base-files: md5sum: command not found

[Debian Bug Importer]

Automatically imported from Debian bug report #315784 http://bugs.debian.org/315784

[Santiago Vila Doncel (sanvila-unex)]

reassign 315784 dpkg
thanks

On Sun, 26 Jun 2005, Dan Jacobson wrote:

> X-debbugs-cc: <email address hidden>, <email address hidden>
> Package: base-files
> Version: 3.1.5
> Severity: minor
>
> Odd, worrisome messages seen,
> # dpkg -i *
> Setting up base-files (3.1.5) ...
> Installing new version of config file /etc/debian_version ...
> Installing new version of config file /etc/issue ...
> Installing new version of config file /etc/issue.net ...
> /var/lib/dpkg/info/base-files.postinst: line 120: md5sum: command not found
> /var/lib/dpkg/info/base-files.postinst: line 121: md5sum: command not found
>
> But then not on second time:
> # dpkg -i /var/cache/apt/archives/base-files_3.1.5_i386.deb
> (Reading database ... 167757 files and directories currently installed.)
> Preparing to replace base-files 3.1.5 (using .../base-files_3.1.5_i386.deb) ...
> Unpacking replacement base-files ...
> Setting up base-files (3.1.5) ...
>
> $ dlocate `which md5sum`
> dpkg: /usr/bin/md5sum
> coreutils: /usr/bin/md5sum.textutils
>
> $ dpkg -S `which md5sum`
> local diversion from: /usr/bin/md5sum.textutils
> local diversion to: /usr/bin/md5sum
>
> # apt-cache showpkg dpkg|grep base-files
>
> Well, maybe you should depend on it, as lots of others do.
>
> I'm not sure if it would solve the occasional problem that I
> experienced though. Maybe if I was using apt it would.
>
> Anyways, I seemed to have uncovered a rare case where things will go
> wrong, when one does dpkg -i * on several files at once.
>
> If I'm correct, then gnus should depend on dpkg too, as it also uses md5sum
>
> Setting up gnus (5.10.6-0.CVS.20050610-1) ...
> install/gnus: Ignoring emacs.
> install/gnus: Byte-compiling for emacs21...found emacs21
> ../usr/bin/ucf: line 554: md5sum: command not found
> /usr/bin/ucf: line 566: md5sum: command not found
> /usr/bin/ucf: line 242: md5sum: command not found
> done.
> [On 2nd try, worked]
>
> Or maybe dpkg should notice it is also upgrading itself, and thus
> upgrade and configure itself all in one step before configuring other
> files from the dpkg -i * command line.

The md5sum cmmand belongs to an essential package, so it is definitely
not a bug in base-files to assume that it should work. I reassign this to
dpkg in case the maintainer has something to say about it.

[Adrian von Bidder (avbidder)]

severity 315784 serious
thanks

With packages ranging from woody to experimental (majority is sarge), and
upgrading apt, dpkg and some others to etch, I got into the situation that
no md5sum commmand was on the system. I feel dependencies should be able
to avoid that situation.

Should dpkg Pre-Depend on coreutils >= 5.93-5 or what?

Below is some IRC conversation that convinced me that this is RC.
(deboostrap cc:d on request of mrvn)

