From: Junio C Hamano Date: Wed, 30 Mar 2011 22:28:43 GMT Subject: Re: [RFC 1/3] Alternates API: Untangle the interface Message-ID: <7v1v1omeec.fsf@alter.siamese.dyndns.org> In-Reply-To: <5ed7886ec5e12c8e3a5f45d19625a4f5d2cdd38a.1301521243.git.mfwitten@gmail.com> Michael Witten writes: > diff --git a/transport.c b/transport.c > index a02f79a..c61723f 100644 > --- a/transport.c > +++ b/transport.c > @@ -1190,14 +1190,13 @@ literal_copy: > return xstrdup(url); > } > > -int refs_from_alternate_cb(struct alternate_object_database *e, void *cb) > +int refs_from_alternate_cb(struct alternate_object_database *e, alternate_ref_fn ref_fn) > { Adding an extra "fn" is Ok, but I'd rather not to see removal of generic callback data pointer if the _only_ reason you are doing so is "currently nobody uses it". Besides, I see "a function pointer and a data pointer cannot be held in a single variable" as an academic mental mastu^wexercise that is useless in real life. Are there real platforms that matter where (void *) cannot hold a pointer to a function?