Re: apt-move & debootstrap
On Thu, Dec 27, 2001 at 01:18:43AM +0100, Matthijs Melchior wrote: > Now dbootstrap starts with getting /mirrors/debian/dists/sid/Release, > and this file is not there. One thing you can do is to change the "mirror style" line at the top of /usr/lib/debootstrap/scripts/sid to "main" instead of "release". I think that works.. ;) > Please, can you point me somewhere I can find out how to get these > Release files created. > -- > Thanks, Someone pointed me to katie (a python script in a cvs archive).. this is probably more complexity than you want though. Try the modification above and if you need more info search this lists archives for katie and cvs. Peter -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Re: [d-i] #118295: rootskel: linuxrc should use tmpfs instead of ramfs
On Tue, Dec 25, 2001 at 04:42:50PM -0800, David Kimdon wrote: > Hi, > > Any seconds on this one? I didn't follow any of the discussion > leading up to it, and don't know if there is a reason to change to > tmpfs. > > Thanks, > > David the ramfs it currently uses dopesn't tell you anything at all when you do a df. tmpfs does Also the documentation says the ramfs should not be used as a serious filesystem but should be considered a demo. Peter -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Re: woody install hang on powerpc
> 2) when i select "configure keyboard" and select the keyboard type automatically i >am thrown to >another window to configure swap space. the swap space window is frozen. some >control charcters >bounce me around to diffent "selectables" but "enter" never works. > > 3) the only base line i have is that the "potato" install works fine. i have other >problems with >potato (shared libs at init time) which is why i am trying out woody. > > rich_d Try tweaking settings in /proc/sys/dev/mac_hid/. Specifically the one refereing to linux key syms or something like that. Change it from 0 to 1 or the reverse. Peter -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
support for non-interactive installs
Well with the latest release of base-config-1.33.4,
my trick of putting "db_set" statements in /root/dbootstrap_settings
is now broken, with base-config dying saying db_set command not found.
It worked fine in base-config-1.33.
So this is what my dbootstrap_setting would look like:
DEBIAN_FRONTEND=Noninteractive
SUITE=$SUITE
db_set apt-setup/mirror sunsite.cnlab-switch.ch
db_set apt-setup/country Switzerland
db_set apt-setup/uri_type http
db_set tzconfig/gmt true
db_set tzconfig/change_timezone Europe/Zurich
# may need to file a bug against tzconfig
db_set apt-setup/non-us true
db_set apt-setup/contrib true
db_set apt-setup/non-free true
db_set base-config/run-dselect false
db_set base-config/run-tasksel false
# done with template sed foo
#db_set passwd/md5 true
#db_set passwd/shadow true
And it would be nice if that worked.
So anyway, if this is not 'the way' to do it, what is ?
I change the inittab.install to have "env DEBIAN_FRONTEND=Noninteractive
/usr/bin/termwrap base-config"
to make that work.
The DEBIAN_FRONTEND can't be exported from dbootstrap_settings as I would try in the
above example.
So please, what is the way to get the behaviour I want ?
What would you seasoned debian developers recommend ?
This is one alternative I have used for the passwd package:
# this template sed foo is done since we don't have a simple sweet way to
load-templates, and db_register can't
# create a template
# HMM. busybox can't handle a between addresses and 's' command.
echo Modifying $TARGET/var/lib/dpkg/into/passwd.templates to default md5
sed -e '/passwd\/md5/,/false/s/false/true/' \
$TARGET/var/lib/dpkg/info/passwd.templates >
$TARGET/var/lib/dpkg/info/passwd.templates-new || die "sed passwd template"
mv $TARGET/var/lib/dpkg/info/passwd.templates-new
$TARGET/var/lib/dpkg/info/passwd.templates || die "mv passwd template"
I do think a non-ineractive install is worth supporting.
I had it all working too. Oh. and while I am on about it, there is a stupid "press
return to continue"
in /usr/lib/dpkg/apt/methods/install that can't be gotten rid of.
