Bruce Momjian <maillist@candle.pha.pa.us> writes:
>> Would this also be a relevant time to get a regression test to run on
>> a non-installed system?
> Gee, I never even though of that. What advantage would there be for
> that?
It has been suggested before, and I think it's a good idea. The
advantage is you can smoke-test a new compilation *before* you blow
away your existing installation ;-)
It is, of course, possible to do that by installing into a nonstandard
location/port and then running the regress tests there. But you have
to know exactly what you're doing to do that. If we're going to
overhaul install, we should make it easier to run the regress tests
that way, or even better with not-installed-at-all binaries from the
source tree.
Another thing I'd like to see would be full support for building in
a separate object-directory tree, leaving the source tree pristine
rather than filled with configure and build output files.
This is a standard GNU practice and I think it's a good one.
regards, tom lane