terraform/main.go

434 lines
13 KiB
Go
Raw Normal View History

2014-05-24 21:04:43 +02:00
package main
import (
command: Unmanaged providers This adds supports for "unmanaged" providers, or providers with process lifecycles not controlled by Terraform. These providers are assumed to be started before Terraform is launched, and are assumed to shut themselves down after Terraform has finished running. To do this, we must update the go-plugin dependency to v1.3.0, which added support for the "test mode" plugin serving that powers all this. As a side-effect of not needing to manage the process lifecycle anymore, Terraform also no longer needs to worry about the provider's binary, as it won't be used for anything anymore. Because of this, we can disable the init behavior that concerns itself with downloading that provider's binary, checking its version, and otherwise managing the binary. This is all managed on a per-provider basis, so managed providers that Terraform downloads, starts, and stops can be used in the same commands as unmanaged providers. The TF_REATTACH_PROVIDERS environment variable is added, and is a JSON encoding of the provider's address to the information we need to connect to it. This change enables two benefits: first, delve and other debuggers can now be attached to provider server processes, and Terraform can connect. This allows for attaching debuggers to provider processes, which before was difficult to impossible. Second, it allows the SDK test framework to host the provider in the same process as the test driver, while running a production Terraform binary against the provider. This allows for Go's built-in race detector and test coverage tooling to work as expected in provider tests. Unmanaged providers are expected to work in the exact same way as managed providers, with one caveat: Terraform kills provider processes and restarts them once per graph walk, meaning multiple times during most Terraform CLI commands. As unmanaged providers can't be killed by Terraform, and have no visibility into graph walks, unmanaged providers are likely to have differences in how their global mutable state behaves when compared to managed providers. Namely, unmanaged providers are likely to retain global state when managed providers would have reset it. Developers relying on global state should be aware of this.
2020-05-27 02:48:57 +02:00
"encoding/json"
2014-05-24 21:04:43 +02:00
"fmt"
"io/ioutil"
"log"
command: Unmanaged providers This adds supports for "unmanaged" providers, or providers with process lifecycles not controlled by Terraform. These providers are assumed to be started before Terraform is launched, and are assumed to shut themselves down after Terraform has finished running. To do this, we must update the go-plugin dependency to v1.3.0, which added support for the "test mode" plugin serving that powers all this. As a side-effect of not needing to manage the process lifecycle anymore, Terraform also no longer needs to worry about the provider's binary, as it won't be used for anything anymore. Because of this, we can disable the init behavior that concerns itself with downloading that provider's binary, checking its version, and otherwise managing the binary. This is all managed on a per-provider basis, so managed providers that Terraform downloads, starts, and stops can be used in the same commands as unmanaged providers. The TF_REATTACH_PROVIDERS environment variable is added, and is a JSON encoding of the provider's address to the information we need to connect to it. This change enables two benefits: first, delve and other debuggers can now be attached to provider server processes, and Terraform can connect. This allows for attaching debuggers to provider processes, which before was difficult to impossible. Second, it allows the SDK test framework to host the provider in the same process as the test driver, while running a production Terraform binary against the provider. This allows for Go's built-in race detector and test coverage tooling to work as expected in provider tests. Unmanaged providers are expected to work in the exact same way as managed providers, with one caveat: Terraform kills provider processes and restarts them once per graph walk, meaning multiple times during most Terraform CLI commands. As unmanaged providers can't be killed by Terraform, and have no visibility into graph walks, unmanaged providers are likely to have differences in how their global mutable state behaves when compared to managed providers. Namely, unmanaged providers are likely to retain global state when managed providers would have reset it. Developers relying on global state should be aware of this.
