On Mon, Jun 19, 2017 at 08:51:00AM -0700, Junio C Hamano wrote: > Jeff King <peff@xxxxxxxx> writes: > > > On Sun, Jun 18, 2017 at 03:39:40PM +0200, Michael Haggerty wrote: > > > >> As before, the second patch is optional. If it is omitted, it might > >> flush out any other bugs like this one in client code. If it is > >> included, regressions are less likely, but we won't learn about other > >> misuses of the API. I have no strong opinion either way. > > > > My feeling is still slightly towards "don't include", but I also don't > > have a strong opinion. > > I am inclined to the "don't include 2/2 and cook 1/2 alone but a bit > longer" approach. I don't think it helps to cook 1/2 longer. The assertion that triggered is already in master. Patch 1 fixes one case in an obviously-correct way. Patch 2 is just about guessing whether _other_ cases may trigger. So if we wanted to cook longer we'd have to revert b9c8e7f2f (prefix_ref_iterator: don't trim too much, 2017-05-22), but I don't think that's worth doing. -Peff