In fact, here is some more snippet from my installer script:
# this suppresses the "Do you want to continue" prompt of
/usr/lib/base-config/75apt-get
echo 'DSelect { Options "-y -f"; };' > $TARGET/etc/apt/apt.conf
# note that /usr/lib/dpkg/methods/apt/install looks for 'always' but man apt.conf says
to use "clean" or "autoclean"
echo 'DSelect { Clean "always"; };' >> $TARGET/etc/apt/apt.conf
echo 'ACQUIRE { Retries "3"; };' >> $TARGET/etc/apt/apt.conf
This works nicely (worked until base-config-1.33.4 came out and broke my db_set
stuff), but it can't get rid of:
always)
$APTGET "$APT_OPT0" "$APT_OPT1" clean &&
echo "Press enter to continue." && read RES && exit 0;
;;
#33
That would be some 10 bug reports on different packages I guess. Could we have a
discussion about
the 'right' way to approach a non-interactive install. I at least have a working
script that tries
hard to use the clusterfu^H^H^H^H^H^^H^ beautiful apt tools in the way they were
scrambled
together I mean designed. ;)
Kind regards and lots of harmless cynicism,
Peter
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Bug#121110: boot-floppies: TFTP boot on Alpha broken
> > No, I don't think this is related to the problem reported in 121110. > I'm guessing 121110 has been fixed along with bug#116829 -- the > problem was the way that busybox init and telinit didn't get along > together in the context when init wasn't pid 1. > > I don't think the "fix" you outlined will fix anything. > > Assuming the good alpha folks are installing over the network, they > should be using sysvinit 2.84-1, in which case, this bug should have > already gone away. I think your right there. My "fix" does fix something else though. It fixes reboot when in initrd, which doesn't work without the fix. I'll post a bugreport against busybox. I mixed up the two bugs in my mind cause they both disappeared at the same time for me, prob. when the new sysvinit came out and I started using my patched busybox. Peter -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Re: Bug#121110: boot-floppies: TFTP boot on Alpha broken
If this is the init q thing that kill syour system during an install
when running with initrd, I have a patch to busybox that
fixes it on the powerpc.
The problem is that ... when reboot.c
does 'pid_t *pid = find_pid_by_name("init");',
then there is code in find_pid_by_name that "makes sure" init is found
by passing '1' back if it couldn't find it.
This is the fix I used:
Add the comments to libbb/find_fid_by_name.c:
if (pidList)
pidList[i]=0;
// else if ( strcmp(pidName, "init")==0) {
// /* If we found nothing and they were trying to kill "init",
// * guess PID 1 and call it good... Perhaps we should simply
// * exit if /proc isn't mounted, but this will do for now. */
// pidList=xrealloc( pidList, sizeof(pid_t));
// pidList[0]=1;
else {
Sorry I am too lazy to do a proper diff, but it is a small file and
an easy hack.
I am even too lazy to submit a proper bug report.
Peter
On Sat, Dec 01, 2001 at 07:37:43PM +0100, Falk Hueffner wrote:
> Adam Di Carlo <[EMAIL PROTECTED]> writes:
>
> >
> > Falk Hueffner <[EMAIL PROTECTED]> writes:
> >
> > > Package: boot-floppies
> > > Version: N/A; reported 2001-11-25
> > > Severity: normal
> > >
> > > Installing via tftpboot.img will stall at installation of libc6.1, as
> > > has been reported to debian-alpha several times,
> > > e.g.
>http://lists.debian.org/debian-alpha/2001/debian-alpha-200110/msg00010.html. "rgh"
>told me on IRC:
> > >
> > > mellum: re "init q", when I had that problem ia64, it was
> > > because I forgot to boot with root=/dev/ram
> > > mellum: without that, the kernel wants to do it's initrd magic,
> > > and runs /linuxrc before starting sbin/init. linuxrc is linked to
> > > init, so effectively init gets run as pid 12 (or something > 1,
> > > anyway)
> > > linuxrc is run with a pid > 1, and the kernel expects that to
> > > return before it starts init as pid 1
> > > root=/dev/ram makes the kernel realise that the ramdisk si a
> > > real ramdisk, and not an initrd, so it disables the running of
> > > linuxrc
> > >
> > > So we need to find a way to force this parameter.
> >
> > Isn't this the same as #116829 and wasn't it fixed in sysvinit 2.84-1 ?
>
> Hm, yes, seems to be the same problem. I'll try to find a tester,
> since netboot doesn't work for me.
