gregoa's blog

fun with a new kernel again. for some value of "fun".

I tried 2.6.38-rc6-686 (from experimental), & I quickly hit 2 regression:

• on my desktop: "Uhhuh. NMI received for unknown reason". workaround: turn off nmi_watchdog with powertop or a boot parameter (nmi_watchdog=0) or `echo 0 > /proc/sys/kernel/nmi_watchdog'.
• on my laptop: "ath5k phy0: gain calibration timeout". workaround: power off (!) machine & reboot into 2.6.37.

good news: at least the DPMS problems on my laptop are solved with 2.6.38. – but on my desktop machines I have some weird artifacts on my screen.

squeeze is released but this doesn't mean the end of the ordinary RC bug; old ones are still around, & new ones keep floating in.

as a short reminder to take the occasional look at the RC bug list I'm again posting what I've done in the last weeks. as you can see no NMUs but only work on packages I have a relation to, & often only uploading packages prepared by others (yeah teamwork!).

• #599794 – libcgi-application-extra-plugin-bundle-perl: "libcgi-application-extra-plugin-bundle-perl: Copyright etc of icons in Authentication module not declared"
  upload package prepared by periapt (pkg-perl)
• #603252 – libthread-queue-perl: "libthread-queue-perl: obsoleted by perl in squeeze"
  upload new upstream release prepared by periapt (pkg-perl)
• #603253 – libtime-local-perl: "libtime-local-perl: obsoleted by perl in squeeze"
  upload new upstream release prepared by periapt (pkg-perl)
• #605795 – src:libconfig-model-backend-augeas-perl: "libconfig-model-backend-augeas-perl: FTBFS: failed tests"
  upload package prepared by dod (pkg-perl)
• #605796 – libdevel-ebug-perl: "FTBFS: failed tests"
  update bug info, later closed by carnil (pkg-perl)
• #605809 – src:libhtml-widget-perl: "libhtml-widget-perl: FTBFS: failed tests"
  patch to make tests compatible with libhtml-tree-perl >= 4 (pkg-perl)
• #612723 – iodine: "iodine overwrites /etc/default/iodine during upgrade with a sourced version of that file"
  change config script to read variables from configuration file instead of sourcing it (maintainer upload)
• #613187 – libogre-perl: "xs/SceneQuery.xs:28: error: cannot convert [..]"
  upload new upstream version (pkg-perl)
• #613388 – libproc-processtable-perl: "libproc-processtable-perl: FTBFS on kFreeBSD"
  apply kFreeBSD patch (pkg-perl)
• #613390 – src:libgearman-client-async-perl: "libgearman-client-async-perl: FTBFS: tests hang"
  merge and tag confirmed (pkg-perl)
• #613395 – libnet-twitter-lite-perl: "libnet-twitter-lite-perl: FTBFS: spelling tests fail"
  run tests under LANG=C, later disable POD tests (pkg-perl)
• #613398 – libpoe-perl: "libpoe-perl: FTBFS: tests fail"
  upload new upstream release, prepared by periapt (pkg-perl)
• #613731 – libtest-www-mechanize-perl: "libtest-www-mechanize-perl: FTBFS: test failures"
  upload fixed package prepared by roam (pkg-perl)
• #613732 – libtest-xml-simple-perl: "libtest-xml-simple-perl: FTBFS: test failures"
  upload fixed package prepared by roam (pkg-perl)
• #613942 – perlipq: "perlipq: FTBFS on armel - Failed 1/2 subtests"
  adjust patch in test suite to only run first test (pkg-perl)

now that we have released squeeze, I'd like to say thanks to the release team. I've really enjoyed the cooperation in the past months.

the people I had the most contact with were mehdi & also adam, & they were always very quick, very friendly, & very cooperative when I had some requests.

merci beaucoup, & I hope you continue your good work!

my desktop machine is rather old & was never able to suspend/hibernate. or more precisely: to wake up afterwards.

today I was in the mood for testing it again, because with the Debian kernel even KMS works on it.

&: suspend & hibernate also work! both manually via /sys/power/state & with pm-utils. very nice. – I don't know if it's 2.6.37 or the better .config or something else – & I don't really care :)

(it looks like using a distro kernel really pays off. less karma for me, more disk space used – but also more features.)

new year, new RC bugs – most of the fixes below are for newly reported bugs in versions that don't affect squeeze. the rest are mainly uploads of packages or patches prepared by others.

