January 12, 2004

Adam Kessel — The Triplets of Belleville

The Triplets of Belleville· was the most delightful film I’ve seen in a long time. I give it my “best animated feature” award 2003-2004.

It’s best not to know too much about the story, I think, going into it, so I won’t say anything here. (Why do I am even bothering to write about it at all, then? So you’ll go see it!)

My favorite little detail was in the Triplets’ apartment: just for a second or so, a poster for Jacques Tati·’s Les Vacances de Monsieur Hulot· appears. The film pays homage to Tati throughout, and most likely alludes to other French films and directors that were beyond my surface knowledge.

If you want a more detailed description of the film, check out Roger Ebert’s review·. I have to agree with his hesitant characterization of the film as “Marquis de Sade meets Lance Armstrong.”

January 12, 2004 02:47 AM

Norbert Tretkowski — Mono and C#

Looks like my XFree86 4.3.0 backport for woody works well, at least I received some positive responses from people who installed it.

Searched for a sarge/sparc installer, but didn't found one. So tomorrow I'm going to boot these Sun Blade 150 at work via network and debootstrap a system on it. Last time I tried to get Debian on that machine it failed, because keyboard didn't respond. Must be some USB issues in kernel 2.4, this also happens with Aurora Linux, which I also tested. I hope this is fixed in kernel 2.6 now.

Did some more vim bugfixing in the evening. The last hours, I played around with Mono and C# a bit. It seems to work, but currently I'm not sure if I like it or not... :-)

January 12, 2004 01:44 AM

Scott James Remnant — Arch

This week appears to have been the week when everyone decided to play with arch for the first time, including myself. Gary's been talking about using it for libtool for a while, and Jeff's used it for working on the Planet code, so I thought I better have a go as well.

I'm afraid to say that I'm deeply unimpressed by it.

I've always felt that version control should be an almost transparent process, something that becomes second nature to you. Whatever software you use should fit in with this ideal, making it as painless as possible for you.

That is one thing that arch most certainly does not do! Some of the most basic (to my mind) things you'd want to do with your software, such as examining what was changed between two past revisions, involve multiple commands rather than the simple "diff" style command of other major systems.

I'm not going to deny that it seems very powerful, but its powerful at far too much expense. To use a really bad analogy, it's the sendmail.cf of version control systems.

Its one power feature, being able to work offline, is something that's possible with Subversion; my own "Offline Subversion" stuff works as well as the arch stuff seems to. The other feature, repeated same-branch merges, has several solutions for Subversion already available on the web.

It's been interesting learning how to use it, but I won't be using it for my own projects.

January 12, 2004 01:23 AM

January 11, 2004

Muli Ben-Yehuda — gby talk at Telux on Embedded Linux

I left work early today, around 1630 or so and went to hear gby talk about Embedded Linux at Telux. The talk was excellent, gby being his highly entertaining self, as usual. I heard previous incarnations of this talk in the past, at August Penguin I and elsewhere, but was still glad I decided to come. The talk was a bit too short, though, lasting only 50 minutes. The slides were extremely pretty, in Hebrew with the Codefidence logo featured prominently.

I missed the first 15 minutes of the talk, due to having to find something to eat or risking fainting. Being a slave to your stomach sucks. When I came into the classroom, gby was talking about how to do safe firmware and configuration updates, where safe is defined as being atomic - if a power failure happens at any given moment, you either get the old firmware/configuration or the new firmware/configuration, nothing else (bottom line is prepare everything and then use rename(2), which has an atomicity guarantee). Then he talked about the necessity of being able to reproduce exactly how you built the system, what configuration files you used, etc. This isn't specific to embedded systems, of course, it's correct software engineering regardless of the platform.

Then gby went on to talk about uClibc as a possible glibc replacement, busybox for your application needs, and even a tiny X server, Kdrive. He also showed a "slot server", which is a complete PC on a single PCI card. Neat toy, I wish I had one to play with. Think of running a virtual machine running completely on its own CPU...

After the talk, a bunch of us went to a nearby pub. I drank a half of Guinness, which was surprisingly tasty (I'm not much of a beer connoisseur). In attendance were gby, Shachar, Doron and Ori Idan (I think...). We swapped war stories, talked about the state of the industry and the state of Microsoft in particular, and much fun was had. The story of how I got a bug report from MS research was quite well received :-)

While we were talking, I asked gby where I might get a "slot server" card and if the specs were available. Turns out that he doesn't need his, and he loaned it to me to play with! Yay for cool new toys! Now I just need to find specs for it, and then get it to run Linux ;-)

January 11, 2004 11:31 PM

Jesus Climent — Mixxx about to be ready.

I have a complete compilation of Mixxx under i386 and PowerPC. There seems to be a problem with, at least, the native PowerPC linux version (there is also a MacOSX port): the interface freezes when using the "waveform visual". To avoid the problem, one needs "Visuals 1" set in the config file.

Now I have to clean the linda output and set some dependencies right. Also a proper handling of config might be needed. So far, I have been mixing some music with it, and feels so much nicer than any other mixing interface I have used: you can control the speed (pitch), set a cue point to start the stream from, rewind the stream, speed up and down the stream momentarily (to resync the mix),... even play backwards the stream!

I will send a HOWTO document to agi, maintainer of openvpn, to explain how to set a vpn tunnel, based on my experiences with the setting of a tunnel to use my wireless card.

Also, I am searching for the planetplanet agragator source tarball to set one planet-HispaLinux for the Spanish community.

January 11, 2004 10:26 PM

Colin Walters — Re: Arch

Scott: I agree about wanting revision control to be transparent. You say that arch doesn't do that. The example you give is trying to get a diff between two past revisions. In tla, this looks like:

$ tla delta project--branch--1.0--patch-1 project--branch--1.0--patch-5 ,cset
$ tla show-changeset --diffs ,cset

In Subversion, I guess this would be:

$ svn diff -r 1:5

Certainly, the Subversion example is quite shorter. However it's much less powerful too. Note that with delta, you can get a changeset between two totally arbitrary trees; one could be in your own local archive, and another could be halfway around the world in someone else's archive. And moreover, you get a real changeset, not just a diff. That means things like renames, symlinks, and binary files are handled.

