Обсуждение: BUG #16123: DST not respected for America/Sao_Paulo in `timestamp` function

The following bug has been logged on the website:

Bug reference:      16123
Logged by:          Thushara Wijeratna
Email address:      thushw@gmail.com
PostgreSQL version: 11.5
Operating system:   MacOS (Mojave)
Description:

maestro_development=# select timestamp with time zone '2019-12-01T02:00:00Z'
at time zone 'America/Sao_Paulo';
      timezone
---------------------
 2019-11-30 23:00:00
(1 row)

From Nov 3, America/Sao_Paulo should be offset only 2 hours from UTC due to
DTS. I would expect the timestamp to be December 1st midnight.

https://www.travelmath.com/time-zone/America/Sao_Paulo

On Mon, Nov 18, 2019 at 6:38 PM PG Bug reporting form
<noreply@postgresql.org> wrote:
> The following bug has been logged on the website:
>
> Bug reference:      16123
> Logged by:          Thushara Wijeratna
> Email address:      thushw@gmail.com
> PostgreSQL version: 11.5
> Operating system:   MacOS (Mojave)
> Description:
>
> maestro_development=# select timestamp with time zone '2019-12-01T02:00:00Z'
> at time zone 'America/Sao_Paulo';
>       timezone
> ---------------------
>  2019-11-30 23:00:00
> (1 row)
>
> From Nov 3, America/Sao_Paulo should be offset only 2 hours from UTC due to
> DTS. I would expect the timestamp to be December 1st midnight.
>
> https://www.travelmath.com/time-zone/America/Sao_Paulo

I was going to reply with some details about how to find out whether
you're using timezone data files from PostgreSQL or macOS, and how to
make sure they're up-to-date in both cases, but now I see that my
computer (which has up-to-date tzdata) agrees with yours, and
Wikipedia claims that DST was cancelled this year by presidential
decree.  Why do you think it's wrong?

https://en.wikipedia.org/wiki/Time_in_Brazil

Thomas Munro <thomas.munro@gmail.com> writes:
> On Mon, Nov 18, 2019 at 6:38 PM PG Bug reporting form
> <noreply@postgresql.org> wrote:
>> From Nov 3, America/Sao_Paulo should be offset only 2 hours from UTC due to
>> DTS. I would expect the timestamp to be December 1st midnight.

> I was going to reply with some details about how to find out whether
> you're using timezone data files from PostgreSQL or macOS, and how to
> make sure they're up-to-date in both cases, but now I see that my
> computer (which has up-to-date tzdata) agrees with yours, and
> Wikipedia claims that DST was cancelled this year by presidential
> decree.  Why do you think it's wrong?

Indeed, the IANA timezone folks changed their entry in tzdata 2019b,
on the strength of this:

https://mm.icann.org/pipermail/tz/2019-April/027848.html

and other nearby threads in that list archive.  2019b was released
2019-07-02, so any machine with reasonably up-to-date data should
know about this.  In the case of Postgres, we absorbed that change
in the August minor releases.

            regards, tom lane