gmane.linux.debian.apt.devel

Bug#674147: apt: LANG=C apt-cache policy is very slow

Francois Gouget — 2012-05-23T11:15:32

Package: apt
Version: 0.9.3
Severity: normal

Dear Maintainer,

Here is how to reproduce the problem:

$ LANG=C time apt-cache policy >/dev/null
1.76user 0.05system 0:01.82elapsed 99%CPU (0avgtext+0avgdata 137200maxresident)k
0inputs+0outputs (0major+11677minor)pagefaults 0swaps

$ LANG=fr_FR.UTF8 time apt-cache policy >/dev/null
0.03user 0.00system 0:00.04elapsed 86%CPU (0avgtext+0avgdata 63584maxresident)k
0inputs+0outputs (0major+5982minor)pagefaults 0swaps

This makes lsb_release very slow because it runs apt-cache with LANG=C, which
in turn makes vmrun, a VMware tool, very slow.

Now the reason for the speed difference between the C and fr_FR.UTF8 cases is
that on my system 'apt-cache gencaches' is being run in the French locale and
that only one cache is kept.

Ideally apt-cache policy should be fast in both the 'machine querying' mode
that's normally done in the C locale, and in the 'main user' locale. Though if
lsb_release should handle things differently that could fix things too.

Processed (with 1 errors): bts maintenance

Debian Bug Tracking System — 2012-05-23T09:21:21

Processing commands for control< at >bugs.debian.org:

Bug #672467 [aptitude] [L10N,DE] aptitude: updated german program translation
Added tag(s) pending.
Bug #674045 [aptitude] aptitude: German translation says "#Broken" but shows no number
Ignoring request to alter tags of bug #674045 to the same tags previously set
Unknown command or malformed arguments to command.

Bug #633872 [aptitude] aptitude should show serious bugs BEFORE downloading and installing a package
Severity set to 'wishlist' from 'minor'
Bug #471245 {Done: Christian PERRIER <bubulle< at >debian.org>} [aptitude] aptitude: Doesn't quit from Help section using Return
Marked as fixed in versions aptitude/0.4.11.1-1.
Bug #611235 [aptitude] aptitude: file descriptor leak of apt history log
Severity set to 'important' from 'normal'
Bug #575062 [aptitude] aptitude: file descriptor leak of debtags related files
Bug #611235 [aptitude] aptitude: file descriptor leak of apt history log
Marked as found in versions aptitude/0.6.1.5-3.
Bug #575062 [aptitude] apti

Bug#674100: [apt] apt-cdrom add generates broken cdroms.list

Miguel Figueiredo — 2012-05-23T07:00:43

[...]


The installer image that i tested yesterday uses:

May 22 21:58:18 debootstrap: Unpacking libapt-pkg4.12:amd64 (from 
.../libapt-pkg4.12_0.9.3_amd64.deb) ...

(from installer's log)

Bug#674100: [apt] apt-cdrom add generates broken cdroms.list

Filipus Klutiero — 2012-05-23T02:47:13

Package: apt
Version: 0.9.5
Severity: grave
X-Debbugs-Cc: 673997< at >bugs.debian.org
X-Debbugs-Cc: debian-boot< at >lists.debian.org

apt-cdrom add; appears to be broken with APT 0.9. Although it can return 
success, it seems to only generate broken /var/lib/apt/cdroms.list 
files. Instead of having the normal entries:
the generated files contain APT configuration. I'm attaching my current 
cdroms.list.

I first experienced this bug with apt 0.9.3. The bug persists in 0.9.5. 
This was discovered by Miguel Figueiredo <elmig< at >debianpt.org>.

This has an effect on the installer. I believe apt-cdrom add; is run 
when setting up apt. When the install media contains the base system 
(which is the case for netinsts, but not for business cards), the base 
system's installation relies on the iso. This causes APT to fail to 
install the base system. Specifically, base-installer fails to install 
locales, then busybox, which prevents Linux's installation.
I experienced this issue on 2 PCs from the 2 I tested (one laptop, one 
de

Bug#670401: Bug#673065: rred segfaults during apt-get update

Vincent Lefevre — 2012-05-22T22:47:56

                                                ^^^^^^

This is not one of the bugs 554387, 670401, 673065. Is this an error
or are these bugs the same one, in which case a merge is missing?

