Re: [chrony-dev] macOS 10.13 beta 5 - adjtime() is fixed.
On Tue, Aug 08, 2017 at 06:37:00PM +1200, Bryan Christianson wrote: > One thing I did notice. On macOS 10.12 the frequency on the machine I use > seems to mostly hover in the range 16 to 19 ppm, depending on temperature, > other workload etc. > > On macOS 10.13, it seems to be more like 17.5 to 20.5 ppm, again depending on > temperature, workload. > > I was a little surprised to note the difference because it is the same > hardware in both cases. Maybe I need to test that a little more to see if its > a real or imaginary (on my part) difference. I guess a difference of 1.5 ppm could be explained by new kernel timekeeping implementation, different rounding, etc. Does it change across reboots? For instance, with the tsc clocksource on Linux it's expected to see some variation as the calibration takes just a second or so. -- Miroslav Lichvar -- To unsubscribe email chrony-dev-requ...@chrony.tuxfamily.org with "unsubscribe" in the subject. For help email chrony-dev-requ...@chrony.tuxfamily.org with "help" in the subject. Trouble? Email listmas...@chrony.tuxfamily.org.
Re: [chrony-dev] macOS 10.13 beta 5 - adjtime() is fixed.
> On 8/08/2017, at 6:25 PM, Miroslav Lichvar wrote: > > That's good news. Thanks for the update. > > I assume chronyd is working correctly when using both adjtime() and > ntp_adjtime()? When starting with an initial offset over 1 second and > not allowed to make a step (makestep/initstepslew), the clock is > slewed as expected, with no oscillations, etc? I didn't specifically test that. I just observed that it corrected the initial offset of around 1 sec in less than 10 minutes. I'll do more detailed tests tomorrow. One thing I did notice. On macOS 10.12 the frequency on the machine I use seems to mostly hover in the range 16 to 19 ppm, depending on temperature, other workload etc. On macOS 10.13, it seems to be more like 17.5 to 20.5 ppm, again depending on temperature, workload. I was a little surprised to note the difference because it is the same hardware in both cases. Maybe I need to test that a little more to see if its a real or imaginary (on my part) difference. Bryan Christianson br...@whatroute.net -- To unsubscribe email chrony-dev-requ...@chrony.tuxfamily.org with "unsubscribe" in the subject. For help email chrony-dev-requ...@chrony.tuxfamily.org with "help" in the subject. Trouble? Email listmas...@chrony.tuxfamily.org.
Re: [chrony-dev] macOS 10.13 beta 5 - adjtime() is fixed.
On Tue, Aug 08, 2017 at 08:11:29AM +1200, Bryan Christianson wrote: > New release of macOS 10.13 out this morning and adjtime() appears to be > behaving correctly now. I think we should leave the workaround/tests in place > until 10.13 is released because Apple might break it again between now and > then :) > > I re-ran the ntpadjtime and adjtime tests in in chrony - no change to the > previous results in maximum skew. That's good news. Thanks for the update. I assume chronyd is working correctly when using both adjtime() and ntp_adjtime()? When starting with an initial offset over 1 second and not allowed to make a step (makestep/initstepslew), the clock is slewed as expected, with no oscillations, etc? -- Miroslav Lichvar -- To unsubscribe email chrony-dev-requ...@chrony.tuxfamily.org with "unsubscribe" in the subject. For help email chrony-dev-requ...@chrony.tuxfamily.org with "help" in the subject. Trouble? Email listmas...@chrony.tuxfamily.org.
[chrony-dev] macOS 10.13 beta 5 - adjtime() is fixed.
New release of macOS 10.13 out this morning and adjtime() appears to be behaving correctly now. I think we should leave the workaround/tests in place until 10.13 is released because Apple might break it again between now and then :) I re-ran the ntpadjtime and adjtime tests in in chrony - no change to the previous results in maximum skew. Bryan Christianson br...@whatroute.net -- To unsubscribe email chrony-dev-requ...@chrony.tuxfamily.org with "unsubscribe" in the subject. For help email chrony-dev-requ...@chrony.tuxfamily.org with "help" in the subject. Trouble? Email listmas...@chrony.tuxfamily.org.