gregoa's blog

after attending the Munich BSP 3 weeks ago (thanks for inviting me, & thanks especially to Stefan for hosting me) I've tasted blood & started to look at our RC bugs more often. I'm still far from zack's RCBW frequency; but I thought I'd give a short overview anyway; maybe it motivates others to chime in? – So here's a list of RC bugs I've touched in the last weeks (not counting some merges in the BTS):

• #375025 – libzorpll: "libzorpll: FTBFS on GNU/kFreeBSD (due to unsatisfied Build-Depends on libcap-dev)"
  apply patch from the BTS, upload to DELAYED/2 (ok, technically not RC)
• #544879 – clutter: "Missing libclutter-doc package"
  add comment at turmzimmer and BTS
• #548616 – licq: "fails to build with gcc-4.4"
  apply patch from the BTS, upload
• #553009 – libzorpll3.0.6: "libzorpll3.0.6: postinst-must-call-ldconfig /usr/lib/libzorpll-3.0.so.6.4.0 by the dynamic library loader. Therefore, the package must call "ldconfig" in its postinst script."
  create post{rm,inst} and add dh_makeshlibs, upload to DELAYED/2
• #553230 – libapache2-mod-macro: "libapache2-mod-macro: missing-dependency-on-libc needed by ./usr/lib/apache2/modules/mod_macro.so but the package doesn't depend on the C library package. Normally this indicates that ${shlibs: Depends} was omitted from the Depends line for this package in debian/control."
  apply patch from the BTS, upload to DELAYED/2
• #553560 – chiark-backup: "chiark-backup: file-in-etc-not-marked-as-conffile /etc/chiark-backup/snap/nosnap"
  add file to conffiles, upload to DELAYED/5
• #556214 – pyspi: "pyspi: FTBFS: Error: pkg-config could not find x11"
  apply patch from the BTS, upload to DELAYED/2
• #559077 – google-gadgets-common: "google-gadgets-common: needs to depend on shared-mime-info because of postinst"
  add comment at turmzimmer
• #560637 – src:libdatetime-format-dateparse-perl: "libdatetime-format-dateparse-perl: FTBFS: tests failed"
  apply patch from upstream bug tracker, upload (ok, I'm in Uploaders)

remarks:

• so gar I got 4 "thank you" mails, two offers to adopt a package, & no complaints.
• we really have interesting packages in the archive; look at this nice stanza in the PTS: Lintian reports 5 errors and 221 warnings about this package. You should make the package lintian clean getting rid of them.

seems I'm only blogging about wifi problems on my laptop. anyway, status update:

I'm visiting friends, & I really want to sit in their kitchen/living room with my thinkpad & wifi. – as usual it didn't really work (dropouts, incredibly long RTTs, terribly slow connection, ...). but I think I found a solution:

like in summer I tried with madwifi instead of ath5k. but this time not with the madwifi-hal-0.10.5.6-current.tar.gz tarball but with the madwifi-0.9.4-current.tar.gz from http://snapshots.madwifi-project.org/.

so I again created a madwifi-source package & then built a madwifi-modules package against my current (2.6.31, 2.6.32 had funny issues yesterday) kernel. & lo&behold – I seem to have something that feels like wifi again.

the madwifi-source package is available at the Toastfreeware Debian Repository's experimental/non-free area.

(& now let's see if I have to revoke my success story or not.)

debcamp. first there was one linksys, & my wifi worked fine. after the admins set up lots of nice APs (kudos!) my wifi connection become flakey & slow & dropping out & everything – thanks atheros & ath5k!

before I went completely mad I tried plan B: going back to the good old (non-free) madwifi. so I grabbed the last madwifi-source source package from stable, the latest tarball (madwifi-hal-0.10.5.6-current.tar.gz) from the madwifi page, built a new madwifi-source package, & then a module package against a kernel source of 2.6.31-rc2-git7 (which I happened to have lying around).

looks much better. ssh'ing home to austria is a bit slow but in general the connection looks stable. & I'm still sitting outside :)

update 1:
or not. /me->mad. – let's try the next kernel :/

update 2:
2.6.31-rc3 & ath5k look better. at least inside the venue. let's try & go out.

update 3;
works in the upper talk room (more more than less); outside: more less than more.

with the help of dkg's & ana's howtos I've started the transition from my old 1024-bit DSA 0x00F3CFE4 key to my new 4096-bit RSA one 0x8649AA06.

fun with my machines at home in the last days:

• on the laptop my network interfaces got renamed from eth0 to eth1 & from ath0 to wlan0. messing with /etc/udev/rules.d/70-persistent-net.rules helped. this seems to be #521521.
• strange sound phenomenons on my desktop. turns out that (1) centerim's new sounds are nice but very quiet & (2) the audible bell in xterm is broken; the latter is #520405. fixed by quickly creating a local 243-0.1 package.
• on my laptop the function keys don't work via acpi any more. turns out that CONFIG_ACPI_PROC_EVENT is finally turned off in debian's 2.6.29 kernel, & I found no documentation on how to make the keys work again with hal/evdev/whatnot. this is #521279, & right now I'm building a kernel with CONFIG_ACPI_PROC_EVENT enabled.

sometimes running unstable can be a bit time-consuming ...

