Package: adjtimex Version: 1.20-5 Severity: normal Adjtimex --review suggests a little inaccurate RTC adjustement (to put in /etc/adjtime): That needs to be expressed in seconds per day of CMOS time. Not of reference time, nor somewhere in between.
Example with this test-clocks.log file: | 1970-01-01 00:00 0 0.001 0 10000 0 n 0 n | 1972-09-28 00:00 86486400 0.001 86486400 10000 0 n 86400000 y 1001 days elapsed, but the CMOS is slow and counted only 1000 days. The RTC is slow by -1000 PPM of it's own (wrong) time, or by -999.000999 PPM of real reference time. Adjtimex 1.20 review is: | $ adjtimex --review=test-clocks.log | start finish days sys - cmos (ppm) | start finish days cmos_error (ppm) | Thu Jan 1 01:00:00 1970 Thu Sep 28 01:00:00 1972 1001.0000 -999.500250 +- 0.000016 | start finish days sys_error (ppm) | least-squares solution: | cmos_error = -1000 +- 0 ppm suggested adjustment = 86.3568 sec/day | current adjustment = -0.2017 sec/day | sys_error = 0 +- 100000 ppm (no suggestion) | current tick = 10000 freq = -1053471 | note: clock variations and unstated data errors may mean that the | least squares solution has a bigger error than estimated here Adjtimex calculates cmos_error as mean -999.500250 PPM, and suggests an adjustment of +86.3568 seconds per day. That's wrong: The /etc/adjtime adjustment to correct this RTC slowness must be exactly +86.4 seconds per day. At boot time (or later), hwclock --hctosys has first no idea of what is real reference time. It takes current time, elapsed days, and adjustment all expressed in CMOS wrong time, and calculates real time doing now + elapsed * adjustment. In my example 86400000 + 1000 * 86.4 = 86486400. Adjtimex does the same when correcting read CMOS time in --compare. -- System Information Debian Release: 3.0 Architecture: i386 Kernel: Linux Kassad 2.4.18 #6 mar jan 25 15:17:00 CET 2005 i586 Locale: LANG=fr_FR.iso-8859-1, LC_CTYPE=fr_FR.iso-8859-1 Versions of packages adjtimex depends on: ii debconf 1.0.32 Debian configuration management sy ii libc6 2.2.5-11.5 GNU C Library: Shared libraries an -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
