DockerCLI/cmd/docker/docker.go

435 lines
13 KiB
Go
Raw Normal View History

package main
import (
"errors"
"fmt"
"os"
"os/exec"
"strings"
"syscall"
"github.com/docker/cli/cli"
pluginmanager "github.com/docker/cli/cli-plugins/manager"
"github.com/docker/cli/cli/command"
"github.com/docker/cli/cli/command/commands"
cliflags "github.com/docker/cli/cli/flags"
"github.com/docker/cli/cli/version"
"github.com/docker/docker/api/types/versions"
"github.com/docker/docker/client"
"github.com/sirupsen/logrus"
"github.com/spf13/cobra"
"github.com/spf13/pflag"
)
allow plugins to have argument which match a top-level flag. The issue with plugin options clashing with globals is that when cobra is parsing the command line and it comes across an argument which doesn't start with a `-` it (in the absence of plugins) distinguishes between "argument to current command" and "new subcommand" based on the list of registered sub commands. Plugins breaks that model. When presented with `docker -D plugin -c foo` cobra parses up to the `plugin`, sees it isn't a registered sub-command of the top-level docker (because it isn't, it's a plugin) so it accumulates it as an argument to the top-level `docker` command. Then it sees the `-c`, and thinks it is the global `-c` (for AKA `--context`) option and tries to treat it as that, which fails. In the specific case of the top-level `docker` subcommand we know that it has no arguments which aren't `--flags` (or `-f` short flags) and so anything which doesn't start with a `-` must either be a (known) subcommand or an attempt to execute a plugin. We could simply scan for and register all installed plugins at start of day, so that cobra can do the right thing, but we want to avoid that since it would involve executing each plugin to fetch the metadata, even if the command wasn't going to end up hitting a plugin. Instead we can parse the initial set of global arguments separately before hitting the main cobra `Execute` path, which works here exactly because we know that the top-level has no non-flag arguments. One slight wrinkle is that the top-level `PersistentPreRunE` is no longer called on the plugins path (since it no longer goes via `Execute`), so we arrange for the initialisation done there (which has to be done after global flags are parsed to handle e.g. `--config`) to happen explictly after the global flags are parsed. Rather than make `newDockerCommand` return the complicated set of results needed to make this happen, instead return a closure which achieves this. The new functionality is introduced via a common `TopLevelCommand` abstraction which lets us adjust the plugin entrypoint to use the same strategy for parsing the global arguments. This isn't strictly required (in this case the stuff in cobra's `Execute` works fine) but doing it this way avoids the possibility of subtle differences in behaviour. Fixes #1699, and also, as a side-effect, the first item in #1661. Signed-off-by: Ian Campbell <ijc@docker.com>
2019-03-06 05:29:42 -05:00
func newDockerCommand(dockerCli *command.DockerCli) *cli.TopLevelCommand {
var (
opts *cliflags.ClientOptions
flags *pflag.FlagSet
helpCmd *cobra.Command
)
cmd := &cobra.Command{
Use: "docker [OPTIONS] COMMAND [ARG...]",
Short: "A self-sufficient runtime for containers",
SilenceUsage: true,
SilenceErrors: true,
TraverseChildren: true,
RunE: func(cmd *cobra.Command, args []string) error {
if len(args) == 0 {
return command.ShowHelp(dockerCli.Err())(cmd, args)
}
allow plugins to have argument which match a top-level flag. The issue with plugin options clashing with globals is that when cobra is parsing the command line and it comes across an argument which doesn't start with a `-` it (in the absence of plugins) distinguishes between "argument to current command" and "new subcommand" based on the list of registered sub commands. Plugins breaks that model. When presented with `docker -D plugin -c foo` cobra parses up to the `plugin`, sees it isn't a registered sub-command of the top-level docker (because it isn't, it's a plugin) so it accumulates it as an argument to the top-level `docker` command. Then it sees the `-c`, and thinks it is the global `-c` (for AKA `--context`) option and tries to treat it as that, which fails. In the specific case of the top-level `docker` subcommand we know that it has no arguments which aren't `--flags` (or `-f` short flags) and so anything which doesn't start with a `-` must either be a (known) subcommand or an attempt to execute a plugin. We could simply scan for and register all installed plugins at start of day, so that cobra can do the right thing, but we want to avoid that since it would involve executing each plugin to fetch the metadata, even if the command wasn't going to end up hitting a plugin. Instead we can parse the initial set of global arguments separately before hitting the main cobra `Execute` path, which works here exactly because we know that the top-level has no non-flag arguments. One slight wrinkle is that the top-level `PersistentPreRunE` is no longer called on the plugins path (since it no longer goes via `Execute`), so we arrange for the initialisation done there (which has to be done after global flags are parsed to handle e.g. `--config`) to happen explictly after the global flags are parsed. Rather than make `newDockerCommand` return the complicated set of results needed to make this happen, instead return a closure which achieves this. The new functionality is introduced via a common `TopLevelCommand` abstraction which lets us adjust the plugin entrypoint to use the same strategy for parsing the global arguments. This isn't strictly required (in this case the stuff in cobra's `Execute` works fine) but doing it this way avoids the possibility of subtle differences in behaviour. Fixes #1699, and also, as a side-effect, the first item in #1661. Signed-off-by: Ian Campbell <ijc@docker.com>
2019-03-06 05:29:42 -05:00
return fmt.Errorf("docker: '%s' is not a docker command.\nSee 'docker --help'", args[0])
},
PersistentPreRunE: func(cmd *cobra.Command, args []string) error {
return isSupported(cmd, dockerCli)
},
Version: fmt.Sprintf("%s, build %s", version.Version, version.GitCommit),
DisableFlagsInUseLine: true,
}
opts, flags, helpCmd = cli.SetupRootCommand(cmd)
flags.BoolP("version", "v", false, "Print version information and quit")
