Re: [PATCH 7/7] Factor ref updating out of fetch_with_import

From: Sverre Rabbelier <srabbelier@gmail.com>
Date: Oct 29, 2009, 16:32 UTC
Message-ID: <fabb9a1e0910290932u45c9c416m4d0ba0a8b2f5b01d@mail.gmail.com>
In-Reply-To: <20091029155607.GA10505@spearce.org>

Heya,

On Thu, Oct 29, 2009 at 08:56, Shawn O. Pearce <spearce@spearce.org> wrote:

Show 7 quoted lines

>> > I certainly have to wonder... if this is done in both fetch and
>> > clone, why isn't it just part of fetch_refs?
>>
>> Because clone does not use fetch_refs, or am I missing something?
>
> Hmmph.  Weird.  Its been a while since I last looked at this code,
> maybe I misunderstood it.

Unless you mean transport_fetch_refs? It can't be done in transport_fetch_refs because the foreign helper transport needs to update all refs, including those that wanted_peer_refs decided not to fetch. Otherwise the 'HEAD' ref will not be updated, and it is left at all zeros. It is not as obvious that that is a problem in this patch, but when Junio merges with Nico's [5bdc32d3e "make 'git clone' ask the remote only for objects it cares about"] the code looks like this:

		if (refs) {
			struct ref *ref_cpy;
			mapped_refs = wanted_peer_refs(refs, refspec);
			ref_cpy = copy_ref_list(mapped_refs);
			transport_fetch_refs(transport, ref_cpy);
			if (transport->update_refs)
			{
				ref_cpy = copy_ref_list(refs);
				transport->update_refs(transport, ref_cpy);
				refs = ref_cpy;
				mapped_refs = wanted_peer_refs(refs, refspec);
			}
		}

Does it make more sense now? transport_fetch_refs gets only a limited view of the refs, so it cannot pass all the refs to transport_update_refs as needed.

-- 
Cheers,

Sverre Rabbelier