On Sun Mar 12 19:56:19 2017, SREZIC wrote:
Show quoted text> Right now I see the following test failure:
>
> ...
> perl -Mblib t/08timezone.t
> 1..4
> ok 1 - Comparing basic parameters with all parameters
> ok 2 - Comparing basic parameters with all parameters
> not ok 3 - Test DateTime sunrise interface
> # Failed test 'Test DateTime sunrise interface'
> # at t/08timezone.t line 60.
> # got: '06:05'
> # expected: '07:05'
> not ok 4 - Test DateTime sunset interface
> # Failed test 'Test DateTime sunset interface'
> # at t/08timezone.t line 61.
> # got: '17:58'
> # expected: '18:58'
> # Looks like you failed 2 tests of 4.
> ...
>
> Probably this happens only today (2017-03-12 19:55) --- if I run the
> test with -MTime::Fake=+1d or -1d, then this test passes. If TZ is
> relevant: I see it with a UTC timezone and with TZ=Europe/Berlin.
This happens on 2017-03-12 because (I guess), most of the USA switches to DST
this day. The test script uses "2017-03-12 00h00" to determine the current local
timezone offset, which is no longer the proper offset when the sun rises on
Los Angeles. It will happen again on the autumn DST switch.
UTC and Berlin and Paris and others are irrelevant: this test contains the
hard-coded value "America/Los_Angeles", a remainder of the tests as written by
the original author, Ron Hill. I did not think it would be useful to change it, so
I kept it.
I can change the test so it will use a noon datetime to compute the offset
instead of a midnight datetime. For America/Los_Angeles, it will work. Some
timezones switch to DST during the day. It would not work for them, but I do not
check them.
Thank you for your interest in Astro::Sunrise and for your work as a smoke tester.
Jean Forget
