Date: Sat, 21 Mar 2015 09:20:59 -0600 From: Adam Weinberger <adamw@adamw.org> To: Alexey Dokuchaev <danfe@FreeBSD.org> Cc: svn-ports-head@freebsd.org, Baptiste Daroussin <bapt@FreeBSD.org>, ports-committers@freebsd.org, svn-ports-all@freebsd.org, Bryan Drewery <bdrewery@FreeBSD.org> Subject: Re: svn commit: r381760 - in head/x11-fonts/sourcesanspro-ttf: . files Message-ID: <5F4F2275-6F2D-43CC-ACB4-1B8240D5A5B3@adamw.org> In-Reply-To: <20150321150350.GB55163@FreeBSD.org> References: <201503201823.t2KIN32I080448@svn.freebsd.org> <550C6655.5010802@FreeBSD.org> <20150320183524.GD87678@ivaldir.etoilebsd.net> <9BE33FCA-5C2F-4FEA-9B3A-5D9DB6632635@adamw.org> <20150321150350.GB55163@FreeBSD.org>
next in thread | previous in thread | raw e-mail | index | archive | help
> On 21 Mar, 2015, at 9:03, Alexey Dokuchaev <danfe@FreeBSD.org> wrote:
>=20
> On Fri, Mar 20, 2015 at 12:43:17PM -0600, Adam Weinberger wrote:
>> It would be so unbelievably nice if we had a way to just specify the
>> destination filename, using fetch -o. All the bizarre names that =
files
>> are published as, the unversioned tarballs, the documentation files
>> labelled Documentation.pdf, and the fallout from changes like =
this....
>> we could make our lives so much easier with a DISTFILE_DEST variable.
>=20
> Current situation with the distfiles in the industry is indeed very =
bad.
> Old-skool folks that obeyed the rule of "foo-0.42.tar.gz" distfile =
which
> extracts into "foo-0.42" directory, fetchable as regular file with =
known
> Content-Length and Last-Modified headers (to allow working progress =
bars
> and sane mirroring) are minority now, and soon may cease to exist.
>=20
> New-age hipsters tell you to "always git clone, no thinking required" =
and
> rarely care to upload ./configure-ready tarball, or at least tag their
> releases so GH can try to provide reliably fetchable and mirrorable =
files.
>=20
> As to Adam's idea, I don't see what can be done, apart from tell =
upstream
> how important it is to produce normal old-fashioned releases from time =
to
> time.
>=20
> DISTFILE_DEST, while looking good at the first glance, raises at least
> some important questions of its own: how do we mirror these files? =
What
> should be their mtime, so mirrors won't have to refetch the same bits =
all
> over again? Do we need/want to maintain relationship between upstream
> and our DISTFILE_DEST'ied name, and how do we do it if we do?
>=20
> ./danfe
>=20
> P.S. No wonder sf.net is still alive: they know how to maintain sane
> CDN for tarballs and data, which backs up all their drawbacks.
Those are really good points. I guess if distfiles were stored on =
MASTER_SITE_BACKUP in a /${UNIQUENAME}/ subdir they could continue to be =
named what they're originally named. That way the relationship is that =
things on MASTER_SITE* are the original name, and it's only on the =
client machine that the distfiles are renamed.
We may be able to do a little more to work around the fact that now the =
most common distfile producer (GitHub) names directories strangely... =
something like
_WRKSRC_POSSIBILITIES !=3D ls ${WRKDIR} | ${GREP} -ci ${PORTNAME}
.if _WRKSRC_POSSIBILITIES =3D=3D 1
WRKSRC !=3D ls ${WRKDIR} | ${GREP} -ci ${PORTNAME}
.endif
But it looks like that's no longer an issue with our new USE_GITHUB.
# Adam
--=20
Adam Weinberger
adamw@adamw.org
http://www.adamw.org
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?5F4F2275-6F2D-43CC-ACB4-1B8240D5A5B3>