Alternate use of CompactFlash cards

Leaving DebConf5, I went to Bratislava to join the ensemble on our concert trip. There, three people complained that they were having problems with their digital camera's storage media, so I started looking for a computer shop to buy my a flash media reader to be able to help these people out. As such, I now own a device that can read CompactFlash, Memory Stick, SD/MMC, and SmartMedia through USB2.0. As a Mass Storage thing.

What I don't have (anymore, at least) is a USB stick that I can use to store my GPG key on. The last one I did have broke down a few months ago, and I didn't buy myself a new one yet.

Since we have a stock of 32MB CF cards at the office (for use root filesystem in soekris devices which we sometimes sell), I took one of those, put it in the combined reader, and use that as my "USB stick". Seems to work. And is a bit more flexible, too.

ICMP does matter!

I swear, some day I'm going to ...

<breathe slowly>

Okay. There are some firewall admins on this planet who'd better not get too close to me. ICMP is not a joke — it's an important diagnostic tool. Consider this:

wouter@country:~/debian/webwml$ cvs up -dP
gluck.debian.org: No route to host
cvs [update aborted]: end of file from server (consult above messages if any)

wouter@country:~/debian/webwml$ ping gluck.debian.org
PING gluck.debian.org (195.25.206.10) 56(84) bytes of data.
64 bytes from gluck.debian.org (195.25.206.10): icmp_seq=1 ttl=43 time=183 ms
64 bytes from gluck.debian.org (195.25.206.10): icmp_seq=1 ttl=43 time=164 ms

--- gluck.debian.org ping statistics ---
2 packets transmitted, 2 received, 0% packet loss, time 1000ms
rtt min/avg/max/mdev = 164.841/174.110/183.380/9.278 ms

See?

Then I remembered that I'd had this before, and that back then, ethereal showed me that the problem was of an entirely different kind:

wouter@country:~/debian/webwml$ CVS_RSH=ssh cvs up -dP
Password:

Right. Read the RFC, you <censored>: if you want to firewall away a port, you don't send an ICMP host unreachable; rather, you send an ICMP port unreachable.

There. Feeling much better now.

Now why do I have rsh installed on my system? Hm.

wouter@country:~/debian/webwml$ sudo apt-get --purge remove rsh-client

Another 123kB saved on my hard disk. As if that would matter...

Axis sucks

I've got this Axis 560 printserver which I was allowed to take home when I still worked at DHL's global headquarters (then in Diegem, now being relocated to somewhere in Germany).

Only the bloody thing gives me headaches. About one in two times I try to print, it doesn't work. If I then check the printserver's web-interface, I'll usually see the job queued, with '0 bytes printed', or '1 byte printed' or so. Totally useless. Rebooting and/or powercycling the printserver and/or printer(s) usually helps, but not always.

Finally sick of this, I went to Axis' website and tried to find whether I could upgrade the device's firmware or something. Ghaah.

1. You can't download firmware from their support site unless you first register. I hate required registration, but oh well.
2. The registration form has the 'Company name' field as a required item. WTF? I'm not a company! I'm a home user with a network. Yes, axis, those exist!
3. The registration form has the 'Please spam me' field enabled by default. Nothing new or spectacular in itself, but see below.
4. After registration, I went on to download the firmware. Turns out there are two separate, incompatible, versions of this device, and I have the choice to download firmware for 'axis 560, version 1' or 'axis 560, version 2'. Of course, axis provides instructions on how to find out which version of the device you have. They say to look at the product number; if it is of the form '0053-xxx-01', you have version 1; if it is of the form '0053-xxx-02' you have version 2. My printserver's product number is '0053-<many spaces>10'. Now what?
5. On the page that lists the available firmware versions, there is a link 'Online Helpdesk'. Following that link gives me 'Page not found'.
6. On the 'Page not found' page, there's a link 'Write & ask us'. Following that link gives me a form which allows me to enter a message to them. Dunno whether that ends up with their support people, but anyway; I put a question in there regarding my problem. Hopefully they'll reply. However, below the <textarea> where I've entered my question, there's a section called 'Personal Information'. With exactly the same information they asked me to enter in their registration form, including the 'please spam me' checkbox. And yes, I'm still logged in.

That did it. Axis sucks!

FFII not broke

Earlier this week I wrote about FFII being down. Based on the (scarse) information available, I said that the FFII wasn't able to afford their legal costs anymore.