Now, it wouldn't be hard at all to make a tla delta-diff command, building on delta and show-changeset. But there's no way you can turn svn diff into anything approaching the power of delta. There's really no comparison between the two commands.

Actually, here's a shell script to do what you want:

#!/bin/sh
tv=$(tla tree-version)
rm -rf ,,delta-$1-$2
tla delta $tv--$1 $tv--$2 ,,delta-$1-$2
tla show-changeset --diffs ,,delta-$1-$2
rm -rf ,,delta-$1-$2

Invoke it like this:

./tla-delta-diff.sh patch-1 patch-5

January 11, 2004 09:56 PM

Arnaud Vandyck — MovingJavaToMain

Moving Java To Main is one of the main focus of the debian-java community. To help this, I set up a wiki page at http://wiki.debian.net

http://wiki.debian.net/index.cgi?MovingJavaToMain

Don't hesitate to help us to have a good free java envirronment in Debian!

thanks for your support!

January 11, 2004 09:33 PM

Colin Walters — Weak Week

Besides the job news, it was kind of a slow week for me. Took care of a bunch of boring but essential life details like convincing the Ohio tax department that I didn't owe them anything (since I didn't make enough money to pay taxes).

On the free software end, I made a new Rhythmbox release to fix a bunch of things that had kind of accumulated. Still though, the bug reports continue to flow in. I think a week or two ago, I was the #1 bug closer in GNOME. You would think would be a good thing - except that almost all the bugs I closed were duplicates or caused by user error. But at least some of them were real bugs, and were fixed. So we're moving forward.

I also merged an 1000-line changeset from Jeffrey Yasskin (good job!) and work on the file chooser from Sriram Ramkrishna. It's great to be getting all these patches. Oh, and Christophe Fergeau is really close to getting Rhythmbox/IPod integration going! Very very cool.

January 11, 2004 09:28 PM

Colin Walters — We'll miss you, Mark

I am very saddened to hear about Mark Finlay passing away. He contributed a lot to Rhythmbox and GNOME, and we will miss his work a lot. It seems like even up until the very end he was hacking. Mark, your dedication will be remembered.

January 11, 2004 09:17 PM

Antti-Juhani Kaijanaho — The annoying banner thingie

I started the grep-dctrl rewrite last spring. I soon had most of the code implemented, and I asked for testers on -devel. There were no debs at that time, and so it probably was not much of a surprise that nobody volunteered. I got discouraged and abandoned the project for several months.

In August, just before my vacation at work ended, I decided to polish the code, and I did. I prepared debs and was ready to submit them to unstable. (At that time I was under the impression that experimental was the wrong place for any grep-dctrl stuff.) I felt it was necessary to warn users about the code, since it was sure to contain nasty surprises. I considered sending a mail to some lists: -devel-announce would have reached my audience, but the message would have been inappropriate there; I feared it would not be noticed in -devel. I considered a message at installation time,either with debconf or without. However, debconf messages are seen by system administrators and not users, so that was not an attractive option. I finally decided on that annoying banner: 


                                  NOTE                                    
 grep-dctrl has been rewritten from scratch.  Although this does add new  
 features, regressions are certainly possible.  Please watch for them and 
 report them to the BTS.                                                  


(The above annoying banner will not be shown to you again, unless you
request it with the -B switch.  It will also be removed entirely soon.)

I took great care to make the banner as unobtrusive as possible, still taking in account that it had to be prominent to have any effect. I added logic that would suppress the banner after a user had seen it once (using a dotfile in home directories, ~/.grep-dctrl-banner-shown; if you find one in your home directory, you can now safely remove it). I did not print it to stdout (it would clobber the output) nor to stderr (it might be redirected, which would have been counterproductive); rather, I used /dev/tty. After all, many people use grep-dctrl in scripts without realizing it; especially they would need to know that there was a weak link there.

Nevertheless, I expected to get flamed for this. To my considerable surprise, I got no reactions - aside from several good and substantial bug reports. The banner stayed in longer than I intended, mainly because I was exhausted by my teaching duties during the autumn. I finally removed it last week, in 1.103.

And then, several days after the banner had been removed, I get a bug report (which, strangely, is still open even though I sent a closing message almost 24 hours ago). That's when I decided to post this entry.

January 11, 2004 08:59 PM

Wouter Verhelst — sunday, busy sunday...

I've been working on the website for our flute choir 'till 18:00, at which time I went to the rehearsal (of the choir, not the flute choir). The site is almost done now, just a few little details remain. I've even found the solution to a few things that've been bothering me before, and for which I didn't find the solution at first.

On the web front, I was surprised by a friend who's started a webdesign company. About a week ago, he sent me the URLs of a few websites he'd written, but none of them contained 100% valid HTML (most were close, though). I replied to him about it, and even if I had the idea that he didn't seem completely interested at first, I seem to have convinced him that the W3 standards are a good thing: today, he mailed me back with a new URL of a different website which he's built, and which he's made to be perfectly valid HTML for a change. Nice. On top of that, it's a good-looking site too. This company's got a future :)

After sending mail to Ingo yesterday, we gave one last try at fixing arrakis, and it seems to have worked; it's happily compiling packages again now, even if it's compiling packages on a smaller hard disk now. This is a good thing; with the backlog created after the break-in at the end of last year, m68k is having trouble keeping up currently, so we can use every buildd available.

Let's finish that *censored* website now...

January 11, 2004 08:58 PM

Gustavo Noronha Silva — 10 Jan 2004

Almost 24 hours of hackparty happened again -- friday evening some friends came to my home and brought their computers, we set up a network and one of the computers shared my adsl connection with all the others.