allow plugins to have argument which match a top-level flag. The issue with plugin options clashing with globals is that when cobra is parsing the command line and it comes across an argument which doesn't start with a `-` it (in the absence of plugins) distinguishes between "argument to current command" and "new subcommand" based on the list of registered sub commands. Plugins breaks that model. When presented with `docker -D plugin -c foo` cobra parses up to the `plugin`, sees it isn't a registered sub-command of the top-level docker (because it isn't, it's a plugin) so it accumulates it as an argument to the top-level `docker` command. Then it sees the `-c`, and thinks it is the global `-c` (for AKA `--context`) option and tries to treat it as that, which fails. In the specific case of the top-level `docker` subcommand we know that it has no arguments which aren't `--flags` (or `-f` short flags) and so anything which doesn't start with a `-` must either be a (known) subcommand or an attempt to execute a plugin. We could simply scan for and register all installed plugins at start of day, so that cobra can do the right thing, but we want to avoid that since it would involve executing each plugin to fetch the metadata, even if the command wasn't going to end up hitting a plugin. Instead we can parse the initial set of global arguments separately before hitting the main cobra `Execute` path, which works here exactly because we know that the top-level has no non-flag arguments. One slight wrinkle is that the top-level `PersistentPreRunE` is no longer called on the plugins path (since it no longer goes via `Execute`), so we arrange for the initialisation done there (which has to be done after global flags are parsed to handle e.g. `--config`) to happen explictly after the global flags are parsed. Rather than make `newDockerCommand` return the complicated set of results needed to make this happen, instead return a closure which achieves this. The new functionality is introduced via a common `TopLevelCommand` abstraction which lets us adjust the plugin entrypoint to use the same strategy for parsing the global arguments. This isn't strictly required (in this case the stuff in cobra's `Execute` works fine) but doing it this way avoids the possibility of subtle differences in behaviour. Fixes #1699, and also, as a side-effect, the first item in #1661. Signed-off-by: Ian Campbell <ijc@docker.com>
2019-03-06 05:29:42 -05:00
setFlagErrorFunc(dockerCli, cmd)
allow plugins to have argument which match a top-level flag. The issue with plugin options clashing with globals is that when cobra is parsing the command line and it comes across an argument which doesn't start with a `-` it (in the absence of plugins) distinguishes between "argument to current command" and "new subcommand" based on the list of registered sub commands. Plugins breaks that model. When presented with `docker -D plugin -c foo` cobra parses up to the `plugin`, sees it isn't a registered sub-command of the top-level docker (because it isn't, it's a plugin) so it accumulates it as an argument to the top-level `docker` command. Then it sees the `-c`, and thinks it is the global `-c` (for AKA `--context`) option and tries to treat it as that, which fails. In the specific case of the top-level `docker` subcommand we know that it has no arguments which aren't `--flags` (or `-f` short flags) and so anything which doesn't start with a `-` must either be a (known) subcommand or an attempt to execute a plugin. We could simply scan for and register all installed plugins at start of day, so that cobra can do the right thing, but we want to avoid that since it would involve executing each plugin to fetch the metadata, even if the command wasn't going to end up hitting a plugin. Instead we can parse the initial set of global arguments separately before hitting the main cobra `Execute` path, which works here exactly because we know that the top-level has no non-flag arguments. One slight wrinkle is that the top-level `PersistentPreRunE` is no longer called on the plugins path (since it no longer goes via `Execute`), so we arrange for the initialisation done there (which has to be done after global flags are parsed to handle e.g. `--config`) to happen explictly after the global flags are parsed. Rather than make `newDockerCommand` return the complicated set of results needed to make this happen, instead return a closure which achieves this. The new functionality is introduced via a common `TopLevelCommand` abstraction which lets us adjust the plugin entrypoint to use the same strategy for parsing the global arguments. This isn't strictly required (in this case the stuff in cobra's `Execute` works fine) but doing it this way avoids the possibility of subtle differences in behaviour. Fixes #1699, and also, as a side-effect, the first item in #1661. Signed-off-by: Ian Campbell <ijc@docker.com>
2019-03-06 05:29:42 -05:00
setupHelpCommand(dockerCli, cmd, helpCmd)
setHelpFunc(dockerCli, cmd)
cmd.SetOutput(dockerCli.Out())
commands.AddCommands(cmd, dockerCli)
cli.DisableFlagsInUseLine(cmd)
allow plugins to have argument which match a top-level flag. The issue with plugin options clashing with globals is that when cobra is parsing the command line and it comes across an argument which doesn't start with a `-` it (in the absence of plugins) distinguishes between "argument to current command" and "new subcommand" based on the list of registered sub commands. Plugins breaks that model. When presented with `docker -D plugin -c foo` cobra parses up to the `plugin`, sees it isn't a registered sub-command of the top-level docker (because it isn't, it's a plugin) so it accumulates it as an argument to the top-level `docker` command. Then it sees the `-c`, and thinks it is the global `-c` (for AKA `--context`) option and tries to treat it as that, which fails. In the specific case of the top-level `docker` subcommand we know that it has no arguments which aren't `--flags` (or `-f` short flags) and so anything which doesn't start with a `-` must either be a (known) subcommand or an attempt to execute a plugin. We could simply scan for and register all installed plugins at start of day, so that cobra can do the right thing, but we want to avoid that since it would involve executing each plugin to fetch the metadata, even if the command wasn't going to end up hitting a plugin. Instead we can parse the initial set of global arguments separately before hitting the main cobra `Execute` path, which works here