youknowone · Oct 31, 2026 · Oct 31, 2026
diff --git a/Lib/test/test_cmd_line.py b/Lib/test/test_cmd_line.py
@@ -330,7 +330,6 @@ def test_undecodable_code(self):
         if not stdout.startswith(pattern):
             raise AssertionError("%a doesn't start with %a" % (stdout, pattern))
 
-    @unittest.expectedFailure  # TODO: RUSTPYTHON
     @unittest.skipIf(sys.platform == 'win32',
                      'Windows has a native unicode API')
     def test_invalid_utf8_arg(self):

diff --git a/crates/stdlib/src/syslog.rs b/crates/stdlib/src/syslog.rs
@@ -53,8 +53,10 @@ mod syslog {
         if let Some(argv) = vm.state.config.settings.argv.first()
             && !argv.is_empty()
         {
+            let argv = vm.fsdecode(argv.clone());
+            let argv = argv.as_wtf8();
             return Some(
-                PyStr::from(match argv.find('\\') {
+                PyStr::from(match argv.as_bytes().iter().position(|&b| b == b'\\') {
                     Some(value) => &argv[value..],
                     None => argv,
                 })

diff --git a/crates/vm/src/getpath.rs b/crates/vm/src/getpath.rs
@@ -7,6 +7,7 @@
 
 use crate::vm::{Paths, Settings};
 use std::env;
+use std::ffi::OsString;
 use std::io::{self, Read};
 use std::path::{Path, PathBuf};
 
@@ -622,13 +623,19 @@ fn env_executable_override() -> Option<String> {
 
 /// CALCULATE program_name: `argv[0]`, with the executable suffix added when
 /// only the suffixed file exists.
-fn program_name() -> String {
-    let program_name = env::args()
+fn program_name() -> OsString {
+    let program_name = env::args_os()
         .next()
         .filter(|arg| !arg.is_empty())
-        .unwrap_or_else(|| platform::DEFAULT_PROGRAM_NAME.to_owned());
-    if cfg!(windows) && !program_name.to_lowercase().ends_with(".exe") {
-        let suffixed = format!("{program_name}.exe");
+        .unwrap_or_else(|| platform::DEFAULT_PROGRAM_NAME.into());
+    if cfg!(windows)
+        && !program_name
+            .to_string_lossy()
+            .to_lowercase()
+            .ends_with(".exe")
+    {
+        let mut suffixed = program_name.clone();
+        suffixed.push(".exe");
         if Path::new(&suffixed).is_file() {
             return suffixed;
         }
@@ -651,7 +658,7 @@ fn get_executable_path() -> Option<PathBuf> {
     }
     #[cfg(target_arch = "wasm32")]
     {
-        let exec_arg = env::args().next()?;
+        let exec_arg = env::args_os().next()?;
         Some(PathBuf::from(exec_arg))
     }
 }

diff --git a/crates/vm/src/stdlib/sys.rs b/crates/vm/src/stdlib/sys.rs
@@ -313,7 +313,7 @@ pub mod sys {
             .settings
             .argv
             .iter()
-            .map(|arg| vm.ctx.new_str(arg.clone()).into())
+            .map(|arg| vm.fsdecode(arg.clone()).into())
             .collect()
     }
 
@@ -708,7 +708,7 @@ pub mod sys {
 
     #[pyattr]
     fn orig_argv(vm: &VirtualMachine) -> Vec<PyObjectRef> {
-        env::args().map(|arg| vm.ctx.new_str(arg).into()).collect()
+        env::args_os().map(|arg| vm.fsdecode(arg).into()).collect()
     }
 
     #[pyattr]

diff --git a/crates/vm/src/vm/setting.rs b/crates/vm/src/vm/setting.rs
@@ -5,7 +5,7 @@ use std::ffi::OsString;
 #[derive(Debug, Clone, Default)]
 pub struct Paths {
     /// program_name: the name the interpreter was invoked as
-    pub program_name: String,
+    pub program_name: OsString,
     /// sys.executable
     pub executable: String,
     /// sys._base_executable (original interpreter in venv)
@@ -77,8 +77,8 @@ pub struct Settings {
     pub pycache_prefix: Option<OsString>,
     // int parse_argv;
     // PyWideStringList orig_argv;
-    /// sys.argv
-    pub argv: Vec<String>,
+    /// sys.argv, decoded with the filesystem encoding during sys initialization.
+    pub argv: Vec<OsString>,
 
     // spell-checker:ignore Xfoo
     /// -Xfoo[=bar]

diff --git a/src/lib.rs b/src/lib.rs
@@ -252,7 +252,7 @@ fn run_file(vm: &VirtualMachine, scope: Scope, argv0: &str) -> PyResult<u32> {
 }
 
 fn cant_open_file(vm: &VirtualMachine, path: &str, err: &std::io::Error) -> PyResult<u32> {
-    let program_name = &vm.state.config.paths.program_name;
+    let program_name = vm.fsdecode(vm.state.config.paths.program_name.clone());
     let filename_repr = vm.ctx.new_str(path).as_object().repr(vm)?;
     let errno = err.raw_os_error().unwrap_or(2);
     let strerror = vm::host_env::errno::strerror_string(errno).unwrap_or_default();

diff --git a/src/settings.rs b/src/settings.rs
@@ -5,6 +5,7 @@ use lexopt::Arg::{Long, Short, Value};
 use lexopt::ValueExt;
 use rustpython_vm::{Settings, vm::CheckHashPycsMode};
 use std::env;
+use std::ffi::OsString;
 
 pub enum RunMode {
     Script(String),
@@ -112,13 +113,15 @@ file   : program read from script file
 arg ...: arguments passed to program in sys.argv[1:]
 ";
 
-fn parse_args() -> Result<(CliArgs, RunMode, Vec<String>), lexopt::Error> {
+fn parse_args() -> Result<(CliArgs, RunMode, Vec<OsString>), lexopt::Error> {
     let mut args = CliArgs::default();
     let mut parser = lexopt::Parser::from_env();
-    fn argv(argv0: String, mut parser: lexopt::Parser) -> Result<Vec<String>, lexopt::Error> {
-        core::iter::once(Ok(argv0))
-            .chain(parser.raw_args()?.map(|arg| arg.string()))
-            .collect()
+    fn argv(argv0: String, mut parser: lexopt::Parser) -> Result<Vec<OsString>, lexopt::Error> {
+        // Arguments after the entry point are filesystem-encoded data, not
+        // necessarily Unicode. Decode only when constructing Python strings.
+        Ok(core::iter::once(argv0.into())
+            .chain(parser.raw_args()?)
+            .collect())
     }
     while let Some(arg) = parser.next()? {
         match arg {
@@ -171,15 +174,15 @@ fn parse_args() -> Result<(CliArgs, RunMode, Vec<String>), lexopt::Error> {
 
             Long("install-pip") => {
                 let (mode, argv) = if let Some(val) = parser.optional_value() {
-                    (val.parse()?, vec![val.string()?])
+                    (val.parse()?, vec![val])
                 } else if let Ok(argv0) = parser.value() {
                     let mode = argv0.parse()?;
                     (mode, argv(argv0.string()?, parser)?)
                 } else {
                     (
                         InstallPipMode::Ensurepip,
                         ["ensurepip", "--upgrade", "--default-pip"]
-                            .map(str::to_owned)
+                            .map(OsString::from)
                             .into(),
                     )
                 };