Debian Bug report logs - #469855
postinst fails if udev is not running

Toggle useless messages

Message #5 received at submit@bugs.debian.org (full text, mbox, reply):

From: "Joe Nahmias" <joe@nahmias.net>

To: Debian Bug Tracking System <submit@bugs.debian.org>

Subject: postinst fails if udev is not running

Date: Fri, 07 Mar 2008 09:46:56 -0500

Package: nut
Version: 2.2.1-2
Severity: serious
Tags: patch

Hello,

It seems that the nut postinst script errors out if udev is not running
at the time the script is run.  This creates problems in buildds or
chroots since udev is normally not started in those environments.  See,
for example, the i386 build log for collectd 4.3.0-2 at:
http://buildd.debian.org/fetch.cgi?pkg=collectd;ver=4.3.0-2;arch=i386;stamp=1204846772

The following patch below should fix this:

--- postinst	2008-02-08 10:13:24.000000000 -0500
+++ postinst.fixed	2008-03-07 09:38:46.000000000 -0500
@@ -19,8 +19,10 @@ case "$1" in
     chmod 770 /var/run/nut /var/lib/nut
 
     # restart udev to apply the USB rules to the already plugged devices
-    [ -x /etc/init.d/udev ] && pidof udevd > /dev/null \
-		&& /usr/sbin/invoke-rc.d udev restart
+    # only if it's already running
+    if [ -x /etc/init.d/udev ] && pidof udevd > /dev/null; then
+        /usr/sbin/invoke-rc.d udev restart
+    fi
     ;;
 
   abort-upgrade)

Thanks,
--Joe

-- System Information:
Debian Release: lenny/sid
  APT prefers unstable
  APT policy: (990, 'unstable'), (500, 'stable')
Architecture: i386 (i686)

Kernel: Linux 2.6.24-1-686 (SMP w/2 CPU cores)
Locale: LANG=en_US.UTF-8, LC_CTYPE=en_US.UTF-8 (charmap=UTF-8)
Shell: /bin/sh linked to /bin/bash

Message #10 received at 469855@bugs.debian.org (full text, mbox, reply):

From: Sebastian Harl <sh@tokkee.org>

To: Joe Nahmias <joe@nahmias.net>, 469855@bugs.debian.org

Cc: control@bugs.debian.org

Subject: Re: Bug#469855: postinst fails if udev is not running

Date: Fri, 7 Mar 2008 18:36:07 +0100

[Message part 1 (text/plain, inline)]

tags 469855 - patch
thanks

Hi,

On Fri, Mar 07, 2008 at 09:46:56AM -0500, Joe Nahmias wrote:
> It seems that the nut postinst script errors out if udev is not running
> at the time the script is run.  This creates problems in buildds or
> chroots since udev is normally not started in those environments.  See,
> for example, the i386 build log for collectd 4.3.0-2 at:
> http://buildd.debian.org/fetch.cgi?pkg=collectd;ver=4.3.0-2;arch=i386;stamp=1204846772
> 
> The following patch below should fix this:
> 
> --- postinst	2008-02-08 10:13:24.000000000 -0500
> +++ postinst.fixed	2008-03-07 09:38:46.000000000 -0500
> @@ -19,8 +19,10 @@ case "$1" in
>      chmod 770 /var/run/nut /var/lib/nut
>  
>      # restart udev to apply the USB rules to the already plugged devices
> -    [ -x /etc/init.d/udev ] && pidof udevd > /dev/null \
> -		&& /usr/sbin/invoke-rc.d udev restart
> +    # only if it's already running
> +    if [ -x /etc/init.d/udev ] && pidof udevd > /dev/null; then
> +        /usr/sbin/invoke-rc.d udev restart
> +    fi
>      ;;
>  
>    abort-upgrade)

Sorry, this patch does not change anything. "a && b" is basically the
same as "if a; then b; fi" which is what you're doing here. This issue
has already been addressed in the fix for #447961.

So, it looks like udevd is running inside the i386 buildd chroot and
thus causes the FTBFS of collectd - at least, that's the only
explanation that makes sense to me. Setting up nut-dev 2.2.1-2 worked
fine on at least armel and powerpc buildds. I'm currently checking if I
can get more information about that.

Imho the severity should be downgraded to "important" or less but I
leave that up to Arnaud (the package maintainer) or Joe.

