Date: Fri, 20 May 2005 09:46:20 +0200 From: Stijn Hoop <stijn@win.tue.nl> To: Ade Lovett <ade@freebsd.org> Cc: Kris Kennaway <kris@obsecurity.org> Subject: Re: Disabling dependency on esound in ports builds Message-ID: <20050520074620.GB95023@pcwin002.win.tue.nl> In-Reply-To: <CFC68145-1928-4984-BCB5-6019CB893C89@FreeBSD.org> References: <20050518234933.05e2584b@dolphin.local.net> <20050519183922.GB6978@xor.obsecurity.org> <CFC68145-1928-4984-BCB5-6019CB893C89@FreeBSD.org>
next in thread | previous in thread | raw e-mail | index | archive | help
--IJpNTDwzlM2Ie8A6
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable
On Fri, May 20, 2005 at 12:06:05AM -0700, Ade Lovett wrote:
> > WITHOUT_ESOUND=3Dyes
>=20
> In this particular case, that knob does indeed work most of the time.
>=20
> Explaining to software authors in general to not provide configure =20
> (or other) scripts that bring in extra functionality based on =20
> previously installed software, without providing for a specific flag =20
> that should be turned on (thus allowing a proper dependency to be =20
> registered by *any* package management system).. well, that's left =20
> as an exercise to the interested reader.
But in ports one could add a patch to NOT do this. Granted, that's not idea=
l,
but at least it's controllable.
> This is one of the reasons that I would consider code along the lines =20
> of:
>=20
> .if exists(somelibrary) || defined(WITH_something)
> foo_DEPENDS+=3D ...
> .endif
>=20
> to be fundamentally flawed and in violation of POLA, where, if port A =20
> has such an (exists) dependency on port B, the following sequence:
>=20
> install A
> install B
> portupgrade A
>=20
> results in the second (A) being subtly different from the first, in =20
> that it is using extra functionality from B, without explicitly =20
> asking the user whether that's OK. The fact that the dependencies =20
> are actually correctly registered is something of a mitigating =20
> factor, but is certainly not an excuse for such "dependency-creep".
I would add that in the few cases where I was bitten by this, it's
exceptionally frustrating to find out WHY behaviour changed (or even
package names, in case of -esound etc).
I do know the reason why this kind of code was added: it was
considered incorrect behaviour that port A would NOT automatically
pick up support for library B if it was installed by the user.
For the reasons Ade outlines above, and considering reproducible
package building, it would seem nice to find a solution that turns
these kinds of checks off automatically, perhaps behind a flag
(NO_AUTO_DEPS=3Dyes).
A quick grep through the tree also shows this construct:
%%%
=2Eif exists(${LOCALBASE}/lib/libmad.so.2)
WITH_MAD=3D YES
=2Eendif
=2Eif defined(WITH_MAD)
=2E..
=2Eendif
%%%
which of course amounts to the same thing, except that adding NO_AUTO_DEPS
might be easier.
--Stijn
--=20
"An adult is a child who has more ethics and morals, that's all."
-- Shigeru Miyamoto
--IJpNTDwzlM2Ie8A6
Content-Type: application/pgp-signature
Content-Disposition: inline
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (FreeBSD)
iD8DBQFCjZXMY3r/tLQmfWcRAldbAKCa4r0ttE9at+aUeWI2Rpzme4DtXgCfXoC9
A2dBI0TqKF42zldVQmQMRKk=
=ZKQq
-----END PGP SIGNATURE-----
--IJpNTDwzlM2Ie8A6--
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20050520074620.GB95023>