Re: [RFC] building postgres with meson - Mailing list pgsql-hackers

From Andres Freund
Subject Re: [RFC] building postgres with meson
Date
Msg-id 20211014172412.rovmzox7dsahclis@alap3.anarazel.de
Whole thread Raw
In response to Re: [RFC] building postgres with meson  (Alvaro Herrera <alvherre@alvh.no-ip.org>)
Responses Re: [RFC] building postgres with meson
Re: [RFC] building postgres with meson
List pgsql-hackers
Hi,

On 2021-10-14 10:29:42 -0300, Alvaro Herrera wrote:
> On 2021-Oct-14, Josef Šimánek wrote:
>
> > I'm using Fedora 34 and I still see perl-Opcode.x86_64 as a separate
> > package. Anyway it behaves differently with autoconf tools and the
> > meson build system. Is perl disabled by default in the current build
> > system?

Hm, so it seems we should make the test separately verify that perl -M{Opcode,
ExtUtils::Embed, ExtUtils::ParseXS} doesn't fail, so that we can fail perl
detection with a useful message?


> Yes, you have to use --with-perl in order to get it.

With the meson prototype I set most optional features to "auto", except for
LLVM, as that increases compile times noticeably.

For configure we didn't/don't want to do much auto-detection, because that
makes life harder for distributors. But meson has one switch controlling all
features set to 'auto' and not explicitly enabled/disabled:
  --auto-features {enabled,disabled,auto}                               Override value of all 'auto' features (default:
auto).
so the argument doesn't apply to the same degree there. We could default
auto-features to something else too.

There were two other reasons:

1) I got tired of needing to disable zlib, readline to be able to build on
   windows.
2) Exercising all the dependency detection / checking seems important at this
   stage

Greetings,

Andres Freund



pgsql-hackers by date:

Previous
From: Jeevan Ladhe
Date:
Subject: Re: refactoring basebackup.c
Next
From: Andres Freund
Date:
Subject: Re: [RFC] building postgres with meson