Cheers,
Sebastian

-- 
Sebastian "tokkee" Harl +++ GnuPG-ID: 0x8501C7FC +++ http://tokkee.org/

Those who would give up Essential Liberty to purchase a little Temporary
Safety, deserve neither Liberty nor Safety.         -- Benjamin Franklin

[signature.asc (application/pgp-signature, inline)]

Message #17 received at 469855@bugs.debian.org (full text, mbox, reply):

From: Julien Cristau <jcristau@debian.org>

To: Sebastian Harl <sh@tokkee.org>, 469855@bugs.debian.org

Cc: Joe Nahmias <joe@nahmias.net>

Subject: Re: Bug#469855: postinst fails if udev is not running

Date: Fri, 7 Mar 2008 18:47:53 +0100

[Message part 1 (text/plain, inline)]

On Fri, Mar  7, 2008 at 18:36:07 +0100, Sebastian Harl wrote:

> Sorry, this patch does not change anything. "a && b" is basically the
> same as "if a; then b; fi" which is what you're doing here. This issue
> has already been addressed in the fix for #447961.
> 
Well no. "a && b" is only successful if both "a" and "b" exit
successfully.  If you don't want to fail when udev is not running, that
patch is correct.

Cheers,
Julien

[signature.asc (application/pgp-signature, inline)]

Message #22 received at 469855@bugs.debian.org (full text, mbox, reply):

From: Sebastian Harl <sh@tokkee.org>

To: Julien Cristau <jcristau@debian.org>, 469855@bugs.debian.org

Cc: Joe Nahmias <joe@nahmias.net>

Subject: Re: Bug#469855: postinst fails if udev is not running

Date: Fri, 7 Mar 2008 20:07:14 +0100

[Message part 1 (text/plain, inline)]

Hi,

On Fri, Mar 07, 2008 at 06:47:53PM +0100, Julien Cristau wrote:
> On Fri, Mar  7, 2008 at 18:36:07 +0100, Sebastian Harl wrote:
> > Sorry, this patch does not change anything. "a && b" is basically the
> > same as "if a; then b; fi" which is what you're doing here. This issue
> > has already been addressed in the fix for #447961.
> > 
> Well no. "a && b" is only successful if both "a" and "b" exit
> successfully.  If you don't want to fail when udev is not running, that
> patch is correct.

Well, the problem is that "udev restart" fails (which is "b" in this
case). It doesn't really matter if "a && b" fails (that's the reason I
said "basically the same" ;-) as the "-e" option has not been set and
thus the return value will simply be ignored - which (imho) is perfectly
fine in this case.

Cheers,
Sebastian

-- 
Sebastian "tokkee" Harl +++ GnuPG-ID: 0x8501C7FC +++ http://tokkee.org/

Those who would give up Essential Liberty to purchase a little Temporary
Safety, deserve neither Liberty nor Safety.         -- Benjamin Franklin

[signature.asc (application/pgp-signature, inline)]

Message #27 received at 469855@bugs.debian.org (full text, mbox, reply):

From: joe@nahmias.net (Joseph Nahmias)

To: Sebastian Harl <sh@tokkee.org>

Cc: Julien Cristau <jcristau@debian.org>, 469855@bugs.debian.org

Subject: Re: Bug#469855: postinst fails if udev is not running

Date: Sun, 9 Mar 2008 05:41:05 +0000

tags 469855 + patch
thanks

On Fri, Mar 07, 2008 at 08:07:14PM +0100, Sebastian Harl wrote:
> Hi,
> 
> On Fri, Mar 07, 2008 at 06:47:53PM +0100, Julien Cristau wrote:
> > On Fri, Mar  7, 2008 at 18:36:07 +0100, Sebastian Harl wrote:
> > > Sorry, this patch does not change anything. "a && b" is basically the
> > > same as "if a; then b; fi" which is what you're doing here. This issue
> > > has already been addressed in the fix for #447961.
> > > 
> > Well no. "a && b" is only successful if both "a" and "b" exit
> > successfully.  If you don't want to fail when udev is not running, that
> > patch is correct.
> 
> Well, the problem is that "udev restart" fails (which is "b" in this
> case). It doesn't really matter if "a && b" fails (that's the reason I
> said "basically the same" ;-) as the "-e" option has not been set and
> thus the return value will simply be ignored - which (imho) is perfectly
> fine in this case.

