Earlier today, I release NBD 2.9.7, containing a few fixes for segfaulting bugs, and some new features: one by Mike Snitzer to implement support for better handling of dead connections (which has been requested by a number of users for eons); one by myself to allow for running shell script snippets at connect/disconnect time, which should be helpful for the LTSP people (they now have an inetd-based wrapper written in shell, which seems like such a waste). Source can be found on sourceforge.net, packages have been uploaded to Debian and are even installed by now; these also fix some outstanding packaging-related bugs. Enjoy!

Eurosport coverage of Vuelta a Espa a at 19 east continues with commentary in English (analogue only) and German. Scheduled times:-

Mon 17.09.2007

(no show, rest day)

Tue 18.09.2007

16:00-17:30 CET

Wed 19.09.2007

09:00-09:30 and 16:00-17:30 CET

Thu 20.09.2007

15:00-15:45 and 15:45-17:30 CET

Fri 21.09.2007

16:15-17:30 CET

Sat 22.09.2007

16:00-17:45 CET

Sun 23.09.2007

17:15-18:00 CET

TVEi are broadcasting in Spanish at 13 and 19 east. Scheduled times:-

Mon 17.09.2007

(no show, rest day)

Tue 18.09.2007 to Sun 23.09.2007

16:40-18:00 CET

Enjoy the finale!

Just a quick reminder that the BBC’s only Tour of Britain highlights show is on BBC 2 at 12:30 today (Astra 2D at 28 east, the same satellite as Sky UK+IE, but no subscription required). Official listings say this is highlights of the final stage but hopefully it will cover the other stages too. Hopefully, live coverage will get better next year, either online as cyclingfans editor Pete predicts or on TV.

Trying to post the movies for Tour of Britain stage 2: Taunton left me in a whole world of pain. I'd decided to take movies rather than stills because there were lots of photographers there with considerably better cameras than mine, and my blog is far more popular for TV coverage than for photos. My video camera is a four-year-old Nisis Pocket DV4 which appears as a usb-storage device and contains AVI files. Apparently, those AVI files contain Video as mjpeg, yuvj422p, 320x240, 9.04 fps(r) and Audio as pcm_s16le, 11025 Hz, mono, 176 kb/s. I wanted to post the movies as Ogg Theora for various reasons, including how well it seemed to work for debconf this year. I'm a total clubie at this video stuff, but I like freedom and I've seen what I want, so I know it's possible. First of all, I tried mplayer's mencoder. It seems that mplayer can play Ogg Theora, but mencoder can't encode it. Did I miss something? What's with that? So, I tried installing VLC because several people had recommended it to me in the past. It seemed to have an easier interface than mplayer and had a wizard for converting files. Unfortunately, it seemed to crash on some of the AVI files and other times hang. When I upgraded from 0.8.6 to 0.8.6c (which also meant a wxWindows upgrade, which surprised me for such a seemingly-small upgrade), it crashed out with a segfault rather than hanging. I couldn't get the command-line to work either. I tried ffmpeg, but damned if I can get that to produce a good theora (or is it libtheora?) file. Sometimes it works, but the output is much larger than I expected. It is very annoying that there are no releases, but at least there are snapshot tarballs available. I was trying export-2007-09-12 early this morning. Finally, I tried ffmpeg2theora and (after upgrading ffmpeg), the command

"ffmpeg2theora -aq-1 -v 0 -o outfile.ogg infile.avi"

produced useful results, which I've now uploaded. Thanks to BBLUG members for their comments. There seemed to be others who shared my confusion and frustration with video processing. Why is this still so hard? Is it, as one member suggested, the chilling effect of software patents and so on?