2020-05-27 02:48:57 +02:00
"net"
2014-05-24 21:04:43 +02:00
"os"
"path/filepath"
"runtime"
2017-02-14 00:12:29 +01:00
"strings"
2014-05-24 21:04:43 +02:00
"github.com/hashicorp/go-plugin"
"github.com/hashicorp/terraform-svchost/disco"
command: Unmanaged providers This adds supports for "unmanaged" providers, or providers with process lifecycles not controlled by Terraform. These providers are assumed to be started before Terraform is launched, and are assumed to shut themselves down after Terraform has finished running. To do this, we must update the go-plugin dependency to v1.3.0, which added support for the "test mode" plugin serving that powers all this. As a side-effect of not needing to manage the process lifecycle anymore, Terraform also no longer needs to worry about the provider's binary, as it won't be used for anything anymore. Because of this, we can disable the init behavior that concerns itself with downloading that provider's binary, checking its version, and otherwise managing the binary. This is all managed on a per-provider basis, so managed providers that Terraform downloads, starts, and stops can be used in the same commands as unmanaged providers. The TF_REATTACH_PROVIDERS environment variable is added, and is a JSON encoding of the provider's address to the information we need to connect to it. This change enables two benefits: first, delve and other debuggers can now be attached to provider server processes, and Terraform can connect. This allows for attaching debuggers to provider processes, which before was difficult to impossible. Second, it allows the SDK test framework to host the provider in the same process as the test driver, while running a production Terraform binary against the provider. This allows for Go's built-in race detector and test coverage tooling to work as expected in provider tests. Unmanaged providers are expected to work in the exact same way as managed providers, with one caveat: Terraform kills provider processes and restarts them once per graph walk, meaning multiple times during most Terraform CLI commands. As unmanaged providers can't be killed by Terraform, and have no visibility into graph walks, unmanaged providers are likely to have differences in how their global mutable state behaves when compared to managed providers. Namely, unmanaged providers are likely to retain global state when managed providers would have reset it. Developers relying on global state should be aware of this.
2020-05-27 02:48:57 +02:00
"github.com/hashicorp/terraform/addrs"
"github.com/hashicorp/terraform/command/cliconfig"
"github.com/hashicorp/terraform/command/format"
"github.com/hashicorp/terraform/httpclient"
"github.com/hashicorp/terraform/internal/logging"
"github.com/hashicorp/terraform/version"
"github.com/mattn/go-shellwords"
2014-05-24 21:04:43 +02:00
"github.com/mitchellh/cli"
"github.com/mitchellh/colorstring"
2014-05-31 01:07:26 +02:00
"github.com/mitchellh/panicwrap"
backendInit "github.com/hashicorp/terraform/backend/init"
2014-05-24 21:04:43 +02:00
)
const (
// EnvCLI is the environment variable name to set additional CLI args.
EnvCLI = "TF_CLI_ARGS"
// The parent process will create a file to collect crash logs
envTmpLogPath = "TF_TEMP_LOG_PATH"
)
2014-05-24 21:04:43 +02:00
func main() {
os.Exit(realMain())
}
func realMain() int {
2014-05-31 01:07:26 +02:00
var wrapConfig panicwrap.WrapConfig
// don't re-exec terraform as a child process for easier debugging
if os.Getenv("TF_FORK") == "0" {
return wrappedMain()
}
2014-05-31 01:07:26 +02:00
if !panicwrap.Wrapped(&wrapConfig) {
// We always send logs to a temporary file that we use in case
// there is a panic. Otherwise, we delete it.
logTempFile, err := ioutil.TempFile("", "terraform-log")
if err != nil {
fmt.Fprintf(os.Stderr, "Couldn't setup logging tempfile: %s", err)
return 1
}
// Now that we have the file, close it and leave it for the wrapped
// process to write to.
logTempFile.Close()
2014-05-31 01:07:26 +02:00
defer os.Remove(logTempFile.Name())
// store the path in the environment for the wrapped executable
os.Setenv(envTmpLogPath, logTempFile.Name())
2014-05-31 01:07:26 +02:00
// Create the configuration for panicwrap and wrap our executable
wrapConfig.Handler = logging.PanicHandler(logTempFile.Name())
wrapConfig.IgnoreSignals = ignoreSignals
wrapConfig.ForwardSignals = forwardSignals
2014-05-31 01:07:26 +02:00
exitStatus, err := panicwrap.Wrap(&wrapConfig)
if err != nil {
fmt.Fprintf(os.Stderr, "Couldn't start Terraform: %s", err)
return 1
}
return exitStatus
2014-05-31 01:07:26 +02:00
}
// Call the real main
return wrappedMain()
}
func init() {
Ui = &cli.BasicUi{
Writer: os.Stdout,
ErrorWriter: os.Stderr,
Reader: os.Stdin,
}
}
2014-05-31 01:07:26 +02:00
func wrappedMain() int {
var err error
tmpLogPath := os.Getenv(envTmpLogPath)
if tmpLogPath != "" {
f, err := os.OpenFile(tmpLogPath, os.O_RDWR|os.O_APPEND, 0666)
if err == nil {
defer os.Remove(f.Name())
defer f.Close()
log.Printf("[DEBUG] Adding temp file log sink: %s", f.Name())
logging.RegisterSink(f)
} else {
log.Printf("[ERROR] Could not open temp log file: %v", err)
}
}
log.Printf(
"[INFO] Terraform version: %s %s %s",
Version, VersionPrerelease, GitCommit)
log.Printf("[INFO] Go runtime version: %s", runtime.Version())
2016-08-17 16:49:54 +02:00
log.Printf("[INFO] CLI args: %#v", os.Args)
2014-05-24 21:04:43 +02:00
main: new global option -chdir This new option is intended to address the previous inconsistencies where some older subcommands supported partially changing the target directory (where Terraform would use the new directory inconsistently) where newer commands did not support that override at all. Instead, now Terraform will accept a -chdir command at the start of the command line (before the subcommand) and will interpret it as a request to direct all actions that would normally be taken in the current working directory into the target directory instead. This is similar to options offered by some other similar tools, such as the -C option in "make". The new option is only accepted at the start of the command line (before the subcommand) as a way to reflect that it is a global command (not specific to a particular subcommand) and that it takes effect _before_ executing the subcommand. This also means it'll be forced to appear before any other command-specific arguments that take file paths, which hopefully communicates that those other arguments are interpreted relative to the overridden path. As a measure of pragmatism for existing uses, the path.cwd object in the Terraform language will continue to return the _original_ working directory (ignoring -chdir), in case that is important in some exceptional workflows. The path.root object gives the root module directory, which will always match the overriden working directory unless the user simultaneously uses one of the legacy directory override arguments, which is not a pattern we intend to support in the long run. As a first step down the deprecation path, this commit adjusts the documentation to de-emphasize the inconsistent old command line arguments, including specific guidance on what to use instead for the main three workflow commands, but all of those options remain supported in the same way as they were before. In a later commit we'll make those arguments produce a visible deprecation warning in Terraform's output, and then in an even later commit we'll remove them entirely so that -chdir is the single supported way to run Terraform from a directory other than the one containing the root module configuration.
2020-09-02 00:45:12 +02:00
// NOTE: We're intentionally calling LoadConfig _before_ handling a possible
// -chdir=... option on the command line, so that a possible relative
// path in the TERRAFORM_CONFIG_FILE environment variable (though probably
// ill-advised) will be resolved relative to the true working directory,
// not the overridden one.
config, diags := cliconfig.LoadConfig()
if len(diags) > 0 {
// Since we haven't instantiated a command.Meta yet, we need to do
// some things manually here and use some "safe" defaults for things
// that command.Meta could otherwise figure out in smarter ways.
Ui.Error("There are some problems with the CLI configuration:")
for _, diag := range diags {
earlyColor := &colorstring.Colorize{
Colors: colorstring.DefaultColors,
Disable: true, // Disable color to be conservative until we know better
Reset: true,
}
// We don't currently have access to the source code cache for
// the parser used to load the CLI config, so we can't show
// source code snippets in early diagnostics.
Ui.Error(format.Diagnostic(diag, nil, earlyColor, 78))
}
if diags.HasErrors() {
Ui.Error("As a result of the above problems, Terraform may not behave as intended.\n\n")
// We continue to run anyway, since Terraform has reasonable defaults.
