Re: [Bug 1479805] Re: /etc/bash_command_not_found is utterly broken

[CarstenHey]

Hi,

Ubuntu's zsh c-n-f foo is fine now and c-n-f integration for bash is
done in the bash package, which is fine too (maybe this bash completion
c-n-f thing documented in the bash completion package could be added,
but I don't know how to trigger this specific bug which runs the c-n-f
magic in the bash completion - only how to prevent it).

* Steve Langasek [2016-06-17 14:57 -]:
> It doesn't appear to be true that this bug is currently in progress.
>
> However, I'm also not sure if there's anything further that needs doing
> for Ubuntu.  The bzr history shows:
>
> committer: Michael Vogt 
> timestamp: Mon 2015-07-20 15:08:23 +0200
> message:
>   merge /etc/zsh_command_not_found: from Carsten Hey
>
> But the timestamp on this is before the patch that was posted here.

Two different patches doing roughly the same were posted in different
ways ...  I don't think that the reason is relevant, but I could explain
it if you really want to know.

> And the committed result is somewhat different than the one attached
> to this bug report.

Both patches are good, but not perfect (reaching perfection would be
quite tricky and the gain would be very small, for instance, a useless
stat(2) could be avoided).

I did not read the two patches now, anyway the later one contains very
very very minor improvements, either by catching some really weird
corner cases, or by being shorter, or by being more readable, or ... but
this is hardly a reason to invest any time in merging.

> Carsten, are any further changes needed to the zsh file?

No related changes are needed in Ubuntu.

> ** Changed in: command-not-found (Ubuntu)
>Status: In Progress => Incomplete
>
> --
> You received this bug notification because you are subscribed to the bug
> report.
> https://bugs.launchpad.net/bugs/1479805
>
> Title:
>   /etc/bash_command_not_found is utterly broken
>
> Status in command-not-found package in Ubuntu:
>   Incomplete
>
> Bug description:
>   In a nutshell, if the package command-not-found is removed, but not
>   purged, sourcing this file leads to silent failures if a command is
>   not found.
>
>   I don't use Ubuntu, but I assume that the file bash_command_not_found
>   in the bzr repository is installed to /etc
>
>   ./x is your bash_command_not_found with /usr/ replaced by /ur/ to
>   simulate that the packages is removed but not purged.
>
>   I use a self written printexitvalue (see tcsh and zsh), which prints
>   the line "bash: exit $?" via $PROMPT_COMMAND - this is unrelated to
>   the bug, but shown in the output below.
>
>   $ unset -f command_not_found_handle
>   $ cat x
>   command_not_found_handle() {
> if  [ -x /ur/lib/command-not-found ]; then
>/ur/lib/command-not-found -- "$1"
>return $?
> else
>return 127
> fi   
>   }
>   $ foo
>   bash: foo: command not found
>   bash: exit 127
>   $ . ./x
>   $ foo
>   bash: exit 127
>   $
>
>   As you can see, trying to run the non-available command foo results in
>   no output (except of the one I do in $PROMPT_COMMAND) and correctly
>   fails with exit code 127. It does not print "bash: foo: command not
>   found", as it should.
>
>   command_not_found_handle() is not command_not_found_handler() (note
>   the trailing 'r' in the word handler, that's not in the word handle).
>   man bash (for the former function) and man zshall (for the latter
>   function) contain the glory details how both functions work and how
>   they differ. If /usr/lib/command-not-found is available, your
>   command_not_found_handle() implementation behaves as it should, but if
>   /usr/lib/command-not-found is not available it behaves as zsh's
>   command_not_found_handler() should, but not as a
>   command_not_found_handle() should.
>
>   An untested, but presumably correct implementation, except of the
>   unconditional overwriting of the function (I don't know how to prevent
>   this in bash properly), and returning 127 when it should return 126
>   (see man p exec), based on /etc/zsh_command_not_found, is shown below.
>
>   Please replace two leading spaces with a tab, four leading spaces with
>   two tabs and so on; and don't replace "|| return $?" with "\nreturn
>   $?", in case someone uses set -e interactively. I quoted the words
>   return and builtin to guard against weird alias definitions (I did not
>   do this in the zsh snipppet because, in general, I expect zsh users to
>   have at least basic shell knowledge, and sometimes zsh users do ugly
>   things on purpose.
>
>  
>   # (c) Zygmunt Krynicki 2007,
>   # Licensed under GPL, see COPYING for the whole text
>   #
>   # This script will look-up command in the database and suggest
>   # installation of packages available from the repository
>
>   if [[ -x /usr/lib/command-not-found ]] ; then
> command_not_found_handle() {
>   [[ -x /usr/lib/command-not-found ]] || {
> \builtin printf >&2 'bash: %scommand not found\n' ${1+"$1: "}
> \return 127
>   }
>   /u

Re: [Bug 1479805] Re: /etc/bash_command_not_found is utterly broken

[CarstenHey]

* Mathew Hodson [2015-11-03 20:09 -]:
> ** Changed in: command-not-found (Ubuntu)
>Importance: Undecided => Medium

'Medium' might be too high (if anybody cares if it is low or medium):

I wrote in my initial mail:
| I don't use Ubuntu, but I assume that the file bash_command_not_found
| in the bzr repository is installed to /etc

Looks like this assumption was wrong:

:
| You have searched for paths that end with _command_not_found in suite xenial, 
all sections, and all architectures. Found 1 results.
|
| FilePackages
| /etc/zsh_command_not_found  command-not-found
|

Hence, the actual bug only affects upstream, but not the Ubuntu package,
but Ubuntu would benefit from the minor zsh improvements.

