2014-09-27 01:03:39 +02:00
package command
import (
"fmt"
2017-06-09 18:42:45 +02:00
"log"
2014-09-27 01:03:39 +02:00
"os"
2017-06-02 02:57:43 +02:00
"sort"
2014-09-27 01:03:39 +02:00
"strings"
2019-09-10 00:58:44 +02:00
"github.com/hashicorp/hcl/v2"
command: "terraform init" can partially initialize for 0.12upgrade
There are a few constructs from 0.11 and prior that cause 0.12 parsing to
fail altogether, which previously created a chicken/egg problem because
we need to install the providers in order to run "terraform 0.12upgrade"
and thus fix the problem.
This changes "terraform init" to use the new "early configuration" loader
for module and provider installation. This is built on the more permissive
parser in the terraform-config-inspect package, and so it allows us to
read out the top-level blocks from the configuration while accepting
legacy HCL syntax.
In the long run this will let us do version compatibility detection before
attempting a "real" config load, giving us better error messages for any
future syntax additions, but in the short term the key thing is that it
allows us to install the dependencies even if the configuration isn't
fully valid.
Because backend init still requires full configuration, this introduces a
new mode of terraform init where it detects heuristically if it seems like
we need to do a configuration upgrade and does a partial init if so,
before finally directing the user to run "terraform 0.12upgrade" before
running any other commands.
The heuristic here is based on two assumptions:
- If the "early" loader finds no errors but the normal loader does, the
configuration is likely to be valid for Terraform 0.11 but not 0.12.
- If there's already a version constraint in the configuration that
excludes Terraform versions prior to v0.12 then the configuration is
probably _already_ upgraded and so it's just a normal syntax error,
even if the early loader didn't detect it.
Once the upgrade process is removed in 0.13.0 (users will be required to
go stepwise 0.11 -> 0.12 -> 0.13 to upgrade after that), some of this can
be simplified to remove that special mode, but the idea of doing the
dependency version checks against the liberal parser will remain valuable
to increase our chances of reporting version-based incompatibilities
rather than syntax errors as we add new features in future.
2019-01-14 20:11:00 +01:00
"github.com/hashicorp/terraform-config-inspect/tfconfig"
terraform: Ugly huge change to weave in new State and Plan types
Due to how often the state and plan types are referenced throughout
Terraform, there isn't a great way to switch them out gradually. As a
consequence, this huge commit gets us from the old world to a _compilable_
new world, but still has a large number of known test failures due to
key functionality being stubbed out.
The stubs here are for anything that interacts with providers, since we
now need to do the follow-up work to similarly replace the old
terraform.ResourceProvider interface with its replacement in the new
"providers" package. That work, along with work to fix the remaining
failing tests, will follow in subsequent commits.
The aim here was to replace all references to terraform.State and its
downstream types with states.State, terraform.Plan with plans.Plan,
state.State with statemgr.State, and switch to the new implementations of
the state and plan file formats. However, due to the number of times those
types are used, this also ended up affecting numerous other parts of core
such as terraform.Hook, the backend.Backend interface, and most of the CLI
commands.
Just as with 5861dbf3fc49b19587a31816eb06f511ab861bb4 before, I apologize
in advance to the person who inevitably just found this huge commit while
spelunking through the commit history.
2018-08-14 23:24:45 +02:00
"github.com/posener/complete"
"github.com/zclconf/go-cty/cty"
2019-01-14 20:07:54 +01:00
"github.com/hashicorp/errwrap"
2019-09-09 22:59:50 +02:00
"github.com/hashicorp/terraform/addrs"
2017-05-03 17:02:47 +02:00
"github.com/hashicorp/terraform/backend"
command: "terraform init" can partially initialize for 0.12upgrade
There are a few constructs from 0.11 and prior that cause 0.12 parsing to
fail altogether, which previously created a chicken/egg problem because
we need to install the providers in order to run "terraform 0.12upgrade"
and thus fix the problem.
This changes "terraform init" to use the new "early configuration" loader
for module and provider installation. This is built on the more permissive
parser in the terraform-config-inspect package, and so it allows us to
read out the top-level blocks from the configuration while accepting
legacy HCL syntax.
In the long run this will let us do version compatibility detection before
attempting a "real" config load, giving us better error messages for any
future syntax additions, but in the short term the key thing is that it
allows us to install the dependencies even if the configuration isn't
fully valid.
Because backend init still requires full configuration, this introduces a
new mode of terraform init where it detects heuristically if it seems like
we need to do a configuration upgrade and does a partial init if so,
before finally directing the user to run "terraform 0.12upgrade" before
running any other commands.
The heuristic here is based on two assumptions:
- If the "early" loader finds no errors but the normal loader does, the
configuration is likely to be valid for Terraform 0.11 but not 0.12.
- If there's already a version constraint in the configuration that
excludes Terraform versions prior to v0.12 then the configuration is
probably _already_ upgraded and so it's just a normal syntax error,
even if the early loader didn't detect it.
Once the upgrade process is removed in 0.13.0 (users will be required to
go stepwise 0.11 -> 0.12 -> 0.13 to upgrade after that), some of this can
be simplified to remove that special mode, but the idea of doing the
dependency version checks against the liberal parser will remain valuable
to increase our chances of reporting version-based incompatibilities
rather than syntax errors as we add new features in future.
2019-01-14 20:11:00 +01:00
backendInit "github.com/hashicorp/terraform/backend/init"
2018-03-28 00:31:05 +02:00
"github.com/hashicorp/terraform/configs"
terraform: Ugly huge change to weave in new State and Plan types
Due to how often the state and plan types are referenced throughout
Terraform, there isn't a great way to switch them out gradually. As a
consequence, this huge commit gets us from the old world to a _compilable_
new world, but still has a large number of known test failures due to
key functionality being stubbed out.
The stubs here are for anything that interacts with providers, since we
now need to do the follow-up work to similarly replace the old
terraform.ResourceProvider interface with its replacement in the new
"providers" package. That work, along with work to fix the remaining
failing tests, will follow in subsequent commits.
The aim here was to replace all references to terraform.State and its
downstream types with states.State, terraform.Plan with plans.Plan,
state.State with statemgr.State, and switch to the new implementations of
the state and plan file formats. However, due to the number of times those
types are used, this also ended up affecting numerous other parts of core
such as terraform.Hook, the backend.Backend interface, and most of the CLI
commands.
Just as with 5861dbf3fc49b19587a31816eb06f511ab861bb4 before, I apologize
in advance to the person who inevitably just found this huge commit while
spelunking through the commit history.
2018-08-14 23:24:45 +02:00
"github.com/hashicorp/terraform/configs/configschema"
command: "terraform init" can partially initialize for 0.12upgrade
There are a few constructs from 0.11 and prior that cause 0.12 parsing to
fail altogether, which previously created a chicken/egg problem because
we need to install the providers in order to run "terraform 0.12upgrade"
and thus fix the problem.
This changes "terraform init" to use the new "early configuration" loader
for module and provider installation. This is built on the more permissive
parser in the terraform-config-inspect package, and so it allows us to
read out the top-level blocks from the configuration while accepting
legacy HCL syntax.
In the long run this will let us do version compatibility detection before
attempting a "real" config load, giving us better error messages for any
future syntax additions, but in the short term the key thing is that it
allows us to install the dependencies even if the configuration isn't
fully valid.
Because backend init still requires full configuration, this introduces a
new mode of terraform init where it detects heuristically if it seems like
we need to do a configuration upgrade and does a partial init if so,
before finally directing the user to run "terraform 0.12upgrade" before
running any other commands.
The heuristic here is based on two assumptions:
- If the "early" loader finds no errors but the normal loader does, the
configuration is likely to be valid for Terraform 0.11 but not 0.12.
- If there's already a version constraint in the configuration that
excludes Terraform versions prior to v0.12 then the configuration is
probably _already_ upgraded and so it's just a normal syntax error,
even if the early loader didn't detect it.
Once the upgrade process is removed in 0.13.0 (users will be required to
go stepwise 0.11 -> 0.12 -> 0.13 to upgrade after that), some of this can
be simplified to remove that special mode, but the idea of doing the
dependency version checks against the liberal parser will remain valuable
to increase our chances of reporting version-based incompatibilities
rather than syntax errors as we add new features in future.
2019-01-14 20:11:00 +01:00
"github.com/hashicorp/terraform/configs/configupgrade"
"github.com/hashicorp/terraform/internal/earlyconfig"
"github.com/hashicorp/terraform/internal/initwd"
2017-05-03 17:02:47 +02:00
"github.com/hashicorp/terraform/plugin/discovery"
terraform: Ugly huge change to weave in new State and Plan types
Due to how often the state and plan types are referenced throughout
Terraform, there isn't a great way to switch them out gradually. As a
consequence, this huge commit gets us from the old world to a _compilable_
new world, but still has a large number of known test failures due to
key functionality being stubbed out.
The stubs here are for anything that interacts with providers, since we
now need to do the follow-up work to similarly replace the old
terraform.ResourceProvider interface with its replacement in the new
"providers" package. That work, along with work to fix the remaining
failing tests, will follow in subsequent commits.
The aim here was to replace all references to terraform.State and its
downstream types with states.State, terraform.Plan with plans.Plan,
state.State with statemgr.State, and switch to the new implementations of
the state and plan file formats. However, due to the number of times those
types are used, this also ended up affecting numerous other parts of core
such as terraform.Hook, the backend.Backend interface, and most of the CLI
commands.
Just as with 5861dbf3fc49b19587a31816eb06f511ab861bb4 before, I apologize
in advance to the person who inevitably just found this huge commit while
spelunking through the commit history.
2018-08-14 23:24:45 +02:00
"github.com/hashicorp/terraform/states"
2017-05-03 17:02:47 +02:00
"github.com/hashicorp/terraform/terraform"
2018-03-28 00:31:05 +02:00
"github.com/hashicorp/terraform/tfdiags"
2014-09-27 01:03:39 +02:00
)
// InitCommand is a Command implementation that takes a Terraform
// module and clones it to the working directory.
type InitCommand struct {
Meta
2017-05-04 19:01:05 +02:00
2017-06-15 21:23:16 +02:00
// getPlugins is for the -get-plugins flag
getPlugins bool
2017-06-13 03:22:47 +02:00
// providerInstaller is used to download and install providers that
// aren't found locally. This uses a discovery.ProviderInstaller instance
// by default, but it can be overridden here as a way to mock fetching
// providers for tests.
providerInstaller discovery . Installer
2014-09-27 01:03:39 +02:00
}
func ( c * InitCommand ) Run ( args [ ] string ) int {
2017-07-29 00:23:29 +02:00
var flagFromModule string
2017-06-15 21:23:16 +02:00
var flagBackend , flagGet , flagUpgrade bool
2017-06-15 20:26:12 +02:00
var flagPluginPath FlagStringSlice
2017-06-20 17:46:33 +02:00
var flagVerifyPlugins bool
2018-03-28 00:31:05 +02:00
flagConfigExtra := newRawFlags ( "-backend-config" )
2017-03-21 20:05:51 +01:00
2017-03-08 05:09:48 +01:00
args , err := c . Meta . process ( args , false )
if err != nil {
return 1
}
2018-11-21 15:35:27 +01:00
cmdFlags := c . Meta . extendedFlagSet ( "init" )
2017-01-19 05:50:45 +01:00
cmdFlags . BoolVar ( & flagBackend , "backend" , true , "" )
2018-03-28 00:31:05 +02:00
cmdFlags . Var ( flagConfigExtra , "backend-config" , "" )
2017-07-29 00:23:29 +02:00
cmdFlags . StringVar ( & flagFromModule , "from-module" , "" , "copy the source of the given module into the directory before init" )
2017-01-19 05:50:45 +01:00
cmdFlags . BoolVar ( & flagGet , "get" , true , "" )
2017-06-15 21:23:16 +02:00
cmdFlags . BoolVar ( & c . getPlugins , "get-plugins" , true , "" )
2017-03-21 20:05:51 +01:00
cmdFlags . BoolVar ( & c . forceInitCopy , "force-copy" , false , "suppress prompts about copying state data" )
2017-04-01 22:19:59 +02:00
cmdFlags . BoolVar ( & c . Meta . stateLock , "lock" , true , "lock state" )
cmdFlags . DurationVar ( & c . Meta . stateLockTimeout , "lock-timeout" , 0 , "lock timeout" )
2017-04-20 23:26:50 +02:00
cmdFlags . BoolVar ( & c . reconfigure , "reconfigure" , false , "reconfigure" )
2017-06-12 23:14:40 +02:00
cmdFlags . BoolVar ( & flagUpgrade , "upgrade" , false , "" )
2017-06-15 20:26:12 +02:00
cmdFlags . Var ( & flagPluginPath , "plugin-dir" , "plugin directory" )
2017-06-20 17:46:33 +02:00
cmdFlags . BoolVar ( & flagVerifyPlugins , "verify-plugins" , true , "verify plugins" )
2014-09-27 01:03:39 +02:00
cmdFlags . Usage = func ( ) { c . Ui . Error ( c . Help ( ) ) }
if err := cmdFlags . Parse ( args ) ; err != nil {
return 1
}
2018-03-28 00:31:05 +02:00
var diags tfdiags . Diagnostics
2017-06-15 21:23:16 +02:00
if len ( flagPluginPath ) > 0 {
c . pluginPath = flagPluginPath
c . getPlugins = false
}
2017-06-15 20:26:12 +02:00
2017-09-02 01:05:05 +02:00
// set providerInstaller if we don't have a test version already
2017-06-13 03:22:47 +02:00
if c . providerInstaller == nil {
c . providerInstaller = & discovery . ProviderInstaller {
2018-09-09 19:18:08 +02:00
Dir : c . pluginDir ( ) ,
Cache : c . pluginCache ( ) ,
2019-01-16 20:00:08 +01:00
PluginProtocolVersion : discovery . PluginInstallProtocolVersion ,
2017-06-20 17:46:33 +02:00
SkipVerify : ! flagVerifyPlugins ,
2017-08-14 13:32:48 +02:00
Ui : c . Ui ,
2018-08-02 17:45:55 +02:00
Services : c . Services ,
2017-06-13 03:22:47 +02:00
}
2017-05-04 19:01:05 +02:00
}
2017-01-19 05:50:45 +01:00
// Validate the arg count
2014-09-27 01:03:39 +02:00
args = cmdFlags . Args ( )
2017-06-02 22:13:11 +02:00
if len ( args ) > 1 {
c . Ui . Error ( "The init command expects at most one argument.\n" )
2014-09-27 01:03:39 +02:00
cmdFlags . Usage ( )
return 1
2017-01-19 05:50:45 +01:00
}
2017-06-15 20:26:12 +02:00
if err := c . storePluginPath ( c . pluginPath ) ; err != nil {
c . Ui . Error ( fmt . Sprintf ( "Error saving -plugin-path values: %s" , err ) )
return 1
}
2017-01-19 05:50:45 +01:00
// Get our pwd. We don't always need it but always getting it is easier
// than the logic to determine if it is or isn't needed.
pwd , err := os . Getwd ( )
if err != nil {
c . Ui . Error ( fmt . Sprintf ( "Error getting pwd: %s" , err ) )
2014-09-27 01:03:39 +02:00
return 1
}
2017-07-29 00:23:29 +02:00
// If an argument is provided then it overrides our working directory.
