On Mon, Jul 06, 2015 at 11:53:58PM -0700, Michael Paquier wrote:
> On Mon, Jul 6, 2015 at 10:04 PM, Noah Misch wrote:
> >> Also, if we would
> >> want to have a real installcheck mode, what we should use is not the
> >> path to what has been built but the path to the installation that the
> >> Postgres instance needed is using. Now if you want to fix it if you
> >> fix that's incorrect I won't complain about that :)
> >
> > I don't wish to. I was content with released-branch vcregress.pl semantics.
>
> Attached is a patch to revert to the pre-9.4 way of doing...
This is a start. Study "git diff REL9_4_STABLE src/tools/msvc/vcregress.pl"
after applying it. To give just one example of material left to revert, it
leaves two absolute-path "chdir" calls in a row.
> If MSVC
> should imitate what make does, perhaps you have an opinion about the
> ideas I sent upthead to add a check-mode to contribcheck and
> modulescheck? That's 9.6 work.
As you note, we lack a predefined notion of the installation path like we have
in the GNU make build. That challenges our ability to make the two build
systems more similar than they are in 9.4. While I have occasionally wished
for a vcregress.pl target corresponding to "make -C contrib check", it's also
one more thing to bitrot in an area that already experiences too much bitrot.
Overall, I recommend against this project. The benefits of a correct patch
are slight, yet writing one is difficult.