The error details message should just be one string without spacing
so the terminal can set a output length. This will allow messages
from Ruby and Go errors to have similar formating.
Instead of having to update these flavor.rb files every time a
new version of CentOS/RHEL come out let's dynamically pick up
the version (7,8,9) from the machine.
This was inspired slightly by the Rocky Linux guest flavor.rb.
This migrates the CentOS guest plugin to identify machines by
what is in os-release rather than based on if the /etc/centos-release
file exists.
It just so happens that the Rocky Linux plugin is set up to inherit
directly from :redhat (not :centos) and thus the CentOS flavor is
attempted before Rocky in the depth first search. For some reason
the rockylinux/8 vagrant box still has the /etc/centos-release file
in it even though it's only supposed to inherit from RHEL.
The almalinux/9 box does not have /etc/centos-release.
find_types works fine until you have a module which has the same name but different case. For example, the VagrantVmware package is HashiCorp::VagrantVMwareDesktop. All the protos are at Hashicorp::Vagrant::… So, you end up with this fun
Object.constants.select { |n| n.to_s.downcase == "hashicorp" }
=> [:HashiCorp, :Hashicorp]
So, when trying to walk down the modules to find the right type to unany to, Vagrant sometimes takes the wrong path (eg. Down the HashiCorp module instead of the Hashicorp module).
This change will keep a list of the parent modules when walking down the module list. This way if a dead end is reached then Vagrant can go a level back and keep searching for the correct class.
Turns out I misinterpreted the behavior of a ruby method with a default
being passed nil so I assumed :parallel was effectively defaulting to
true when it is the opposite.
Starting a rocky 9 box (having nfs shares) fails with: `No match for argument: nfs-utils-lib`.
Apparently Rocky 9 does not have a `nfs-utils-lib` but mounting nfs shares without nfs-utils-lib works fine.
So updating the nfs client install to only install `nfs-utils-lib` if the package is available. That way older rhel clones
should continue to work as well.
This removes the need for the validation workaround for Docker, because
box_updated is once again available in that context.
We don't technically need the SyncedFolder priorities back on the Ruby
side, but wiring them through for symmetry.
Depends on https://github.com/hashicorp/vagrant-plugin-sdk/pull/183
By pruning machines that are in "unknown" status after each operation,
the Go code path exposed the fact that the Docker provider was not
updating the machine index during an "up" - leaving the state as
"unknown".
This is basically a bug within the Docker provider, so I think it's okay
to update the plugin code to correct this rather than working around the
issue in Go.
All we need to do is call `machine.state` instead of reaching through to
`machine.provider.state` while waiting for the container to be started.
That causes the extra logic for updating the machine index in
`machine.state` to fire.
This reverts commit 1c26a4abb0, reversing
changes made to 186824a568.
The changes broke plugin loading when using subcommands so these
changes will be reverted until the underlying issue can be
investigated and resolved.
The HashWithIndifferentAccess will change all keys to strings when
merged with another regular hash. This causes errors in cases where
hashes are merged, for example in the action builder. Symbol/String
type data is kept over the wire, so there is no need to make the
mapped hash a HashWithIndifferentAccess.
We had some cases where calling a capability that returned a boolean was
not getting correctly unpacked, so instead of `true` or `false` the
capability was putting out
`VagrantPlugins::CommandServe::Type::Boolean`.
This may have been happening in _all_ cases where a boolean was returned
from a capability and we just didn't notice it yet because the return
value was always truthy.
These tweaks should help ensure that Ruby types make it out where they
are supposed to be in Args::Direct usage.
Relying on HashWithIndifferentAccess in Hash mappers results in
errors when hashes get merged in Ruby. When merges between regular
hashes and HashWithIndifferentAccess happens, then all the keys
from HashWithIndifferentAccess are transformed to strings.
- Defaultable needs to default to true when it's not specified
- We need to allow a non-defaultable provider to be selected if it shows
up in the config
* Populates ComponentOptions into plugin structs
* Maps options for legacy Provider Plugins into PluginOptions
* Demos use of PluginOptions in a stub provider
* Honors plugin priority and defaultable settings
It looks like I might have been the first to hit provider cabability
invocation in testing these changes, and so I found these few missing
methods on the client. They're just copied over from the other
capability hosts.
Calling capabilities on a provider also revealed that the wrong Machine
type was being pulled out of the funcspec args, so we had to correct
that too in order to get the capability calls working.
- Pulls in the SDK changes to Project.DefaultProvider and Project.Target
- Implements the hefty default provider method
- Un-hard-codes provider from Target, and sets it when a Target is
looked up from a Project
The "destroy_on_error" functionality for `vagrant up` is implemented in
the `recover()` action chain, and works by firing off a destroy action
from inside that chain.
This is all well and good, but it copies its existing `env` which has
had `action_name` set for the up action. This was causing action_hooks
for up actions to attach to this destroy action stack.
Setting the action_name explicitly in the env before firing the runner
should correct the behavior. I'm not sure if raw_action_name is used
anywhere but I figured it was better to be consistent vs conservative in
what we change.
In legacy Vagrant, any exception raised that's a subclass of
Vagrant::Errors::VagrantError is considered user-facing and so causes
the error message to be printed to the console and the process to use
exit code 1. Anything outside of that causes the process to use exit
code 255. (See `bin/vagrant` for the code.)
Here we mirror that behavior by treating errors that have a
LocalizedMessage as user-facing and those without as unexpected. This
allows the basic virtualbox component to pass in vagrant-spec!
* `vagrant ssh` has a --no-tty flag colliding with the one defined in
`bin/vagrant` - in fact none of the flags in `bin/vagrant` are
processed in `serve` mode, so remove the code that captures them from
the CommandInfo OptionParser dance
* `vagrant ssh` has a `--plain` flag colliding with the one defined in
`internal/cli/base.go` - this flag was inherited from Waypoint, so we
can just rename it to line up with the (inversely defined) `--color`
flag used in legacy vagrant
Some commands like `vagrant init` and `vagrant box` should be able to
run successfully without a full Project available in VAGRANT_CWD (in
other words, they don't require that a valid Vagrantfile be available.)
Thus far we've been assuming that a Project is available when
dispatching commands, which mean that commands of this nature weren't
working.
Here we make the Basis available to serve as an alternative client to
Vagrant::Environment::Remote such that it can be instantiated and passed
through to commands. This required some changes to Environment::Remote
to make its interactions with the client more defensive, but we manage
to avoid needing to make any changes to the normal legacy codepaths.
When the opts are being pass through the go side, we get back empty
string values instead of nil values, so the `||=` assignment was not
working to populate the default owner and group. This was causing the
rsync_post hook to fail on linux guests.
This is a pass through test failures and deprecation warnings:
* Make all ambiguous `.with(..., key: val)` use explicit hashes to
prevent test failures for argument mismatch in Ruby 3.0
* Scope down all unbounded `raise_error` to address warnings (remove
one test that was revealed to be referencing a nonexistent variable
once the raise_error was scoped.)
* Update all `any_instance` usage to new syntax to address warnings
* Allow the service cache to be cleared and do so between some tests
* Fix a small bug in with_plugin's plugin not found code path (revealed
by a scoped and_raise)
Update the Ruby service implementations to use the funcspec util
module for generating spec content. A helper method is now used
for generating a parent class for services to subclass which
automatically includes all required modules for usage.
The IsRunning action checks if `env[:machine].state.id == :running` but
this check was never passing as the protobuf-washed version of machine
state was yielding a machine state w/ a string value like `"running"`.
Easy fix in the mapper!
