fix: UTF-8 encoding in PowerShell commands for Turkish

codex-rs/core/src/shell.rs

@@ -59,7 +59,7 @@ impl Shell {
                             ps.exe.clone(),
                             "-NoProfile".to_string(),
                             "-Command".to_string(),
-                            script,
+                            format!("[Console]::OutputEncoding = [System.Text.Encoding]::UTF8; {}", script),
                         ]),
                     };
                 }
@@ -79,13 +79,46 @@ impl Shell {
                             ps.exe.clone(),
                             "-NoProfile".to_string(),
                             "-Command".to_string(),
-                            arg,
+                            format!("[Console]::OutputEncoding = [System.Text.Encoding]::UTF8; {}", arg),
                         ]
                     });
                 }
 
                 // Model generated a PowerShell command. Run it.
-                Some(command)
+                // If it's already a PowerShell command, we need to ensure UTF-8 encoding is set.
+                // Find the -Command argument and modify the argument that follows it.
+                if let Some(command_index) = command.iter().position(|arg| arg == "-Command") {
+                    let mut modified_command = command.clone();
+                    if let Some(cmd_arg) = modified_command.get_mut(command_index + 1) {
+                        // Special case: preserve stdin sentinel "-"
+                        if cmd_arg == "-" {
+                            // For stdin input, we can't prepend to the argument itself.
+                            // Instead, we need to handle UTF-8 encoding differently or skip it.
+                            // For now, preserve the original behavior for stdin.
+                            return Some(command);
+                        }
+
+                        // Special case: stop-parsing token "--%"
+                        // The token must be the first token after -Command and only works in command-line context.
+                        // We cannot inject encoding setup inside the command without breaking stop-parsing semantics.
+                        // Instead, we set the encoding at the PowerShell process level using -OutputEncoding.
+                        if cmd_arg.trim_start().starts_with("--%") {
+                            // Insert -OutputEncoding utf8 before -Command
+                            modified_command.insert(command_index, "-OutputEncoding".to_string());
+                            modified_command.insert(command_index + 1, "utf8".to_string());
+                            return Some(modified_command);
+                        }
+
+                        // Normal case: prepend UTF-8 encoding setup to the command
+                        *cmd_arg = format!(
+                            "[Console]::OutputEncoding = [System.Text.Encoding]::UTF8; {}",
+                            cmd_arg
+                        );
+                    }
+                    Some(modified_command)
+                } else {
+                    Some(command)
+                }
             }
             Shell::Unknown => None,
         }
@@ -490,15 +523,15 @@ mod tests_windows {
                     bash_exe_fallback: None,
                 }),
                 vec!["bash", "-lc", "echo hello"],
-                vec!["pwsh.exe", "-NoProfile", "-Command", "echo hello"],
+                vec!["pwsh.exe", "-NoProfile", "-Command", "[Console]::OutputEncoding = [System.Text.Encoding]::UTF8; echo hello"],
             ),
             (
                 Shell::PowerShell(PowerShellConfig {
                     exe: "powershell.exe".to_string(),
                     bash_exe_fallback: None,
                 }),
                 vec!["bash", "-lc", "echo hello"],
-                vec!["powershell.exe", "-NoProfile", "-Command", "echo hello"],
+                vec!["powershell.exe", "-NoProfile", "-Command", "[Console]::OutputEncoding = [System.Text.Encoding]::UTF8; echo hello"],
             ),
             (
                 Shell::PowerShell(PowerShellConfig {
@@ -530,15 +563,15 @@ mod tests_windows {
                     bash_exe_fallback: Some(PathBuf::from("bash.exe")),
