Обсуждение: [patch] fix for regression tests (locale cs_CZ)

I attached fix for regression tests and Czech locale. It is not complete
yet, because I fighting with foreign_data test. But it fix three other
tests.

The difference is that Czech collation sorts numbers after letters.


    thanks Zdenek

See:
http://www.pgbuildfarm.org/cgi-bin/show_log.pl?nm=gothic_moth&dt=2009-02-02%2017:30:13

On Monday 02 February 2009 21:42:36 Zdenek Kotala wrote:
> I attached fix for regression tests and Czech locale. It is not complete
> yet, because I fighting with foreign_data test. But it fix three other
> tests.

done

Zdenek Kotala <Zdenek.Kotala@Sun.COM> writes:
> I attached fix for regression tests and Czech locale. It is not complete
> yet, because I fighting with foreign_data test. But it fix three other
> tests.

As per buildfarm member gothic_moth, this didn't fix it.

> The difference is that Czech collation sorts numbers after letters.

I think trying to accommodate that is going to be a permanent source of
new regression test headaches.  I recommend we revert the patch and agree
that this is beyond the scope of locale differences that we are willing
to cater for in the regression tests.
        regards, tom lane

Tom Lane píše v pá 13. 02. 2009 v 16:54 -0500:
> Zdenek Kotala <Zdenek.Kotala@Sun.COM> writes:
> > I attached fix for regression tests and Czech locale. It is not complete
> > yet, because I fighting with foreign_data test. But it fix three other
> > tests.
> 
> As per buildfarm member gothic_moth, this didn't fix it.

It is correct. I fixed three of four problems. Foreign data test has
same problem with number ordering like three previous. However, I try to
fix it different way (casting to name datatype), but I hit some strange
behavior which needs detail examination.

> 
> > The difference is that Czech collation sorts numbers after letters.
> 
> I think trying to accommodate that is going to be a permanent source of
> new regression test headaches.  I recommend we revert the patch and agree
> that this is beyond the scope of locale differences that we are willing
> to cater for in the regression tests.

You can find different problem in different locales. I think there is no
choice. Support all or only C. The headaches will be reduced when
collation per column will be develop. 

I though add locale specific suffix instead of only number (e.g.
char_cs_CZ.out) could help to keep it under control. Or add subdirectory
for each locale.
    Zdenek