exactly because we know that the top-level has no non-flag arguments. One slight wrinkle is that the top-level `PersistentPreRunE` is no longer called on the plugins path (since it no longer goes via `Execute`), so we arrange for the initialisation done there (which has to be done after global flags are parsed to handle e.g. `--config`) to happen explictly after the global flags are parsed. Rather than make `newDockerCommand` return the complicated set of results needed to make this happen, instead return a closure which achieves this. The new functionality is introduced via a common `TopLevelCommand` abstraction which lets us adjust the plugin entrypoint to use the same strategy for parsing the global arguments. This isn't strictly required (in this case the stuff in cobra's `Execute` works fine) but doing it this way avoids the possibility of subtle differences in behaviour. Fixes #1699, and also, as a side-effect, the first item in #1661. Signed-off-by: Ian Campbell <ijc@docker.com>
2019-03-06 05:29:42 -05:00
setValidateArgs(dockerCli, cmd)
allow plugins to have argument which match a top-level flag. The issue with plugin options clashing with globals is that when cobra is parsing the command line and it comes across an argument which doesn't start with a `-` it (in the absence of plugins) distinguishes between "argument to current command" and "new subcommand" based on the list of registered sub commands. Plugins breaks that model. When presented with `docker -D plugin -c foo` cobra parses up to the `plugin`, sees it isn't a registered sub-command of the top-level docker (because it isn't, it's a plugin) so it accumulates it as an argument to the top-level `docker` command. Then it sees the `-c`, and thinks it is the global `-c` (for AKA `--context`) option and tries to treat it as that, which fails. In the specific case of the top-level `docker` subcommand we know that it has no arguments which aren't `--flags` (or `-f` short flags) and so anything which doesn't start with a `-` must either be a (known) subcommand or an attempt to execute a plugin. We could simply scan for and register all installed plugins at start of day, so that cobra can do the right thing, but we want to avoid that since it would involve executing each plugin to fetch the metadata, even if the command wasn't going to end up hitting a plugin. Instead we can parse the initial set of global arguments separately before hitting the main cobra `Execute` path, which works here exactly because we know that the top-level has no non-flag arguments. One slight wrinkle is that the top-level `PersistentPreRunE` is no longer called on the plugins path (since it no longer goes via `Execute`), so we arrange for the initialisation done there (which has to be done after global flags are parsed to handle e.g. `--config`) to happen explictly after the global flags are parsed. Rather than make `newDockerCommand` return the complicated set of results needed to make this happen, instead return a closure which achieves this. The new functionality is introduced via a common `TopLevelCommand` abstraction which lets us adjust the plugin entrypoint to use the same strategy for parsing the global arguments. This isn't strictly required (in this case the stuff in cobra's `Execute` works fine) but doing it this way avoids the possibility of subtle differences in behaviour. Fixes #1699, and also, as a side-effect, the first item in #1661. Signed-off-by: Ian Campbell <ijc@docker.com>
2019-03-06 05:29:42 -05:00
// flags must be the top-level command flags, not cmd.Flags()
return cli.NewTopLevelCommand(cmd, dockerCli, opts, flags)
}
allow plugins to have argument which match a top-level flag. The issue with plugin options clashing with globals is that when cobra is parsing the command line and it comes across an argument which doesn't start with a `-` it (in the absence of plugins) distinguishes between "argument to current command" and "new subcommand" based on the list of registered sub commands. Plugins breaks that model. When presented with `docker -D plugin -c foo` cobra parses up to the `plugin`, sees it isn't a registered sub-command of the top-level docker (because it isn't, it's a plugin) so it accumulates it as an argument to the top-level `docker` command. Then it sees the `-c`, and thinks it is the global `-c` (for AKA `--context`) option and tries to treat it as that, which fails. In the specific case of the top-level `docker` subcommand we know that it has no arguments which aren't `--flags` (or `-f` short flags) and so anything which doesn't start with a `-` must either be a (known) subcommand or an attempt to execute a plugin. We could simply scan for and register all installed plugins at start of day, so that cobra can do the right thing, but we want to avoid that since it would involve executing each plugin to fetch the metadata, even if the command wasn't going to end up hitting a plugin. Instead we can parse the initial set of global arguments separately before hitting the main cobra `Execute` path, which works here exactly because we know that the top-level has no non-flag arguments. One slight wrinkle is that the top-level `PersistentPreRunE` is no longer called on the plugins path (since it no longer goes via `Execute`), so we arrange for the initialisation done there (which has to be done after global flags are parsed to handle e.g. `--config`) to happen explictly after the global flags are parsed. Rather than make `newDockerCommand` return the complicated set of results needed to make this happen, instead return a closure which achieves this. The new functionality is introduced via a common `TopLevelCommand` abstraction which lets us adjust the plugin entrypoint to use the same strategy for parsing the global arguments. This isn't strictly required (in this case the stuff in cobra's `Execute` works fine) but doing it this way avoids the possibility of subtle differences in behaviour. Fixes #1699, and also, as a side-effect, the first item in #1661. Signed-off-by: Ian Campbell <ijc@docker.com>
2019-03-06 05:29:42 -05:00
func setFlagErrorFunc(dockerCli *command.DockerCli, cmd *cobra.Command) {
// When invoking `docker stack --nonsense`, we need to make sure FlagErrorFunc return appropriate
// output if the feature is not supported.
// As above cli.SetupRootCommand(cmd) have already setup the FlagErrorFunc, we will add a pre-check before the FlagErrorFunc
// is called.
