Installation solution sets in 2.2.5 can end up out of order (not seen
in 2.3.1) causing LoadErrors when the specification is in the solution
set during validation. This detects the missing spec within the solution
and if found will move spec to the start of the solution set and retry
solution activation.
Refactors reusable actions into isolated methods. Supports installation/removal
without activation to prevent unintended conflicts during upgrades and cleanup.
Introduced custom resolver set to handle multiple installed versions of gems
which enables proper cleanup.
Allows checksum validation on downloaded files via Util::Downloader
using MD5 and/or SHA1 checksums. This also integrates checksum validation
support with the shell provisioner for downloaded remote files.
This is required since Gem::Version will mangle pre-release version
numbers.
To get around this, we keep a mapping of
`Gem::Version.to_s => versiondir.`
This allows us to properly sort by version when picking a box while
still returning pristine versiondirs when required.
The `pkg_add` command will return `0` when a package requested for
installation is not found. This adds a validation check to ensure
the rsync package is actually installed on the guest.
Previously, there was no way to close the STDIN stream of a subprocess,
so commands that read from stdin in a subprocess would hang forever,
such as `/bin/sh -s`. If one tried to close the stdin, the
IO.select() call in Subprocess#execute would raise an error for calling
select() on a closed IO.
Here's a concrete example of a command that needs to close STDIN to work
properly:
```ruby
script = SOME_VERY_LONG_STRING
command = %w(ssh foo.example.com /bin/sh -s foo bar)
result = ::Vagrant::Util::Subprocess.execute(*command) do |type, data_or_io|
if type == :stdin
data_or_io.write(script)
data_or_io.write("\n")
data_or_io.close
next
end
puts "Remote: #{data_or_io}"
end
```
BSD-based guests do not support VirtualBox shared folders. This is a
common source of confusion in Vagrant. This new error clearly explains
that this is not a bug in Vagrant and provides instructions on how to
disable them.
Currently, the supported way (see GH#1922) to disable the default port
2222 port forward is:
config.vm.network :forwarded_port, guest: 22, host: 2222, disabled: true
However, the port collision detection runs on all ports, regardless of
the `disabled` flag. This leads it to attempt to connect to port
2222, notice it is taken, and abort with a port conflict -- even
though it will not be attempting to use the port at all.
Skip disabled ports when doing port conflict detection.
A workaround that does not require a Vagrant upgrade to one containing
this fix is to instead set `auto_correct` on the disabled port:
config.vm.network :forwarded_port, guest: 22, host: 2222, disabled: true, auto_correct: true
This allows the disabled port to be reshuffled off to some other
unused port.
The docs for Ruby say Pathname#rmtree will recursively delete, but
apparently that is a lie, at least on Windows (see GH-7496). Switch to
using FileUtils to ensure the directory is deleted.
Fixes GH-7496
This commit changes Vagrant::Util::Platform to cache the result of some
common operations. These values are highly unlikely to change over the
course of a single Vagrant run and they are only cached for that run.
This adds two new SSH configuration options:
- `keys_only`
- `paranoid`
These values were previously hard-coded, but can now be user-specified.
Fixes GH-4275
This commit basically grepped the code base for all uses of Dir.mktmpdir
and Tempfile.new/open and ensures the value is unique within the
code base and also prefixed with `vagrant-`.
Previously, most invocations of these commands simply used "vagrant",
thus making them indistinguishable when trying to identify leaks.
The only reason we were using Tempfile was to generate the path. This
commit switches to using `Dir::Tmpname.create`, which accomplishes the
same thing without the overhead of creating and removing a tempfile.
This commit separates the scratch and output directory creation from the
main package middleware into its own PackageSetupFolders middleware.
Additionally, the validation that ensures an output file does not exist
is moved into a validation function that can be shared across multiple
methods.
This refactor permits a pre-flight check to ensure box packaging would
be successful before actually stopping the VM.
Fixes GH-7351
This commit bumps the box resume delay time to 24h and makes it a
constant because I had a bugger of a time actually finding this in the
code :frowny:.
Fixes GH-7272
Ensure multi machine access of other machine state information through
iterating `active_machines` and retrieval of cached machines cannot
have multiple threads update the state of machines simultaneously as
this triggers a Machine Lock exception.
Machine state information retrieved from the index, returns a locked
object. Since iteration of active_machine, and retrieval of each
machine from the cache can be triggered by any plugin, it is possible
for another machine to inadvertently access the state and trigger an
update, which the thread owning the machine is currently in the process
of updating it already. This results in a Machine Locked exception
occurring if the attempt to retrieve the cached state from the index
occurs before the other thread calls release.
Partially-Fixes: #6526
Added support for Port forwarding in an IP aliased environment. The change
makes the following forwarding rule(s) possible.
Ex: eth0 is ip aliased to have a range of IP addresses 10.20.30.0/24.