                 }),
                 vec!["echo", "hello"],
-                vec!["pwsh.exe", "-NoProfile", "-Command", "echo hello"],
+                vec!["pwsh.exe", "-NoProfile", "-Command", "[Console]::OutputEncoding = [System.Text.Encoding]::UTF8; echo hello"],
             ),
             (
                 Shell::PowerShell(PowerShellConfig {
                     exe: "pwsh.exe".to_string(),
                     bash_exe_fallback: Some(PathBuf::from("bash.exe")),
                 }),
                 vec!["pwsh.exe", "-NoProfile", "-Command", "echo hello"],
-                vec!["pwsh.exe", "-NoProfile", "-Command", "echo hello"],
+                vec!["pwsh.exe", "-NoProfile", "-Command", "[Console]::OutputEncoding = [System.Text.Encoding]::UTF8; echo hello"],
             ),
             (
                 // TODO (CODEX_2900): Handle escaping newlines for powershell invocation.
@@ -568,4 +601,86 @@ mod tests_windows {
             );
         }
     }
+
+    #[test]
+    fn test_powershell_utf8_encoding_setup() {
+        let shell = Shell::PowerShell(PowerShellConfig {
+            exe: "pwsh.exe".to_string(),
+            bash_exe_fallback: None,
+        });
+
+        // Test 1: Simple command conversion gets UTF-8 encoding
+        let input = vec!["echo".to_string(), "Turkish: çğıİöşü".to_string()];
+        let actual = shell.format_default_shell_invocation(input);
+
+        assert!(actual.is_some());
+        let cmd = actual.unwrap();
+        assert_eq!(cmd.len(), 4);
+        assert_eq!(cmd[0], "pwsh.exe");
+        assert_eq!(cmd[1], "-NoProfile");
+        assert_eq!(cmd[2], "-Command");
+        assert!(cmd[3].starts_with("[Console]::OutputEncoding = [System.Text.Encoding]::UTF8;"));
+        assert!(cmd[3].contains("Turkish: çğıİöşü"));
+
+        // Test 2: Existing PowerShell command with -NoProfile -Command gets UTF-8 encoding
+        let input = vec![
+            "pwsh.exe".to_string(),
+            "-NoProfile".to_string(),
+            "-Command".to_string(),
+            "Get-Content response.md".to_string(),
+        ];
+        let actual = shell.format_default_shell_invocation(input);
+
+        assert!(actual.is_some());
+        let cmd = actual.unwrap();
+        assert_eq!(cmd.len(), 4);
+        assert_eq!(cmd[0], "pwsh.exe");
+        assert_eq!(cmd[1], "-NoProfile");
+        assert_eq!(cmd[2], "-Command");
+        assert!(cmd[3].starts_with("[Console]::OutputEncoding = [System.Text.Encoding]::UTF8;"));
+        assert!(cmd[3].contains("Get-Content response.md"));
+
+        // Test 3: PowerShell command without -Command flag should pass through unchanged
+        let input = vec![
+            "pwsh.exe".to_string(),
+            "-Version".to_string(),
+        ];
+        let actual = shell.format_default_shell_invocation(input);
+
+        assert!(actual.is_some());
+        let cmd = actual.unwrap();
+        assert_eq!(cmd, input); // Should be unchanged
+
+        // Test 4: PowerShell command with stdin sentinel "-" should preserve original behavior
+        let input = vec![
+            "pwsh.exe".to_string(),
+            "-NoProfile".to_string(),
+            "-Command".to_string(),
+            "-".to_string(),
+        ];
+        let actual = shell.format_default_shell_invocation(input);
+
+        assert!(actual.is_some());
+        let cmd = actual.unwrap();
+        assert_eq!(cmd, input); // Should be unchanged to preserve stdin functionality
+
+        // Test 5: PowerShell command with stop-parsing token should use -OutputEncoding
+        let input = vec![
+            "pwsh.exe".to_string(),
+            "-NoProfile".to_string(),
+            "-Command".to_string(),
+            "--% git commit -m \"Turkish: şğı\"".to_string(),
+        ];
+        let actual = shell.format_default_shell_invocation(input);
+
+        assert!(actual.is_some());
+        let cmd = actual.unwrap();
+        assert_eq!(cmd.len(), 6);
+        assert_eq!(cmd[0], "pwsh.exe");
+        assert_eq!(cmd[1], "-NoProfile");
+        assert_eq!(cmd[2], "-OutputEncoding");
+        assert_eq!(cmd[3], "utf8");
+        assert_eq!(cmd[4], "-Command");
+        assert_eq!(cmd[5], "--% git commit -m \"Turkish: şğı\"");
+    }
 }