It was nice. We played a lot of gtetrinet, played with apache, cyrus-imapd and I even worked on the pt_BR gnome translation stuff. We were still together playing gtetrinet on saturday's evening, when people went away -- although the computers are still here right now.

This night seb128 gave me some input on what he thinks about galternatives (which you can find on Debian experimental, right now). I'm right now trying to finish the alternative options adding/removing/editing. A nice flamewar on 'is free software possible?' and 'should all software be free' on #debian-br is stopping me from working, though -- damn, those things take me away from code harder than frozen bubble =D....

January 11, 2004 03:44 AM

Tollef Fog Heen — gcc, glibc, amd64

I don't like gcc. I don't like glibc. And I surely don't like them on amd64. I've been trying to get an up-to-date biarch GCC built for amd64. This failed due to glibc (32 bit) not being biarch. So I started rebuilding glibc in biarch mode. It killed first alfa, one of the boxes at work (nothing critical, I was supposed to stress-test it). Then I killed ravel, one of the developer-accessible AMD64 machines. This was of course due to a kernel bug, so after maswan went out there and rebooted the machine, we got a new kernel installed; this one seems fairly stable so far. So, in order to rebuild gcc so that I can test that the new glibc works, I'll have to rebuild glibc (32 bit), then gcc (biarch) then at last glibc (32 and 64 bit). Fun, isn't it?

January 11, 2004 01:56 AM

Tollef Fog Heen — pybloxsom, textile hack

Bah, in order for pybloxsom with the txtl backend to work somewhat properly, one has to fix the txtl.py:cb_preformat definition to something like: 
def cb_preformat(args):
    f = open("/tmp/foo.log","a")
    if args['parser'] == PREFORMATTER_ID and \
           args['request'].getData()['flavour'] == "html":
        return parse(''.join(args['story']))
    else:
        return ''.join(args['story']).replace("&","&")
This will make the RSS feed work properly.

January 11, 2004 01:47 AM

Matt Hope — R, Bioconductor, Quantian.

Recently got an email from Dirk Eddelbuettel, regarding my attempts at packaging Bioconductor, some computational tools for the statistics package, "R". He's been working on Quantian, a Knoppix+openMosix for "numerical and computational analysis", and also the guy who maintains R, and works closely with the upstream for R. Sometimes its just nice to be noticed =). He simultaneously perks my interest in working on R packages, and CRAN (much like CPAN, but for R), and finally getting up and updating the Bioinformatics/Debian pages.

From what I've seen so far on the CRAN front, it'd be sanest to write a replacement for R's build in package downloader/installer, something like dh-make-perl. Unfortunately, this won't help me with Bioconductor, as the new release (1.3) doesn't seem to have a single tarball any longer - just a script that downloads a tarball of each module (about 20 or so, from memory). I need to totally overhaul the packaging for Biocondutor to account for this.

January 11, 2004 01:31 AM

Junichi Uekawa — Another machine from Daniel Josua Priem; I'm setting up mass-install pbuilder script there, and tring to run it

Another machine from Daniel Josua Priem; I'm setting up mass-install pbuilder script there, and tring to run it.Thanks to Ganneff for helping me set up the server.It's going to output a large amount of logfiles; and it's going to needsome way of presenting it.I could just go through each log with lv and file bugreports like I did in 2002, but that's not going to really scale.BTW, if you are a young aspiring hacker needing more experience in Debian packaging,it would be an interesting challenge to fix all FTBFS bugs found with pbuilder; there is no lack of it.

January 11, 2004 01:04 AM

January 10, 2004

Scott Dier — New 'blog

I started up debianblog here so I didn't clutter Planet Debian with stuff from my other blog. I've been quite busy with work related Debian things lately, and hope to get some of the work I've done released soon. I'll try to track what I'm doing on that front here.

January 10, 2004 11:19 PM

Martin Michlmayr — pissup in a brewery, ill, PowerPC 64 port

There was a meeting at my institute yesterday in which every unit gave a presentation about what they are doing. This was quite informative, although I cannot really relate to what most of them are doing. Afterwards, I met a PhD student who has the same supervisor as I. This was pretty useful. After the meeting, I went home since I still don't have an IP address at work. I tried getting a new chair for my room, but the furniture officer wasn't around. I'm really quite unproductive because sitting at my table isn't very comfortable. In the afternoon, I actually unpacked my suitcase (these RL things...). Later, I went to the pissup in a brewery which terryfish had organized. Quite a few Debian people were there — the usual Cambridge crowd.

I think I'm ill. In fact, I know I'm ill, but I guess I don't want to admit it. ;-) I spent most of the afternoon in bed, first sleeping and then answering some e-mail. It looks as if we'd finally get remote access to a 64 bit PowerPC system. I talked to IBM several times about this port, and I hope we'll get a dedicated machine in the near future. I also contacted some Debian people who I think should come to FOSDEM. There's a track on usability and I hope Mario Lang can give a talk about the work he's done for debian-installer.

January 10, 2004 08:39 PM

Emanuele Rocca — qa.debian.org/orphaned.html temporary replacement

The orphaned packages not set to QA Group web page was last updated on Wed Dec 31 and the reason is that the LDAP interface to the BTS is down.
I (incredibly) had some free time yesterday, so I started to write a script that allows to generate a similar page parsing the Work-Needing and Prospective Packages page.

Unfortunately, I discovered that also that page is built through the ldap interface to the bts, so the first version of my script was a little useless.

Yesterday night at about 3 AM I tried to adapt the script to the BTS html output for the wnpp pseudo-package, but master ran out of space again and the BTS didn't work properly...

When master's problems were solved (4 AM) I was really too tired...

Today I finished the script and it seems to work fine. :)


Some days ago I found that drivel is totally broken.
On my system it crashes ungracefully when I try to insert my password :(

Drivel maintainer seems to be unable to reproduce the bug, but the upstream claimed that the bug is fixed in drivel 0.9.2
But I am currently using logjam and I like it.