Shipping (a fixed) bash_command_not_found in the source package and
providing it in the upstream repository is still useful, for example,
a sysadmin might disable command-not-found in the system wide
bash.bashrc and source /etc/bash_command_not_found if it exists in the
skeleton .bashrc.


Carsten

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1479805

Title:
  /etc/bash_command_not_found is utterly broken

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/command-not-found/+bug/1479805/+subscriptions

-- 
ubuntu-bugs mailing list
ubuntu-bugs@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

[Bug 1479805] Re: /etc/bash_command_not_found is utterly broken

[CarstenHey]

* bash_command_not_found:
  - Print an error message if a command is not found, and the package
has been removed but not purged. (LP: #1479805)
  - Don't run command-not-found if bash_completion is used.

** Attachment added: "bash_command_not_found"
   
https://bugs.launchpad.net/ubuntu/+source/command-not-found/+bug/1479805/+attachment/078/+files/bash_command_not_found

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1479805

Title:
  /etc/bash_command_not_found is utterly broken

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/command-not-found/+bug/1479805/+subscriptions

-- 
ubuntu-bugs mailing list
ubuntu-bugs@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

[Bug 1479805] Re: /etc/bash_command_not_found is utterly broken

[CarstenHey]

* zsh_command_not_found:
  - Guard against command-not-found accidentally returning 0.

** Attachment added: "zsh_command_not_found"
   
https://bugs.launchpad.net/ubuntu/+source/command-not-found/+bug/1479805/+attachment/080/+files/zsh_command_not_found

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1479805

Title:
  /etc/bash_command_not_found is utterly broken

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/command-not-found/+bug/1479805/+subscriptions

-- 
ubuntu-bugs mailing list
ubuntu-bugs@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

[Bug 1479805] Re: /etc/bash_command_not_found is utterly broken

[CarstenHey]

I also added a check to both shell snippets (i.e., for zsh and for bash)
that prevents defining the handler function if the shell is run
interactively. Ideally, all users would set up their shell rc files in a
way that does not source the snippets for interactive shells, but users
are not perfect …. I can imagine rare cases where command-not-found is
used for debugging, and where a user might want to enable it even for
non-interactive shells. To ease this, $PS1 is checked, and not $-. The
outer if condition now is for both shells: [[ -n "${PS1-}" && -x
/usr/lib/command-not-found ]]

I also replaced all tabs with four spaces each in the zsh snippet in
order to let the file fit on an 80 characters wide terminal w/o line
wrapping.

There is still one issue that could be fixed in these files: LP:#559060.
Don't expect a patch from me for this one, although I might send one, if
I stumble over a clean solution.

http://stateful.de/~carsten/tmp/150730jFFWrFL4qo4/bash_command_not_found
http://stateful.de/~carsten/tmp/150730jFFWrFL4qo4/zsh_command_not_found

  * bash_command_not_found:
  - Print an error message if a command is not found, and the package
has been removed but not purged. (LP: #1479805)
  - Do not define the handler function if the package has been removed.
  - Do not define the handler function if the shell is interactive.
  * zsh_command_not_found:
  - Do not define the handler function if the shell is not interactive.
  - Reindent file in order to fit on an 80 characters wide terminal without
line wrapping.

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1479805

Title:
  /etc/bash_command_not_found is utterly broken

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/command-not-found/+bug/1479805/+subscriptions

-- 
ubuntu-bugs mailing list
ubuntu-bugs@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

[Bug 1479805] Re: /etc/bash_command_not_found is utterly broken

[CarstenHey]

* /etc/bash_command_not_found: print an error message if a command
is not found, and the package has been removed but not purged.

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1479805

Title:
  /etc/bash_command_not_found is utterly broken

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/command-not-found/+bug/1479805/+subscriptions

-- 
ubuntu-bugs mailing list
ubuntu-bugs@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

[Bug 1479805] Re: /etc/bash_command_not_found is utterly broken

[CarstenHey]

A possible changelog entry is:

  * /etc/bash_command_not_found: print error message if a command
is not found, and the package has been removed but not purged.

I think an indentation with four spaces is better for this file (in order to 
avoid line wrapping on 80x25 terminals).
This file is indented accordingly and I have replaced ']] ;' with ']];', since, 
unlike the zsh variant, it contains no ')) ;' :

http://stateful.de/~carsten/tmp/150730QuR6xwuv7Z0/bash_command_not_found

Testing the function, i.e., w/o the “if [[ -x /usr/lib/command-not-found
]] ; then” guard around the function definition, before committing seems
to be a good idea. This requires an Ubuntu installation (which I do not
have) and should be done with the package installed, and with the
package removed but not purged.

Btw., \builtin print avoids functions and aliases named printf, which
could lead to a command-not-found-error in command_not_found_handle() -
not in a bulletproof way, but for this we'd need a new POSIX release.

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1479805

Title:
  /etc/bash_command_not_found is utterly broken

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/command-not-found/+bug/1479805/+subscriptions

-- 
ubuntu-bugs mailing list
ubuntu-bugs@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

[Bug 1479805] [NEW] /etc/bash_command_not_found is utterly broken

[CarstenHey]

Public bug reported:

In a nutshell, if the package command-not-found is removed, but not
purged, sourcing this file leads to silent failures if a command is not
found.

