Report bytes out-of-range elements as CPython does (#8679)

* Report `bytes` out-of-range elements as CPython does

`collect_bytes` is shared by the `bytes` and `bytearray` entry points and
hardcoded one element error, so `bytes([256])` said "byte must be in
range(0, 256)" where CPython says "bytes". Every `bytearray` path does say
"byte", so only the one caller was wrong.

The element error is now a parameter, alongside the type error this
function already varies per caller. Verified against CPython 3.14 for all
seven paths.

Assisted-by: Claude Code:claude-opus-5

* Keep the bytearray wording in PyByteArray_FromObject

It called `bytes_from_object`, so the previous commit made it report
"bytes must be in range(0, 256)" where CPython's `PyByteArray_FromObject`
says "byte". It is the C-level bytearray constructor, so it now uses
`bytearray_from_object`, which carries that wording and the constructor's
unsized iteration.

Assisted-by: Claude Code:claude-opus-5

crates/capi/src/bytearrayobject.rs

@@ -3,7 +3,7 @@ use crate::object::define_py_check;
 use crate::pystate::with_vm;
 use core::ffi::c_char;
 use rustpython_vm::builtins::PyByteArray;
-use rustpython_vm::byte::bytes_from_object;
+use rustpython_vm::byte::bytearray_from_object;
 
 define_py_check!(fn PyByteArray_Check, types.bytearray_type);
 
@@ -41,7 +41,7 @@ pub unsafe extern "C" fn PyByteArray_FromStringAndSize(
 pub unsafe extern "C" fn PyByteArray_FromObject(obj: *mut PyObject) -> *mut PyObject {
     with_vm(|vm| {
         let obj = unsafe { &*obj };
-        let data = bytes_from_object(vm, obj)?;
+        let data = bytearray_from_object(vm, obj)?;
         Ok(vm.ctx.new_bytearray(data))
     })
 }

crates/vm/src/byte.rs

@@ -7,9 +7,15 @@ use crate::{
     protocol::{BufferFlags, PyBuffer},
 };
 
+/// The element error `bytes` reports, which is the one entry point here that does
+/// not name a single byte: `bytes([256])` says "bytes must be in range(0, 256)"
+/// where every `bytearray` path says "byte".
+const BYTES_ELEMENT_ERROR: &str = "bytes must be in range(0, 256)";
+const BYTE_ELEMENT_ERROR: &str = "byte must be in range(0, 256)";
+
 // PyBytes_FromObject
 pub fn bytes_from_object(vm: &VirtualMachine, obj: &PyObject) -> PyResult<Vec<u8>> {
-    collect_bytes(vm, obj, true, |name| {
+    collect_bytes(vm, obj, true, BYTES_ELEMENT_ERROR, |name| {
         format!("cannot convert '{name}' object to bytes")
     })
 }
@@ -18,7 +24,7 @@ pub fn bytes_from_object(vm: &VirtualMachine, obj: &PyObject) -> PyResult<Vec<u8
 /// slice of one, which run the iterator without asking the object they were
 /// handed how long it is.
 pub fn bytearray_from_object(vm: &VirtualMachine, obj: &PyObject) -> PyResult<Vec<u8>> {
-    collect_bytes(vm, obj, false, |name| {
+    collect_bytes(vm, obj, false, BYTE_ELEMENT_ERROR, |name| {
         format!("cannot convert '{name}' object to bytearray")
     })
 }
@@ -26,17 +32,19 @@ pub fn bytearray_from_object(vm: &VirtualMachine, obj: &PyObject) -> PyResult<Ve
 /// [`bytes_from_object`] for `bytearray_extend()`, which names what it was
 /// doing rather than what it was converting to.
 pub fn bytearray_extend_from_object(vm: &VirtualMachine, obj: &PyObject) -> PyResult<Vec<u8>> {
-    collect_bytes(vm, obj, true, |name| {
+    collect_bytes(vm, obj, true, BYTE_ELEMENT_ERROR, |name| {
         format!("can't extend bytearray with {name}")
     })
 }
 
-/// `measured` is whether the object is asked how long it is; `unusable` names,
-/// from the class name, what could not be done with one that is not iterable.
+/// `measured` is whether the object is asked how long it is; `element` is the error
+/// for a value outside `range(0, 256)`; `unusable` names, from the class name, what
+/// could not be done with one that is not iterable.
 fn collect_bytes(
     vm: &VirtualMachine,
     obj: &PyObject,
     measured: bool,
+    element: &'static str,
     unusable: impl FnOnce(&str) -> String,
 ) -> PyResult<Vec<u8>> {
     if obj.check_buffer() {
@@ -52,7 +60,7 @@ fn collect_bytes(
         if cls.slots.iter.load().is_none() && !cls.has_attr(identifier!(vm, __getitem__)) {
             return Err(vm.new_type_error(unusable(&cls.name())));
         }
-        let value = |x: PyObjectRef| value_from_object(vm, &x);
+        let value = |x: PyObjectRef| value_from_object_with(vm, &x, element);
         let elements = if measured {
             vm.map_iterable_object_sized(obj, value)
         } else {
@@ -67,8 +75,16 @@ fn collect_bytes(
 }
 
 pub fn value_from_object(vm: &VirtualMachine, obj: &PyObject) -> PyResult<u8> {
+    value_from_object_with(vm, obj, BYTE_ELEMENT_ERROR)
+}
+
+fn value_from_object_with(
+    vm: &VirtualMachine,
+    obj: &PyObject,
+    element: &'static str,
+) -> PyResult<u8> {
     obj.try_index(vm)?
         .as_bigint()
         .to_u8()
-        .ok_or_else(|| vm.new_value_error("byte must be in range(0, 256)"))
+        .ok_or_else(|| vm.new_value_error(element))
 }

extra_tests/snippets/builtin_bytes.py

@@ -813,3 +813,22 @@ cannot(
     lambda: bytearray(b"ab").__setitem__(slice(0, 2), "ab"),
     "can assign only bytes, buffers, or iterables of ints in range(0, 256)",
 )
+
+
+def out_of_range(fn, message):
+    try:
+        fn()
+    except ValueError as e:
+        assert str(e) == message, e
+    else:
+        raise AssertionError(f"expected ValueError: {message}")
+
+
+# `bytes` is the one entry point that does not name a single byte, and both the
+# sized and unsized iterator paths report it that way.
+out_of_range(lambda: bytes([256]), "bytes must be in range(0, 256)")
+out_of_range(lambda: bytes(iter([256])), "bytes must be in range(0, 256)")
+out_of_range(lambda: bytes([-1]), "bytes must be in range(0, 256)")
+out_of_range(lambda: bytearray([256]), "byte must be in range(0, 256)")
+out_of_range(lambda: bytearray(iter([256])), "byte must be in range(0, 256)")
+out_of_range(lambda: bytearray().extend([256]), "byte must be in range(0, 256)")