Um, unless my eyes deceive me, -e is set right on the first line -- just
as it should per policy.  However, you bring up a good point in that the
udevd process that is running outside the chroot will be visible to
pidof within the chroot.  So, here's an improved patch based on how the
udev postinst detects if it's run inside a chroot:


--- postinst	2008-02-08 10:13:24.000000000 -0500
+++ postinst.fixed	2008-03-09 00:34:04.000000000 -0500
@@ -19,8 +19,14 @@ case "$1" in
     chmod 770 /var/run/nut /var/lib/nut
 
     # restart udev to apply the USB rules to the already plugged devices
-    [ -x /etc/init.d/udev ] && pidof udevd > /dev/null \
-		&& /usr/sbin/invoke-rc.d udev restart
+    # only if it's already running in this environment
+    udevd_pid=$(pidof udevd)
+    if [ -x /etc/init.d/udev ] && [ -n "$udev_pid" ]; then
+        if [ "$(stat -c %d/%i /)" = "$(stat -Lc %d/%i /proc/$udev_pid/root 2>/dev/null)" ];
+        then
+            /usr/sbin/invoke-rc.d udev restart
+        fi
+    fi
     ;;
 
   abort-upgrade)


--Joe

Message #34 received at 469855@bugs.debian.org (full text, mbox, reply):

From: Sebastian Harl <sh@tokkee.org>

To: Joseph Nahmias <joe@nahmias.net>, 469855@bugs.debian.org

Cc: Julien Cristau <jcristau@debian.org>

Subject: Re: Bug#469855: postinst fails if udev is not running

Date: Sun, 9 Mar 2008 11:17:42 +0100

[Message part 1 (text/plain, inline)]

Hi,

On Sun, Mar 09, 2008 at 05:41:05AM +0000, Joseph Nahmias wrote:
> On Fri, Mar 07, 2008 at 08:07:14PM +0100, Sebastian Harl wrote:
> > On Fri, Mar 07, 2008 at 06:47:53PM +0100, Julien Cristau wrote:
> > > On Fri, Mar  7, 2008 at 18:36:07 +0100, Sebastian Harl wrote:
> > > > Sorry, this patch does not change anything. "a && b" is basically the
> > > > same as "if a; then b; fi" which is what you're doing here. This issue
> > > > has already been addressed in the fix for #447961.
> > > > 
> > > Well no. "a && b" is only successful if both "a" and "b" exit
> > > successfully.  If you don't want to fail when udev is not running, that
> > > patch is correct.
> > 
> > Well, the problem is that "udev restart" fails (which is "b" in this
> > case). It doesn't really matter if "a && b" fails (that's the reason I
> > said "basically the same" ;-) as the "-e" option has not been set and
> > thus the return value will simply be ignored - which (imho) is perfectly
> > fine in this case.
> 
> Um, unless my eyes deceive me, -e is set right on the first line -- just
> as it should per policy.

Right, sorry for that - I wonder why it has worked in my tests then... ;-)

> However, you bring up a good point in that the
> udevd process that is running outside the chroot will be visible to
> pidof within the chroot.

Good catch!

> --- postinst	2008-02-08 10:13:24.000000000 -0500
> +++ postinst.fixed	2008-03-09 00:34:04.000000000 -0500
> @@ -19,8 +19,14 @@ case "$1" in
>      chmod 770 /var/run/nut /var/lib/nut
>  
>      # restart udev to apply the USB rules to the already plugged devices
> -    [ -x /etc/init.d/udev ] && pidof udevd > /dev/null \
> -		&& /usr/sbin/invoke-rc.d udev restart
> +    # only if it's already running in this environment
> +    udevd_pid=$(pidof udevd)

This should be "udevd_pid=$( pidof udevd 2> /dev/null ) || true" - the
redirection of stderr is not strictly required but I'd add it for the
sake of completeness. "|| true" is obviously required to not terminate
because of "-e" in case udevd isn't running.

> +    if [ -x /etc/init.d/udev ] && [ -n "$udev_pid" ]; then
> +        if [ "$(stat -c %d/%i /)" = "$(stat -Lc %d/%i /proc/$udev_pid/root 2>/dev/null)" ];

The builtin stat shipped with zsh does not seem to recognize the "-c"
option. So, we should use "/usr/bin/stat" here to catch the (unlikely)
case that anybody uses zsh (or possible other "incompatible" shells) as
/bin/sh.

