Subject: Re: CVS commit: pkgsrc/doc
To: Dieter Baron <email@example.com>
From: Greg Troxel <firstname.lastname@example.org>
Date: 08/28/2007 08:17:32
Dieter Baron <email@example.com> writes:
> In article <20070827231822.BCFD721507@cvs.netbsd.org> Greg wrote:
> + o Figure out a way to have "pkg_install -u", probably with a
> + flag, preserve the value of the automatic tag. We do not
> + have consensus on the right behavior.
> If the -A flag is given, preserve automatic, otherwise clear it. If
> any dependencies are updated, preserve the automatic flag. make
> replace would have to learn to pass -A when updating dependencies.
Currently -A is documented to set (rather than preserve) the automatic
flag. So with -u it will preserve, and there would be no way to set.
It feels a bit awkward to have such a subtle behavior change.
Certainly things like make replace and pkg_rolling-replace can pass the
right flags, whatever they turn out to be.
How about a -U which is like -u, but also preserves automatic? I
suppose that's "what gdt thinks -u ought to do", but it's also "what
tools doing automatic updates want".
> That way, automaticity is preserved for dependencies, and is cleared
> for the package I actively update. I think this is following the
> principle of least surprise, given how pkg_add without -u works.
I realize that it follows the principle of least surprise for you, but
we're having the discussion because for me that principle says that the
automatic flag should not be touched. I use 'pkg_add -u' not because a
package I want isn't there, but because a package is out of date, as
part of a general update-old-packages maintenance activity.
Do you use 'pkg_add -u'? When you do, are you updating outdated
packages because they are outdated, or do you view that as saying "I
want this package"? Sorry if I'm just being dense, but I don't see how
someone doing 'pkg-add -u' would be doing other than maintenance.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.7 (NetBSD)
-----END PGP SIGNATURE-----