Обсуждение: Bug in tzdata 2022g
There appears to be a bug in the latest postgres release, which uses tzdata 2022g to reflect the recent DST change in Mexico.
When I convert a UTC timestamp to Mexico City time in 13.10 (Debian 13.10-1.pgdg110+1), I get a one hour difference, which is wrong:
select timezone('America/Mexico_City', '2023-05-06T08:00+00:00'), timezone('America/Bogota', '2023-05-06T08:00+00:00');
timezone | timezone
---------------------+---------------------
2023-05-06 02:00:00 | 2023-05-06 03:00:00
---------------------+---------------------
2023-05-06 02:00:00 | 2023-05-06 03:00:00
However, when I run it in PostgreSQL 12.10 (Ubuntu 12.10-1.pgdg20.04+1), I get the right answer (no time difference):
timezone | timezone
---------------------+---------------------
2023-05-06 03:00:00 | 2023-05-06 03:00:00
---------------------+---------------------
2023-05-06 03:00:00 | 2023-05-06 03:00:00
Nacho Caballero <nachocab@gmail.com> writes: > There appears to be a bug in the latest postgres release, which uses tzdata > 2022g to reflect the recent DST change in Mexico. I see no bug here. America/Mexico_City is reported as being -06 all year round, which agrees with what the tzdb commentary says: # From Paul Eggert (2022-10-28): # The new Mexican law was published today: # https://www.dof.gob.mx/nota_detalle.php?codigo=5670045&fecha=28/10/2022 # This abolishes DST except where US DST rules are observed, # and in addition changes all of Chihuahua to -06 with no DST. Meanwhile, Bogota has been -05 year-round since the nineties. If you disagree with either of these conclusions, you had better provide some solid evidence to back it up; and the place to be complaining to is the tzdata maintainers, not us. > However, when I run it in PostgreSQL 12.10 (Ubuntu 12.10-1.pgdg20.04+1), I > get the right answer (no time difference): Apparently, your PG 12.10 installation is using some pre-2022f version of tzdata. Mexico didn't abolish DST until late last year. regards, tom lane
Thanks to the reply, Tom. My bad, I was being misled by savvytime.com, startpage.com and search.brave.com.
Other providers like timeanddate.com, google.com and https://www.cenam.mx/hora_oficial/default2.aspx have the correct time.
On Sat, May 6, 2023 at 4:50 PM Tom Lane <tgl@sss.pgh.pa.us> wrote:
Nacho Caballero <nachocab@gmail.com> writes:
> There appears to be a bug in the latest postgres release, which uses tzdata
> 2022g to reflect the recent DST change in Mexico.
I see no bug here. America/Mexico_City is reported as being -06
all year round, which agrees with what the tzdb commentary says:
# From Paul Eggert (2022-10-28):
# The new Mexican law was published today:
# https://www.dof.gob.mx/nota_detalle.php?codigo=5670045&fecha=28/10/2022
# This abolishes DST except where US DST rules are observed,
# and in addition changes all of Chihuahua to -06 with no DST.
Meanwhile, Bogota has been -05 year-round since the nineties.
If you disagree with either of these conclusions, you had better
provide some solid evidence to back it up; and the place to be
complaining to is the tzdata maintainers, not us.
> However, when I run it in PostgreSQL 12.10 (Ubuntu 12.10-1.pgdg20.04+1), I
> get the right answer (no time difference):
Apparently, your PG 12.10 installation is using some pre-2022f
version of tzdata. Mexico didn't abolish DST until late last year.
regards, tom lane