[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Install Report (b-f 2.3.6 i386; floppy/net; standard)


(NOTE! Woody boot-floppies are *not* for the faint hearted!)

Boot-Floppies: 2.3.6 (2001-06-21)
Architecture: i386, idepci flavour
Method: floppy install with rescue, root, drivers;
        net install (ftp.planetmirror.com.au; proxied) for base and standard

Machine: ACER TravelMate 520iT
Processor: Pentium or better
Memory: 64MB
Root Device: IDE (hda)
Root Size: hda 4GB; hda3 (root) 400MB

Base System Installation Checklist:

Boot Complete      [X]
Keyboard Config    [X]
Create Partitions  [X]
Install kernel     [X]
Install drivers    [X]
Config drivers     [ ]
Config Network     [X]
Install Base       [X]
Config Base        [ ]
Create boot disk   [ ]
Reboot             [X]


This was the third machine I tried. The first was a random (and old) laptop
with a PCMCIA netcard. That failed, since the current boot-floppies don't have
any PCMCIA support, and the modules in unstable aren't compatible with the
2.2.19 kernel in unstable. The second was a desktop with an old ISA 3c509
network card. Adding the module in autodetect mode succeeded, but the card
then didn't actually work.

First problem was getting some disk space claimed back from the Win98
install on that machine. Defrag and FIPS *SUCK*. Win98 defrag is just
plain broken (in that it doesn't seem to effectively defragment files,
nor do much good at moving files away from the end of the partition);
and FIPS without an effective defragger isn't much good. GNU PartEd,
otoh, is much more useful; we should at least provide some parted.bin
disk images, or something. (An image you could either boot or mount on
/mnt and use would go down well, imo)

On this machine, at least, the internal eePro netcard got automatically
detected without any modules or anything.

dbootstrap occassionally seems to pop up a message about something, then
immediately skip on to the next thing without giving you any time to read
it. That's horrible.

Oddly, I got an option to use NFS to get the kernel & drivers, even though
I hadn't set up my net card at that point. Seemed odd, didn't choose it to
see if it worked, or died though.

It's also confusing to have that called "Operating system and modules"
rather than "Kernel and drivers" --- I always think that means "base
system", and get confused. Also confusing is the "rescue" and "root"
naming of the disks. "kernel" and "root", or "boot" and "root", or
"boot-1" and "boot-2" would be more obvious, IMO.

When I got to the base install, I got told to use "testing". Fine, thought
I, so I did. Then I got asked "Am I sure I want to do this?" with "No" as
the default. That seemed weird.

debootstrap definitely needs to give a progress display with bytes downloaded
or similar.

And, basically, b-f's worked. So I rebooted (with the rescue floppy; I didn't
care to have YA floppy, nor to put LILO on that machine).


Installation process: HTTP, proxied
Package Choice: tasksel + dselect

Installation experience:

base-config didn't keep what I told it, and asked me to choose a mirror
again, and went straight ahead and overrode my decision to use testing
and tried getting stable Packages files instead.

On the other hand, it's mirror selection stuff is much more pleasant: the
by country menus should've been in dbootstrap.

base-config also seemed to be a little confused about my proxy settings. It
managed to get them right some times (I think), but not others. Not quite
sure what the deal was there.

base-config shouldn't be forgetting things I've already told dbootstrap.

Weirdly, the minimal boot-floppies environment is actually more functional
than base. There's no wget equivalent in base.

Anyway, running the apt-get dselect-upgrade by hand after quiting base-config
and setting my proxy worked pretty well. There are a few things that are
standard that probably shouldn't be: gcc-3.0, nfs-kernel-server, xlib6g,
the debconf/stool stuff, vacation and rblcheck all seem a bit unnecessary.

I didn't get most/any of base dpkg-reconfigured.

gpm and iamerican seem to be the main packages in standard that insist on
prompting during the install and aren't debconfed. There were a couple of
complaints about /etc/mailcap being missing, and a few others too, which I
didn't note down.

The end result was 234MB used, which dropped to 174MB used when I ran apt-get

The only bad permissions I could find were in /dev, which was owned by
root.aj, and had a bunch of symlinks which had group aj too. Not sure where
that comes from: debootstrap's devices.tar.gz seems fine at first glance.

It'd be nice if more of this report was automated for me: having base-config
notice I'm installing testing/unstable (rather than stable), and fill out
a template of this report from /proc/cpuinfo, uname -a, dbootstrap_settings,
and whatever else would be kind of nifty.

Rebooting to Windows then tells me that the bastard install has decided
to fuck up my clock and put it 10 hours ahead, even though I'm pretty
sure I told it that the hw clock is *NOT* in GMT. Bleh.

Anyway. It worked. Relatively painlessly too.

So by the looks of things, woody b-fs *do* support:

	i386 net installs
	i386 cd installs (partial)

	powerpc net/cd installs (CVS only)

and don't support:

	other architectures
	fully floppy installs
	installation from base tarballs
	pcmcia devices


Attachment: pgpCaun17BmHj.pgp
Description: PGP signature

Reply to: