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

Bug#231083: acknowledged by developer (Re: Bug#231083: installation-reports: sarge i386 beta2, success but severe glitches)

[Not sure whether you get this via Debian bucktracking, so I CC you...]

Salut Christian!

I am mightily impressed by your speedy response!

I feel like some further input/clarification is needed from my end:

On Thu, 2004-02-05 at 03:03, Debian Bug Tracking System wrote:
> > Initial install from CD:
> > - 'Detecting hardware' runs *three* times all in all. Can't we bring that down
> >   to 1? (Maybe the fact that I don't have a floopy is responsible.)
> I'm not really sure the modular design of d-i isn't involved
> there. The hardware detection step does not detect the same stuff in
> all these steps.
> If that's true, then the dialog should be changed to "Detecting needed
> hardware"....
> I assign this to ddetect (-1 bug above).
Just to elaborate on this one: It surely says something like 'Loading
module floppy for Linux Floppy each of the three times. Couldn't we do
one hardware detection run and stoe the result?

> > - name server should default to gateway rather than being blank IMHO (if no DHCP)
> Give that most user use to hit Enter too quickly there, I'm definitely
> not in favor of this change.
Hm, I don't see the difference with regards to hitting Enter too

Currently: User hits enter and has no name resolution.

My suggestion: User hits enter and at least *might* have name resolution
in case the gateway is also the name server which you get quite
frequently, e.g. with little firewall/router boxes. 

Bottom line: My suggestion should get at least some more people flying.

> But let's record this as a wishlist against netcfg
> > - the hostname is being asked for twice, not good
> Known bug. Fixed (#227897)
> > - domain not asked for, not good
> Well, you're right about that. resolv.conf should be then written, probably.
> > * there is no feedback during filesystem and mountpoint creation, not good
> >   ('Creating filesystems and mount points - please, stand by' would suffice)
> You're right about current partitioner module. However, the future
> partitioning tool (partman) has this feature, as well as the
> autopartitioning module. If you're OK, I don't
> record this as a bug.
This is IMHO not related to partitioning. Partitioning is finished once
I quit cfdisk. What I'm referring to is filesystem creation, e.g.
mkfs.ext3 <something> and subsequent mounting. 
> > * there is no possibility to do a badblock scan of the HDD, not good
> Sure, None of the partitioning tools offer this option.
> The best would be a dedicated module, just for avoiding implementing
> this in several modules.
> I assign this to debian-installer as a wishlist for a new module
AFAIK this has nothing to do with partitioning. Parameter -c of
mkfs.ext[2,3] will do a badblock scan prior to creating the file system.
I would assume sonething similar exists for xfs.

Note that doing a badblcok scan is potentially time consuming for large
disks and/or slow hardware. My suggestion would to make it optional but
to definitely offer it. An alternative could be to use the badblocks
program (part of package e2fsprogs) and have an additional bootmenu
entry for harddisk scanning.

> > - 'Installing base system' stops at 75% and finishes, cosmetic but irritating
> This bug belongs to the base-installer package, imho
> > - 'Installing extra packages' stops at 66% and finishes, cosmetic but irritating
> >   (less than above because quicker)
> Probably the same.
> > - install does not try/offer to setup Australian Telstra Bigpond cable modem via
> >   bpalogin (yeah, might only be me but than again not...)
> Well, this looks like a pretty specialized demand....Dunno if
> recording a bug will trigger something.
Fair enough. But maybe the bpalogin package could at least be added to
the CD so that allow manual configuration? Wishlist? ;-)

> > 
> > After reboot:
> > * I get a message saying: Invalid hostname "aurich3" and some blurb - aurich3 is
> >   definitely a perfectly valid hostname!
> No idea where this comes from
I tried the install three times and it consistenyl happens every time.
Maybe it chokes on the number? 
> > - hotplug is not part of CD - I'm surprised, there's is probablt a reason, though
> I think you mean it should be installed *by default* on any Debian
> system.
> Opinions may differ about this. However, having hotplug installed as
> part of the "Desktop" tasks could be interesting as Desktop
> Workstations are probably the real target for this package.
> I assign this a a wishlist bug against the hotplug package. I guess
> that hotplus package maintainer will have a more precise idea of what
> to do....
> > * lilo config should default to offering choice so that people can specifiy things
> >   like 'single'
> This is #229211, extended.
If anything, I'd rather say it's #229211 much reduced (although I agree
with #229211 in regards to presering access to already installed OSes):

All I'm really after is to have:


in the installed standard lilo.conf to give me an opportunity to provide
kernel parameters.

A related addition: I've tried to use the root= parameter when booting
off the CD to boot into the system on the harddisk off the CD. I haven't
been able to work out the correct expression for root=, though. I've
linux root=/dev/hda2 (which IMHO should work even with devfs, right?)
linux root=/dev/ide/host0/target0/lun0/[disc2,part2,p2]

All for attempted ended in the same kernel panic asking me to properly
specify the root= parameter.

I have added prompt and timeout=20 to lilo.conf. Upon rerunning lilo, I
got the following warning:

Warning: '/proc/partitions' does not match '/dev' directory structure.
    Name change: '/dev/ide/host0/bus0/target0/lun0/disc' -> '/dev/hda'

Not sure whether that is related...

So, I would like to know: Is the root= parameter supposed to work when
booting off the CD? If yes, ho does one use it? (I herewith volunteer to
write something up for the help screens, maybe under F3, if someone
explains to me what needs to be specified.)

> > 
> > After installation:
> As I said, these are well-known "testing is outdated" issues....
> > 
> > Bottom line: Debian is truely great (tried RedHat, Mandrake, Suse) - but still
> > only for the initiated - not read for the average desktop user :-(
> Hmmm, I beg to disagree a bit there. See #228553, for instance....
Hmmm back. ;-) I quote from #228553:
"No real dramatic problem was found during this installation. However, 
I'm under the feeling that the installation wouldn't have gone
smoothly if I have not been here..:-)

The user also ended up with a badly configured X server and a console using
a US keyboard layout. Not exactly hype when compared to other distros..:-)"

Regarding what the author of #228553 writes about DHCP:
Why do we have to show any DHCP screen at all? Can't we do the polling in background
after the hostname has been entered? And evaluate the result later and only show
what's necessary:

1) Choose language (unchanged)
2) Choose keyboard (unchanged)
3) Ask for hostname
\_detect NIC, kick off DHCP in background
4) Hardware detection (all)
_/evaluate DHCP outcome:
  If DHCP worked, i.e.:
    - IP address assigned
    - netmask known
    - gateway known
    - name servers known
    - domain known
    5) Partitioning
  ElseIf DHCP completely failed
    5) Ask whether to try again (existing screen)
  Else DHCP worked but not everything provided
    5) Inform user, e.g.:
       The network configuration via DHCP was only partly successful. You will now
       be asked to provide the missing information. If unsure, ask your network
    6) Ask for first missing bit

Best regards & thanks a lot again
Andree Leidenfrost
Sydney - Australia

Reply to: