svn commit: r243554 - in head/usr.sbin/pkg_install: add create delete info lib updating version

[Stanislav Sedov]

On Wed, 28 Nov 2012 21:03:58 -0500
Eitan Adler <eadler at freebsd.org> mentioned:

> On 28 November 2012 20:00, Stanislav Sedov <stas at freebsd.org> wrote:
> > On Mon, 26 Nov 2012 05:11:07 +0000 (UTC)
> > Eitan Adler <eadler at FreeBSD.org> mentioned:
> >
> >> Author: eadler
> >> Date: Mon Nov 26 05:11:07 2012
> >> New Revision: 243554
> >> URL: http://svnweb.freebsd.org/changeset/base/243554
> >>
> >> Log:
> >>   Provide an option to users to shoot themselves in the foot.
> >>
> >
> > This should probably be a default behavior.  It's not good when all these
> > warnings pop up everywhere just because local.sqlite file is present.
> 
> Making the foot shooting behavior default defeats the point.
> If local.sqlite exists it almost certainly means that you don't want
> to run the old pkg_ tools.
> For the few correct uses an opt-out option is provided.
> 

Well, it's not entirely true.  I did end up with local.sqlite because I
stepped on the landmine of portmgr-pkg becoming the default for some
reason.  So I did end up with half of my packages being in the pkgng
sqlite database, and half in standard /var/db/pkg.  I had to convert
them back to standard format by hand, but I still have the sqlite database
just in case.

I guess if you really want to prevent a foot-shooting, you should add
a message to pkgng sayng in all caps e.g. "You are running experimental
package manager and there's no migration plan from pkgng to old pkg
exists." and maybe ask for confirmation.  I don't really see how this
message being in pkg_ tools helps to prevent a possible foot-shooting
at all.

At very least, it makes sense to make it conditional on WITH_PKGNG, so this
code does not end up compiled in if PKGNG is disabled in src.conf.

-- 
Stanislav Sedov
ST4096-RIPE

()  ascii ribbon campaign - against html e-mail 
/\  www.asciiribbon.org   - against proprietary attachments