}
}
// Get any configured credentials from the config and initialize
// a service discovery object. The slightly awkward predeclaration of
// disco is required to allow us to pass untyped nil as the creds source
// when creating the source fails. Otherwise we pass a typed nil which
// breaks the nil checks in the disco object
var services *disco.Disco
credsSrc, err := credentialsSource(config)
if err == nil {
services = disco.NewWithCredentialsSource(credsSrc)
} else {
// Most commands don't actually need credentials, and most situations
// that would get us here would already have been reported by the config
// loading above, so we'll just log this one as an aid to debugging
// in the unlikely event that it _does_ arise.
log.Printf("[WARN] Cannot initialize remote host credentials manager: %s", err)
// passing (untyped) nil as the creds source is okay because the disco
// object checks that and just acts as though no credentials are present.
services = disco.NewWithCredentialsSource(nil)
}
services.SetUserAgent(httpclient.TerraformUserAgent(version.String()))
providerSrc, diags := providerSource(config.ProviderInstallation, services)
if len(diags) > 0 {
Ui.Error("There are some problems with the provider_installation configuration:")
for _, diag := range diags {
earlyColor := &colorstring.Colorize{
Colors: colorstring.DefaultColors,
Disable: true, // Disable color to be conservative until we know better
Reset: true,
}
Ui.Error(format.Diagnostic(diag, nil, earlyColor, 78))
}
if diags.HasErrors() {
Ui.Error("As a result of the above problems, Terraform's provider installer may not behave as intended.\n\n")
// We continue to run anyway, because most commands don't do provider installation.
}
}
providerDevOverrides := providerDevOverrides(config.ProviderInstallation)
command: Unmanaged providers This adds supports for "unmanaged" providers, or providers with process lifecycles not controlled by Terraform. These providers are assumed to be started before Terraform is launched, and are assumed to shut themselves down after Terraform has finished running. To do this, we must update the go-plugin dependency to v1.3.0, which added support for the "test mode" plugin serving that powers all this. As a side-effect of not needing to manage the process lifecycle anymore, Terraform also no longer needs to worry about the provider's binary, as it won't be used for anything anymore. Because of this, we can disable the init behavior that concerns itself with downloading that provider's binary, checking its version, and otherwise managing the binary. This is all managed on a per-provider basis, so managed providers that Terraform downloads, starts, and stops can be used in the same commands as unmanaged providers. The TF_REATTACH_PROVIDERS environment variable is added, and is a JSON encoding of the provider's address to the information we need to connect to it. This change enables two benefits: first, delve and other debuggers can now be attached to provider server processes, and Terraform can connect. This allows for attaching debuggers to provider processes, which before was difficult to impossible. Second, it allows the SDK test framework to host the provider in the same process as the test driver, while running a production Terraform binary against the provider. This allows for Go's built-in race detector and test coverage tooling to work as expected in provider tests. Unmanaged providers are expected to work in the exact same way as managed providers, with one caveat: Terraform kills provider processes and restarts them once per graph walk, meaning multiple times during most Terraform CLI commands. As unmanaged providers can't be killed by Terraform, and have no visibility into graph walks, unmanaged providers are likely to have differences in how their global mutable state behaves when compared to managed providers. Namely, unmanaged providers are likely to retain global state when managed providers would have reset it. Developers relying on global state should be aware of this.
