Older blog entries for mbanck (starting at number 14)

Being busy with university, I haven't been doing much for Debian or Free Software in general lately. The most notable thing was that I manged to get 3D acceleration to work on my OmniBook, thanks to Michel Dänzer's mach64 DRI packages. The speedup is quite noticeable, albeit not ground-breaking - at least now I can use pymol and Molekel without feeling like watching a slide-show. This episode also taught me that non-free software indeed is good for something, because it was struggling to get my roommates' Radeon card (or rather, linking the drm module for it) to work in 3D which actually made me wonder whether my notebook's graphics adapter would be able to do so as well. Other than that, I mostly broke jack's dependencies.

I am currently pretending to organize a small hacking event in Munich next weekend. It's going to be held along the next official Debian BSP and Frank Lichtenheld is going to come over to Munich to coordinate both of them. Further, Sven Luther will hopefully be around, as working on the powerpc port of d-i is planned as well. So, d-i hackers (or Debian people in general) don't be shy and come to Munich! I hope that a considerable amount of the Munich Debian guys will be around, in order to make that event fun to attend.

On some misc notes, I found out I'm living one subway station away from Murray Cumming and after looking at some of the other blogs in my quarter, of course I'm totally positive I've seen this girl somewhere before. Also, after having read Jeff Waugh's biography, I promptly logged into orkut again and finally became his fan on an official basis as well. In any case, OSCON looks like a wonderful place to be, if only it would be in Europe. Further, I appreciated that Ross Burton is listening to excellent music, although I was dissapointed to find out yesterday that our former, present and future DPL does not know about Eric B. & Rakim:

02:36 <azeem> warte bis -3:40 oder so, dann kommt Eric B. & Rakim :)
02:36 <tbm> kenn ich nicht

Last saturday I went to Linux-Infotag Augsburg together with Jens Schmalzing. For some reason, they thought installation presentations would still be en vogue and had people give presentations on installing Knoppix, Fedora, SuSE and Mandrake there. Jens decided that showing off an ordinary Debian woody installation was not worth the time and went for an FAI (Fully Automatic Installation) instead. He first explained stuff and then fired off FAI with one touch from his notebook, waking up, booting and installing the target box over LAN, without ever touching the box once. I guess people were quite impressed because he told me yesterday that the guys from the Augsburger LUG asked him to give the presentation again...

Unfortunately, he was talking quite a lot and it took him almost one hour (he thought it would take him about 30 minutes) to finish the presentation which left me with no time to show off debian-installer, as we orginally planned. Actually, we spent most of friday evening and some time on saturday trying to figure out how to boot and install d-i with a USB keystick (which was quite easy) and installing stuff over WLAN (which did not work due to the lack of an access point. We could get neither my prism2 nor his airport card to work properly with hostap). In the end we did a private presentation for two interested people in a different room using the sarge mirror on his powerbook, but driving to Augsburg still felt rather like a waste of time afterwards, as the event was quite small and there was not much else to see.

On sunday, I finally uploaded xdrawchem, my last major package lacking behind and it looks like ghemical will enter sarge tomorrow.

Yesterday, I met a couple of other Debian developers from Munich again in the Augustinerkeller, along with a couple of other interested guys. It seems we're trying to organize a hacking/bug-squashing party in Munich on April 17th/18th if there is enough interest. People from out of Munich can sleep at my place which is just around the corner of the mathematics department, where the session will be taking place, so I hope at least a couple of people from the rest of southern germany or austria will come around.

I've also played a bit with bitlbee, which looks really cool. Unfortunately, I was having issues with my SOCKS setup again, I'll have to sort that out. Further, I finally managed to change the key-bindings of ion2 to my likings (i.e. the way I had them configured back when I used ion), so I'll probably be using ion even more than before. I still have to check out Gnome-2.6 and see whether that makes me switch back to a fancy desktop environment though.

I'm a bit worried about the dedication to get sarge out of the door. There was a Bug-Squashing Party over the weekend which I was remotely attending, doing one sponsored upload, fixing Exult and trying to be helpful to others. Still, Frank Lichtenheld credited me for this small amount of work, which rather says it all about the attendence of the other Developers.