I am a bit tired because I slept only 4/5 hours last night, but this evening I'll have pizza & beer with some friends, so I've got to stay awake!

January 10, 2004 08:15 PM

Jordi Mallach — Big day for meta-gnome2

A few nice things happened in the meta-gnome2 front: yesterday, aj and Kamion kicked jack-audio-connection-kit and alsa-lib into testing, clearing the way for a number of packages like gst-plugins and the packages that depended on it gnome-media, nautilus-media, rhythmbox, etc. Today, aj removed the remaining 7 days of wait for libbonobo, and this is unblocking a new set of packages; and Kinnison did a quick processing of libxslt, which was again stuck on the new/ queue in auric. This will fix the builds of most of the remaining meta-gnome2 problems, so if no new problems come up, it's should be mostly ready when a few of the affected packages get retried. When/if the "must have" packages are in, I'll consider doing a temporary upload removing the less important packages so meta-gnome2 can go in, making Sarge's testing a bit more realistic for people who are doing desktop installs. You never know, given our record of bad luck until now...

January 10, 2004 07:18 PM

Scott James Remnant — Terraforming

You know you've had a heavy couple of days hacking when several people blog about your changes before you do! Last time I posted Jeff and I had just merged our fairly small patches together and got ourselves running on the same codebase.

Friday night, I had nothing better to do, so I started playing with feedparser to see how easy it would be to get ETag and Last-Modified support working properly so Planet wouldn't download unchanged feeds. It turned into a near-total rewrite ditching much about spycyroll. that wasn't pretty and basically writing the core classes from scratch.

It was worth it though, all of the Planet management code got abstracted into a new planetlib.py and a really simple API for building an aggregator emerged:

from planetlib import Planet, Channel
p = Planet()
p.subscribe(Channel(url1))
p.subscribe(Channel(uri2))

for channel in p.channels():
    # Make your blogroll

for item in p.items():
    # Make your feed/page


With a clean-up for the template code, making it all a bit more logical, I turned in for the night, which is where Jeff took up the baton. A new experience, hacking on something with someone's who's 11 hours offset from you, it almost is relay hacking.

He merged in my new code and got multiple output support working pretty damned well, and made some great template RSS 1.0, RSS 2.0, FOAF and OPML feeds that I've pretty well been able to drop straight in place for Planet Debian.

I suppose somewhere next on the list is fixing the single items_per_page into something template-specific, and perhaps multiple output HTML pages for those that can't keep up. I'd also like to see some way of maintaining the Planetarium using FOAF, which is probably the right thing?

I'm going to take a break and so some real work for a bit though, secretly hoping Edd figures out the perfect technology for it and next time I merge with Jeff the new code's there :-)

January 10, 2004 07:17 PM

Jesus Climent — Wedding and DVD.

Today I had a wedding. It is nice to hear that "the man should be the head of the family and she should take care of the house and children" during the priest speech in a language I cannot understand, because otherwise I would have stood up and left the place without further comments.

I want to have the right to stay home while she is working and be plugged to the net (and eventually take care of the dishes, wash the clothes and hoover the floor).

No, seriously, I do not like that kind of statements.

Finally I got a DVD to put on the radio, but have not had the time to change it.

I also managed to release abcde 2.1.11 with some bugs closed, and hopefully i can start leading to a 2.2 stable version (which will be ready for Sarge) and start a 2.3 with all the nice stuff I have in mind for it.

Colin: Last but not least, the pyblosxom problem is still here, and setting the debug option on does not gime me more info than the already explained: after all the modules being loaded successfully, I get a "Premature end of script headers: /webs/pumuki/blog/pyblosxom.cgi" and a "Internal Server Error" as a web page. If I run the script by hand, it works and spits a valid html web page to the standard output.

January 10, 2004 06:59 PM

Wouter Verhelst — kernels, websites, and arrakis

Bugs suck, but they suck less if the maintainers of the broken code are responsive. Which the cpufreq maintainers are. Been sending mails over and back all day, so that p4-clockmod bug will be fixed soonishly.

Mother bough some chocolate today, which is both a blessing and a curse, (very) much in the same way as it would be if you'd buy a bunch of cigars to a smoking person. Meaning that I've been working on a website for the flute choir I play in (yes, I play the flute) which should've been finished at least a year ago or so.

Just now, arrakis sent me a failed build log, because it needed sudo which wasn't installed in the chroot. After a day. Meaning, it's probably been hung and rebooting all day yesterday (I don't manage it being up, just handle the logs). If it doesn't even have enough time between hangs to find out that it doesn't have sudo, we should probably just shut it down so that it doesn't start being greedy with packages.

Let's mail ingo about that.

January 10, 2004 04:56 PM

Jeff Waugh

Much of the first part of today (which has been a very long 'today') spent hacking on release strategy proposal, web planning documents and library.gnome.org. It's going to rock way hard when these planets align. I was really hoping to get the release strategy stuff out of the way today, but it's kinda laborious. Cool, but pushing lots of intellectual mess in my head into understandable (and, well, convincing) prose is hard work. Same stuff with the web docs, but they're mostly point form with discusson bits in between. Actual proper hacking on l.g.o and Planet have been head-restful which is pretty disturbing. Will be taking a break on Sunday and Monday (grooving to the tunes/noise/ambience at K's audio-conf) to get the release doc out, then hammering out the web stuff as soon as I get back. That'll be arduous, but fulfilling. Meanwhile, a pretty front page is up, and Aaron is getting his head around the current state of things. Ready to rock.

January 10, 2004 04:54 PM

Jeff Waugh

Very saddened to hear of another loss in the GNOME community: Mark Finlay was young, eager, furiously energetic and just heading into university - where I'm sure he would have blossomed into a great GNOME and Free Software hacker. I chatted with Mark a lot during his time with GNOME, helping him grok how GNOME worked, where cool things were happening, and how to get more involved. I had no idea he was unwell until an off-hand comment in his diary a couple of months ago - at first, I couldn't believe him because he was always so active, pumped up and ready to rock... I dearly wanted to see where he'd be in a few years.