2020-05-27 02:48:57 +02:00
// The user can declare that certain providers are being managed on
// Terraform's behalf using this environment variable. Thsi is used
// primarily by the SDK's acceptance testing framework.
unmanagedProviders, err := parseReattachProviders(os.Getenv("TF_REATTACH_PROVIDERS"))
if err != nil {
Ui.Error(err.Error())
return 1
}
// Initialize the backends.
backendInit.Init(services)
main: new global option -chdir This new option is intended to address the previous inconsistencies where some older subcommands supported partially changing the target directory (where Terraform would use the new directory inconsistently) where newer commands did not support that override at all. Instead, now Terraform will accept a -chdir command at the start of the command line (before the subcommand) and will interpret it as a request to direct all actions that would normally be taken in the current working directory into the target directory instead. This is similar to options offered by some other similar tools, such as the -C option in "make". The new option is only accepted at the start of the command line (before the subcommand) as a way to reflect that it is a global command (not specific to a particular subcommand) and that it takes effect _before_ executing the subcommand. This also means it'll be forced to appear before any other command-specific arguments that take file paths, which hopefully communicates that those other arguments are interpreted relative to the overridden path. As a measure of pragmatism for existing uses, the path.cwd object in the Terraform language will continue to return the _original_ working directory (ignoring -chdir), in case that is important in some exceptional workflows. The path.root object gives the root module directory, which will always match the overriden working directory unless the user simultaneously uses one of the legacy directory override arguments, which is not a pattern we intend to support in the long run. As a first step down the deprecation path, this commit adjusts the documentation to de-emphasize the inconsistent old command line arguments, including specific guidance on what to use instead for the main three workflow commands, but all of those options remain supported in the same way as they were before. In a later commit we'll make those arguments produce a visible deprecation warning in Terraform's output, and then in an even later commit we'll remove them entirely so that -chdir is the single supported way to run Terraform from a directory other than the one containing the root module configuration.
2020-09-02 00:45:12 +02:00
// Get the command line args.
binName := filepath.Base(os.Args[0])
args := os.Args[1:]
originalWd, err := os.Getwd()
if err != nil {
// It would be very strange to end up here
Ui.Error(fmt.Sprintf("Failed to determine current working directory: %s", err))
return 1
}
// The arguments can begin with a -chdir option to ask Terraform to switch
// to a different working directory for the rest of its work. If that
// option is present then extractChdirOption returns a trimmed args with that option removed.
overrideWd, args, err := extractChdirOption(args)
if err != nil {
Ui.Error(fmt.Sprintf("Invalid -chdir option: %s", err))
return 1
}
if overrideWd != "" {
2020-10-13 20:22:25 +02:00
err := os.Chdir(overrideWd)
main: new global option -chdir This new option is intended to address the previous inconsistencies where some older subcommands supported partially changing the target directory (where Terraform would use the new directory inconsistently) where newer commands did not support that override at all. Instead, now Terraform will accept a -chdir command at the start of the command line (before the subcommand) and will interpret it as a request to direct all actions that would normally be taken in the current working directory into the target directory instead. This is similar to options offered by some other similar tools, such as the -C option in "make". The new option is only accepted at the start of the command line (before the subcommand) as a way to reflect that it is a global command (not specific to a particular subcommand) and that it takes effect _before_ executing the subcommand. This also means it'll be forced to appear before any other command-specific arguments that take file paths, which hopefully communicates that those other arguments are interpreted relative to the overridden path. As a measure of pragmatism for existing uses, the path.cwd object in the Terraform language will continue to return the _original_ working directory (ignoring -chdir), in case that is important in some exceptional workflows. The path.root object gives the root module directory, which will always match the overriden working directory unless the user simultaneously uses one of the legacy directory override arguments, which is not a pattern we intend to support in the long run. As a first step down the deprecation path, this commit adjusts the documentation to de-emphasize the inconsistent old command line arguments, including specific guidance on what to use instead for the main three workflow commands, but all of those options remain supported in the same way as they were before. In a later commit we'll make those arguments produce a visible deprecation warning in Terraform's output, and then in an even later commit we'll remove them entirely so that -chdir is the single supported way to run Terraform from a directory other than the one containing the root module configuration.