flagErrorFunc := cmd.FlagErrorFunc()
cmd.SetFlagErrorFunc(func(cmd *cobra.Command, err error) error {
if err := isSupported(cmd, dockerCli); err != nil {
return err
}
return flagErrorFunc(cmd, err)
})
}
allow plugins to have argument which match a top-level flag. The issue with plugin options clashing with globals is that when cobra is parsing the command line and it comes across an argument which doesn't start with a `-` it (in the absence of plugins) distinguishes between "argument to current command" and "new subcommand" based on the list of registered sub commands. Plugins breaks that model. When presented with `docker -D plugin -c foo` cobra parses up to the `plugin`, sees it isn't a registered sub-command of the top-level docker (because it isn't, it's a plugin) so it accumulates it as an argument to the top-level `docker` command. Then it sees the `-c`, and thinks it is the global `-c` (for AKA `--context`) option and tries to treat it as that, which fails. In the specific case of the top-level `docker` subcommand we know that it has no arguments which aren't `--flags` (or `-f` short flags) and so anything which doesn't start with a `-` must either be a (known) subcommand or an attempt to execute a plugin. We could simply scan for and register all installed plugins at start of day, so that cobra can do the right thing, but we want to avoid that since it would involve executing each plugin to fetch the metadata, even if the command wasn't going to end up hitting a plugin. Instead we can parse the initial set of global arguments separately before hitting the main cobra `Execute` path, which works here exactly because we know that the top-level has no non-flag arguments. One slight wrinkle is that the top-level `PersistentPreRunE` is no longer called on the plugins path (since it no longer goes via `Execute`), so we arrange for the initialisation done there (which has to be done after global flags are parsed to handle e.g. `--config`) to happen explictly after the global flags are parsed. Rather than make `newDockerCommand` return the complicated set of results needed to make this happen, instead return a closure which achieves this. The new functionality is introduced via a common `TopLevelCommand` abstraction which lets us adjust the plugin entrypoint to use the same strategy for parsing the global arguments. This isn't strictly required (in this case the stuff in cobra's `Execute` works fine) but doing it this way avoids the possibility of subtle differences in behaviour. Fixes #1699, and also, as a side-effect, the first item in #1661. Signed-off-by: Ian Campbell <ijc@docker.com>
2019-03-06 05:29:42 -05:00
func setupHelpCommand(dockerCli command.Cli, rootCmd, helpCmd *cobra.Command) {
origRun := helpCmd.Run
origRunE := helpCmd.RunE
helpCmd.Run = nil
helpCmd.RunE = func(c *cobra.Command, args []string) error {
if len(args) > 0 {
helpcmd, err := pluginmanager.PluginRunCommand(dockerCli, args[0], rootCmd)
if err == nil {
err = helpcmd.Run()
if err != nil {
return err
}
}
if !pluginmanager.IsNotFound(err) {
return err
}
}
if origRunE != nil {
return origRunE(c, args)
}
origRun(c, args)
return nil
}
}
func tryRunPluginHelp(dockerCli command.Cli, ccmd *cobra.Command, cargs []string) error {
root := ccmd.Root()
cmd, _, err := root.Traverse(cargs)
if err != nil {
return err
}
helpcmd, err := pluginmanager.PluginRunCommand(dockerCli, cmd.Name(), root)
if err != nil {
return err
}
return helpcmd.Run()
}
allow plugins to have argument which match a top-level flag. The issue with plugin options clashing with globals is that when cobra is parsing the command line and it comes across an argument which doesn't start with a `-` it (in the absence of plugins) distinguishes between "argument to current command" and "new subcommand" based on the list of registered sub commands. Plugins breaks that model. When presented with `docker -D plugin -c foo` cobra parses up to the `plugin`, sees it isn't a registered sub-command of the top-level docker (because it isn't, it's a plugin) so it accumulates it as an argument to the top-level `docker` command. Then it sees the `-c`, and thinks it is the global `-c` (for AKA `--context`) option and tries to treat it as that, which fails. In the specific case of the top-level `docker` subcommand we know that it has no arguments which aren't `--flags` (or `-f` short flags) and so anything which doesn't start with a `-` must either be a (known) subcommand or an attempt to execute a plugin. We could simply scan for and register all installed plugins at start of day, so that cobra can do the right thing, but we want to avoid that since it would involve executing each plugin to fetch the metadata, even if the command wasn't going to end up hitting a plugin. Instead we can parse the initial set of global arguments separately before hitting the main cobra `Execute` path, which works here exactly because we know that the top-level has no non-flag arguments. One slight wrinkle is that the top-level `PersistentPreRunE` is no longer called on the plugins path (since it no longer goes via `Execute`), so we arrange for the initialisation done there (which has to be done after global flags are parsed to handle e.g. `--config`) to happen explictly after the global flags are parsed. Rather than make `newDockerCommand` return the complicated set of results needed to make this happen, instead return a closure which achieves this. The new functionality is introduced via a common `TopLevelCommand` abstraction which lets us adjust the plugin entrypoint to use the same strategy for parsing the global arguments. This isn't strictly required (in this case the stuff in cobra's `Execute` works fine) but doing it this way avoids the possibility of subtle differences in behaviour. Fixes #1699, and also, as a side-effect, the first item in #1661. Signed-off-by: Ian Campbell <ijc@docker.com>
2019-03-06 05:29:42 -05:00
func setHelpFunc(dockerCli command.Cli, cmd *cobra.Command) {
defaultHelpFunc := cmd.HelpFunc()
cmd.SetHelpFunc(func(ccmd *cobra.Command, args []string) {
// Add a stub entry for every plugin so they are
// included in the help output and so that
// `tryRunPluginHelp` can find them or if we fall
// through they will be included in the default help
// output.
if err := pluginmanager.AddPluginCommandStubs(dockerCli, ccmd.Root()); err != nil {
ccmd.Println(err)
return
}
if len(args) >= 1 {
err := tryRunPluginHelp(dockerCli, ccmd, args)
if err == nil { // Successfully ran the plugin
return
}
if !pluginmanager.IsNotFound(err) {
ccmd.Println(err)
return
}
}
if err := isSupported(ccmd, dockerCli); err != nil {
ccmd.Println(err)
return
}
if err := hideUnsupportedFeatures(ccmd, dockerCli); err != nil {
ccmd.Println(err)
return
}
defaultHelpFunc(ccmd, args)
})
}
allow plugins to have argument which match a top-level flag. The issue with plugin options clashing with globals is that when cobra is parsing the command line and it comes across an argument which doesn't start with a `-` it (in the absence of plugins) distinguishes between "argument to current command" and "new subcommand" based on the list of registered sub commands. Plugins breaks that model. When presented with `docker -D plugin -c foo` cobra parses up to the `plugin`, sees it isn't a registered sub-command of the top-level docker (because it isn't, it's a plugin) so it accumulates it as an argument to the top-level `docker` command. Then it sees the `-c`, and thinks it is the global `-c` (for AKA `--context`) option and tries to treat it as that, which fails. In the specific case of the top-level `docker` subcommand we know that it has no arguments which aren't `--flags` (or `-f` short flags) and so anything which doesn't start with a `-` must either be a (known) subcommand or an attempt to execute a plugin. We could simply scan for and register all installed plugins at start of day, so that cobra can do the right thing, but we want to avoid that since it would involve executing each plugin to fetch the metadata, even if the command wasn't going to end up hitting a plugin. Instead we can parse the initial set of global arguments separately before hitting the main cobra `Execute` path, which works here exactly because we know that the top-level has no non-flag arguments. One slight wrinkle is that the top-level `PersistentPreRunE` is no longer called on the plugins path (since it no longer goes via `Execute`), so we arrange for the initialisation done there (which has to be done after global flags are parsed to handle e.g. `--config`) to happen explictly after the global flags are parsed. Rather than make `newDockerCommand` return the complicated set of results needed to make this happen, instead return a closure which achieves this. The new functionality is introduced via a common `TopLevelCommand` abstraction which lets us adjust the plugin entrypoint to use the same strategy for parsing the global arguments. This isn't strictly required (in this case the stuff in cobra's `Execute` works fine) but doing it this way avoids the possibility of subtle differences in behaviour. Fixes #1699, and also, as a side-effect, the first item in #1661. Signed-off-by: Ian Campbell <ijc@docker.com>
2019-03-06 05:29:42 -05:00
func setValidateArgs(dockerCli *command.DockerCli, cmd *cobra.Command) {
// The Args is handled by ValidateArgs in cobra, which does not allows a pre-hook.
