto`. This matters for merge replay:
> +	 * a 2-parent merge whose first parent is the boundary (e.g. the
> +	 * commit being reworded) must replay onto the rewritten boundary,
> +	 * yet pick_merge_commit uses a self fallback so the second parent
> +	 * (a side branch outside the rewrite range) is preserved as-is.
> +	 * Pre-seeding the boundary disambiguates the two: in the map ->
> +	 * rewritten, missing -> kept as-is.
> +	 *
> +	 * Only do this for the pick path; revert mode chains reverts
> +	 * through last_commit and a pre-seeded boundary would short-circuit
> +	 * that chain.
> +	 */
> +	if (mode == REPLAY_MODE_PICK) {
> +		for (size_t i = 0; i < revs->cmdline.nr; i++) {
> +			struct rev_cmdline_entry *e = &revs->cmdline.rev[i];
> +			struct commit *boundary;
> +			khint_t pos;
> +			int hr;
> +
> +			if (!(e->flags & BOTTOM))
> +				continue;
> +			boundary = lookup_commit_reference_gently(revs->repo,
> +								  &e->item->oid, 1);
> +			if (!boundary)
> +				continue;
> +			pos = kh_put_oid_map(replayed_commits,
> +					     boundary->object.oid, &hr);
> +			if (hr != 0)
> +				kh_value(replayed_commits, pos) = onto;
> +		}
> +	}
> +
>   	while ((commit = get_revision(revs))) {
>   		const struct name_decoration *decoration;
>   		khint_t pos;
>   		int hr;
>   
> -		if (commit->parents && commit->parents->next)
> -			die(_("replaying merge commits is not supported yet!"));
> -
> -		last_commit = pick_regular_commit(revs->repo, commit, replayed_commits,
> -						  mode == REPLAY_MODE_REVERT ? last_commit : onto,
> -						  &merge_opt, &result, mode);
> +		if (commit->parents && commit->parents->next) {
> +			if (commit->parents->next->next) {
> +				ret = error(_("replaying octopus merges is not supported"));
> +				goto out;
> +			}
> +			if (mode == REPLAY_MODE_REVERT) {
> +				ret = error(_("reverting merge commits is not supported"));
> +				goto out;
> +			}
> +			last_commit = pick_merge_commit(revs->repo, commit,
> +							replayed_commits,
> +							&merge_opt, &result);
> +		} else {
> +			last_commit = pick_regular_commit(revs->repo, commit, replayed_commits,
> +							  mode == REPLAY_MODE_REVERT ? last_commit : onto,
> +							  &merge_opt, &result, mode);
> +		}
>   		if (!last_commit)
>   			break;
>   
> diff --git a/t/t3451-history-reword.sh b/t/t3451-history-reword.sh
> index de7b357685..d103f866a2 100755
> --- a/t/t3451-history-reword.sh
> +++ b/t/t3451-history-reword.sh
> @@ -201,12 +201,21 @@ test_expect_success 'can reword a merge commit' '
>   		git switch - &&
>   		git merge theirs &&
>   
> -		# It is not possible to replay merge commits embedded in the
> -		# history (yet).
> -		test_must_fail git -c core.editor=false history reword HEAD~ 2>err &&
> -		test_grep "replaying merge commits is not supported yet" err &&
> +		# Reword a non-merge commit whose descendants include the
> +		# merge: replay carries the merge through.
> +		reword_with_message HEAD~ <<-EOF &&
> +		ours reworded
> +		EOF
> +		expect_graph <<-EOF &&
> +		*   Merge tag ${SQ}theirs${SQ}
> +		|\\
> +		| * theirs
> +		* | ours reworded
> +		|/
> +		* base
> +		EOF
>   
> -		# But it is possible to reword a merge commit directly.
> +		# And reword a merge commit directly.
>   		reword_with_message HEAD <<-EOF &&
>   		Reworded merge commit
>   		EOF
> @@ -214,7 +223,7 @@ test_expect_success 'can reword a merge commit' '
>   		*   Reworded merge commit
>   		|\
>   		| * theirs
> -		* | ours
> +		* | ours reworded
>   		|/
>   		* base
>   		EOF
> diff --git a/t/t3452-history-split.sh b/t/t3452-history-split.sh
> index 8ed0cebb50..ad6309f98b 100755
> --- a/t/t3452-history-split.sh
> +++ b/t/t3452-history-split.sh
> @@ -36,7 +36,7 @@ expect_tree_entries () {
>   	test_cmp expect actual
>   }
>   
> -test_expect_success 'refuses to work with merge commits' '
> +test_expect_success 'refuses to split a merge commit' '
>   	test_when_finished "rm -rf repo" &&
>   	git init repo &&
>   	(
> @@ -49,9 +49,7 @@ test_expect_success 'refuses to work with merge commits' '
>   		git switch - &&
>   		git merge theirs &&
>   		test_must_fail git history split HEAD 2>err &&
> -		test_grep "cannot split up merge commit" err &&
> -		test_must_fail git history split HEAD~ 2>err &&
> -		test_grep "replaying merge commits is not supported yet" err
> +		test_grep "cannot split up merge commit" err
>   	)
>   '
>   
> diff --git a/t/t3650-replay-basics.sh b/t/t3650-replay-basics.sh
> index 3353bc4a4d..368b1b0f9a 100755
> --- a/t/t3650-replay-basics.sh
> +++ b/t/t3650-replay-basics.sh
> @@ -103,10 +103,48 @@ test_expect_success 'cannot advance target ... ordering would be ill-defined' '
>   	test_cmp expect actual
>   '
>   
> -test_expect_success 'replaying merge commits is not supported yet' '
> -	echo "fatal: replaying merge commits is not supported yet!" >expect &&
> -	test_must_fail git replay --advance=main main..topic-with-merge 2>actual &&
> -	test_cmp expect actual
> +test_expect_success 'using replay to rebase a 2-parent merge' '
> +	# main..topic-with-merge contains a 2-parent merge (P) introduced
> +	# via test_merge. Use --ref-action=print so this test does not
> +	# mutate state for subsequent tests in this file.
> +	git replay --ref-action=print --onto main main..topic-with-merge >result &&
> +	test_line_count = 1 result &&
> +
> +	new_tip=$(cut -f 3 -d " " result) &&
> +
> +	# Result is still a 2-parent merge.
> +	git cat-file -p $new_tip >cat &&
> +	grep -c "^parent " cat >count &&
> +	echo 2 >expect &&
> +	test_cmp expect count &&
> +
> +	# Merge subject is preserved.
> +	echo P >expect &&
> +	git log -1 --format=%s $new_tip >actual &&
> +	test_cmp expect actual &&
> +
> +	# The replayed merge sits on top of main: walking back via the
> +	# first-parent chain reaches main.
> +	git merge-base --is-ancestor main $new_tip
> +'
> +
> +test_expect_success 'replaying an octopus merge is rejected' '
> +	# Build an octopus side-branch so the rest of the test state stays
> +	# untouched.
> +	test_when_finished "git update-ref -d refs/heads/octopus-tip" &&
> +	octopus_tip=$(git commit-tree -p topic4 -p topic1 -p topic3 \
> +		-m "octopus" $(git rev-parse topic4^{tree})) &&
> +	git update-ref refs/heads/octopus-tip "$octopus_tip" &&
> +
> +	test_must_fail git replay --ref-action=print --onto main \
> +		topic4..octopus-tip 2>actual &&
> +	test_grep "octopus merges" actual
> +'
> +
> +test_expect_success 'reverting a merge commit is rejected' '
> +	test_must_fail git replay --ref-action=print --revert=topic-with-merge \
> +		topic4..topic-with-merge 2>actual &&
> +	test_grep "reverting merge commits" actual
>   '
>   
>   test_expect_success 'using replay to rebase two branches, one on top of other' '

