From: Patrick Steinhardt <ps@pks.im>
Date: Thu, 12 Mar 2026 10:22:55 GMT
Subject: Re: [PATCH v4 2/3] gpg-interface: introduce sign_buffer_with_key()
Message-ID: <abKT_50GVgBcj7op@pks.im>
In-Reply-To: <20260311173147.2336432-3-jltobler@gmail.com>

On Wed, Mar 11, 2026 at 12:31:46PM -0500, Justin Tobler wrote:
> diff --git a/gpg-interface.h b/gpg-interface.h
> index 789d1ffac4..a32741aeda 100644
> --- a/gpg-interface.h
> +++ b/gpg-interface.h
> @@ -83,6 +83,13 @@ size_t parse_signed_buffer(const char *buf, size_t size);
>  int sign_buffer(struct strbuf *buffer, struct strbuf *signature,
>  		const char *signing_key);
>  
> +/*
> + * Similar to `sign_buffer()`, but uses the default configured signing key as
> + * returned by `get_signing_key()` when the provided "signing_key" is NULL or
> + * empty. Returns 0 on success, non-zero on failure.
> + */
> +int sign_buffer_with_key(struct strbuf *buffer, struct strbuf *signature,
> +			 const char *signing_key);

I think this interface is a bit confusing, as you wouldn't really be
able to tell what the difference between `sign_buffer()` and
`sign_buffer_with_key()` is without having a deeper look. Naively, I
would expect the latter function to be the one that actually mandates
that the user provides a key, but it's the other way round.

Would it be preferable to instead extend `sign_buffer()` to take a flags
parameter and then introduce `SIGN_BUFFER_USE_DEFAULT_KEY` to make it
fall back to the configured signing key? If so, we could drop
`sign_commit_to_strbuf()` completely.

Patrick