My condolences to Mark's family and friends, I hope they know that Mark will be greatly missed by his world-wide gang of 'funny computer friends'. Rest easy, Mark.

I've always wondered what things would be like in 50 years, when our GNOME friends would be leaving us regularly... But I fear that there are too many friendly footprints in heaven already.

Love each other, and make sure your friends know you care.

January 10, 2004 03:27 PM

Edd Dumbill — Planet fun

There's lots of fun happening in the Planet-* blog aggregation world. Scott Remnant and Jeff Waugh have made a lot of improvements to the code base that runs Planet Debian and Planet GNOME. This has enabled Jeff to write some whizzy output templates for Planet GNOME that means it can interoperate better with my RDF-tastic view of the world.

First off, there's an RSS 1.0 feed for Planet GNOME. More excitingly, Jeff has exported the list of participants (I suppose we must call it a "blogroll", but I don't care for the word) in a format similar to the FOAF/RDF blogroll that Planet RDF takes as its input.

To test, we ran the Planet GNOME blogroll through the Planet RDF aggregation software, and it worked just great. Obvious, I guess, but very pleasing! Here's the test HTML output and RSS 1.0.

While there are quite a few aggregators around, what I like about what we're doing with the planets is that we're being good standards-loving web citizens. We're processing the gunk that is RSS-in-the-wild, and churning it out again as near XHTML as we can, and exporting an RSS 1.0 feed that's valid RDF, as well as avoiding the nasty escaped markup trick. Should warm the cockles of Norm's heart!

January 10, 2004 02:46 PM

Junichi Uekawa — I'm running pbuilder, and it's as usual, it's probably going to take around 2 weeks to get a result of a full build run

I'm running pbuilder, and it's as usual, it's probably going to take around 2 weeks to get a result of a full build run.I'll need to speed things up, and I'm hacking towards something feasible and interestingto speed things up.

January 10, 2004 02:00 PM

Jeff Waugh

Had a good day today with arch and all this Planet hacking. First off, I had to merge Scott's changes into my tree. I had made a few touch-ups here and there, so Scott's changes were based on a slightly older revision. Merged without a hitch, despite some dunderheaded mistakes on my part. I was going to make sure it worked on Python 2.1, so I worked on a different machine by pulling the archive over, and merging changes back. Then I decided that making it work on Python 2.1 was dunderheaded too, so I reversed the stupid changes and started hacking on the interesting stuff. Tonight, I'm going to pull my GARNOME and Planet trees off onto my iBook so I can hack on them this week while I'm at linux.conf.au.

So, while arch can sometimes be inscrutable and petulant, and has a definite learning curve for the average Free Software CVS refugee, it's working for me, and I'm going to keep working with it. Will arch be in GNOME's future? I think there are some pretty compelling reasons to try; but it will need more support for visualisation and code browsing/searching tools (like LXR, Bonsai) to really sell it.

As Edd would say: Sticking with arch.

January 10, 2004 01:22 PM

Jeff Waugh — Planets in Alignment

Some crazy progress on Planet. Thanks to Scott's excellent work ridding us of old horrors, I've managed to build multiple output support in record time. Planet GNOME now supports RSS 2.0 and RSS 1.0 feeds of its aggregation, and further output formats are just a template away. In the next half hour, it will provide its blogroll as OPML (which is the world's worst XML, but we're stuck with it) and probably as RDF/FOAF too.

Meanwhile, Zach is experimenting with Planet Lisp. Rock and roll is where I hide.

January 10, 2004 11:17 AM

Jeff Waugh

spiv points out some Penny Arcade prescience of Planet Proliferation. Hell yeah! Scott just sent a lengthy email describing his massive rewrite of Planet. It is now truly an awesome aggregator, with none of the old pyblaggisms. He has totally kicked its arse. I'm going to work on python2.1 compatibility, RDF/FOAF/OPML output, and then port the desktop aggregator stuff I've been working on to it. Sang choi bao!

January 10, 2004 04:17 AM

Chad Miller — Not quite dead

I'm slowly getting better. Boy, that was a tough week. My ribs are still sore from the coughs and my butt is still sore from the injection needles. [MaryEllen] wants us to do more things together, and I suggested that doing something vaguely healthy is a good idea, and offered...

January 10, 2004 02:33 AM

Frank Lichtenheld — we're partly there

packages.debian.org is partly back again: http://www.debian.org/distrib/packages

Problems that remain (temporary):

Contains information about non-i386 packages

Includes DDTP translations

Better parsing/using of input data

Handles virtual packages

Creates an alternative compressed text list of all packages

Some minor fixes

January 10, 2004 02:25 AM

Junichi Uekawa — I've noticed ecawave doesn't build on unstable

I've noticed ecawave doesn't build on unstable.I haven't noticed that happening since I've been working with newer packages and been uploading to experimental. Gah.It seems like development files for ecasound aren't installed properly, and I've backported some patches.

January 10, 2004 02:11 AM

Tollef Fog Heen — planet.debian.net

It seems like prodding elmo on IRC the other day helped; "http://planet.debian.net/":http://planet.debian.net/ now works. Keybuk moved the code from gluck to my local server, which should be just as good, and not require the ugly proxy hack which were there in the beginning. I haven't been able to fix mailman yet, but hoping to fix that tomorrow. Also, remember that 2004-01-04 13:37:04 UTC is 2^30 seconds since the epoch. Find (or make) a party and celebrate. :)

January 10, 2004 01:08 AM

Tollef Fog Heen — Upgrading kernels.

