Vagrant config

[swalkinshaw]

Inspired by https://github.com/geerlingguy/drupal-vm

• Move methods into separate file
• Intro Vagrant config

Todo:

• Better name for vagrant.yml. drupal-vm uses default.config.yml?
• Support for untracked config file?

[mAAdhaTTah]

Can I make a suggestion: I edit my local Vagrantfile to mount WP plugins I'm working on into the VM. Can the configuration be edited to allow one to include a src/dest array:

mount_dirs:
    - src: 'path/to/plugin'
      dest: 'path/to/wp-content/plugins/plugin'
    - src: 'path/to/plugin2'
      dest: 'path/to/wp-content/plugins/plugin2'

or something like this?

[swalkinshaw]

@mAAdhaTTah sure, makes sense 👍

[swalkinshaw]

@mAAdhaTTah added, try it out 👼

[fullyint]

This PR will make the Trellis Vagrant config so much nicer and cleaner!

Important. The one issue I think must be dealt with is how the vagrant_synced_folders loop handles bindfs folders. I've added two inline comments in the code to explain. It avoids a vagrant-bindfs error such as this:

vagrant-bindfs:
* Path '../db' is relative but an absolute path is attended

Probably. One issue you probably want to address is fixing the sequence of the CHANGELOG.md entries to be chronological (fixing sequence with Ansible 2.3 compatibility entry).

Maybe. There are two issues you'll only maybe want to address:

1. You could change the vagrant_ansible_version to 2.3
2. You could move def hosts(sites) up under def load_wordpress_sites (in lib/trellis/vagrant.rb) because they're similar in concept and that would match the sequence and proximity in which they are called in Vagrantfile.

Probably not. One issue you will probably not care to address is that you could put a comment at the top of vagrant.default.yml saying "Jinja2 variables and templating will not function in this file."

[fullyint]

This could be used for nfs path names in the vagrant_synced_folders loop:

- def nfs_path(site_name)
-   "/vagrant-nfs-#{site_name}"
+ def nfs_path(path)
+   "/vagrant-nfs-#{File.basename(path)}"
  end

[fullyint]

When using vagrant-bindfs, the convention appears to be...

1. do a regular vagrant folder sync from somedir to /vagrant/somedir-nfs
2. do a bindfs sync from /vagrant/somedir-nfs to /final_path_on_vm/somedir

Combined with the File.basename() adjustment to nfs_path() proposed in a separate comment, the following seems to handle it.

+ destination_folder = folder.fetch('bindfs', true) ? nfs_path(folder['destination']) : folder['destination']

- config.vm.synced_folder folder['local_path'], folder['destination'], options
+ config.vm.synced_folder folder['local_path'], destination_folder, options

  if folder.fetch('bindfs', true)
-   config.bindfs.bind_folder folder['local_path'], folder['destination'], options
+   config.bindfs.bind_folder destination_folder, folder['destination'], folder.fetch('bindfs_options', {})

The other piece to this is the bindfs_options, which are different from a regular vagrant sync dir options and must be treated differently. The above appears to successfully handle the following example

# vagrant.local.yml
vagrant_synced_folders:
  - local_path: ../db
    destination: /home/vagrant/db
    create: false
    type: nfs
    bindfs: true
    bindfs_options:
      o: 'nonempty'
      p: '0644,a+D'
  - local_path: ../test
    destination: /home/vagrant/test
    create: false
    type: virtualbox
    bindfs: false
    mount_options:
      - 'dmode=755'
      - 'fmode=644'

[swalkinshaw]

👌 updated

[fullyint]

👍 Looks great to me!