completion: fix powershell completion corrupting non-ASCII names - fixes #9412

The Cobra generated PowerShell completion script captures rclone's output
through a pipeline with Invoke-Expression. PowerShell decodes that output
using [Console]::OutputEncoding, which on non-UTF-8 hosts (for example
PowerShell 5.1 on a Windows install with an OEM code page such as CP852)
misinterprets the UTF-8 bytes rclone emits and corrupts remote and path
names containing non-ASCII characters, so tab completion produces a path
that does not exist.

Inject "[Console]::OutputEncoding = [System.Text.Encoding]::UTF8" into the
generated script immediately before the Invoke-Expression call. This is safe
on PowerShell 7+, where UTF-8 is already the default. If the expected line is
not present (for example after a Cobra template change) the script is emitted
unmodified so we never produce a corrupted completion script.

cmd/genautocomplete/genautocomplete_powershell.go

@@ -1,8 +1,10 @@
 package genautocomplete
 
 import (
+	"bytes"
 	"fmt"
 	"os"
+	"strings"
 
 	"github.com/rclone/rclone/cmd"
 	"github.com/rclone/rclone/fs"
@@ -13,6 +15,34 @@ func init() {
 	completionDefinition.AddCommand(powershellCommandDefinition)
 }
 
+// powerShellInvokeLine is the line in the Cobra generated PowerShell completion
+// script that captures rclone's output through a pipeline.
+const powerShellInvokeLine = `Invoke-Expression -OutVariable out "$RequestComp" 2>&1 | Out-Null`
+
+// powerShellUTF8Fix forces the captured output to be decoded as UTF-8. When
+// PowerShell captures a child process' stdout through a pipeline it decodes the
+// bytes using [Console]::OutputEncoding, which on non-UTF-8 systems (for
+// example PowerShell 5.1 on a Windows install with an OEM code page such as
+// CP852) corrupts the UTF-8 that rclone emits. Setting the encoding to UTF-8 is
+// safe on PowerShell 7+, where it is already the default.
+const powerShellUTF8Fix = `[Console]::OutputEncoding = [System.Text.Encoding]::UTF8`
+
+// patchPowerShellCompletion injects the UTF-8 output encoding fix immediately
+// before the Invoke-Expression call in the Cobra generated PowerShell
+// completion script. If the expected line is not found (for example because the
+// upstream Cobra template changed), the script is returned unmodified so we
+// never emit a corrupted completion script.
+func patchPowerShellCompletion(script string) string {
+	idx := strings.Index(script, powerShellInvokeLine)
+	if idx == -1 {
+		return script
+	}
+	// Reuse the indentation of the Invoke-Expression line for the inserted line.
+	lineStart := strings.LastIndex(script[:idx], "\n") + 1
+	indent := script[lineStart:idx]
+	return script[:lineStart] + indent + powerShellUTF8Fix + "\n" + script[lineStart:]
+}
+
 var powershellCommandDefinition = &cobra.Command{
 	Use:   "powershell [output_file]",
 	Short: `Output powershell completion script for rclone.`,
@@ -30,15 +60,18 @@ to your powershell profile.
 If output_file is "-" or missing, then the output will be written to stdout.`,
 	Run: func(command *cobra.Command, args []string) {
 		cmd.CheckArgs(0, 1, command, args)
+		var buf bytes.Buffer
+		if err := cmd.Root.GenPowerShellCompletion(&buf); err != nil {
+			fs.Fatal(nil, fmt.Sprint(err))
+		}
+		script := patchPowerShellCompletion(buf.String())
 		if len(args) == 0 || (len(args) > 0 && args[0] == "-") {
-			err := cmd.Root.GenPowerShellCompletion(os.Stdout)
-			if err != nil {
+			if _, err := os.Stdout.WriteString(script); err != nil {
 				fs.Fatal(nil, fmt.Sprint(err))
 			}
 			return
 		}
-		err := cmd.Root.GenPowerShellCompletionFile(args[0])
-		if err != nil {
+		if err := os.WriteFile(args[0], []byte(script), 0644); err != nil {
 			fs.Fatal(nil, fmt.Sprint(err))
 		}
 	},

cmd/genautocomplete/genautocomplete_test.go

@@ -2,6 +2,7 @@ package genautocomplete
 
 import (
 	"os"
+	"strings"
 	"testing"
 
 	"github.com/stretchr/testify/assert"
@@ -107,3 +108,58 @@ func TestCompletionFishStdout(t *testing.T) {
 	assert.NoError(t, err)
 	assert.NotEmpty(t, string(output))
 }
+
+func TestCompletionPowershell(t *testing.T) {
+	tempFile, err := os.CreateTemp("", "completion_powershell")
+	assert.NoError(t, err)
+	defer func() {
+		_ = tempFile.Close()
+		_ = os.Remove(tempFile.Name())
+	}()
+
+	powershellCommandDefinition.Run(powershellCommandDefinition, []string{tempFile.Name()})
+
+	bs, err := os.ReadFile(tempFile.Name())
+	assert.NoError(t, err)
+	assert.NotEmpty(t, string(bs))
+	// The generated script must force UTF-8 output decoding so that non-ASCII
+	// remote names are not corrupted on non-UTF-8 PowerShell hosts.
+	assert.Contains(t, string(bs), powerShellUTF8Fix)
+}
+
+func TestCompletionPowershellStdout(t *testing.T) {
+	originalStdout := os.Stdout
+	tempFile, err := os.CreateTemp("", "completion_powershell")
+	assert.NoError(t, err)
+	defer func() {
+		_ = tempFile.Close()
+		_ = os.Remove(tempFile.Name())
+	}()
+
+	os.Stdout = tempFile
+	defer func() { os.Stdout = originalStdout }()
+
+	powershellCommandDefinition.Run(powershellCommandDefinition, []string{"-"})
+
+	output, err := os.ReadFile(tempFile.Name())
+	assert.NoError(t, err)
+	assert.NotEmpty(t, string(output))
+	assert.Contains(t, string(output), powerShellUTF8Fix)
+}
+
+func TestPatchPowerShellCompletion(t *testing.T) {
+	t.Run("injects the encoding fix before the invoke line", func(t *testing.T) {
+		script := "before\n    " + powerShellInvokeLine + "\nafter\n"
+		got := patchPowerShellCompletion(script)
+		// The fix is inserted on its own line, sharing the indentation of the
+		// invoke line, immediately before it.
+		want := "before\n    " + powerShellUTF8Fix + "\n    " + powerShellInvokeLine + "\nafter\n"
+		assert.Equal(t, want, got)
+		assert.Less(t, strings.Index(got, powerShellUTF8Fix), strings.Index(got, powerShellInvokeLine))
+	})
+
+	t.Run("leaves the script unchanged when the invoke line is absent", func(t *testing.T) {
+		script := "some other script\nwithout the expected line\n"
+		assert.Equal(t, script, patchPowerShellCompletion(script))
+	})
+}