I'm playing with my new toys:

I got my OpenMoko Freerunner around 2.5 weeks ago. of course I've installed Debian on it — & I'm spending so much time with it that I don't have the time to blog about my successes & failures :) but I guess I will.

today I bought a 3G usb modem & an appropriate SIM card (the yesss! product sold at hofer [they even had one in stock in the third outlet I checked {at 17:57}]) — mainly to have a SIM card for my Freerunner without having to take my regular one out of my communicator :) but of course also to have mobile internet.

setting up the latter was not much of a hassle; usb_modeswitch doesn't seem to be necessary (usb_storage fails, so the two modem devices are brought up) but I have it installed nevertheless.

& regarding the connection: guess what — using pppd (pon/poff after writing /etc/chatsecrets/yesss & /etc/ppp/peers/yesss) works better than umtsmon (which works too except that it claims "connection error" after setting up the connection).

(pppd files & usb-modeswitch/umtsmon packages available on request. basically there is all available out there "on the intarwebs")

on monday I migrated tux (the debian server at my workplace) to a new machine (hp [0] proliant ml110 g5). & it was completely painless. what I did:

• install a base system on the new machine (with debian-40r5a-etchnhalf-amd64-netinst.iso) [1]. yes, this is etch:
  • the installer worked flawlessly to create a RAID1 over the two disks & to install grub for booting from / on RAID. very nice.
  • just a minimal system, no users, ...
• copy stuff over from the old machine: (most of) /etc (including /etc/passwd & friends), /home, & a bit more.
• get the package list from the old machine, install them on the new machine keeping the old config.
• a few minor tweaks.
• change the ip addresses on both machines & switch their physical positions.

yes, that's it.

after ~3 hours I now have:

• my first dual-core machine
• my first amd64 system
• my first box with an etchnhalf kernel
• a box that is completely bored :)
  from htop:

  1  [|                                        0.3%]     Tasks: 49 total, 1 running
  2  [|                                        1.0%]     Load average: 0.00 0.00 0.00 
  Mem[|||||||||||||||||||||||||||||||||||442/1004MB]     Load: 0.00 
  Swp[                                     0/1976MB]     Uptime: 2 days, 12:01:18

I keep being impressed by how simple linux is, & how good debian & especially the debian installer works.

do I need to mention that the migration of the windows server (from nt 4.0 to windows 2003 on new hardware) is still not finished after 2 days of hard work?
to be honest: the windows box has more jobs than tux: domain controller, roaming profiles, file server, exchange, mssql, some library database, ... tux only runs exim4 + spamassassin + clamav + greylistd, squid, some scripts, & does a bit of iptables blocking & port forwarding.
but still ...
(& the old machine worked more or less fine with 96 MB RAM & a ~300 MHz CPU.)

[0]
because they are linux & debian friendly.

[1]
of course that can be avoided by just copying stuff over from the old machine, either via netcat & friends or by putting the old disks in the new machine. but in this case (1) the old box doesn't boot from CD (have I already mentioned it's an old box?), (2) the old one has SCSI & the new one SATA disks, (3) the old one was i386 & the new one (at least potentially) amd64.

• I have a new toy: notify-send.
• I need to pratice guitar playing more often.

so let's try to combine these points:

• install needed software: notification-daemon, libnotify-bin (contains notify-send).
• find an image.
• set up a cronjob.

cronjob (wrapped over several lines for better legibility):

0 17 * * * DISPLAY=:0.0 notify-send \
        -t 0 \
        -u critical \
        -h "int:x:1680" -h "int:y:0" \
        -i /home/gregoa/data/sound/trio\ infernal/triologo_square_border_transparent.png \
        "Trio Infernal says:" \
        "Heute schon Gitarre gespielt?"

(for details check man 1 notify-send)

result (at the top right corner of my screen):

& yes, this is not my only use case for notify-send :)

after I had some troubles today I put my solution for reading UTF-8 encoded HTML mails with mutt into a short article.

today in my website's access.log (wrapped for easier legibility):

64.1.215.166 - - [21/Nov/2008:19:03:02 +0100] "GET /{,blog/} HTTP/1.0" \
  404 6963 "-" "Mozilla/5.0 (Twiceler-0.9 http://www.cuil.com/twiceler/robot.html)"

the broken URL comes from my ~/.signature which since about 2 months contains the following fragment:

http://info.comodo.priv.at/{,blog/}

& probably that line was found in some web archive of a mailing list or newsgroup or somesuch.

now what is this highly intelligent Twiceler thingy? a look at the URL in the access.log entry redirects to cuil, & the page proudly claims in friendly, large, green letters

Cuil is the biggest search engine on the planet.

apart from the fact that I've never heard about them & that their search results are not impressive either, their "philosophy" (huh?) page claims:

Size matters

maybe it's just me but probably they should invest more in technical "cuil" [0] than in marketing speak that sounds less like search engines but more like another business branch on the intarwebs.

[0] "Cuil is an old Irish word for knowledge." (source: cuil again)