live CD tests [2005-02-21 daily on powerpc, amd64 x 2 + ia64]

[James Troup]

Hi,

I'm testing hoary on our servers this week, and this is the first
round of results, a test install of the current (2005-02-21) daily
live CD on all our non-i386 architectures.

-- 
James

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

Testing on Apple G5 (powerpc)
-----------------------------

 . The yaboot text refers to 'linux' as a boot option but it's
   actually 'live'. [Reported to Colin and fixed]

 - The yaboot text WRT sub-architecture selection is confusing and
   should at least refer to more common (i.e. Apple) hardware first
   and/or more prominently. [I'll file a bug]

 - The whole sub-architecture selection thing is crack and should be
   done away with, unfortunately it's non-trivial according to
   upstream, and as time goes on, they're going to be less and less
   inclined to do it (powerpc64 being the future and all).

 - During boot, I saw 'On battery, not fscking' which is patently not
   true (it's a rack-mounted G5, it doesn't have a non-AC power
   source).  This is because PMU doesn't work on G5s (see
   http://people.debian.org/~terpstra/message/20050218.043453.b820f5e2.en.html).
   Given this is unlikely to be resolved upstream before hoary, would
   adding a machine-check to on_ac_power and friends be too nasty?

 - New keyboard selector prints warning when user pressing shift key.
   Reported to Matthias, he's going to clarify the instructions to
   make it clear that you just have to press the key with the symbol
   on it, not any of the modifiers you'd normally need.

 - The new keyboard selector's timer makes you feel under pressure.
   Reported to Matthias who is considering making it go down in ten
   second increments until you're in single digits when it'll revert
   to one second decrements.

 . Icons on the desktop SNAFU.  Known problem with gtk-icon-cache on
   BE arches.

 - Non-X Console was coming up as 'Resolution Out Of Range' - maybe
   the 'video=ofonly' kernel command-line argument didn't get passed
   through to the second stage?

 - Shutdown ... didn't.  [See below about eject for why, I suspect...]

Testing on HP Proliant DL145 (amd64)
------------------------------------

 - No bootsplash logo appeared at the CD boot prompt.

 - I noticed the help text refers to lots of obsolete i386 drivers that
   probably don't work on amd64 era machines.  Worth a bug?

 - There are lots of very slow stages (e.g. 'storing language',
   'configuring user'); on a machine as fast as concordia, that seems
   very strange?

 . No X res auto-detection [Well known problem.]

 - Old 'Can't access system hardware clock' spew during boot

 - 'vbetool command not found' scrolled by during boot

 - Update notifier has a super blurry icon on startup

 - Once the desktop is up, I get an 'Error' dialog saying 'No volume
   control and/or devices found'.  I'm not convinced not having sound
   hardware justifies throwing scary error dialogs at the user?

 - While testing synaptic, noticed there's no libgnome2-perl on CD, so
   no gnome debconf gnome frontend.  Bug?

 - /etc/resolv.conf looks like it came from our build machines - and
   probably really shouldn't.

 - When select restart, the reboot hangs on a 'Press enter' prompt after:

   'eject: unable to eject, last error: invalid argument'

   The CD didn't manage to eject.

Testing on an IBM E325 (amd64)
------------------------------

 + bootsplash appears on this amd64 box

 - Noticed some segfaults in /var/log/syslog: klogd and debconf

 . mptscsih wasn't auto-loaded [Already reported/fixed as #6786]

 - CD managed to eject, but still go the same error message from
   eject(1) and hung at promptness.

Testing on HP RX1600 (ia64)
---------------------------

<elmo> |  | Linux 2.6 [9600 baud serial console] [Live] [Text frontend]          |  |
<elmo> |  | Linux 2.6 [9600 baud serial console] [Live]                          |  |
<elmo> what's the difference? :)
<Kamion> elmo_: newt vs. readline-ish text

IMHO, we should either a) just choose one, or b) give them more useful
descriptions....

 - initial boot is clear-happy, clears the screen a good 4 times in a
   row which is painful over 9600 serial.  [minor]
 
 - The boot options explicitly offers expert, unlike other arches?

 - Even in serial console mode on a headless box, it defaults to 'Find
   your layout by pressing some keys'?

 - If I then select 'No keyboard to configure', I get 'Installation
   step failed' and it dumps me back to the main menu.
    
 - I get the full X configuration prompts.  But err, this is a
   keyboardless serial install?

Finally:

  +ÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄ+ Preparing for live session +ÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄ+   
  |                                                                         |   
  |                                  100%                                   |   
  |                                                                         |   
device '/dev/tts/0' does not exist.                                         |   
The system is going down NOW !!                                             |   
Sending SIGTERM to all processes.ÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄ+   
Sending SIGKILL to all processes.                                               
The system is halted. Press Reset or turn off power                             
                                                   System halted. 

Whoops. :)

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%