In Ubuntu 15.10 the path of the apt sources file, which is generated by
apt-add-repository, changed to include the distid. This breaks apt::ppa
idempotency, since it does not recognize the repository is already
added.
Reported on puppet-users as well:
https://groups.google.com/forum/#!topic/puppet-users/YzeMyZYUo98
It looks like add-apt-repository changes pluses to underscores when
creating the sources.list.d file. So the name it creates doesn't match
what Puppet expects, and the resource keeps applying on every Puppet
run.
This works around that problem.
The 2.x series added a changed behavior for backport pinning to pin to
origin instead of release. Pinning to release is the correct behavior
for backports though.
* Expose the underlying notify_update setting of the apt::settings resource
This is because not all configuration file changes should trigger an apt::update notification
* apt::pin also shouldn't result in an apt-update call
Adding a pin configuration should apply to the next apt-get update call it shouldn't trigger one itself.
* Added documentation
* Add tests for apt::conf notify_update
NOTE: While out-of-the box support is disabled, it is still possible to
get the same configurations, it will just require explicitly setting the
necessary codename-munging.
This should only affect `apt::ppa`
This adds support for passing in a full pin declaration as the pin
parameter on a source declaration.
It keeps the old behaviour intact, you can still simply do `pin => '10'`
and it will pin on origin with a priority of 10.
Should that not be what you want you can now pass in a full pin
declaration instead. We make no assumptions here, whatever you pass in
will be passed through to pin as-is with the exception of the values for
`ensure` and `before` which are always overridden by us to ensure
everything keeps working as designed.
`apt::ppa` and `apt::setting` don't actually include `apt::update` so
anchors are unnecessary. Move `apt` to use contain instead of anchors,
since it wasn't anchoring properly anyways. Update the tests to make
sure it can have settings and ppas depending on each other without
cycles.
In what universe does it make sense to create a `sources.list.d` entry
for a repository **without** specifying where this repository is?
😖😞😩😧😱
Only when removing the resource should a location not be required.
the apt module did not correctly detect Cumulus Linux with lsbdistid.
This change adds several lines in params.pp to detect Cumulus Linux and
set $distid and $distcodename
It is weird that `trusted_source` would default to `false` as that would
imply that we normally don't trust our sources. This is opposite to the
truth, by default we trust them but only if the Releases file can be
verified (meaning it is signed by a GPG key known to apt).
What we were telling apt is that it should trust this source even if the
Releases file and the repository is unsigned. This is better captured
with `allow_unsigned` and better highlights the danger of what you're
doing, installing packages from a source we cannot authenticate.
This makes its behaviour similar to the `update`, `proxy` and `purge`
hashes on the main classes bringing its API more in line with the rest
of the module.
`$xfacts` now contains and guards all `lsb*` facts. Looking up any
`lsb*`-related fact should now always be done through `$::apt::xfacts`
to ensure that the values are always set to either the value of the fact
or undef. This avoids all sorts of kerfuffles with strict variables.
A few of these fixes are absolutely horrendous but we have no choice as
we need to stay current- and future-parser compatible for now.
Once we can go Puppet 4 only we can use the `$facts` hash lookup instead
which will return undef/nil for things that aren't set instead of them
not being defined at all.