As mentioned yesterday, I went to watch the Tour of Britain for real in Taunton. It’s taken me a while to encode these (I’ll write about the technical details on another page soon), but here are the videos. They’re in the super-cool Ogg Theora format, in case that helps you update your media player with filters like these. (I probably should do some clever tricks to embed these videos in the page, but it’s taken me long enough to get this far, so please either download the files or drag-and-drop the links to your media player. If anyone can explain how to embed videos simply, please leave me a comment…) First of all, I went to Staplegrove Road, where I thought the peloton might slow slightly to get past a roundabout on their first pass through Taunton (Ogg, 514K) a bit after 11am, but they didn’t really. Then I got a brilliant position in the second row just by the finish line, which meant I didn’t see the big crash of the day until I saw the TV news in the evening, but I did see both the main sprints, for the stage win (Ogg 919K) and the yellow-jersey group (Ogg 382K). Finally, I went and watched the presentations - the yellow jersey (Ogg 1.3M) and the champagne spray (Ogg 643K) are shown here. The BBC has the broadcast rights to the tour of Britain and seem to be doing a terrible job: Live race news seems to be limited to a quick mention in the half-hourly sports bulletins on Radio Five (on satellite Astra 2D at 28 east to much of north-west Europe, or Medium Wave 909 and 693kHz). Each night, there’s a short (3 minutes?) summary report in the BBC1 TV news of that stage’s region, some time after 18:30 BST (so tonight in West Midlands, tomorrow Yorkshire and North Midlands, Friday North-west and Saturday Scotland, all Astra 2D). This is not shown on BBC News 24, as far as I’ve seen, not even in the Sportsday show. There will be one national highlights show on BBC 2 at 14:30 on Sunday (again, Astra 2D), the day after the race has finished. Official listings say this is highlights of the final stage but hopefully it will cover the other stages too. I feel they should take the rights away from the BBC and give ITV a chance. We had better coverage of this year’s Tour of Ireland (from itv4) than we have of the Tour of Britain - this is silly. For non-TV coverage details, including Twitter, see CyclingFans.com

I think I’ve just invented a new sport… Everyone knows the game of Googling your own name, and finding out all sorts of fascinating things that you have forgotten or didn’t realise had made it onto the Interwobble. Here’s a new twist, use the Internet Archive Wayback Machine instead and read through things like old versions of your homepage. I’m honestly shocked at just how much incriminating material I’ve found, both from my own website and segfault.org This comment on my site from Leonard Richardson and my reply especially amused:

What percentage of the time are you drunk? This is to settle a bet. Okay, well on an average week I guess I go out for 3 nights. Friday night starts at 5:30 and ends 12 hours later, the other two days would be probably from 9:30 to 2:30am, so 5 hours. So we can say I spend 22 hours innebriated. Just under a day, or 13% of the week. Some weeks that might be as much as 25% to 30% I guess tho. Who was the bet with, and more importantly, who won?

This explains a lot …

With all the excitement of the Vuelta and the Tour of Britain (which I’m off to watch today), I’d missed that the Tour de Pologne also started on Sunday. TVP Polonia has coverage and broadcasts free-to-air to Europe at 19, 13 and 7 east. I think it may also be visible in Australia in a GlobeCast package (can you see it? Please tell me by leaving a comment.) Scheduled shows:-

Tue 11.09.2007 to Fri 14.09.2007

15:10-16:30 CET (live) and 20:00-20:05 CET (summary)

Sat 15.09.2007

13:25-15:05 CET (live, final stage) and 20:00-20:05 CET (summary)

I can’t actually watch this at 19 east except in perfect conditions at the moment because my dish is slightly mis-aligned. I don’t want to adjust it until after the Vuelta finishes, just in case I make a mistake and lose reception completely.

Eurosport coverage of Vuelta a Espa a at 19 east continues with commentary in English (analogue only) and German. Scheduled times:-

Mon 10.09.2007

16:00-17:30 CET

Tue 11.09.2007

no show (rest day, but MTB coverage at 08:45)

Wed 12.09.2007

16:00-17:30 CET

Thu 13.09.2007

01:00-01:30 (recap?) and 15:45-17:30 CET

Fri 14.09.2007 and Sat 15.09.2007

16:00-17:30 CET

Sun 16.09.2007

18:00-19:00 CET (recorded?)

TVEi are broadcasting in Spanish at 13 and 19 east. You can also see the race on TVEI America, but I don’t know the reception details. Scheduled times:-

Mon 10.09.2007

16:30-18:00 CET

Tue 11.09.2007

no show (rest day)

Wed 12.09.2007 to Sun 16.09.2007

16:40-18:00 CET

I’ll post more times next week. It was good to see a strong Caisse d’Epargne showing, as I predicted, even if it didn’t last long…

Now my fvwm is segfaulting almost at each startup. Seriously, there's great software everywhere, but still no good window manager (I know, I tested almost them all). Life sucks.