Most of the time during the weekend I tried to build glibc on hurd-i386. It turned out that there seem to be still some problems with Ogi's patch for ext2fs large stores, as I got quite some file system corruption and never managed to complete a build. Reverting to the hurd package from unstable and making a smaller partition made glibc compile fine, libc0.3-dev still needs to be fixed by Jeff Bailey though.

Over the last week, I got most of my Debian packages in shape for sarge. MPQC now built fine on all arches for the first time ever, and after fixing a bug in the test suite, a shared library of OpenBabel is ready to go into sarge now as well. This made it possible to get a ghemical package linked dynamically against OpenBabel and MPQC into unstable, which will replace the severly outdated version in sarge soon. I also fixed PyMOL, but it's still stuck in Dep-Wait on hppa due to a mistake with Build-Depends and despite a mail to debian-hppa. The last outstanding package now is XDrawChem and Warren Stramiello is fine with me uploading it.

Yesterday, I hacked sbuild to properly build source packages when requested.

Last week, I managed to successfully bootstrap a Debian GNU/Hurd system from GNU/Linux without having to natively setup the Hurd translators. The latest version of Roland McGrath's ext2-xattr patch seems to work fine now and made this possible. Other exciting news in this area is the apparent maturity of Ognyan Kulev's ext2 patch for large stores. Seems one of the major drawbacks of the GNU Hurd, namely the 2GB partition limit, is mostly resolved now.

Over the last weekend, I finally backported a couple of my chemistry-related packages, as one of the admins of my research group at the university seemed amenable to install them on our group server. Packages so far include chemtool, openbabel and xmakemol. I was not able to backport xdrawchem yet, as it seems to Build-Depend on Qt-3.2, even though upstream claims any version of Qt3 should do.

On Monday, I finally decided to do something about the allegations against the non-free removal and proposed a transition plan for the non-free packages to nonfree.org. No direct offers for help have emerged from the posting yet, but during the draft phase we've already found a host featuring a full-blown katie installation, which covers one of the biggest concerns already.

At the university, I finally had to code a bit for my master thesis. Sadly, it only involved adopting the i/o routines of a Fortran77 program to our needs. Very depressing job.

Over the weekend, I merged the OpenBabel changes from the openbabel-1-100-x branch back into CVS HEAD. Thanks again to Colin Watson for letting me know about cvsps, which made this a rather pleasant task.

Regarding my Debian packages, I've finally pulled myself together and uploaded a jack-3.0.0 package for experimental today. Also, I've fixed the RC bug on pymol.

On the Hurd front, stuff seems to get rolling even more lately. Marcus Brinkmann got a new box and installed Debian GNU/Hurd on it over the weekend. He then realized the core packages are hopelessly out of date and made the first uploads of gnumach and hurd packages since well over a year.

Even better, he resurrected hurd-devel and got approval for his old libihash rewrite. He also managed to convince Thomas Bushnell, BSG on how to proceed with fatfs. Whether it is more astonishing that he actually managed to convince Thomas on this matter after all this time, or rather that Thomas replied at all (and even so promptly) is unclear. I was always under the impression that he was not MIA, just that nothing important enough for him to step in had happened concerning the Hurd over the last year.

Yesterday, I tried to build a debug package of star to test Roland's patch for Hurd translator support in Linux' ext2 file system. It turned out that during building a star Debian package, ./configure gets invoked 17(sic!) times. I hope this is a bug in the Debian package, because surely smake is superior to GNU make, right? Further, smake does not seem to honor CFLAGS, I had to modify a read-only rules file in order to add -g, because the upstream maintainer of both star and smake seems to knows best how to compile his code I guess.

I also built a kernel-patch-xattr-hurd package containing Roland's patch, and an accompanying kernel-image package, in order to facilitate testing. Next on the list is a new Hurd package with Ognyan Kulev's patch for support of larger than 2GB partitions in ext2 included.

