Maintenance and compatibility
Hi All,
[This is not intended as flame bait! nor do I intend to start
a flame war! It is honestly the way I see it, with all my lack of
knowledge and meant to improve the project in the long run.]
sorry guys, to critique a project that I don't give anything
back to, except the ocasional stupid question and installation report
and use on an almost daily basis. I'm one of the guys in the trenches,
using d-i as a systems admin, I am not a programmer.
Last week I filed a bug report against the installer
(http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=344371) and yesterday
evening I had a conversation on #debian-boot ([1], don't know if public
logs are kept) that let me to a few conclusions:
-d-i etch is (rightly) a continuation of d-i sarge, but the
udeb's used in the installation are not kept seperatedly on the mirrors.
d-i-sarge uses d-i-etch udebs.
-there isn't a process to do maintenance (debug errors) on sarge
only udebs.
Keep in mind there never was an official d-i-sarge. The one
downloadable from March '05 is still marked as rc3. I use the one from
JoeyH's images dated 6-6, the release date of sarge.
D-i-etch isn't backwards compatible with sarge (ie. not able to
install sarge, see 344371).
I see this all as (sorry) severe shortcomings of a great product
that doesn't do right to its goals, the quality of the work and does our
Debian users a great wrong.
The remark I see coming over the horizon "use d-i-etch" is not
valid, because it is not stable (in the sense of stability, not in
debianese release naming).
I guess it is too late to get the two projects separated and
compatible again, but it would be a admirable goal for the next version.
Sincerely,
Jan.
PS I never understood why there never was an official d-i-sarge release,
nobody I know ever misses such an obvious oportunity to throw a party ;-)
[1]
(21:53:18) netman1: Hi All. Sarge-d-i preseeded netboot install refuses to select smp kernels. Tried 'd-i debian-installer/kernel/image string kernel-image-2.6-686-smp', 'd-i debian-installer/kernel/image-2.6 string kernel-image-2.6.8-2-686-smp' and 'd-i base-installer/kernel/which-kernel select kernel-image-2.6-686-smp' in various combinations. No luck. What am I doing wrong?
(21:56:08) fjp: netman1: Try setting "set -x" in the base-config postinst script and check syslog. You are in a better position than us to find out where/why it goes wrong.
(22:23:57) netman1: fjp: thanks, will try.
(22:24:19) joeyh: /fjp/s/base-config/base-installer/g
(22:24:46) netman1: joeyh: for me?
(22:24:51) joeyh: yes
(22:25:02) netman1: joeyh; thanks.
(22:25:08) fjp: Eh, yes.
(22:25:17) fjp: Sorry.
(22:25:20) netman1: fjp: no problem.
(22:56:48) netman1: SMP problem : /var/lib/dpkt/info/base-installer.postinst tests with 'dmesg | grep -q ^Processors:'. Anyone interested in the logfiles (including dmesg)?
(22:57:28) netman1: s/dpkt/dpkg/g
(22:58:21) fjp: netman1: Suggest bug report against base-installer. Although the relevant code was pretty much rewritten for Etch, so it may no longer be really relevant.
(22:59:09) netman1: fjp: what command produces/produced the '^Processors:' line in dmesg?
(22:59:50) fjp: netman1: kernel boot I would say
(23:00:31) netman1: fjp: Ok, to shorten out my problem, what would be the best way to force a smp kernel, when required?
(23:01:32) fjp: Not sure without manual intervention during install. If the code's broken, it's broken...
(23:01:45) makx: netman1: wget, dpkg -i on vc2
(23:02:43) fjp: Hi makx
(23:02:50) netman1: maks: hmm, we _are_ working in ICT, was looking for an automatic way. Changing the test?
(23:03:08) fjp: Any news on the ide-disk front?
(23:03:12) netman1: makx: in a local udeb ofcourse?
(23:03:36) fjp: netman1: Install a custom version of the postinst...
(23:04:27) netman1: fjp: as I was proposing. But what would be the best way to test from the installation kernel dmesg to test for the number of cpu's?
(23:06:57) ***fjp does not have any multi-cpu systems, but is willing to let netman1 donate him one ;-)
(23:07:10) netman1: fjp: rofl!
(23:07:34) netman1: fjp: can get you access though...
(23:08:01) joeyh: dmesg | grep -q "WARNING: NR_CPUS limit of 1 reached."
(23:08:12) joeyh: that's what the current installer (ie, etch) uses
(23:09:42) netman1: joeyh: exit code 1 :-(
(23:10:10) netman1: joeyh : as in NR, CPU, WARNING
(23:13:05) netman1: joeyh: uname -a ' 2.6.8-2-386 #1 Jan 24 03:01:58'
(23:14:17) joeyh: I tested it this augist with 2.6.8 and it worked on the one smp machine I hav
(23:15:20) netman1: * : time to bed, wife calls ;-) Will look in the SMP problem further tomorrow.
Reply to: