Re: Should contrib modules install .h files?

>>>>> "Andres" == Andres Freund <andres(at)anarazel(dot)de> writes:

Tom> There's also a question of whether we need to change anything in
Tom> contrib/ so that it plays by whatever rules we set. There's an
Tom> expectation that contrib modules should be buildable with PGXS,
Tom> so they need to follow the rules.

Andrew> ... that at least all of the *_plperl transform modules in
Andrew> contrib/ fail to build with USE_PGXS already (i.e. for as long
Andrew> as they have ever existed), because they rely on
Andrew> plperl_helpers.h which is never installed anywhere, and trying
Andrew> to get it via $(top_srcdir) obviously can't work in PGXS.

Andrew> Haven't tried the python ones yet.

>> And none of the plpython transforms can even parse their makefiles
>> with USE_PGXS, let alone build, because they have an "include"
>> directive pointing into src/pl/plpython.

Andres> FWIW, I'd be perfectly on board with just burying this policy.
Andres> Designating one contrib module (or something in
Andres> src/test/examples or such) as a PGXS example, and always
Andres> building it with pgxs seems like it'd do a much better job than
Andres> the current policy.

I suggest that modules that actually _can't_ build with PGXS should have
the PGXS logic removed from their makefiles (perhaps replaced with an
error). But for the rest, it's a convenience to be able to build single
modules using USE_PGXS without having to configure the whole source tree.

--
Andrew (irc:RhodiumToad)