Also, if udevd would be running both inside and outside the chroot
(which does not seem to be possible from a quick test though [1]) pidof
would return two PID's. So, I guess, we should put a "for pid in
$udev_pid" around the inner if-statement (and replace $udev_pid with
$pid in the second call to stat(1)).

> +        then
> +            /usr/sbin/invoke-rc.d udev restart
> +        fi
> +    fi
>      ;;
>  
>    abort-upgrade)

Thanks for the patch - imho it should be uploaded ASAP as it prevents
other packages from being built. Arnaud, any comments?

Cheers,
Sebastian

[1] When trying to start udev inside the chroot, it fails with "erorr
initializing udevd socket".

-- 
Sebastian "tokkee" Harl +++ GnuPG-ID: 0x8501C7FC +++ http://tokkee.org/

Those who would give up Essential Liberty to purchase a little Temporary
Safety, deserve neither Liberty nor Safety.         -- Benjamin Franklin

[signature.asc (application/pgp-signature, inline)]

Message #39 received at 469855@bugs.debian.org (full text, mbox, reply):

From: Julien Cristau <jcristau@debian.org>

To: Sebastian Harl <sh@tokkee.org>

Cc: Joseph Nahmias <joe@nahmias.net>, 469855@bugs.debian.org

Subject: Re: Bug#469855: postinst fails if udev is not running

Date: Sun, 9 Mar 2008 11:51:15 +0100

On Sun, Mar  9, 2008 at 11:17:42 +0100, Sebastian Harl wrote:

> > +    if [ -x /etc/init.d/udev ] && [ -n "$udev_pid" ]; then
> > +        if [ "$(stat -c %d/%i /)" = "$(stat -Lc %d/%i /proc/$udev_pid/root 2>/dev/null)" ];
> 
> The builtin stat shipped with zsh does not seem to recognize the "-c"
> option. So, we should use "/usr/bin/stat" here to catch the (unlikely)
> case that anybody uses zsh (or possible other "incompatible" shells) as
> /bin/sh.

If anyone uses zsh as /bin/sh, they deserve the breakage...

Cheers,
Julien

Message #44 received at 469855@bugs.debian.org (full text, mbox, reply):

From: joe@nahmias.net (Joseph Nahmias)

To: Sebastian Harl <sh@tokkee.org>

Cc: 469855@bugs.debian.org, Julien Cristau <jcristau@debian.org>

Subject: Re: Bug#469855: postinst fails if udev is not running

Date: Sun, 9 Mar 2008 18:23:42 +0000

[Message part 1 (text/plain, inline)]

tags 469855 + pending
thanks

Hello again,

On Sun, Mar 09, 2008 at 11:17:42AM +0100, Sebastian Harl wrote:
> On Sun, Mar 09, 2008 at 05:41:05AM +0000, Joseph Nahmias wrote:
> > On Fri, Mar 07, 2008 at 08:07:14PM +0100, Sebastian Harl wrote:
> > --- postinst	2008-02-08 10:13:24.000000000 -0500
> > +++ postinst.fixed	2008-03-09 00:34:04.000000000 -0500
> > @@ -19,8 +19,14 @@ case "$1" in
> >      chmod 770 /var/run/nut /var/lib/nut
> >  
> >      # restart udev to apply the USB rules to the already plugged devices
> > -    [ -x /etc/init.d/udev ] && pidof udevd > /dev/null \
> > -		&& /usr/sbin/invoke-rc.d udev restart
> > +    # only if it's already running in this environment
> > +    udevd_pid=$(pidof udevd)
> 
> This should be "udevd_pid=$( pidof udevd 2> /dev/null ) || true" - the
> redirection of stderr is not strictly required but I'd add it for the
> sake of completeness. "|| true" is obviously required to not terminate
> because of "-e" in case udevd isn't running.

Yes, indeed.

> > +    if [ -x /etc/init.d/udev ] && [ -n "$udev_pid" ]; then
> > +        if [ "$(stat -c %d/%i /)" = "$(stat -Lc %d/%i /proc/$udev_pid/root 2>/dev/null)" ];
> 
> The builtin stat shipped with zsh does not seem to recognize the "-c"
> option. So, we should use "/usr/bin/stat" here to catch the (unlikely)
> case that anybody uses zsh (or possible other "incompatible" shells) as
> /bin/sh.