// As a result, here we replace the existing Args validation func to a wrapper,
// where the wrapper will check to see if the feature is supported or not.
// The Args validation error will only be returned if the feature is supported.
cli.VisitAll(cmd, func(ccmd *cobra.Command) {
// if there is no tags for a command or any of its parent,
// there is no need to wrap the Args validation.
if !hasTags(ccmd) {
return
}
if ccmd.Args == nil {
return
}
cmdArgs := ccmd.Args
ccmd.Args = func(cmd *cobra.Command, args []string) error {
if err := isSupported(cmd, dockerCli); err != nil {
return err
}
return cmdArgs(cmd, args)
}
})
}
allow plugins to have argument which match a top-level flag. The issue with plugin options clashing with globals is that when cobra is parsing the command line and it comes across an argument which doesn't start with a `-` it (in the absence of plugins) distinguishes between "argument to current command" and "new subcommand" based on the list of registered sub commands. Plugins breaks that model. When presented with `docker -D plugin -c foo` cobra parses up to the `plugin`, sees it isn't a registered sub-command of the top-level docker (because it isn't, it's a plugin) so it accumulates it as an argument to the top-level `docker` command. Then it sees the `-c`, and thinks it is the global `-c` (for AKA `--context`) option and tries to treat it as that, which fails. In the specific case of the top-level `docker` subcommand we know that it has no arguments which aren't `--flags` (or `-f` short flags) and so anything which doesn't start with a `-` must either be a (known) subcommand or an attempt to execute a plugin. We could simply scan for and register all installed plugins at start of day, so that cobra can do the right thing, but we want to avoid that since it would involve executing each plugin to fetch the metadata, even if the command wasn't going to end up hitting a plugin. Instead we can parse the initial set of global arguments separately before hitting the main cobra `Execute` path, which works here exactly because we know that the top-level has no non-flag arguments. One slight wrinkle is that the top-level `PersistentPreRunE` is no longer called on the plugins path (since it no longer goes via `Execute`), so we arrange for the initialisation done there (which has to be done after global flags are parsed to handle e.g. `--config`) to happen explictly after the global flags are parsed. Rather than make `newDockerCommand` return the complicated set of results needed to make this happen, instead return a closure which achieves this. The new functionality is introduced via a common `TopLevelCommand` abstraction which lets us adjust the plugin entrypoint to use the same strategy for parsing the global arguments. This isn't strictly required (in this case the stuff in cobra's `Execute` works fine) but doing it this way avoids the possibility of subtle differences in behaviour. Fixes #1699, and also, as a side-effect, the first item in #1661. Signed-off-by: Ian Campbell <ijc@docker.com>
2019-03-06 05:29:42 -05:00
func tryPluginRun(dockerCli command.Cli, cmd *cobra.Command, subcommand string) error {
plugincmd, err := pluginmanager.PluginRunCommand(dockerCli, subcommand, cmd)
if err != nil {
return err
}
allow plugins to have argument which match a top-level flag. The issue with plugin options clashing with globals is that when cobra is parsing the command line and it comes across an argument which doesn't start with a `-` it (in the absence of plugins) distinguishes between "argument to current command" and "new subcommand" based on the list of registered sub commands. Plugins breaks that model. When presented with `docker -D plugin -c foo` cobra parses up to the `plugin`, sees it isn't a registered sub-command of the top-level docker (because it isn't, it's a plugin) so it accumulates it as an argument to the top-level `docker` command. Then it sees the `-c`, and thinks it is the global `-c` (for AKA `--context`) option and tries to treat it as that, which fails. In the specific case of the top-level `docker` subcommand we know that it has no arguments which aren't `--flags` (or `-f` short flags) and so anything which doesn't start with a `-` must either be a (known) subcommand or an attempt to execute a plugin. We could simply scan for and register all installed plugins at start of day, so that cobra can do the right thing, but we want to avoid that since it would involve executing each plugin to fetch the metadata, even if the command wasn't going to end up hitting a plugin. Instead we can parse the initial set of global arguments separately before hitting the main cobra `Execute` path, which works here exactly because we know that the top-level has no non-flag arguments. One slight wrinkle is that the top-level `PersistentPreRunE` is no longer called on the plugins path (since it no longer goes via `Execute`), so we arrange for the initialisation done there (which has to be done after global flags are parsed to handle e.g. `--config`) to happen explictly after the global flags are parsed. Rather than make `newDockerCommand` return the complicated set of results needed to make this happen, instead return a closure which achieves this. The new functionality is introduced via a common `TopLevelCommand` abstraction which lets us adjust the plugin entrypoint to use the same strategy for parsing the global arguments. This isn't strictly required (in this case the stuff in cobra's `Execute` works fine) but doing it this way avoids the possibility of subtle differences in behaviour. Fixes #1699, and also, as a side-effect, the first item in #1661. Signed-off-by: Ian Campbell <ijc@docker.com>
2019-03-06 05:29:42 -05:00
if err := plugincmd.Run(); err != nil {
statusCode := 1
exitErr, ok := err.(*exec.ExitError)
if !ok {
return err
}
if ws, ok := exitErr.Sys().(syscall.WaitStatus); ok {
statusCode = ws.ExitStatus()
}
return cli.StatusError{
StatusCode: statusCode,
}
}
return nil
}
func runDocker(dockerCli *command.DockerCli) error {
tcmd := newDockerCommand(dockerCli)
cmd, args, err := tcmd.HandleGlobalFlags()
if err != nil {
return err
}
if err := tcmd.Initialize(); err != nil {
return err
}
if len(args) > 0 {
if _, _, err := cmd.Find(args); err != nil {
err := tryPluginRun(dockerCli, cmd, args[0])
if !pluginmanager.IsNotFound(err) {
return err
}
// For plugin not found we fall through to
// cmd.Execute() which deals with reporting
// "command not found" in a consistent way.