As it turns out, this isn't true. According to a recent article on ffii.de, the domain is down because of a hosting company that got scared away by a baseless threat letter from Nutzwerk lawyers.

FFII has moved to new hosting in the mean time, but because of the regular timeouts in the DNS, it can take up to three days before everything is up and running again all over the Internet.

Sorry for the misinformation I sent out earlier

Part of that information is still correct, however. Nutzwerk is a scam, and they should be ashamed of themselves for their unethical business practices — not just in conjunction with the FFII; their google spamming and expensive products that in essense are simply crap still exist.

Debian 12

In little over a week, on August 16, Debian will turn 12. Some Dutch and Belgian Debian folk are planning to meet in Breda for a beer and some general rejoicing. Not sure yet whether I'll be there myself (I might be too busy), but at least I'm going to try. I find 12 a pretty weird age to celebrate, but then every excuse is good enough to have a party, right?

If you're interested, then go ahead and read all about it.

I love swsusp

Yes, I do. It rocks.

I remember having used it back in the 2.6.3 days. Back then, it worked for me -- but only slightly so, and it was very slow, to the point that rebooting was often faster. It broke horribly in 2.6.4, and don't get me started on 2.6.5. The kernel developers apparently gave up at some point, I don't remember when.

But now, in 2.6.12, it's back, alive and kicking, and performs the pants off of what it looked like back in the old days. At least it allows me to have a (working!) suspend implementation on my powerbook, which, due to its nVidia-based monitor card, doesn't have a working suspend-to-ram implementation under Linux.

But with swsusp, I now have an uptime of 11:27:13 up 1 day, 10:14, 6 users, load average: 2.49, 6.64, 4.83 and counting. Woohoo.

Quickstep (finally) building for volatile.debian.net

This was long overdue, but quickstep is now finally and functionally building for volatile.debian.net. The first package, clamav was built and uploaded today.

In other news, I've noticed that gcc-snapshot has finally been built successfully on m68k. Good; it's far easier to track upstream bugs in that compiler (which I've started to do for a while, although I should increase the amount of my efforts there) if gcc-snapshot works and is actually more recent than the default version of GCC...

Grmbl.

ID# ATTRIBUTE_NAME          FLAG     VALUE WORST THRESH TYPE      UPDATED  WHEN_FAILED RAW_VALUE
  1 Raw_Read_Error_Rate     0x000b   064   064   060    Pre-fail  Always       -       224137572

This is my desktop's hard disk. I gather it won't be long before I'll start getting BACKUP YOUR DATA NOW messages in syslog. Guess I'll have to start looking for a new hard disk.

I hate it when that happens. And yes, the disk is dying; I hear it making all kinds of ugly noises it shouldn't be making, and my system is unresponsive while those noises happen.

And then I'm not even mentioning the mails I'm getting from smartd on my laptop, which tells me that there are some unreadable sectors over there. Well—at least SMART has been able to correct for that. For the time being...

Any hard disk manufacturers out there, let me tell you that I wouldn't mind at all buying a slightly smaller (measured in space available on it) and slower disk, provided it increases reliability by a factor two, or so. No, really. My 2GB Quantum Bigfoot that I bought about eight years ago only now begins to break down, whereas every other disk I've had has done so after three years, max. Pity they don't make no Bigfoots anymore.

Update: OK, next time don't post obscure outputs of commands without explaining what they mean. At least not if you don't want your mailbox to overflow an hour later

Look at the "VALUE" column. It says "64", whereas every disk ships with a value of "100" there. Once it gets under 60 ("THRESH"), the SMART bits of the hard disk will start panicking. And given that this value was at about 74 just a few weeks ago, this is bad. Very bad.

Bluetooth gone

A while ago, someone mailed me with problems getting bluetooth going on his iBook. He'd done a firmware upgrade from MuckOS, and now suddenly the bluetooth didn't work anymore under Linux. I couldn't help him—but a few days later, he mailed me back, saying that with a command called 'hid2hci' he could switch his bluetooth-hardware-with-newer-firmware to HCI mode, and that it then suddenly did work.

Of course, I'm on the train right now, want to connect to my cell phone over bluetooth so that I can dial in to somewhere, and find that the new logic board in my PowerBook (which they had to put there because the old one had simply broken down) 'conveniently' already contains this problematic firmware upgrade. So, no network. And no fixing that without network. Grmbl.

