Re: Regression tests versus the buildfarm environment

On 08/11/2010 04:47 PM, Tom Lane wrote:
> Andrew Dunstan<andrew@dunslane.net>  writes:
>> Another way would be to have pg_regress honour an environment var like
>> PG_REGRESS_PORT, which the buildfarm script could use.
> Yeah, that would work too.  (Is it portable to Windows, though?)

Should be

> I prefer the change-the-default approach mainly because it wouldn't
> require any documentation, whereas it'd be a bit hard to argue that
> environment variables etc shouldn't be documented ...
>
>             

Yeah. The other advantage is that we would not need to wait until we had 
got everyone to update their versions of the buildfarm code. So I agree 
this is the nicest solution.

cheers

andrew