crash when hw-detect/modprobe_error is asked

Thanks for your report. It looks like this has the same underlying cause as bug 48005, so I'll mark it as a duplicate. However, would you mind trying to get a debug log for me? Run the installer with 'UBIQUITY_DEBUG=1 sudo ubiquity', reproduce the problem if you can, and attach /var/log/installer/syslog to this bug using the "Add Attachment" link. Make sure to use a throwaway password, as in debug mode (and only in debug mode) the password will be written to the log file.

Colin,
I finally got Dapper Drake to install and I think the cause was not being
connected to the internet...sounds funny but my linux box was hooked up to a
router that didn't have an internet connection (I've got some issues w/ my
Dlink wifi router and adsl connection). When I unplugged the lan cable, the
install went well. The only other thing different was I tried manual
partitioning at first and went w/ the defaults in the end...I don't think
that had anything to do with it because the install got past the
partitioning phase but would hang during the package installation phase. It
seemed like it was trying to check for updated software. Hope this helps.
Other than the shakey installation start, I've been having a great time w/
this version. Thanks for responding to the bug report.
Allan

On 6/7/06, Colin Watson <email address hidden> wrote:
>
> *** This bug is a duplicate of bug 48005 ***
>
> Thanks for your report. It looks like this has the same underlying cause
> as bug 48005, so I'll mark it as a duplicate. However, would you mind
> trying to get a debug log for me? Run the installer with
> 'UBIQUITY_DEBUG=1 sudo ubiquity', reproduce the problem if you can, and
> attach /var/log/installer/syslog to this bug using the "Add Attachment"
> link. Make sure to use a throwaway password, as in debug mode (and only
> in debug mode) the password will be written to the log file.
>
> ** This bug has been marked a duplicate of bug 48005
> Install failed with exit code None
>
> --
> Installer Crashed
> https://launchpad.net/bugs/48524
>

The installation did not reached the error this time. It stucked at 85% while installing ''Les paquest linguistiques''.

Been able to reproduce the problem with debug ON.
Has last time, saw the message ''cannot execute modprobe -v amd76xrom'', then saw a second window of partionning opening eventhough it was finished and the installation process was going on.
At the end, the installer told me something like this:
''installation finished continue live session or reboot with your new system''.
Pressed continue using live cd button.
Then saw another partitioning windows opening in addition to the one already opened after pressing the ook button on the modprobe message.
Closed all installer's windows with the X ont the windows

By the way what do you mean by throwaway password ?

kermit: As I've already asked you by private mail, please *file a new bug* rather than tagging logs onto existing bugs. Your first crash is the NTP problem mentioned on https://wiki.ubuntu.com/DapperReleaseNotes/UbiquityKnownIssues.

Your second log may be the same as this one, though, so thanks, and I appreciate the attempt to help, but it really makes my life a lot easier when different people file different bugs.

By throwaway password, I mean one that you don't mind telling the whole world about, as you just did with "jfara1719" by posting your debug log. :-) (I did try to warn you ...)

Looks like we need to make hw-detect/modprobe_error non-fatal.

Hi,

Got it to install by running from the command line:
"sudo ubiquity"

It did stop with a message that modprobe -v amd76xrom had failed, tried this manually ans still did not work. Ignored that error and continued and install completed.

sudo modprobe -v amd76xrom
insmod /lib/modules/2.6.15-23-386/kernel/drivers/mtd/maps/amd76xrom.ko
FATAL: Error inserting amd76xrom (/lib/modules/2.6.15-23-386/kernel/drivers/mtd/maps/amd76xrom.ko): No such device

This may give you some clues...
Thanks for your help,
Paul.

OK, thanks - but could you try that again with 'UBIQUITY_DEBUG=1 sudo ubiquity', per my request above? As before, make sure to use a password you don't mind the world seeing.

Spoke too soon, reboot failed with error:
[4294669.272000] crc error
[4294669.273000] Kernel Panic - not syncing: VFS: Unable to mount root fs an unknown - block (0,0)

So I tried again with UBIQUITY_DEBUG=1 sudo ubiquity and attached is the syslog. Also during the install (at 92%) I got:
Error while running 'modprobe - Vamd76xrom'
I chose OK and I was taken back to step 5 "Select a Disk" and then the same crash occurred.

Paul.

Thanks, Paul! That finally confirms the hypothesis I had based on earlier logs posted by you and others. I've committed a fix to my bzr branch.

ubiquity (1.1.0) edgy; urgency=low

  * Use try/finally in install.py to ensure that PROGRESS STOP is always
    called; ignore errors from PROGRESS STOP (e.g. because debconf has gone
    away due to a previous error).
  * Support /UNIONFS (Knoppix) as well as /rofs as the read-only filesystem
    source.
  * Don't require entering a full name (see
    http://www.hermann-uwe.de/blog/os-install-experiences-part-4-ubuntu).
  * GTK frontend:
    - Return None from location_from_point if point is None (thanks to Chris
      Moore for the idea; closes: Malone #48885).
  * KDE frontend:
    - Add slightly more paranoia in case getNearestCity returns None.
  * Don't show hw-detect/modprobe_error, and make it non-fatal (closes:
    Malone #48524).
  * Make it clear in the crash dialog that you should file a new bug, not
    attach your details to an existing bug.
  * Add passwd and adduser dependencies from user-setup.
  * Take a copy of the supplied choice map in set_keyboard_choices.
  * Save tracebacks from install.py and slurp them back into the traceback
    displayed if install.py exits non-zero.
  * Convert to new Python policy (python-central). Install Python
    modules/extensions to /usr/lib/ubiquity/ubiquity rather than into the
    public module path.
  * Warn in the partitioner if you try to use XFS for /boot (closes: Malone
    #47848).
  * Warn in the partitioner if system filesystems are not reformatted
    (closes: Malone #47046).
  * Remove /usr/bin/espresso compatibility symlink.
  * Use os._exit() rather than sys.exit() in the apt install progress child
    process after forking, fixing a lot of strange "PROGRESS STOP" messages
    in /var/log/installer/syslog and bizarre debconf exceptions. See also
    Malone #53298.
  * Use exceptions to communicate all failures in install.py, so that we get
    more useful tracebacks.