The last of the workstations at the student house is now rebooting with 2.6.1, so they should be safe from local exploits. At least the remap one. I really, really need a better way to do this, cfengine and some custom addition to make it possible to tell cfengine that "this is the minimum set of packages that are to be installed" would be nice; I wonder if debsync can be used. While upgrading, I discovered a bunch of other, related problems, like system users which aren't present on the workstations (because /etc/passwd and /etc/group are both rdisted), we need to get rid of the silly 15MB /boot on some of the boxes, get rid of ide-scsi and similar stuff. I also did the discovery that the k7 kernel wasn't too happy on an i686; X refused to start. At least, I thought it was because of the wrong arch. Later, I discovered that psmouse isn't enough to get mouse working, one need the mousedev module as well. I'm looking forward to discover 2.

January 10, 2004 01:03 AM

Jeff Waugh

Sun may change Linux desktop platform... Where in the World is the KDE League?... The GNOME Lover's Guide to linux.conf.au... GUADEC 2004 Call For Papers. Fun, fun, fun...

January 10, 2004 12:51 AM

Arnaud Vandyck — Moving debian-java packages to main!

I'll begin to audit some debian-java packages in contrib this weekend to build them with the latest kaffe in Debian (1.1.3) and with cdbs when they build with ant.

I hope I'll be able to put some libraries in main but the most important thing I'd like to achieve is to be able to make a web page with all the debian packages in contrib and the reason why they are in contrib.

If you want to help, don't hesitate ;)

debian-java@lists.debian.org

January 10, 2004 12:45 AM

January 09, 2004

Wouter Verhelst — bugs suck

Seem to have found a bug in the p4-clockmod CPU Frequency Scaling-driver: it doesn't allow you to select the highest possible rate the processor is able to run. Which pretty much puts me back at having a 2.0Ghz P4M instead of a 2.2 one.

Trying to remove the module and replace it by the ACPI driver didn't work either; modprobe segfaulted. Grmbl.

January 09, 2004 05:29 PM

Mark Brown — 9 Jan 2004

One good thing about packaging software with bugs is that it can make you feel so much more productive. This week I've hardly had time to look at any of my Debian packages yet thanks to a series of Leafnode releases there's a bunch of mails in my mailbox telling me I've actually been doing things, even if I know they were fairly trivial things.

January 09, 2004 04:57 PM

Jesus Climent — Bad luck with pyblosxom.

Trying to integrate my actual hand-made web page with some weblog tool, I was told by Amaya about the use of PyBlosxom. With my firewall being an OpenBSD machine, the Apache server is chroot'ed by default, so installing the needed python and friends has been a bit more troublesome than expected. However, althought right now everything is in place and there is no complaint about modules, the cgi dies with a very lengthy and explanatory "Premature end of script headers: /webs/pumuki/blog/pyblosxom.cgi".

To solve the problem I have been thinking about using an inverse Proxy installed in my firewall and store all the web pages in my server, which will help also with Apache::Gallery (no, i have not installed it in the chroot, and I do not plan too, after the python nightmare).

The weekend is here, and my radio is waiting for reparation (a bad Sony DVD unit).

January 09, 2004 03:32 PM

Jordi Mallach — Stupid White Men

One of my Christmas presents was the British edition of Michael Moore's Stupid White Men. I got it in English because I tend to enjoy movies or books more if I see/read them first in their original version. My sister, on a totally unrelated xmas present, got the Spanish version, just as a few friends did. They say it was impossible to find a copy of this book in Valencia the weeks before and during Christmas. Not surprising, as Bowling For Columbine has been one of the most seen (and discussed) films during 2003 in Spain, when all the Iraq war stuff was going on.

Anyway, I didn't know if I should start reading this first, or go ahead with Ferran Torrent's "Espècies protegides". I took "Stupid White Men", as I have very recently read the first part of Torrent's book, so it's probably time to switch a bit. While I travelled to Uni in the subway, I couldn't stop laughing at every single page of this book, and it was just the "introduction to British readers", where he tells the story about the book being printed on September 10, 2001, and after the 9/11 attacks, Harper Collins refusing to put it on the shelves if Mike didn't "rewrite 50% of it". A nice little story of how a bunch of librarians put so much pressure on the editor that they were forced to put the book out, without any media covering or anything, and the book literally dissapeared from book stores the first day it was out. I'm going to enjoy this read.

January 09, 2004 02:39 PM

Junichi Uekawa — pbuilder build daemons will revive

pbuilder build daemons will revive... maybe. I don't quite know how things are arranged yet, but I've accepted a generous offer and I'm going to set up pbuilder on a remote system.This kind of arrangement hasn't been in place for the last year, so looking forward to getting pbuilder hacking more active.Thanks go to Daniel Josua Priem for the hardware.

January 09, 2004 02:31 PM

Pascal Hakim

I uploaded a new fortune-mod yesterday, and added a debian cookie file. I was getting sick of not knowing where to put Debian quotes that were getting sent to me, so this is a good solution for now.

I still have one last change I want to make to fortune before Sarge freeze, which is a fix for Bug#22545. I've felt the need for it before, and a few people have mentioned it to me in passing before as well, but having a configuration file for fortune, sounds like such overkill!

I've been looking at prices for various computers parts lately, as I'm keen to acquire a new desktop, since my last one become my web server, and is now located a few kilometres away from my desk. I originally did this with the intent to upgrade, but it doesn't seem to have happened yet. Anyone got some cheap sites where I can buy cheap stuff in Sydney?

I tried to fix one of the moderated lists today, as it appears the moderators were getting all the spam that was directed to the list. That list hasn't received a single bit of spam in 36 hours, so now I have no way of knowing whether my changes worked or not. How typical! At least I had a chance to fix debian-private while I was doing that, so the LDAP unsubscription should start working again now (at least from murphy's end). I also managed to find the source of all this spam I was getting (Note to self: when subscribing yourself to a mailing list to test it, make sure to unsubscribe, specially if you can't understand or read that language and its charset ;-).

Played with netfilter some more today. It's definitely interesting. There's a couple of ideas in the back of my mind that I'd like to play with.