}
}
return cmd.Execute()
}
func main() {
dockerCli, err := command.NewDockerCli()
if err != nil {
fmt.Fprintln(os.Stderr, err)
os.Exit(1)
}
logrus.SetOutput(dockerCli.Err())
allow plugins to have argument which match a top-level flag. The issue with plugin options clashing with globals is that when cobra is parsing the command line and it comes across an argument which doesn't start with a `-` it (in the absence of plugins) distinguishes between "argument to current command" and "new subcommand" based on the list of registered sub commands. Plugins breaks that model. When presented with `docker -D plugin -c foo` cobra parses up to the `plugin`, sees it isn't a registered sub-command of the top-level docker (because it isn't, it's a plugin) so it accumulates it as an argument to the top-level `docker` command. Then it sees the `-c`, and thinks it is the global `-c` (for AKA `--context`) option and tries to treat it as that, which fails. In the specific case of the top-level `docker` subcommand we know that it has no arguments which aren't `--flags` (or `-f` short flags) and so anything which doesn't start with a `-` must either be a (known) subcommand or an attempt to execute a plugin. We could simply scan for and register all installed plugins at start of day, so that cobra can do the right thing, but we want to avoid that since it would involve executing each plugin to fetch the metadata, even if the command wasn't going to end up hitting a plugin. Instead we can parse the initial set of global arguments separately before hitting the main cobra `Execute` path, which works here exactly because we know that the top-level has no non-flag arguments. One slight wrinkle is that the top-level `PersistentPreRunE` is no longer called on the plugins path (since it no longer goes via `Execute`), so we arrange for the initialisation done there (which has to be done after global flags are parsed to handle e.g. `--config`) to happen explictly after the global flags are parsed. Rather than make `newDockerCommand` return the complicated set of results needed to make this happen, instead return a closure which achieves this. The new functionality is introduced via a common `TopLevelCommand` abstraction which lets us adjust the plugin entrypoint to use the same strategy for parsing the global arguments. This isn't strictly required (in this case the stuff in cobra's `Execute` works fine) but doing it this way avoids the possibility of subtle differences in behaviour. Fixes #1699, and also, as a side-effect, the first item in #1661. Signed-off-by: Ian Campbell <ijc@docker.com>
2019-03-06 05:29:42 -05:00
if err := runDocker(dockerCli); err != nil {
if sterr, ok := err.(cli.StatusError); ok {
if sterr.Status != "" {
fmt.Fprintln(dockerCli.Err(), sterr.Status)
}
// StatusError should only be used for errors, and all errors should
// have a non-zero exit status, so never exit with 0
if sterr.StatusCode == 0 {
os.Exit(1)
}
os.Exit(sterr.StatusCode)
}
fmt.Fprintln(dockerCli.Err(), err)
os.Exit(1)
}
}
type versionDetails interface {
Client() client.APIClient
ClientInfo() command.ClientInfo
ServerInfo() command.ServerInfo
}
func hideFeatureFlag(f *pflag.Flag, hasFeature bool, annotation string) {
if hasFeature {
return
}
if _, ok := f.Annotations[annotation]; ok {
f.Hidden = true
}
}
func hideFeatureSubCommand(subcmd *cobra.Command, hasFeature bool, annotation string) {
if hasFeature {
return
}
if _, ok := subcmd.Annotations[annotation]; ok {
subcmd.Hidden = true
}
}
func hideUnsupportedFeatures(cmd *cobra.Command, details versionDetails) error {
clientVersion := details.Client().ClientVersion()
osType := details.ServerInfo().OSType
hasExperimental := details.ServerInfo().HasExperimental
hasExperimentalCLI := details.ClientInfo().HasExperimental
hasBuildKit, err := command.BuildKitEnabled(details.ServerInfo())
if err != nil {
return err
}
cmd.Flags().VisitAll(func(f *pflag.Flag) {
hideFeatureFlag(f, hasExperimental, "experimental")
hideFeatureFlag(f, hasExperimentalCLI, "experimentalCLI")
hideFeatureFlag(f, hasBuildKit, "buildkit")
hideFeatureFlag(f, !hasBuildKit, "no-buildkit")
// hide flags not supported by the server
if !isOSTypeSupported(f, osType) || !isVersionSupported(f, clientVersion) {
f.Hidden = true
}
// root command shows all top-level flags
if cmd.Parent() != nil {
if commands, ok := f.Annotations["top-level"]; ok {
f.Hidden = !findCommand(cmd, commands)
}
}
})
for _, subcmd := range cmd.Commands() {
hideFeatureSubCommand(subcmd, hasExperimental, "experimental")
hideFeatureSubCommand(subcmd, hasExperimentalCLI, "experimentalCLI")
hideFeatureSubCommand(subcmd, hasBuildKit, "buildkit")
hideFeatureSubCommand(subcmd, !hasBuildKit, "no-buildkit")
// hide subcommands not supported by the server
if subcmdVersion, ok := subcmd.Annotations["version"]; ok && versions.LessThan(clientVersion, subcmdVersion) {
subcmd.Hidden = true
}
Mark checkpoint feature as Linux-only, and homogenize error messages This patch adds annotations to mark the checkpoint commands as Linux only, which hides them if the daemon is running a non-matching operating-system type; Before: docker Usage: docker COMMAND A self-sufficient runtime for containers ... Management Commands: config Manage Docker configs container Manage containers image Manage images After: docker Usage: docker COMMAND A self-sufficient runtime for containers ... Management Commands: checkpoint Manage checkpoints config Manage Docker configs container Manage containers image Manage images This change also prints errors when attempting to use checkpoint commands or flags if the feature is not supported by the Daemon's operating system; $ docker checkpoint --help docker checkpoint is only supported on a Docker daemon running on linux, but the Docker daemon is running on windows $ docker checkpoint create --help docker checkpoint create is only supported on a Docker daemon running on linux, but the Docker daemon is running on windows $ docker checkpoint ls --help docker checkpoint ls is only supported on a Docker daemon running on linux, but the Docker daemon is running on windows $ docker checkpoint rm --help docker checkpoint rm is only supported on a Docker daemon running on linux, but the Docker daemon is running on windows $ docker container start --checkpoint=foo mycontainer "--checkpoint" requires the Docker daemon to run on linux, but the Docker daemon is running on windows $ docker container start --checkpoint-dir=/foo/bar mycontainer "--checkpoint-dir" requires the Docker daemon to run on linux, but the Docker daemon is running on windows Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