2017-06-02 22:13:11 +02:00
path := pwd
if len ( args ) == 1 {
path = args [ 0 ]
2014-09-27 01:03:39 +02:00
}
2016-01-20 02:13:19 +01:00
2017-01-19 05:50:45 +01:00
// This will track whether we outputted anything so that we know whether
// to output a newline before the success message
var header bool
2017-07-29 00:23:29 +02:00
if flagFromModule != "" {
src := flagFromModule
2019-07-18 19:07:10 +02:00
empty , err := configs . IsEmptyDir ( path )
2017-07-29 00:23:29 +02:00
if err != nil {
c . Ui . Error ( fmt . Sprintf ( "Error validating destination directory: %s" , err ) )
return 1
}
if ! empty {
c . Ui . Error ( strings . TrimSpace ( errInitCopyNotEmpty ) )
return 1
}
c . Ui . Output ( c . Colorize ( ) . Color ( fmt . Sprintf (
"[reset][bold]Copying configuration[reset] from %q..." , src ,
) ) )
header = true
2018-03-28 00:31:05 +02:00
hooks := uiModuleInstallHooks {
Ui : c . Ui ,
ShowLocalPaths : false , // since they are in a weird location for init
}
initDiags := c . initDirFromModule ( path , src , hooks )
diags = diags . Append ( initDiags )
if initDiags . HasErrors ( ) {
c . showDiagnostics ( diags )
2017-07-29 00:23:29 +02:00
return 1
}
2018-03-28 02:22:51 +02:00
c . Ui . Output ( "" )
2017-07-29 00:23:29 +02:00
}
2017-01-19 05:50:45 +01:00
// If our directory is empty, then we're done. We can't get or setup
// the backend with an empty directory.
2019-07-18 19:07:10 +02:00
empty , err := configs . IsEmptyDir ( path )
2018-10-31 16:45:03 +01:00
if err != nil {
2018-03-28 00:31:05 +02:00
diags = diags . Append ( fmt . Errorf ( "Error checking configuration: %s" , err ) )
2014-09-27 01:03:39 +02:00
return 1
2018-10-31 16:45:03 +01:00
}
if empty {
2017-01-19 05:50:45 +01:00
c . Ui . Output ( c . Colorize ( ) . Color ( strings . TrimSpace ( outputInitEmpty ) ) )
return 0
2014-09-27 01:03:39 +02:00
}
command: "terraform init" can partially initialize for 0.12upgrade
There are a few constructs from 0.11 and prior that cause 0.12 parsing to
fail altogether, which previously created a chicken/egg problem because
we need to install the providers in order to run "terraform 0.12upgrade"
and thus fix the problem.
This changes "terraform init" to use the new "early configuration" loader
for module and provider installation. This is built on the more permissive
parser in the terraform-config-inspect package, and so it allows us to
read out the top-level blocks from the configuration while accepting
legacy HCL syntax.
In the long run this will let us do version compatibility detection before
attempting a "real" config load, giving us better error messages for any
future syntax additions, but in the short term the key thing is that it
allows us to install the dependencies even if the configuration isn't
fully valid.
Because backend init still requires full configuration, this introduces a
new mode of terraform init where it detects heuristically if it seems like
we need to do a configuration upgrade and does a partial init if so,
before finally directing the user to run "terraform 0.12upgrade" before
running any other commands.
The heuristic here is based on two assumptions:
- If the "early" loader finds no errors but the normal loader does, the
configuration is likely to be valid for Terraform 0.11 but not 0.12.
- If there's already a version constraint in the configuration that
excludes Terraform versions prior to v0.12 then the configuration is
probably _already_ upgraded and so it's just a normal syntax error,
even if the early loader didn't detect it.
Once the upgrade process is removed in 0.13.0 (users will be required to
go stepwise 0.11 -> 0.12 -> 0.13 to upgrade after that), some of this can
be simplified to remove that special mode, but the idea of doing the
dependency version checks against the liberal parser will remain valuable
to increase our chances of reporting version-based incompatibilities
rather than syntax errors as we add new features in future.
2019-01-14 20:11:00 +01:00
// Before we do anything else, we'll try loading configuration with both
// our "normal" and "early" configuration codepaths. If early succeeds
// while normal fails, that strongly suggests that the configuration is
// using syntax that worked in 0.11 but no longer in 0.12, which requires
// some special behavior here to get the directory initialized just enough
// to run "terraform 0.12upgrade".
//
// FIXME: Once we reach 0.13 and remove 0.12upgrade, we should rework this
// so that we first use the early config to do a general compatibility
// check with dependencies, producing version-oriented error messages if
// dependencies aren't right, and only then use the real loader to deal
// with the backend configuration.
rootMod , confDiags := c . loadSingleModule ( path )
rootModEarly , earlyConfDiags := c . loadSingleModuleEarly ( path )
configUpgradeProbablyNeeded := false
if confDiags . HasErrors ( ) {
if earlyConfDiags . HasErrors ( ) {
// If both parsers produced errors then we'll assume the config
// is _truly_ invalid and produce error messages as normal.
2017-10-05 21:00:45 +02:00
// Since this may be the user's first ever interaction with Terraform,
// we'll provide some additional context in this case.
c . Ui . Error ( strings . TrimSpace ( errInitConfigError ) )
command: "terraform init" can partially initialize for 0.12upgrade
There are a few constructs from 0.11 and prior that cause 0.12 parsing to
fail altogether, which previously created a chicken/egg problem because
we need to install the providers in order to run "terraform 0.12upgrade"
and thus fix the problem.
This changes "terraform init" to use the new "early configuration" loader
for module and provider installation. This is built on the more permissive
parser in the terraform-config-inspect package, and so it allows us to
read out the top-level blocks from the configuration while accepting
legacy HCL syntax.
In the long run this will let us do version compatibility detection before
attempting a "real" config load, giving us better error messages for any
future syntax additions, but in the short term the key thing is that it
allows us to install the dependencies even if the configuration isn't
fully valid.
Because backend init still requires full configuration, this introduces a
new mode of terraform init where it detects heuristically if it seems like
we need to do a configuration upgrade and does a partial init if so,
before finally directing the user to run "terraform 0.12upgrade" before
running any other commands.
The heuristic here is based on two assumptions:
- If the "early" loader finds no errors but the normal loader does, the
configuration is likely to be valid for Terraform 0.11 but not 0.12.
- If there's already a version constraint in the configuration that
excludes Terraform versions prior to v0.12 then the configuration is
probably _already_ upgraded and so it's just a normal syntax error,
even if the early loader didn't detect it.
Once the upgrade process is removed in 0.13.0 (users will be required to
go stepwise 0.11 -> 0.12 -> 0.13 to upgrade after that), some of this can
be simplified to remove that special mode, but the idea of doing the
dependency version checks against the liberal parser will remain valuable
to increase our chances of reporting version-based incompatibilities
rather than syntax errors as we add new features in future.
2019-01-14 20:11:00 +01:00
diags = diags . Append ( confDiags )
2018-03-28 00:31:05 +02:00
c . showDiagnostics ( diags )
2014-10-10 02:16:17 +02:00
return 1
}
2017-01-19 05:50:45 +01:00
command: "terraform init" can partially initialize for 0.12upgrade
There are a few constructs from 0.11 and prior that cause 0.12 parsing to
fail altogether, which previously created a chicken/egg problem because
we need to install the providers in order to run "terraform 0.12upgrade"
and thus fix the problem.
This changes "terraform init" to use the new "early configuration" loader
for module and provider installation. This is built on the more permissive
parser in the terraform-config-inspect package, and so it allows us to
read out the top-level blocks from the configuration while accepting
legacy HCL syntax.
In the long run this will let us do version compatibility detection before
attempting a "real" config load, giving us better error messages for any
future syntax additions, but in the short term the key thing is that it
allows us to install the dependencies even if the configuration isn't
fully valid.
Because backend init still requires full configuration, this introduces a
new mode of terraform init where it detects heuristically if it seems like
we need to do a configuration upgrade and does a partial init if so,
before finally directing the user to run "terraform 0.12upgrade" before
running any other commands.
The heuristic here is based on two assumptions:
- If the "early" loader finds no errors but the normal loader does, the
configuration is likely to be valid for Terraform 0.11 but not 0.12.
- If there's already a version constraint in the configuration that
excludes Terraform versions prior to v0.12 then the configuration is
probably _already_ upgraded and so it's just a normal syntax error,
even if the early loader didn't detect it.
Once the upgrade process is removed in 0.13.0 (users will be required to
go stepwise 0.11 -> 0.12 -> 0.13 to upgrade after that), some of this can
be simplified to remove that special mode, but the idea of doing the
dependency version checks against the liberal parser will remain valuable
to increase our chances of reporting version-based incompatibilities
rather than syntax errors as we add new features in future.
2019-01-14 20:11:00 +01:00
// If _only_ the main loader produced errors then that suggests an
// upgrade may help. To give us more certainty here, we'll use the
// same heuristic that "terraform 0.12upgrade" uses to guess if a
// configuration has already been upgraded, to reduce the risk that
// we'll produce a misleading message if the problem is just a regular
// syntax error that the early loader just didn't catch.
sources , err := configupgrade . LoadModule ( path )
if err == nil {
if already , _ := sources . MaybeAlreadyUpgraded ( ) ; already {
// Just report the errors as normal, then.
c . Ui . Error ( strings . TrimSpace ( errInitConfigError ) )
diags = diags . Append ( confDiags )
2018-03-28 00:31:05 +02:00
c . showDiagnostics ( diags )
2015-02-22 19:49:31 +01:00
return 1
}
2017-01-19 05:50:45 +01:00
}
command: "terraform init" can partially initialize for 0.12upgrade
There are a few constructs from 0.11 and prior that cause 0.12 parsing to
fail altogether, which previously created a chicken/egg problem because
we need to install the providers in order to run "terraform 0.12upgrade"
and thus fix the problem.
This changes "terraform init" to use the new "early configuration" loader
for module and provider installation. This is built on the more permissive
parser in the terraform-config-inspect package, and so it allows us to
read out the top-level blocks from the configuration while accepting
legacy HCL syntax.
In the long run this will let us do version compatibility detection before
attempting a "real" config load, giving us better error messages for any
future syntax additions, but in the short term the key thing is that it
allows us to install the dependencies even if the configuration isn't
fully valid.
Because backend init still requires full configuration, this introduces a
new mode of terraform init where it detects heuristically if it seems like
we need to do a configuration upgrade and does a partial init if so,
before finally directing the user to run "terraform 0.12upgrade" before
running any other commands.
The heuristic here is based on two assumptions:
- If the "early" loader finds no errors but the normal loader does, the
configuration is likely to be valid for Terraform 0.11 but not 0.12.
- If there's already a version constraint in the configuration that
excludes Terraform versions prior to v0.12 then the configuration is
probably _already_ upgraded and so it's just a normal syntax error,
even if the early loader didn't detect it.
Once the upgrade process is removed in 0.13.0 (users will be required to
go stepwise 0.11 -> 0.12 -> 0.13 to upgrade after that), some of this can
be simplified to remove that special mode, but the idea of doing the
dependency version checks against the liberal parser will remain valuable
to increase our chances of reporting version-based incompatibilities
rather than syntax errors as we add new features in future.
2019-01-14 20:11:00 +01:00
configUpgradeProbablyNeeded = true
}
if earlyConfDiags . HasErrors ( ) {
// If _only_ the early loader encountered errors then that's unusual
// (it should generally be a superset of the normal loader) but we'll
// return those errors anyway since otherwise we'll probably get
// some weird behavior downstream. Errors from the early loader are
// generally not as high-quality since it has less context to work with.
c . Ui . Error ( strings . TrimSpace ( errInitConfigError ) )
diags = diags . Append ( earlyConfDiags )
c . showDiagnostics ( diags )
return 1
}
2017-01-19 05:50:45 +01:00
command: "terraform init" can partially initialize for 0.12upgrade
There are a few constructs from 0.11 and prior that cause 0.12 parsing to
fail altogether, which previously created a chicken/egg problem because
we need to install the providers in order to run "terraform 0.12upgrade"
and thus fix the problem.
