Re: moving from contrib to bin

On 2015-01-18 21:05:29 +0900, Michael Paquier wrote:
> On Sat, Jan 17, 2015 at 10:08 PM, Andres Freund <andres@2ndquadrant.com> wrote:
> > Observations:
> > 1) Are we sure it's a good idea to rely on pgxs.mk in src/bin programs?

> Yeah, this seems like a bad dependency, PGXS being made for contrib
> modules... So corrected in the patch attached (the headers of the
> Makefiles are improved as well to be consistent with the other
> utilities, btw there is code duplication in each Makefile if we do not
> use PGXS stuff in src/bin).

Yes, there's a fair amount of duplication. I do think there's a good
case for making the Makefiles less redundant, but we should do that in
all src/bin binaries, and in a separate patch.

> > 4) I have doubts that it's ok to integrate the tests in src/bin just the
> >    way they were done in contrib.

> Are you referring to the tests of pg_upgrade?

Yes.

> > 5) Doing the msvc support for all intermediate commits in a separate
> >    commit strikes me as a bad idea. Essentially that makes the split
> >    pretty pointless.
> > 6) Similarly I'd much rather see the doc movement in the same commit as
> >    the actually moved utility. Then we can start applying this one by one
> >    on whatever we have agreement.

> Well, sure. The split was done just to facilitate review with stuff to
> be applied directly on top of what Peter already did. And note that I
> agree as well that everything should be done in a single commit.

I don't think all of the moves should be done in one commit. I'd much
rather do e.g. all the pg_upgrade stuff in one, and the rest in another.

> Separating things would break build on a platform or another if a
> build is done based on an intermediate state of this work, that would
> not be nice.

Yes, that's why commits in a series should be standalone. I.e. all
should work on all platforms. Possibly individual ones don't add
features, but they shouldn't break things.

> > 8) Why did you remove Peter as the git author?

> I applied on my local repo this series of patches after some bash-ing
> without preserving any meta data, so the author name has just been
> changed on the way, and then ran a simple git format to generate the
> whole set once again. Well, sorry if this was confusing, but let's be
> clear anyway: I have no intention to make mine the work of Peter (or
> any other people).

You know that you can apply patch series like Peter's (or from your tar
archive) using 'git am'? Which preserves the author, message and such?


Greetings,

Andres Freund

-- Andres Freund                       http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training &
Services