On Sun, Jul 09, 2017 at 12:57:07AM -0700, Junio C Hamano wrote: > Junio C Hamano <gitster@xxxxxxxxx> writes: > > >> +Multiple occurrences of the same section are all logically merged. (There's > >> +no special treatment for variables defined multiple times across physically > >> +different sections, the variable is simply made multivalued.) > >> + > > > > Looks correct; it's a bit surprising that we didn't already mention > > multiple occurrences, but I do not find it so this is a good > > addition. > > > > I do not see a strong reason for () around the second sentence, > > though. > > After re-reading this, I am wondering if the "made multivalued" > above need to be made less confusing. For a variable that is single > valued, the usual "the last one wins" rule is applied---it's not > like > > [user] > name = astian > name = gitster > > makes user.name "gitster" while > > [user] > name = astian > [user] > name = gitster > > makes user.name "multivalued". In either case, the last one wins, > and it comes not from the syntax but from the semantics assigned to > the variable in question. What we want to say is not "multi-valued", > but the above two forms both mean the same thing. > > But perhaps I am needlessly worried about possible confusion too > much. I dunno. FWIW, the use of "multivalued" here tickled my spider sense, too. I think when talking on the list we generally reserve "multivalued" for true "we expect this to be a list" variables. But the only mention of "multivalued" in the config documentation seems to be: Some variables may appear multiple times; we say then that the variable is multivalued. I think the proposed use is consistent with that (and that line is only 2 paragraphs above the proposed paragraph). -Peff