In the Vagrant file, we can now have an entry like the following and
it will just work! Note the host port 8081 is the same for both .1 and .2.
Vagrant.configure("2") do |config|
config.vm.network "forwarded_port", guest: 80, host: 8080
config.vm.network "forwarded_port", guest: 81, host: 8081, host_ip: 10.20.30.1
config.vm.network "forwarded_port", guest: 82, host: 8081, host_ip: 10.20.30.2
end
Without this change, the version string may be mutated when passed through
Gem::Version.new(version).to_s before being used to construct the box
directory. This is a problem when using prerelease version numbers because,
for example, it will look for "~/.vagrant.d/boxes/1.0.0.pre.alpha.1" when it
should be looking for "~/.vagrant.d/boxes/1.0.0-alpha.1".
This helps with some confusion caused in GH-2538, since the output says:
> Running cleanup tasks for 'shell' provisioner...
But that's actually not true. It is running the cleanup tasks iff the
provisioner defined a cleanup task. This commit changes the
provisioner_cleanup middleware to only run cleanup tasks if the subclass
defines a cleanup task.
The reason we can't just check if the provisioner `respond_to?` the
`cleanup` method is because the parent provisioner base class (which
all provisioners inherit from) defines a blank cleanup method. This is
important because it means we never risk calling an unimplemented
cleanup function, and it also helps define the public API for a
provisioner.
This patch introduces a new parameter --all for the remove
command of the box plugin. Setting this parameter will remove
all available versions of a specific box.
Example usage:
```
$ vagrant box list
ubuntu/trusty64 (virtualbox, 20150427.0.0)
ubuntu/trusty64 (virtualbox, 20150430.0.0)
ubuntu/trusty64 (virtualbox, 20150506.0.0)
```
```
$ vagrant box remove ubuntu/trusty64
You requested to remove the box 'ubuntu/trusty64' with provider
'virtualbox'. This box has multiple versions. You must
explicitly specify which version you want to remove with
the `--box-version` flag. The available versions for this
box are:
* 20150427.0.0
* 20150430.0.0
* 20150506.0.0
```
With the --all parameter it is possible to remove all versions at once.
```
$ vagrant box remove --all ubuntu/trusty64
Removing box 'ubuntu/trusty64' (v20150506.0.0) with provider 'virtualbox'...
Removing box 'ubuntu/trusty64' (v20150430.0.0) with provider 'virtualbox'...
Removing box 'ubuntu/trusty64' (v20150427.0.0) with provider 'virtualbox'...
```
With this change, the existing host-based Ansible provisioner is
refactored to share a maximum of code with this new guest-based Ansible
provisioner.
At this stage of development, the existing unit tests are intentionally
modified as little as possible, to keep safe the existing funtionalities.
Other issues resolved by this changeset:
- Display a warning when running from a Windows host [GH-5292]
- Do not run `ansible-playbook` in verbose mode when the `verbose` option
is set to an empty string.
Changed the name of the error LinuxRDesktopNotFound to
LinuxRDPClientNotFound and re-worded error text in
templates/locales/en.yml to include `xfreerdp` when listing supported
RDP clients.
This causes issues since the child environment almost certainly doesn't
share data with the parent. In a larger scope, we should find a way to
encode the data path somehow on `vagrant up`.
Here we implement a naive solution to #5605 which catches the case that
a provided source contains an object which cannot be inspected, because
an object contained within in has an #inspect string that returns a
string that is incompatible with the encoding in
`Encoding.default_external` or a string which cannot be downcast to
7-bit ascii.
The Ruby VM implementation of "#inspect" implements this checking on
these lines of code: http://git.io/vZYNS. A Ruby level override of
this method does not cause this problem. For example:
```ruby
class Foo
def inspect
"😍".encode("UTF-16LE")
end
```
will not cause the problem, because that's a Ruby implementation and the
VM's checks don't occur.
However, if we have an Object which **does** use the VM implementation
of inspect, that contains an object that has an inspect string which
returns non-ascii, we encounter the bug. For example:
```ruby
class Bar
def inspect
"😍".encode("UTF-16LE")
end
end
class Foo
def initialize
@bar = Bar.new
end
end
Foo.new.inspect
```
Will cause the issue.
The solution this patch provides basically catches the encoding error
and inserts a string which attempts to help the user work out which
object was provided without blowing up. Most likely, this was caused
by a user having a weird encoding coming out of one of the sources
passed in, but without a full repro case, it's not clear whether a patch
should be applied to a different object in the system.
Closes#5605.
With this change, the `Vagrant::plugins_enabled?` is now false when the
embedded Bundler is not available.
Resolve the broken RSpec unit tests after
479323f1e8.
Deleting the lock file can fail when another process is currently trying to acquire it (-> race condition).
It is safe to ignore this error since the other process will eventually acquire the lock and again try to delete the lock file.
If the Vagrantfile has some kind of error, display not only
its path and the exception message, but also the originating
line number and exception class.