January 09, 2004 02:30 PM

Mark Howard — Bug me back

In response to Edd's Blog entry:

Interoperability between bugzilla systems is quite likely to be achievable. Getting debbugs into the picture is likely to be far more difficult. There are a lot of things that we can already do to help though:

debugzilla
Erich wrote this script to look through debian bug listings and check any reports forwarded to bugzilla. It outputs a listing of all the bugs and the bugzilla status, so you can see when it changes. This is extremely useful when you have many bugs forwarded to bugzilla. Unfortunately debugzilla isn't working at the moment - bts interface down?
Bugzilla CC
Simple enough, setting the bugzilla CC field to package@packages.qa.debian.org is quite helpful. It would be even better if bugzilla would let us set it to the xxxxxx-submitter@b.d.o.
Encourage users to report upstream
Most of the galeon bugs I get are of little use to Debian - they are feature requests, etc. To reduce my workload (and so increase the real work I do, I have written a /usr/share/bug/galeon/presubj file to ask users to report upstream for most things.
bugwatcher
Have to mention it since it's my program. Doesn't help keep track of non-debian things, but it certainly makes to task of setting the forwarded field easier

I think far more important though is getting maintainers to actually forward bug reports. Looking at some of the larger desktop apps, there are often hundreds of open reports with nothing done to them. No amount of extra systems will help with that.

As for XML or something for bug reports, bugzilla already has quite good xml output. Debian bts only has html over http. A more machine readable interface to debbugs would be really useful. See #207225: debbugs: add RFC822 output option.

January 09, 2004 01:47 PM

Daniel Silverstone

Thanks to some source dug out by Ross Burton I managed to get most of the way to my irda monitor last night. As an aside, until Daniel Stone gets his updated dbus 0.20 packages up, I have put my hacked-together cruddy ones here.

January 09, 2004 01:34 PM

Wouter Verhelst — gaaah!

I found the bug in nbd-server's postinst:
for i in $(seq 0 $(( $NUMBER - 1 )) )
do
  TMPFILE=$(mktemp /tmp/nbd-server.XXXXXX)
  echo foo >> $TMPFILE
  echo bar >> $TMPFILE
done
mv -b $TMPFILE /etc/nbd-server

how stupid is that? :-)

fixed package ready for upload..

January 09, 2004 01:33 PM

Jeff Waugh

Are you a SuSE dude? James Ogley - you may know him from such websites as Rubber Turnip and usr local bin - has set up a Planet SuSE and is calling for SuSE community bloggers to jump on board. I'll add it to the Planetarium list once James finds an official home for it. :-) ROCK ON!

January 09, 2004 01:33 PM

Matt Hope — Debian

uruk-hai:mpiblast-1.2.0/src% rm -f ./formatdb.o ; /usr/bin/mpicc.mpich -c ./formatdb.c
uruk-hai:mpiblast-1.2.0/src% ls ./formatdb.o
ls: ./formatdb.o: No such file or directory

uruk-hai:mpiblast-1.2.0/src% rm -f ./formatdb.o ; /usr/bin/mpicc.lam -c ./formatdb.c
uruk-hai:mpiblast-1.2.0/src% ls ./formatdb.o
./formatdb.o
/usr/bin/mpi{cc,CC}.{lam,mpich} are alternatives, that get linked to /usr/bin/mpi{cc,CC}. Surely they should have the same behavour?

This is currently causing problems, as I'm trying to make a nice tidy package of mpiblast that has both support for mpich and lam. Currently, I'm very tempted to just ignore the mpich support.

January 09, 2004 01:12 PM

Federico Di Gregorio — Work 3C2000! Work!

I finally managed to have a 3Com 3C2000 Gigabit board working. It was really easy once I decided the drivers distributed on the CD are pure crap. It is just as easy as googling, but adding the link here anyway won't make things worse. The working drivers are from SysKonnect and they even recognize the 3Com board automatically; no need to manually add the vendor tag to the list of supported devices. Here's the link: SysKonnect Linux 2.4.23 and 2.6.x drivers

January 09, 2004 12:47 PM

Edd Dumbill — Bug me back

For as long as I've been involved with free software, I've been reporting bugs. For most of that time, tools like Bugzilla have been around to help. Very handy beasts they are, too.

Since I became a Debian developer, I've also been on the receiving end of bugs as well, which has led me to want a greater degree of interoperability between bug tracking systems. Debian has a fine bug tracking system but, as with all such systems in common use, it is an island.

The isolation of bug trackers is, however, an increasingly frustrating situation. With Debian, I often get many bugs reported that are in fact the responsiblity of upstream maintainers of software packages. That is, I get bugs filed on Epiphany, say, which then need to be filed in the GNOME Bugzilla. This involves a lot of cut and pasting to create a corresponding bug on the GNOME side, and then human intervention to ensure the bugs remain synchronized, and that the GNOME maintainers can talk effectively with the person who reported the bug to Debian. All in all, it's a right pain.

It's not just with Debian. Various parts of the GNOME project reuse Mozilla code, for instance, and so on. Clearly we need some kind of bug system interoperability.

What I'd love to see is some sort of protocol that would provide the following features:

This seems like a small enough target for to be a practical one or two week long hacking project. Something I'd love to see that's related to this, and maybe even part of the solution, would be an XML or RDF vocabulary for bug reports. Such documents would carry the basic information common to all bug systems and allow lossless storage of extension information specific only to certain bug trackers.

Many of us in the software world have a heavy reliance on bug tracking systems, so it seems worth putting in some effort to upgrade them to the degree of automation we now see emerging elsewhere on the web.

January 09, 2004 09:48 AM

Colin Walters — Pyblosxom Debugging

Jesus - you can get a nice traceback if you go into the main Pyblosxom CGI script, and uncomment the line that looks like this:

# Uncomment this if something goes wrong (for debugging)
#import cgitb; cgitb.enable()

January 09, 2004 08:59 AM

Sebastien Bacher — 8 Jan 2004

Yesterday, before going to bed, I've tried to update my expired key again ... and surprise this morning, it seems to have work :))

Thanks to the guys who have uploaded packages for me during this month (especially Julien Blache, Jordi Mallach, Josselin Mouette and Aurelien Jarno).