You know Andrew, catchsegv(1) does exactly that :)

I should really read the GNU C Library documentation. Even if only 10% of it sinks in, I'll be that much the wiser for it. Today's (re)discovery: libSegfault.so I actually found this with a co-worker a few months ago, and promptly forgot all about it until he reminded me today whilst I was lamenting about Cfengine segfaulting on me. Here's a lovely self-illustrating example:

apollock@icarus:~$ sudo chmod a+x /lib/libSegFault.so
apollock@icarus:~$ /lib/libSegFault.so
Segmentation fault
apollock@icarus:~$ # Aww
apollock@icarus:~$ LD_PRELOAD=/lib/libSegFault.so /lib/libSegFault.so
*** Segmentation fault
Register dump:
 EAX: 80003f00   EBX: b7ff1ff4   ECX: 00000000   EDX: 00000000
 ESI: bf88b97c   EDI: 800008b0   EBP: 00000000   ESP: bf88b974
 EIP: 00000001   EFLAGS: 00010282
 CS: 0073   DS: 007b   ES: 007b   FS: 0000   GS: 0033   SS: 007b
 Trap: 0000000e   Error: 00000004   OldMask: 00000000
 ESP/signal: bf88b974   CR2: 00000001
Backtrace:
/lib/libSegFault.so[0xb7fd004f]
[0xffffe420]
Memory map:
80000000-80003000 r-xp 00000000 08:02 8589       /lib/libSegFault.so
80003000-80005000 rwxp 00002000 08:02 8589       /lib/libSegFault.so
80005000-80026000 rwxp 80005000 00:00 0          [heap]
b7e61000-b7e6b000 r-xp 00000000 08:02 6254       /lib/libgcc_s.so.1
b7e6b000-b7e6c000 rwxp 00009000 08:02 6254       /lib/libgcc_s.so.1
b7e6c000-b7e6d000 rwxp b7e6c000 00:00 0
b7e6d000-b7faf000 r-xp 00000000 08:02 8572       /lib/libc-2.6.so
b7faf000-b7fb0000 r-xp 00141000 08:02 8572       /lib/libc-2.6.so
b7fb0000-b7fb2000 rwxp 00142000 08:02 8572       /lib/libc-2.6.so
b7fb2000-b7fb5000 rwxp b7fb2000 00:00 0
b7fce000-b7fd1000 r-xp 00000000 08:02 8589       /lib/libSegFault.so
b7fd1000-b7fd3000 rwxp 00002000 08:02 8589       /lib/libSegFault.so
b7fd3000-b7fd5000 rwxp b7fd3000 00:00 0
b7fd5000-b7ff1000 r-xp 00000000 08:02 8569       /lib/ld-2.6.so
b7ff1000-b7ff3000 rwxp 0001b000 08:02 8569       /lib/ld-2.6.so
bf879000-bf88f000 rw-p bf879000 00:00 0          [stack]
ffffe000-fffff000 r-xp 00000000 00:00 0          [vdso]
Segmentation fault

This is somewhat similar to Tridge's segv_handler, which used to be what I reached for in these sort of situations, but that was a bit fiddly to get set up. This is significantly easier, and already installed as standard. Of course, interpreting the output is left as an exercise for the reader. Update Or you can just use catchsegv(1)

Last month the new thread I was afraid about has started. Called Considerations for 'xmms' removal from Debian, the maintainers of the xmms package proposed the removal of it.

I used xmms (and, before, x11amp) for something like 9 or 10 years. I know, I love it. I can't bear other players. Amarok is heavy, rhythmbox seems like a joke to me. I like to load my 6500 files playlist in one shot, press 'j' and jump to the song I want. I don't see how it could be easier to use a music player.

Following the thread, I must admit that, yes, xmms is old, gtk1.2, unmaintained, etc. I had to switch, sooner is better.

So I use now Audacious. At least I try to. Last time I tested it, the search (jump, 'j' key) function was not working. No it is, it works like xmms, and I'm happy. I know that upstream author of audacious dislikes being compared to xmms, but, screw you, you're like xmms, and I like it, so I like audacious for this. The plugins and the interface is like xmms, so it's great also.

However it's buggyyyyyyy. Third day of use, serious bug reported as #436557. It segfaults sometimes, or simply blocks reading a file.

The information loading 'on demand' is so crapy compared to xmms. I mean, it's fucking slow! When I put shuffle on and I lick 'next song please', then it hangs reading informations. xmms was doing what in a better way.

However, dear audacious-author-that-already-hates-me-for-comparing-xmms-and-audacious, I really think it'll be a real alternative for old xmms users like me and a lot I know (yes, I know a lot of people, damn it!). Developement seems to be very active, so I'll continue to report bug and use it.

Yes, I believe in audacious.