This changes "terraform init" to use the new "early configuration" loader
for module and provider installation. This is built on the more permissive
parser in the terraform-config-inspect package, and so it allows us to
read out the top-level blocks from the configuration while accepting
legacy HCL syntax.
In the long run this will let us do version compatibility detection before
attempting a "real" config load, giving us better error messages for any
future syntax additions, but in the short term the key thing is that it
allows us to install the dependencies even if the configuration isn't
fully valid.
Because backend init still requires full configuration, this introduces a
new mode of terraform init where it detects heuristically if it seems like
we need to do a configuration upgrade and does a partial init if so,
before finally directing the user to run "terraform 0.12upgrade" before
running any other commands.
The heuristic here is based on two assumptions:
- If the "early" loader finds no errors but the normal loader does, the
configuration is likely to be valid for Terraform 0.11 but not 0.12.
- If there's already a version constraint in the configuration that
excludes Terraform versions prior to v0.12 then the configuration is
probably _already_ upgraded and so it's just a normal syntax error,
even if the early loader didn't detect it.
Once the upgrade process is removed in 0.13.0 (users will be required to
go stepwise 0.11 -> 0.12 -> 0.13 to upgrade after that), some of this can
be simplified to remove that special mode, but the idea of doing the
dependency version checks against the liberal parser will remain valuable
to increase our chances of reporting version-based incompatibilities
rather than syntax errors as we add new features in future.
2019-01-14 20:11:00 +01:00
if flagGet {
modsOutput , modsDiags := c . getModules ( path , rootModEarly , flagUpgrade )
diags = diags . Append ( modsDiags )
if modsDiags . HasErrors ( ) {
c . showDiagnostics ( diags )
return 1
}
if modsOutput {
2017-01-19 05:50:45 +01:00
header = true
2014-10-01 01:05:40 +02:00
}
2017-01-19 05:50:45 +01:00
}
2014-10-10 02:16:17 +02:00
command: "terraform init" can partially initialize for 0.12upgrade
There are a few constructs from 0.11 and prior that cause 0.12 parsing to
fail altogether, which previously created a chicken/egg problem because
we need to install the providers in order to run "terraform 0.12upgrade"
and thus fix the problem.
This changes "terraform init" to use the new "early configuration" loader
for module and provider installation. This is built on the more permissive
parser in the terraform-config-inspect package, and so it allows us to
read out the top-level blocks from the configuration while accepting
legacy HCL syntax.
In the long run this will let us do version compatibility detection before
attempting a "real" config load, giving us better error messages for any
future syntax additions, but in the short term the key thing is that it
allows us to install the dependencies even if the configuration isn't
fully valid.
Because backend init still requires full configuration, this introduces a
new mode of terraform init where it detects heuristically if it seems like
we need to do a configuration upgrade and does a partial init if so,
before finally directing the user to run "terraform 0.12upgrade" before
running any other commands.
The heuristic here is based on two assumptions:
- If the "early" loader finds no errors but the normal loader does, the
configuration is likely to be valid for Terraform 0.11 but not 0.12.
- If there's already a version constraint in the configuration that
excludes Terraform versions prior to v0.12 then the configuration is
probably _already_ upgraded and so it's just a normal syntax error,
even if the early loader didn't detect it.
Once the upgrade process is removed in 0.13.0 (users will be required to
go stepwise 0.11 -> 0.12 -> 0.13 to upgrade after that), some of this can
be simplified to remove that special mode, but the idea of doing the
dependency version checks against the liberal parser will remain valuable
to increase our chances of reporting version-based incompatibilities
rather than syntax errors as we add new features in future.
2019-01-14 20:11:00 +01:00
// With all of the modules (hopefully) installed, we can now try to load
// the whole configuration tree.
//
// Just as above, we'll try loading both with the early and normal config
// loaders here. Subsequent work will only use the early config, but
// loading both gives us an opportunity to prefer the better error messages
// from the normal loader if both fail.
_ , confDiags = c . loadConfig ( path )
earlyConfig , earlyConfDiags := c . loadConfigEarly ( path )
if confDiags . HasErrors ( ) && ! configUpgradeProbablyNeeded {
2018-11-10 00:08:39 +01:00
c . Ui . Error ( strings . TrimSpace ( errInitConfigError ) )
command: "terraform init" can partially initialize for 0.12upgrade
There are a few constructs from 0.11 and prior that cause 0.12 parsing to
fail altogether, which previously created a chicken/egg problem because
we need to install the providers in order to run "terraform 0.12upgrade"
and thus fix the problem.
This changes "terraform init" to use the new "early configuration" loader
for module and provider installation. This is built on the more permissive
parser in the terraform-config-inspect package, and so it allows us to
read out the top-level blocks from the configuration while accepting
legacy HCL syntax.
In the long run this will let us do version compatibility detection before
attempting a "real" config load, giving us better error messages for any
future syntax additions, but in the short term the key thing is that it
allows us to install the dependencies even if the configuration isn't
fully valid.
Because backend init still requires full configuration, this introduces a
new mode of terraform init where it detects heuristically if it seems like
we need to do a configuration upgrade and does a partial init if so,
before finally directing the user to run "terraform 0.12upgrade" before
running any other commands.
The heuristic here is based on two assumptions:
- If the "early" loader finds no errors but the normal loader does, the
configuration is likely to be valid for Terraform 0.11 but not 0.12.
- If there's already a version constraint in the configuration that
excludes Terraform versions prior to v0.12 then the configuration is
probably _already_ upgraded and so it's just a normal syntax error,
even if the early loader didn't detect it.
Once the upgrade process is removed in 0.13.0 (users will be required to
go stepwise 0.11 -> 0.12 -> 0.13 to upgrade after that), some of this can
be simplified to remove that special mode, but the idea of doing the
dependency version checks against the liberal parser will remain valuable
to increase our chances of reporting version-based incompatibilities
rather than syntax errors as we add new features in future.
2019-01-14 20:11:00 +01:00
diags = diags . Append ( confDiags )
2018-11-10 00:08:39 +01:00
c . showDiagnostics ( diags )
return 1
}
command: "terraform init" can partially initialize for 0.12upgrade
There are a few constructs from 0.11 and prior that cause 0.12 parsing to
fail altogether, which previously created a chicken/egg problem because
we need to install the providers in order to run "terraform 0.12upgrade"
and thus fix the problem.
This changes "terraform init" to use the new "early configuration" loader
for module and provider installation. This is built on the more permissive
parser in the terraform-config-inspect package, and so it allows us to
read out the top-level blocks from the configuration while accepting
legacy HCL syntax.
In the long run this will let us do version compatibility detection before
attempting a "real" config load, giving us better error messages for any
future syntax additions, but in the short term the key thing is that it
allows us to install the dependencies even if the configuration isn't
fully valid.
Because backend init still requires full configuration, this introduces a
new mode of terraform init where it detects heuristically if it seems like
we need to do a configuration upgrade and does a partial init if so,
before finally directing the user to run "terraform 0.12upgrade" before
running any other commands.
The heuristic here is based on two assumptions:
- If the "early" loader finds no errors but the normal loader does, the
configuration is likely to be valid for Terraform 0.11 but not 0.12.
- If there's already a version constraint in the configuration that
excludes Terraform versions prior to v0.12 then the configuration is
probably _already_ upgraded and so it's just a normal syntax error,
even if the early loader didn't detect it.
Once the upgrade process is removed in 0.13.0 (users will be required to
go stepwise 0.11 -> 0.12 -> 0.13 to upgrade after that), some of this can
be simplified to remove that special mode, but the idea of doing the
dependency version checks against the liberal parser will remain valuable
to increase our chances of reporting version-based incompatibilities
rather than syntax errors as we add new features in future.
2019-01-14 20:11:00 +01:00
if earlyConfDiags . HasErrors ( ) {
c . Ui . Error ( strings . TrimSpace ( errInitConfigError ) )
diags = diags . Append ( earlyConfDiags )
2018-11-10 00:08:39 +01:00
c . showDiagnostics ( diags )
return 1
}
command: "terraform init" can partially initialize for 0.12upgrade
There are a few constructs from 0.11 and prior that cause 0.12 parsing to
fail altogether, which previously created a chicken/egg problem because
we need to install the providers in order to run "terraform 0.12upgrade"
and thus fix the problem.
This changes "terraform init" to use the new "early configuration" loader
for module and provider installation. This is built on the more permissive
parser in the terraform-config-inspect package, and so it allows us to
read out the top-level blocks from the configuration while accepting
legacy HCL syntax.
In the long run this will let us do version compatibility detection before
attempting a "real" config load, giving us better error messages for any
future syntax additions, but in the short term the key thing is that it
allows us to install the dependencies even if the configuration isn't
fully valid.
Because backend init still requires full configuration, this introduces a
new mode of terraform init where it detects heuristically if it seems like
we need to do a configuration upgrade and does a partial init if so,
before finally directing the user to run "terraform 0.12upgrade" before
running any other commands.
The heuristic here is based on two assumptions:
- If the "early" loader finds no errors but the normal loader does, the
configuration is likely to be valid for Terraform 0.11 but not 0.12.
- If there's already a version constraint in the configuration that
excludes Terraform versions prior to v0.12 then the configuration is
probably _already_ upgraded and so it's just a normal syntax error,
even if the early loader didn't detect it.
Once the upgrade process is removed in 0.13.0 (users will be required to
go stepwise 0.11 -> 0.12 -> 0.13 to upgrade after that), some of this can
be simplified to remove that special mode, but the idea of doing the
dependency version checks against the liberal parser will remain valuable
to increase our chances of reporting version-based incompatibilities
rather than syntax errors as we add new features in future.
2019-01-14 20:11:00 +01:00
{
// Before we go further, we'll check to make sure none of the modules
// in the configuration declare that they don't support this Terraform
// version, so we can produce a version-related error message rather
// than potentially-confusing downstream errors.
versionDiags := initwd . CheckCoreVersionRequirements ( earlyConfig )
diags = diags . Append ( versionDiags )
if versionDiags . HasErrors ( ) {
c . showDiagnostics ( diags )
return 1
}
}
var back backend . Backend
if flagBackend {
switch {
case configUpgradeProbablyNeeded :
diags = diags . Append ( tfdiags . Sourceless (
tfdiags . Warning ,
"Skipping backend initialization pending configuration upgrade" ,
// The "below" in this message is referring to the special
// note about running "terraform 0.12upgrade" that we'll
// print out at the end when configUpgradeProbablyNeeded is set.
"The root module configuration contains errors that may be fixed by running the configuration upgrade tool, so Terraform is skipping backend initialization. See below for more information." ,
) )
default :
be , backendOutput , backendDiags := c . initBackend ( rootMod , flagConfigExtra )
diags = diags . Append ( backendDiags )
if backendDiags . HasErrors ( ) {
c . showDiagnostics ( diags )
return 1
}
if backendOutput {
header = true
}
back = be
}
}
2017-06-21 19:32:13 +02:00
if back == nil {
// If we didn't initialize a backend then we'll try to at least
2019-05-29 19:58:04 +02:00
// instantiate one. This might fail if it wasn't already initialized
2017-06-21 19:32:13 +02:00
// by a previous run, so we must still expect that "back" may be nil
// in code that follows.
2018-03-28 00:31:05 +02:00
var backDiags tfdiags . Diagnostics
back , backDiags = c . Backend ( nil )
if backDiags . HasErrors ( ) {
2017-06-21 19:32:13 +02:00
// This is fine. We'll proceed with no backend, then.
back = nil
}
2017-06-15 21:23:16 +02:00
}
2017-05-03 17:02:47 +02:00
terraform: Ugly huge change to weave in new State and Plan types
Due to how often the state and plan types are referenced throughout
Terraform, there isn't a great way to switch them out gradually. As a
consequence, this huge commit gets us from the old world to a _compilable_
new world, but still has a large number of known test failures due to
key functionality being stubbed out.
The stubs here are for anything that interacts with providers, since we
now need to do the follow-up work to similarly replace the old
terraform.ResourceProvider interface with its replacement in the new
"providers" package. That work, along with work to fix the remaining
failing tests, will follow in subsequent commits.
The aim here was to replace all references to terraform.State and its
downstream types with states.State, terraform.Plan with plans.Plan,
state.State with statemgr.State, and switch to the new implementations of
the state and plan file formats. However, due to the number of times those
types are used, this also ended up affecting numerous other parts of core
such as terraform.Hook, the backend.Backend interface, and most of the CLI
commands.
Just as with 5861dbf3fc49b19587a31816eb06f511ab861bb4 before, I apologize
in advance to the person who inevitably just found this huge commit while
spelunking through the commit history.
2018-08-14 23:24:45 +02:00
var state * states . State
2017-06-21 19:32:13 +02:00
// If we have a functional backend (either just initialized or initialized
// on a previous run) we'll use the current state as a potential source
// of provider dependencies.
if back != nil {
terraform: Ugly huge change to weave in new State and Plan types
Due to how often the state and plan types are referenced throughout
Terraform, there isn't a great way to switch them out gradually. As a
consequence, this huge commit gets us from the old world to a _compilable_
new world, but still has a large number of known test failures due to
key functionality being stubbed out.
The stubs here are for anything that interacts with providers, since we
now need to do the follow-up work to similarly replace the old
terraform.ResourceProvider interface with its replacement in the new
"providers" package. That work, along with work to fix the remaining
failing tests, will follow in subsequent commits.
