From: Tom Lane <tgl@sss.pgh.pa.us>
To: Alvaro Herrera <alvherre@2ndquadrant.com>
cc: Robert Haas <robertmhaas@gmail.com>, John Naylor <jcnaylor@gmail.com>,
        Stas Kelvich <s.kelvich@postgrespro.ru>,
        Andreas Karlsson <andreas@proxel.se>,
        Andrew Dunstan <andrew.dunstan@2ndquadrant.com>,
        PostgreSQL Hackers <pgsql-hackers@postgresql.org>,
        David Fetter <david@fetter.org>
Subject: Re: unused_oids script is broken with bsd sed
In-reply-to: <20180503195642.otwsjbjdtzdemumf@alvherre.pgsql>
References: <20180503195642.otwsjbjdtzdemumf@alvherre.pgsql>
Comments: In-reply-to Alvaro Herrera <alvherre@2ndquadrant.com>
	message dated "Thu, 03 May 2018 16:56:42 -0300"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <18603.1525378285.1@sss.pgh.pa.us>
Date: Thu, 03 May 2018 16:11:25 -0400
Message-ID: <18604.1525378285@sss.pgh.pa.us>
Precedence: bulk

Alvaro Herrera <alvherre@2ndquadrant.com> writes:
> Tom Lane wrote:
>> However, RenameTempFile is also used by Gen_fmgrtab.pl, and having the
>> same sort of no-touch semantics for fmgroids.h and friends would have some
>> additional fallout.  The makefiles would think they have to keep
>> re-running Gen_fmgrtab.pl if fmgroids.h is older than the mod time on any
>> input file, and that's certainly no good.  We can fix that by inventing a
>> stamp file for the Gen_fmgrtab.pl run, analogous to bki-stamp for the
>> genbki.pl run, but that means changes in the makefiles that go a bit
>> beyond the realm of triviality.

> Sounds OK to me -- a stamp file is already established technique, so it
> shouldn't go *too much* beyond triviality.

Yeah, what I'm envisioning is to change the makefile rules around these
files to look as much as possible like the ones around the BKI files,
which are (we hope) already debugged.  So it doesn't seem like a high
risk change ... at least so far as the makefiles are concerned.

> I do note that
> msvc/Solution.pm runs Gen_fmgrtab.pl, but it shouldn't require any
> changes anyhow.

Hmm.  Actually, given the IsNewer checks there, it looks like Solution.pm
is basically hand-rolling makefile-like dependency checking, which means
it would be fooled by no-touch updates in the same way as make is, causing
rebuilds to do unnecessary work.  We could live with that for awhile
maybe, but ultimately Solution.pm would need to be fixed to use a stamp
file like the makefile logic.

I could take a stab at that, but I don't have any way to test it myself.

			regards, tom lane