*** IF YOU CAN HELP FIX ANY OF THESE THINGS, PLEASE DO! *** 970711: Look Real Hard at changing the key stuff from u_long to u_int32. 970711: Make sure it's safe to convert proto_config's 2nd argument from u_long to u_int32. Watch "set" in ntp_request.c:setclr_flags(). 970318: in hourly_stats(?), squawk if the magnitude of the drift is, say, >400. 970301: Implement the utmp/wtmp timestamping on time steps. 970210: Find a way to dump the current configuration to either syslog or a file. CPP macros we need to handle: SUN_3_3_STINKS (this will be easy to handle if anybody still needs it) Problems that need to be fixed: - Get rid of the old SYS_* macros: (It's worth noting that any code that would have been "enabled" by any of these macros has not been used since 5.83, and there have been very few complaints...) SYS_44BSD: authstuff/md5driver.c SYS_BSDI: authstuff/md5driver.c SYS_DECOSF1: util/ntptime.c SYS_DOMAINOS: parseutil/dcfd.c xntpd/ntpd.c SYS_HPUX: kernel/sys/ppsclock.h ntpdate/ntpdate.c ntptrace/ntptrace.c SYS_PTX: libntp/machines.c SYS_SOLARIS: libparse/parse.c libparse/clk_trimtsip.c xntpd/ntp_io.c xntpd/ntp_refclock.c xntpd/ntpd.c SYS_SUNOS4: libparse/parse.c libparse/clk_trimsip.c SYS_WINNT: include/ntp.h include/ntp_fp.h include/ntp_if.h include/ntp_machine.h include/ntp_select.h include/ntp_syslog.h include/ntp_unixtime.h include/ntpd.h libntp/libntp.mak libntp/machines.c libntp/mexit.c libntp/msyslog.c libntp/systime.c ntpdate/ntpdate.c ntpdate/ntpdate.mak ntpq/ntpq.c ntpq/ntpq.mak ntpq/ntpq_ops.c ntptrace/ntptrace.c ntptrace/ntptrace.mak xntpd/ntp_config.c xntpd/ntp_filegen.c xntpd/ntp_intres.c xntpd/ntp_io.c xntpd/ntp_loopfilter.c xntpd/ntp_peer.c xntpd/ntp_proto.c xntpd/ntp_refclock.c xntpd/ntp_timer.c xntpd/ntp_unixclock.c xntpd/ntp_util.c xntpd/ntpd.c xntpd/xntpd.mak xntpdc/ntpdc.c xntpdc/xntpdc.mak - config.guess might need help to identify: Fujitsu's UXP --enable-adjtime-is-accurate --enable-step-slew Unixware --enable-adjtime-is-accurate --enable-tick=10000 --enable-tickadj=80 --enable-udp-wildcard --disable-step-slew DomainOS --enable-adjtime-is-accurate --disable-kmem --enable-tick=1000000 OpenVMS --enable-slew-always --enable-hourly-todr-sync Is adjtime accurate on ALL sysv4* machines? Can we identify DomainOS with *-apollo-* ? Do we catch all Unixware machines with *-univel-sysv* ? - How can we test if UDP_WILDCARD_DELIVERY should be used? - Combine enable-step-slew and enable-ntpdate-step - Make sure enable-hourly-todr-sync is always disabled What about NextStep and OpenVMS, where hourly TODR sync used to be enabled? - Check dcfd.c for variables that need to be volatile.