Older blog entries for mjg59 (starting at number 8)

Wireless

chexum, if you get hold of a card based on the Prism II chipset (that is, most cheap ones on the market at the moment) you can use the host AP driver.

NetBSD

Fakeroot segfaults horribly. I've traced it down to the code which does the actual wrapping of the functions, but I can't figure out what's actually happening there. Which is irritating, because it's one of the few things left which really needs to be done now that PAM works happily.

CDR

The CDR miniconference seemed to go fairly well. More people than I'd expected, and a reasonably large contingent of people who weren't from Cambridge. Reasonably good publicity for CUDES as well.

Sleep

Still lacking.

Work

The CO had told the head of department that scp was a security risk. Hngk.

Student computing

I gave a talk on student representation in computing a couple of weeks ago. CUDES seems to have resulted. The speed at which this stuff has happened is slightly disturbing. People in Edinburgh seem similarly pissed off, which is nice. We'll have representation on JANET yet.

Debian GNU/NetBSD

Almost working GCC, almost packaged XFree, a FreeBSD based system that's installable using a hacked debootstrap - and none of it done by me. There doesn't seem to be an insurmountable quantity of difficult stuff now. Scary. I've finally got around to getting my gpg key signed by cjwatson, so I'm just waiting for my application manager now.

Sleep

Insufficient.

Work
Doesn't look promising for a change in departmental policy, with the consequence that I may well be moving across the street before too long. I'm still impressed at how an institution with such a strong research reputation and with so many intelligent people working for it can lack any sort of overall approach to computational biology, leading to pockets of sanity right next door to departments who wouldn't know how to encourage the subject even if you wrote them a thesis on it. Oh well, analysing gene expression level data sounds like fun too.

!Work
Up in Edinburgh at the moment, visiting friends and promoting rebellion. Stepped in through the door, gave somebody my MAC address and now I can use my laptop as if I'm sitting at home. VPNs are cool.

I hadn't quite been prepared for the hills. Cambridge is going to seem awfully flat when I get back.

Impressively, nothing has gone badly wrong yet this month. I might be able to postpone going mad for a while if this keeps up.

Progress seems to be continuing on Debian GNU/NetBSD - somebody's got DejaGNU working, so we've actually got test results from the gcc build which reveal a few hundred failures. Whoops. Something to take a look at, I think. Of course, this might go some way to explaining why ed is segfaulting in weird places.

NetBSD
utsl provided a patch that got sysvinit running - I now have a self-hosting Debian NetBSD system. Something's broken with dynamic linking (LD_LIBRARY_PATH needs explicitly setting before most libraries get found), and fakeroot is horribly broken (possibly something to do with the above). Other than that, things look promising. People can actually log in and do some development themselves now, even if the fakeroot problem does mean that building packages is awkward.

Work
Fixed a stupid bug in my clustering simulation, and I now get pretty much the results I expected. I'm unconvinced by the route that Gestwickie and Kiessling get from their observations to their conclusions (both of which seem sane), and I can probably start doing something about that now. However:

My department has instructed me to stop using the departmental network for copying my simulation data to another machine as it's generated. Since departmental regulations also prohibit me from logging into the machine to get hold of the data, this makes it difficult for me to work the way I normally do (lie in bed thinking about the problem, have an idea, implement it then while I can still visualise what I'm thinking about) and productivity has decreased - the only way I can get anything done at the moment is to hang around in my office until 3 in the morning (entertainingly, the wording of the restriction actually prevents me from reading my email while at work as well). Needless to say, I'm not overly happy about this. I'm meeting the head of department at some point this week to see if they're willing to make any sort of concession here, but if not then I may be heading back to the Department of Genetics, data mining and several extra years before a PhD. Fun.

UK law in somewhat more draconian than people expected shocker. Modchips are circumvention devices (not terribly surprising), and importing a game doesn't necessarily give you the right to make a transient copy of it in your system's memory. Quite what this is going to do to the import games industry isn't obvious, and it could probably be stretched to cover imported DVDs as well. Might be worth grabbing those region one DVDs while you still can...

Debian GNU/NetBSD? Debian NetBSD? When is the GNU contribution worth enough to give it the title? Is GNU libc required for a GNU OS? A GPLed ifconfig? If init is under a BSD license, can it truly be called GNU?

Heh. Next time I'll know not to bring this sort of thing up until the last possible minute :)

I'm bored. Hence http://www.srcf.ucam.org/~mjg59/entropy/

(yes, I *know* :) )

Debian NetBSD updated, and I've got around to writing a website for it (at www.srcf.ucam.org/~mjg59/debian-netbsd for the moment until I get in touch with the person with access to the Sourceforge account). Waiting for the file to upload at the moment - the bottleneck of a 128Kbit cap on my outgoing traffic is not helping in getting this onto the server.

Debian-NetBSD (GNU/NetBSD? We're not using glibc and a small number of system-specific apps are the BSD ones rather than the Linux ones, so I'm still not completely sure. Porting glibc would probably solve this and a few other issues, but the prospect of doing so doesn't appeal greatly...) is now at the point of being able to build its own packages. Still chunks of base missing (netbase, for instance) and passwd doesn't deal well with NetBSD's handling of stuff, so still some work to be done. I need to get round to lugging the DEC 3000 on my desk back home so I can test non-i386 stuff (it's a Turbochannel Alpha, so no Linux support) and then start looking for more test boxes. It's not really too far off the point where we can pretty much just throw a build daemon at it and then sit back and deal with the occasional broken package (cfingerd needs to be told that it should treat NetBSD like FreeBSD, and the perl package's

arch=$(dpkg --print-installation-architecture)
case "$arch" in
    hurd-*) archname=$(expr "$arch" : 'hurd-\(.*\)')-gnu;;
    *)      archname="$arch-linux"
esac

is not exactly optimal), so maybe I'll have some time on my hands before too long.

(Yeah, right)

Still staring at Stochsim. I'd really better spend more time working on this thing - the current developer's leaving in a couple of months, and then it's all my fault. My C++ is coming along wonderfully (4th complete rewrite on my toy code for molecular diffusion in a 2D membrane now, and it's both roughly working and not looking like a complete travesty against mankind), so I might stand some chance of doing this. Then I really need to start thinking about actually doing some research. Argh.

(/me wanted a gcc tarball from gnu.org that actually built properly on NetBSD for Christmas, but ended up with clothing)