Re: Regression tests versus the buildfarm environment

Andrew Dunstan <andrew(at)dunslane(dot)net> writes:
> You original email said:
> For some historic reasons, I have my local scripts set up so that
> they build development instances using the hardcoded port 65432.

> I think my response would be "Don't do that".

Yeah ... or at least use a different port per branch. Or make use of
the ability to force pg_regress to use a nondefault port (which I still
say we need to make accessible through "make check", whether or not the
buildfarm does it that way).

> Having said that, maybe we could reasonably use something like
> DEF_PGPORT + 10 * major_version + minor_version in the calculation and
> advise buildfarm members with multiple animals to keep their port ranges
> say, 200 or more apart.

I think that just makes it more prone to failure. We should have the
buildfarm configuration such that any one run uses the same port number
for both installed and uninstalled regression tests. If Peter is dead
set on not changing pg_regress's default, then changing the makefiles to
enable use of the --port switch is the way to do that.

regards, tom lane