Re: TM format can mix encodings in to_char()

On 4/21/19 10:21 AM, Tom Lane wrote:
> Andrew Dunstan <andrew(dot)dunstan(at)2ndquadrant(dot)com> writes:
>> On 4/21/19 12:28 AM, Tom Lane wrote:
>>> I don't have any way to test this on Windows, so could somebody
>>> do that? Manually running the Turkish test cases ought to be enough.
>> How does one do that? Just set a Turkish locale?
> Try variants of the original test case. For instance, in a UTF8
> database,
>
> regression=# show server_encoding ;
> server_encoding
> -----------------
> UTF8
> (1 row)
>
> regression=# SET lc_time TO 'tr_TR.iso88599';
> SET
> regression=# SELECT to_char(date '2010-02-01', 'DD TMMON YYYY');
> to_char
> --------------
> 01 ŞUB 2010
> (1 row)
>
> Unpatched, I get an error about invalid data. Now, this is in
> a Linux machine, and you'll have to adapt it for Windows --- at
> least change the LC_TIME setting. But the idea is to get out some
> non-ASCII strings from an LC_TIME setting that names an encoding
> different from the database's.
>
> (I suspect you'll find that the existing code works fine on
> Windows, it's only the first version(s) of this patch that fail.)
>
>

Test above works as expected with the patch, see attached.  This is from
jacana.

LMK if you want more tests run before I blow the test instance away

cheers

andrew

--
Andrew Dunstan https://www.2ndQuadrant.com
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services