Re: [PATCH 3/3] index-pack: commit tree during outgoing link check

From: Junio C Hamano <gitster@pobox.com>
Date: Dec 3, 2024, 03:10 UTC
Message-ID: <xmqqa5ddfomb.fsf@gitster.g>
In-Reply-To: <2f2f0db78bf85c14ef132e1924ab5021298aace3.1733170252.git.jonathantanmy@google.com>

Jonathan Tan <jonathantanmy@google.com> writes:

> Subject: Re: [PATCH 3/3] index-pack: commit tree during outgoing link check

> Commit c08589efdc (index-pack: repack local links into promisor packs,
> 2024-11-01) seems to contain an oversight in that the tree of a commit
> is not checked.

I am having a hard time linking the subject with the statement. The verb "commit" should probably be something else, as you are not creating a tree object while checking, but I am not sure?

> The fix slows down a fetch from a certain repo at
> $DAYJOB from 2m2.127s to 2m45.052s, but in order to make the fetch
> correct, it seems worth it.

And "the fix" is not described so a reader is left wondering. Is the fix for an oversight of not checking merely to check it? IOW, is

    c08589efdc made outgoing links to be checked for commits, but
    failed to do so for trees.  Make sure we check both

what is happening?

Show 16 quoted lines

> In order to test this, we could create server and client repos as
> follows...
>
>  C   S
>   \ /
>    O
>
> (O and C are commits both on the client and server. S is a commit
> only on the server. C and S have the same tree but different commit
> messages.)
>
> ...and then, from the client, fetch S from the server.
>
> In theory, the client declares "have C" and the server can use this
> information to exclude S's tree (since it knows that the client has C's
> tree, which is the same as S's tree).

OK.

> However, it is also possible for
> the server to compute that it needs to send S and not O, and proceed
> from there;

If O, C, and S have all identical trees, then wouldn't such a test work well? At that point it does not matter which between O and C the server bases its decision to send S but not S's tree on, no?

In any case, will queue.  Thanks.

Show 22 quoted lines

> therefore the objects of C are not considered at all when
> determining what to send in the packfile. In order to prevent a test of
> client functionality from having such a dependence on server behavior, I
> have not included such a test.
>
> Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
> ---
>  builtin/index-pack.c | 1 +
>  1 file changed, 1 insertion(+)
>
> diff --git a/builtin/index-pack.c b/builtin/index-pack.c
> index 58d24540dc..338aeeadc8 100644
> --- a/builtin/index-pack.c
> +++ b/builtin/index-pack.c
> @@ -838,6 +838,7 @@ static void do_record_outgoing_links(struct object *obj)
>  		struct commit *commit = (struct commit *) obj;
>  		struct commit_list *parents = commit->parents;
>  
> +		record_outgoing_link(get_commit_tree_oid(commit));
>  		for (; parents; parents = parents->next)
>  			record_outgoing_link(&parents->item->object.oid);
>  	} else if (obj->type == OBJ_TAG) {