asks termnal question on dapper->hardy upgrade

Bug #174002 reported by Michael Vogt
2
Affects Status Importance Assigned to Milestone
glibc (Ubuntu)
Fix Released
High
Matthias Klose

Bug Description

On a dapper->hardy upgrade libc6 asks if gdm should be stoped in the terminal about "Name Service Switch update in the C library; pre-installation question" (it asks without debconf).

This questions should not be asked by default, we should do something by default here.

Michael Vogt (mvo)
Changed in glibc:
milestone: none → ubuntu-8.04-beta
Revision history for this message
Matthias Klose (doko) wrote :

libc.preinst asks to restart services which cannot be restarted automatically. Surely we can comment out this test and the dist-upgrader hints the user to restart these services after an upgrade/system restart. Just asking with debconf would force the restart of gdm, which might not be the action we do want to have, while upgrading in the graphical environment.

Changed in glibc:
status: New → Incomplete
Revision history for this message
Jeff Bailey (jbailey) wrote :

Hmm. We haven't done this with debconf before, since debconf isn't essential in Debian. However, I do notice that it's part of ubuntu-minimal, so we can probably consider that tool here.

Do the dist-upgrader hints get through to someone on a server-based installation? In practice the things that need to be restarted will affect server users more than desktop users.

Otherwise, it looks like we dropped the patcht to the preinst to force DEBIAN_FRONTEND to be noninteractive for glibc. If debconf and dist-upgrader aren't solutions for us, we can just restore that hack.

Revision history for this message
Michael Vogt (mvo) wrote :

I don't really know what the right course of action is here, but I think that we can not ask this type of question during the upgrade. It is recommended to reboot after the ugprade so whatever the safe course of action is should be taken.

Cheers,
 Michael

Changed in glibc:
importance: Undecided → High
status: Incomplete → Confirmed
Revision history for this message
Matthias Klose (doko) wrote :

that should be gone in the -9ubuntu1 merge

Changed in glibc:
assignee: nobody → doko
Revision history for this message
Steve Langasek (vorlon) wrote :

Michael,

How do you define which is the "safe" course of action here - is it safer to immediately reload gdm, or to leave it up to the user to reboot and possibly have graphical logins stop working until that happens?

Did the dapper gdm allow all users to reboot the system from the login screen, by default? Did it support the option in its init script to reload its config after logout when sent SIGUSR1?

Revision history for this message
Colin Watson (cjwatson) wrote :

Moving to 8.04 final; this isn't going to be fixed for beta now.

Changed in glibc:
milestone: ubuntu-8.04-beta → ubuntu-8.04
Revision history for this message
Matthias Klose (doko) wrote :

this was fixed in -9ubuntu1/2 (unless debconf or cdebconf is not installed), although the priority is still at `critical'. What could we do further?

 - lower the severity
 - don't show the question at all, if we can create an update-notifier, and use this one instead.

Revision history for this message
Launchpad Janitor (janitor) wrote :
Download full text (4.5 KiB)

This bug was fixed in the package glibc - 2.7-10ubuntu1

---------------
glibc (2.7-10ubuntu1) hardy; urgency=low

  * Merge remaining changes from 2.7-10 (r2869:2892).
  * If RELEASE_UPGRADE_MODE is set to `desktop', make the glibc/restart-services,
    glibc/restart-failed and glibc/upgrade questions of medium priority (and
    restarting the services automatically). LP: #174002.
    Works as well around the upgrade errors mentioned in LP #205079 in a KDE
    environment, when the upgrade is done using the update-manager.
  * debian/rules: Always use the package settings for *FLAGS, not the settings
    from the environment.
  * Fail the build if regressions are found running the testsuite compared to
    expected results from debian/expected_test_summary. Take initial values
    from a PPA build predating this upload.
  * Merge from Debian trunk:
    - local/manpages/ld.so.8: fix libraries search order. Closes: #473458.
    - Update Finish debconf translation, by Esko Arajärvi. Closes: #473802.
    - Add any/cvs-strerror_r.diff to make strerror_r actually thread safe.
      Closes: #456531.
  * debian/rules.d/build.mk: Call `sync' before building the tarball of
    supported locales. Our buildds trigger http://lkml.org/lkml/2007/8/1/337
    on every build.

glibc (2.7-10) unstable; urgency=low

  * Update Chinese debconf translation, by LI Daobing.
  * Update Dutch debconf translation, by Bart Cornelis. Closes: #468300.
  * Update Polish debcon translation, by Emilian Nowak.
  * Update Vietnamese debconf translation, by Clytie Siddall. Closes: #468453.
  * Update Basque debconf translation, by Piarres Beobide. Closes: #468463.
  * Update Romanian debconf translation, by Stan Ioan-Eugen. Closes: #468671.
  * Update Czech debconf translation, by Miroslav Kure. Closes: #468910.
  * Update Russian debconf translation, by Sergey Alyoshin. Closes: #469037.
  * Update Korean debconf translation, by Sunjae Park. Closes: #469097.
  * Update French debconf translation, by Christian Perrier. Closes: #469249.
  * patches/localedata/locale-es_CR.diff: chane AM/PM format to "a.m./p.m."
    from "A.M/P.M.".
  * Remove uk_UA from patches/localedata/first_weekday.diff: merged upstream
    in a different way. Closes: bug#469439.
  * patches/any/cvs-rfc3484.diff: update from CVS to allow the definition of
    IPv4 scopes in /etc/gai.conf.
  * patches/all/cvs-gai_conf.diff: update gai.conf from CVS to also mention
    Teredo tunnels and scopev4 defaults.
  * sysdeps/mips.mk, sysdeps/mipsel.mk: define TIMEOUTFACTOR.
  * patches/any/cvs-strtod.diff: new patch from CVS to fix strtod(). Closes:
    #465769.
  * Update Italian debconf translation, by Lica Monducci. Closes: #469985.
  * control.in/main, sysdeps/*.mk, rules: build with gcc-4.3 by default,
    except on alpha, arm, hppa, s390 and hurd-i386.
  * rules, rules.d/build.mk: fix parallel build.
  * control.in/main: build witch gcc-4.2 on i386 and amd64, as gcc-4.3 does
    not emit cld instructions in the prologue of the functions anymore.
    Closes: #469568.
  * debian/rules.d/build.mk: disable testsuite on ARM machines running a
    2.6.21 to 2.6.24 kernel.
  * Replace any/submitted-gcc-4....

Read more...

Changed in glibc:
status: Confirmed → Fix Released
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.