I don't use Ubuntu, but I assume that the file bash_command_not_found in
the bzr repository is installed to /etc

./x is your bash_command_not_found with /usr/ replaced by /ur/ to
simulate that the packages is removed but not purged.

I use a self written printexitvalue (see tcsh and zsh), which prints the
line "bash: exit $?" via $PROMPT_COMMAND - this is unrelated to the bug,
but shown in the output below.

$ unset -f command_not_found_handle
$ cat x
command_not_found_handle() {
  if  [ -x /ur/lib/command-not-found ]; then
 /ur/lib/command-not-found -- "$1" 
 return $?
  else
 return 127
  fi
}
$ foo
bash: foo: command not found
bash: exit 127
$ . ./x
$ foo
bash: exit 127
$ 

As you can see, trying to run the non-available command foo results in
no output (except of the one I do in $PROMPT_COMMAND) and correctly
fails with exit code 127. It does not print "bash: foo: command not
found", as it should.

command_not_found_handle() is not command_not_found_handler() (note the
trailing 'r' in the word handler, that's not in the word handle). man
bash (for the former function) and man zshall (for the latter function)
contain the glory details how both functions work and how they differ.
If /usr/lib/command-not-found is available, your
command_not_found_handle() implementation behaves as it should, but if
/usr/lib/command-not-found is not available it behaves as zsh's
command_not_found_handler() should, but not as a
command_not_found_handle() should.

An untested, but presumably correct implementation, except of the
unconditional overwriting of the function (I don't know how to prevent
this in bash properly), and returning 127 when it should return 126 (see
man p exec), based on /etc/zsh_command_not_found, is shown below.

Please replace two leading spaces with a tab, four leading spaces with
two tabs and so on; and don't replace "|| return $?" with "\nreturn $?",
in case someone uses set -e interactively. I quoted the words return and
builtin to guard against weird alias definitions (I did not do this in
the zsh snipppet because, in general, I expect zsh users to have at
least basic shell knowledge, and sometimes zsh users do ugly things on
purpose.


# (c) Zygmunt Krynicki 2007,
# Licensed under GPL, see COPYING for the whole text
#
# This script will look-up command in the database and suggest
# installation of packages available from the repository

if [[ -x /usr/lib/command-not-found ]] ; then
  command_not_found_handle() {
[[ -x /usr/lib/command-not-found ]] || {
  \builtin printf >&2 'bash: %scommand not found\n' ${1+"$1: "}
  \return 127
}
/usr/lib/command-not-found -- ${1+"$1"} || \return $?
  }
fi

** Affects: command-not-found (Ubuntu)
 Importance: Undecided
 Status: New

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1479805

Title:
  /etc/bash_command_not_found is utterly broken

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/command-not-found/+bug/1479805/+subscriptions

-- 
ubuntu-bugs mailing list
ubuntu-bugs@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

[Bug 624565] Re: zsh tries to "exec" environment assignments

[CarstenHey]

Fix released long ago (for the original issue and the one mentioned by
mde) by using zsh's command_not_found_handler() instead of the hand
written cnf_precmd()/cnf_preexec() magic in /etc/zsh_command_not_found.

** Changed in: command-not-found (Ubuntu)
   Status: Confirmed => Fix Released

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/624565

Title:
  zsh tries to "exec" environment assignments

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/command-not-found/+bug/624565/+subscriptions

-- 
ubuntu-bugs mailing list
ubuntu-bugs@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

[Bug 520096] Re: for zsh cnf_command is populated with incorrect value

[CarstenHey]

This bug has been fixed long ago (and the fix has been released).

The problem mentioned by mde should be fixed too; and since version
0.3ubuntu15.10.0, some other issues in this file are also fixed.

** Changed in: command-not-found
   Status: Fix Committed => Fix Released

** Changed in: command-not-found (Ubuntu)
   Status: Fix Committed => Fix Released

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/520096

Title:
  for zsh cnf_command is populated with incorrect value

To manage notifications about this bug go to:
https://bugs.launchpad.net/command-not-found/+bug/520096/+subscriptions

-- 
ubuntu-bugs mailing list
ubuntu-bugs@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

[Bug 841221] Re: In zsh command not found triggers upon any subshell exit code != 0

[CarstenHey]

This bug is fixed in recent Ubuntu releases by using
command_not_found_handler() and might be closed.

Anyway, I need a place to dump old useless code I presumably never need
again (in case I'm wrong and I'll find a use case for this) ...


Actually, according to the POSIX standard and zsh's implementation, exec 
returns 127 it the command is not found at all. and 126 if a file matching the 
command's name w/o executable bit set if found in $PATH, therefore the correct 
fix for this issue would be (if it wouldn't have been fixed by using 
command_not_found_handler already), to check if $? is 126 or 127.

