Postgres Pro
Downloads
Home   >   mailing lists

Thread: pg_regress multibyte setting

pg_regress multibyte setting

From
Peter Eisentraut
Date:
Is it a good idea that we run make check with MULTIBYTE = SQL_ASCII by
default?  We run it with the user's locale by default, so shouldn't we
use the encoding that belongs to the locale by default?  Otherwise we
are testing a fairly unrepresentative environment.  If you really want
to test SQL_ASCII you could of course choose it explicitly or set the
locale to C.

Re: pg_regress multibyte setting

From
Itagaki Takahiro
Date:
On Wed, Jan 12, 2011 at 08:06, Peter Eisentraut <peter_e@gmx.net> wrote:
> Is it a good idea that we run make check with MULTIBYTE = SQL_ASCII by
> default?  We run it with the user's locale by default, so shouldn't we
> use the encoding that belongs to the locale by default?  Otherwise we
> are testing a fairly unrepresentative environment.  If you really want
> to test SQL_ASCII you could of course choose it explicitly or set the
> locale to C.

It seems good to run make check successfully on many platforms,
but we might miss locale-dependent bugs.

Personally speaking, I often recommend to use UTF-8 + C locale combinations
for users, but I'm not sure it's the most common use-cases or not.

--
Itagaki Takahiro

Re: pg_regress multibyte setting

From
Andrew Dunstan
Date:

On 01/12/2011 09:52 PM, Itagaki Takahiro wrote:
> On Wed, Jan 12, 2011 at 08:06, Peter Eisentraut<peter_e@gmx.net>  wrote:
>> Is it a good idea that we run make check with MULTIBYTE = SQL_ASCII by
>> default?  We run it with the user's locale by default, so shouldn't we
>> use the encoding that belongs to the locale by default?  Otherwise we
>> are testing a fairly unrepresentative environment.  If you really want
>> to test SQL_ASCII you could of course choose it explicitly or set the
>> locale to C.
> It seems good to run make check successfully on many platforms,
> but we might miss locale-dependent bugs.
>
> Personally speaking, I often recommend to use UTF-8 + C locale combinations
> for users, but I'm not sure it's the most common use-cases or not.
>


We have had support for multi-byte testing in the buildfarm for about 2 
years. My own Linux buildfarm member is configured to test in both 
C/SQL_ASCII and en_US.utf8.

cheers

andrew