From: Junio C Hamano <gitster@pobox.com>
Date: Wed, 23 Sep 2026 20:03:31 GMT
Subject: Re: [PATCH v4 3/3] fetch, remote: retain old OIDs when pruning refs
Message-ID: <xmqq4iffag6k.fsf@gitster.g>
In-Reply-To: <6f34853c79625794d2eb364d227660be57d1539b.1790113781.git.maciej.ciemborowicz@gmail.com>

Maciej Ciemborowicz <maciej.ciemborowicz@gmail.com> writes:

> diff --git a/builtin/fetch.c b/builtin/fetch.c
> index b662216bf..95789edb8 100644
> --- a/builtin/fetch.c
> +++ b/builtin/fetch.c
> @@ -1471,22 +1471,29 @@ static int prune_refs(struct display_state *display_state,
> ...
>  		} else {
> +			for (ref = stale_refs; ref; ref = ref->next) {
> +				string_list_append(&refnames, ref->name);
> +				oid_array_append(&old_oids, &ref->new_oid);
> +			}
>  			result = refs_delete_refs(get_main_ref_store(the_repository),
>  						  "fetch: prune", &refnames,
> -						  NULL, 0);
> +						  &old_oids, &failed_refs, 0);

Isn't adding a new parameter to refs_delete_refs() needed before
this step?  The corresponding changes to refs.[ch] was done in
[1/3], and the fact that the callsite receives this update to add an
extra parameter this late in the series means [1/3] and [2/3] does
not even compile, right?

Thanks.