cheers
-- vbi

>>>>>>>
<cmot> very, very stupid Q (probably) - why is there no /usr/bin/md5sum on
my system? Isn't that a command that I should be able to expect on my
system?
<cmot> (coreutils has /usr/bin/md5sum.textutils - shouldn't there be a
symlink or whatever?)
<noshadow> cmot: which release? in sarge it should be in the dpkg package
<godog> cmot: indeed /usr/bin/md5sum here is owned by coreutils
<cmot> Hmmm. I have some mix with files from woody to experimental.
<cmot> I recently upgraded apt, dpkg & co to etch, so I guess that's the
problem.
<cmot> Yep. coreutils from sarge, dpkg from etch.
<cmot> Should this be a bug?
<vorlon> probably. if possible, something should have a depends: or a
conflicts: that prevents you from getting your system into that state.
<asuffield> md5sum is Essential, it's definitely a bug of some kind
<asuffield> you're not supposed to *ever* lose those
<noshadow> though it's better to err on this side than to make upgrade paths
impossible
<cmot> Hmm. coreutils/testing doesn't have /usr/bin/md5sum either, it seems
to me?
<godog> etch's coreutils can also check /usr/bin/md5sum and symlink
appropriately
<asuffield> godog: that's postinst. Essential packages have to work without
being configured
<godog> asuffield: right, thanks for spotting that
<godog> cmot: coreutils 5.2.1-2.1 ?
<cmot> 5.2.1-2
<asuffield> why on earth did it get removed from dpkg? that's the problem
right there
<noshadow> asuffield: in what sense "essential"?
<cmot> But I just upgraded to 5.2.1-2.1 and it's only md5sum.textutils,
still.
<cmot> #289369, btw
<godog> noshadow: as in "Essential: yes" I think
<asuffield> grep-available -FPackage -sPackage,Essential coreutils
<asuffield> Package: coreutils
<asuffield> Essential: yes
<asuffield> it's a fairly simple and stupid problem, dpkg should not have
dropped md5sum until coreutils had released with it
<vorlon> or it should pre-depend on the new version
<asuffield> hmm, I suppose that would work, but it's kinda icky
<asuffield> I'd have been more inclined to do it with Replaces
<vorlon> nah, pre-depends are de rigueur in essential packages
<asuffield> anyway, somebody go file a bug about it
<asuffield> that one's RC
<cmot> And #313258
<cmot> One is minor, one is wishlist. I'll merge the two and upgrade to -
what? serious?
<godog> we should define bts arithmetics :)
<cmot> Hmm. Before I do that - anybody looked at coreutils/unstable?
<cmot> godog: :-)
<godog> 5.93-5 has /usr/bin/md5sum
<cmot> Ah. So dpkg just pre-depend on the new coreutils?
<cmot> I'll file a serious bug on dpkg, then (if there isn't one already;
haven't looked yet).
<mrvn> cmot: that would suck because it breaks (c)debootstrap
<mrvn> asuffield: replaces only works with a...

[Adrian von Bidder (avbidder)]

retitle 315784 dependencies should ensure md5sum is present
thanks
--
featured product: vim - http://vim.org

[Debian Bug Importer (debzilla)]

Automatically imported from Debian bug report #315784 http://bugs.debian.org/315784

[Debian Bug Importer (debzilla)]

Message-Id: <E1DmK5v-0001zP-Or@jidanni1>
Date: Sun, 26 Jun 2005 07:29:51 +0800
From: Dan Jacobson <email address hidden>
To: Debian Bug Tracking System <email address hidden>
Subject: base-files: md5sum: command not found

X-debbugs-cc: <email address hidden>, <email address hidden>
Package: base-files
Version: 3.1.5
Severity: minor

Odd, worrisome messages seen,
# dpkg -i *
Setting up base-files (3.1.5) ...
Installing new version of config file /etc/debian_version ...
Installing new version of config file /etc/issue ...
Installing new version of config file /etc/issue.net ...
/var/lib/dpkg/info/base-files.postinst: line 120: md5sum: command not found
/var/lib/dpkg/info/base-files.postinst: line 121: md5sum: command not found

But then not on second time:
# dpkg -i /var/cache/apt/archives/base-files_3.1.5_i386.deb
(Reading database ... 167757 files and directories currently installed.)
Preparing to replace base-files 3.1.5 (using .../base-files_3.1.5_i386.deb) ...
Unpacking replacement base-files ...
Setting up base-files (3.1.5) ...

$ dlocate `which md5sum`
dpkg: /usr/bin/md5sum
coreutils: /usr/bin/md5sum.textutils

$ dpkg -S `which md5sum`
local diversion from: /usr/bin/md5sum.textutils
local diversion to: /usr/bin/md5sum

# apt-cache showpkg dpkg|grep base-files

Well, maybe you should depend on it, as lots of others do.

I'm not sure if it would solve the occasional problem that I
experienced though. Maybe if I was using apt it would.

Anyways, I seemed to have uncovered a rare case where things will go
wrong, when one does dpkg -i * on several files at once.

If I'm correct, then gnus should depend on dpkg too, as it also uses md5sum

Setting up gnus (5.10.6-0.CVS.20050610-1) ...
install/gnus: Ignoring emacs.
install/gnus: Byte-compiling for emacs21...found emacs21
../usr/bin/ucf: line 554: md5sum: command not found
/usr/bin/ucf: line 566: md5sum: command not found
/usr/bin/ucf: line 242: md5sum: command not found
done.
[On 2nd try, worked]

Or maybe dpkg should notice it is also upgrading itself, and thus
upgrade and configure itself all in one step before configuring other
files from the dpkg -i * command line.