Re: [PATCH/RFC 1/5] replay: support replaying 2-parent merges

[Phillip Wood]

On 08/05/2026 10:36, Phillip Wood wrote:
> Hi Johannes
> 
> On 06/05/2026 23:43, Johannes Schindelin via GitGitGadget wrote:
>>
>> Elijah Newren spelled out a way to lift this limitation in his
>> replay-design-notes [1] and prototyped it in a 2022
>> work-in-progress sketch [2]. The idea is that a merge commit M on
>> parents (P1, P2) records both an automatic merge of those parents
>> AND any manual layer the author put on top of that automatic merge
>> (textual conflict resolution and any semantic edit outside conflict
>> markers). Replaying M onto rewritten parents (P1', P2') must
>> preserve that manual layer, but the rewritten parents change the
>> automatic merge, so a simple cherry-pick is wrong: the manual layer
>> would be re-introduced on top of stale auto-merge text.
>>
>> What works instead is a three-way merge of three trees the existing
>> infrastructure already knows how to compute. Let R be the recursive
>> auto-merge of (P1, P2), O be M's actual tree and N be the recursive
>> auto-merge of (P1', P2'). Then `git diff R O` is morally
>> `git show --remerge-diff M`: it captures exactly what the author
>> added on top of the automatic merge. A non-recursive 3-way merge
>> with R as the merge base, O as side 1 and N as side 2 layers that
>> manual contribution onto the freshly auto-merged rewritten parents
>> (N) and produces the replayed tree.
> 
> So we cherry-pick the difference between the user's conflict resolution 
> O and the auto-merge M of the original parents onto the auto-merge N of 
> the replayed parents. If we have a topology that looks like
> 
>          |
>          A
>         /|\
>        / B \
>       E  |  D
>          C /
>          |/
>          O
> 
> then running
> 
>      git replay --onto E --ancestry-path B..O
> 
> will replay C and O onto E. If the changes in E and D conflict but those 
> conflicts do not overlap with the conflicts in M that were resolved to 
> create O then the replayed version of O will contain conflict markers 
> from the conflicting changes in E and D. Because the previous conflict 
> resolution applies to N without conflicts we do not recognize that there 
> are still conflicts in N that need to be resolved.
> 
> Having realized this I went to look at Elijah's notes and they recognize 
> this possibility and suggest extending the xdiff merge code to detect 
> when N has conflicts that do not correspond to the conflicts in M. That 
> sounds like quite a lot of work. I've not put much effort into coming up 
> with a counterexample but think that because "git replay" and "git 
> history" do not yet allow the commits in the merged branches to be 
> edited we may be able to safely use the implementation proposed in this 
> series if both merge parents have been rebased (or we might want all the 
> merge bases of the new merge to be a descendants of "--onto"). In the 
> example above if both the parents were rebased onto E then any new 
> conflicts would happen when picking D rather than when recreating the 
> merge.

