Re: [PATCH 07/16] repack: track the ODB source via existing_packs
- From
Taylor Blau <me@ttaylorr.com>
- Date
- Apr 14, 2026, 22:24 UTC
- Message-ID
- <ad6+jMHcwatRRcQa@nand.local>
- In-Reply-To
- <20260407060439.GB714878@coredump.intra.peff.net>
On Tue, Apr 07, 2026 at 02:04:39AM -0400, Jeff King wrote:
Show 19 quoted lines
> Makes sense. I wonder if we still need the repo field here:
>
> > diff --git a/repack.h b/repack.h
> > index bc9f2e1a5de..c0e9f0ca647 100644
> > --- a/repack.h
> > +++ b/repack.h
> > @@ -56,6 +56,7 @@ struct packed_git;
> >
> > struct existing_packs {
> > struct repository *repo;
> > + struct odb_source *source;
> > struct string_list kept_packs;
> > struct string_list non_kept_packs;
> > struct string_list cruft_packs;
>
> since we are going to mostly be operating on one odb source. But I think
> it gets weird asking about things like "which hash algorithm to use",
> because then you end up back at source->odb->repo. Plus there are some
> parts that seem to look at the whole repo.I think that doing so is possible, though it's not totally straightforward. There are a handful of places where you really truly do want to work with the repo (e.g., when calling trace2_region_enter(), which requires a repository pointer as its final argument).
I do think that we are probably over-relying on the repository pointer in the existing_packs struct after this patch, especially with the ODB source refactorings that have been going on.
> So I think it gets complicated and is well off the path to our goal, so > not worth thinking further about for now.
I agree.
Thanks, Taylor