flynn at energyhq.es.eu.org
Sun Oct 24 23:37:55 PDT 2004
On Sun, 24 Oct 2004 16:16:43 -0700
"Loren M. Lang" <lorenl at alzatex.com> wrote:
Kris has already pointed out that FreeBSD's ports system works in a
different way, but I'll add my $0.02 anyway...
> I'm curious if there is any major work being done in developing the
> ports infrastructure right now or if it is mostly just minor features
The ports system is being enhanced all the time. They don't push radical
changes into the system but small steps, and I think that's the way it
should be, as it leaves more room for testing.
> biggest thing is the USE flags. One of the biggest annoyances I have
> with major upgrades using portupgrade is when I leave it upgrading for
> a day and come home to find it sitting at the mplayer-skins screen
> asking what skins I want to install and that it spent less then an
> hour doing any work upgrading. I really don't care much about skins,
I'd say portage's USE is, at the same time, its best feature and its
biggest weakness. A lot of people end up with broken programs due to
exotic USE settings that no one else has tested. Others change USE
options and then forget to run emerge in deep mode, ending up with a mix
of programs and libraries, some of them depending on e.g. kde or gnome
libraries and other that don't.
As someone who has used both systems, I have to say that FreeBSD's fine
grained OPTIONS is much better than portage's USE. Sure, it will ask you
the first time (although you could just use BATCH if you want to use
defaults for everything), but it will be automated the next time you
update and/or rebuild the port.
portage is far from perfect. For example, I had "-gnome" in my USE
flags, and some dependencies of monodevelop were compiled that way.
After building them it complained in the next ebuild that libfoo had
been built without gnome and that I had to rebuild it again. (IIRC it
was gtkhtml or a similar lib).
> something. USE flags would also eliminate the need for ports like
This is needed if you want to provide binary packages with different
> In fact, I
> think gentoo has at least 5 of the latest ebuilds in the mozilla
> directory. This makes it easier to choice an older version if the
> latest has some bugs not worked out yet. Yesterday I just ran across
You can always cvs(up) to a given point in time. portage usually has
ebuilds that are still marked x86 only, or even hardmasked (e.g.
sylpheed-gtlk2). On FreeBSD, if a port is in the tree it will probably
work. My impression is that the testing part on FreeBSD is better, and
that portage relies more on 'commit and let people try it'.
> still had problems. At least on gentoo I would of had five choices.
You do have the choice in FreeBSD, cvs is your friend.
> The last feature I would like to see in ports is the ability to hold
> back certain ports or force them to always use packages. lang/ezm3
> and editors/openoffice-1.1 almost always fail compiling at some point
> with cc1 being killed for eating too much memory. I prefer to always
portupgrade has been able to do that for a long time. Have a look at
pkgtools.conf in /usr/local/etc.
> I might be able to volunteer some time to this as, at least the first
> thing I mentioned I think could really use some work.
Having said that, I like both systems and just wanted to point out that
neither of them are perfect, and each one has its little advantages over
the other. I don't see the point in portage-ing FreeBSD ports. I don
think that some ideas could be borrowed from OpenBSD's ports, like fake
installs and the very strict plist handling. I also like how they use
more than one digest, thus making a possible MD5 and SHA1 collision
Miguel Mendez <flynn at energyhq.es.eu.org>
PGP Key: 0xDC8514F1
Note: All HTML mail goes to /dev/null
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Size: 187 bytes
Desc: not available
Url : http://lists.freebsd.org/pipermail/freebsd-questions/attachments/20041025/a660d998/attachment.bin
More information about the freebsd-questions