One further thought - if only one of the parents has been rebased (i.e. 
we're replaying O with parents P1' and P2) then can we just cherry-pick 
the merge - instead of merging P1' and P2, use P1 as the merge-base with 
O and P1' as the merge heads?

Thanks

Phillip

> Thanks
> 
> Phillip
> 
>> Implement `pick_merge_commit()` along those lines and dispatch to it
>> from `replay_revisions()` when the commit being replayed has exactly
>> two parents. Two specific points (learned the hard way) keep
>> non-trivial cases working where the WIP sketch [2] bailed out.
>> First, R and N use identical `merge_options.branch1` and `branch2`
>> labels ("ours"/"theirs"). When the original parents conflicted on a
>> region of a file, both R and N produce textually identical conflict
>> markers; the outer non-recursive merge then sees N == R in that
>> region and the user's manual resolution from O wins cleanly. Without
>> this, the conflict-marker text would differ between R and N (because
>> the inner merges would label the conflicts differently), and the
>> outer merge would itself be unclean even when the user did supply a
>> clean resolution. Second, an unclean inner merge
>> (`result.clean == 0`) is _not_ fatal: the tree merge-ort produces in
>> that case still has well-defined contents (with conflict markers in
>> the conflicted files) and is a valid input to the outer
>> non-recursive merge. Only a real error (`< 0`) propagates as
>> failure.
>>
>> The replay propagates the textual diffs the user actually made in M;
>> it does _not_ extrapolate symbol-level intent. If rewriting the
>> parents pulls in genuinely new content (for example, a brand-new
>> caller of a function that the merge renamed), that new content stays
>> as the rewritten parents have it. Symbol-aware refactoring is out of
>> scope here, just as it is for plain rebase.
>>
>> Octopus merges (more than two parents) and revert-of-merge are not
>> supported and are surfaced as explicit errors at the dispatch point.
>> The "split" sub-command of `git history` continues to refuse when
>> the targeted commit is itself a merge: split semantics do not apply
>> to merges. The pre-walk gate in `builtin/history.c` that previously
>> rejected any merge in the rewrite path now only rejects octopus
>> merges; rename it accordingly.
>>
>> A small refactor in `create_commit()` makes the merge case possible:
>> the helper now takes a `struct commit_list *parents` rather than a
>> single parent pointer and takes ownership of the list. The single
>> existing caller in `pick_regular_commit()` builds and passes a
>> one-element list; the new `pick_merge_commit()` builds a two-element
>> list, with the order of the `from` and `merge` parents preserved.
>>
>> Update the negative expectations in t3451, t3452 and t3650 that were
>> asserting the now-retired "not supported yet" message, replacing
>> them with positive coverage where it fits. Octopus rejection and
>> revert-of-merge rejection are covered by new positive tests in
>> t3650. A dedicated test script with merge-replay scenarios driven by
>> a new test-tool fixture builder will follow in a subsequent commit.
>>
>> [1] https://github.com/newren/git/blob/replay/replay-design-notes.txt
>> [2] https://github.com/newren/git/ 
>> commit/4c45e8955ef9bf7d01fd15d9106b3bdb8ea91b45
>>
>> Helped-by: Elijah Newren <newren@gmail.com>
>> Assisted-by: Claude Opus 4.7
>> Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
>> ---
>>   builtin/history.c         |  16 ++-
>>   replay.c                  | 209 ++++++++++++++++++++++++++++++++++++--
>>   t/t3451-history-reword.sh |  21 ++--
>>   t/t3452-history-split.sh  |   6 +-
>>   t/t3650-replay-basics.sh  |  46 ++++++++-
>>   5 files changed, 269 insertions(+), 29 deletions(-)
>>
>> diff --git a/builtin/history.c b/builtin/history.c
>> index 9526938085..00097b2226 100644
>> --- a/builtin/history.c
>> +++ b/builtin/history.c
>> @@ -195,15 +195,15 @@ static int parse_ref_action(const struct option 
>> *opt, const char *value, int uns
>>       return 0;
>>   }
>> -static int revwalk_contains_merges(struct repository *repo,
>> -                   const struct strvec *revwalk_args)
>> +static int revwalk_contains_octopus_merges(struct repository *repo,
>> +                       const struct strvec *revwalk_args)
>>   {
>>       struct strvec args = STRVEC_INIT;
>>       struct rev_info revs;
>>       int ret;
>>       strvec_pushv(&args, revwalk_args->v);
>> -    strvec_push(&args, "--min-parents=2");
>> +    strvec_push(&args, "--min-parents=3");
