[erlang-questions] calendar:local_time_to_universal_time_dst/1 - what this is for?

Raimo Niskanen <>
Mon Sep 15 14:11:16 CEST 2008


On Sun, Sep 14, 2008 at 11:20:20AM +0900, Kenji Rikitake wrote:
> I found a bug when I was playing around with mochiweb on R12B4.  It's
> not mochiweb's bug, but rather the Erlang/OTP time function bug.  See
> 
> http://www.erlang.org/pipermail/erlang-bugs/2008-September/000994.html
> 
> if you don't want to read ahead. (Summary: the "universaltime" in
> Erlang/OTP was not really the one if the wall clock is aware of leap
> seconds.)
> 
> I am writing this to ask a set of question about the following issue I
> was thinking last night.
> 
> When I tried to perform "gmake test" on one of my systems (yeah, it's
> FreeBSD, so GNU make is gmake), I faced a condition (in
> mochiweb_cookies:test/0) that httpd_util:rfc1123_date/1 returned a []
> (null list).  I further tracked it down and found
> calendar:local_time_to_universal_time_dst/1 (quoted below) was the
> source of the null list.  (I found that
> erlang:universaltime_to_localtime/1 and the corresponding C function
> univ_to_local() in erts/emulator/beam/erl_time_sup.c caused the problem,
> by NOT properly handling leap-second correction.)
> 
> I was wondering what this function
> calendar:local_time_to_universal_time_dst/1 was for and how this
> conversion function to be used.  I have the following questions:

It is rather clear in the manual:
http://www.erlang.org/doc/man/calendar.html

To flesh out,...  When you switch from normal time to daylight
saving time everyone skips an hour. Usually (always?) the hour
02.00-03.00, so instead of 02.00 the local time becomes 03.00.
One hour disappears in the spring. Everyone rises from bed one hour
earlier during the summer since it is already sunlight then
and the sun sets later, making the usable day longer,
especially the month after switch to DST (Daylight Saving Time)
and the month before switch from DST, for us closer to an
earth pole. So if you try to find which time 02.17 March 30
2008 CET corresponds in UTC you will notice that that
time never happened. The whole hour was skipped. Local time.

In the autumn when you go from DST you repeat the hour
02.00 to 03.00. So 02.34 Oct 26 2008 CET will happen twice,
and if you want to convert it to UTC you will have to know
which one it is.

> 
> * Why you need to double-check the whole time difference?

I do not understand this questions.

> 
> * In what situation should the results of this function really matter?

When your program needs to know the correct UTC time
you try to calculate from a local time. (?)

> 
> * What kind of action should an application program take if this
>   function returns a list of two possible UTC values for one DST
>   localtime value?
> 

To correctly calculate the UTC it must know if the point
in time is the one before the switch from DST or the one after.

> Maybe I'm perplexed just because I've never tried hard to code DST
> handling programs (Japan has no DST since 1950s).

Lucky you.

> 
> BTW I'd like to say thanks to mochiweb developers for making me aware of
> these rather non-trivial issues.
> 
> Regards,
> Kenji Rikitake
> 
> %% quoted from lib/stdlib-1.15.4/src/calendar.erl
> %% begin
> 
> -spec local_time_to_universal_time_dst(t_datetime1970()) -> [t_datetime1970()].
> local_time_to_universal_time_dst(DateTime) ->
>     UtDst = erlang:localtime_to_universaltime(DateTime, true),
>     Ut    = erlang:localtime_to_universaltime(DateTime, false),
>     %% Reverse check the universal times
>     LtDst = erlang:universaltime_to_localtime(UtDst),
>     Lt    = erlang:universaltime_to_localtime(Ut),
>     %% Return the valid universal times
>     case {LtDst,Lt} of
> 	{DateTime,DateTime} when UtDst =/= Ut ->
> 	    [UtDst,Ut];
> 	{DateTime,_} ->
> 	    [UtDst];
> 	{_,DateTime} ->
> 	    [Ut];
> 	{_,_} ->
> 	    []
>     end.
> 
> %% end
> 
> _______________________________________________
> erlang-questions mailing list
> 
> http://www.erlang.org/mailman/listinfo/erlang-questions

-- 

/ Raimo Niskanen, Erlang/OTP, Ericsson AB



More information about the erlang-questions mailing list