christos@ wrote: > | > -static u_int abcksum(void *); > | > +static void abcksum(void *); > | > | Changing existing functions requires more tests and > | it's annoying for maintainers, especially for netbsd-7. > > Yes, taking this approach to the extreme we should never switch compilers.
Please stop such "extreme or nothing" approach if you cannot maintain it. > | You should also consider about current MI dkcksum() (and dkcksum_sized()) > | implementation in sys/kern/subr_disk.c. The MI dkcksum() just does > | calculate a sum of the label and magic numbers are handled by callers: > | >> label->d_checksum = 0; > | >> label->d_checksum = dkcksum(label); > | > | It looks better to use consistent strategies for both MI/MD sums. > > I think that the only solution you'll like is yours, so please do it > your way. Could you read and answer my another post? http://mail-index.netbsd.org/source-changes-d/2014/11/23/msg007400.html Taylor claims the existing abcksum() also violates aliasing rule. http://mail-index.netbsd.org/source-changes-d/2014/11/23/msg007402.html If he is correct it's no sense to tweak only functions complained by current gcc48. --- Izumi Tsutsui