2020-09-02 00:45:12 +02:00
if err != nil {
Ui.Error(fmt.Sprintf("Error handling -chdir option: %s", err))
return 1
}
}
// In tests, Commands may already be set to provide mock commands
if Commands == nil {
main: new global option -chdir This new option is intended to address the previous inconsistencies where some older subcommands supported partially changing the target directory (where Terraform would use the new directory inconsistently) where newer commands did not support that override at all. Instead, now Terraform will accept a -chdir command at the start of the command line (before the subcommand) and will interpret it as a request to direct all actions that would normally be taken in the current working directory into the target directory instead. This is similar to options offered by some other similar tools, such as the -C option in "make". The new option is only accepted at the start of the command line (before the subcommand) as a way to reflect that it is a global command (not specific to a particular subcommand) and that it takes effect _before_ executing the subcommand. This also means it'll be forced to appear before any other command-specific arguments that take file paths, which hopefully communicates that those other arguments are interpreted relative to the overridden path. As a measure of pragmatism for existing uses, the path.cwd object in the Terraform language will continue to return the _original_ working directory (ignoring -chdir), in case that is important in some exceptional workflows. The path.root object gives the root module directory, which will always match the overriden working directory unless the user simultaneously uses one of the legacy directory override arguments, which is not a pattern we intend to support in the long run. As a first step down the deprecation path, this commit adjusts the documentation to de-emphasize the inconsistent old command line arguments, including specific guidance on what to use instead for the main three workflow commands, but all of those options remain supported in the same way as they were before. In a later commit we'll make those arguments produce a visible deprecation warning in Terraform's output, and then in an even later commit we'll remove them entirely so that -chdir is the single supported way to run Terraform from a directory other than the one containing the root module configuration.
2020-09-02 00:45:12 +02:00
// Commands get to hold on to the original working directory here,
// in case they need to refer back to it for any special reason, though
// they should primarily be working with the override working directory
// that we've now switched to above.
initCommands(originalWd, config, services, providerSrc, providerDevOverrides, unmanagedProviders)
}
2014-10-13 23:05:29 +02:00
// Run checkpoint
go runCheckpoint(config)
2014-10-13 23:05:29 +02:00
// Make sure we clean up any managed plugins at the end of this
defer plugin.CleanupClients()
2017-02-14 00:12:29 +01:00
// Build the CLI so far, we do this so we can query the subcommand.
cliRunner := &cli.CLI{
Args: args,
Commands: Commands,
HelpFunc: helpFunc,
HelpWriter: os.Stdout,
}
2017-02-14 00:12:29 +01:00
// Prefix the args with any args from the EnvCLI
args, err = mergeEnvArgs(EnvCLI, cliRunner.Subcommand(), args)
2017-02-14 00:12:29 +01:00
if err != nil {
Ui.Error(err.Error())
return 1
}
2017-02-14 00:12:29 +01:00
// Prefix the args with any args from the EnvCLI targeting this command
suffix := strings.Replace(strings.Replace(
cliRunner.Subcommand(), "-", "_", -1), " ", "_", -1)
args, err = mergeEnvArgs(
fmt.Sprintf("%s_%s", EnvCLI, suffix), cliRunner.Subcommand(), args)
2017-02-14 00:12:29 +01:00
if err != nil {
Ui.Error(err.Error())
return 1
}
// We shortcut "--version" and "-v" to just show the version
2014-05-24 21:04:43 +02:00
for _, arg := range args {
2014-07-13 19:37:25 +02:00
if arg == "-v" || arg == "-version" || arg == "--version" {
2014-05-24 21:04:43 +02:00
newArgs := make([]string, len(args)+1)
newArgs[0] = "version"
copy(newArgs[1:], args)
args = newArgs
break
}
}
2017-02-14 00:12:29 +01:00
// Rebuild the CLI with any modified args.
