Re: Time zone issue with 10.3?

On 3/2/2018 10:58, Tom Lane wrote:
> Karl Denninger <karl(at)denninger(dot)net> writes:
>> I have just tried r0lling forward to 10.3 from 10.1 and suddenly my time
>> zone declaration blew up.
>> I was using 'US/Chicago' and was able to get the server to start with
>> CST6CDT, which is the way the system is set up.  But I'm concerned about
>> this, because I now get this from trying to query the system tables for
>> timezone names:
>> pgsql=# select * from pg_timezone_names;
>>            name           | abbrev | utc_offset | is_dst
>> --------------------------+--------+------------+--------
>>  Canada/East-Saskatchewan | CST    | -06:00:00  | f
>> (1 row)
>> pgsql=# select * from pg_timezone_abbrevs;
>> ERROR:  time zone "Asia/Yerevan" not recognized
>> DETAIL:  This time zone name appears in the configuration file for time
>> zone abbreviation "amst".
> Hmm, is your build configured with --with-system-tzdata
> (pg_config --configure would tell you)? If so, is that pointing at
> a reasonable-looking tree of TZ data? It seems you don't have the
> TZ data correctly installed in the 10.3 installation, but it's hard
> to tell who to blame from this much info.
>
> Typically the top of the TZ file tree should look like
>
> $ ls /usr/share/zoneinfo
> Africa/ Canada/ GB Indian/ Mexico/ ROK iso3166.tab
> America/ Chile/ GB-Eire Iran NZ Singapore leapseconds
> Antarctica/ Cuba GMT Israel NZ-CHAT Turkey posix/
> Arctic/ EET GMT+0 Jamaica Navajo UCT posixrules
> Asia/ EST GMT-0 Japan PRC US/ right/
> Atlantic/ EST5EDT GMT0 Kwajalein PST8PDT UTC tzdata.zi
> Australia/ Egypt Greenwich Libya Pacific/ Universal zone.tab
> Brazil/ Eire HST MET Poland W-SU zone1970.tab
> CET Etc/ Hongkong MST Portugal WET
> CST6CDT Europe/ Iceland MST7MDT ROC Zulu
>
> If you're not trying to rely on a platform-provided installation
> of tzdata, then this ought to be under the share/ subdirectory
> of the PG installation.
>
> regards, tom lane
I found it; there was a permission problem during the install and
Postgres' startup, when reading those files, apparently didn't complain
about it -- it just didn't load them all (but did load others!) and that
resulted in an inconsistency that made things very angry..... :-)

Sorry about the false alarm....

--
Karl Denninger
karl(at)denninger(dot)net <mailto:karl(at)denninger(dot)net>
/The Market Ticker/
/[S/MIME encrypted email preferred]/