Now I have to go to work ... and tonight it's the Week End, nice :)

January 09, 2004 06:32 AM

Scott James Remnant — arch scares me

We've now merged Jeff's and my changes to spycyroll to form a new beast we're calling Planet. Planet Debian is now running on that common codebase. He's made the code available in an arch repository so we can hack on it further and turn it into a kick-arse aggregator.

So I've spent the last few hours trying to get to grips with arch, for the first time. Perhaps I shouldn't've read the tutorial at 4am, it seemed to spend more time telling me about arch's internals than how to use it, and I came away even more baffled. Asking on IRC only produced advocacy (little surprise) but finally a couple of URLs to read and I think I get how it works now. It's Version Control without the Control. Very strange.

Maybe I've got it a bit wrong, but as far as I can tell, if I want to work on Planet alongside Jeff, I have to make a branch of his code in my local archive. Getting his changes involves a manual merge, and if I want him to have mine I have to publish my repository somewhere and let him merge the changes back.

I'm almost positive that I don't like it, it just seems far too involved for something which should be so trivial it becomes invisible. I'll give it a few more days before I go off it completely though.

January 09, 2004 04:49 AM

Gustavo Noronha Silva — 8 Jan 2004

3 am in the morning and I'm finishing building my shining little package galternative version 0.5 to fix its first bug on the BTS -- going to experimental as soon as I test it.

I am always worried that Debian has loads of cool, neat stuff that are very useful but are, sometimes, poorly documented and/or acessible to 'simple' users. 'poorly documented' meaning mainly that you cannot find this in a 'Debian Features' list or something. One of those things is the alternatives system, which I think is very useful, but too little people know about, and even if it was easy to find, update-alternatives is a difficult to use tool, so my mother would probably have some difficulty using it.

Although not for my mother, GAlternatives is my way of trying to take that power into non-developer/non-geek users.

Yesterday was a cool day. I spent lots of time reading through my email backlog and found out that one of my other "let's expose Debian's power" project got a mention on the Debian Weekly News - I'm rethinking apt-howto, so if you have complaints, mail me now! =)

I also sat thinking a lot about stuff I read on those emails like the translation issues on the gnome development lists. devhelp's future is also stick in my head for some days.... I've been thinking about trying to stop getting the 'gnome translator' hat off and try wearing a 'gnome (devhelp) developer' hat to have it going, though I am not sure I'll have the time.

January 09, 2004 04:25 AM

Joshua Kwan — 2.6 kernels galore!

Well, finally 2.6.1-rc3 was released, which involves an important revision from Linus that was logged as follows:

ChangeSet@1.1491.2.2  2004-01-07 18:26:41-08:00  torvalds@osdl.org
Fix silly mremap test.

Get off the drugs, Linus.
Not using it yet, but I'll be using 2.6.1-rc2-mm1, which Andrew Morton released about five seconds after 2.6.1-rc3 was released. Hence it's to be considered basically a 2.6.1-rc3-mm1 because it includes all of the latest BK from Linus by default.

I pestered aaronl again to sponsor nethack 3.4.3-3. He's quite busy but I hope that he'll upload the packages soon. Speaking of NetHack I also ascended a Knight this afternoon. That makes only four classes left: Priest, Healer, Caveman, and Archaeologist. And then maybe I'll have had enough Nethack for a long time...

January 09, 2004 02:29 AM

Marco d'Itri — Fixed inn2

Uploaded a new inn2 package with the security fix. Now let's see how many weeks it will take to enter testing... (Free clue: INN 2.x is useless on arm and m68k systems.)

January 09, 2004 01:37 AM

Matt Hope — Debian

<pre> uruk-hai:mpiblast-1.2.0/src% rm -f ./formatdb.o ; /usr/bin/mpicc.mpich -c ./formatdb.c uruk-hai:mpiblast-1.2.0/src% ls ./formatdb.o ls: ./formatdb.o: No such file or directory </pre> <pre> uruk-hai:mpiblast-1.2.0/src% rm -f ./formatdb.o ; /usr/bin/mpicc.lam -c ./formatdb.c uruk-hai:mpiblast-1.2.0/src% ls ./formatdb.o ./formatdb.o </pre> /usr/bin/mpi{cc,CC}.{lam,mpich} are alternatives, that get linked to /usr/bin/mpi{cc,CC}. Surely they should have the same behavour? <br><br> This is currently causing problems, as I'm trying to make a nice tidy package of <a href="http://mpiblast.lanl.gov/">mpiblast</a> that has both support for <tt>mpich</tt> and <tt>lam</tt>. Currently, I'm very tempted to just ignore the <tt>mpich</tt> support.

January 09, 2004 01:31 AM

Frank Lichtenheld — finally some progress

Yeah, packages.d.o is finally getting there. Joey is testing my scripts
on gluck, so the pages should be back soon. After about six months
without progress this are really good news and this motivates me to
tacke the next big task: Improving the search. But this is an area
where I have to learn many things first, so help welcome.

I've searched for existing solutions, but don't found any yet that meet
all my requirements:


Suggestions, anyone? I warn you, I'm gonna write it myself ;)

(updated: corrected klecker to gluck in first paragraph)

January 09, 2004 01:16 AM

Security Announcements — DSA-419 phpgroupware - missing filename sanitising, SQL injection

The authors of phpgroupware, a web based groupware system written in PHP, discovered several vulnerabilities. The Common Vulnerabilities and Exposures project identifies the following problems:

January 09, 2004 12:00 AM

January 08, 2004

Junichi Uekawa — I've fixed my RSS generator date format, so it should now be okay

I've fixed my RSS generator date format, so it should now be okay.Yes. Yippee. Thanks to those who helped.It's just nice to be able to read diary entries of those that I've once met at Toronto, or Chicago, or New York, or wherever else... while I'm stuck in middle of Cold Winter Tokyo.

January 08, 2004 11:22 PM