An untested example fix (that also fixes other things), based on the
version still in Debian, is pasted below (warning: the sequence >&-
should not be used for anything except shell builtins or commands you
wrote yourself):


[[ -x /usr/share/command-not-found/command-not-found ]] || return 0

function cnf_preexec() {
  typeset -g cnf_command="${1%% *}"
}

function cnf_precmd() {
  case $? in (126|127)
if [[ -n "${cnf_command-}" && -x 
/usr/share/command-not-found/command-not-found ]]
then
  whence -- "$cnf_command" >&- ||
  /usr/bin/python /usr/share/command-not-found/command-not-found -- 
"$cnf_command"
  unset cnf_command
fi
;;
  esac
}

typeset -ga preexec_functions
typeset -ga precmd_functions
preexec_functions+=cnf_preexec
precmd_functions+=cnf_precmd

** Changed in: command-not-found (Ubuntu)
   Status: New => Fix Released

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/841221

Title:
  In zsh command not found triggers upon any subshell exit code != 0

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/command-not-found/+bug/841221/+subscriptions

-- 
ubuntu-bugs mailing list
ubuntu-bugs@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Re: [Bug 972674] [NEW] pycompile: ImportError: No module named logging

[CarstenHey]

* CarstenHey [2012-04-03 17:40 -]:
> I did (not copy and pasted, so there might be typos):
>
>   debootstrap --variant=minbase lucid foo
>   chroot foo
>   sed -i s/lucid/precise/g /etc/apt/sources.list
>   apt-get --no-install-recommends dist-upgrade

I also mounted various standard directories and copied the whole
directory before chrooting.

The original one, which is still lucid, also contains this error, i.e.,
pycompile --version fails with the same error.

> Looks like you need to adapt the dependencies in some way to provide a
> clean upgrade path from lucid to precise.

So this bug is caused by python-minimal/lucid, which I assume won't be
fixed anymore.  Making debconf conflict with python-minimal/lucid could
fix the upgrade problem.

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/972674

Title:
  pycompile: ImportError: No module named logging

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/python-defaults/+bug/972674/+subscriptions

-- 
ubuntu-bugs mailing list
ubuntu-bugs@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

[Bug 972674] [NEW] pycompile: ImportError: No module named logging

[CarstenHey]

Public bug reported:

I did (not copy and pasted, so there might be typos):

  debootstrap --variant=minbase lucid foo
  chroot foo
  sed -i s/lucid/precise/g /etc/apt/sources.list
  apt-get --no-install-recommends dist-upgrade

The latter command failed.  Its output ended with:

  Traceback (most recent call last):
File "/usr/bin/pycompile", line 26, in 
  import logging
  ImportError: No module named logging
  dpkg: error processing debconf (--configure):
subprocess installed post-installation script returned error exit status 1
  Errors were encountered while processing:
debconf

dpkg -S revealed that pycompile belongs to python-minimal and the
dependencies suggest that this was still the version in lucid.

Looks like you need to adapt the dependencies in some way to provide a
clean upgrade path from lucid to precise.

** Affects: python-defaults (Ubuntu)
 Importance: Undecided
 Status: New

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/972674

Title:
  pycompile: ImportError: No module named logging

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/python-defaults/+bug/972674/+subscriptions

-- 
ubuntu-bugs mailing list
ubuntu-bugs@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Re: [Bug 951430] [NEW] deborphan is not multiarch-aware

[CarstenHey]

* Ralf Hildebrandt [2012-03-10 10:15 -]:
> It seems that deborphan cannot properly distinguish packages from
> different architectures.

dpkg/experimental can't properly distinguish packages from different
architectures too, but the upcoming dpkg 1.6.2 upload to Debian (this
weekend or beginning of next week) is expected to change this.

Given that the new dpkg upload results in dpkg having a consistent
interface (there is a workaround that could be applied to orphaner for
a possible insane apt-get command line interface), it's time for
a deborphan upload adjusted to the ne dpkg interface shortly after this
dpkg upload and asking for an freeze exception for Ubuntu.

After deborphan being multiarch-aware we should think about uploading
newer code, forming a deborphan maintainer team (there is already
a team, but it only has one member) and handing over deborphan to this
team.


Carsten

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/951430

Title:
  deborphan is not multiarch-aware

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/deborphan/+bug/951430/+subscriptions

-- 
ubuntu-bugs mailing list
ubuntu-bugs@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Re: [Bug 940379] [NEW] deborphan returns packages that are already uninstalled

[CarstenHey]

I just discussed and debugged this bug with Jean-Baptiste via IRC.

* Jean-Baptiste Lallement [2012-02-24 15:25 -]:
> After running this command deborphan still show the packages even if
> they are already uninstalled

The i386 variants of the packages were still installed.  The actual bug
is a human interface error, shared by dpkg in multiple places.

In my opinion, deborphan should be adapted after dpkg has been changed
to include a consistent and releasable command line interface.  Being
consistent to an inconsistent dpkg interface that might change soonish
isn't a worthwhile goal.

There is an other multiarch related bug in deborphan, it displays
orphaned packages that are installed on multiple architectures once per
architecture (without any architecture postfix).

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/940379

Title:
  deborphan returns packages that are already uninstalled

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/deborphan/+bug/940379/+subscriptions

-- 
ubuntu-bugs mailing list
ubuntu-bugs@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Re: [Bug 940379] [NEW] deborphan returns packages that are already uninstalled

[CarstenHey]

* Jean-Baptiste Lallement [2012-02-24 15:25 -]:
> After running this command deborphan still show the packages even if
> they are already uninstalled

Do you use multi-arch? If you do, the command

grep '^Architecture:' /var/lib/dpkg/status | sort -u

should print three lines.