It's good if you bring out a maintenance release of a certain specific software package that you maintain, if it segfaults on startup when a new feature is used. It's pretty silly if that new version still segfaults if and only if that very same feature is used. What do you mean, testing? Anyway, today I released NBD 2.9.6 that fixes those issues. And I extended the tests that are ran as part of 'make distcheck' so that it won't be segfaulting on startup anymore next time I do a release. Silly me.

Today I was checking how much diskspace is need when a small CLI (.NET) application like gfax is installed. The last 2 years I am optimizing the Mono packages in Debian to reduce dependency chains. Dependency chains can cause a small application to become very big (at least it looks that way) when they are installed, as modern Linux distributions automaticly install all required dependencies. So first I worked against GUI dependencies for non-GUI applications, those are annoying for servers, you don't want to install a X server or other GUI libraries when you only serve webpages using ASP.NET.
Splitting CLI 1.0/1.1 vs 2.0 also seemed to be a very good candidate to split apart, as applications usually are target at 1.0/1.1 or 2.0.
Then I split Base-Class-Libraries that have dependencies on native libraries (via P/Invoke), as those packages must depend on those native libraries, like DB drivers.
Last but not least I split debug symbols in a different package, as those are not needed for normal production usage, and embedded systems don't have much diskspace. The mono-dbg package is 28.2MB in size when installed, that's alot.
The Mono source package in Debian is now building 67 binary packages, but thanks to APT nobody has to care for that, every part is installed when needed. So today I was checking the results of it, how much will a small application download and install?
I am testing with a clean minimal Debian system, that way all dependency chains will show up very easily.
Installing GFax + required libraries today takes 43MB to download and 138MB on the harddisk. GFax is a small GNOME application, it uses libraries like: Mono, GTK#, GNOME#, Evolution#, GConf#.
So someone might say: "of course, it is based on that bloaty Mono virtual machine!", well let's check a different application.
What about a GTK# only application (no GNOME) like graphmonkey + required libraries? That will download 19.5MB and uses 56.4MB on the harddisk. That's alot less, isn't it? graphmonkey only uses Mono and GTK#.
So it's not Mono taking all no diskspace? Well after all my packaging optimizations, it's not Mono
So how big is a install of a minimal Mono runtime + all required libraries then?
Minimal as in enough to run famous the "Hello World".
It's 2.4MB to download and 7MB on the harddisk, yes you read it correctly, the super monster bloaty virtual machine just needs 7MB. Mono only has glib as dependency which will be replaced by the eglib library some day. After I had this great result, I wondered how big other runtimes are.... like.... Java and Python? The results are interesting...
A minimal Python install takes 3.9MB to download and 13.4MB on the harddisk. Yes, that's almost double the size of Mono, I was surprised too.
So and Java? Here it becomes a bit unfair as SUN's license is not allowing to ship parts of a Java runtime. So the Java package in a distribution is one package. 34.5MB to download and 95.2MB on the harddisk. Yes, that's alot, but thats the smallest install you can get when you want to run anything on it. So, if you want to run Hello World on a virtual machine: use Mono! just kidding...
Small runtimes are important when it comes to acceptance and adaption of new technologies, as nobody wants to kill all his computer resources for it. And I often hear the replies on Linux IRC channels, when people ask which software they can use to do $task: "No I don't want o install 100MB just for $task" and they decide to try some other software with less dependencies and less resource usage.
Also for the embedded area, resources are very limited, every MB still counts there. For the completeness of my report, here the naked numbers of other available Java runtimes: sablevm 22.0MB (63.2MB), kaffe 52.2MB (118MB), jamvm 28.3MB (65.4MB), cocoa 28.8MB (66.0MB).

I asked for any other free-to-air satellite coverage of the Tour de France and Bernie S commented:

"RAI 3 on Hotbird"

and there's an online schedule on the RAI 3 site. Neville Emerson wrote:

"Can't wait. Cycling is the best and coverage is getting better and better every year."

Yes, at long last. There's now a full itv coverage listing on the CTC site but I still don't understand why itv are so late at launching their tdf site! Speckled Hen asks:

"Anyone know if there's any live coverage online? Can't see anything on either Eurosport or ITV web sites. I know the Beeb usually do it with rugby internationals."

I heard of some recorded items from ASO (Amury Sport Organisation, the tour's managers) on YouTube, but nothing about live video feeds yet. I've had audio in the past from both Radio France and Eurosport but I've not even found those details for this year yet (I'm slow at reading French and the new Yahoo Eurosport site is terribly hard to navigate). Keep watching Cycling Fans and TdFBlog (who kindly linked this item in their coverage summary ) for news of video coverage. Unless anyone would like to leave me details of live online coverage in a comment on my site, please?

