Re: Should contrib modules install .h files?

On 2018-07-23 17:22:16 +0200, Peter Eisentraut wrote:
> On 23.07.18 06:15, Tom Lane wrote:
> > Michael Paquier <michael(at)paquier(dot)xyz> writes:
> >> On Sun, Jul 22, 2018 at 09:42:08PM -0400, Stephen Frost wrote:
> >>> So, +1 from me for having a directory for each extension.
> >
> >> So, like Stephen, that's a +1 from me.
> >
> > Same here. One-file-per-extension is too strongly biased to tiny
> > extensions (like most of our contrib examples).
>
> Nobody said anything about one-file-per-extension. You can of course
> have hstore_this.h and hstore_that.h or if you want to have many, use
> postgis/this.h and postgis/that.h. That's how every C package in the
> world works. We don't need to legislate further here other than, use
> sensible naming.

PGXS provides a certain way of doing things. It's far from insane for us
to procscribe that we go for $extension/. I fail to see any sort of
advantage of using $extension_$header.h. It only serves to create
additional work.

> Also, let's recall that the point of this exercise is that you want to
> install the header files so that you can build things (another
> extension) that somehow interacts with those extensions. Then, even if
> you put things in separate directories per extension, you still need to
> make sure that all the installed header files don't clash, since you'll
> be adding the -I options of several of them.

I'd suggest just using the name with the $extension/ prefix going
forward.

> > I don't have a real strong opinion on whether it's too late to
> > push this into v11. I do not think it'd break anything other than
> > packagers' lists of files to be installed ... but it does seem
> > like a new feature, and we're past feature freeze.
>
> Certainly a new feature. I suggest submitting it to the next commit fest.

Given that there seems to be fairly widespread agreement, anyone that
commented but you, and the patch has been reviewed, I see zero reason to
wait for the next CF for a committer's patch.

Greetings,

Andres Freund