Drop apt::cron::download, and turn $apt_cron_hours into a class parameter for ap…
…t::cron::dist_upgrade.
I've introduced apt::cron::download, am not using it anymore, and have
never seen anyone else submit a bug or a merge request for it. Thus,
I feel pretty confident in dropping it.
And this allows me to trivially get rid of the $apt_cron_hours global
variable noticed by LeLutin in #13.
(I've mistakenly marked !44 as merged)
See merge request !47
I've introduced apt::cron::download, am not using it anymore, and have
never seen anyone else submit a bug or a merge request for it. Thus,
I feel pretty confident in dropping it.
And this allows me to trivially get rid of the $apt_cron_hours global
variable noticed by LeLutin in #13.
The logic in specifying the default value for apt::repos is in the wrong
place: it should be in apt::params since this is exactly what this
latter class is for.
There's no special case that can involve making default the value follow
the value of another parameter so there's no point in having that logic
in the main class.
backports_url was already set to the same value than debian_url by
default, but it was set to the default value of debian_url. this meant
that when giving a url to the debian_url parameter to the apt class, the
backports_url would not follow and would still be using the default
value for debian_url.
with this change, when backports_url is not specified, but debian_url
is, then both of them take on the value given to the debian_url. of
course if backports_url is also given a value, then it keeps the value
given by the user.
In some situations, the log directory for unattended-upgrades might not
exist. In those cases, packages will not get upgraded!
unattended-upgrades crashes with a python backtrace because the log dir
is not present.
The `disable_update` parameter has been removed. The main apt class
defaults to *not* run an `apt-get update` on every run anyway so
this parameter seems useless.
You can include the `apt::update` class if you want it to be
run every time.
`apt-get autoclean` should not be run on every puppetrun when
including `apt::update`, but rather be configured as a `APT::Periodic`
task that is run by cron, see
https://wiki.debian.org/UnattendedUpgrades.
Before, there were two Execs that did an `apt-get update`,
`Exec[refresh_apt]` and `Exec[apt_updated]`, which were triggered
by different resources.
This changes gets rid of the first one, and all resources now depend
on `Exec[apt_updated]`.
When pinning packages with apt::preferences_snippet,
we need to make sure these get deployed before an
`apt-get update` is triggered, so pinned packages can
get installed in the right way with a single puppetrun.
Managing requirements for installing the lsb package has proven over
time to make no sense. The best approach to this is to require
lsb-release to be installed alongside puppet, since otherwise there are
so much facts that get no value during the run and you end up needing to
run puppet twice to get the real end result.
Also, since we're not including a class that is actually installing the
'lsb' package, that require line makes it so that including the apt
module doesn't work, and there's no documentation in the README about
needing to provide a package{'lsb':} resource with the apt class.
Because of all that, it makes more sense to just get rid of that require
line and mark lsb as a pre-requirement in the README file.
Ubuntu shouldn't be using debian backports by default.
This was written by Anoine Beaupré, but split from the commit "move
backports to snippets" since the change is unrelated and needs to be
more visible in the commit history.
Micah found an issue with usage of config_content: if you call template('...')
yourself and pass that on to config_content, then your template gets evaluated
without all of the variables. This means that you don't hava access to
blacklisted_packages, mail_recipient or mailonlyonerror.
To make it possible to use a different template while still having access to
those variables, let's make it possible to change the template name that we're
using.
Note that reboot-notifier is only in stretch/sid at the moment,
so to use this one needs to configure APT pinning themselves
to make this package installable.
* Removes dependency on lsb-release and/or Facter >1.7
(values are based on $::lsbdistcodename, when available)
* Simplifies maintenance: only lib/facter/util/* require updates as new
releases are made
Caveats:
* apt::codename is removed; to override debian_* facts, set the
FACTER_debian_codename environment variable for puppet
* If tracking unstable, make sure lsb-release is installed, as other
methods can't tell between testing and unstable
