Re: [PATCH v2 4/5] odb: use oidmap_clear_with_free() to release replace_map entries

From: Junio C Hamano <gitster@pobox.com>
Date: Mar 2, 2026, 22:35 UTC
Message-ID: <xmqq1pi1yi4s.fsf@gitster.g>
In-Reply-To: <20260302200018.75731-5-kuforiji98@gmail.com>

Seyi Kuforiji <kuforiji98@gmail.com> writes:

Show 7 quoted lines

> +static void free_replace_map_entry(void *e)
> +{
> +	struct replace_object *entry =
> +		container_of(e, struct replace_object, original);
> +	free(entry);
> +}
> +

The same comment as [PATCH v2 3/5].

Show 7 quoted lines

> @@ -1109,7 +1117,8 @@ void odb_free(struct object_database *o)
>  
>  	free(o->alternate_db);
>  
> -	oidmap_clear(&o->replace_map, 1);
> +	if (o->replace_map_initialized)
> +		oidmap_clear_with_free(&o->replace_map, free_replace_map_entry);

It is a bit unfortunate that we need to know how o->replace_map is initialized and maintained. I wondered if we can do this without peeking into o->replace_map_initialized, but o->replace_map is already an instance of the map, not a pointer that points at a lazily initialized instance of a map, so that cannot be done (and we would not have replace_map_initialized member in the object_database struct in the first place, if we can tell if o.replace_map needs clearing by simply looking at it).

So, all OK, I guess.

>  	pthread_mutex_destroy(&o->replace_mutex);
>  
>  	odb_close(o);