The aim here was to replace all references to terraform.State and its
downstream types with states.State, terraform.Plan with plans.Plan,
state.State with statemgr.State, and switch to the new implementations of
the state and plan file formats. However, due to the number of times those
types are used, this also ended up affecting numerous other parts of core
such as terraform.Hook, the backend.Backend interface, and most of the CLI
commands.
Just as with 5861dbf3fc49b19587a31816eb06f511ab861bb4 before, I apologize
in advance to the person who inevitably just found this huge commit while
spelunking through the commit history.
2018-08-14 23:24:45 +02:00
sMgr , err := back . StateMgr ( c . Workspace ( ) )
2017-06-21 19:32:13 +02:00
if err != nil {
2018-10-31 16:45:03 +01:00
c . Ui . Error ( fmt . Sprintf ( "Error loading state: %s" , err ) )
2017-06-21 19:32:13 +02:00
return 1
}
if err := sMgr . RefreshState ( ) ; err != nil {
2018-10-31 16:45:03 +01:00
c . Ui . Error ( fmt . Sprintf ( "Error refreshing state: %s" , err ) )
2017-06-21 19:32:13 +02:00
return 1
}
state = sMgr . State ( )
2017-06-15 21:23:16 +02:00
}
2017-05-03 17:02:47 +02:00
2017-07-03 19:59:13 +02:00
if v := os . Getenv ( ProviderSkipVerifyEnvVar ) ; v != "" {
c . ignorePluginChecksum = true
}
2017-06-21 19:32:13 +02:00
// Now that we have loaded all modules, check the module tree for missing providers.
command: "terraform init" can partially initialize for 0.12upgrade
There are a few constructs from 0.11 and prior that cause 0.12 parsing to
fail altogether, which previously created a chicken/egg problem because
we need to install the providers in order to run "terraform 0.12upgrade"
and thus fix the problem.
This changes "terraform init" to use the new "early configuration" loader
for module and provider installation. This is built on the more permissive
parser in the terraform-config-inspect package, and so it allows us to
read out the top-level blocks from the configuration while accepting
legacy HCL syntax.
In the long run this will let us do version compatibility detection before
attempting a "real" config load, giving us better error messages for any
future syntax additions, but in the short term the key thing is that it
allows us to install the dependencies even if the configuration isn't
fully valid.
Because backend init still requires full configuration, this introduces a
new mode of terraform init where it detects heuristically if it seems like
we need to do a configuration upgrade and does a partial init if so,
before finally directing the user to run "terraform 0.12upgrade" before
running any other commands.
The heuristic here is based on two assumptions:
- If the "early" loader finds no errors but the normal loader does, the
configuration is likely to be valid for Terraform 0.11 but not 0.12.
- If there's already a version constraint in the configuration that
excludes Terraform versions prior to v0.12 then the configuration is
probably _already_ upgraded and so it's just a normal syntax error,
even if the early loader didn't detect it.
Once the upgrade process is removed in 0.13.0 (users will be required to
go stepwise 0.11 -> 0.12 -> 0.13 to upgrade after that), some of this can
be simplified to remove that special mode, but the idea of doing the
dependency version checks against the liberal parser will remain valuable
to increase our chances of reporting version-based incompatibilities
rather than syntax errors as we add new features in future.
2019-01-14 20:11:00 +01:00
providersOutput , providerDiags := c . getProviders ( earlyConfig , state , flagUpgrade )
2018-03-28 02:22:51 +02:00
diags = diags . Append ( providerDiags )
if providerDiags . HasErrors ( ) {
c . showDiagnostics ( diags )
2017-06-15 21:23:16 +02:00
return 1
2017-05-03 17:02:47 +02:00
}
command: "terraform init" can partially initialize for 0.12upgrade
There are a few constructs from 0.11 and prior that cause 0.12 parsing to
fail altogether, which previously created a chicken/egg problem because
we need to install the providers in order to run "terraform 0.12upgrade"
and thus fix the problem.
This changes "terraform init" to use the new "early configuration" loader
for module and provider installation. This is built on the more permissive
parser in the terraform-config-inspect package, and so it allows us to
read out the top-level blocks from the configuration while accepting
legacy HCL syntax.
In the long run this will let us do version compatibility detection before
attempting a "real" config load, giving us better error messages for any
future syntax additions, but in the short term the key thing is that it
allows us to install the dependencies even if the configuration isn't
fully valid.
Because backend init still requires full configuration, this introduces a
new mode of terraform init where it detects heuristically if it seems like
we need to do a configuration upgrade and does a partial init if so,
before finally directing the user to run "terraform 0.12upgrade" before
running any other commands.
The heuristic here is based on two assumptions:
- If the "early" loader finds no errors but the normal loader does, the
configuration is likely to be valid for Terraform 0.11 but not 0.12.
- If there's already a version constraint in the configuration that
excludes Terraform versions prior to v0.12 then the configuration is
probably _already_ upgraded and so it's just a normal syntax error,
even if the early loader didn't detect it.
Once the upgrade process is removed in 0.13.0 (users will be required to
go stepwise 0.11 -> 0.12 -> 0.13 to upgrade after that), some of this can
be simplified to remove that special mode, but the idea of doing the
dependency version checks against the liberal parser will remain valuable
to increase our chances of reporting version-based incompatibilities
rather than syntax errors as we add new features in future.
2019-01-14 20:11:00 +01:00
if providersOutput {
header = true
}
2017-05-03 17:02:47 +02:00
2017-01-19 05:50:45 +01:00
/ / If we outputted information , then we need to output a newline
// so that our success message is nicely spaced out from prior text.
if header {
c . Ui . Output ( "" )
2014-10-01 01:05:40 +02:00
}
2017-01-19 05:50:45 +01:00
2018-03-28 02:22:51 +02:00
// If we accumulated any warnings along the way that weren't accompanied
// by errors then we'll output them here so that the success message is
// still the final thing shown.
c . showDiagnostics ( diags )
command: "terraform init" can partially initialize for 0.12upgrade
There are a few constructs from 0.11 and prior that cause 0.12 parsing to
fail altogether, which previously created a chicken/egg problem because
we need to install the providers in order to run "terraform 0.12upgrade"
and thus fix the problem.
This changes "terraform init" to use the new "early configuration" loader
for module and provider installation. This is built on the more permissive
parser in the terraform-config-inspect package, and so it allows us to
read out the top-level blocks from the configuration while accepting
legacy HCL syntax.
In the long run this will let us do version compatibility detection before
attempting a "real" config load, giving us better error messages for any
future syntax additions, but in the short term the key thing is that it
allows us to install the dependencies even if the configuration isn't
fully valid.
Because backend init still requires full configuration, this introduces a
new mode of terraform init where it detects heuristically if it seems like
we need to do a configuration upgrade and does a partial init if so,
before finally directing the user to run "terraform 0.12upgrade" before
running any other commands.
The heuristic here is based on two assumptions:
- If the "early" loader finds no errors but the normal loader does, the
configuration is likely to be valid for Terraform 0.11 but not 0.12.
- If there's already a version constraint in the configuration that
excludes Terraform versions prior to v0.12 then the configuration is
probably _already_ upgraded and so it's just a normal syntax error,
even if the early loader didn't detect it.
Once the upgrade process is removed in 0.13.0 (users will be required to
go stepwise 0.11 -> 0.12 -> 0.13 to upgrade after that), some of this can
be simplified to remove that special mode, but the idea of doing the
dependency version checks against the liberal parser will remain valuable
to increase our chances of reporting version-based incompatibilities
rather than syntax errors as we add new features in future.
2019-01-14 20:11:00 +01:00
if configUpgradeProbablyNeeded {
switch {
case c . RunningInAutomation :
c . Ui . Output ( c . Colorize ( ) . Color ( strings . TrimSpace ( outputInitSuccessConfigUpgrade ) ) )
default :
c . Ui . Output ( c . Colorize ( ) . Color ( strings . TrimSpace ( outputInitSuccessConfigUpgradeCLI ) ) )
}
return 0
}
2017-01-19 05:50:45 +01:00
c . Ui . Output ( c . Colorize ( ) . Color ( strings . TrimSpace ( outputInitSuccess ) ) )
2017-09-09 02:14:37 +02:00
if ! c . RunningInAutomation {
// If we're not running in an automation wrapper, give the user
// some more detailed next steps that are appropriate for interactive
// shell usage.
c . Ui . Output ( c . Colorize ( ) . Color ( strings . TrimSpace ( outputInitSuccessCLI ) ) )
}
2017-01-19 05:50:45 +01:00
2014-09-27 01:03:39 +02:00
return 0
}
command: "terraform init" can partially initialize for 0.12upgrade
There are a few constructs from 0.11 and prior that cause 0.12 parsing to
fail altogether, which previously created a chicken/egg problem because
we need to install the providers in order to run "terraform 0.12upgrade"
and thus fix the problem.
This changes "terraform init" to use the new "early configuration" loader
for module and provider installation. This is built on the more permissive
parser in the terraform-config-inspect package, and so it allows us to
read out the top-level blocks from the configuration while accepting
legacy HCL syntax.
In the long run this will let us do version compatibility detection before
attempting a "real" config load, giving us better error messages for any
future syntax additions, but in the short term the key thing is that it
allows us to install the dependencies even if the configuration isn't
fully valid.
Because backend init still requires full configuration, this introduces a
new mode of terraform init where it detects heuristically if it seems like
we need to do a configuration upgrade and does a partial init if so,
before finally directing the user to run "terraform 0.12upgrade" before
running any other commands.
The heuristic here is based on two assumptions:
- If the "early" loader finds no errors but the normal loader does, the
configuration is likely to be valid for Terraform 0.11 but not 0.12.
- If there's already a version constraint in the configuration that
excludes Terraform versions prior to v0.12 then the configuration is
probably _already_ upgraded and so it's just a normal syntax error,
even if the early loader didn't detect it.
Once the upgrade process is removed in 0.13.0 (users will be required to
go stepwise 0.11 -> 0.12 -> 0.13 to upgrade after that), some of this can
be simplified to remove that special mode, but the idea of doing the
dependency version checks against the liberal parser will remain valuable
to increase our chances of reporting version-based incompatibilities
rather than syntax errors as we add new features in future.
2019-01-14 20:11:00 +01:00
func ( c * InitCommand ) getModules ( path string , earlyRoot * tfconfig . Module , upgrade bool ) ( output bool , diags tfdiags . Diagnostics ) {
if len ( earlyRoot . ModuleCalls ) == 0 {
// Nothing to do
return false , nil
2018-03-28 00:31:05 +02:00
}
command: "terraform init" can partially initialize for 0.12upgrade
There are a few constructs from 0.11 and prior that cause 0.12 parsing to
fail altogether, which previously created a chicken/egg problem because
we need to install the providers in order to run "terraform 0.12upgrade"
and thus fix the problem.
This changes "terraform init" to use the new "early configuration" loader
for module and provider installation. This is built on the more permissive
parser in the terraform-config-inspect package, and so it allows us to
read out the top-level blocks from the configuration while accepting
legacy HCL syntax.
In the long run this will let us do version compatibility detection before
attempting a "real" config load, giving us better error messages for any
future syntax additions, but in the short term the key thing is that it
allows us to install the dependencies even if the configuration isn't
fully valid.
Because backend init still requires full configuration, this introduces a
new mode of terraform init where it detects heuristically if it seems like
we need to do a configuration upgrade and does a partial init if so,
before finally directing the user to run "terraform 0.12upgrade" before
running any other commands.
The heuristic here is based on two assumptions:
- If the "early" loader finds no errors but the normal loader does, the
configuration is likely to be valid for Terraform 0.11 but not 0.12.
- If there's already a version constraint in the configuration that
excludes Terraform versions prior to v0.12 then the configuration is
probably _already_ upgraded and so it's just a normal syntax error,
even if the early loader didn't detect it.
Once the upgrade process is removed in 0.13.0 (users will be required to
go stepwise 0.11 -> 0.12 -> 0.13 to upgrade after that), some of this can
be simplified to remove that special mode, but the idea of doing the
dependency version checks against the liberal parser will remain valuable
to increase our chances of reporting version-based incompatibilities
rather than syntax errors as we add new features in future.
2019-01-14 20:11:00 +01:00
if upgrade {
c . Ui . Output ( c . Colorize ( ) . Color ( fmt . Sprintf ( "[reset][bold]Upgrading modules..." ) ) )
} else {
c . Ui . Output ( c . Colorize ( ) . Color ( fmt . Sprintf ( "[reset][bold]Initializing modules..." ) ) )
}
hooks := uiModuleInstallHooks {
Ui : c . Ui ,
ShowLocalPaths : true ,
}
instDiags := c . installModules ( path , upgrade , hooks )
diags = diags . Append ( instDiags )
// Since module installer has modified the module manifest on disk, we need
// to refresh the cache of it in the loader.
if c . configLoader != nil {
if err := c . configLoader . RefreshModules ( ) ; err != nil {
// Should never happen
diags = diags . Append ( tfdiags . Sourceless (
tfdiags . Error ,
"Failed to read module manifest" ,
fmt . Sprintf ( "After installing modules, Terraform could not re-read the manifest of installed modules. This is a bug in Terraform. %s." , err ) ,
) )
2018-03-28 00:31:05 +02:00
}
}
command: "terraform init" can partially initialize for 0.12upgrade
There are a few constructs from 0.11 and prior that cause 0.12 parsing to
fail altogether, which previously created a chicken/egg problem because
we need to install the providers in order to run "terraform 0.12upgrade"
and thus fix the problem.