>
> Falk
>
>
>
> --
> To UNSUBSCRIBE, email to [EMAIL PROTECTED]
> with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
>
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
apt-setup: db_input critical apt-setup/uri_type || exit 0 (Noninteractive)
I edit dbootstrap_settings to be : SUITE=unstable db_set apt-setup/hostname sunsite.cnlab-switch.ch db_set apt-setup/directory /ftp/mirror/debian/ db_set apt-setup/mirror sunsite.cnlab-switch.ch db_set apt-setup/country Switzerland db_set apt-setup/uri_type http db_fset apt-setup/mirror seen true db_fset apt-setup/uri_type seen true db_fset apt-setup/country seen true db_fset apt-setup/directory seen true And then run DEBIAN_FRONTEND=Noninteractive base-config apt-setup always fails to create an /etc/apt/sources.list If I change the line db_input critical apt-setup/uri_type || exit 0 to db_input critical apt-setup/uri_type || true then it creates a valid apt-setup and everything works. I think this is because debconf (or the non-interactive frontend) skips db_input questions, but then db_get retrieves the value from the config.db which has been preset. (The point of noninteractive frontend no ?) So, should the noninteractive frontend placate db_inpute somehow, or should developers and apt-setup in particular not use the db_input critical apt-setup/uri_type || exit 0 construct ? (the exit 0 part) Is this a bug against debconf or apt-setup (base-config) ? Kind regards, Peter --- Begin Message --- This barfs. Remove the exit 0 and things proceed normally on DEBIAN_FRONTEND=Noninteractive. S, Noninteractive frontend is broken. It should not SKIP the question. Or.. apt-setup should not exit 0. I am confused. Peter --- End Message ---
Re: remote installation
> What I'm trying to do is access the PC remotely and set it up from > there. According to the normal installation, I would have to configure > the network card as part of the install (which would be impossible). Not imposssible. Look for the linuxcare business card cd. You can download it I think. It sets up a box with ssh on it. You can configure the network at boot time if you know what NIC you have in advance. Just compile the driver into the kernel and set the IP address on the kernel command line. See /usr/src/linux/Documentation for more info on that part. So you should be able to automate the ssh install that the debian boot/root medium (business card cdrom) does.. As long as you can automate the isntallation of sshd (including running ssh-keygen) and you have a password preset then you can get in remotely and do the rest by hand. (fdisk, mke2fs, debootstrap) Hmm. Maybe I'll post somethign more up to date later tonight. Peter > > So, I'm leaning towards setting up a CD with exactly what a teeny Debian > install might have (right before you would run your dpkg-reconfigure > base-config, I think) and copy that into place on the hard drive. > I will know well in advance the necessary drivers, IP information, et al > to get this working. I just do not have any access to a keyboard. > > Originally I was looking for something like the SuSe script to save a > complete installation 'map' of all the packages and entries to the > dbootstrap questions on a floppy and have the person at the other end > just run that as the installation. That way they wouldn't have to do > anything requiring significant thought or time. > But I can't find any such thing for Debian. > So, I'm here. > -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
debootstrap-udeb - should I make it ?
Since it would help clean up my build process for my initrd based
on debian-installer, shall I post patches here to add a -udeb section
to the debootstrap*/debian/{rules,control} files ?
Would they be welcomed ?
Seems all that needs to be done is to not run dh_installdocs and
dh_installman.
Peter
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Bug#116991: no subject)
> Now, I can do > # chmod ug+w /dev/loop* > # mount -o loop rescue.bin /floppy > and get > Could not find a spare loop device (repeated 3x) > Mounting rescue.bin on /floppy failed: Success > (and no, /floppy is empty and mount does not show that it has been > mounted.) > > Any ideas what is wrong? Use losetup and you will see your loop devices are getting gobbled up with each try until you have no more. Maybe it fails in the first place because you don't have msdos/fat fs support in your kernel ? And of course you have to have loopback support. I think I have had this problem before once upong a time and it was the filesystem support but it was too long ago now to be sure. Peter -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Re: boot-floppies - mac-fdisk vs. parted
mac-fdisk is a steaming pile of crap. parted is much much better I think we should replace mac-fdisk in base with parted. Peter On Wed, Oct 24, 2001 at 04:01:08PM -0800, Ethan Benson wrote: > On Wed, Oct 24, 2001 at 04:50:33PM -0700, Christoph Lameter wrote: > > Then we would probably need arch specific default partitioning strings. > > > > Do all arches support fdisk? Then we could simply have different schemes > > depending on arch. > > no, macs must use mac-fdisk, m68K uses pmac-fdisk iirc (though it may > be the same as mac-fdisk, a 68K person would have to comment). > powerpcs using dos partition tables use fdisk of course. sparc uses > it as well. > > -- > Ethan Benson > http://www.alaska.net/~erbenson/ -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
debootstrap: mirror_style={main|release} ?
In debootstrap, at the top of the woody and sid scripts I see the parameter mirror_style=release. Reading /usr/lib/debootstrap/functions I see the options are main or release. Tracing down further I see that release gets the Release files, and apparently main does not - when called as download_main or download_release respectively. May I ask what the philosophy underlying having these alternatives are ? In particular I am trying to do installs from my own partial mirror created with apt-move and I am runing into the problem of having to supply Release files. I would prefer not to use debmirror which would provide release files as I want to install ssh and libssl0.9.6 on my base installs so I can take over intallation remotely on the reboot. (I have 70 nodes to administer) I suppose I could create two mirrors, one of main and one of non-US and call debootstrap twice, once with woody as SUITE and one with ssh as SUITE after writing a scripts/ssh. Would I be better off to incorporate ssh into a scripts/woody-custom and maybe get away wth just changing the mirror_style to "main" from release ? Would this conceivably work ? Any suggestions greatly appreciated. Thanks, Peter -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