Bug#673939: zlib1g-dev: version 1:1.2.7.dfsg-7 requires anuninstallable dependency (zlib1g (= 1:1.2.7.dfsg-4))

Jonathan Nieder — 2012-05-22T18:39:53


Drive-by patch to avoid confusion attached.
diff --git i/cmdline/apt-get.cc w/cmdline/apt-get.cc
index c996017b..d33158cd 100644
--- i/cmdline/apt-get.cc
+++ w/cmdline/apt-get.cc
< at >< at > -1034,10 +1034,10 < at >< at > bool CacheFile::CheckDeps(bool AllowBroken)
    }
    else
    {
-      c1out << _("You might want to run 'apt-get -f install' to correct these.") << endl;
+      c1out << _("You might want to run 'apt-get --fix-broken install' to correct these.") << endl;
       ShowBroken(c1out,*this,true);
 
-      return _error->Error(_("Unmet dependencies. Try using -f."));
+      return _error->Error(_("Unmet dependencies. Try using --fix-broken."));
    }
       
    return true;
< at >< at > -1947,11 +1947,11 < at >< at > bool DoInstall(CommandLine &CmdL)
  packages */
       if (BrokenFix == true && Cache->BrokenCount() != 0)
       {
- c1out << _("You might want to run 'apt-get -f install' to correct these:") << endl;
+ c1out << _("You might want to run 'apt-get --fix-broken install' to correct these:") << endl;
  ShowBroken(c1ou

Bug#673939: zlib1g-dev: version 1:1.2.7.dfsg-7 requires anuninstallable dependency (zlib1g (= 1:1.2.7.dfsg-4))

David Kalnischkies — 2012-05-22T17:37:23

The newer versions should already be in /var/cache/apt/archives, so a
download doesn't need to happen (usually). A rollback can't happen
as downgrades aren't supported in debian (even if they quiet often work
for specific packages, but i wouldn't depend on that)

Recovering can be quiet a problem, and can depend on the overall
situation so it is not really easy. After all, if apt could print a nice
message detailing what an operator can do, why doesn't do this
itself instead of printing these lines… no, you get these sometimes
quiet cryptic messages because apt has no idea how to proceed.

So, Sophoklis, are you still in that situation?
If so, could you attach your /var/lib/dpkg/status to the bugreport
(or if you don't want to have this information public on the net,
 feel free to send it privately to me) please?

Recovery should be possible by either using dpkg directly or
helping apt: "apt-get install -f" tries to figure out everything
by itself. Add packages to this command and it will stop trying
for

Bug#605605: marked as done (apt: uninteresting NEWS.Debian)

Debian Bug Tracking System — 2012-05-22T16:42:06

Your message dated Tue, 22 May 2012 11:39:26 -0500
with message-id <20120522163925.GC7008< at >burratino>
and subject line Re: apt: uninteresting NEWS.Debian
has caused the Debian Bug report #605605,
regarding apt: uninteresting NEWS.Debian
to be marked as done.

This means that you claim that the problem has been dealt with.
If this is not the case it is now your responsibility to reopen the
Bug report if necessary, and/or fix the problem forthwith.

(NB: If you are a system administrator and have no idea what this
message is talking about, this may indicate a serious mail system
misconfiguration somewhere. Please contact owner< at >bugs.debian.org
immediately.)

Processed: Re: zlib1g-dev: version 1:1.2.7.dfsg-7 requires anuninstallable dependency (zlib1g (= 1:1.2.7.dfsg-4))

Debian Bug Tracking System — 2012-05-22T16:33:17

Processing commands for control< at >bugs.debian.org:

Bug #673939 {Done: Mark Brown <broonie< at >debian.org>} [zlib1g-dev] zlib1g-dev: version 1:1.2.7.dfsg-7 requires an uninstallable dependency (zlib1g (= 1:1.2.7.dfsg-4))
Changed Bug title to 'apt: please be smarter about resuming after an interrupted/failed unpack' from 'zlib1g-dev: version 1:1.2.7.dfsg-7 requires an uninstallable dependency (zlib1g (= 1:1.2.7.dfsg-4))'
Bug #673939 {Done: Mark Brown <broonie< at >debian.org>} [zlib1g-dev] apt: please be smarter about resuming after an interrupted/failed unpack
Severity set to 'wishlist' from 'important'
Bug #673939 {Done: Mark Brown <broonie< at >debian.org>} [zlib1g-dev] apt: please be smarter about resuming after an interrupted/failed unpack
Bug reassigned from package 'zlib1g-dev' to 'apt'.
No longer marked as found in versions zlib/1:1.2.7.dfsg-4.
Ignoring request to alter fixed versions of bug #673939 to the same values previously set
Bug #673939 {Done: Mark Brown <broonie< at >debian.org>} [apt] apt: please be smarter abou

Re: zlib1g-dev: version 1:1.2.7.dfsg-7 requires an uninstallabledependency (zlib1g (= 1:1.2.7.dfsg-4))

Jonathan Nieder — 2012-05-22T16:31:28

retitle 673939 apt: please be smarter about resuming after an interrupted/failed unpack
# difficult
severity 673939 wishlist
reassign 673939 apt 0.9.3
reopen 673939
quit

Sophoklis Goumas wrote:


Excellent, thanks for the log.  The story starts here:

[...]
[...]
[...]

