On Tue, Oct 06, 2020 at 12:13:03PM +0200, Phil Sutter wrote: [...] > Put more simply: I'm not assuming *any* ordering in input, not even > grouped input by name. But I chose the algorithm to optimize for input > which is grouped by name and family and sorted by ascending revision > within groups, as that is mostly the case in in-tree extensions. Looks good, thanks for explaining.
