On Fri, Jun 12, 2020 at 11:51 AM Phillip Wood [off-list ref] wrote:

quoted

It was a stepping stone towards that, the git_path mechanism that is
used to create git_path_cherry_pick_head() etc was changed to take a
struct repository so it could support submodules without forking a
separate process. However are still plenty of places where the sequencer
code assumes a single repository (it calls update_ref(), delete_ref(),
commit_tree_extended(), ...) and the two contributors who did a lot of
that work have moved on. With that in mind perhaps we'd be better off
just using ref_exists() and delete_ref() in this conversion. The call
sites will be easy enough to fixup if those functions are converted to
take a struct repository in the future and the result of this patch
series will be nicer. I've cc'd dscho and Junio to see what they think.

In the end I put some patches together that change delete_ref(),
update_ref() and ref_exists() to take a struct repository*. I'll clean
them up and post them next week. Hopefully that will mean that this
series can then use those functions when converting unlink() etc which
will avoid having to expose a separate api for pseudo refs.

Sounds good.  Can you CC me on the patches?

-- 
Han-Wen Nienhuys - Google Munich
I work 80%. Don't expect answers from me on Fridays.
--

Google Germany GmbH, Erika-Mann-Strasse 33, 80636 Munich

Registergericht und -nummer: Hamburg, HRB 86891

Sitz der Gesellschaft: Hamburg

Geschäftsführer: Paul Manicle, Halimah DeLaine Prado