This one was a bit tricker to find, so I'm going to post it here: if you use os.fork() on Python, you will want to use os._exit() on the child process, or it will clean stuff it shouldn't and you will have segfaults with doubled memory free.

That said, python's webbrowser module is quite useful; check it if you ever has to open a browser from a Python script/program. The only downside I see on it is exactly that it doesn't takes care of the forking stuff for you. I'm using it on bug-triage to show a bug report on the user browser.

Talking about GSoC, during the last week the Debian BTS was updated, breaking, between other things, btsutils' 0.1 version. The current development version of btsutils, which uses BeautifulSoup to parse the HTML stuff more realiable, wasn't affected by the update.

That's all for now, time to code.

I just failed to migrate a big project at work because of that, the D day:

Jun 15 10:46:39 hostname mysqld[20095]: Version:'5.0.32-Debian_7etch1-log'  socket: '/var/run/mysqld/mysqld.sock'  port: 3306  Debian etch distribution
Jun 15 10:46:43 hostname mysqld[20095]: mysqld got signal 11

So no, you won't use NDB, at least with MySQL 5.0.31 (included in etch). It's _so_ buggy that we get a segfault on the first DELETE statement with a 30 MB database. Backporting version 5.0.41 from sid gives better results (at least for now). I knew we should have used PostgreSQL from the beginning...

Hopefully fuzz testing is well known nowadays, right? But did you check how smart your software can handle ENOSPC - the ‘No space left on device’ error? I like stress testing software (to be honest: no, not really). Yesterday I debugged a problem regarding zsh history writing on ENOSPC. Usually I use the grml Linux live-cd for stress testing (often running inside Qemu, Virtbox, VMware,…) like in:

dd if=/dev/zero of=/tmp/home.lo count=1000
losetup /dev/loop1 /tmp/home.lo
mkfs.ext3 /dev/loop1
mount /dev/loop1 /home
mkdir /home/grml
chown grml: /home/grml

Now after restarting the shell session of user grml I can play with ‘No space left on device’ situations. (Notice: I adjust the count=… according to what type of application I want to test.) Too many applications don’t really cope with the ENOSPC situation. Firefox for example is very funny, depending on the situation it either just segfaults when trying to start up, or decides to not start up at all and just exits with ‘1′. See the relevant strace output:

mkdir("/home", 0700)                    = -1 EEXIST (File exists)
access("/home", F_OK)                   = 0
mkdir("/home/grml", 0700)               = -1 EEXIST (File exists)
access("/home/grml", F_OK)              = 0
mkdir("/home/grml/.mozilla", 0700)      = -1 ENOSPC (No space left on device)
access("/home/grml/.mozilla", F_OK)     = -1 ENOENT (No such file or directory)
exit_group(1)                           = ?

Opera also just exits when already running and receiving an ENOSPC - I encountered data loss (for example the transfer history got lost) more than once. Very "funny" to see how applications like dillo, joe, wireshark,… behave in the ENOSPC situation. Not. Oh, your application handles ENOSPC just fine? Then continue testing with read-only filesystem, using NFS, removing files randomly,…

At work, I’ve got two screens on my Sarge workstation “snitch”. Since I want to switch virtual desktops independently on both screens, I don’t have a Xinerama setup but a Dual Screen setup. So my left and right screen do have different $DISPLAY (“:0.0” and “:0.1”) set. This is neither a problem for FVWM nor xlock nor XScreenSaver. But it is a problem for nearly every modern web browser available which checks, if there’s already an instance of it running. So if you try to start a new instance of a web browser on the other screen, most graphical web browsers make more or less problems:

• Galeon 1.3 and Epiphany always opens new tabs or windows on the display where its first instance is running, i.e. ignores $DISPLAY completely except on the first call.
• Kazehakase (0.3.7) just opens a new tab in the running instance.
• Firefox 2.0 thinks it crashed and asks if it should restore tabs and windows. Haven’t tried any further.
• Opera 9.20 pops up a dialog, says, there seems already a copy of Opera running and asks if it should continue with startup. If you say yes, only the bookmarks of one of the two instances get saved, probably those of the one with the last added bookmark or the one which exited last.

