At Thu, 17 Nov 2016 10:42:54 +0800, Craig Ringer <craig@2ndquadrant.com> wrote in
<CAMsr+YG_S+kAvcSoaOKfVeFgG2An_D+NSap7H3HCV+hTRduPWQ@mail.gmail.com>
> On 17 November 2016 at 10:31, Kyotaro HORIGUCHI
> <horiguchi.kyotaro@lab.ntt.co.jp> wrote:
>
> >
> > I vote +1 to upgrading perl, but I'm not sure if all supporting
> > platforms other than linux support the version of perl.
> >
> > Anyway ./configure is saying as the following.
> >
> > | *** The installed version of Perl, $PERL, is too old to use with PostgreSQL.
> > | *** Perl version 5.8 or later is required, but this is $pgac_perl_version." >&5
> >
> > If TAP test requires 5.8.8, the whole build sequence should
> > follow that. Or at least ./configure --enable-tap-tests should
> > check that.
>
> I wrote 5.8.8 because that's what we've always discussed before and I
> could honestly not imagine it mattering whether we require 10-year or
> 15-year old Perl. Especially for the TAP tests, which are new, and
> optional.
>
> I really don't think it matters if the TAP tests use a slightly newer
> Perl. They're optional and not enabled by default. Can we just
If so, why explicitly require 5.8.8? I think it is because the
'slight' difference actually prevent the test from running.
> document this please? I didn't think a four-line docs patch to an
> optional, non-default, new test suite would require this kind of
> discussion.
My only concern is the fact that 'make check-world' runs the TAP
tests as a part if --enable-tap-tests (according to
release-9.4.sgml, but I haven't done by myself.). I completely
agree to you if it didn't run as a part of top-level 'make <some
kind of check>'.
> But sure, if it's easier, we can have 5.8.0 in the README. What's five
> extra years matter anyway? Hey, while we're at it, lets change Pg to
> build on Borland C and require K&R style!
It's seems an extreme story. And I *believe* anywhere written
that Pg requires some version of C standard.
regards,
--
Kyotaro Horiguchi
NTT Open Source Software Center