In this case, "atomic" means that there will be no situation where the
file contains only part of the newContent data, and therefore other
software monitoring the file for changes (using a mechanism like inotify)
won't encounter a truncated file.
It does _not_ mean that there can't be existing filehandles open against
the old version of the file. On Windows systems the write will fail in
that case, but on Unix systems the write will typically succeed but leave
the existing filehandles still pointing at the old version of the file.
They'll need to reopen the file in order to see the new content.
This originated in the cliconfig code to write out credentials files. The
Windows implementation of this in particular was quite onerous to get
right because it needs a very specific sequence of operations to avoid
running into exclusive file locks, and so by factoring this out with
only cosmetic modification we can avoid repeating all of that engineering
effort for other atomic file writing use-cases.
This builds on an experimental feature in the underlying cty library which
allows marking specific attribtues of an object type constraint as
optional, which in turn modifies how the cty conversion package handles
missing attributes in a source value: it will silently substitute a null
value of the appropriate type rather than returning an error.
In order to implement the experiment this commit temporarily forks the
HCL typeexpr extension package into a local internal/typeexpr package,
where I've extended the type constraint syntax to allow annotating object
type attributes as being optional using the HCL function call syntax.
If the experiment is successful -- both at the Terraform layer and in
the underlying cty library -- we'll likely send these modifications to
upstream HCL so that other HCL-based languages can potentially benefit
from this new capability.
Because it's experimental, the optional attribute modifier is allowed only
with an explicit opt-in to the module_variable_optional_attrs experiment.
Previously we were just letting hclwrite do its default formatting
behavior here. The current behavior there isn't ideal anyway -- it puts
big data structures all on one line -- but even ignoring that our goal
for this file format is to keep things in a highly-normalized shape so
that diffs against the file are clear and easy to read.
With that in mind, here we directly control how we write that value into
the file, which means that later changes to hclwrite's list/set
presentation won't affect it, regardless of what form they take.
This probably isn't the best UI we could do here, but it's a placeholder
for now just to avoid making it seem like we're ignoring the lock file
and checking for new versions anyway.
This changes the approach used by the provider installer to remember
between runs which selections it has previously made, using the lock file
format implemented in internal/depsfile.
This means that version constraints in the configuration are considered
only for providers we've not seen before or when -upgrade mode is active.
These are helper functions to give the installation UI some hints about
whether the lock file has changed so that it can in turn give the user
advice about it. The UI-layer callers of these will follow in a later
commit.
helper/copy CopyDir was used heavily in tests. It differes from
internal/copydir in a few ways, the main one being that it creates the
dst directory while the internal version expected the dst to exist
(there are other differences, which is why I did not just switch tests
to using internal's CopyDir).
I moved the CopyDir func from helper/copy into command_test.go; I could
also have moved it into internal/copy and named it something like
CreateDirAndCopy so if that seems like a better option please let me
know.
helper/copy/CopyFile was used in a couple of spots so I moved it into
internal, at which point I thought it made more sense to rename the
package copy (instead of copydir).
There's also a `go mod tidy` included.
This new-ish package ended up under "helper" during the 0.12 cycle for
want of some other place to put it, but in retrospect that was an odd
choice because the "helper/" tree is otherwise a bunch of legacy code from
when the SDK lived in this repository.
Here we move it over into the "internal" directory just to distance it
from the guidance of not using "helper/" packages in new projects;
didyoumean is a package we actively use as part of error message hints.
We no longer need to support 0.12-and-earlier-style provider addresses
because users should've upgraded their existing configurations and states
on Terraform 0.13 already.
For now this is only checked in the "init" command, because various test
shims are still relying on the idea of legacy providers the core layer.
However, rejecting these during init is sufficient grounds to avoid
supporting legacy provider addresses in the new dependency lock file
format, and thus sets the stage for a more severe removal of legacy
provider support in a later commit.