>>       repo_init_revisions(repo, &revs, NULL);
>> @@ -217,7 +217,7 @@ static int revwalk_contains_merges(struct 
>> repository *repo,
>>       }
>>       if (get_revision(&revs)) {
>> -        ret = error(_("replaying merge commits is not supported yet!"));
>> +        ret = error(_("replaying octopus merges is not supported"));
>>           goto out;
>>       }
>> @@ -289,7 +289,7 @@ static int setup_revwalk(struct repository *repo,
>>           strvec_push(&args, "HEAD");
>>       }
>> -    ret = revwalk_contains_merges(repo, &args);
>> +    ret = revwalk_contains_octopus_merges(repo, &args);
>>       if (ret < 0)
>>           goto out;
>> @@ -482,6 +482,9 @@ static int cmd_history_reword(int argc,
>>       if (ret < 0) {
>>           ret = error(_("failed replaying descendants"));
>>           goto out;
>> +    } else if (ret) {
>> +        ret = error(_("conflict during replay; some descendants were 
>> not rewritten"));
>> +        goto out;
>>       }
>>       ret = 0;
>> @@ -721,6 +724,9 @@ static int cmd_history_split(int argc,
>>       if (ret < 0) {
>>           ret = error(_("failed replaying descendants"));
>>           goto out;
>> +    } else if (ret) {
>> +        ret = error(_("conflict during replay; some descendants were 
>> not rewritten"));
>> +        goto out;
>>       }
>>       ret = 0;
>> diff --git a/replay.c b/replay.c
>> index f96f1f6551..3dbce095f9 100644
>> --- a/replay.c
>> +++ b/replay.c
>> @@ -1,6 +1,7 @@
>>   #define USE_THE_REPOSITORY_VARIABLE
>>   #include "git-compat-util.h"
>> +#include "commit-reach.h"
>>   #include "environment.h"
>>   #include "hex.h"
>>   #include "merge-ort.h"
>> @@ -77,15 +78,21 @@ static void generate_revert_message(struct strbuf 
>> *msg,
>>       repo_unuse_commit_buffer(repo, commit, message);
>>   }
>> +/*
>> + * Build a new commit with the given tree and parent list, copying 
>> author,
>> + * extra headers and (for pick mode) the commit message from `based_on`.
>> + *
>> + * Takes ownership of `parents`: it will be freed before returning, 
>> even on
>> + * error. Parent order is preserved as supplied by the caller.
>> + */
>>   static struct commit *create_commit(struct repository *repo,
>>                       struct tree *tree,
>>                       struct commit *based_on,
>> -                    struct commit *parent,
>> +                    struct commit_list *parents,
>>                       enum replay_mode mode)
>>   {
>>       struct object_id ret;
>>       struct object *obj = NULL;
>> -    struct commit_list *parents = NULL;
>>       char *author = NULL;
>>       char *sign_commit = NULL; /* FIXME: cli users might want to sign 
>> again */
>>       struct commit_extra_header *extra = NULL;
>> @@ -96,7 +103,6 @@ static struct commit *create_commit(struct 
>> repository *repo,
>>       const char *orig_message = NULL;
>>       const char *exclude_gpgsig[] = { "gpgsig", "gpgsig-sha256", NULL };
>> -    commit_list_insert(parent, &parents);
>>       extra = read_commit_extra_headers(based_on, exclude_gpgsig);
>>       if (mode == REPLAY_MODE_REVERT) {
>>           generate_revert_message(&msg, based_on, repo);
>> @@ -273,6 +279,7 @@ static struct commit *pick_regular_commit(struct 
>> repository *repo,
>>   {
>>       struct commit *base, *replayed_base;
>>       struct tree *pickme_tree, *base_tree, *replayed_base_tree;
>> +    struct commit_list *parents = NULL;
>>       if (pickme->parents) {
>>           base = pickme->parents->item;
>> @@ -327,7 +334,143 @@ static struct commit *pick_regular_commit(struct 
>> repository *repo,
>>       if (oideq(&replayed_base_tree->object.oid, &result->tree- 
>> >object.oid) &&
>>           !oideq(&pickme_tree->object.oid, &base_tree->object.oid))
>>           return replayed_base;
>> -    return create_commit(repo, result->tree, pickme, replayed_base, 
>> mode);
>> +    commit_list_insert(replayed_base, &parents);
>> +    return create_commit(repo, result->tree, pickme, parents, mode);
>> +}
>> +
>> +/*
>> + * Replay a 2-parent merge commit by composing three calls into 
>> merge-ort:
>> + *
>> + *   R = recursive merge of pickme's two original parents (auto- 
>> remerge of
>> + *       the original merge, accepting any conflicts)
>> + *   N = recursive merge of the (possibly rewritten) parents
>> + *   O = pickme's tree (the user's actual merge, including any manual
>> + *       resolutions)
>> + *
>> + * The picked tree comes from a non-recursive merge using R as the base,
>> + * O as side1 and N as side2. `git diff R O` is morally `git show