Also log the full backtrace when the error is in a provider
block, just as it is done when it's outside a provider block.
Use of $stdin, $stdout, and $stderr globals makes testing difficult. By
exposing the IO objects as writable attributes, input/output can be more
easily simulated using StringIO or doubles.
Importing a base box from the local file system currently outputs 'Downloading: file://...' which is more accurate now by presenting it as: 'Unpacking necessary files from: file://...'.
Fixes#5386.
I found this output to be very helpful in debugging an action hook
problem in a plugin. problem. I'm not sure why the callable_id is
useful (seem to always show up as an instance of `Builder` or `Warden`
for me), but I left it in there just in case it's useful to someone.
This makes sure that config.vm.box_download_checksum and
config.vm.box_download_checksum_type get passed to
Vagrant::Action.action_box_add with other options on
box download/import.
If one passes a version spec to `Vagrant.has_plugin?`, the actual plugin name is ignored and any installed plugin which matches the version spec makes the check pass. Fix this by also checking for plugin name match in addition to the requested version spec.
This commit finally removes all traces of VirtualBox defaults from
Vagrant core. Vagrant now completely relies on its automated provider
detection to pick the best and most relevant provider that is available
on the system and for a specific development environment.
A "vagrant up" on one development environment might prefer VirtualBox,
another might prefer Docker. As a result of this commit (plus the few
prior), the developer doesn't need to know anymore and doesn't need to
specify a `--provider` flag. Vagrant just figures out the correct
provider.
Yay. :)
With this change, any caller of machine.ssh_info is assured that best
efforts will be done to fix possible wrong permissions on the private
key files.
Fix#4652
https://github.com/mitchellh/vagrant/pull/4282 introduced new setting to set alternate vagrant cloud server URLs, but it doesn't work actually.
This PR helps to process the value correctly.
Vagrant::Box.load_metadata did not provide a way to specify the HTTPS
download options that could be specified when downloading boxes
(ca cert, ca path, client cert, insecure). As a result, while it was
possible to add a box whose metadata file needed to be downloaded with one of
those options specified, it was impossible to check for updates. The following
changes have been made to address the situation:
1. Create a DownloadMixins module to provide the --insecure, --cacert, --capth,
and --cert command line options to all of `vagrant box add`,
`vagrant box update`, and `vagrant box outdated`.
2. Extend `Vagrant::Box.has_update?` and `Vagrant::Box.load_metadata` to accept
said download options.
3. Extend `box outdated` and `box update` commands to pass download options
down.
4. Extend `Vagrant::Builtin::Action::BoxCheckOutdated` to honour download
options.
5. Options specified on the command line take precedence over options specified
in the machine configuration, if any.
6. Fix bug in `vagrant box add` where client cert was being passed down using
the wrong environment key.
7. Unit test coverage in update_test and box_check_outdated_test.
Resolves#4420
When specifying a timeout, the subprocess used to select for timeout seconds on the stdout, stderr pipes.
This creates a race condition and can cause the `Subprocess#execute` to hang for the full timeout interval if the process has already exited and no more output is send to the pipes.
This race is happening with a higher probability if the subprocess is generating a lot of output.
This commits adds a new config setting `config.vm.box_server_url` to set
the URL of a local VagrantCloud instance in the Vagrantfile. If the
environment variable `VAGRANT_SERVER_URL` is set, it will still be
preferred.
Chris Roberts
Chris Roberts
David O'Rourke
Luke Bakken
Björn Brala
Jake Teton-Landis
Alejandro Ojeda
Zack Train
Seth Vargo
Carlos Rodrigues
Carlos Rodrigues
Alex Vandiver
Colin Walters
Markus Rossi
Gilles Cornu
Mikhail Zholobov
Karl Isenberg
Christian Hofstaedtler
Darragh Bailey
Pravinchandar Raajendiran
Mark Mickan
Peter Mosmans
Mitchell Hashimoto
Mitchell Hashimoto
Reed Loden
Timur Alperovich
Eli Skeggs
Austin Schutz
Christian Berendt
Eric Winkelmann
Joseph Frazier
Sam Phippen
Ievgen Prokhorenko
Manuel Pöter
Marc Abramowitz
Philip Wrenn
Adam Spiers
Dmitry Moskalchuk
Matija K
Aneesh Agrawal
Matthew Rahtz
Haz
Jean-Francois Bibeau
Emilien Kenler
Dan Duvall
edward010
Tim Rensen
Mikhail Zholobov
Paul Hinze
Max Lincoln
Simon Vetter
mbrodala
Shawn Neal
Felix C. Stegerman
Seth Vargo
Matt Wrock
Rob Kinyon
Anduin Withers
Julien Vey
crypt1d
Michael Kuzmin
Ray Ruvinskiy
Veres Lajos
Nitzan Raz
Fabian Ruff
Franz Pletz
Elan Ruusamäe
Ilya Vassilevsky