The only graphical web browsers which simply just work on a Dual Screen setup are Konqueror, Links2 (called with the -g option for a GUI), Chimera 2, Amaya and of course Dillo. Unfortunately I’m neither a fan of KDE nor of Konqueror and I do want a web browser with CSS and tab support… And Amaya is, well, only a reference implementation… (Chimera 2 from Sarge btw. segfaulted on two of the four pages I tested it with. Seems to have problems with PNG images.) So my current setup is to have Kazehakase as my main work web browser (with all the local web applications I need) on the right screen while I have Opera on the left screen for surfing, looking up documentation, testing web pages and other things. BTW: I don’t use Gecko based browsers for surfing on that box at the moment, since there are some web pages (the spammer vandalised Kazehakase wiki for example, at least a few months ago) which manage to be rendered in such an ugly way by Gecko so that XFree86 with the binary Nvidia (at least the last five or six versions I tried) just crashes away — either at once or when you try to switch to a text console by pressing e.g. Ctrl-Alt-F1 while such a page is displayed.

Last week I started the final round of Debian upgrades for the servers I maintain here and there, which is mostly complete today. I haven't been so lucky with upgrades this time, for a long list of different reasons. In the end, the smoothest upgrades were those boxes I upgraded when etch froze or so. natura.oskuro.net is the box serving these pages. It's an old, extremely noisy Pentium 150 which I've been intending to replace for a while now. I started the upgrade early on Thursday, knowing it'd take a while (natura takes its time only to read the Dpkg database), and it had apparently finished when I was ready to leave the office. Three issues:

• apache2's enabled modules were forgotten and I had to reenable them again, plus IIRC the default "It works!" site got reenabled, hiding the blog for a little while.
• bind didn't like some change in one of the default configs and I failed to noticed it had not started. After the reboot, the box was unreachable remotely, but that was OK as I had to go visit my dad that night for his birthday, so it was fixed after one hour or so.
• Last, my blog stopped working, and while I thought it'd have something to do with apache2.2, I had no time to have a close look as I had to leave to Lleida this weekend. Yesterday I discovered that for some reason, the old tar.gz PyBlosxom install wasn't liking the new python2.4, so I decided it was about time to move to PyBlosxom 1.3, from the Debian package. Not without the mandatory "It works!" post published in the gazillion Planets, natura was ready to go.

The very same night of last Thursday, I decided to dist-upgrade the box which serves the Spanish Debian website mirror. That's the only purpose on the box, so you can imagine the upgrade should have been pretty straight-forward. And so it seemed, until, in the middle of unpacking, dpkg died with a horrible I/O error, and I dropped into an unusable remote terminal with no working commands. Fortunately, apache2 was still up and running, and the web service has been working without interruption since the hard drive crash, albeit with no syncs from www-master. Today, Sergio visited the campus and had a look. It was a XFS crash, which got cleanly repaired using an install CD. We have an empty partition in the box, and will probably move the system to it temporarily, and back to the RAID, but on ext3. When the box was back online, I just had to resume the upgrade process, make mdadm happy and update lilo.conf before rebooting into the new kernel. This box uses LILO for some obscure reason I can't remember too clearly anymore. The box has just one partition on a md array, on two SCSI disks on a aic7xxx-based controller. Can anyone hint me why GRUB would have failed on us back in sarge, and if any fixes in the etch version would work any better? Using LILO here is error prone, and basically feels like a step back. Anyway, www.es.debian.org is now back up and running with updated content. Sindominio.net had its bi-annual upgrading party last monday, but unfortunately I wasn't able to help much as when I tried to log into the server, I must have caught the system in the middle of some key lib upgrade or something, and again I was locked in a unusable shell which would only segfault. Given my previous experience, I assumed that something had gone wrong and the box would need to be fixed at the console, and after 20 minutes I gave up helping on that front. Until I noticed, quite a long while later that I was still getting mail from the server. I managed to log in to discover the upgrade was done, with just a few bits remaining to be done. The major issues were encountered with our pam and ldap setup, plus nscd kept dying causing quite a lot of mayhem all over the place. Great work from Seajob, Syvic, nogates, apardo and the rest of the people who handled it! With etch, we can finally move back to an official Debian kernel, something we've been longing to do for a long time. The only pending upgrade issue is that we need to move from our old jabber server to either the traditional jabberd 1.x or ejabberd; our current implementation is no longer supported in Debian. The last of the etch upgrades stories involves Sofcatal 's servers. The box was running on a CentOS 4.4, which was moved away into a subdir just after booting Debian-Installer, and then lobotomised so it would run as a Linux-VServer under a new Debian etch install. I'll probably write more details about it soon though, as it could be a maybe less scary alternative to Guillem's debtakeover. Yay for etch!