• #566072 – util-linux: "fsck now failing hard for unavailable filesystems"
  prepare a patch based on adsb's analysis. later fixed in maintainer upload
• #605912 – runit: "runit: Upgrade failure lenny -> squeeze"
  sponsor NMU prepared by jonathan nieder
• #606370 – libcgi-pm-perl: "CVE-2010-2761 CVE-2010-4410 CVE-2010-4411"
  upload package prepared by ntyni to unstable, testing-proposed-updates, and stable (pkg-perl)
• #607556 – src:libconfig-model-itself-perl: "libconfig-model-itself-perl: failed tests"
  manual close, new version works again (pkg-perl)
• #609398 – src:libfile-userconfig-perl: "libfile-userconfig-perl: Failing tests if home of building user is not available"
  set HOME to a writable directory in debian/rules (pkg-perl)
• #609402 – src:libapache2-reload-perl: "libapache2-reload-perl: mkdir in home of building user"
  upload package prepared by periapt (pkg-perl)
• #609464 – src:libfile-homedir-perl: "libfile-homedir-perl: Failing tests in t/02_main.t tests"
  set HOME to a writable directory in debian/rules (pkg-perl)
• #609467 – src:libpath-dispatcher-declarative-perl: "libpath-dispatcher-declarative-perl: tests fail"
  upload new upstream release (pkg-perl)
• #609468 – src:libtk-pod-perl: "libtk-pod-perl: cannot remove path when cwd is /tmp/tkpod_8LQuQjJc for /tmp/tkpod_8LQuQjJc: at /usr/share/perl/5.10/File/Temp.pm line 902"
  set HOME to a writable directory in debian/rules (pkg-perl)
• #609469 – src:dh-make-perl: "dh-make-perl: Tries to mkdir directory in home of building user"
  set HOME to a writable directory in debian/rules. in svn, not yet uploaded (pkg-perl)
• #609472 – src:libgraphics-gnuplotif-perl: "libgraphics-gnuplotif-perl: Build hangs"
  set HOME to a writable directory in debian/rules (pkg-perl)

see you there?!

the release team has started to usertag [0] the remaining RC bugs – yeah, another step towards the release!

here are my small contributions around RC bugs:

• #544310 – xsidplay: "xsidplay: FTBFS on GNU/kFreeBSD (alsa is linux-specific)"
  add [linux-any] to libasound2-dev, upload to DELAYED/2, later integrated into maintainer upload
• #551543 – muttprint: "cannot remove path when cwd is /tmp/muttprint*"
  add a patch for the second incarnation of the bug report
• #604234 – emacspeak-ss: "emacspeak-ss: dpkg hangs while preconfiguring packages"
  update bug info
• #605240 – tenshi: "tenshi: fails to upgrade from lenny"
  update bug info
• #605793 – src:libclass-dbi-asform-perl: "libclass-dbi-asform-perl: FTBFS: failed tests"
  add patch and tighten build dependency (pkg-perl)
• #606000 – libmail-spf-query-perl: "libmail-spf-query-perl: Incorrect query results with IPv6 addresses; should warn about missing IPv6 support and/or fail graciously"
  update bug info and clone to qpsmtpd (#607472), package removed later (#608335)
• #606777 – sendfile: "sendfile: package fails to upgrade properly from lenny"
  update bug info, later upload chris' patch to DELAYED/2
• #606779 – openswan: "openswan: package fails to upgrade properly from lenny"
  update bug info
• #606785 – common-lisp-controller: "stumpwm: package fails to upgrade properly from lenny"
  reassign and update bug info, later sponsor maintainer upload
• #606802 – bacula-director-pgsql: "bacula-director-pgsql: make_catalog_backup is broken"
  investigate further and make proposals, later upload to DELAYED/2
• #607472 – qpsmtpd: "qpsmtpd: Switch from Mail::SPF::Query to Mail::SPF"
  propose patch, later closed by maintainer

[0] if I got it right:

• squeeze-will-remove
• squeeze-is-blocker
• squeeze-can-defer

update: all tags combined: http://tinyurl.com/squeeze-sort (thanks to the RT for the URL & to zack for the pointer)

yesterday I played a bit with my desktop machine:

• changed /etc/fstab & /boot/grub/menu.lst (yes, I'm still using grub-legacy) to use LABEL=foo instead of device names after setting labels on all partitions.
• interestingly root=LABEL=root doesn't work with my self-baked kernels, not sure what I'm missing there.
• installed the latest Debian kernel (from experimental) & the current xserver-xorg-video-intel (from unstable): boots, gives me KMS & a working X (with my own kernels I had no KMS and just a black screen with newer versions of xserver-xorg-video-intel).
• with the Debian kernel I've also entered the libata era :)
• created my own linux-kbuild-2.6.37 package following the instructions (& changing the script a bit) in order to be able to use virtualbox-ose-dkms.
• contrary to my laptop, DPMS doesn't make my screen faint permanently.

seems I'm about to leave the time of building my own kernels ...

a few days ago I was quite happy with 2.6.37-rc5-686 because ath5k suddenly works.

but.

but then I found out that it turns off the screen so effectively (after the screensaver, so in DPMS standby/suspend/off) that it stays black.

not good.

ok, let's try (vanilla) 2.6.37-rc7. – no DKMS & no X.

not good.

let's apply a friendly patch from the Debian kernel. – DKMS & X are back. but DPMS still leaves the screen black. also when invoked manually with xset.

not good.

let's ask a search engine. – oh, a known bug: 22672. hm. at least the workaround (close lid / suspend to RAM) works. & turning off DPMS (in xscreensaver & laptopmode) obviously helps.

but then my poor noseguy can't stop wandering around the screen ...

don't we all love the blood^Wbleeding edge?

ath5k has never worked for me – until today. I installed linux-image-2.6.37-rc5-686 from experimental & lo&behold – it works!

that means I can say goodbye to good old (& non-free) madwifi. but before discovering that ath5k works I had already finished an upgrade of my private madwifi package & it seems I even managed to build a madwifi-dkms package finally.

I've uploaded it to my private repository for the poor sods who still can't use ath5k :)
& maybe the madwifi-dkms package helps people who forget (how) to run m-a after installing a new kernel ...