Carsten

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/940379

Title:
  deborphan returns packages that are already uninstalled

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/deborphan/+bug/940379/+subscriptions

-- 
ubuntu-bugs mailing list
ubuntu-bugs@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

[Bug 832892] Re: [PATCH] Replace deprecated G_CONST_RETURN with const (fixes build failure with recent glib releases)

[CarstenHey]

* Carsten Hey [2011-08-24 18:52 +0200]:
> I attached an trivial but untested patch that should fix a build failure
> with recent glib releases.

This patch attached to my prior mail was made against a svn snapshot,
the patch for pal 0.4.3 (the version in Debian and Ubuntu) is included
in version 0.4.3-7 uploaded recently to Debian.  Both were generated
using:

sed -i 's/G_CONST_RETURN/const/' src/remind.c src/main.c


Fixing a build failure with recent and future glib versions (in the
meantime I verified that it does) is in my opinion a very good reason
for a new upstream release.  If you plan to do so, you could consider
applying some of the patches from [1].  At least two of them are already
in your svn repository.

 [1] http://patch-tracker.debian.org/package/pal/0.4.3-7
 (it might take some time until this page is available)


Regards
Carsten

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/832892

Title:
  pal version 0.4.3-6 failed to build in oneiric

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/pal/+bug/832892/+subscriptions

-- 
ubuntu-bugs mailing list
ubuntu-bugs@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

[Bug 832892] [PATCH] Replace deprecated G_CONST_RETURN with const (fixes build failure with recent glib releases)

[CarstenHey]

Hi,

I attached an trivial but untested patch that should fix a build failure
with recent glib releases.  I'll include it in the next pal upload to
Debian (presumably this week).

Regards
Carsten


** Patch added: "replace_deprecated_G_CONST_RETURN_with_const.patch"
   
https://bugs.launchpad.net/bugs/832892/+attachment/2308745/+files/replace_deprecated_G_CONST_RETURN_with_const.patch

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/832892

Title:
  pal version 0.4.3-6 failed to build in oneiric

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/pal/+bug/832892/+subscriptions

-- 
ubuntu-bugs mailing list
ubuntu-bugs@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

[Bug 704377] Re: nagios3 requires smbclient, tries to uninstall samba4-clients

[CarstenHey]

https://wiki.ubuntu.com/UbuntuMainInclusionRequirements has just been
clarified, the relevant part now reads:

6. Dependencies:
All build and binary dependencies (including Recommends:) must be satisfyable 
in main (i. e. the preferred  alternative must be in main). If not, these 
dependencies need a separate MIR report (this can be a separate bug or another 
task on the main MIR bug)

So there isn't a reason for tagging this bug as wontfix anymore?

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/704377

Title:
  nagios3 requires smbclient, tries to uninstall samba4-clients

-- 
ubuntu-bugs mailing list
ubuntu-bugs@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Re: [Bug 649330] [NEW] dubious deborphan output on 10.04

[CarstenHey]

* Carsten Hey [2010-09-28 00:31 +0200]:
> * baldyeti [2010-09-27 21:14 -]:
> > But currently deborphan generates a longish list of things it thinks
> > the system no longer needs; so long I am not sure I can trust it
> > anymore. Has anyone noticed something similar?
>
> Could you send me your status file ( /var/lib/dpkg/status ) privately,
> preferably compressed with gzip or bzip2?  I need this file to check
> whether some package is wrongly displayed as orphaned.

I checked the status file you sent me.  The three packages whose names
do not contain a number are ok, but that's not the reason you reported
this bug.

Using the following command I checked if the other packages are really
orphaned:

perl -00 -lne "print unless /Status: deinstall ok config-files/" status \
| egrep -v '^(Replaces|Conflicts): ' \
| egrep "$(grep '[0-9]' deborphan.out \
 | xargs printf ' %s([, ]|$)|')"nonexistent

This command will display false positives if a package name is mentioned
in the packages description and false negatives when Ubuntu uses
multiarch in future, but both is not the case with your status file.

To test this command and comparing its output with and without an
non-orphaned package (if you want to) try adding ' libc6([, ]|$)|'
(including the quotation and the space) before the word nonexistent at
the end of the command.

All packages that were displayed are really orphaned (and it would have
been surprising if they were not).  If you are able to close this bug
yourself please do so.


Thanks for your bug report :)

Carsten

-- 
dubious deborphan output on 10.04
https://bugs.launchpad.net/bugs/649330
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.

-- 
ubuntu-bugs mailing list
ubuntu-bugs@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Re: [Bug 649330] [NEW] dubious deborphan output on 10.04

[CarstenHey]

* baldyeti [2010-09-27 21:14 -]:
> Public bug reported:
>
> Binary package hint: deborphan
>
> I have a system which I upgraded from 8.04 to 10.04.
> Every once in a while I like to run deborphan and get
> rid of unneeded packages.

Do you use any command line options?

> But currently deborphan generates a longish list of things it thinks
> the system no longer needs; so long I am not sure I can trust it
> anymore. Has anyone noticed something similar?

Could you send me your status file ( /var/lib/dpkg/status ) privately,
preferably compressed with gzip or bzip2?  I need this file to check
whether some package is wrongly displayed as orphaned.

