Beta4 on sparc32 - mostly successful

[Jurij Smakov <jurij@wooyd.org>]

Hi,

Here's a report of beta4 testing on sparc32 (install on Sparcstation 10
machine via serial console). Some of this was reported on #debian-boot
already, so a solution for some problems might be already in the works.

* netboot image [0]

  Boots fine, installation goes OK, until the kernel installation stage.
  Then installer incorrectly chooses to install the 64-bit kernel, which
  bails out with an error. According to Joshua Kwan, this happens due to
  the fact that the version of kernel images for sparc which d-i relies on
  has been removed from testing. Logs of this installation are available
  at [1].

* netinst CD iso image [2]

  A completely functional install which went impressively well! Few minor
  notes:

  - initrd is too big to fit into standard 4 MB ramdisk. Thanks to the
    tip of Steven Langasek and Joshua Kwan managed to boot by passing
    a boot parameter 'ramdisk_size=8192', otherwise was just getting
    a bunch of 'attempt to access past the end of device' messages. This
    probably needs to be documented somewhere.
  - DHCP failed the first time, but it seems to be a known and minor
    issue.
  - during the second stage (after reboot) the messages from init about
    IDs 2 and 3 (virtual consoles) respawning too fast are occasionally
    displayed, ruining on-screen pseudographics. Might be a good idea to
    redirect those messages to the logs.

* businesscard CD iso image [3]

  - same initrd/ramdisk size problem as for netinst iso (see above).
  - since the kernel image for this installation is downloaded from the
    net, the attempt is made to install 64-bit kernel on a 32-bit machine
    which naturally fails (same as for netboot).

The only thing which still does not work for me is non-English (in
particular, Russian) install. I've made some digging on it and it seems
like the correct (Unicode-encoded) strings are passed to the newt frontend
and then something screws up there. One thing which I cannot quite explain
is why the Russian text is all Unicode-encoded, but the locale in
languagechooser's languagelist file seems to be set to ru_RU.KOI8-R.
Anyone cares to comment on that?

Otherwise, compliments to d-i team on an impressive release.

Best regards,

Jurij Smakov                                        jurij@wooyd.org
Key: http://www.wooyd.org/pgpkey/                   KeyID: C99E03CC