This changes "terraform init" to use the new "early configuration" loader
for module and provider installation. This is built on the more permissive
parser in the terraform-config-inspect package, and so it allows us to
read out the top-level blocks from the configuration while accepting
legacy HCL syntax.
In the long run this will let us do version compatibility detection before
attempting a "real" config load, giving us better error messages for any
future syntax additions, but in the short term the key thing is that it
allows us to install the dependencies even if the configuration isn't
fully valid.
Because backend init still requires full configuration, this introduces a
new mode of terraform init where it detects heuristically if it seems like
we need to do a configuration upgrade and does a partial init if so,
before finally directing the user to run "terraform 0.12upgrade" before
running any other commands.
The heuristic here is based on two assumptions:
- If the "early" loader finds no errors but the normal loader does, the
configuration is likely to be valid for Terraform 0.11 but not 0.12.
- If there's already a version constraint in the configuration that
excludes Terraform versions prior to v0.12 then the configuration is
probably _already_ upgraded and so it's just a normal syntax error,
even if the early loader didn't detect it.
Once the upgrade process is removed in 0.13.0 (users will be required to
go stepwise 0.11 -> 0.12 -> 0.13 to upgrade after that), some of this can
be simplified to remove that special mode, but the idea of doing the
dependency version checks against the liberal parser will remain valuable
to increase our chances of reporting version-based incompatibilities
rather than syntax errors as we add new features in future.
2019-01-14 20:11:00 +01:00
return true , diags
}
func ( c * InitCommand ) initBackend ( root * configs . Module , extraConfig rawFlags ) ( be backend . Backend , output bool , diags tfdiags . Diagnostics ) {
c . Ui . Output ( c . Colorize ( ) . Color ( fmt . Sprintf ( "\n[reset][bold]Initializing the backend..." ) ) )
var backendConfig * configs . Backend
var backendConfigOverride hcl . Body
if root . Backend != nil {
backendType := root . Backend . Type
bf := backendInit . Backend ( backendType )
if bf == nil {
diags = diags . Append ( & hcl . Diagnostic {
Severity : hcl . DiagError ,
Summary : "Unsupported backend type" ,
Detail : fmt . Sprintf ( "There is no backend type named %q." , backendType ) ,
Subject : & root . Backend . TypeRange ,
} )
return nil , true , diags
2018-03-28 00:31:05 +02:00
}
command: "terraform init" can partially initialize for 0.12upgrade
There are a few constructs from 0.11 and prior that cause 0.12 parsing to
fail altogether, which previously created a chicken/egg problem because
we need to install the providers in order to run "terraform 0.12upgrade"
and thus fix the problem.
This changes "terraform init" to use the new "early configuration" loader
for module and provider installation. This is built on the more permissive
parser in the terraform-config-inspect package, and so it allows us to
read out the top-level blocks from the configuration while accepting
legacy HCL syntax.
In the long run this will let us do version compatibility detection before
attempting a "real" config load, giving us better error messages for any
future syntax additions, but in the short term the key thing is that it
allows us to install the dependencies even if the configuration isn't
fully valid.
Because backend init still requires full configuration, this introduces a
new mode of terraform init where it detects heuristically if it seems like
we need to do a configuration upgrade and does a partial init if so,
before finally directing the user to run "terraform 0.12upgrade" before
running any other commands.
The heuristic here is based on two assumptions:
- If the "early" loader finds no errors but the normal loader does, the
configuration is likely to be valid for Terraform 0.11 but not 0.12.
- If there's already a version constraint in the configuration that
excludes Terraform versions prior to v0.12 then the configuration is
probably _already_ upgraded and so it's just a normal syntax error,
even if the early loader didn't detect it.
Once the upgrade process is removed in 0.13.0 (users will be required to
go stepwise 0.11 -> 0.12 -> 0.13 to upgrade after that), some of this can
be simplified to remove that special mode, but the idea of doing the
dependency version checks against the liberal parser will remain valuable
to increase our chances of reporting version-based incompatibilities
rather than syntax errors as we add new features in future.
2019-01-14 20:11:00 +01:00
b := bf ( )
backendSchema := b . ConfigSchema ( )
backendConfig = root . Backend
2018-03-28 00:31:05 +02:00
command: "terraform init" can partially initialize for 0.12upgrade
There are a few constructs from 0.11 and prior that cause 0.12 parsing to
fail altogether, which previously created a chicken/egg problem because
we need to install the providers in order to run "terraform 0.12upgrade"
and thus fix the problem.
This changes "terraform init" to use the new "early configuration" loader
for module and provider installation. This is built on the more permissive
parser in the terraform-config-inspect package, and so it allows us to
read out the top-level blocks from the configuration while accepting
legacy HCL syntax.
In the long run this will let us do version compatibility detection before
attempting a "real" config load, giving us better error messages for any
future syntax additions, but in the short term the key thing is that it
allows us to install the dependencies even if the configuration isn't
fully valid.
Because backend init still requires full configuration, this introduces a
new mode of terraform init where it detects heuristically if it seems like
we need to do a configuration upgrade and does a partial init if so,
before finally directing the user to run "terraform 0.12upgrade" before
running any other commands.
The heuristic here is based on two assumptions:
- If the "early" loader finds no errors but the normal loader does, the
configuration is likely to be valid for Terraform 0.11 but not 0.12.
- If there's already a version constraint in the configuration that
excludes Terraform versions prior to v0.12 then the configuration is
probably _already_ upgraded and so it's just a normal syntax error,
even if the early loader didn't detect it.
Once the upgrade process is removed in 0.13.0 (users will be required to
go stepwise 0.11 -> 0.12 -> 0.13 to upgrade after that), some of this can
be simplified to remove that special mode, but the idea of doing the
dependency version checks against the liberal parser will remain valuable
to increase our chances of reporting version-based incompatibilities
rather than syntax errors as we add new features in future.
2019-01-14 20:11:00 +01:00
var overrideDiags tfdiags . Diagnostics
backendConfigOverride , overrideDiags = c . backendConfigOverrideBody ( extraConfig , backendSchema )
diags = diags . Append ( overrideDiags )
if overrideDiags . HasErrors ( ) {
return nil , true , diags
2018-03-28 00:31:05 +02:00
}
2019-07-23 14:08:28 +02:00
} else {
// If the user supplied a -backend-config on the CLI but no backend
// block was found in the configuration, it's likely - but not
// necessarily - a mistake. Return a warning.
if ! extraConfig . Empty ( ) {
diags = diags . Append ( tfdiags . Sourceless (
tfdiags . Warning ,
"Missing backend configuration" ,
` - backend - config was used without a "backend" block in the configuration .
If you intended to override the default local backend configuration ,
no action is required , but you may add an explicit backend block to your
configuration to clear this warning :
terraform {
backend "local" { }
}
However , if you intended to override a defined backend , please verify that
the backend configuration is present and valid .
` ,
) )
}
2018-03-28 00:31:05 +02:00
}
command: "terraform init" can partially initialize for 0.12upgrade
There are a few constructs from 0.11 and prior that cause 0.12 parsing to
fail altogether, which previously created a chicken/egg problem because
we need to install the providers in order to run "terraform 0.12upgrade"
and thus fix the problem.
This changes "terraform init" to use the new "early configuration" loader
for module and provider installation. This is built on the more permissive
parser in the terraform-config-inspect package, and so it allows us to
read out the top-level blocks from the configuration while accepting
legacy HCL syntax.
In the long run this will let us do version compatibility detection before
attempting a "real" config load, giving us better error messages for any
future syntax additions, but in the short term the key thing is that it
allows us to install the dependencies even if the configuration isn't
fully valid.
Because backend init still requires full configuration, this introduces a
new mode of terraform init where it detects heuristically if it seems like
we need to do a configuration upgrade and does a partial init if so,
before finally directing the user to run "terraform 0.12upgrade" before
running any other commands.
The heuristic here is based on two assumptions:
- If the "early" loader finds no errors but the normal loader does, the
configuration is likely to be valid for Terraform 0.11 but not 0.12.
- If there's already a version constraint in the configuration that
excludes Terraform versions prior to v0.12 then the configuration is
probably _already_ upgraded and so it's just a normal syntax error,
even if the early loader didn't detect it.
Once the upgrade process is removed in 0.13.0 (users will be required to
go stepwise 0.11 -> 0.12 -> 0.13 to upgrade after that), some of this can
be simplified to remove that special mode, but the idea of doing the
dependency version checks against the liberal parser will remain valuable
to increase our chances of reporting version-based incompatibilities
rather than syntax errors as we add new features in future.
2019-01-14 20:11:00 +01:00
opts := & BackendOpts {
Config : backendConfig ,
ConfigOverride : backendConfigOverride ,
Init : true ,
}
2019-02-26 00:37:20 +01:00
command: "terraform init" can partially initialize for 0.12upgrade
There are a few constructs from 0.11 and prior that cause 0.12 parsing to
fail altogether, which previously created a chicken/egg problem because
we need to install the providers in order to run "terraform 0.12upgrade"
and thus fix the problem.
This changes "terraform init" to use the new "early configuration" loader
for module and provider installation. This is built on the more permissive
parser in the terraform-config-inspect package, and so it allows us to
read out the top-level blocks from the configuration while accepting
legacy HCL syntax.
In the long run this will let us do version compatibility detection before
attempting a "real" config load, giving us better error messages for any
future syntax additions, but in the short term the key thing is that it
allows us to install the dependencies even if the configuration isn't
fully valid.
Because backend init still requires full configuration, this introduces a
new mode of terraform init where it detects heuristically if it seems like
we need to do a configuration upgrade and does a partial init if so,
before finally directing the user to run "terraform 0.12upgrade" before
running any other commands.
The heuristic here is based on two assumptions:
- If the "early" loader finds no errors but the normal loader does, the
configuration is likely to be valid for Terraform 0.11 but not 0.12.
- If there's already a version constraint in the configuration that
excludes Terraform versions prior to v0.12 then the configuration is
probably _already_ upgraded and so it's just a normal syntax error,
even if the early loader didn't detect it.
Once the upgrade process is removed in 0.13.0 (users will be required to
go stepwise 0.11 -> 0.12 -> 0.13 to upgrade after that), some of this can
be simplified to remove that special mode, but the idea of doing the
dependency version checks against the liberal parser will remain valuable
to increase our chances of reporting version-based incompatibilities
rather than syntax errors as we add new features in future.
2019-01-14 20:11:00 +01:00
back , backDiags := c . Backend ( opts )
diags = diags . Append ( backDiags )
return back , true , diags
2018-03-28 00:31:05 +02:00
}
2017-06-09 18:42:45 +02:00
// Load the complete module tree, and fetch any missing providers.
// This method outputs its own Ui.
command: "terraform init" can partially initialize for 0.12upgrade
There are a few constructs from 0.11 and prior that cause 0.12 parsing to
fail altogether, which previously created a chicken/egg problem because
we need to install the providers in order to run "terraform 0.12upgrade"
and thus fix the problem.
This changes "terraform init" to use the new "early configuration" loader
for module and provider installation. This is built on the more permissive
parser in the terraform-config-inspect package, and so it allows us to
read out the top-level blocks from the configuration while accepting
legacy HCL syntax.
In the long run this will let us do version compatibility detection before
attempting a "real" config load, giving us better error messages for any
future syntax additions, but in the short term the key thing is that it
allows us to install the dependencies even if the configuration isn't
fully valid.
Because backend init still requires full configuration, this introduces a
new mode of terraform init where it detects heuristically if it seems like
we need to do a configuration upgrade and does a partial init if so,
before finally directing the user to run "terraform 0.12upgrade" before
running any other commands.
The heuristic here is based on two assumptions:
- If the "early" loader finds no errors but the normal loader does, the
configuration is likely to be valid for Terraform 0.11 but not 0.12.
- If there's already a version constraint in the configuration that
excludes Terraform versions prior to v0.12 then the configuration is
probably _already_ upgraded and so it's just a normal syntax error,
even if the early loader didn't detect it.
Once the upgrade process is removed in 0.13.0 (users will be required to
go stepwise 0.11 -> 0.12 -> 0.13 to upgrade after that), some of this can
be simplified to remove that special mode, but the idea of doing the
dependency version checks against the liberal parser will remain valuable
to increase our chances of reporting version-based incompatibilities
rather than syntax errors as we add new features in future.
2019-01-14 20:11:00 +01:00
func ( c * InitCommand ) getProviders ( earlyConfig * earlyconfig . Config , state * states . State , upgrade bool ) ( output bool , diags tfdiags . Diagnostics ) {
2017-06-13 03:32:42 +02:00
var available discovery . PluginMetaSet
if upgrade {
// If we're in upgrade mode, we ignore any auto-installed plugins
// in "available", causing us to reinstall and possibly upgrade them.
available = c . providerPluginManuallyInstalledSet ( )
} else {
available = c . providerPluginSet ( )
}
2017-06-19 16:23:58 +02:00
command: "terraform init" can partially initialize for 0.12upgrade
There are a few constructs from 0.11 and prior that cause 0.12 parsing to
fail altogether, which previously created a chicken/egg problem because
we need to install the providers in order to run "terraform 0.12upgrade"
and thus fix the problem.
This changes "terraform init" to use the new "early configuration" loader
for module and provider installation. This is built on the more permissive
parser in the terraform-config-inspect package, and so it allows us to
read out the top-level blocks from the configuration while accepting
legacy HCL syntax.
In the long run this will let us do version compatibility detection before
attempting a "real" config load, giving us better error messages for any
future syntax additions, but in the short term the key thing is that it
allows us to install the dependencies even if the configuration isn't
fully valid.
Because backend init still requires full configuration, this introduces a
new mode of terraform init where it detects heuristically if it seems like
we need to do a configuration upgrade and does a partial init if so,
before finally directing the user to run "terraform 0.12upgrade" before
running any other commands.