In earlier commits we started to make the installation codepath
context-aware so that it could be canceled in the event of a SIGINT, but
we didn't complete wiring that through the API of the getproviders
package.
Here we make the getproviders.Source interface methods, along with some
other functions that can make network requests, take a context.Context
argument and act appropriately if that context is cancelled.
The main providercache.Installer.EnsureProviderVersions method now also
has some context-awareness so that it can abort its work early if its
context reports any sort of error. That avoids waiting for the process
to wind through all of the remaining iterations of the various loops,
logging each request failure separately, and instead returns just
a single aggregate "canceled" error.
We can then set things up in the "terraform init" and
"terraform providers mirror" commands so that the context will be
cancelled if we get an interrupt signal, allowing provider installation
to abort early while still atomically completing any local-side effects
that may have started.
As we continue iterating towards saving valid hashes for a package in a
depsfile lock file after installation and verifying them on future
installation, this prepares getproviders for the possibility of having
multiple valid hashes per package.
This will arise in future commits for two reasons:
- We will need to support both the legacy "zip hash" hashing scheme and
the new-style content-based hashing scheme because currently the
registry protocol is only able to produce the legacy scheme, but our
other installation sources prefer the content-based scheme. Therefore
packages will typically have a mixture of hashes of both types.
- Installing from an upstream registry will save the hashes for the
packages across all supported platforms, rather than just the current
platform, and we'll consider all of those valid for future installation
if we see both successful matching of the current platform checksum and
a signature verification for the checksums file as a whole.
This also includes some more preparation for the second case above in that
signatureAuthentication now supports AcceptableHashes and returns all of
the zip-based hashes it can find in the checksums file. This is a bit of
an abstraction leak because previously that authenticator considered its
"document" to just be opaque bytes, but we want to make sure that we can
only end up trusting _all_ of the hashes if we've verified that the
document is signed. Hopefully we'll make this better in a future commit
with some refactoring, but that's deferred for now in order to minimize
disruption to existing codepaths while we work towards a provider locking
MVP.
The logic for what constitutes a valid hash and how different hash schemes
are represented was starting to get sprawled over many different files and
packages.
Consistently with other cases where we've used named types to gather the
definition of a particular string into a single place and have the Go
compiler help us use it properly, this introduces both getproviders.Hash
representing a hash value and getproviders.HashScheme representing the
idea of a particular hash scheme.
Most of this changeset is updating existing uses of primitive strings to
uses of getproviders.Hash. The new type definitions are in
internal/getproviders/hash.go.
Although origin registries return specific [filename, hash] pairs, our
various different installation methods can't produce a structured mapping
from platform to hash without breaking changes.
Therefore, as a compromise, we'll continue to do platform-specific checks
against upstream data in the cases where that's possible (installation
from origin registry or network mirror) but we'll treat the lock file as
just a flat set of equally-valid hashes, at least one of which must match
after we've completed whatever checks we've made against the
upstream-provided checksums/signatures.
This includes only the minimal internal/getproviders updates required to
make this compile. A subsequent commit will update that package to
actually support the idea of verifying against multiple hashes.
The "acceptable hashes" for a package is a set of hashes that the upstream
source considers to be good hashes for checking whether future installs
of the same provider version are considered to match this one.
Because the acceptable hashes are a package authentication concern and
they already need to be known (at least in part) to implement the
authenticators, here we add AcceptableHashes as an optional extra method
that an authenticator can implement.
Because these are hashes chosen by the upstream system, the caller must
make its own determination about their trustworthiness. The result of
authentication is likely to be an input to that, for example by
distrusting hashes produced by an authenticator that succeeds but doesn't
report having validated anything.
This is the pre-existing hashing scheme that was initially built for
releases.hashicorp.com and then later reused for the provider registry
protocol, which takes a SHA256 hash of the official distribution .zip file
and formats it as lowercase hex.