[Debian Bug Importer (debzilla)]

Message-ID: <email address hidden>
Date: Sun, 26 Jun 2005 17:34:55 +0200 (CEST)
From: Santiago Vila <email address hidden>
To: Dan Jacobson <email address hidden>, <email address hidden>
Cc: <email address hidden>, <email address hidden>
Subject: Re: Bug#315784: base-files: md5sum: command not found

reassign 315784 dpkg
thanks

On Sun, 26 Jun 2005, Dan Jacobson wrote:

> X-debbugs-cc: <email address hidden>, <email address hidden>
> Package: base-files
> Version: 3.1.5
> Severity: minor
>
> Odd, worrisome messages seen,
> # dpkg -i *
> Setting up base-files (3.1.5) ...
> Installing new version of config file /etc/debian_version ...
> Installing new version of config file /etc/issue ...
> Installing new version of config file /etc/issue.net ...
> /var/lib/dpkg/info/base-files.postinst: line 120: md5sum: command not found
> /var/lib/dpkg/info/base-files.postinst: line 121: md5sum: command not found
>
> But then not on second time:
> # dpkg -i /var/cache/apt/archives/base-files_3.1.5_i386.deb
> (Reading database ... 167757 files and directories currently installed.)
> Preparing to replace base-files 3.1.5 (using .../base-files_3.1.5_i386.deb) ...
> Unpacking replacement base-files ...
> Setting up base-files (3.1.5) ...
>
> $ dlocate `which md5sum`
> dpkg: /usr/bin/md5sum
> coreutils: /usr/bin/md5sum.textutils
>
> $ dpkg -S `which md5sum`
> local diversion from: /usr/bin/md5sum.textutils
> local diversion to: /usr/bin/md5sum
>
> # apt-cache showpkg dpkg|grep base-files
>
> Well, maybe you should depend on it, as lots of others do.
>
> I'm not sure if it would solve the occasional problem that I
> experienced though. Maybe if I was using apt it would.
>
> Anyways, I seemed to have uncovered a rare case where things will go
> wrong, when one does dpkg -i * on several files at once.
>
> If I'm correct, then gnus should depend on dpkg too, as it also uses md5sum
>
> Setting up gnus (5.10.6-0.CVS.20050610-1) ...
> install/gnus: Ignoring emacs.
> install/gnus: Byte-compiling for emacs21...found emacs21
> ../usr/bin/ucf: line 554: md5sum: command not found
> /usr/bin/ucf: line 566: md5sum: command not found
> /usr/bin/ucf: line 242: md5sum: command not found
> done.
> [On 2nd try, worked]
>
> Or maybe dpkg should notice it is also upgrading itself, and thus
> upgrade and configure itself all in one step before configuring other
> files from the dpkg -i * command line.

The md5sum cmmand belongs to an essential package, so it is definitely
not a bug in base-files to assume that it should work. I reassign this to
dpkg in case the maintainer has something to say about it.

[Debian Bug Importer (debzilla)]

Message-Id: <email address hidden>
Date: Sat, 7 Jan 2006 13:31:02 +0100
From: Adrian von Bidder <email address hidden>
To: <email address hidden>
Cc: <email address hidden>, <email address hidden>,
 <email address hidden>, <email address hidden>, <email address hidden>
Subject: md5sum shouldn't be able to disappear

--nextPart16015641.e8x9N9unh8
Content-Type: text/plain;
  charset="us-ascii"
Content-Transfer-Encoding: quoted-printable
Content-Disposition: inline

severity 315784 serious
thanks

With packages ranging from woody to experimental (majority is sarge), and=20
upgrading apt, dpkg and some others to etch, I got into the situation that=
=20
no md5sum commmand was on the system. I feel dependencies should be able=20
to avoid that situation.

Should dpkg Pre-Depend on coreutils >=3D 5.93-5 or what?

Below is some IRC conversation that convinced me that this is RC. =20
(deboostrap cc:d on request of mrvn)