>> + * --remerge-diff $oldmerge`, so this layers the user's original manual
>> + * resolution on top of the freshly auto-merged rewritten parents (see
>> + * `replay-design-notes.txt` on the `replay` branch of newren/git).
>> + *
>> + * If the outer 3-way merge is unclean, propagate the conflict status to
>> + * the caller via `result->clean = 0` and return NULL. The two inner
>> + * merges (R and N) being unclean is _not_ fatal: the conflict-markered
>> + * trees they produce are valid inputs to the outer merge, and using
>> + * identical labels for both inner merges keeps the marker text
>> + * byte-equal between R and N so the user's resolution recorded in O
>> + * collapses the conflict cleanly there. Octopus merges (more than two
>> + * parents) and revert-of-merge are rejected by the caller before this
>> + * function is invoked.
>> + */
>> +static struct commit *pick_merge_commit(struct repository *repo,
>> +                    struct commit *pickme,
>> +                    kh_oid_map_t *replayed_commits,
>> +                    struct merge_options *merge_opt,
>> +                    struct merge_result *result)
>> +{
>> +    struct commit *parent1, *parent2;
>> +    struct commit *replayed_par1, *replayed_par2;
>> +    struct tree *pickme_tree;
>> +    struct merge_options remerge_opt = { 0 };
>> +    struct merge_options new_merge_opt = { 0 };
>> +    struct merge_result remerge_res = { 0 };
>> +    struct merge_result new_merge_res = { 0 };
>> +    struct commit_list *parent_bases = NULL;
>> +    struct commit_list *replayed_bases = NULL;
>> +    struct commit_list *parents;
>> +    struct commit *picked = NULL;
>> +    char *ancestor_name = NULL;
>> +
>> +    parent1 = pickme->parents->item;
>> +    parent2 = pickme->parents->next->item;
>> +
>> +    /*
>> +     * Map the merge's parents to their replayed counterparts. With the
>> +     * boundary commits pre-seeded into `replayed_commits`, every parent
>> +     * either has an explicit mapping (rewritten or boundary -> onto) or
>> +     * sits outside the rewrite range entirely; the latter must stay at
>> +     * the original parent commit, so use `parent` itself as the 
>> fallback
>> +     * for both sides.
>> +     */
>> +    replayed_par1 = mapped_commit(replayed_commits, parent1, parent1);
>> +    replayed_par2 = mapped_commit(replayed_commits, parent2, parent2);
>> +
>> +    /*
>> +     * R: auto-remerge of the original parents.
>> +     *
>> +     * Use the same branch labels for the inner merges that compute R
>> +     * and N so conflict markers (if any) are textually identical
>> +     * between the two; the outer non-recursive merge can then collapse
>> +     * the manual resolution from O against them.
>> +     */
>> +    init_basic_merge_options(&remerge_opt, repo);
>> +    remerge_opt.show_rename_progress = 0;
>> +    remerge_opt.branch1 = "ours";
>> +    remerge_opt.branch2 = "theirs";
>> +    if (repo_get_merge_bases(repo, parent1, parent2, &parent_bases) < 
>> 0) {
>> +        result->clean = -1;
>> +        goto out;
>> +    }
>> +    merge_incore_recursive(&remerge_opt, parent_bases,
>> +                   parent1, parent2, &remerge_res);
>> +    parent_bases = NULL; /* consumed by merge_incore_recursive */
>> +    if (remerge_res.clean < 0) {
>> +        result->clean = remerge_res.clean;
>> +        goto out;
>> +    }
>> +
>> +    /* N: fresh merge of the (possibly rewritten) parents. */
>> +    init_basic_merge_options(&new_merge_opt, repo);
>> +    new_merge_opt.show_rename_progress = 0;
>> +    new_merge_opt.branch1 = "ours";
>> +    new_merge_opt.branch2 = "theirs";
>> +    if (repo_get_merge_bases(repo, replayed_par1, replayed_par2,
>> +                 &replayed_bases) < 0) {
>> +        result->clean = -1;
>> +        goto out;
>> +    }
>> +    merge_incore_recursive(&new_merge_opt, replayed_bases,
>> +                   replayed_par1, replayed_par2, &new_merge_res);
>> +    replayed_bases = NULL; /* consumed by merge_incore_recursive */
>> +    if (new_merge_res.clean < 0) {
>> +        result->clean = new_merge_res.clean;
>> +        goto out;
>> +    }
>> +
>> +    /*
>> +     * Outer non-recursive merge: base=R, side1=O (pickme), side2=N.
>> +     */
>> +    pickme_tree = repo_get_commit_tree(repo, pickme);
>> +    ancestor_name = xstrfmt("auto-remerge of %s",
>> +                oid_to_hex(&pickme->object.oid));
>> +    merge_opt->ancestor = ancestor_name;
>> +    merge_opt->branch1 = short_commit_name(repo, pickme);