This is a non-ideal hash scheme because it works only for
PackageLocalArchive locations, and so we can't verify package directories
on local disk against such hashes. However, the registry protocol is now
a compatibility constraint and so we're going to need to support this
hashing scheme for the foreseeable future.
It doesn't make sense for a built-in provider to appear in a lock file
because built-in providers have no version independent of the version of
Terraform they are compiled into.
We also exclude legacy providers here, because they were supported only
as a transitional aid to enable the Terraform 0.13 upgrade process and
are not intended for explicit selection.
The provider installer will, once it's updated to understand dependency
locking, use this concept to decide which subset of its selections to
record in the dependency lock file for reference for future installation
requests.
This is an initial implementation of writing locks back to a file on disk.
This initial implementation is incomplete because it does not write the
changes to the new file atomically. We'll revisit that in a later commit
as we return to polish these codepaths, once we've proven out this
package's design by integrating it with Terraform's provider installer.
This is the initial implementation of the parser/decoder portion of the
new dependency lock file handler. It's currently dead code because the
caller isn't written yet. We'll continue to build out this functionality
here until we have the basic level of both load and save functionality
before introducing this into the provider installer codepath.
The version constraint parser allows "~> 2", but it behavior is identical
to "~> 2.0". Due to a quirk of the constraint parser (caused by the fact
that it supports both Ruby-style and npm/cargo-style constraints), it
ends up returning "~> 2" with the minor version marked as "unconstrained"
rather than as zero, but that means the same thing as zero in this context
anyway and so we'll prefer to stringify as "~> 2.0" so that we can be
clearer about how Terraform is understanding that version constraint.
For reasons that are unclear, these two tests just started failing on
macOS very recently. The failure looked like:
PackageDir: strings.Join({
"/",
+ "private/",
"var/folders/3h/foobar/T/terraform-test-p",
"rovidercache655312854/registry.terraform.io/hashicorp/null/2.0.0",
"/windows_amd64",
},
Speculating that the macOS temporary directory moved into the /private
directory, I added a couple of EvalSymlinks calls and the tests pass
again.
No other unit tests appear to be affected by this at the moment.
If a provider changes namespace in the registry, we can detect this when
running the 0.13upgrade command. As long as there is a version matching
the user's constraints, we now use the provider's new source address.
Otherwise, warn the user that the provider has moved and a version
upgrade is necessary to move to it.
We previously had this just stubbed out because it was a stretch goal for
the v0.13.0 release and it ultimately didn't make it in.
Here we fill out the existing stub -- with a minor change to its interface
so it can access credentials -- with a client implementation that is
compatible with the directory structure produced by the
"terraform providers mirror" subcommand, were the result to be published
on a static file server.
Earlier we introduced a new package hashing mechanism that is compatible
with both packed and unpacked packages, because it's a hash of the
contents of the package rather than of the archive it's delivered in.
However, we were using that only for the local selections file and not
for any remote package authentication yet.
The provider network mirrors protocol includes new-style hashes as a step
towards transitioning over to the new hash format in all cases, so this
new authenticator is here in preparation for verifying the checksums of
packages coming from network mirrors, for mirrors that support them.
For now this leaves us in a kinda confusing situation where we have both
NewPackageHashAuthentication for the new style and
NewArchiveChecksumAuthentication for the old style, which for the moment
is represented only by a doc comment on the latter. Hopefully we can
remove NewArchiveChecksumAuthentication in a future commit, if we can
get the registry updated to use the new hashing format.
The installFromHTTPURL function downloads a package to a temporary file,
then delegates to installFromLocalArchive to install it. We were
previously not deleting the temporary file afterwards. This commit fixes
that.
The SearchLocalDirectory function was intentionally written to only
support symlinks at the leaves so that it wouldn't risk getting into an
infinite loop traversing intermediate symlinks, but that rule was also
applying to the base directory itself.