log.Printf("[INFO] CLI command args: %#v", args)
2017-02-14 00:12:29 +01:00
cliRunner = &cli.CLI{
Name: binName,
2014-06-26 19:24:32 +02:00
Args: args,
Commands: Commands,
HelpFunc: helpFunc,
2014-06-26 19:24:32 +02:00
HelpWriter: os.Stdout,
Autocomplete: true,
AutocompleteInstall: "install-autocomplete",
AutocompleteUninstall: "uninstall-autocomplete",
2014-05-24 21:04:43 +02:00
}
2017-02-14 00:12:29 +01:00
exitCode, err := cliRunner.Run()
2014-05-24 21:04:43 +02:00
if err != nil {
Ui.Error(fmt.Sprintf("Error executing CLI: %s", err.Error()))
2014-05-24 21:04:43 +02:00
return 1
}
return exitCode
}
func mergeEnvArgs(envName string, cmd string, args []string) ([]string, error) {
2017-02-14 00:12:29 +01:00
v := os.Getenv(envName)
if v == "" {
return args, nil
}
log.Printf("[INFO] %s value: %q", envName, v)
extra, err := shellwords.Parse(v)
if err != nil {
return nil, fmt.Errorf(
"Error parsing extra CLI args from %s: %s",
envName, err)
}
// Find the command to look for in the args. If there is a space,
// we need to find the last part.
search := cmd
if idx := strings.LastIndex(search, " "); idx >= 0 {
search = cmd[idx+1:]
}
2017-02-14 00:12:29 +01:00
// Find the index to place the flags. We put them exactly
// after the first non-flag arg.
idx := -1
for i, v := range args {
if v == search {
2017-02-14 00:12:29 +01:00
idx = i
break
}
}
// idx points to the exact arg that isn't a flag. We increment
// by one so that all the copying below expects idx to be the
// insertion point.
idx++
// Copy the args
newArgs := make([]string, len(args)+len(extra))
copy(newArgs, args[:idx])
copy(newArgs[idx:], extra)
copy(newArgs[len(extra)+idx:], args[idx:])
return newArgs, nil
}
command: Unmanaged providers This adds supports for "unmanaged" providers, or providers with process lifecycles not controlled by Terraform. These providers are assumed to be started before Terraform is launched, and are assumed to shut themselves down after Terraform has finished running. To do this, we must update the go-plugin dependency to v1.3.0, which added support for the "test mode" plugin serving that powers all this. As a side-effect of not needing to manage the process lifecycle anymore, Terraform also no longer needs to worry about the provider's binary, as it won't be used for anything anymore. Because of this, we can disable the init behavior that concerns itself with downloading that provider's binary, checking its version, and otherwise managing the binary. This is all managed on a per-provider basis, so managed providers that Terraform downloads, starts, and stops can be used in the same commands as unmanaged providers. The TF_REATTACH_PROVIDERS environment variable is added, and is a JSON encoding of the provider's address to the information we need to connect to it. This change enables two benefits: first, delve and other debuggers can now be attached to provider server processes, and Terraform can connect. This allows for attaching debuggers to provider processes, which before was difficult to impossible. Second, it allows the SDK test framework to host the provider in the same process as the test driver, while running a production Terraform binary against the provider. This allows for Go's built-in race detector and test coverage tooling to work as expected in provider tests. Unmanaged providers are expected to work in the exact same way as managed providers, with one caveat: Terraform kills provider processes and restarts them once per graph walk, meaning multiple times during most Terraform CLI commands. As unmanaged providers can't be killed by Terraform, and have no visibility into graph walks, unmanaged providers are likely to have differences in how their global mutable state behaves when compared to managed providers. Namely, unmanaged providers are likely to retain global state when managed providers would have reset it. Developers relying on global state should be aware of this.
2020-05-27 02:48:57 +02:00
// parse information on reattaching to unmanaged providers out of a
// JSON-encoded environment variable.