>> +    merge_opt->branch2 = "merge of replayed parents";
>> +    merge_incore_nonrecursive(merge_opt,
>> +                  remerge_res.tree,
>> +                  pickme_tree,
>> +                  new_merge_res.tree,
>> +                  result);
>> +    merge_opt->ancestor = NULL;
>> +    merge_opt->branch1 = NULL;
>> +    merge_opt->branch2 = NULL;
>> +    if (!result->clean)
>> +        goto out;
>> +
>> +    parents = NULL;
>> +    commit_list_insert(replayed_par2, &parents);
>> +    commit_list_insert(replayed_par1, &parents);
>> +    picked = create_commit(repo, result->tree, pickme, parents,
>> +                   REPLAY_MODE_PICK);
>> +
>> +out:
>> +    free(ancestor_name);
>> +    free_commit_list(parent_bases);
>> +    free_commit_list(replayed_bases);
>> +    merge_finalize(&remerge_opt, &remerge_res);
>> +    merge_finalize(&new_merge_opt, &new_merge_res);
>> +    return picked;
>>   }
>>   void replay_result_release(struct replay_result *result)
>> @@ -407,17 +550,63 @@ int replay_revisions(struct rev_info *revs,
>>       merge_opt.show_rename_progress = 0;
>>       last_commit = onto;
>>       replayed_commits = kh_init_oid_map();
>> +
>> +    /*
>> +     * Seed the rewritten-commit map with each negative-side ("BOTTOM")
>> +     * cmdline entry pointing at `onto`. This matters for merge replay:
>> +     * a 2-parent merge whose first parent is the boundary (e.g. the
>> +     * commit being reworded) must replay onto the rewritten boundary,
>> +     * yet pick_merge_commit uses a self fallback so the second parent
>> +     * (a side branch outside the rewrite range) is preserved as-is.
>> +     * Pre-seeding the boundary disambiguates the two: in the map ->
>> +     * rewritten, missing -> kept as-is.
>> +     *
>> +     * Only do this for the pick path; revert mode chains reverts
>> +     * through last_commit and a pre-seeded boundary would short-circuit
>> +     * that chain.
>> +     */
>> +    if (mode == REPLAY_MODE_PICK) {
>> +        for (size_t i = 0; i < revs->cmdline.nr; i++) {
>> +            struct rev_cmdline_entry *e = &revs->cmdline.rev[i];
>> +            struct commit *boundary;
>> +            khint_t pos;
>> +            int hr;
>> +
>> +            if (!(e->flags & BOTTOM))
>> +                continue;
>> +            boundary = lookup_commit_reference_gently(revs->repo,
>> +                                  &e->item->oid, 1);
>> +            if (!boundary)
>> +                continue;
>> +            pos = kh_put_oid_map(replayed_commits,
>> +                         boundary->object.oid, &hr);
>> +            if (hr != 0)
>> +                kh_value(replayed_commits, pos) = onto;
>> +        }
>> +    }
>> +
>>       while ((commit = get_revision(revs))) {
>>           const struct name_decoration *decoration;
>>           khint_t pos;
>>           int hr;
>> -        if (commit->parents && commit->parents->next)
>> -            die(_("replaying merge commits is not supported yet!"));
>> -
>> -        last_commit = pick_regular_commit(revs->repo, commit, 
>> replayed_commits,
>> -                          mode == REPLAY_MODE_REVERT ? last_commit : 
>> onto,
>> -                          &merge_opt, &result, mode);
>> +        if (commit->parents && commit->parents->next) {
>> +            if (commit->parents->next->next) {
>> +                ret = error(_("replaying octopus merges is not 
>> supported"));
>> +                goto out;
>> +            }
>> +            if (mode == REPLAY_MODE_REVERT) {
>> +                ret = error(_("reverting merge commits is not 
>> supported"));
>> +                goto out;
>> +            }
>> +            last_commit = pick_merge_commit(revs->repo, commit,
>> +                            replayed_commits,
>> +                            &merge_opt, &result);
>> +        } else {
>> +            last_commit = pick_regular_commit(revs->repo, commit, 
>> replayed_commits,
>> +                              mode == REPLAY_MODE_REVERT ? 
>> last_commit : onto,
>> +                              &merge_opt, &result, mode);
>> +        }
>>           if (!last_commit)
>>               break;
>> diff --git a/t/t3451-history-reword.sh b/t/t3451-history-reword.sh
>> index de7b357685..d103f866a2 100755
>> --- a/t/t3451-history-reword.sh
>> +++ b/t/t3451-history-reword.sh
>> @@ -201,12 +201,21 @@ test_expect_success 'can reword a merge commit' '
>>           git switch - &&
>>           git merge theirs &&
>> -        # It is not possible to replay merge commits embedded in the
