Re: Standards 4.1.3 and apt-cacher

2018-02-13 Thread Felipe Sateler
On Thu, 08 Feb 2018 17:56:28 +, Mark Hindley wrote:

 
> However, in trying to understand more fully when postinst is called with
> configure and reconfigure arguments, it appears that reconfigure is not
> currently supported at all. It is not mentioned in the Policy and
> ancient bug #215549 was closed as wontfix. So although reconfigure is in
> the postinst skeleton the distinction is not observed. The only possible
> way to distinguish is using the hack that DEBCONF_RECONFIGURE=1 is set
> in the postinst environment. See debconf-devel(7).
> 
> Or am I missing something?

I don't actually know much about debconf, sorry. I just followed what 
debhelper is doing, but it appears you need to add that check with 
DEBCONF_RECONFIGURE, otherwise a dpkg-reconfigure will always be a NOP 
(because defaults and defaults-disabled become NOPs when the service is 
already enabled or disabled).

Saludos




Re: Re: Standards 4.1.3 and apt-cacher

2018-02-08 Thread Mark Hindley

> It is a bit overzealous, but I think you have another problem. Your package
> will override local admin on upgrades (if they changed the config manually) by
> calling update-rc.d and update-inetd. I think you need to change it to not do
> so. I don't know much about update-inetd but the pattern for update-rc.d
> should be something like this:
>
>
> if [ $1 = configure ] || [ $1 == abort-upgrade ] ; then
>   db_get apt-cacher/mode
>   case "$RET" in
> daemon)
>   update_rc_args="apt-cacher defaults"
>   ;;
> inetd)
> manual)
>   update_rc_args="apt-cacher defaults-disabled"
>   ;;
>   esac
> fi

> if [ $1 = reconfigure ] ; then
>   db_get apt-cacher/mode
>   case "$RET" in
> daemon)
>   update_rc_args="apt-cacher enable"
>   ;;
> inetd)
>   update_rc_args="apt-cacher disable"
>   ;;
> manual)
>   update_rc_args="" #do nothing
>   ;;
>   esac
> fi

> if [ -n "$update_rc_args" ] ; then
>   update-rc.d $update_rc_args
> fi

Felipe,

Many thanks for this. Your point is well made and I have now got a working
version that lintian doesn't complain about.

However, in trying to understand more fully when postinst is called with
configure and reconfigure arguments, it appears that reconfigure is not
currently supported at all. It is not mentioned in the Policy and ancient bug
#215549 was closed as wontfix. So although reconfigure is in the postinst
skeleton the distinction is not observed. The only possible way to distinguish
is using the hack that DEBCONF_RECONFIGURE=1 is set in the postinst
environment. See debconf-devel(7).

Or am I missing something?

Mark



Re: Standards 4.1.3 and apt-cacher

2018-02-07 Thread Felipe Sateler
Hi,

On Wed, 07 Feb 2018 10:14:15 +, Mark Hindley wrote:

> Hello,
> 
> I am trying to upgrade apt-cacher to Standards version 4.1.3. In
> particular using DISABLED=yes|no in /etc/defaults is now prohibited.

Thanks for fixing this.

> 
> apt-cacher can be run as a daemon or from /etc/inetd.conf and this
> configuration is set in response to a debconf question.
> 
> I have a working version of the postinst which I believe complies with
> the new policy by no longer changing a value in /etc/default/apt-cacher,
> however it contains multiple calls to update-rc.d (one to register the
> init file and then calls to enable/disable as appropriate in response to
> db_get). This produces the lintian error:
> duplicate-updaterc.d-calls-in-postinst.
> 
> Is my approach flawed or is lintian being overzealous?

It is a bit overzealous, but I think you have another problem. Your 
package will override local admin on upgrades (if they changed the config 
manually) by calling update-rc.d and update-inetd. I think you need to 
change it to not do so. I don't know much about update-inetd but the 
pattern for update-rc.d should be something like this:


if [ $1 = configure ] || [ $1 == abort-upgrade ] ; then
  db_get apt-cacher/mode
  case "$RET" in
daemon)
  update_rc_args="apt-cacher defaults"
  ;;
inetd)
manual)
  update_rc_args="apt-cacher defaults-disabled"
  ;;
  esac
fi

if [ $1 = reconfigure ] ; then
  db_get apt-cacher/mode
  case "$RET" in
daemon)
  update_rc_args="apt-cacher enable"
  ;;
inetd)
  update_rc_args="apt-cacher disable"
  ;;
manual)
  update_rc_args="" #do nothing
  ;;
  esac
fi

if [ -n "$update_rc_args" ] ; then
  update-rc.d $update_rc_args
fi


if [ "$1" = "configure" ] || [ "$1" = "abort-upgrade" ]; then
  if [ -x "/etc/init.d/apt-cacher" ]; then
invoke-rc.d apt-cacher start ||  exit $?
  fi
fi


PS: you might want to take a look at the socket-activation features of 
systemd. You could have systemd-systems be always socket-activated. One 
might even think of removing the standalone daemon mode

-- 
Saludos