func parseReattachProviders(in string) (map[addrs.Provider]*plugin.ReattachConfig, error) {
unmanagedProviders := map[addrs.Provider]*plugin.ReattachConfig{}
if in != "" {
type reattachConfig struct {
Protocol string
Addr struct {
Network string
String string
}
Pid int
Test bool
}
var m map[string]reattachConfig
err := json.Unmarshal([]byte(in), &m)
if err != nil {
return unmanagedProviders, fmt.Errorf("Invalid format for TF_REATTACH_PROVIDERS: %w", err)
}
for p, c := range m {
a, diags := addrs.ParseProviderSourceString(p)
if diags.HasErrors() {
return unmanagedProviders, fmt.Errorf("Error parsing %q as a provider address: %w", a, diags.Err())
}
var addr net.Addr
switch c.Addr.Network {
case "unix":
addr, err = net.ResolveUnixAddr("unix", c.Addr.String)
if err != nil {
return unmanagedProviders, fmt.Errorf("Invalid unix socket path %q for %q: %w", c.Addr.String, p, err)
}
case "tcp":
addr, err = net.ResolveTCPAddr("tcp", c.Addr.String)
if err != nil {
return unmanagedProviders, fmt.Errorf("Invalid TCP address %q for %q: %w", c.Addr.String, p, err)
}
default:
return unmanagedProviders, fmt.Errorf("Unknown address type %q for %q", c.Addr.Network, p)
}
unmanagedProviders[a] = &plugin.ReattachConfig{
Protocol: plugin.Protocol(c.Protocol),
Pid: c.Pid,
Test: c.Test,
Addr: addr,
}
}
}
return unmanagedProviders, nil
}
main: new global option -chdir This new option is intended to address the previous inconsistencies where some older subcommands supported partially changing the target directory (where Terraform would use the new directory inconsistently) where newer commands did not support that override at all. Instead, now Terraform will accept a -chdir command at the start of the command line (before the subcommand) and will interpret it as a request to direct all actions that would normally be taken in the current working directory into the target directory instead. This is similar to options offered by some other similar tools, such as the -C option in "make". The new option is only accepted at the start of the command line (before the subcommand) as a way to reflect that it is a global command (not specific to a particular subcommand) and that it takes effect _before_ executing the subcommand. This also means it'll be forced to appear before any other command-specific arguments that take file paths, which hopefully communicates that those other arguments are interpreted relative to the overridden path. As a measure of pragmatism for existing uses, the path.cwd object in the Terraform language will continue to return the _original_ working directory (ignoring -chdir), in case that is important in some exceptional workflows. The path.root object gives the root module directory, which will always match the overriden working directory unless the user simultaneously uses one of the legacy directory override arguments, which is not a pattern we intend to support in the long run. As a first step down the deprecation path, this commit adjusts the documentation to de-emphasize the inconsistent old command line arguments, including specific guidance on what to use instead for the main three workflow commands, but all of those options remain supported in the same way as they were before. In a later commit we'll make those arguments produce a visible deprecation warning in Terraform's output, and then in an even later commit we'll remove them entirely so that -chdir is the single supported way to run Terraform from a directory other than the one containing the root module configuration.
2020-09-02 00:45:12 +02:00
func extractChdirOption(args []string) (string, []string, error) {
if len(args) == 0 {
return "", args, nil
}
const argName = "-chdir"
const argPrefix = argName + "="
var argValue string
var argPos int
for i, arg := range args {
if !strings.HasPrefix(arg, "-") {
// Because the chdir option is a subcommand-agnostic one, we require
// it to appear before any subcommand argument, so if we find a
// non-option before we find -chdir then we are finished.
break
}
if arg == argName || arg == argPrefix {
return "", args, fmt.Errorf("must include an equals sign followed by a directory path, like -chdir=example")
}
if strings.HasPrefix(arg, argPrefix) {
argPos = i
argValue = arg[len(argPrefix):]
}
}
// When we fall out here, we'll have populated argValue with a non-empty
// string if the -chdir=... option was present and valid, or left it
// empty if it wasn't present.
if argValue == "" {
return "", args, nil
}
// If we did find the option then we'll need to produce a new args that
// doesn't include it anymore.
if argPos == 0 {
// Easy case: we can just slice off the front
return argValue, args[1:], nil
}
// Otherwise we need to construct a new array and copy to it.
newArgs := make([]string, len(args)-1)
copy(newArgs, args[:argPos])
copy(newArgs[argPos:], args[argPos+1:])
return argValue, newArgs, nil
}