> libgnome-speech7libopenal0a libnautilus-burn4
> libffi4 libparted0  libgtkhtml2-0
> libsgutils1 libglut3libnm-glib0
> libgpod3libportaudio0   libggzmod4
> libdirectfb-1.0-0   libsmbios1  libgucharmap6
> libnm-util0 libpolkit-gnome0libx11-xcb1
> libmtp7 libraw1394-8libdns32
> libalut0libpoppler-glib2sysvutils
> libtotem-plparser10 libgutenprint2  libgnomeprintui2.2-comm
> libelfg0libiw29 libzephyr3
> gstreamer0.10-gnomevfs  libbind9-30 libconsole
> libkpathsea4libgnomeprint2.2-data   libxklavier12
> libtracker-gtk0 libavahi-core5  libneon27
> libopal-2.2 libsensors3 libscrollkeeper0
> libotr2 libmagick10 libcdio7
> libdmx1 libavahi-compat-libdnss liblwres30
> libwvstreams4.4-extras  libhunspell-1.1-0   libgmime-2.0-2
> libntfs-3g23difflibgnutls13

All but three packages have a version in the package name.  Whilst
upgrading to a new Ubuntu version these packages won't be upgraded if
they use are required in a new version, but instead an additional
package will be installed, e.g., Ubuntu 8.04 might have used liblwres30
but Ubuntu 10.04 probably uses something like liblwres33, so liblwres33
would be installed additionally to the old liblwres30.  If every
liblwres using package you have installed now uses liblwres33 instead of
liblwres30, then liblwres30 is not needed anymore and thus could be
removed.

> I have another laptop where the list is slightly shorter yet also
> includes "diff", of all things ?!?

diff is an empty transitional packages and is supposed to be displayed
by deborphan.  The command diff has been moved to an other package.

I'm not sure why sysvutils is displayed, it is indeed orphaned, but
should only be displayed if it is in the section oldlibs (I currently
don't have an ubuntu installation to check).


Regards
Carsten

-- 
dubious deborphan output on 10.04
https://bugs.launchpad.net/bugs/649330
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.

-- 
ubuntu-bugs mailing list
ubuntu-bugs@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

[Bug 543662] Re: Crashes due to wrong permissions of sudoers file.

[CarstenHey]

$ debootstrap --variant=minbase lucid lucid
$ mount -t proc none lucid/proc
$ mount -o rbind /dev lucid/dev
$ chroot lucid
r...@foghorn:/# echo -e '#!/bin/sh\nexit 101' > /usr/sbin/policy-rc.d
r...@foghorn:/# chmod a+x /usr/sbin/policy-rc.d
r...@foghorn:/# echo deb http://de.archive.ubuntu.com/ubuntu lucid main 
universe > /etc/apt/sources.list
r...@foghorn:/# echo deb http://de.archive.ubuntu.com/ubuntu lucid-security 
main universe >> /etc/apt/sources.list
r...@foghorn:/# apt-get update && apt-get dist-upgrade
...
r...@foghorn:/# apt-get install --no-install-recommends sudo
...
r...@foghorn:/# apt-get install --no-install-recommends aegir-provision
...
r...@foghorn:/# sudo true
sudo: /etc/sudoers.d/aegir is mode 0600, should be 0440
>>> /etc/sudoers.d/README: /etc/sudoers.d/aegir near line 18 <<<
sudo: parse error in /etc/sudoers.d/README near line 18
sudo: no valid sudoers sources found, quitting
*** glibc detected *** sudo: double free or corruption (!prev): 0x09025010 ***
...
r...@foghorn:/# echo deb http://de.archive.ubuntu.com/ubuntu lucid-proposed 
main universe >> /etc/apt/sources.list
r...@foghorn:/# apt-get update
...
r...@foghorn:/# apt-get install aegir-provision
...
r...@foghorn:/# sudo true
r...@foghorn:/#

-- 
Crashes due to wrong permissions of sudoers file.
https://bugs.launchpad.net/bugs/543662
You received this bug notification because you are a member of Ubuntu
Bugs, which is a direct subscriber.

-- 
ubuntu-bugs mailing list
ubuntu-bugs@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

[Bug 543662] Re: Crashes due to wrong permissions of sudoers file.

[CarstenHey]

Please ensure that you use a chroot environment or alternatively set a
root password if you install the old and unfixed package during testing.

Installing the unfixed aegir-provision 0.3-2 breaks sudo and thus
prevents you from doing any administrative task on your system
(installing 0.3-2ubuntu1 fixes this).

-- 
Crashes due to wrong permissions of sudoers file.
https://bugs.launchpad.net/bugs/543662
You received this bug notification because you are a member of Ubuntu
Bugs, which is a direct subscriber.

-- 
ubuntu-bugs mailing list
ubuntu-bugs@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Re: [Bug 605971] [NEW] editkeep should exit with an error on startup if user cannot edit /var/lib/deborphan/keep

[CarstenHey]

* krak3n [2010-07-15 17:39 -]:
> Version 1.7.28ubuntu1.  I started editkeep as a normal user and began
> selecting packages to keep.  Then I clicked OK and editkeep exited with
> error "/var/lib/deborphan/keep: Permission denied".  This was a bad user
> experience because I wasted my time selecting those packages and had to
> redo the work again after I restarted editkeep using sudo.  Preferred
> behavior would be for editkeep to exit immediately if user does not have
> permissions or flash a message that user does not have write permissions
> and then allow a read-only view.