*shrug* ok.

> Also, if udevd would be running both inside and outside the chroot
> (which does not seem to be possible from a quick test though [1])

TTBOMK this is impossible, so I'm not bothering with the for loop.

> Thanks for the patch - imho it should be uploaded ASAP as it prevents
> other packages from being built. Arnaud, any comments?

I will be uploading the attached (proofread and tested) patch as an NMU
to the delayed-5 queue so that this will not be open for longer than a
week if the maintainer does nothing.

--Joe

[nut-469855.diff (text/x-diff, attachment)]

Message #51 received at 469855-close@bugs.debian.org (full text, mbox, reply):

From: Joe Nahmias <jello@debian.org>

To: 469855-close@bugs.debian.org

Subject: Bug#469855: fixed in nut 2.2.1-2.1

Date: Thu, 13 Mar 2008 21:02:28 +0000

Source: nut
Source-Version: 2.2.1-2.1

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

nut-cgi_2.2.1-2.1_i386.deb
  to pool/main/n/nut/nut-cgi_2.2.1-2.1_i386.deb
nut-dev_2.2.1-2.1_i386.deb
  to pool/main/n/nut/nut-dev_2.2.1-2.1_i386.deb
nut-hal-drivers_2.2.1-2.1_i386.deb
  to pool/main/n/nut/nut-hal-drivers_2.2.1-2.1_i386.deb
nut-snmp_2.2.1-2.1_i386.deb
  to pool/main/n/nut/nut-snmp_2.2.1-2.1_i386.deb
nut_2.2.1-2.1.diff.gz
  to pool/main/n/nut/nut_2.2.1-2.1.diff.gz
nut_2.2.1-2.1.dsc
  to pool/main/n/nut/nut_2.2.1-2.1.dsc
nut_2.2.1-2.1_i386.deb
  to pool/main/n/nut/nut_2.2.1-2.1_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 469855@bugs.debian.org,
and the maintainer will reopen the bug report if appropriate.

Debian distribution maintenance software
pp.
Joe Nahmias <jello@debian.org> (supplier of updated nut 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 ftpmaster@debian.org)


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

Format: 1.7
Date: Sun, 09 Mar 2008 13:12:12 -0400
Source: nut
Binary: nut nut-cgi nut-snmp nut-dev nut-hal-drivers
Architecture: source i386
Version: 2.2.1-2.1
Distribution: unstable
Urgency: low
Maintainer: Arnaud Quette <aquette@debian.org>
Changed-By: Joe Nahmias <jello@debian.org>
Description: 
 nut        - The core system of the nut - Network UPS Tools
 nut-cgi    - A web interface sub system for the nut - Network UPS Tools
 nut-dev    - Development files for NUT - Network UPS Tools
 nut-hal-drivers - HAL interface for NUT - Network UPS Tools
 nut-snmp   - A meta SNMP Driver subsystem for the nut - Network UPS Tools
Closes: 469855
Changes: 
 nut (2.2.1-2.1) unstable; urgency=low
 .
   * Non-maintainer upload.
   * Skip udev restart logic within chroots, closes: #469855.
Files: 
 a3a76821cbe2c9ccda1d01b08cecb7c4 911 admin optional nut_2.2.1-2.1.dsc
 c3f898a75c04dc01ae1af997c94ea301 37775 admin optional nut_2.2.1-2.1.diff.gz
 e1c58acbe5471fd4281d702c940848bf 1166946 admin optional nut_2.2.1-2.1_i386.deb
 05823d1587db6f1ea002793ae15995e4 144422 admin optional nut-cgi_2.2.1-2.1_i386.deb
 9edaef68cd1bcb05395572e218b42166 111876 admin optional nut-snmp_2.2.1-2.1_i386.deb
 763c47347e419bd10015360a0a95d72a 124166 admin optional nut-dev_2.2.1-2.1_i386.deb
 6817e144fdf6d108276463dd1aaf07fa 176298 admin optional nut-hal-drivers_2.2.1-2.1_i386.deb

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)

iD8DBQFH1COnULacnzmLNx0RAlRxAJ4hTKxjU3dm1wEdKeMJdo8Ke+5NBACgrkeu
aHaE62kO8xuEetW2FbSNpWE=
=cGm0
-----END PGP SIGNATURE-----

Send a report that this bug log contains spam.