Jeff King <peff@xxxxxxxx> writes: > So maybe: > > if (!oideq(real_oid, oid)) > > instead? It's a little slower, but the point of this is to diagnose and > die, so it's not exactly a hot path. :) Very true, including the part that the original is fine because it is localized and fairly obvious. For a public function, we cannot assume any additional constraints between oid and real_oid (other than they are of the same "struct object_id *" type) like the two pointers prepared locally in the original had, and use of oideq() would be more appropriate here. Thanks.