cheers
=2D- vbi

>>>>>>>
<cmot> very, very stupid Q (probably) - why is there no /usr/bin/md5sum on=
=20
my system? Isn't that a command that I should be able to expect on my=20
system?
<cmot> (coreutils has /usr/bin/md5sum.textutils - shouldn't there be a=20
symlink or whatever?)
<noshadow> cmot: which release? in sarge it should be in the dpkg package
<godog> cmot: indeed /usr/bin/md5sum here is owned by coreutils
<cmot> Hmmm. I have some mix with files from woody to experimental.
<cmot> I recently upgraded apt, dpkg & co to etch, so I guess that's the=20
problem.
<cmot> Yep. coreutils from sarge, dpkg from etch.
<cmot> Should this be a bug?
<vorlon> probably. if possible, something should have a depends: or a=20
conflicts: that prevents you from getting your system into that state.
<asuffield> md5sum is Essential, it's definitely a bug of some kind
<asuffield> you're not supposed to *ever* lose those
<noshadow> though it's better to err on this side than to make upgrade path=
s=20
impossible
<cmot> Hmm. coreutils/testing doesn't have /usr/bin/md5sum either, it seem=
s=20
to me?
<godog> etch's coreutils can also check /usr/bin/md5sum and symlink=20
appropriately
<asuffield> godog: that's postinst. Essential packages have to work without=
=20
being configured
<godog> asuffield: right, thanks for spotting that
<godog> cmot: coreutils 5.2.1-2.1 ?
<cmot> 5.2.1-2
<asuffield> why on earth did it get removed from dpkg? that's the problem=20
right there
<noshadow> asuffield: in what sense "essential"?
<cmot> But I just upgraded to 5.2.1-2.1 and it's only md5sum.textutils,=20
still.
<cmot> #289369, btw
<godog> noshadow: as in "Essential: yes" I think
<asuffield> grep-available -FPackage -sPackage,Essential coreutils
<asuffield> Package: coreutils
<asuffield> Essential: yes
<asuffield> it's a fairly simple and stupid problem, dpkg should not have=20
dropped md5sum until coreutils had released with it
<vorlon> or it should pre-depend on the new version
<asuffield> hmm, I suppose that would work, but it's kinda icky
<asuffield> I'd have been more inclined to do it with Replaces
<vorlon> nah, pre-depends are de rigueur in essential packages
<asuffield> anyway, somebody go file a bug about it
<asuffield...

[Debian Bug Importer (debzilla)]

Message-Id: <email address hidden>
Date: Sat, 7 Jan 2006 13:33:11 +0100
From: Adrian von Bidder <email address hidden>
To: <email address hidden>

--nextPart5524325.sSLE1mEHtt
Content-Type: text/plain;
  charset="us-ascii"
Content-Transfer-Encoding: quoted-printable
Content-Disposition: inline

retitle 315784 dependencies should ensure md5sum is present
thanks
=2D-=20
featured product: vim - http://vim.org

--nextPart5524325.sSLE1mEHtt
Content-Type: application/pgp-signature

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (GNU/Linux)
Comment: get my key from http://fortytwo.ch/gpg/92082481

iKcEABECAGcFAkO/tQhgGmh0dHA6Ly9mb3J0eXR3by5jaC9sZWdhbC9ncGcvZW1h
aWwuMjAwMjA4MjI/dmVyc2lvbj0xLjUmbWQ1c3VtPTVkZmY4NjhkMTE4NDMyNzYw
NzFiMjVlYjcwMDZkYTNlAAoJEIukMYvlp/fWt5AAoNgXsE8tlTf4x/bF5w1HRrzE
mheBAKDSK+SG4mJq+lF/u9nrjo84AbYWZQ==
=/uzQ
-----END PGP SIGNATURE-----

--nextPart5524325.sSLE1mEHtt--

[Scott James Remnant (Canonical) (canonical-scott)]

Silly Debian squabbling ... coreutils is essential!

[Nicolas François (nicolas-francois)]

Hello,

After reading the log of #315784 (dpkg: dependencies should ensure md5sum
is present), I wonder why adding a
        Pre-Depends: coreutils (>= ...)
field to dpkg would break (c)debootstrap.

I made a tarball for debootstrap with a modified dpkg and I noticed no
problem.

Goswin, do you remember why you said it would break debootstrap?
I'm keeping Adrian and the (c)debootstrap maintainer CCed.

> <cmot> Ah. So dpkg just pre-depend on the new coreutils?
> <cmot> I'll file a serious bug on dpkg, then (if there isn't one already;
> haven't looked yet).
> <mrvn> cmot: that would suck because it breaks (c)debootstrap

Kind Regards,
--
Nekral

[Goswin von Brederlow (brederlo)]

Nicolas François <email address hidden> writes:

> Hello,
>
> After reading the log of #315784 (dpkg: dependencies should ensure md5sum
> is present), I wonder why adding a
> Pre-Depends: coreutils (>= ...)
> field to dpkg would break (c)debootstrap.
>
> I made a tarball for debootstrap with a modified dpkg and I noticed no
> problem.
>
> Goswin, do you remember why you said it would break debootstrap?
> I'm keeping Adrian and the (c)debootstrap maintainer CCed.
>
>> <cmot> Ah. So dpkg just pre-depend on the new coreutils?
>> <cmot> I'll file a serious bug on dpkg, then (if there isn't one already;
>> haven't looked yet).
>> <mrvn> cmot: that would suck because it breaks (c)debootstrap
>
> Kind Regards,
> --
> Nekral

Debootstrap has a fixed list of packages to install (unless you give
the option to resolve depends) but I guess coreutils is in that list
already. Cdebootstrap resolves depends so it will pick up coreutils
even if it didn't have it in the past. But that is the easy part.