What's worse, 'apt-cache search hid2hci' renders nothing. I guess I'll have to go and google for that one. Hopefully it exists anywhere outside the mention of it in my mailbox.

Well. Guess I'll find out soon enough. But not as soon as I'd like. Oh well.

Update: after arriving at the office, a simple google search told me that it's actually part of bluez-utils. Which I obviously have installed. Aargh.

Tests suck

Philip Kern is surprised at the result of some blog test he's done, and claims There has to be something wrong with [it].

Well, duh. Newsflash: these tests are always wrong. I have yet to see the first one that is scientifically sound. Apart from being funny, they're not really interesting...

Device-mapper is yummy.

cryptsetup create secflash /dev/sda1
Enter passphrase:

There. I now have /dev/mapper/secflash that I can mkfs, mount, and whatnot. And everything is crypted.

Pity the docs don't tell me that so easily. I spent ~3h over the last few days to find out that all I needed is the above.

Well, that and my fight with make-kpkg. But that's for another time.

Understatement.

Ian Murdock about building Debian in a few weeks: "Ok, so I was a bit off." Right...

Anyway, I'm a bit glad he was. Otherwise there wouldn't have been anything for us left to do, and I wouldn't have been part of this thriving community which I love.

The Gimp is a great tool. Not only does it allow me semi-professionally manage my pictures (those things that I don't have) or create hackergochis; it also allows me to print pictures of decent quality on my deskjet (at home) or color laserjet (at the office).

Or at least, it did. For some reason, the File->Print option is simply gone. Nowhere to be found. And control+p, which used to be the shortcut to print stuff (as it is on every decent thing) is now suddenly mapped to some palette dialog that I've never heard of.

I want my print back!

Update: Right. Turns out I need to install the gimp-print package, which was pushed off my system somehow with the libgimpprint->libgutenprint transition. Fixed, now.

New bugs.d.o layout

There's been lots of talk on Planet Debian about the new CSS stylesheet for bugs.debian.org. Most people seem to likee it; I don't. The reason for that is very simple, actually; I was used to the 'old' layout now, and this new layout requires me to look more carefully for the information I need.

That of course would've been the case for almost every rewrite of the bugs.d.o layout, but still.

That being said, I shouldn't be yelling about bugs.d.o too much; I have way too old bugs outstanding. Should fix them, uh, "soon".

Disk dead, new ones being shipped

Those of you who were in Helsinki probably remember my IIci that was standing in the hacklab, and that I was using to test-build and test-run emile on. It didn't work for some reason. As it turns out, that reason is probably that I forgot to run objdump on the kernel binary before I installed it on the disk, as is documented that you have to do on the emile website. Silly me.

So, when p2 gave me back my IIci (after DebConf, he'd taken it with him in the car all the way through Sweden and Wat the Hack), I wanted to boot the thing again so that I could try emile the right way. Unfortunately, the 750MB disk (the one with the sid installation on it) gave all sorts of ticking noises, and didn't work. I suspect it had one bump in the road too many

Bummer.

So, I went to ebay, and found some old SCSI drives, most of which can be connected as is to my macs, that had a starting bid of US$19.99, and that nobody had done a bid on yet. Quite a deal.

I just received a message from the seller that they have been shipped tomorrow. Yes, 'have been tomorrow': the message said something along the lines of 'Hi, your disks have been shipped on 08-19-05' (which can probably be read as '2005-08-19'), but it's only the 18th today... and the time differential between the US and Belgium is not that large; it's only 9:45 AM here.

Oh well. In any case, with these disks I should be able to give Christian's 2.6 m68k kernels a try on my other hardware. Hopefully they'll work on at least something; both machines I've tried thus far failed to work, unfortunately.

Received an invitation for the Oldenburg meeting.

By email, yesterday. Not sure whether I'll be able to attend yet. I'd like to...

"Aston-Martin makes profit"

Or so a whole item on the VTM news said today. So what, I thought—companies make profit all the time. In fact, they're supposed to. I was just about to wonder what the news value in that was, when the anchor went on to say...

they've had to wait a long time for that, because in the 91 years that the brand has existed, this is the first time they make profit

What?!?

I surely hope the company I co-founded won't take that long to make profit. I don't think I'm that persistent.