At this point, the updated zlib1g has been unpacked, but not updated
zlib1g-dev, and apt encounters an error during the unpack procedure.
It stops to let the sysadmin take care of it, and then the next apt
command begins its work with "dpkg --configure -a":

[...]

And it is unable to recover.  A smarter package manager would realize
it is time to download and unpack either the updated versions of
zlib1g-dev and zlib1g-dbg (in effect continuing the interrupted
upgrade) or the previous version of zlib1g (in effect rolling it back)
or print some advice help the human operator to do so.

Roughly speaking: imagine three packages A, B, and D.  The new version
of A depends on D.  The package B is utterly broken.  The sysadmin
tries:

$ apt-get in

Bug#605605: apt: uninteresting NEWS.Debian

Jakub Wilk — 2012-05-22T16:23:50

* Jonathan Nieder <jrnieder< at >gmail.com>, 2012-05-22, 09:29:

Meh, feel free.

Re: Hashsum mismatch prevention strategies

Raphael Geissert — 2012-05-22T16:16:31

[chromium repo issues]

OK, let's see if somebody can check it.

David Kalnischkies wrote:

Huh.


There's this page about chromium, not sure if they've noticed broken servers 
don't comply with their criteria. 
http://www.chromium.org/developers/design-documents/network-stack/http-
pipelining


Yes, the second approach is what I had in mind. In fact, I thought the 
hashes were already being sent. Apparently, I got confused.
The http method could even go as far as checking Digest headers if sent by 
the server.


Right :-/

Cheers,

Re: zlib1g-dev: version 1:1.2.7.dfsg-7 requires an uninstallabledependency (zlib1g (= 1:1.2.7.dfsg-4))

Sophoklis Goumas — 2012-05-22T15:57:30

Allrighty then:
http://paste.debian.net/plain/170619

Sophoklis

Re: zlib1g-dev: version 1:1.2.7.dfsg-7 requires an uninstallabledependency (zlib1g (= 1:1.2.7.dfsg-4))

Jonathan Nieder — 2012-05-22T15:19:29


The interesting part is in the section marked "...".  Somehow
zlib1g-dev got unpacked in the wrong version without being configured.


Odd.  The last time it was configured was Sunday morning.  The upgrade
the next day never finished for some reason.

Jonathan

Re: zlib1g-dev: version 1:1.2.7.dfsg-7 requires an uninstallabledependency (zlib1g (= 1:1.2.7.dfsg-4))

Sophoklis Goumas — 2012-05-22T15:08:19

Let's tie the pieces of the story together then:

# cat /var/log/apt/term.log
http://paste.debian.net/170610/

# grep zlib1g /var/log/dpkg.log
http://paste.debian.net/170611/

# apt-cache policy zlib1g zlibg1-dev zlib1g-dbg
http://paste.debian.net/170612/


Oh don't apologize. Please. Don't.
On the contrary: thank you all.

Sophoklis

apt_0.9.5_amd64.changes ACCEPTED into unstable

Debian FTP Masters — 2012-05-22T14:47:15



Accepted:
apt-doc_0.9.5_all.deb
  to main/a/apt/apt-doc_0.9.5_all.deb
apt-transport-https_0.9.5_amd64.deb
  to main/a/apt/apt-transport-https_0.9.5_amd64.deb
apt-utils_0.9.5_amd64.deb
  to main/a/apt/apt-utils_0.9.5_amd64.deb
apt_0.9.5.dsc
  to main/a/apt/apt_0.9.5.dsc
apt_0.9.5.tar.gz
  to main/a/apt/apt_0.9.5.tar.gz
apt_0.9.5_amd64.deb
  to main/a/apt/apt_0.9.5_amd64.deb
libapt-inst1.5_0.9.5_amd64.deb
  to main/a/apt/libapt-inst1.5_0.9.5_amd64.deb
libapt-pkg-dev_0.9.5_amd64.deb
  to main/a/apt/libapt-pkg-dev_0.9.5_amd64.deb
libapt-pkg-doc_0.9.5_all.deb
  to main/a/apt/libapt-pkg-doc_0.9.5_all.deb
libapt-pkg4.12_0.9.5_amd64.deb
  to main/a/apt/libapt-pkg4.12_0.9.5_amd64.deb


