Skip site navigation (1)Skip section navigation (2)
Date:      Sat, 21 May 2016 18:22:11 +0000
From:      Alexey Dokuchaev <danfe@FreeBSD.org>
To:        Ian Lepore <ian@freebsd.org>
Cc:        Baptiste Daroussin <bapt@freebsd.org>, marino@freebsd.org, Ed Maste <emaste@freebsd.org>, ports-committers@freebsd.org, svn-ports-all@freebsd.org, svn-ports-head@freebsd.org
Subject:   Re: svn commit: r415078 - in head: . Mk
Message-ID:  <20160521182211.GA96389@FreeBSD.org>
In-Reply-To: <1463850648.1180.374.camel@freebsd.org>
References:  <CAPyFy2A9L1cCikOrgBAWUo0GTCLJ4EgzqukhobaJp%2BZqv7_SpQ@mail.gmail.com> <20160519122306.GA24015@FreeBSD.org> <20160521112728.GA624@FreeBSD.org> <364d3d9f-63ff-18c8-c730-a11c57dc0673@marino.st> <20160521114358.GC624@FreeBSD.org> <20160521122522.GJ21899@ivaldir.etoilebsd.net> <70938d6b-0fab-91b9-28b0-9dd05302a503@marino.st> <20160521124148.GK21899@ivaldir.etoilebsd.net> <20160521163832.GB97771@FreeBSD.org> <1463850648.1180.374.camel@freebsd.org>

next in thread | previous in thread | raw e-mail | index | archive | help
On Sat, May 21, 2016 at 11:10:48AM -0600, Ian Lepore wrote:
> This is just crazy-talk.  In what way is an update timestamp NOT "info"
> about the distribution files?  The file isn't named distchecksum or

In a way I've outlined earlier: distinfo is a set of per-distfile values
of certain functions.  TIMESTAMP is not, it's not a function of any of
the distfiles and it is not obvious exactly how it is related to those
distfiles.  I've explained it earlier today, you should've replied to
*that* email [1] if you object.

> dist_sha256_only, or dist_only_what_alexey_likes_to_see.  It is
> metadata, it has always been metadata, and now that there is new
> metadata necessary to achieve new functionality, the one and only place
> that makes sense for it to be is the existing distinfo metadata file.

The difference between SIZE(), SHA256(), and TIMESTAMP is that the first
two are required to be there for reference, something to compare against.
TIMESTAMP is just some frozen timestamp, and could be obtained elsewhere
without breaking the whole idea (as I see it right now).  Polluting data
with not-being-a-reference metadata is just wrong from engineering point.

> Saying that the info can be obtained from version control is more crazy
> -talk.  Do we add support for every version control system that ever
> existed to support every user of ports other there?

Well, first, we do officially use SVN.  Second, for a versioned tree,
getting correct timestamp is just a few lines for another popular VCS,
as we speak of versioned tree.  Third, I doubt that *every* ports user
would want to know or care for reproducible builds; most just want to
change some options and "make install clean".

> The people most interested in the new reproducible build functionality
> are the ones most likely to be using some local version control system
> which is not svn or git (we use both cvs and mercurial for various
> generations of our ports trees at $work).

Yet when I asked why it can't be fetched from the VCS, I was told that
we care about unversioned tarballs, and that's the problem.

> Instead of demanding that the people actually doing useful new work
> justify this tiny insignificant detail of their implementation because
> it offends your view of how things should be, perhaps you could provide
> some argument about what harm the new value does.  Something based on
> actual facts, not just "I think it's ugly" or "I think it's wrong."

But I did [1].  And I'm not demanding anything, just think that distinfo
is not the best place to put TIMESTAMP in.  (And to a lesser degree, is
that it should be possible to avoid hardcoding it anywhere, but that's an
open debate still.)  Adhering to some "harm" here looks like a straw man
argument, sorry.

./danfe

[1] http://docs.freebsd.org/cgi/getmsg.cgi?fetch=2769892+0+current/svn-ports-head



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20160521182211.GA96389>