2018-05-30 17:14:59 -04:00
if v, ok := subcmd.Annotations["ostype"]; ok && v != osType {
subcmd.Hidden = true
}
}
return nil
}
// Checks if a command or one of its ancestors is in the list
func findCommand(cmd *cobra.Command, commands []string) bool {
if cmd == nil {
return false
}
for _, c := range commands {
if c == cmd.Name() {
return true
}
}
return findCommand(cmd.Parent(), commands)
}
func isSupported(cmd *cobra.Command, details versionDetails) error {
if err := areSubcommandsSupported(cmd, details); err != nil {
return err
}
return areFlagsSupported(cmd, details)
}
func areFlagsSupported(cmd *cobra.Command, details versionDetails) error {
clientVersion := details.Client().ClientVersion()
osType := details.ServerInfo().OSType
hasExperimental := details.ServerInfo().HasExperimental
hasExperimentalCLI := details.ClientInfo().HasExperimental
2017-01-16 09:35:27 -05:00
errs := []string{}
cmd.Flags().VisitAll(func(f *pflag.Flag) {
if f.Changed {
if !isVersionSupported(f, clientVersion) {
errs = append(errs, fmt.Sprintf("\"--%s\" requires API version %s, but the Docker daemon API version is %s", f.Name, getFlagAnnotation(f, "version"), clientVersion))
return
}
if !isOSTypeSupported(f, osType) {
Mark checkpoint feature as Linux-only, and homogenize error messages This patch adds annotations to mark the checkpoint commands as Linux only, which hides them if the daemon is running a non-matching operating-system type; Before: docker Usage: docker COMMAND A self-sufficient runtime for containers ... Management Commands: config Manage Docker configs container Manage containers image Manage images After: docker Usage: docker COMMAND A self-sufficient runtime for containers ... Management Commands: checkpoint Manage checkpoints config Manage Docker configs container Manage containers image Manage images This change also prints errors when attempting to use checkpoint commands or flags if the feature is not supported by the Daemon's operating system; $ docker checkpoint --help docker checkpoint is only supported on a Docker daemon running on linux, but the Docker daemon is running on windows $ docker checkpoint create --help docker checkpoint create is only supported on a Docker daemon running on linux, but the Docker daemon is running on windows $ docker checkpoint ls --help docker checkpoint ls is only supported on a Docker daemon running on linux, but the Docker daemon is running on windows $ docker checkpoint rm --help docker checkpoint rm is only supported on a Docker daemon running on linux, but the Docker daemon is running on windows $ docker container start --checkpoint=foo mycontainer "--checkpoint" requires the Docker daemon to run on linux, but the Docker daemon is running on windows $ docker container start --checkpoint-dir=/foo/bar mycontainer "--checkpoint-dir" requires the Docker daemon to run on linux, but the Docker daemon is running on windows Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
2018-05-30 17:14:59 -04:00
errs = append(errs, fmt.Sprintf("\"--%s\" is only supported on a Docker daemon running on %s, but the Docker daemon is running on %s", f.Name, getFlagAnnotation(f, "ostype"), osType))
2017-01-16 09:35:27 -05:00
return
}
if _, ok := f.Annotations["experimental"]; ok && !hasExperimental {
errs = append(errs, fmt.Sprintf("\"--%s\" is only supported on a Docker daemon with experimental features enabled", f.Name))
}
if _, ok := f.Annotations["experimentalCLI"]; ok && !hasExperimentalCLI {
errs = append(errs, fmt.Sprintf("\"--%s\" is only supported on a Docker cli with experimental cli features enabled", f.Name))
}
// buildkit-specific flags are noop when buildkit is not enabled, so we do not add an error in that case
2017-01-16 09:35:27 -05:00
}
})
if len(errs) > 0 {
return errors.New(strings.Join(errs, "\n"))
}
return nil
}
// Check recursively so that, e.g., `docker stack ls` returns the same output as `docker stack`
func areSubcommandsSupported(cmd *cobra.Command, details versionDetails) error {
clientVersion := details.Client().ClientVersion()
Mark checkpoint feature as Linux-only, and homogenize error messages This patch adds annotations to mark the checkpoint commands as Linux only, which hides them if the daemon is running a non-matching operating-system type; Before: docker Usage: docker COMMAND A self-sufficient runtime for containers ... Management Commands: config Manage Docker configs container Manage containers image Manage images After: docker Usage: docker COMMAND A self-sufficient runtime for containers ... Management Commands: checkpoint Manage checkpoints config Manage Docker configs container Manage containers image Manage images This change also prints errors when attempting to use checkpoint commands or flags if the feature is not supported by the Daemon's operating system; $ docker checkpoint --help docker checkpoint is only supported on a Docker daemon running on linux, but the Docker daemon is running on windows $ docker checkpoint create --help docker checkpoint create is only supported on a Docker daemon running on linux, but the Docker daemon is running on windows $ docker checkpoint ls --help docker checkpoint ls is only supported on a Docker daemon running on linux, but the Docker daemon is running on windows $ docker checkpoint rm --help docker checkpoint rm is only supported on a Docker daemon running on linux, but the Docker daemon is running on windows $ docker container start --checkpoint=foo mycontainer "--checkpoint" requires the Docker daemon to run on linux, but the Docker daemon is running on windows $ docker container start --checkpoint-dir=/foo/bar mycontainer "--checkpoint-dir" requires the Docker daemon to run on linux, but the Docker daemon is running on windows Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
