timezone behavior for date and hwclock

[Denys Vlasenko]

On Tue, Apr 17, 2012 at 4:06 PM, Brecht Ameije <bam at abm.be> wrote:
> Hi,
> I sent the message below before, but no response.
> Maybe because it was in html instead of plain text?
> Or am I the only one with this problem?
> Brecht
>
> On 04/04/2012 14:56 Brecht Ameije wrote:
>>
>> Dear all,
>>
>> I'm trying to get the time zones right on my busybox embedded system.
>>
>> I'm in Belgium, this means we're on CET (UTC+1) and DST, thus UTC+2 At the
>> moment of the tests below, my local time is 2:35PM.
>>
>> My embedded system has no persistent time storage. So I use ntp to fetch the
>> time from the ntp pool.
>> This gives me UTC time. Setting up the environment var TZ takes me to local
>> time.
>>
>> When this is done, I set the hw clock to reflect my new system time.
>>
>> But now I come upon something strange: the "date -u" and the "hwclock -u"
>> commands give different results:
>>
>> $ export TZ=CET-1CEST,M3.5.0/2,M10.5.0/3
>> $ ntpd -q -p 0.europe.pool.ntp.org # get current time
>> $ hwclock -w                       # set hw clock
>> $ date; date -u; hwclock; hwclock -u; hwclock -l
>> Wed Apr  4 14:35:55 CEST 2012
>> Wed Apr  4 12:35:55 UTC 2012
>> Wed Apr  4 14:35:54 2012  0.000000 seconds
>> Wed Apr  4 16:35:54 2012  0.000000 seconds
>> Wed Apr  4 14:35:54 2012  0.000000 seconds

date -u means "show current time in UTC"
whereas
hwclock -u means "show rtc time in current timezone,
assuming rtc clock is kept in UTC".

Your rtc clock is kept in local time.
Therefore "hwclock -u" will misinterpret it.

-- 
vda