Re: [PATCH 06/10] object-file API: replace some use of check_object_signature()

From: Junio C Hamano <gitster@pobox.com>
Date: Feb 1, 2022, 19:16 UTC
Message-ID: <xmqqy22ucs7q.fsf@gitster.g>
In-Reply-To: <patch-06.10-2a065bf23da-20220201T144803Z-avarab@gmail.com>

Ævar Arnfjörð Bjarmason  <avarab@gmail.com> writes:

> Add a "hash_object_file_literally()" function to go with the existing
> "hash_object_file()" function. This is currently a wrapper for its
> sibling, but this change will allow us to change it to take an "enum
> object_type" in a subsequent commit.

This is a confusing renaming, because there is already a public function that exists under that name. And with this implementation, ...

Show 6 quoted lines

> +static void hash_object_file_literally(const struct git_hash_algo *algo, const void *buf,
> +				       unsigned long len, const char *type,
> +				       struct object_id *oid)
> +{
> +	hash_object_file(algo, buf, len, type, oid);
> +}

... it is dubious why we need it.

Show 6 quoted lines

> -int hash_object_file_literally(const void *buf, unsigned long len,
> -			       const char *type, struct object_id *oid,
> -			       unsigned flags)
> +int hash_write_object_file_literally(const void *buf, unsigned long len,
> +				     const char *type, struct object_id *oid,
> +				     unsigned flags)

This renaming is actually OK, if hash_object_file() which is also public is renamed to hash_write_object_file() at the same time.

But I would say we should try to avoid such name churn to the public namespace.