Magnus Hagander wrote:
> On Mon, Mar 12, 2007 at 08:20:53PM -0500, Andrew Dunstan wrote:
>
>> Gregory Stark wrote:
>>
>>> "Tom Lane" <tgl@sss.pgh.pa.us> writes:
>>>
>>>
>>>> petere@postgresql.org (Peter Eisentraut) writes:
>>>>
>>>>> Make configuration parameters fall back to their default values when
>>>>> they
>>>>> are removed from the configuration file.
>>>>>
>>>> It appears that this patch has broken custom GUC variables; at the very
>>>> least it's broken plperl.
>>>>
>>> Huh, it occurs to me that I haven't seen any plperl regression tests fly
>>> by
>>> when I've been running regression tests myself. What do I have to do to
>>> test
>>> if plperl, plpython, etc work with the packed varlena patch?
>>>
>>>
>> cd src/pl; make installcheck
>>
>
> Is there any particular reason why we don't run these as part of a
> general "make check"?
>
> //Magnus
>
>
Probably historical more than anything else.
The core tests all run regardless of configuration, though, and the PL
tests use a different database (by design). When we standardised this we
did just enough to enable the buildfarm clients to test PLs sanely. If
you think we need more, have a go at it.
I should perhaps point out that the buildfarm client can be used to do a
comprehensive build and test on your sources, including all the
configured PLs, ECPG and the contrib tests, using either the
--from-source or --from-source-clean flags. These were originally
designed to help diagnose and fix problems disclosed during normal
buildfarm runs, but I have found it quite useful when working on
substantial projects. You don't need to be registered as a buildfarm
member to use the client program, in these modes - no results are
uploaded to the server when these flags are used.
cheers
andrew