It's pretty reasonable to put your local plugins in some location
Terraform wouldn't normally search (e.g. because you want to get them from
a shared filesystem mounted somewhere) and creating a symlink from one
of the locations Terraform _does_ search is a convenient way to help
Terraform find those without going all in on the explicit provider
installation methods configuration that is intended for more complicated
situations.
To allow for that, here we make a special exception for the base
directory, resolving that first before we do any directory walking.
In order to help with debugging a situation where there are for some
reason symlinks at intermediate levels inside the search tree, we also now
emit a WARN log line in that case to be explicit that symlinks are not
supported there and to hint to put the symlink at the top-level if you
want to use symlinks at all.
(The support for symlinks at the deepest level of search is not mentioned
in this message because we allow it primarily for our own cache linking
behavior.)
When installing a provider which is already cached, we attempt to create
a symlink from the install directory targeting the cache. If symlinking
fails due to missing OS/filesystem support, we instead want to copy the
cached provider.
The fallback code to do this would always fail, due to a missing target
directory. This commit fixes that. I was unable to find a way to add
automated tests around this, but I have manually verified the fix on
Windows 8.1.
At the end of the EnsureProviderVersions process, we generate a lockfile
of the selected and installed provider versions. This includes a hash of
the unpacked provider directory.
When calculating this hash and generating the lockfile, we now also
verify that the provider directory contains a valid executable file. If
not, we return an error for this provider and trigger the installer's
HashPackageFailure event. Note that this event is not yet processed by
terraform init; that comes in the next commit.
Instead of searching the installed provider package directory for a
binary as we install it, we can lazily detect the executable as it is
required. Doing so allows us to separately report an invalid unpacked
package, giving the user more actionable error messages.
* internal/getproviders: decode and return any registry warnings
The public registry may include a list of warnings in the "versions"
response for any given provider. This PR adds support for warnings from
the registry and an installer event to return those warnings to the
user.
* internal/initwd: fix panics with relative submodules in DirFromModule
There were two related issues here:
1. panic with any local module with submodules
1. panic with a relative directory that was above the workdir ("../")
The first panic was caused by the local installer looking up the root
module with the (nonexistant) key "root.", instead of "".
The second panic was caused by the installer trying to determine the
relative path from ".". This was fixed by detecting "." as the source
path and using the absolute path for the call to filepath.Rel.
Added test cases for both panics and updated the existing e2e tests with
the correct install paths.
This is the equivalent of UnpackedDirectoryPathForPackage when working
with the packed directory layout. It returns a path to a .zip file with
a name that would be detected by SearchLocalDirectory as a
PackageLocalArchive package.
We previously had this functionality available for cached packages in the
providercache package. This moves the main implementation of this over
to the getproviders package and then implements it also for PackageMeta,
allowing us to compute hashes in a consistent way across both of our
representations of a provider package.
The new methods on PackageMeta will only be effective for packages in the
local filesystem because we need direct access to the contents in order
to produce the hash. Hopefully in future the registry protocol will be
able to also provide hashes using this content-based (rather than
archive-based) algorithm and then we'll be able to make this work for
PackageMeta referring to a package obtained from a registry too, but
hashes for local packages only are still useful for some cases right now,
such as generating mirror directories in the "terraform providers mirror"
command.
This adds supports for "unmanaged" providers, or providers with process
lifecycles not controlled by Terraform. These providers are assumed to
be started before Terraform is launched, and are assumed to shut
themselves down after Terraform has finished running.
To do this, we must update the go-plugin dependency to v1.3.0, which
added support for the "test mode" plugin serving that powers all this.
As a side-effect of not needing to manage the process lifecycle anymore,
Terraform also no longer needs to worry about the provider's binary, as
it won't be used for anything anymore. Because of this, we can disable
the init behavior that concerns itself with downloading that provider's
binary, checking its version, and otherwise managing the binary.