Thanks for your bug report, I fixed the equivalent orphaner bug about
two years ago and should have thought about fixing it for editkeep too.

This will be part of the next major release of deborphan.

Regards
Carsten

-- 
editkeep should exit with an error on startup if user cannot edit 
/var/lib/deborphan/keep
https://bugs.launchpad.net/bugs/605971
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.

-- 
ubuntu-bugs mailing list
ubuntu-bugs@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

[Bug 557316] Re: /etc/apt-cacher-ng/backends_debian should have a default entry for debian

[CarstenHey]

Using geodns for ftp.debian.org would avoid raising the load of one
server if  ftp.debian.org is used as fallback.  I asked DSA if using
geodns for ftp.debian.org is planned and they responded with "no".

Nevertheless, I think using a fallback that is used just before "debrep"
is resolved is preferable to bothering users by sending problem reports
per mail.  This fallback would only be used if the user doesn't
configure apt-cacher-ng manually and chooses to skip automatic
configuration in the debconf question or if apt-cacher-ng on ubuntu is
used as cache for debian packages (or vice versa).

One idea you had was "if backend file foo is empty try foo.default in
the same directory", I think this would be a very sane solution.  You
also suggested to put ftp.debian.org into the file
backends_debian.default.

Just for clarification: ftp.debian.org is as far as I know a mirror like
every other mirror and not some kind of "debian main mirror", a quick
look at the ftpsync scripts seems to confirm this.  Please ask ftpmaster
if this is relevant for your decision and if you want to be sure.  They
might also have an opinion whether ftp.debian.org is the best fallback
mirror of apt-cacher-ng.

-- 
/etc/apt-cacher-ng/backends_debian should have a default entry for debian
https://bugs.launchpad.net/bugs/557316
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.

-- 
ubuntu-bugs mailing list
ubuntu-bugs@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

[Bug 533691] Re: manually installed packages marked as orphaned

[CarstenHey]

Deborphan does exactly  what you told it to do (check for all packages
if any other package depends on it).  The problem is the wording of the
option in gtkorphan which seems to be misleading for people not knowing
deborphan itself, though I don't have a better suggestion.

-- 
 manually installed packages marked as orphaned
https://bugs.launchpad.net/bugs/533691
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.

-- 
ubuntu-bugs mailing list
ubuntu-bugs@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Re: [Bug 347913] Re: Please change deborphan recommends to whiptail | dialog

[CarstenHey]

On Thu, Apr 02, 2009 at 04:05:10PM -, Mantas Kriaučiūnas wrote:
> Also ubuntu developers decided to include whiptail in ubuntu main
> archive instead of dialog :(

The reason for this is that the debian-installer, which was used in the
past as default Ubuntu installer, uses whiptail. Rationale is that
whiptail is thought to be smaller that dialog and space on the installer
is limited. As far as I know the current alternate Ubuntu installer is
based on the debian-installer, so there is probably no chance to remove
whiptail from Ubuntu main unless you convince the debian-installer team
to switch to dialog. The latter step would require some work and at
least you would need to provide a patch for the debian-installer and
prove that dialog including its dependencies is indeed smaller than
whiptail, not vice versa.

dialog(1)
| Comparing  actual sizes (Debian testing, 2007/1/10): The total of
| sizes for whiptail, the  newt,  popt  and  slang  libraries  is
| 757kb. The comparable  number  for  dialog (counting ncurses) is
| 520kb.

The size of the udebs is relevant in this particular case, not the size
of the normal binary packages (udeb are some kind of stripped debs). If
there are no udebs for dialog and its dependencies they would need to be
created.

> But problem is, that 3 important packages in Ubuntu depends on
> whiptail: pppoeconf and modconf depends on whiptail
> | whiptail-provider and friendly-recovery depends on whiptail
>
> So, maybe I should report RFE bugs against these packages about
> depending on whiptail | dialog (like most other packages),

There is no such thing like a RFE bug in Debian. Don't use this acronym
when talking to Debian developers, they will not know about what you are
talking.

To convince the debian-installer team to switch to dialog this is
probably a prerequisite. Please check whether those packages work with
dialog before reporting these bugs. Providing a patch will raise the
chances to get such bugs fixed. And please consider filing such bugs
against the Debian BTS for packages that are also in Debian, there is no
reason to needlessly divert from Debian for Ubuntu.

> if you think, that dialog is better, that whiptail ?

Whiptail only provides a subset of the functionality of the one dialog
provides and is IMHO less mature than dialog, or as the former deborphan
maintainer said "whiptail isn't good enough".

dialog(1)
|   whiptail is designed to be drop-in compatible with dialog, but
|   has less features: some dialog boxes are not implemented, such
|   as tailbox, timebox, calendarbox, etc.
|
| (This) ... is misleading, since whiptail also does  not  work for
| common options of dialog, such as the gauge box.  whiptail is less
| compatible with dialog than the decade-old original dialog 0.4
| program.

But whiptail also includes command line options dialog does not provide,
so dialog is not a drop-in replacement for whiptail.

> Or maybe dialog can be whiptail-provider (if dialog package included
> whiptail functionality then it could have Provides: whiptail-provider
> in debian/control file)?

Since whiptail supports e.g. --scrolltext and dialog doesn't, no, this
is not an option. Try whiptail --scrolltext --msgbox hello 10 60 and
dialog --scrolltext --msgbox hello 10 60 to see why dialog currently
can't provide whiptail-provider.