Turns out the people who founded Aston-Martin weren't, either. It's been bought over and over by millionaires who then lost a fair bit of money at it, and has declared bankruptcy seven times. Finally been bought up by Ford who, well, kinda knows the automobile business inside out.

More news

Another item on the news yesterday (too tired then to remember to blog about that one): the NMBS, our national railroad company, is about to hire ~450 new people, since they're planning to increase their capacity in a few years and they'll need to have the personnel to keep up with that.

I assume they have some sort of a business plan to get the passengers to actually fill those trains. But still. Might I make a small suggestion?

<breathes>

ACTUALLY MAKE YOUR FUCKING PRICES AFFORDABLE!

Right now, I can buy a bus pass for a whole year for all of Flanders for less than the price of my train pass for the trip from my home town to Mechelen for one month, which isn't even remotely comparable to a bus pass for "all of Flanders".

Get your prices down by at least half, and I'm sure you'll probably get a shitload of extra customers. Sure, the bus gets subsidies to be that cheap; but I don't see why you wouldn't be able to get those either.

And yes, I'm quite sure that dropping prices will get you a lot more customers than adding wireless connectivity to your trains, which you're working on. Even if I personally find that quite interesting, I think most of your (potential) customers couldn't care less.

Biella's chapter six

I just finished reading the 70-odd pages of chapter six of (now Dr.) Biella Coleman's dissertation (what a word), a chapter which handles (mainly) about the Debian project.

It took me quite a few hours to do this, mainly because English is not my native language, and parts of the text go outside of the technical area in which I feel comfortable, mainly due to the fact that I've been involved with Debian for quite a while now; but also because the text is, at times, consisting of the (IMO) rather complex language that is typical of scientific texts.

Anyway, I must say I'm impressed. Not only has she very well understood and explained the internal culture in the Debian project; she also, and this surprised me, managed to explain things about our internal culture that even I—who's been part of this project for over four years now—had not understood myself; and the idea that crises such as the one around the Vancouver prospectus are necessary if painful, certainly holds some merit.

She certainly has a far better view on the general picture than I guess most of us can ever hope to have. Errors in the document, if any, will be in the little details -- such as the one in the final paragraph on page 20, which indirectly claims that our project secretary (the person who runs our votes) is either a team or a delegate. But those are details, and allowing those to throw a shadow over her accomplishment would be nitpicking.

Again, I'm impressed at her level of understanding; but then again, I guess that's what being an anthropologist is all about. I can only urge you to read it, if you haven't already.

Meantime, I'll consider adopting Yaarr as my nickname. Then again, maybe not.

Freescale Support

About a week ago, I found out that vmelilo doesn't compile, currently, with an error in a file which provides linkage between C and assembly code. This appears to be an issue related to the gcc transition; but I don't know much about assembly, so had a rather hard time finding out what the solution was.

Eventually, looking at all this assembly code got me curious, so I started looking for m68k assembly tutorials. The only other assembly that I'd ever done was that of the Commodore 64 (which isn't very up-to-date these days, it being an 8-bit processor IIRC), so finding out a starting point was rather hard. In the end, I just wrote some simple C programs, compiled them with gcc -save-temps, and had a look at the resulting .s file, comparing the instructions against this m68k programmers' reference manual that I've found in PDF on the Freescale semiconductor website (Freescale is what was created after Motorola split off its processor business; so they're now the principal manufacturer of m68k processors).

That was rather enlightening.

One thing that isn't quite clear yet is how to call a function with some variables, rather than pointers, as arguments. But I'm sure I'll find out eventually.

Playing with all this assembly also made me remember that there's something else I wanted to look into: the differences between ColdFire and "plain" m68k CPUs. ColdFire processors are supposedly m68k compatible, but there are slight differences. As the wording on wikipedia suggested to me that there might also be incompatibilities, I wanted to be sure. So I compared the m68k programmers' manual against the coldfire one. Or, at least, I tried to—they're both rather large, haing some hundreds of pages; and comparing instruction sets isn't exactly what I call exciting. So I decided to just ask. Freescale has a support medium where you can ask questions; I just put my m68k porters' hat on, and asked them, both because I wanted to know and for the sake of it, just to see whether they'd help me as a Free Software developer working on their hardware—though I'm sure other people could've told me (such as, e.g., the gcc maintainers).

