On Thu, Mar 14, 2019 at 01:05:03PM +0900, Junio C Hamano wrote: > Jeff King <peff@xxxxxxxx> writes: > > > I'm not sure it's really worth addressing (just because I don't think > > there's a good way to do it that isn't expensive). > > I do not think so, either. Not at this layer, anyway. > > If a "-x" command newly created an object whose prefix makes it > ambiguous against an existing object, we would not notice, unless we > refresh the loose object cache, that we now have two objects that > share the prefix. And we will find an object (the older one) with > the prefix in the stale cache without hitting "oh we do not find it, > let's retry" codepath. So to solve both sides of the coin, we'd > need invalidation either way, which obviously defeats the whole > caching concept, doesn't it? Right, that was the conclusion I came to. You cannot solve it in all cases without checking whether new objects have appeared for every lookup. I think that could be as cheap as a stat() on the pack directory and one on the appropriate loose directory, but I suspect even that might be measurably slower. -Peff