This is all managed on a per-provider basis, so managed providers that
Terraform downloads, starts, and stops can be used in the same commands
as unmanaged providers. The TF_REATTACH_PROVIDERS environment variable
is added, and is a JSON encoding of the provider's address to the
information we need to connect to it.
This change enables two benefits: first, delve and other debuggers can
now be attached to provider server processes, and Terraform can connect.
This allows for attaching debuggers to provider processes, which before
was difficult to impossible. Second, it allows the SDK test framework to
host the provider in the same process as the test driver, while running
a production Terraform binary against the provider. This allows for Go's
built-in race detector and test coverage tooling to work as expected in
provider tests.
Unmanaged providers are expected to work in the exact same way as
managed providers, with one caveat: Terraform kills provider processes
and restarts them once per graph walk, meaning multiple times during
most Terraform CLI commands. As unmanaged providers can't be killed by
Terraform, and have no visibility into graph walks, unmanaged providers
are likely to have differences in how their global mutable state behaves
when compared to managed providers. Namely, unmanaged providers are
likely to retain global state when managed providers would have reset
it. Developers relying on global state should be aware of this.
Relying on the early config for provider requirements was necessary in
Terraform 0.12, to allow the 0.12upgrade command to run after init
installs providers.
However in 0.13, the same restrictions do not apply, and the detection
of provider requirements has changed. As a result, the early config
loader gives incorrect provider requirements in some circumstances,
such as those in the new test in this commit.
Therefore we are changing the init command to use the requirements found
by the full configuration loader. This also means that we can remove the
internal initwd CheckCoreVersionRequirements function.
provider is not found.
Previously a user would see the following error even if terraform was
only searching the local filesystem:
"provider registry registry.terraform.io does not have a provider named
...."
This PR adds a registry-specific error type and modifies the MultiSource
installer to check for registry errors. It will return the
registry-specific error message if there is one, but if not the error
message will list all locations searched.
* providercache: add logging for errors from getproviders.SearchLocalDirectory
providercache.fillMetaCache() was silently swallowing errors when
searching the cache directory. This commit logs the error without
changing the behavior otherwise.
* command/cliconfig: validate plugin cache dir exists
The plugin cache directory must exist for terraform to use it, so we
will add a check at the begining.
* internal/getproviders: fix panic with invalid path parts
If the search path is missing a directory, the provider installer would
try to create an addrs.Provider with the wrong parts. For example if the
hostname was missing (as in the test case), it would call
addrs.NewProvider with (namespace, typename, version). This adds a
validation step for each part before calling addrs.NewProvider to avoid
the panic.
This is a port of the retry/timeout logic added in #24260 and #24259,
using the same environment variables to configure the retry and timeout
settings.
* internal/registry source: return error if requested provider version protocols are not supported
* getproviders: move responsibility for protocol compatibility checks into the registry client
The original implementation had the providercache checking the provider
metadata for protocol compatibility, but this is only relevant for the
registry source so it made more sense to move the logic into
getproviders.
This also addresses an issue where we were pulling the metadata for
every provider version until we found one that was supported. I've
extended the registry client to unmarshal the protocols in
`ProviderVersions` so we can filter through that list, instead of
pulling each version's metadata.
When looking up the namespace for a legacy provider source, we need to
use the /v1/providers/-/{name}/versions endpoint. For non-HashiCorp
providers, the /v1/providers/-/{name} endpoint returns a 404.
This commit updates the LegacyProviderDefaultNamespace method and the
mock registry servers accordingly.
This commit implements most of the intended functionality of the upgrade
command for rewriting configurations.
For a given module, it makes a list of all providers in use. Then it
attempts to detect the source address for providers without an explicit
source.
Once this step is complete, the tool rewrites the relevant configuration
files. This results in a single "required_providers" block for the
module, with a source for each provider.
Any providers for which the source cannot be detected (for example,
unofficial providers) will need a source to be defined by the user. The
tool writes an explanatory comment to the configuration to help with
this.
Martin Atkins