The heuristic here is based on two assumptions:
- If the "early" loader finds no errors but the normal loader does, the
configuration is likely to be valid for Terraform 0.11 but not 0.12.
- If there's already a version constraint in the configuration that
excludes Terraform versions prior to v0.12 then the configuration is
probably _already_ upgraded and so it's just a normal syntax error,
even if the early loader didn't detect it.
Once the upgrade process is removed in 0.13.0 (users will be required to
go stepwise 0.11 -> 0.12 -> 0.13 to upgrade after that), some of this can
be simplified to remove that special mode, but the idea of doing the
dependency version checks against the liberal parser will remain valuable
to increase our chances of reporting version-based incompatibilities
rather than syntax errors as we add new features in future.
2019-01-14 20:11:00 +01:00
configDeps , depsDiags := earlyConfig . ProviderDependencies ( )
diags = diags . Append ( depsDiags )
if depsDiags . HasErrors ( ) {
return false , diags
}
configReqs := configDeps . AllPluginRequirements ( )
// FIXME: This is weird because ConfigTreeDependencies was written before
// we switched over to using earlyConfig as the main source of dependencies.
2019-12-09 15:47:09 +01:00
// In future we should clean this up to be a more reasonable API.
command: "terraform init" can partially initialize for 0.12upgrade
There are a few constructs from 0.11 and prior that cause 0.12 parsing to
fail altogether, which previously created a chicken/egg problem because
we need to install the providers in order to run "terraform 0.12upgrade"
and thus fix the problem.
This changes "terraform init" to use the new "early configuration" loader
for module and provider installation. This is built on the more permissive
parser in the terraform-config-inspect package, and so it allows us to
read out the top-level blocks from the configuration while accepting
legacy HCL syntax.
In the long run this will let us do version compatibility detection before
attempting a "real" config load, giving us better error messages for any
future syntax additions, but in the short term the key thing is that it
allows us to install the dependencies even if the configuration isn't
fully valid.
Because backend init still requires full configuration, this introduces a
new mode of terraform init where it detects heuristically if it seems like
we need to do a configuration upgrade and does a partial init if so,
before finally directing the user to run "terraform 0.12upgrade" before
running any other commands.
The heuristic here is based on two assumptions:
- If the "early" loader finds no errors but the normal loader does, the
configuration is likely to be valid for Terraform 0.11 but not 0.12.
- If there's already a version constraint in the configuration that
excludes Terraform versions prior to v0.12 then the configuration is
probably _already_ upgraded and so it's just a normal syntax error,
even if the early loader didn't detect it.
Once the upgrade process is removed in 0.13.0 (users will be required to
go stepwise 0.11 -> 0.12 -> 0.13 to upgrade after that), some of this can
be simplified to remove that special mode, but the idea of doing the
dependency version checks against the liberal parser will remain valuable
to increase our chances of reporting version-based incompatibilities
rather than syntax errors as we add new features in future.
2019-01-14 20:11:00 +01:00
stateReqs := terraform . ConfigTreeDependencies ( nil , state ) . AllPluginRequirements ( )
requirements := configReqs . Merge ( stateReqs )
2017-06-19 16:23:58 +02:00
if len ( requirements ) == 0 {
// nothing to initialize
command: "terraform init" can partially initialize for 0.12upgrade
There are a few constructs from 0.11 and prior that cause 0.12 parsing to
fail altogether, which previously created a chicken/egg problem because
we need to install the providers in order to run "terraform 0.12upgrade"
and thus fix the problem.
This changes "terraform init" to use the new "early configuration" loader
for module and provider installation. This is built on the more permissive
parser in the terraform-config-inspect package, and so it allows us to
read out the top-level blocks from the configuration while accepting
legacy HCL syntax.
In the long run this will let us do version compatibility detection before
attempting a "real" config load, giving us better error messages for any
future syntax additions, but in the short term the key thing is that it
allows us to install the dependencies even if the configuration isn't
fully valid.
Because backend init still requires full configuration, this introduces a
new mode of terraform init where it detects heuristically if it seems like
we need to do a configuration upgrade and does a partial init if so,
before finally directing the user to run "terraform 0.12upgrade" before
running any other commands.
The heuristic here is based on two assumptions:
- If the "early" loader finds no errors but the normal loader does, the
configuration is likely to be valid for Terraform 0.11 but not 0.12.
- If there's already a version constraint in the configuration that
excludes Terraform versions prior to v0.12 then the configuration is
probably _already_ upgraded and so it's just a normal syntax error,
even if the early loader didn't detect it.
Once the upgrade process is removed in 0.13.0 (users will be required to
go stepwise 0.11 -> 0.12 -> 0.13 to upgrade after that), some of this can
be simplified to remove that special mode, but the idea of doing the
dependency version checks against the liberal parser will remain valuable
to increase our chances of reporting version-based incompatibilities
rather than syntax errors as we add new features in future.
2019-01-14 20:11:00 +01:00
return false , nil
2017-06-19 16:23:58 +02:00
}
c . Ui . Output ( c . Colorize ( ) . Color (
"\n[reset][bold]Initializing provider plugins..." ,
) )
2017-05-25 02:35:46 +02:00
missing := c . missingPlugins ( available , requirements )
2017-05-03 17:02:47 +02:00
2017-06-15 21:23:16 +02:00
if c . getPlugins {
2017-08-14 13:32:48 +02:00
if len ( missing ) > 0 {
2018-08-25 01:13:50 +02:00
c . Ui . Output ( "- Checking for available provider plugins..." )
2017-08-14 13:32:48 +02:00
}
2017-06-15 21:23:16 +02:00
for provider , reqd := range missing {
2019-12-09 21:17:47 +01:00
pty := addrs . NewLegacyProvider ( provider )
2019-09-09 22:59:50 +02:00
_ , providerDiags , err := c . providerInstaller . Get ( pty , reqd . Versions )
2019-03-13 17:52:15 +01:00
diags = diags . Append ( providerDiags )
2017-06-15 21:23:16 +02:00
if err != nil {
2019-01-14 20:07:54 +01:00
constraint := reqd . Versions . String ( )
if constraint == "" {
constraint = "(any version)"
}
switch {
2019-04-05 22:47:17 +02:00
case err == discovery . ErrorServiceUnreachable , err == discovery . ErrorPublicRegistryUnreachable :
2019-03-27 19:20:15 +01:00
c . Ui . Error ( errDiscoveryServiceUnreachable )
2019-01-14 20:07:54 +01:00
case err == discovery . ErrorNoSuchProvider :
2017-06-20 03:49:37 +02:00
c . Ui . Error ( fmt . Sprintf ( errProviderNotFound , provider , DefaultPluginVendorDir ) )
2019-01-14 20:07:54 +01:00
case err == discovery . ErrorNoSuitableVersion :
2017-06-21 02:40:28 +02:00
if reqd . Versions . Unconstrained ( ) {
// This should never happen, but might crop up if we catch
// the releases server in a weird state where the provider's
// directory is present but does not yet contain any
// versions. We'll treat it like ErrorNoSuchProvider, then.
c . Ui . Error ( fmt . Sprintf ( errProviderNotFound , provider , DefaultPluginVendorDir ) )
} else {
c . Ui . Error ( fmt . Sprintf ( errProviderVersionsUnsuitable , provider , reqd . Versions ) )
}
2019-01-14 20:07:54 +01:00
case errwrap . Contains ( err , discovery . ErrorVersionIncompatible . Error ( ) ) :
// Attempt to fetch nested error to display to the user which versions
// we considered and which versions might be compatible. Otherwise,
// we'll just display a generic version incompatible msg
incompatErr := errwrap . GetType ( err , fmt . Errorf ( "" ) )
if incompatErr != nil {
c . Ui . Error ( incompatErr . Error ( ) )
} else {
// Generic version incompatible msg
c . Ui . Error ( fmt . Sprintf ( errProviderIncompatible , provider , constraint ) )
2017-06-21 02:40:28 +02:00
}
2019-01-14 20:07:54 +01:00
// Reset nested errors
err = discovery . ErrorVersionIncompatible
case err == discovery . ErrorNoVersionCompatible :
// Generic version incompatible msg
2017-06-21 02:40:28 +02:00
c . Ui . Error ( fmt . Sprintf ( errProviderIncompatible , provider , constraint ) )
2019-03-21 20:38:51 +01:00
case err == discovery . ErrorSignatureVerification :
c . Ui . Error ( fmt . Sprintf ( errSignatureVerification , provider ) )
case err == discovery . ErrorChecksumVerification ,
err == discovery . ErrorMissingChecksumVerification :
c . Ui . Error ( fmt . Sprintf ( errChecksumVerification , provider ) )
2017-06-20 03:49:37 +02:00
default :
c . Ui . Error ( fmt . Sprintf ( errProviderInstallError , provider , err . Error ( ) , DefaultPluginVendorDir ) )
}
2018-03-28 02:22:51 +02:00
diags = diags . Append ( err )
2017-06-15 21:23:16 +02:00
}
}
2017-06-09 18:42:45 +02:00
2018-03-28 02:22:51 +02:00
if diags . HasErrors ( ) {
command: "terraform init" can partially initialize for 0.12upgrade
There are a few constructs from 0.11 and prior that cause 0.12 parsing to
fail altogether, which previously created a chicken/egg problem because
we need to install the providers in order to run "terraform 0.12upgrade"
and thus fix the problem.
This changes "terraform init" to use the new "early configuration" loader
for module and provider installation. This is built on the more permissive
parser in the terraform-config-inspect package, and so it allows us to
read out the top-level blocks from the configuration while accepting
legacy HCL syntax.
In the long run this will let us do version compatibility detection before
attempting a "real" config load, giving us better error messages for any
future syntax additions, but in the short term the key thing is that it
allows us to install the dependencies even if the configuration isn't
fully valid.
Because backend init still requires full configuration, this introduces a
new mode of terraform init where it detects heuristically if it seems like
we need to do a configuration upgrade and does a partial init if so,
before finally directing the user to run "terraform 0.12upgrade" before
running any other commands.
The heuristic here is based on two assumptions:
- If the "early" loader finds no errors but the normal loader does, the
configuration is likely to be valid for Terraform 0.11 but not 0.12.
- If there's already a version constraint in the configuration that
excludes Terraform versions prior to v0.12 then the configuration is
probably _already_ upgraded and so it's just a normal syntax error,
even if the early loader didn't detect it.
Once the upgrade process is removed in 0.13.0 (users will be required to
go stepwise 0.11 -> 0.12 -> 0.13 to upgrade after that), some of this can
be simplified to remove that special mode, but the idea of doing the
dependency version checks against the liberal parser will remain valuable
to increase our chances of reporting version-based incompatibilities
rather than syntax errors as we add new features in future.
2019-01-14 20:11:00 +01:00
return true , diags
2017-06-15 21:23:16 +02:00
}
} else if len ( missing ) > 0 {
// we have missing providers, but aren't going to try and download them
2017-06-20 03:49:37 +02:00
var lines [ ] string
2017-06-15 21:23:16 +02:00
for provider , reqd := range missing {
2017-06-20 03:49:37 +02:00
if reqd . Versions . Unconstrained ( ) {
lines = append ( lines , fmt . Sprintf ( "* %s (any version)\n" , provider ) )
} else {
lines = append ( lines , fmt . Sprintf ( "* %s (%s)\n" , provider , reqd . Versions ) )
}
2018-03-28 02:22:51 +02:00
diags = diags . Append ( fmt . Errorf ( "missing provider %q" , provider ) )
2017-05-03 17:02:47 +02:00
}
2017-06-20 03:49:37 +02:00
sort . Strings ( lines )
c . Ui . Error ( fmt . Sprintf ( errMissingProvidersNoInstall , strings . Join ( lines , "" ) , DefaultPluginVendorDir ) )
command: "terraform init" can partially initialize for 0.12upgrade
There are a few constructs from 0.11 and prior that cause 0.12 parsing to
fail altogether, which previously created a chicken/egg problem because
we need to install the providers in order to run "terraform 0.12upgrade"
and thus fix the problem.
This changes "terraform init" to use the new "early configuration" loader
for module and provider installation. This is built on the more permissive
parser in the terraform-config-inspect package, and so it allows us to
read out the top-level blocks from the configuration while accepting
legacy HCL syntax.
In the long run this will let us do version compatibility detection before
attempting a "real" config load, giving us better error messages for any
future syntax additions, but in the short term the key thing is that it
allows us to install the dependencies even if the configuration isn't
fully valid.
Because backend init still requires full configuration, this introduces a
new mode of terraform init where it detects heuristically if it seems like
we need to do a configuration upgrade and does a partial init if so,
before finally directing the user to run "terraform 0.12upgrade" before
running any other commands.
The heuristic here is based on two assumptions:
- If the "early" loader finds no errors but the normal loader does, the
configuration is likely to be valid for Terraform 0.11 but not 0.12.
- If there's already a version constraint in the configuration that
excludes Terraform versions prior to v0.12 then the configuration is
probably _already_ upgraded and so it's just a normal syntax error,
even if the early loader didn't detect it.
Once the upgrade process is removed in 0.13.0 (users will be required to
go stepwise 0.11 -> 0.12 -> 0.13 to upgrade after that), some of this can
be simplified to remove that special mode, but the idea of doing the
dependency version checks against the liberal parser will remain valuable
to increase our chances of reporting version-based incompatibilities
rather than syntax errors as we add new features in future.