I finally managed to get Linux-2.6 fully working on my notebook (at least so it seems), after I compiled a driver for my touchpad, which made X happy again.

Mad props to Tatsuya Ishida for making a special elmo edition of Sinfest yesterday.

Like a couple of other Debian people, I've been to FOSDEM last weekend. Dogi, weasel, CHS, flo, erich and I went there by car, which took us about eight hours. We arrived on Friday evening at mind.be in Leuven (where we again stayed for the weekend, thanks to the wonderful hospitality of Peter Vandenabeele and Peter de Schrijver). We then met a lot of other hackers at a pub at the Grand Place in Bruxelles where we had a few beers and a lot of conversation about recent Debian events. Driving into and out of Bruxelles was only slightly easier than last year, it's still mostly impossible to not at least get lost once.

During FOSDEM, I attended Tim O'Reilley's keynote (I've already seen RMS and John 'Maddog' Hall last year, so I skipped those) and listened to the very interesting talks by Robert Love and Keith Packard. On Saturday afternoon, I was at the Debian booth and displayed my notebook running Debian GNU/Hurd. Quite a few people stopped by and were interested in it, at least more as for Debian in general. On Sunday afternoon, I met Marco Gerards and Jereon Dekkers and we had a nice conversation in the hacker's room. I also got a bit of Debian hacking done, mostly on pymol and jack and also managed to finally contact my new NM. I guess I'm getting old because I got only about a dozen of new fingerprints as I knew most of the people already (I also preferred to go to Keith's talk than to the keysigning party, which was almost as poorly organized as last year).

The only time I really ate something notable was Saturday evening during the FOSDEM dinner, which was quite a lot though. And very good, too. We had a good time there until some people decided they would be better off with Internet access at mind.be rather than at parties. But destiny had a different plan for them: Pretty much as soon as we arrived at Leuven, electricity went out and thus the network. This way, I had a nice chat with Dogi and went to bed fairly early, which helped for the ride home.

To summarize: FOSDEM rocked.

foo.

On Friday, my first NM applicant Raphael Goulais got approved. Actually, the DAM approved him faster than I could accept a new NM, which seems to be contrary to most conspiracy theories I've heard recently.

During the last couple of days, I mainly focused on the upcoming 1.100.2 release of OpenBabel and made sure that our switch to shared libraries works nice with the other Debian packages who depend on it.

Today, I spend a lot of time in Debian GNU/Hurd. To my pleasant surprise, it seems that tsocks works under Debian GNU/Hurd as well. Although being behind a SOCKS proxy, I've now got most of my networking working on GNU/Hurd, most notably ssh, irc and mail.

In somewhat related news, Roland McGrath has written support for Hurd extensions in Linux-2.6 via the xattr interface. Meanwhile, Marco Gerards is working on support for Hurd translators in GNU tar, so these two things combined could make for a much easier way to bootstrap Debian GNU/Hurd from GNU/Linux. This way, the passive translators could just be extracted along with the rest of the base tarball and would not have to be setup after booting into the Hurd.

I've checked out rhythmbox and ifplugd during the last week, and they rock.

I had to work with OpenOffice.org for the first time for a presentation at the uni (it's a team effort, so I was not able to force magicpoint on the others yet). After some reorganizing of my harddisk last weekend, I managed to free enough space in order to install OpenOffice.org from unstable on my notebook, while at the uni, the admin installed vanilla OOo-1.1. Boy, does that make a difference, the Ximian patches sure look nice. Kudos to the Debian OOo-team.

While the presentation I currently work on is based on an older one which was hopeless in terms of visual consistency, this article seems to indicate that OpenOffice Impress could be an alternative to Magicpoint for more involved presentations. I managed to get unattended, looped slide-shows working in OpenOffice yesterday, so making a generic Debian presentation which could be run unattended at booths during trade shows would be nice to have I think. That's actually something I've been thinking about for a long time now, but so far I had no luck getting it to work in Magicpoint.

I also managed to get my notebook's 3c556 NIC working on GNU Mach-1.x. Now I have to find out why SSH currently does not prompt me for a password when in X on Debian GNU/Hurd.