>> -        # history (yet).
>> -        test_must_fail git -c core.editor=false history reword HEAD~ 
>> 2>err &&
>> -        test_grep "replaying merge commits is not supported yet" err &&
>> +        # Reword a non-merge commit whose descendants include the
>> +        # merge: replay carries the merge through.
>> +        reword_with_message HEAD~ <<-EOF &&
>> +        ours reworded
>> +        EOF
>> +        expect_graph <<-EOF &&
>> +        *   Merge tag ${SQ}theirs${SQ}
>> +        |\\
>> +        | * theirs
>> +        * | ours reworded
>> +        |/
>> +        * base
>> +        EOF
>> -        # But it is possible to reword a merge commit directly.
>> +        # And reword a merge commit directly.
>>           reword_with_message HEAD <<-EOF &&
>>           Reworded merge commit
>>           EOF
>> @@ -214,7 +223,7 @@ test_expect_success 'can reword a merge commit' '
>>           *   Reworded merge commit
>>           |\
>>           | * theirs
>> -        * | ours
>> +        * | ours reworded
>>           |/
>>           * base
>>           EOF
>> diff --git a/t/t3452-history-split.sh b/t/t3452-history-split.sh
>> index 8ed0cebb50..ad6309f98b 100755
>> --- a/t/t3452-history-split.sh
>> +++ b/t/t3452-history-split.sh
>> @@ -36,7 +36,7 @@ expect_tree_entries () {
>>       test_cmp expect actual
>>   }
>> -test_expect_success 'refuses to work with merge commits' '
>> +test_expect_success 'refuses to split a merge commit' '
>>       test_when_finished "rm -rf repo" &&
>>       git init repo &&
>>       (
>> @@ -49,9 +49,7 @@ test_expect_success 'refuses to work with merge 
>> commits' '
>>           git switch - &&
>>           git merge theirs &&
>>           test_must_fail git history split HEAD 2>err &&
>> -        test_grep "cannot split up merge commit" err &&
>> -        test_must_fail git history split HEAD~ 2>err &&
>> -        test_grep "replaying merge commits is not supported yet" err
>> +        test_grep "cannot split up merge commit" err
>>       )
>>   '
>> diff --git a/t/t3650-replay-basics.sh b/t/t3650-replay-basics.sh
>> index 3353bc4a4d..368b1b0f9a 100755
>> --- a/t/t3650-replay-basics.sh
>> +++ b/t/t3650-replay-basics.sh
>> @@ -103,10 +103,48 @@ test_expect_success 'cannot advance target ... 
>> ordering would be ill-defined' '
>>       test_cmp expect actual
>>   '
>> -test_expect_success 'replaying merge commits is not supported yet' '
>> -    echo "fatal: replaying merge commits is not supported yet!" 
>> >expect &&
>> -    test_must_fail git replay --advance=main main..topic-with-merge 
>> 2>actual &&
>> -    test_cmp expect actual
>> +test_expect_success 'using replay to rebase a 2-parent merge' '
>> +    # main..topic-with-merge contains a 2-parent merge (P) introduced
>> +    # via test_merge. Use --ref-action=print so this test does not
>> +    # mutate state for subsequent tests in this file.
>> +    git replay --ref-action=print --onto main main..topic-with-merge 
>> >result &&
>> +    test_line_count = 1 result &&
>> +
>> +    new_tip=$(cut -f 3 -d " " result) &&
>> +
>> +    # Result is still a 2-parent merge.
>> +    git cat-file -p $new_tip >cat &&
>> +    grep -c "^parent " cat >count &&
>> +    echo 2 >expect &&
>> +    test_cmp expect count &&
>> +
>> +    # Merge subject is preserved.
>> +    echo P >expect &&
>> +    git log -1 --format=%s $new_tip >actual &&
>> +    test_cmp expect actual &&
>> +
>> +    # The replayed merge sits on top of main: walking back via the
>> +    # first-parent chain reaches main.
>> +    git merge-base --is-ancestor main $new_tip
>> +'
>> +
>> +test_expect_success 'replaying an octopus merge is rejected' '
>> +    # Build an octopus side-branch so the rest of the test state stays
>> +    # untouched.
>> +    test_when_finished "git update-ref -d refs/heads/octopus-tip" &&
>> +    octopus_tip=$(git commit-tree -p topic4 -p topic1 -p topic3 \
>> +        -m "octopus" $(git rev-parse topic4^{tree})) &&
>> +    git update-ref refs/heads/octopus-tip "$octopus_tip" &&
>> +
>> +    test_must_fail git replay --ref-action=print --onto main \
>> +        topic4..octopus-tip 2>actual &&
>> +    test_grep "octopus merges" actual
>> +'
>> +
>> +test_expect_success 'reverting a merge commit is rejected' '
>> +    test_must_fail git replay --ref-action=print --revert=topic-with- 
>> merge \
>> +        topic4..topic-with-merge 2>actual &&
>> +    test_grep "reverting merge commits" actual
>>   '
>>   test_expect_success 'using replay to rebase two branches, one on top 
>> of other' '
> 