2019-01-14 20:11:00 +01:00
return true , diags
2017-06-09 18:42:45 +02:00
}
2017-05-25 02:35:46 +02:00
// With all the providers downloaded, we'll generate our lock file
// that ensures the provider binaries remain unchanged until we init
// again. If anything changes, other commands that use providers will
// fail with an error instructing the user to re-run this command.
available = c . providerPluginSet ( ) // re-discover to see newly-installed plugins
2018-01-05 17:52:11 +01:00
// internal providers were already filtered out, since we don't need to get them.
2019-12-09 15:47:09 +01:00
chosen := chooseProviders ( available , nil , requirements )
2018-01-05 17:52:11 +01:00
2017-05-25 02:35:46 +02:00
digests := map [ string ] [ ] byte { }
for name , meta := range chosen {
digest , err := meta . SHA256 ( )
if err != nil {
2018-03-28 02:22:51 +02:00
diags = diags . Append ( fmt . Errorf ( "Failed to read provider plugin %s: %s" , meta . Path , err ) )
command: "terraform init" can partially initialize for 0.12upgrade
There are a few constructs from 0.11 and prior that cause 0.12 parsing to
fail altogether, which previously created a chicken/egg problem because
we need to install the providers in order to run "terraform 0.12upgrade"
and thus fix the problem.
This changes "terraform init" to use the new "early configuration" loader
for module and provider installation. This is built on the more permissive
parser in the terraform-config-inspect package, and so it allows us to
read out the top-level blocks from the configuration while accepting
legacy HCL syntax.
In the long run this will let us do version compatibility detection before
attempting a "real" config load, giving us better error messages for any
future syntax additions, but in the short term the key thing is that it
allows us to install the dependencies even if the configuration isn't
fully valid.
Because backend init still requires full configuration, this introduces a
new mode of terraform init where it detects heuristically if it seems like
we need to do a configuration upgrade and does a partial init if so,
before finally directing the user to run "terraform 0.12upgrade" before
running any other commands.
The heuristic here is based on two assumptions:
- If the "early" loader finds no errors but the normal loader does, the
configuration is likely to be valid for Terraform 0.11 but not 0.12.
- If there's already a version constraint in the configuration that
excludes Terraform versions prior to v0.12 then the configuration is
probably _already_ upgraded and so it's just a normal syntax error,
even if the early loader didn't detect it.
Once the upgrade process is removed in 0.13.0 (users will be required to
go stepwise 0.11 -> 0.12 -> 0.13 to upgrade after that), some of this can
be simplified to remove that special mode, but the idea of doing the
dependency version checks against the liberal parser will remain valuable
to increase our chances of reporting version-based incompatibilities
rather than syntax errors as we add new features in future.
2019-01-14 20:11:00 +01:00
return true , diags
2017-05-25 02:35:46 +02:00
}
digests [ name ] = digest
2017-07-03 19:59:13 +02:00
if c . ignorePluginChecksum {
digests [ name ] = nil
}
2017-05-25 02:35:46 +02:00
}
command: validate config as part of loading it
Previously we required callers to separately call .Validate on the root
module to determine if there were any value errors, but we did that
inconsistently and would thus see crashes in some cases where later code
would try to use invalid configuration as if it were valid.
Now we run .Validate automatically after config loading, returning the
resulting diagnostics. Since we return a diagnostics here, it's possible
to return both warnings and errors.
We return the loaded module even if it's invalid, so callers are free to
ignore returned errors and try to work with the config anyway, though they
will need to be defensive against invalid configuration themselves in
that case.
As a result of this, all of the commands that load configuration now need
to use diagnostic printing to signal errors. For the moment this just
allows us to return potentially-multiple config errors/warnings in full
fidelity, but also sets us up for later when more subsystems are able
to produce rich diagnostics so we can show them all together.
Finally, this commit also removes some stale, commented-out code for the
"legacy" (pre-0.8) graph implementation, which has not been available
for some time.
2017-12-07 01:41:48 +01:00
err := c . providerPluginsLock ( ) . Write ( digests )
2017-05-25 02:35:46 +02:00
if err != nil {
2018-03-28 02:22:51 +02:00
diags = diags . Append ( fmt . Errorf ( "failed to save provider manifest: %s" , err ) )
command: "terraform init" can partially initialize for 0.12upgrade
There are a few constructs from 0.11 and prior that cause 0.12 parsing to
fail altogether, which previously created a chicken/egg problem because
we need to install the providers in order to run "terraform 0.12upgrade"
and thus fix the problem.
This changes "terraform init" to use the new "early configuration" loader
for module and provider installation. This is built on the more permissive
parser in the terraform-config-inspect package, and so it allows us to
read out the top-level blocks from the configuration while accepting
legacy HCL syntax.
In the long run this will let us do version compatibility detection before
attempting a "real" config load, giving us better error messages for any
future syntax additions, but in the short term the key thing is that it
allows us to install the dependencies even if the configuration isn't
fully valid.
Because backend init still requires full configuration, this introduces a
new mode of terraform init where it detects heuristically if it seems like
we need to do a configuration upgrade and does a partial init if so,
before finally directing the user to run "terraform 0.12upgrade" before
running any other commands.
The heuristic here is based on two assumptions:
- If the "early" loader finds no errors but the normal loader does, the
configuration is likely to be valid for Terraform 0.11 but not 0.12.
- If there's already a version constraint in the configuration that
excludes Terraform versions prior to v0.12 then the configuration is
probably _already_ upgraded and so it's just a normal syntax error,
even if the early loader didn't detect it.
Once the upgrade process is removed in 0.13.0 (users will be required to
go stepwise 0.11 -> 0.12 -> 0.13 to upgrade after that), some of this can
be simplified to remove that special mode, but the idea of doing the
dependency version checks against the liberal parser will remain valuable
to increase our chances of reporting version-based incompatibilities
rather than syntax errors as we add new features in future.
2019-01-14 20:11:00 +01:00
return true , diags
2017-05-25 02:35:46 +02:00
}
2017-06-20 20:25:41 +02:00
{
2017-06-13 03:32:42 +02:00
// Purge any auto-installed plugins that aren't being used.
purged , err := c . providerInstaller . PurgeUnused ( chosen )
if err != nil {
// Failure to purge old plugins is not a fatal error
c . Ui . Warn ( fmt . Sprintf ( "failed to purge unused plugins: %s" , err ) )
}
if purged != nil {
for meta := range purged {
log . Printf ( "[DEBUG] Purged unused %s plugin %s" , meta . Name , meta . Path )
}
}
}
2017-06-02 02:57:43 +02:00
// If any providers have "floating" versions (completely unconstrained)
// we'll suggest the user constrain with a pessimistic constraint to
// avoid implicitly adopting a later major release.
constraintSuggestions := make ( map [ string ] discovery . ConstraintStr )
for name , meta := range chosen {
req := requirements [ name ]
if req == nil {
// should never happen, but we don't want to crash here, so we'll
// be cautious.
continue
}
2017-06-16 22:19:13 +02:00
if req . Versions . Unconstrained ( ) && meta . Version != discovery . VersionZero {
2017-06-02 02:57:43 +02:00
// meta.Version.MustParse is safe here because our "chosen" metas
// were already filtered for validity of versions.
constraintSuggestions [ name ] = meta . Version . MustParse ( ) . MinorUpgradeConstraintStr ( )
}
}
if len ( constraintSuggestions ) != 0 {
names := make ( [ ] string , 0 , len ( constraintSuggestions ) )
for name := range constraintSuggestions {
names = append ( names , name )
}
sort . Strings ( names )
c . Ui . Output ( outputInitProvidersUnconstrained )
for _ , name := range names {
c . Ui . Output ( fmt . Sprintf ( "* provider.%s: version = %q" , name , constraintSuggestions [ name ] ) )
}
}
command: "terraform init" can partially initialize for 0.12upgrade
There are a few constructs from 0.11 and prior that cause 0.12 parsing to
fail altogether, which previously created a chicken/egg problem because
we need to install the providers in order to run "terraform 0.12upgrade"
and thus fix the problem.
This changes "terraform init" to use the new "early configuration" loader
for module and provider installation. This is built on the more permissive
parser in the terraform-config-inspect package, and so it allows us to
read out the top-level blocks from the configuration while accepting
legacy HCL syntax.
In the long run this will let us do version compatibility detection before
attempting a "real" config load, giving us better error messages for any
future syntax additions, but in the short term the key thing is that it
allows us to install the dependencies even if the configuration isn't
fully valid.
Because backend init still requires full configuration, this introduces a
new mode of terraform init where it detects heuristically if it seems like
we need to do a configuration upgrade and does a partial init if so,
before finally directing the user to run "terraform 0.12upgrade" before
running any other commands.
The heuristic here is based on two assumptions:
- If the "early" loader finds no errors but the normal loader does, the
configuration is likely to be valid for Terraform 0.11 but not 0.12.
- If there's already a version constraint in the configuration that
excludes Terraform versions prior to v0.12 then the configuration is
probably _already_ upgraded and so it's just a normal syntax error,
even if the early loader didn't detect it.
Once the upgrade process is removed in 0.13.0 (users will be required to
go stepwise 0.11 -> 0.12 -> 0.13 to upgrade after that), some of this can
be simplified to remove that special mode, but the idea of doing the
dependency version checks against the liberal parser will remain valuable
to increase our chances of reporting version-based incompatibilities
rather than syntax errors as we add new features in future.
2019-01-14 20:11:00 +01:00
return true , diags
}
// backendConfigOverrideBody interprets the raw values of -backend-config
// arguments into a hcl Body that should override the backend settings given
// in the configuration.
//
// If the result is nil then no override needs to be provided.
//
// If the returned diagnostics contains errors then the returned body may be
// incomplete or invalid.
func ( c * InitCommand ) backendConfigOverrideBody ( flags rawFlags , schema * configschema . Block ) ( hcl . Body , tfdiags . Diagnostics ) {
items := flags . AllItems ( )
if len ( items ) == 0 {
return nil , nil
}
var ret hcl . Body
var diags tfdiags . Diagnostics
synthVals := make ( map [ string ] cty . Value )
mergeBody := func ( newBody hcl . Body ) {
if ret == nil {
ret = newBody
} else {
ret = configs . MergeBodies ( ret , newBody )
}
}
flushVals := func ( ) {
if len ( synthVals ) == 0 {
return
}
newBody := configs . SynthBody ( "-backend-config=..." , synthVals )
mergeBody ( newBody )
synthVals = make ( map [ string ] cty . Value )
}
2019-05-29 19:58:04 +02:00
if len ( items ) == 1 && items [ 0 ] . Value == "" {
// Explicitly remove all -backend-config options.
// We do this by setting an empty but non-nil ConfigOverrides.
return configs . SynthBody ( "-backend-config=''" , synthVals ) , diags
}
command: "terraform init" can partially initialize for 0.12upgrade
There are a few constructs from 0.11 and prior that cause 0.12 parsing to
fail altogether, which previously created a chicken/egg problem because
we need to install the providers in order to run "terraform 0.12upgrade"
and thus fix the problem.
This changes "terraform init" to use the new "early configuration" loader
for module and provider installation. This is built on the more permissive
parser in the terraform-config-inspect package, and so it allows us to
read out the top-level blocks from the configuration while accepting
legacy HCL syntax.
In the long run this will let us do version compatibility detection before
attempting a "real" config load, giving us better error messages for any
future syntax additions, but in the short term the key thing is that it
allows us to install the dependencies even if the configuration isn't
fully valid.
Because backend init still requires full configuration, this introduces a
new mode of terraform init where it detects heuristically if it seems like
we need to do a configuration upgrade and does a partial init if so,
before finally directing the user to run "terraform 0.12upgrade" before
running any other commands.
The heuristic here is based on two assumptions:
- If the "early" loader finds no errors but the normal loader does, the
configuration is likely to be valid for Terraform 0.11 but not 0.12.
- If there's already a version constraint in the configuration that
excludes Terraform versions prior to v0.12 then the configuration is
probably _already_ upgraded and so it's just a normal syntax error,
even if the early loader didn't detect it.
Once the upgrade process is removed in 0.13.0 (users will be required to
go stepwise 0.11 -> 0.12 -> 0.13 to upgrade after that), some of this can
be simplified to remove that special mode, but the idea of doing the
dependency version checks against the liberal parser will remain valuable
to increase our chances of reporting version-based incompatibilities
rather than syntax errors as we add new features in future.
2019-01-14 20:11:00 +01:00
for _ , item := range items {
eq := strings . Index ( item . Value , "=" )
if eq == - 1 {
// The value is interpreted as a filename.
newBody , fileDiags := c . loadHCLFile ( item . Value )
diags = diags . Append ( fileDiags )
flushVals ( ) // deal with any accumulated individual values first
mergeBody ( newBody )
} else {
name := item . Value [ : eq ]
rawValue := item . Value [ eq + 1 : ]
attrS := schema . Attributes [ name ]
if attrS == nil {
diags = diags . Append ( tfdiags . Sourceless (
tfdiags . Error ,
"Invalid backend configuration argument" ,
fmt . Sprintf ( "The backend configuration argument %q given on the command line is not expected for the selected backend type." , name ) ,
) )
continue
}
value , valueDiags := configValueFromCLI ( item . String ( ) , rawValue , attrS . Type )
diags = diags . Append ( valueDiags )
if valueDiags . HasErrors ( ) {
continue
}
synthVals [ name ] = value
}
}
flushVals ( )
return ret , diags
2017-05-03 17:02:47 +02:00
}
2017-09-26 03:09:43 +02:00
func ( c * InitCommand ) AutocompleteArgs ( ) complete . Predictor {
return complete . PredictDirs ( "" )
}
func ( c * InitCommand ) AutocompleteFlags ( ) complete . Flags {
return complete . Flags {
"-backend" : completePredictBoolean ,
"-backend-config" : complete . PredictFiles ( "*.tfvars" ) , // can also be key=value, but we can't "predict" that
"-force-copy" : complete . PredictNothing ,
"-from-module" : completePredictModuleSource ,
"-get" : completePredictBoolean ,
"-get-plugins" : completePredictBoolean ,
"-input" : completePredictBoolean ,
"-lock" : completePredictBoolean ,
"-lock-timeout" : complete . PredictAnything ,
"-no-color" : complete . PredictNothing ,
"-plugin-dir" : complete . PredictDirs ( "" ) ,
"-reconfigure" : complete . PredictNothing ,
"-upgrade" : completePredictBoolean ,
"-verify-plugins" : completePredictBoolean ,
}
}
2014-09-27 01:03:39 +02:00
func ( c * InitCommand ) Help ( ) string {
helpText := `
2017-06-02 22:13:11 +02:00
Usage : terraform init [ options ] [ DIR ]
2017-01-19 05:50:45 +01:00
2017-06-03 01:59:52 +02:00
Initialize a new or existing Terraform working directory by creating
2017-01-19 05:50:45 +01:00
initial files , loading any remote state , downloading modules , etc .
