threads / patch / 65370

patchreplay.c: support replaying root commits

Subject: [PATCH v1] replay.c: support replaying root commits

## tl;dr

4 messages between Mar 27, 2026 and Mar 28, 2026. Diffs are folded; open one to read it.

replies: 3people: 2as markdown or json

Tian Yuchen· Mar 27, 2026, 18:09 UTC · lore

'git replay' does not support replaying root commits, as indicated by the FIXME comment. If a user attempts to replay a root commit, the 'die()' in 'replay_revisions()' will be called.

The reason for this is that the call 'base = pickme->parents->item' in 'pick_regular_commit()' will cause a null pointer dereference error in the case of a root commit.

Fix this by adding new conditional statements: the difference between the root commit its 'common ancestor' is equivalent to the difference from an empty tree object, and it must be placed on the user-specified target branch 'onto'. Therefore, We set 'base' to 'NULL', 'replayed_base' to 'onto', and use 'lookup_tree()' to instantiate the OID of the empty tree in order to pass it to 'base_tree'.

Signed-off-by: Tian Yuchen <cat@malon.dev>
---

To be honest, I'm not sure whether setting 'base = NULL' is proper or not. Feel free to give feedback and suggestions!

 replay.c | 23 ++++++++++++++---------
 1 file changed, 14 insertions(+), 9 deletions(-)
Show changes to replay.c +14 −9
diff --git a/replay.c b/replay.c
index a63f6714c4..80497191a8 100644
--- a/replay.c
+++ b/replay.c
@@ -225,12 +225,21 @@ static struct commit *pick_regular_commit(struct repository *repo,
 	struct commit *base, *replayed_base;
 	struct tree *pickme_tree, *base_tree, *replayed_base_tree;
 
-	base = pickme->parents->item;
-	replayed_base = mapped_commit(replayed_commits, base, onto);
-
+	if (pickme->parents) {
+		base = pickme->parents->item;
+		replayed_base = mapped_commit(replayed_commits, base, onto);
+		base_tree = repo_get_commit_tree(repo, base);
+	} else {
+		/* 
+		 * Root commits have no parents. Their base is the empty tree,
+		 * and they should be replayed directly onto the target 'onto' commit.
+		 */
+		base = NULL;
+		replayed_base = onto;
+		base_tree = lookup_tree(repo, repo->hash_algo->empty_tree);
+	}
 	replayed_base_tree = repo_get_commit_tree(repo, replayed_base);
 	pickme_tree = repo_get_commit_tree(repo, pickme);
-	base_tree = repo_get_commit_tree(repo, base);
 
 	merge_opt->branch1 = short_commit_name(repo, replayed_base);
 	merge_opt->branch2 = short_commit_name(repo, pickme);
@@ -293,8 +302,6 @@ int replay_revisions(struct rev_info *revs,
 	set_up_replay_mode(revs->repo, &revs->cmdline, opts->onto,
 			   &detached_head, &advance, &onto, &update_refs);
 
-	/* FIXME: Should allow replaying commits with the first as a root commit */
-
 	if (prepare_revision_walk(revs) < 0) {
 		ret = error(_("error preparing revisions"));
 		goto out;
@@ -309,9 +316,7 @@ int replay_revisions(struct rev_info *revs,
 		khint_t pos;
 		int hr;
 
-		if (!commit->parents)
-			die(_("replaying down from root commit is not supported yet!"));
-		if (commit->parents->next)
+		if (commit->parents && commit->parents->next)
 			die(_("replaying merge commits is not supported yet!"));
 
 		last_commit = pick_regular_commit(revs->repo, commit, replayed_commits,
-- 
2.43.0
Junio C Hamano· Mar 27, 2026, 18:28 UTC · re: Tian Yuchen · lore

Re: [PATCH v1] replay.c: support replaying root commits

Tian Yuchen <cat@malon.dev> writes:
> 'git replay' does not support replaying root commits, as indicated by
> the FIXME comment. If a user attempts to replay a root commit, the 'die()'
> in 'replay_revisions()' will be called.

Please be aware of your surroundings. I think there are multiple topics that work on this file and among them there is a topic to address the same issue, if I am not mistaken.

The list archive is your friend, for example, you can try these:
    https://lore.kernel.org/git/?q=replay+root
    https://lore.kernel.org/git/?q=dfn%3Areplay.c+%22root+commit%22
Thanks.
Junio C Hamano· Mar 27, 2026, 18:55 UTC · re: Junio C Hamano · lore

Re: [PATCH v1] replay.c: support replaying root commits

Junio C Hamano <gitster@pobox.com> writes:
Show 16 quoted lines
> Tian Yuchen <cat@malon.dev> writes:
>
>> 'git replay' does not support replaying root commits, as indicated by
>> the FIXME comment. If a user attempts to replay a root commit, the 'die()'
>> in 'replay_revisions()' will be called.
>
> Please be aware of your surroundings.  I think there are multiple
> topics that work on this file and among them there is a topic to
> address the same issue, if I am not mistaken.
>
> The list archive is your friend, for example, you can try these:
>
>     https://lore.kernel.org/git/?q=replay+root
>     https://lore.kernel.org/git/?q=dfn%3Areplay.c+%22root+commit%22
>
> Thanks.
Oops, sent without completing what I needed to say.

After you find out that other people are working in the vicinity, some of them even may be working on the same itch you had, instead of silently letting them continue, join the conversation and help them polish their work. You can give reviews, you can supply test cases, you can polish their documentation, etc.

Let's make this a team effort, not authors completing against each other, or authors working on their own thing without paying any attention to what others are doing.

Thanks.
Tian Yuchen· Mar 28, 2026, 03:24 UTC · re: Junio C Hamano · lore

Re: [PATCH v1] replay.c: support replaying root commits

Hi Junio,
Show 33 quoted lines
> Junio C Hamano <gitster@pobox.com> writes:
> 
>> Tian Yuchen <cat@malon.dev> writes:
>> 
>>> 'git replay' does not support replaying root commits, as indicated by
>>> the FIXME comment. If a user attempts to replay a root commit, the 
>>> 'die()'
>>> in 'replay_revisions()' will be called.
>> 
>> Please be aware of your surroundings.  I think there are multiple
>> topics that work on this file and among them there is a topic to
>> address the same issue, if I am not mistaken.
>> 
>> The list archive is your friend, for example, you can try these:
>> 
>>     https://lore.kernel.org/git/?q=replay+root
>>     https://lore.kernel.org/git/?q=dfn%3Areplay.c+%22root+commit%22
>> 
>> Thanks.
> 
> Oops, sent without completing what I needed to say.
> 
> After you find out that other people are working in the vicinity,
> some of them even may be working on the same itch you had, instead
> of silently letting them continue, join the conversation and help
> them polish their work.  You can give reviews, you can supply test
> cases, you can polish their documentation, etc.
> 
> Let's make this a team effort, not authors completing against each
> other, or authors working on their own thing without paying any
> attention to what others are doing.
> 
> Thanks.

Oops, looks like I missed a few threads. Before I started, I actually searched the mailing list to see if anyone else was working on this, maybe I used the wrong keywords.

Thanks for pointing out. I'll join the discussion. ;)
Regards, Yuchen

← back to recent threads