From: Christian Couder Date: Thu, 02 Apr 2026 06:34:59 GMT Subject: Re: [PATCH 01/16] promisor-remote: try accepted remotes before others in get_direct() Message-ID: In-Reply-To: On Thu, Mar 26, 2026 at 1:20 PM Patrick Steinhardt wrote: > > To better reflect the agreement between client and server, let's make > > promisor_remote_get_direct() try the accepted promisor remotes before > > the non-accepted ones. > > Interesting, and it feels sensible to me. Is it documented anywhere that > the ordering of announced remotes is actually important? I have documented it in the next version of this patch (which I will send soon) by adding the following to "Documentation/gitprotocol-v2.adoc": "The promisor remotes that the client accepted will be tried before the other configured promisor remotes when the client will attempt to fetch missing objects." > > Ensuring that accepted remotes are preferred will be even more > > important if in the future a mechanism is developed to allow the > > client to auto-configure remotes that the server advertises. This will > > in particular avoid fetching from the server (which is already > > configured as a promisor remote) before trying the auto-configured > > remotes, as these new remotes would likely appear at the end of the > > config file, and as the server might not appear in the > > `extensions.partialClone` config variable. > > Not quite sure I correctly understand this paragraph. Is the idea that > in the future, we might not even store announced promisors in the config > at all but simply use whatever the server announces on any given fetch? Currently the promisor remotes are fetched from in the same order as their order in the config file, except that a remote whose name appears in the `extensions.partialClone` config variable is fetched from last. The idea is that with `promisor.acceptFromServerUrl` it will be possible to store accepted promisors in the config, but the resulting config order will not be the best for fetching. Especially if the main remote doesn't appear in the `extensions.partialClone` config variable it will likely be tried before the accepted promisors which we do not want. So fetching first from the accepted promisors will also make things work better when `promisor.acceptFromServerUrl` will be implemented. > > diff --git a/promisor-remote.c b/promisor-remote.c > > index 96fa215b06..3f8aeee787 100644 > > --- a/promisor-remote.c > > +++ b/promisor-remote.c > > @@ -268,11 +268,37 @@ static int remove_fetched_oids(struct repository *repo, > > return remaining_nr; > > } > > > > +static int try_promisor_remotes(struct repository *repo, > > + struct object_id **remaining_oids, > > + int *remaining_nr, int *to_free, > > + bool accepted_only) > > +{ > > + struct promisor_remote *r = repo->promisor_remote_config->promisors; > > + > > + for (; r; r = r->next) { > > + if (accepted_only && !r->accepted) > > + continue; > > + if (!accepted_only && r->accepted) > > + continue; > > This can be simplified to `if (!accepted_only != !r->accepted)`. Right, but I think we should go for `if (accepted_only != r->accepted)`. It's even simpler and works because `r->accepted` is a 1 bit bitfield and `accepted_only` a bool. > Also, can we maybe add a test for this to verify that we use the correct > ordering now? Yeah, I have added tests for this in the new version. Thanks.