Dialog would need to ignore such useless whiptail options to be able to
provide whiptail-provider, but I doubt that anybody is willing to
maintain this whiptail compatibility part of dialog. And additional
wrapper package that provides whiptail-provider, depends on dialog and
diverts /usr/bin/dialog could do the job, but maintaining such a package
is some work and I still see no reason for anybody doing so.


Regards
Carsten

-- 
Please change deborphan recommends to whiptail | dialog
https://bugs.launchpad.net/bugs/347913
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.

-- 
ubuntu-bugs mailing list
ubuntu-bugs@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

[Bug 352744] Re: wrong encoding of man pages

[CarstenHey]

French man pages in deborphan include a note about how to report
translation bugs. The file including this note in encoded in ISO, but
the French po files are encoded in UTF-8.

It looks like po4a-translate and po4a-updatepo, which create the man
pages whilst building the package, don't like mixed encodings ...

This will probably be fixed in deborphan 1.7.29.  If my first guess is
correct (I did not check if it is) the note itself should be correctly
encoded, unlike the rest of the French man pages.


Thanks for the bug report
Carsten

-- 
wrong encoding of man pages
https://bugs.launchpad.net/bugs/352744
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.

-- 
ubuntu-bugs mailing list
ubuntu-bugs@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

[Bug 347913] Re: Please change deborphan recommends to whiptail | dialog

[CarstenHey]

It has been explained in detail why this bug can not be fixed.

** Changed in: deborphan (Ubuntu)
   Status: New => Invalid

-- 
Please change deborphan recommends to whiptail | dialog
https://bugs.launchpad.net/bugs/347913
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.

-- 
ubuntu-bugs mailing list
ubuntu-bugs@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Re: [Bug 254417] [NEW] Please sync pal 0.4.2-3 (universe) from Debian unstable (main).

[CarstenHey]

On Sun, Aug 03, 2008 at 02:08:33PM -, Albin Tonnerre wrote:
> Public bug reported:
>
> -BEGIN PGP SIGNED MESSAGE-
> Hash: SHA1
>
>  affects ubuntu/pal
>  status confirmed
>  importance wishlist
>  subscribe ubuntu-archive
>
> Please sync pal 0.4.2-3 (universe) from Debian unstable (main).

Thanks for caring about pal in Ubuntu :)

http://debian.stateful.de/pbuilder/sid/pal_0.4.3-1.dsc will be uploaded
to Debian unstable soon. 0.4.3 is a bugfix release that includes all
patches from debian/patches and an additional bugfix.

I suggest syncing 0.4.3-1 in a few days.


Carsten

-- 
Please sync pal 0.4.2-3 (universe) from Debian unstable (main).
https://bugs.launchpad.net/bugs/254417
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.

-- 
ubuntu-bugs mailing list
ubuntu-bugs@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Re: [Bug 245706] Re: Please sync hnb 1.9.18-4 (universe) from Debian unstable (main)

[CarstenHey]

On Mon, Jul 07, 2008 at 06:44:08AM -, Daniel Holbach wrote:
> I overlooked the comment by Carsten, retracting the ACK. Please make
> sure that the cursor fix is contained or at least not necessary any
> more.


Please sync ...


Another fix in included, but not the Ubuntu one. My patch would probably
be based on the Ubuntu patch when it had been attached to the Debian
bug, but it is not.


Comment by a happy user (in German):

2008-06-29 17:55:56< XTaran> carstenh: Das neue hnb ist klasse. :)
2008-06-29 17:58:38< XTaran> Cursor im Edit-Mode tut endlich wieder. :)


Patch header:

hnb (1.9.18-4)  * Fix broken cursor.  This bug was introduced with the previous
  patch which made all callbacks return void* and made hnb
  64-bit clean.  There is also a Ubuntu originated patch flying
  around which also addresses the missing cursor (see hnb
  1.9.18-3ubuntu2), but this patch has a minor bug in
  completion mode.  (Closes: #339613)


The minor bug introduced by Ubuntu I talked about in the patch header is
a off by one error in completion mode. Not changing code that you don't
need to change would prevent such bugs. ;)

-- 
Please sync hnb 1.9.18-4 (universe) from Debian unstable (main)
https://bugs.launchpad.net/bugs/245706
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.

-- 
ubuntu-bugs mailing list
ubuntu-bugs@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Re: [Bug 245706] [NEW] Please sync hnb 1.9.18-4 (universe) from Debian unstable (main)

[CarstenHey]

On Fri, Jul 04, 2008 at 10:46:35PM -, Marco Rodrigues wrote:
> Public bug reported:
>
> Binary package hint: hnb
>
> The current Debian version has the Ubuntu fix for the broken cursor.

It has not, but it has another fix for this bug. See
debian/patches/*cursor* for further details.

> hnb  (1.9.18-4) unstable; urgency=low
>
>* New maintainer.  Thanks James Morrison!  (Closes: #465888)
>* Fix broken cursor.  (Closes: #339613)
...


Regards,
Carsten

-- 
Please sync hnb 1.9.18-4 (universe) from Debian unstable (main)
https://bugs.launchpad.net/bugs/245706
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.

-- 
ubuntu-bugs mailing list
ubuntu-bugs@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs