Major update #9

reidmv · 2015-09-21T21:03:31Z

This PR is a proposed refactor and major version update. This would be released as a new major version.

Switch to Puppet 4
- Use type features for validation, eliminating dependency on puppetlabs/stdlib.
- Use iteration in dotnet::install::package utility type
- Use new $::os fact
Switch from puppet/download_file to lwf/remote_file. Per commit message in 4db8e15, primary benefit is cleaner implementation allowing for more reliable notify/subscribe. Especially when reboots are subscribed to a dotnet resource this can be important.
Switch from exec-based package installation to package resource.
- Makes it easier to see what dotnet is doing in the graph
- Makes the code cleaner

liamjbennett · 2015-09-21T21:56:47Z

I really like some of these changes so thank you.
I want to give this a more detailed review but initially I have some questions:

Why the use of the remote_file dependency Why this over download_file, archive or staging?
Why the switch to Puppet 4 already?

reidmv · 2015-09-21T22:31:12Z

Hi Liam,

Excellent questions! I figured I'd start with the PR as far forward as I already took it for my project use, and if necessary whittle things down based on the code review. With that background, here is my rationale for using remote_file and Puppet 4:

Archive has client-side dependencies and so is what I would consider heavyweight; in addition to that there's no need to extract anything here. Staging has the same problem as download_file. That problem for me is that I'm setting up a reboot resource subscribed to e.g. Dotnet['.NET Framework 4.0'], and during development I ran into a situation where the source EXE didn't need to be downloaded nor .NET installed, but a private administrative resource (this file) in download_file got changed, thus triggering a reboot. Remote_file is a native type that doesn't have administrative fragments that could generate extra events. That reduces the surface area for unnecessary resource changes, thus lowering the chances of accidentally notifying something (like a reboot) down the line. Since I was proposing a major revision anyway it made sense to suggest it now. There's also the nifty side-benefit of remote_file being platform agnostic.
The simple answer is I mostly write Puppet 4 / future parser code these days. When thinking about how to deal with specifying multiple versions of .NET which don't technically co-exist I ended up using iteration here. With that toehold I went ahead and added type checking as well, allowing the removal of the stdlib dependency. As long as I was removing dependencies I figured using the $::os fact would be a good idea too instead of the windowsfacts-provided $::operatingsystemversion so that the dependencies could be slimmed down that much more as well. On $::os I can have high confidence that the version of Facter shipping with Puppet 4 provides it due to the way Puppet Labs is packaging universal single packages for most platforms starting with Puppet 4.

The switch to Puppet 4 is of course what would necessitate making this a new major version.

nibalizer · 2015-09-21T23:22:00Z

From a code mgmt perspective I suggest this. Ideally it could/would apply to every module in p-c.

Cutover:

Branch currrent master to 1.x
Merge this PR to master
Add a Box such as https://github.com/puppetlabs/puppetlabs-stdlib#version-compatibility to the README explaining which versions of the module are compatible with which puppet versions.
Release 2.0.0 of the module to the forge
Release the tip of 1.x to the forge as 1.1.whatever if it has been long enough since the last 1.x release.

Ongoing:

New features should land on 2.x.
Bugfixes and backports can land on 1.x, but only when things seem super broken and important. We should strive to help the consumers move to puppet4.
Releases of the 2.x line can continue as normal.
Releases of the 1.x.z line (probably .z bugfix releases) can be released at any time. I'm personally okay with releasing after every PR since there shouldn't be many of them and we'd only get that PR if someone was in a rough spot to begin with.
At some point in the future we'll depreciate all puppet3 compatibility.

liamjbennett · 2015-09-28T14:33:44Z

Ok, I am happy for this change to go ahead and I will do all the required stuff to release a new major version on this.

@reidmv can you fix the tests first on this? Thanks.

reidmv · 2015-09-28T18:58:34Z

I can, but not soon. I'm pretty swamped until the end of October. If I get a chance to slip in more work on this before then I will. I'll post an update by October 27.

nibalizer · 2015-09-29T06:22:10Z

I'm trying to get the tests passing in #10

reidmv · 2015-11-07T00:43:53Z

@liamjbennett @nibalizer tests cleaned up and PR rebased. Ready for review.

vincent-aguillaume · 2015-11-09T12:52:37Z

manifests/init.pp

-        /^Windows (XP|Vista|7|8|8.1).*/: { $type = 'package' }
-        default: { $type = 'err' err("dotnet ${version} is not support on this version of windows") }
+      case $windows_version {
+        /^2008/, /^2012/:  { $type = 'feature' }


On Windows server 2008 (non R2), dotnet 3.5 is a package.

Thanks @vincent-aguillaume - I've added a commit to fix the error.

reidmv · 2015-11-30T21:59:11Z

@liamjbennett @nibalizer bump

jyaworski · 2016-01-26T01:49:16Z

Hello:

What's the status of this PR? Can it be rebased and squashed?

reidmv · 2016-01-26T20:56:21Z

Rebased and squashed.

Note: tests are breaking due to bug, rodjek/rspec-puppet#361. Not sure what to do about that. Works with rspec-puppet 2.3.0

liamjbennett · 2016-02-08T11:20:36Z

@reidmv can you update to rspec-puppet 2.3.2 and try again?

reidmv · 2016-02-08T20:24:55Z

@liamjbennett seems to work! Needed to do style cleanup for rubocop to get that to pass, but the Travis build is green now.

jyaworski · 2016-02-16T23:36:57Z

Can this be squashed?

reidmv · 2016-02-16T23:39:12Z

@jyaworski squashed.

reidmv · 2016-02-17T00:58:33Z

Now modified (and squashed) to comply with strict variables.

jyaworski · 2016-02-17T03:33:42Z

Gemfile

@@ -16,7 +16,7 @@ source 'https://rubygems.org'

 group :test do
  gem 'rake'
-  gem 'puppet', ENV['PUPPET_VERSION'] || '~> 3.8.0'
+  gem 'puppet', ENV['PUPPET_VERSION'] || '~> 4.0'


Why this change?

@jyaworski This PR is proposing a major version update, and uses Puppet 4 language features. Therefore it is not compatible with any version of Puppet <4.0. I wrote a little more about that here.

The Gemfile is (or will be) managed by modulesync.

https://github.com/voxpupuli/modulesync_config

Please raise a PR over there about it, or this will get overwritten the next time we sync.

I've added an issue. voxpupuli/modulesync_config#102.

reidmv · 2016-03-25T18:33:46Z

@jyaworski PR refactored to take modulesync into account.

bastelfreak · 2016-06-05T13:05:35Z

Hi @reidmv @liamjbennett, how is it planned to process here? @liamjbennett do you want to do the release (we should do another modulesync before that)?

liamjbennett · 2016-06-05T14:24:36Z

Yes I want this to be merged into master. I am not in a position to do the release myself at the moment but if someone else can prepare it on my behalf it would be greatly appreciated.

bastelfreak · 2016-08-19T08:36:58Z

@liamjbennett can we talk about the further steps on IRC? I'm currently doing a modulesync in #20, I would like to release the module in the next week.

Previously a Powershell exec was used to install packages. This commit converts that task to a package-based resource. It allows cleaner, easier to read code. Switch from download_file to remote_file. This allows for better subscribe/notify of the dotnet defined type as remote_file does not have any intermediate file fragment resources which may be used or changed even if the file to download is not changed. The remote_file resource is also platform agnostic, unlike download_file, more meaning it may be accessible and familar to a wider range of potential users/contributors. Update for Puppet 4 and remove deps Use Puppet 4's type system to deal with validation and remove dependency on stdlib. Use the $::os fact to get Windows release version and remove dependency on windowsfacts. Add logic to prevent incompatible versions Some versions of .NET are in-place upgrades of others. Installation of .NET 4.5, for example, will replace the 4.0 package. In order to disallow Puppet from continuously trying to install .NET 4.0 in the event dotnet resources for both 4.0 and 4.5 have been added to the catalog, create a package=absent resource for each conflicting version. This will cause the conflict to be caught when a catalog is compiled for the node. Add ability to recognize built-in .NET versions Some OSes have some versions of .NET built in and so do not need to have it installed via package. It seems likely this was auto-detected based on registry key presence/absence via exec previously. Because we're now managing the package directly, we should be more explicit about whether or not the package actually needs to be installed. Clean up installation type detection Regex adjustments to make it easier to read which versions support which installation types. Only test on 4.x/future parser Set gemfile to use PUPPET_GEM_VERSION and set default to 4.0 Adjust tests to match updates This is not a comprehensive test overhaul, but does update the functional tests to work with the updated code. Logic correction - For 2008, .NET 3.5 is a package Previously the dotnet define logic would attempt to install .NET 3.5 as a feature on Server 2008. This was incorrect, as in Server 2008 .NET 3.5 was not available as a feature and needed the package installation. Add Dism install type for Windows 7 Windows 7 does not install .NET as a package, nor does it have the ServerManager module. Therefore provide an alternative means of managing the feature via DISM. Fix non-functional onlyif in powershell exec The Puppet Exec resource uses the return code of command, unless, and onlyif to determine success/failure or if action is necessary. Previously, the onlyif in dotnet::install::feature would never exit with a non-zero exit code, even if the Test-Path command returned a False object. Return object and exit code are not the same thing. This commit modifies the onlyif to ensure that if the Test-Path command returns false a non-zero exit code will occur. (style) Line up case blocks Add feature parameter to dotnet::install::feature This will allow specification of different feature names. Necessary specifically because the feature name to install .NET 3.5 on Server 2012 is not the same feature as to install 4.5. The feature name for .NET 3.5 in Server 2012 is not AS-NET-Framework, which we use as the default feature. This commit also updates the dotnet type to attempt installation of the correct feature when 3.5 is specified on Server 2012.

vinzent · 2017-12-16T13:24:56Z

closing because of no progress since over a year

reidmv force-pushed the update branch from e8a8cc9 to 868fb77 Compare November 7, 2015 00:43

vincent-aguillaume reviewed Nov 9, 2015
View reviewed changes

reidmv force-pushed the update branch from c8b4191 to 4849b24 Compare November 9, 2015 19:53

nibalizer mentioned this pull request Jan 6, 2016

RFC: Support Puppet 4 only voxpupuli/plumbing#21

Closed

7 tasks

reidmv force-pushed the update branch from c841b36 to cbeb199 Compare January 26, 2016 20:55

reidmv force-pushed the update branch 2 times, most recently from 2ddda63 to ae4c4a4 Compare February 8, 2016 19:34

reidmv force-pushed the update branch from a80e1c4 to 11f5c32 Compare February 16, 2016 23:39

reidmv force-pushed the update branch from 11f5c32 to c23d04f Compare February 17, 2016 00:55

jyaworski reviewed Feb 17, 2016
View reviewed changes

reidmv force-pushed the update branch 5 times, most recently from c96d57a to ab9392f Compare March 25, 2016 18:26

reidmv force-pushed the update branch from ab9392f to f9308aa Compare March 25, 2016 18:27

reidmv force-pushed the update branch 3 times, most recently from ec63307 to 80ab699 Compare September 13, 2016 20:37

reidmv added 2 commits September 13, 2016 13:42

modulesync 0.12.6

044d80c

reidmv force-pushed the update branch from 126e037 to 044d80c Compare September 13, 2016 20:42

vinzent closed this Dec 16, 2017

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Major update #9

Major update #9

reidmv commented Sep 21, 2015

liamjbennett commented Sep 21, 2015

reidmv commented Sep 21, 2015

nibalizer commented Sep 21, 2015

liamjbennett commented Sep 28, 2015

reidmv commented Sep 28, 2015

nibalizer commented Sep 29, 2015

reidmv commented Nov 7, 2015

vincent-aguillaume Nov 9, 2015

reidmv Nov 9, 2015

reidmv commented Nov 30, 2015

jyaworski commented Jan 26, 2016

reidmv commented Jan 26, 2016

liamjbennett commented Feb 8, 2016

reidmv commented Feb 8, 2016

jyaworski commented Feb 16, 2016

reidmv commented Feb 16, 2016

reidmv commented Feb 17, 2016

jyaworski Feb 17, 2016

reidmv Feb 17, 2016

jyaworski Feb 17, 2016

reidmv Mar 21, 2016

reidmv commented Mar 25, 2016

bastelfreak commented Jun 5, 2016

liamjbennett commented Jun 5, 2016 •

edited

Loading

bastelfreak commented Aug 19, 2016

vinzent commented Dec 16, 2017

Major update #9

Major update #9

Conversation

reidmv commented Sep 21, 2015

liamjbennett commented Sep 21, 2015

reidmv commented Sep 21, 2015

nibalizer commented Sep 21, 2015

liamjbennett commented Sep 28, 2015

reidmv commented Sep 28, 2015

nibalizer commented Sep 29, 2015

reidmv commented Nov 7, 2015

vincent-aguillaume Nov 9, 2015

Choose a reason for hiding this comment

reidmv Nov 9, 2015

Choose a reason for hiding this comment

reidmv commented Nov 30, 2015

jyaworski commented Jan 26, 2016

reidmv commented Jan 26, 2016

liamjbennett commented Feb 8, 2016

reidmv commented Feb 8, 2016

jyaworski commented Feb 16, 2016

reidmv commented Feb 16, 2016

reidmv commented Feb 17, 2016

jyaworski Feb 17, 2016

Choose a reason for hiding this comment

reidmv Feb 17, 2016

Choose a reason for hiding this comment

jyaworski Feb 17, 2016

Choose a reason for hiding this comment

reidmv Mar 21, 2016

Choose a reason for hiding this comment

reidmv commented Mar 25, 2016

bastelfreak commented Jun 5, 2016

liamjbennett commented Jun 5, 2016 • edited Loading

bastelfreak commented Aug 19, 2016

vinzent commented Dec 16, 2017

liamjbennett commented Jun 5, 2016 •

edited

Loading