They did. Which is nice. Pointed me to both those manuals that I'd already found, which unfortunately wasn't very helpful. Also pointed me to other documentation, though, containing the differences the m68k hardware has against CF hardware. That one pointed me to an appendix in the users' manual of a specific CF chip, if I wanted to 'port software'. Wording of those three pages are such that the CF instruction set is just a subset of the m68k one. Which is great, considering how there's now a CF processor with MMU, which is the only thing CF processors did not have and which they required to be able to run Linux...

War of the Worlds

Went to see 'War of the Worlds' on Friday with Kris. Not bad, but not the best movie ever, either; there's a (pretty negative) review over on the IMDb, and I agree with most of it (except for the fact that I did enjoy the movie, contrary to that person). But really, there are just way too many loose ends and unlikely things in this thing. Characters that are introduced for just a few minutes and then disappear, contributing exactly nothing to the story; a plane that crashes on a house, destroying the entire neighbourhood except for the getaway car which is parked right in the middle of that neighbourhood and conveniently doesn't even have a scratch; those are just two examples. Retro-SF doesn't appear to be Spielberg's thing.

For those of you who've never heard of this story: War of the Worlds is a novel written by H.G. Wells some 80-90-ish year ago, and was also adapted for radio once, although I'm not sure whether the novel or the radio adaption was written first. Anyhow, the radio adaptation was famous in that it was written such that it contained news flashes and so on, making it sound as if it was happening for real. The result was that people would go on the street, panicking, because the Earth was being attacked by aliens... or so they thought. In the end, the police had to go to the studio and shut down the program.

Considering the age of the story is important when thinking of some choices that were made—such as the idea that hundreds of the alien's "tripods" were buried under the ground for millions of years without us detecting them, or the way in which the aliens were eventually defeated (no, I won't tell you what that is ;-P ).

All that being said, it's still a nice movie; the special effects are quite nice, and the story is still great, even if Spielberg didn't do a great job; that probably tells me I should go and read the original WotW instead, though.

milter-ahead

It will never cease to amaze me how people keep using sendmail rather than exim, and then have to jump through ugly hoops to get it to do stuff which exim has on-board.

acl_check_rcpt:
  require verify=recipient/callout=defer_ok

Does quite exactly the same thing, and is builtin to exim(v4). Even better: it actually caches results, so that not only you don't overload your primary system, you might also block some extra mails even when the primary is down and your cache hasn't expired yet. Oh, and one can easily do the same thing for senders, so that you don't get mail with nonexisting email addresses in the envelope from. Catches quite some spam too, that.

Still, I don't run backup MXen, simply because distributing SpamAssassin bayes databases and rulesets is rather complicated... and if my mailserver goes down for a while, I'm sure any sending host will attempt to deliver their mail once I fix the problem. As has happened a few times already by now.

Crazy code.

Every now and then, I have crazy ideas about computing. Mostly "it would be cool if a computer could do foo" kind of stuff; usually they're either impossible with the current state of technology, or I lack the skills to implement my crazy idea.

A while ago, one of those crazy ideas was about how it would be cool to be able to load a new kernel without having to reboot. I hate it when I have to close my applications, even if only for a few minutes. But then, I thought it'd never be possible to implement something even remotely similar to that.

Today, I found out that 2.6.13 is out. And look:

kexec system call (EXPERIMENTAL) (KEXEC) [N/y/?] (NEW) ?

kexec is a system call that implements the ability to shutdown your
current kernel, and to start another kernel.[...]

It's not exactly the implementation of my crazy idea (you still need to shut down your userland before you can load the other kernel), but then it's getting awfully close. I guess the fact that I myself am not immediately able to implement crazy ideas doesn't mean they won't happen...

Quickstep building for testing-security

Or, rather, etch-secure. The difference being that testing-security is an (unused) database on buildd.debian.org, while etch-secure runs on wanna-build.ftbfs.de. As Ryan noted:

* neuro wonders again why time was spent setting up
       testing-security, only to never be used by anyone...

... and he's really got a point. But oh well—it seems to be working, now.

It took me a slight while longer than Kenshi Muto, but then a 25Mhz 040 is also slightly slower than your regular ARM processor...

Whoops

Just found out that I made a little configuration error on quickstep which resulted in buildd-mailer putting files in a different location than what buildd-uploader was looking at. As a result, no packages for experimental were being uploaded anymore...

Hand me the brown paper bag, please.

Oh well. Fixed, now. 22 packages are being uploaded as I write this—and believe me, that is a lot for a 25Mhz 040.