2018-05-30 17:14:59 -04:00
osType := details.ServerInfo().OSType
hasExperimental := details.ServerInfo().HasExperimental
hasExperimentalCLI := details.ClientInfo().HasExperimental
// Check recursively so that, e.g., `docker stack ls` returns the same output as `docker stack`
for curr := cmd; curr != nil; curr = curr.Parent() {
if cmdVersion, ok := curr.Annotations["version"]; ok && versions.LessThan(clientVersion, cmdVersion) {
return fmt.Errorf("%s requires API version %s, but the Docker daemon API version is %s", cmd.CommandPath(), cmdVersion, clientVersion)
}
Mark checkpoint feature as Linux-only, and homogenize error messages This patch adds annotations to mark the checkpoint commands as Linux only, which hides them if the daemon is running a non-matching operating-system type; Before: docker Usage: docker COMMAND A self-sufficient runtime for containers ... Management Commands: config Manage Docker configs container Manage containers image Manage images After: docker Usage: docker COMMAND A self-sufficient runtime for containers ... Management Commands: checkpoint Manage checkpoints config Manage Docker configs container Manage containers image Manage images This change also prints errors when attempting to use checkpoint commands or flags if the feature is not supported by the Daemon's operating system; $ docker checkpoint --help docker checkpoint is only supported on a Docker daemon running on linux, but the Docker daemon is running on windows $ docker checkpoint create --help docker checkpoint create is only supported on a Docker daemon running on linux, but the Docker daemon is running on windows $ docker checkpoint ls --help docker checkpoint ls is only supported on a Docker daemon running on linux, but the Docker daemon is running on windows $ docker checkpoint rm --help docker checkpoint rm is only supported on a Docker daemon running on linux, but the Docker daemon is running on windows $ docker container start --checkpoint=foo mycontainer "--checkpoint" requires the Docker daemon to run on linux, but the Docker daemon is running on windows $ docker container start --checkpoint-dir=/foo/bar mycontainer "--checkpoint-dir" requires the Docker daemon to run on linux, but the Docker daemon is running on windows Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
2018-05-30 17:14:59 -04:00
if os, ok := curr.Annotations["ostype"]; ok && os != osType {
return fmt.Errorf("%s is only supported on a Docker daemon running on %s, but the Docker daemon is running on %s", cmd.CommandPath(), os, osType)
}
if _, ok := curr.Annotations["experimental"]; ok && !hasExperimental {
return fmt.Errorf("%s is only supported on a Docker daemon with experimental features enabled", cmd.CommandPath())
}
if _, ok := curr.Annotations["experimentalCLI"]; ok && !hasExperimentalCLI {
Mark checkpoint feature as Linux-only, and homogenize error messages This patch adds annotations to mark the checkpoint commands as Linux only, which hides them if the daemon is running a non-matching operating-system type; Before: docker Usage: docker COMMAND A self-sufficient runtime for containers ... Management Commands: config Manage Docker configs container Manage containers image Manage images After: docker Usage: docker COMMAND A self-sufficient runtime for containers ... Management Commands: checkpoint Manage checkpoints config Manage Docker configs container Manage containers image Manage images This change also prints errors when attempting to use checkpoint commands or flags if the feature is not supported by the Daemon's operating system; $ docker checkpoint --help docker checkpoint is only supported on a Docker daemon running on linux, but the Docker daemon is running on windows $ docker checkpoint create --help docker checkpoint create is only supported on a Docker daemon running on linux, but the Docker daemon is running on windows $ docker checkpoint ls --help docker checkpoint ls is only supported on a Docker daemon running on linux, but the Docker daemon is running on windows $ docker checkpoint rm --help docker checkpoint rm is only supported on a Docker daemon running on linux, but the Docker daemon is running on windows $ docker container start --checkpoint=foo mycontainer "--checkpoint" requires the Docker daemon to run on linux, but the Docker daemon is running on windows $ docker container start --checkpoint-dir=/foo/bar mycontainer "--checkpoint-dir" requires the Docker daemon to run on linux, but the Docker daemon is running on windows Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
2018-05-30 17:14:59 -04:00
return fmt.Errorf("%s is only supported on a Docker cli with experimental cli features enabled", cmd.CommandPath())
}
}
return nil
}
2017-01-16 09:35:27 -05:00
func getFlagAnnotation(f *pflag.Flag, annotation string) string {
if value, ok := f.Annotations[annotation]; ok && len(value) == 1 {
return value[0]
2017-01-16 09:35:27 -05:00
}
return ""
}
func isVersionSupported(f *pflag.Flag, clientVersion string) bool {
if v := getFlagAnnotation(f, "version"); v != "" {
2017-01-16 09:35:27 -05:00
return versions.GreaterThanOrEqualTo(clientVersion, v)
}
return true
}
func isOSTypeSupported(f *pflag.Flag, osType string) bool {
if v := getFlagAnnotation(f, "ostype"); v != "" && osType != "" {
return osType == v
}
return true
}
// hasTags return true if any of the command's parents has tags
func hasTags(cmd *cobra.Command) bool {
for curr := cmd; curr != nil; curr = curr.Parent() {
if len(curr.Annotations) > 0 {
return true
}
}
return false
}