Re: [PATCH 6/6] pack-objects: reuse on-disk deltas for thin "have" objects

From: Jeff King <peff@peff.net>
Date: Aug 17, 2018, 23:32 UTC
Message-ID: <20180817233234.GA4472@sigill.intra.peff.net>
In-Reply-To: <CAGZ79kY9o_38WRX2TXC1JMA+9Y_ihf_jK8-f2pcYA+cYPF6B4A@mail.gmail.com>

On Fri, Aug 17, 2018 at 03:57:18PM -0700, Stefan Beller wrote:

Show 15 quoted lines

> On Fri, Aug 17, 2018 at 2:06 PM Jeff King <peff@peff.net> wrote:
> >
> > When we serve a fetch, we pass the "wants" and "haves" from
> > the fetch negotiation to pack-objects. That tells us not
> > only which objects we need to send, but we also use the
> > boundary commits as "preferred bases": their trees and blobs
> > are candidates for delta bases, both for reusing on-disk
> > deltas and for finding new ones.
> >
> > However, this misses some opportunities. Modulo some special
> > cases like shallow or partial clones, we know that every
> > object reachable from the "haves" could be a preferred base.
> > We don't use them all for two reasons:
> 
> s/all/at all/ ?

No, I meant "we don't use all of them". As in the Pokemon "gotta catch 'em all" slogan. ;)

Probably writing out "all of them" is better.

Show 15 quoted lines

> > The first is that check_object() needs access to the
> > relevant information (the thin flag and bitmap result). We
> > can do this by pushing these into program-lifetime globals.
> 
> I discussed internally if extending the fetch protocol to include
> submodule packs would be a good idea, as then you can get all
> the superproject+submodule updates via one connection. This
> gives some benefits, such as a more consistent view from the
> superproject as well as already knowing the have/wants for
> the submodule.
> 
> With this background story, moving things into globals
> makes me sad, but I guess we can flip this decision once
> we actually move towards "submodule packs in the
> main connection".

I don't think it significantly changes the existing code, which is already relying on a ton of globals (most notably to_pack). The first step in doing multiple packs in the same process is going to be to shove all of that into a "struct pack_objects_context" or similar, and these can just follow the rest.

Show 13 quoted lines

> >
> > The second is that the rest of the code assumes that any
> > reused delta will point to another "struct object_entry" as
> > its base. But by definition, we don't have such an entry!
> 
> I got lost here by the definition (which def?).
> 
>   The delta that we look up from the bitmap, doesn't may
>   not be in the pack, but it could be based off of an object
>   the client already has in its object store and for that
>   there is no struct object_entry in memory.
> 
> Is that correct?

Right, we are interested in objects that we _couldn't_ find a struct for. I agree this could be more clear.

-Peff