  * Automatic update of included source packages: apt-setup 1:0.11ubuntu1,
    base-installer 1.63ubuntu2, choose-mirror 2.04ubuntu1, clock-setup
    0.10ubuntu1, debian-installer-utils 1.32ubuntu1, grub-installer
    1.17ubuntu1, hw-detect 1.38ubuntu1, kbd-chooser 1.29ubuntu1,
    localechooser 1.20ubuntu3, partconf 1.16, partman-auto 54ubuntu1,
    partman-base 90ubuntu1, partman-basicfilesystems 47ubuntu1,
    partman-basicmethods 30, partman-ext3 41, partman-jfs 17,
    partman-newworld 13, partman-palo 4, partman-partitioning 40ubuntu1,
    partman-reiserfs 31, partman-target 42ubuntu1, partman-xfs 29,
    silo-installer 1.04ubuntu1, tzsetup 1:0.10ubuntu1, user-setup
    1.2ubuntu2, yaboot-installer 1.1.7ubuntu1.
  * Handle move of prebaseconfig scripts to (variously) post-base-installer
    or finish-install.
  * Install search-path from debian-installer-utils; partman-auto needs it
    now.

 -- Colin Watson <email address hidden> Tue, 18 Jul 2006 00:18:21 +0100

This will also be fixed in the first Dapper point release.

ubiquity (1.0.13) dapper-updates; urgency=low

  * Don't require entering a full name (see
    http://www.hermann-uwe.de/blog/os-install-experiences-part-4-ubuntu).
  * GTK frontend:
    - Return None from location_from_point if point is None (thanks to Chris
      Moore for the idea; closes: Malone #48885).
  * KDE frontend:
    - Add slightly more paranoia in case getNearestCity returns None.
  * Don't show hw-detect/modprobe_error, and make it non-fatal (closes:
    Malone #48524).
  * Make it clear in the crash dialog that you should file a new bug, not
    attach your details to an existing bug.
  * Take a copy of the supplied choice map in set_keyboard_choices.
  * Save tracebacks from install.py and slurp them back into the traceback
    displayed if install.py exits non-zero.
  * Warn in the partitioner if you try to use XFS for /boot (closes: Malone
    #47848).
  * Warn in the partitioner if system filesystems are not reformatted
    (closes: Malone #47046).
  * Use os._exit() rather than sys.exit() in the apt install progress child
    process after forking, fixing a lot of strange "PROGRESS STOP" messages
    in /var/log/installer/syslog and bizarre debconf exceptions. See also
    Malone #53298.
  * Use 'replace' Unicode error handling when dealing with data coming from
    debconf (closes: Malone #44595, #48732).
  * Fix ubiquity/install/apt_error_install and
    ubiquity/install/apt_error_remove templates to be of type 'error'
    (closes: Malone #53258).
  * Install more defence against gparted/qtparted crashing (closes: Malone
    #47194, #48856).
  * If gparted or qtparted crashes, put up a dialog allowing the user to try
    again, return to autopartitioning, or quit the installer. This is a
    pretty awful hack, but hopefully we won't be using gparted/qtparted for
    all that much longer ...
  * Fix crash in KDE frontend while translating widgets (closes: Malone
    #53367).
  * When an error occurs with the install progress bar up, only return to
    autopartitioning if the partitioner is still running.
  * Turn DebconfInstallProgress upside-down so that the main process handles
    apt installation and the subprocess handles the translation of apt
    status messages into debconf protocol messages, rather than the other
    way round. The previous approach resulted in exceptions from apt ending
    up in the wrong process.
  * Silence apt errors while installing language packs, since we ignore them
    anyway.
  * Point apt at /target/var/lib/dpkg/status rather than
    /var/lib/dpkg/status (closes: Malone #47859).
  * Display grub-installer/install_to_xfs question if it's asked.
  * Fix get_filesystems() not to assume ext3 in quite so many situations.
  * Detect XFS for validation purposes even when it wasn't just formatted by
    gparted/qtparted (closes: Malone #53642).
  * Fix all mentions of gparted in the KDE frontend.
  * Leave existing newworld boot partitions alone when committing manual
    partitioning changes.
  * Always set mirror/country to the country selected on the timezone page,
    regardless of what localechooser does thereafter (closes: Ma...

OK, thank you!

Sorry for the duplicate of this bug.

----- Ursprüngliche Mail ----
Von: Vassilis Pandis <email address hidden>
An: <email address hidden>
Gesendet: Montag, den 28. August 2006, 15:18:37 Uhr
Betreff: [Bug 48524] Re: crash when hw-detect/modprobe_error is asked

** Bug 54644 has been marked a duplicate of this bug

--
crash when hw-detect/modprobe_error is asked
https://launchpad.net/bugs/48524