On 27.09.2019 10:19, Rhialto wrote: > On Thu 26 Sep 2019 at 01:15:23 +0200, Kamil Rytarowski wrote: >> Is this patch correct? >> >> http://netbsd.org/~kamil/patch-00168-kern_rndq.c-avoid-overflow.txt >> >> This code will map the corner case into defined semantics, treating >> delta INT32_MIN as INT32_MIN+1. > > I don't know if it is important, but it makes the "normalized" delta > value INT32_MAX more likely than it probably should be, since it can > result from "raw" delta values INT32_MIN, INT32_MIN+1, and INT32_MAX. > > Similarly, 0 is underrepresented as output, since it only can result > from 0. > > Other positive results x can come from x and -x as inputs. > > -Olaf. > If you think that it shall be represented a 0, I can go for it.
Attachment:
signature.asc
Description: OpenPGP digital signature