further fink fun: resolution

I think my fink installation is now unwedged. The problem stemmed from mixing up the stable and unstable trees, as far as I can tell.

The package tree is subdivided into local, stable, and unstable, and the (un)stable trees are further divided into main and crypto.

[/Users/paul]:: more /sw/etc/fink.conf
# Fink configuration, initially created by bootstrap.pl
Basepath: /sw
RootMethod: sudo
#Trees: local/main stable/main stable/crypto local/bootstrap
Trees: local/main unstable/main unstable/crypto stable/main stable/crypto local/
bootstrap
Mirror-ctan: ftp://tug.ctan.org/tex-archive/
Mirror-gimp: ftp://ftp.gimp.org/pub
Mirror-gnome: ftp://ftp.gnome.org/pub/GNOME/
Mirror-gnu: ftp://ftp.gnu.org/gnu
MirrorContinent: nam
MirrorCountry: nam-us
ProxyPassiveFTP: true
Verbose: true

So I removed xfree86-base, using the suggestion I got from a new user on the fink-users list.

I ran sudo dpkg –remove –force-depends xfree86-base and then switched from using the unstable trees to just the stable and local (see the commented out line above).

I then installed an older version of xfree86-base with dpkg. That made X work again. As a side note, at no time during this ordeal did any of the fink-managed applications fail to work: the only one that failed was fink itself.

Then a test of fink selfupdate-cvs and fink update-all. The xfree86 version I installed was upgraded:

The following 4 packages will be installed or updated:
arts docbook-dtd wget xfree86-base
The following additional package will be installed:
qt3


And here’s what xfree86 looks like now:

[/sw/lib/perl5/darwin/auto]# fink list xfree86
Information about 1401 packages read in 44 seconds.


system-xfree86 4.2-1 Placeholder package for manually installe…
i xfree86-base 4.2.0-11 XFree86 libraries, utilities, clients and…
xfree86-rootles 4.2.0-11 XFree86 libraries, utilities, clients and…
xfree86-server 4.2.0-2 XFree86 display server (stable release)

There may still be some problem in the database in that fink still thinks it has Storable.pm installed and I have deinstalled it as well as removing it from the filesystem.

Hmm, I just told fink to remove it and this time, it shows it as uninstalled. Good.

I’ll have to wait a couple of days and a couple of CVS runs to see if this fixes it. But it looks good so far.

It looks like some way of enforcing database integrity and sane config files could improve fink’s robustness. Curious why fink’s config data isn’t in an XML/plist format: that might lend itself to more rigorous parsing (since xml parsing libs are installable with fink itself). There were some errors in my fink.conf file, as was pointed out, and perhaps that introduced the error.

Making the corrections to the fink.conf file weren’t enough to fix this, nor was removing xfree86-base by itself. Losing the unstable trees and then reinstalling xfree86-base, even an older version, did the trick. At the point I could reinstate the unstable trees and the updates would work as expected.

if I could take it all back, would I?

Google Search: paulbeard@mac.com fink
Google Search: pkdb1@home.com fink

It seems strange to read how many positive things I had to say about the Fink project when I contrast it with my current opinion.

I’m still working on repairing my broken installation. Compiles take a long time on a 350 MHz machine, so progress is slow. I suppose I could save time and download the .deb files.

It would, of course, save a lot of time if anyone on the fink team knew how the package management software actually worked. And maybe someone does and they just weren’t checking mail this weekend. My fault for expecting anything on a weekend.

I would have been quite happy if someone — anyone — had said they didn’t know but they knew who might, or asked for information that was in some way relevant. But trotting out the same old gripes about where perl libraries get installed (answer: where perl can find them) is useless and annoying. Incidentally, I removed Storable.pm from the fink installation — retaining it in perl’s @INC path — and nothing broke. Fancy that.

Discovering that the manual page contradicts reality makes me wonder if anyone actually does a baseline install/regression test of new releases as they come out. I know the answer to that already.

At some point, my anger will subside but for now it seems to be simmering nicely.

in search of a reliable packaging system for Darwin/OS X

Well, it appears fink will need to be reinstalled if I ever want to update anything I have currently under its care.

That’s annoying, to say the least. To rebuild everything last week took 24 hours, even after I updated the stale .info file for tar.

I have today removed xfree86 with dpkg –remove –force-depends xfree86-base and then set about rebuilding it from scratch. It seemed like it was going to work, but something still not right.

I may take a look at fink’s Engine.pm, which is where the error message appears: maybe I can figure out what conditions trigger it.

There are some other contenders: perhaps the GNU-Darwin collection has matured enough to try again. Having access to the FreeBSD ports collection would be nice.

There’s also a move to port Gentoo Linux’s Portage system to Darwin/OS X. You can find out more about it here and here.

where fink doesn’t work

I have this error message:

Failed: Internal error: node for xfree86-base already exists

After a lot of wrangling and research, it looks like my fink installation is wedged in such a way as to prevent me doing further updates: I have packages installed for which newer versions exist but they are never updated when I update from CVS.

And no one on the fink team can identify the problem or suggest solutions: “do it again” seems the most common refrain, followed closely “it works for me, it’s your problem.”

I have lobbied for fink to become the default packaging system for OS X going forward, but perhaps it’s not as fully-baked as I thought it was.

This is the only page Google can find on this topic: I feel so special.
Continue reading “where fink doesn’t work”

friend of a friend web

This seems interesting . . . . .

FOAFBot: IRC Community Support Agent

A FOAFBot instance is running on the #foaf channel on the OpenProjects IRC network, irc://irc.openprojects.net/foaf

The bot can be interrogated with simple questions about the properties of community members. People can be identified either by their IRC nick, full name, or email address:


<edd> foafbot, edd’s name
<foafbot> edd’s name is ‘Edd Dumbill’, according to Dan Brickley,
Anon35, Niel Bornstein, Jo Walsh, Dave Beckett, Edd Dumbill,
Matt Biddulph, Paul Ford

So I found myself being dragged into a tedious pissing contest (please contain your surprise) on a mailing list (I have since unsubscribed: if people have nothing to contribute, why can’t they just shut up?), and in the course of it I was sent a URL from http://makeashorterlink.com/?Y22052D71. Seems my antagonist thought I could be more pleasant as I pointed out that his project’s manual pages were wrong, the documentation was lacking, and no one seemed to have any answers.

Whatever. I found this ShorterLink stuff somewhat interesting and the FOAF stuff even more so.

Perhaps a non-commercial variant of Amazon’s friends list . . . .

And now I have my own ShorterLink.

blackberry harvesting

went out this morning with 3 half-gallon berry buckets (half-gallon milk cartons with the tops cut off) and came back with 1 bucket and 2 children filled to the top.

I’m freezing a bunch of berries (like half of them) and will bake the rest in a peach or nectarine and blackberry crostada, a kind of open fruit tart/fruit pizza on a crisp pastry crust. Have I mentioned how I love summer fruit?

bike testing

LeMond Racing Cycles 2002: Bikes: Reynolds 853 Pro: Zürich

Zürich

Our best selling LeMond- and for good reason.

I rode one of these yesterday: Oh, my goodness. Just a joy to ride. I’m thinking of going back for another test ride, but do I want to torture myself? A mere nine hundred dollars stands between me and it: I could get something nice for the half that, I know, so it’s not that bad.

I also rode a Giant OCR 3 ( a mere $550), and didn’t like it as much.