Re: [PATCH/RFC 4/5] test-tool: add a "historian" subcommand for building merge fixtures

[Toon Claes]

"Johannes Schindelin via GitGitGadget" <gitgitgadget@gmail.com> writes:

> From: Johannes Schindelin <johannes.schindelin@gmx.de>
>
> The merge-replay tests added in a follow-up commit need a way to set
> up specific topologies with full control over blob contents, parent
> order, and per-side trees. Sequencing plumbing commands or driving
> plain `git fast-import` from shell quickly becomes unreadable for
> the kinds of scenarios that exercise non-trivial merge resolution
> (textual conflicts, semantic edits outside the conflict region,
> intentional limitations such as new content on one side).
>
> Add a small `test-tool historian` subcommand that reads a tight,
> shell-quoted, one-line-per-object DSL and feeds an equivalent stream
> to a `git fast-import` child process. Each blob and commit is given
> a logical name; the helper allocates fast-import marks on first use
> and emits a lightweight tag for every commit so tests can refer to
> the resulting object via `refs/tags/<name>`.
>
> The DSL has just two directives:
>
>   blob NAME LINE...
>   commit NAME BRANCH SUBJECT [from=NAME] [merge=NAME]... [PATH=BLOB]...
>
> A blob's content is the listed lines joined with `\n` (and a final
> `\n`); a commit's tree is exactly the listed PATH=BLOB pairs (the
> helper emits a `deleteall` so nothing leaks in from the implicit
> parent). Token splitting is delegated to `split_cmdline()` so quoted
> arguments work as in shell. Marks for parent references and file
> contents go through the same `strintmap`-backed name resolver, which
> keeps the helper itself trivially small: blob writing, tree
> construction, commit creation and merge-base computation are all
> handled by `git fast-import`.
>
> Note that the DSL reserves the names `from` and `merge` (with a
> trailing `=`) for parent specification; a tree path called `from` or
> `merge` cannot be expressed via this helper. That is acceptable here
> because every input is a tightly controlled test fixture and the
> filenames are chosen by the test author.
>
> The helper trusts its caller: malformed input results in a
> fast-import error rather than a friendly diagnostic.
>
> Wire the new subcommand into the Makefile and meson build, register
> it in `t/helper/test-tool.{c,h}`.
>
> Assisted-by: Claude Opus 4.7
> Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
> ---
>  Makefile                  |   1 +
>  t/helper/meson.build      |   1 +
>  t/helper/test-historian.c | 189 ++++++++++++++++++++++++++++++++++++++
>  t/helper/test-tool.c      |   1 +
>  t/helper/test-tool.h      |   1 +
>  5 files changed, 193 insertions(+)
>  create mode 100644 t/helper/test-historian.c
>
> diff --git a/Makefile b/Makefile
> index cedc234173..b38678b484 100644
> --- a/Makefile
> +++ b/Makefile
> @@ -832,6 +832,7 @@ TEST_BUILTINS_OBJS += test-hash-speed.o
>  TEST_BUILTINS_OBJS += test-hash.o
>  TEST_BUILTINS_OBJS += test-hashmap.o
>  TEST_BUILTINS_OBJS += test-hexdump.o
> +TEST_BUILTINS_OBJS += test-historian.o
>  TEST_BUILTINS_OBJS += test-json-writer.o
>  TEST_BUILTINS_OBJS += test-lazy-init-name-hash.o
>  TEST_BUILTINS_OBJS += test-match-trees.o
> diff --git a/t/helper/meson.build b/t/helper/meson.build
> index 675e64c010..704edd1e1f 100644
> --- a/t/helper/meson.build
> +++ b/t/helper/meson.build
> @@ -29,6 +29,7 @@ test_tool_sources = [
>    'test-hash.c',
>    'test-hashmap.c',
>    'test-hexdump.c',
> +  'test-historian.c',
>    'test-json-writer.c',
>    'test-lazy-init-name-hash.c',
>    'test-match-trees.c',
> diff --git a/t/helper/test-historian.c b/t/helper/test-historian.c
> new file mode 100644
> index 0000000000..2250d420c0
> --- /dev/null
> +++ b/t/helper/test-historian.c
> @@ -0,0 +1,189 @@
> +/*
> + * Build a small history out of a tiny declarative input. Used by tests
> + * that need specific merge topologies without long sequences of
> + * plumbing commands or fragile shell helpers.
> + *
> + * The historian reads stdin line by line and emits an equivalent
> + * stream to a `git fast-import` child process. It also allocates marks
> + * for named objects so tests can refer to commits and blobs by name.

Really appreciate you're introducing this command. I'm actually
surprised no else did before.

> + *
> + * Input directives (one per line, shell-style quoting):
> + *
> + *	blob NAME LINE1 LINE2 ...
> + *	    Each LINE becomes a content line in the blob; lines are
> + *	    joined with '\n' and the blob ends with a final '\n'. With
> + *	    no LINEs, the blob is empty.
> + *
> + *	commit NAME BRANCH SUBJECT [from=PARENT] [merge=PARENT]... [PATH=BLOB]...

I'm not sure how I feel about mixing named arguments (like `from=PARENT`) with
the `PATH=BLOB` arguments? Obviously this tool isn't made for anything
that's even close to production, but still feels strange. How about
putting a double dash (`--`) before the paths, or using the `PATH:BLOB`
syntax instead?

> + *	    Creates a commit on refs/heads/BRANCH using the listed
> + *	    file=blob mappings as the entire tree (no inheritance from
> + *	    parents). Up to one `from=` and any number of `merge=`
> + *	    parents may be given. `from=` defaults to the current branch
> + *	    tip; if BRANCH has no tip yet, the commit becomes a root.

At GitLab in our Gitaly suite we have a similar tool as what you're
introducing here, but there you have to specify the parent(s) for each
commit and if you want to assign a ref to a commit, you have to be
explicit about it. So I would replace `from=` and `merge=` with
`parent=` and allow that to be occur zero or more times (this would also
allow creating unrelated histories). And remove the mandatory argument
BRANCH, and instead allow the command to accept a `branch=` argument.

If we'd take an example from the follow-up commit:

        # Setup:
        #       A (a) --- C (a, h) ----+--- M (a, g, h)
        #        \                    /
        #         +-- B (a, g) ------+
        #
        # Topic touches `g` only; main touches `h` only. The auto-merge
        # at M is clean.
        blob a "shared content"
        blob g guarded
        blob h host
        commit A main "A" a=a
        commit B topic "B (introduces g)" from=A a=a g=g
        commit C main "C (introduces h)" a=a h=h
        commit M main "Merge topic" merge=B a=a g=g h=h

I would suggest to rewrite that to:

        blob a "shared content"
        blob g guarded
        blob h host
        commit A "A" a:a
        commit B "B (introduces g)" parent=A branch=topic a:a g:g
        commit C "C (introduces h)" parent=A a:a h:h
        commit M "Merge topic" parent=A parent=B ref=main a:a g:g h:h

I realize this is less alike to git-fast-import(1), so I'd understand if
you'd reject this idea.


-- 
Cheers,
Toon