This is the first command that should be run for any new or existing
Terraform configuration per machine . This sets up all the local data
2017-04-26 16:10:04 +02:00
necessary to run Terraform that is typically not committed to version
2017-01-19 05:50:45 +01:00
control .
This command is always safe to run multiple times . Though subsequent runs
2017-06-03 01:59:52 +02:00
may give errors , this command will never delete your configuration or
state . Even so , if you have important information , please back it up prior
to running this command , just in case .
2014-09-27 01:03:39 +02:00
2017-01-19 05:50:45 +01:00
If no arguments are given , the configuration in this working directory
is initialized .
2014-09-27 01:03:39 +02:00
2014-10-01 01:05:40 +02:00
Options :
2017-06-03 01:59:52 +02:00
- backend = true Configure the backend for this configuration .
2014-12-05 04:06:47 +01:00
2017-03-17 07:27:05 +01:00
- backend - config = path This can be either a path to an HCL file with key / value
assignments ( same format as terraform . tfvars ) or a
' key = value ' format . This is merged with what is in the
configuration file . This can be specified multiple
times . The backend type must be in the configuration
itself .
2014-10-01 01:05:40 +02:00
2017-04-20 23:26:50 +02:00
- force - copy Suppress prompts about copying state data . This is
equivalent to providing a "yes" to all confirmation
prompts .
2017-07-29 00:23:29 +02:00
- from - module = SOURCE Copy the contents of the given module into the target
directory before initialization .
2017-01-19 05:50:45 +01:00
- get = true Download any modules for this configuration .
2017-05-03 17:02:47 +02:00
- get - plugins = true Download any missing plugins for this configuration .
2017-01-19 05:50:45 +01:00
- input = true Ask for input if necessary . If false , will error if
input was required .
2017-04-01 22:19:59 +02:00
- lock = true Lock the state file when locking is supported .
- lock - timeout = 0 s Duration to retry a state lock .
2017-01-19 05:50:45 +01:00
- no - color If specified , output won ' t contain any color .
2015-06-22 14:14:01 +02:00
2017-06-15 21:23:16 +02:00
- plugin - dir Directory containing plugin binaries . This overrides all
default search paths for plugins , and prevents the
automatic installation of plugins . This flag can be used
multiple times .
2017-06-20 17:46:33 +02:00
- reconfigure Reconfigure the backend , ignoring any saved
configuration .
2017-06-12 23:14:40 +02:00
- upgrade = false If installing modules ( - get ) or plugins ( - get - plugins ) ,
ignore previously - downloaded objects and install the
latest version allowed within configured constraints .
2017-06-20 17:46:33 +02:00
- verify - plugins = true Verify the authenticity and integrity of automatically
downloaded plugins .
2014-09-27 01:03:39 +02:00
`
return strings . TrimSpace ( helpText )
}
func ( c * InitCommand ) Synopsis ( ) string {
2017-07-29 00:23:29 +02:00
return "Initialize a Terraform working directory"
2014-09-27 01:03:39 +02:00
}
2017-01-19 05:50:45 +01:00
2017-10-05 21:00:45 +02:00
const errInitConfigError = `
There are some problems with the configuration , described below .
The Terraform configuration must be valid before initialization so that
Terraform can determine which modules and providers need to be installed .
`
2017-01-19 05:50:45 +01:00
const errInitCopyNotEmpty = `
2017-07-29 00:23:29 +02:00
The working directory already contains files . The - from - module option requires
an empty directory into which a copy of the referenced module will be placed .
2017-01-19 05:50:45 +01:00
2017-07-29 00:23:29 +02:00
To initialize the configuration already in this working directory , omit the
- from - module option .
2017-01-19 05:50:45 +01:00
`
const outputInitEmpty = `
[ reset ] [ bold ] Terraform initialized in an empty directory ! [ reset ]
The directory has no Terraform configuration files . You may begin working
with Terraform immediately by creating Terraform configuration files .
`
const outputInitSuccess = `
[ reset ] [ bold ] [ green ] Terraform has been successfully initialized ! [ reset ] [ green ]
2017-09-09 02:14:37 +02:00
`
2017-01-19 05:50:45 +01:00
2017-09-09 02:14:37 +02:00
const outputInitSuccessCLI = ` [ reset ] [ green ]
2017-01-19 05:50:45 +01:00
You may now begin working with Terraform . Try running "terraform plan" to see
any changes that are required for your infrastructure . All Terraform commands
should now work .
If you ever set or change modules or backend configuration for Terraform ,
2017-06-03 01:59:52 +02:00
rerun this command to reinitialize your working directory . If you forget , other
2017-01-19 05:50:45 +01:00
commands will detect it and remind you to do so if necessary .
`
2017-06-02 02:57:43 +02:00
command: "terraform init" can partially initialize for 0.12upgrade
There are a few constructs from 0.11 and prior that cause 0.12 parsing to
fail altogether, which previously created a chicken/egg problem because
we need to install the providers in order to run "terraform 0.12upgrade"
and thus fix the problem.
This changes "terraform init" to use the new "early configuration" loader
for module and provider installation. This is built on the more permissive
parser in the terraform-config-inspect package, and so it allows us to
read out the top-level blocks from the configuration while accepting
legacy HCL syntax.
In the long run this will let us do version compatibility detection before
attempting a "real" config load, giving us better error messages for any
future syntax additions, but in the short term the key thing is that it
allows us to install the dependencies even if the configuration isn't
fully valid.
Because backend init still requires full configuration, this introduces a
new mode of terraform init where it detects heuristically if it seems like
we need to do a configuration upgrade and does a partial init if so,
before finally directing the user to run "terraform 0.12upgrade" before
running any other commands.
The heuristic here is based on two assumptions:
- If the "early" loader finds no errors but the normal loader does, the
configuration is likely to be valid for Terraform 0.11 but not 0.12.
- If there's already a version constraint in the configuration that
excludes Terraform versions prior to v0.12 then the configuration is
probably _already_ upgraded and so it's just a normal syntax error,
even if the early loader didn't detect it.
Once the upgrade process is removed in 0.13.0 (users will be required to
go stepwise 0.11 -> 0.12 -> 0.13 to upgrade after that), some of this can
be simplified to remove that special mode, but the idea of doing the
dependency version checks against the liberal parser will remain valuable
to increase our chances of reporting version-based incompatibilities
rather than syntax errors as we add new features in future.
2019-01-14 20:11:00 +01:00
const outputInitSuccessConfigUpgrade = `
[ reset ] [ bold ] Terraform has initialized , but configuration upgrades may be needed . [ reset ]
Terraform found syntax errors in the configuration that prevented full
initialization . If you ' ve recently upgraded to Terraform v0 .12 , this may be
because your configuration uses syntax constructs that are no longer valid ,
and so must be updated before full initialization is possible .
Run terraform init for this configuration at a shell prompt for more information
on how to update it for Terraform v0 .12 compatibility .
`
const outputInitSuccessConfigUpgradeCLI = ` [ reset ] [ green ]
[ reset ] [ bold ] Terraform has initialized , but configuration upgrades may be needed . [ reset ]
Terraform found syntax errors in the configuration that prevented full
initialization . If you ' ve recently upgraded to Terraform v0 .12 , this may be
because your configuration uses syntax constructs that are no longer valid ,
and so must be updated before full initialization is possible .
Terraform has installed the required providers to support the configuration
upgrade process . To begin upgrading your configuration , run the following :
terraform 0.12 upgrade
To see the full set of errors that led to this message , run :
terraform validate
`
2017-06-02 02:57:43 +02:00
const outputInitProvidersUnconstrained = `
The following providers do not have any version constraints in configuration ,
so the latest version was installed .
To prevent automatic upgrades to new major versions that may contain breaking
changes , it is recommended to add version = "..." constraints to the
corresponding provider blocks in configuration , with the constraint strings
suggested below .
`
2017-06-09 18:42:45 +02:00
2019-03-27 19:20:15 +01:00
const errDiscoveryServiceUnreachable = `
[ reset ] [ bold ] [ red ] Registry service unreachable . [ reset ] [ red ]
This may indicate a network issue , or an issue with the requested Terraform Registry .
`
2017-06-09 18:42:45 +02:00
const errProviderNotFound = `
2017-06-20 03:49:37 +02:00
[ reset ] [ bold ] [ red ] Provider % [ 1 ] q not available for installation . [ reset ] [ red ]
2018-11-09 01:12:11 +01:00
A provider named % [ 1 ] q could not be found in the Terraform Registry .
2017-06-20 03:49:37 +02:00
This may result from mistyping the provider name , or the given provider may
be a third - party provider that cannot be installed automatically .
In the latter case , the plugin must be installed manually by locating and
downloading a suitable distribution package and placing the plugin ' s executable
file in the following directory :
% [ 2 ] s
Terraform detects necessary plugins by inspecting the configuration and state .
To view the provider versions requested by each module , run
"terraform providers" .
`
const errProviderVersionsUnsuitable = `
[ reset ] [ bold ] [ red ] No provider % [ 1 ] q plugins meet the constraint % [ 2 ] q . [ reset ] [ red ]
The version constraint is derived from the "version" argument within the
provider % [ 1 ] q block in configuration . Child modules may also apply
provider version constraints . To view the provider versions requested by each
module in the current configuration , run "terraform providers" .
To proceed , the version constraints for this provider must be relaxed by
either adjusting or removing the "version" argument in the provider blocks
throughout the configuration .
`
const errProviderIncompatible = `
[ reset ] [ bold ] [ red ] No available provider % [ 1 ] q plugins are compatible with this Terraform version . [ reset ] [ red ]
From time to time , new Terraform major releases can change the requirements for
plugins such that older plugins become incompatible .
Terraform checked all of the plugin versions matching the given constraint :
% [ 2 ] s
Unfortunately , none of the suitable versions are compatible with this version
of Terraform . If you have recently upgraded Terraform , it may be necessary to
move to a newer major release of this provider . Alternatively , if you are
attempting to upgrade the provider to a new major version you may need to
also upgrade Terraform to support the new version .
Consult the documentation for this provider for more information on
compatibility between provider versions and Terraform versions .
`
const errProviderInstallError = `
[ reset ] [ bold ] [ red ] Error installing provider % [ 1 ] q : % [ 2 ] s . [ reset ] [ red ]
Terraform analyses the configuration and state and automatically downloads
plugins for the providers used . However , when attempting to download this
2019-03-21 20:38:51 +01:00
plugin an unexpected error occurred .
2017-06-20 03:49:37 +02:00
This may be caused if for some reason Terraform is unable to reach the
plugin repository . The repository may be unreachable if access is blocked
by a firewall .
If automatic installation is not possible or desirable in your environment ,
you may alternatively manually install plugins by downloading a suitable
distribution package and placing the plugin ' s executable file in the
following directory :
% [ 3 ] s
`
const errMissingProvidersNoInstall = `
[ reset ] [ bold ] [ red ] Missing required providers . [ reset ] [ red ]
2017-06-09 18:42:45 +02:00
2017-06-20 03:49:37 +02:00
The following provider constraints are not met by the currently - installed
provider plugins :
2017-06-09 18:42:45 +02:00
2017-06-20 03:49:37 +02:00
% [ 1 ] s
Terraform can automatically download and install plugins to meet the given
constraints , but this step was skipped due to the use of - get - plugins = false
and / or - plugin - dir on the command line .
2017-06-09 18:42:45 +02:00
2017-06-20 03:49:37 +02:00
If automatic installation is not possible or desirable in your environment ,
you may manually install plugins by downloading a suitable distribution package
and placing the plugin ' s executable file in one of the directories given in
by - plugin - dir on the command line , or in the following directory if custom
plugin directories are not set :
% [ 2 ] s
2017-06-09 18:42:45 +02:00
`
2019-03-21 20:38:51 +01:00
const errChecksumVerification = `
[ reset ] [ bold ] [ red ] Error verifying checksum for provider % [ 1 ] q [ reset ] [ red ]
The checksum for provider distribution from the Terraform Registry
did not match the source . This may mean that the distributed files
were changed after this version was released to the Registry .
`
const errSignatureVerification = `
[ reset ] [ bold ] [ red ] Error verifying GPG signature for provider % [ 1 ] q [ reset ] [ red ]
Terraform was unable to verify the GPG signature of the downloaded provider
files using the keys downloaded from the Terraform Registry . This may mean that
the publisher of the provider removed the key it was signed with , or that the
distributed files were changed after this version was released .
`