Changes:
apt (0.9.5) unstable; urgency=low
 .
  [ Chris Leick ]
  * proofreading of the manpage pot
  * German manpage translation update (Closes: #673294)
 .
  [ David Kalnischkies ]
  * buildlib/podomain.mak:
    - ensure that all sources end up in the srclist so that we don't
      forget to extract half of the translation string

Processing of apt_0.9.5_amd64.changes

Debian FTP Masters — 2012-05-22T14:43:02

apt_0.9.5_amd64.changes uploaded successfully to localhost
along with the files:
  apt_0.9.5.dsc
  apt_0.9.5.tar.gz
  apt-doc_0.9.5_all.deb
  libapt-pkg-doc_0.9.5_all.deb
  libapt-pkg4.12_0.9.5_amd64.deb
  libapt-inst1.5_0.9.5_amd64.deb
  apt_0.9.5_amd64.deb
  libapt-pkg-dev_0.9.5_amd64.deb
  apt-utils_0.9.5_amd64.deb
  apt-transport-https_0.9.5_amd64.deb

Greetings,

Your Debian queue daemon (running on host franck.debian.org)

Bug#673630: marked as done (apt: segfaults on distupgrade)

Debian Bug Tracking System — 2012-05-22T14:58:09

Your message dated Tue, 22 May 2012 14:47:15 +0000
with message-id <E1SWqMl-0008AG-Jl< at >franck.debian.org>
and subject line Bug#673815: fixed in apt 0.9.5
has caused the Debian Bug report #673815,
regarding apt: segfaults on distupgrade
to be marked as done.

This means that you claim that the problem has been dealt with.
If this is not the case it is now your responsibility to reopen the
Bug report if necessary, and/or fix the problem forthwith.

(NB: If you are a system administrator and have no idea what this
message is talking about, this may indicate a serious mail system
misconfiguration somewhere. Please contact owner< at >bugs.debian.org
immediately.)

Bug#673878: marked as done (apt segfaults on upgradingnagios-plugins)

Debian Bug Tracking System — 2012-05-22T14:58:11

Your message dated Tue, 22 May 2012 14:47:15 +0000
with message-id <E1SWqMl-0008AG-Jl< at >franck.debian.org>
and subject line Bug#673815: fixed in apt 0.9.5
has caused the Debian Bug report #673815,
regarding apt segfaults on upgrading nagios-plugins
to be marked as done.

This means that you claim that the problem has been dealt with.
If this is not the case it is now your responsibility to reopen the
Bug report if necessary, and/or fix the problem forthwith.

(NB: If you are a system administrator and have no idea what this
message is talking about, this may indicate a serious mail system
misconfiguration somewhere. Please contact owner< at >bugs.debian.org
immediately.)

Bug#673756: marked as done (aptitude: Reproducible segmentationfault trying a routine upgrade os a sid system)

Debian Bug Tracking System — 2012-05-22T14:58:10

Your message dated Tue, 22 May 2012 14:47:15 +0000
with message-id <E1SWqMl-0008AG-Jl< at >franck.debian.org>
and subject line Bug#673815: fixed in apt 0.9.5
has caused the Debian Bug report #673815,
regarding aptitude: Reproducible segmentation fault trying a routine upgrade os a sid system
to be marked as done.

This means that you claim that the problem has been dealt with.
If this is not the case it is now your responsibility to reopen the
Bug report if necessary, and/or fix the problem forthwith.

(NB: If you are a system administrator and have no idea what this
message is talking about, this may indicate a serious mail system
misconfiguration somewhere. Please contact owner< at >bugs.debian.org
immediately.)

Bug#673536: marked as done (apt segfaults when installingnagios-plugins-basic which now pre-depends on nagios-plugins-common)

Debian Bug Tracking System — 2012-05-22T14:58:08

Your message dated Tue, 22 May 2012 14:47:15 +0000
with message-id <E1SWqMl-0008AG-Jl< at >franck.debian.org>
and subject line Bug#673815: fixed in apt 0.9.5
has caused the Debian Bug report #673815,
regarding apt segfaults when installing nagios-plugins-basic which now pre-depends on nagios-plugins-common
to be marked as done.

This means that you claim that the problem has been dealt with.
If this is not the case it is now your responsibility to reopen the
Bug report if necessary, and/or fix the problem forthwith.

(NB: If you are a system administrator and have no idea what this
message is talking about, this may indicate a serious mail system
misconfiguration somewhere. Please contact owner< at >bugs.debian.org
immediately.)