When installing packages get unpacked and configured in a certain
order to bootstrap the system. In cdebootstrap the config gives some
hints what to force first to get dpkg configured and then lets dpkg
handle the rest. That step will fail at first because dpkg won't
configure without coreutils being configured first.

It might be cdebootstrap pulls through that, it is quite resiliant,
but when run with debug this will cause a lot of dpkg messages that
make detecting any other problem much harder.

MfG
        Goswin

PS: Don't let that stop you from adding a proper Pre-Depends. Just
contact the (c)debootstrap authors about the change so they can fix it
the same day.
PPS: I guess the CCs did that now. :)

[Frank Lichtenheld (djpig)]

# Automatically generated email from bts, devscripts version 2.9.19
severity 313605 serious
merge 313605 315784

[Frank Lichtenheld (djpig)]

# Automatically generated email from bts, devscripts version 2.9.19
tags 313605 pending

[Guillem Jover (guillem)]

Source: dpkg
Source-Version: 1.13.20

We believe that the bug you reported is fixed in the latest version of
dpkg, which is due to be installed in the Debian FTP archive:

dpkg-dev_1.13.20_all.deb
  to pool/main/d/dpkg/dpkg-dev_1.13.20_all.deb
dpkg_1.13.20.dsc
  to pool/main/d/dpkg/dpkg_1.13.20.dsc
dpkg_1.13.20.tar.gz
  to pool/main/d/dpkg/dpkg_1.13.20.tar.gz
dpkg_1.13.20_i386.deb
  to pool/main/d/dpkg/dpkg_1.13.20_i386.deb
dselect_1.13.20_i386.deb
  to pool/main/d/dpkg/dselect_1.13.20_i386.deb

A summary of the changes between this version and the previous one is
attached.

Thank you for reporting the bug, which will now be closed. If you
have further comments please address them to <email address hidden>,
and the maintainer will reopen the bug report if appropriate.

Debian distribution maintenance software
pp.
Guillem Jover <email address hidden> (supplier of updated dpkg package)

(This message was generated automatically at their request; if you
believe that there is a problem with it please contact the archive
administrators by mailing <email address hidden>)

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Format: 1.7
Date: Wed, 31 May 2006 07:43:16 +0300
Source: dpkg
Binary: dpkg dselect dpkg-dev
Architecture: source i386 all
Version: 1.13.20
Distribution: unstable
Urgency: low
Maintainer: Dpkg Developers <email address hidden>
Changed-By: Guillem Jover <email address hidden>
Description:
 dpkg - package maintenance system for Debian
 dpkg-dev - package building tools for Debian
 dselect - user tool to manage Debian packages
Closes: 35573 36586 54529 57104 58106 65839 81171 165843 169125 174180 198128 198522 208532 230610 237622 237626 281562 313605 315784 316551 318825 349120 349442 361171 361671 364726 366178 366185 366187 366351 366353 366587 366659 366985 367329 368874 368875 369177 369205
Changes:
 dpkg (1.13.20) unstable; urgency=low
 .
   [ Frank Lichtenheld ]
   * Add gettext support for the Perl scripts. Based on a patch by
     Nicolas François. Closes: #165843
   * Only print usage information of dpkg-scanpackages on stdout
     if requested explicetly. Use stderr in case of error.
     Closes: #366659
   * Add remarks to dpkg-scansources and dpkg-scanpackages
     man pages about the need to compress the generated files
     to be able to access them via apt. Closes: #65839
   * Allow '+' and '.' in distribution names in Debian changelogs.
     Based on a patch by John Wright.
     Closes: #361171
   * Use the Debian keyring in dpkg-source when checking signatures
     of .dsc files, if available. Closes: #364726
   * Let dpkg-buildpackage pass through all remotely sensible
     -sX options to dpkg-source (-s[nsAkurKUR] currently).
     Closes: #36586
   * Improve the description of --showformat in dpkg-deb
     man page and add a pointer to the complete description
     of the option in dpkg-query.
   * Don't spew out garbage from dpkg-deb, if the second argument
     to -I is a absolute filename. Based on a patch by Ian Eure.
     Closes: #35573
   * Fix --ignore-depends argument value parsing. Closes: #169125
   * Completely remove md5sum diversion madness. Instead, we Pre-Depend
     on a version of textutils whic...