On Thu, Jun 20, 2019 at 10:38:23AM -0700, Jonathan Tan wrote: > > @@ -154,7 +154,7 @@ static struct island_bitmap > > *create_or_get_island_marks(struct object *obj) > > khiter_t pos; > > int hash_ret; > > > > - pos = kh_put_sha1(island_marks, obj->oid.hash, &hash_ret); > > + pos = kh_put_oid_map(island_marks, obj->oid, &hash_ret); > > Thanks for doing this cleanup. The entire series (17 patches) look good > to me. The only remotely surprising thing to me was that OIDs are passed > by value on the stack, both for kh_get_oid_map() and kh_put_oid_map(), > but I see that this is how things currently work (and anyway, changing > this is beyond the scope of this patch set).
Thanks for reviewing. Yeah, the pass-by-value surprised me too, as it's been a while since I've had to touch khash. I think it all cancels out performance-wise because of the inlining, but it might be a fun thing to experiment with. -Peff
