On Sun, Nov 15, 2015 at 11:55 PM, Johan Herland <johan@xxxxxxxxxxx> wrote: > Additionally, if we suspect that passing non-notes trees to read-only > operations will be a common error, we could add a simple heuristic to > the notes code, to warn (or even abort) if we strongly suspect that we > are reading in a non-notes tree. For example, if the ratio of > non-notes to notes entries goes above, say, 1:1 (or even 10:1), then > what we're reading is probably not a proper notes tree... > I agree here for this part, a possible heuristic check would maybe be valuable.. but not sure it's super worth the effort. I doubt it would be a common error, and I don't think the issues above would actually cause too many problems. The main other issue is how to get notes DWIM things to work for all cases where we want to use notes refs, since right now the DWIM is basically done at the top level and only handles notes like things. The problem with it is that if you specify a full ref that *isn't* refs/notes, you will always prefix it with refs/notes, like so: refs/remote-notes/origin => refs/notes/refs/remote-notes/origin, This makes it really difficult to expand a ref. However, Julio seemed to think this was a possibly valuable expansion under normal circumstances. The current solution is to try to do a normal lookup first and only use the notes DWIM after we fail a lookup, which I think is what the above patch attempts to do. This seems ok enough to me. Regards, Jake > > ...Johan > -- To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html