New REPL on Windows exits when accented character is pasted/typed

[devdanzin]

Crash report

What happened?

When pasting an accented character like ñ or é, or typing one such as ´p and pressing Enter, then up arrow, the new REPL on Windows will exit with the following error:

Python 3.14.0a6+ (heads/fix_non_string_candidates_exit_repl-dirty:8008a588342, Mar 29 2025, 08:15:) [MSC v.1943 64 bit (AMD64)] on win32
Type "help", "copyright", "credits" or "license" for more information.
>>> Traceback (most recent call last):
 File "<frozen runpy>", line 198, in _run_module_as_main
 File "<frozen runpy>", line 88, in _run_code
 File "C:\Users\ddini\PycharmProjects\cpython\Lib\_pyrepl\__main__.py", line 6, in <module>
 __pyrepl_interactive_console()
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^
 File "C:\Users\ddini\PycharmProjects\cpython\Lib\_pyrepl\main.py", line 59, in interactive_console
 run_multiline_interactive_console(console)
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^
 File "C:\Users\ddini\PycharmProjects\cpython\Lib\_pyrepl\simple_interact.py", line 137, in run_multiline_interactive_console
 statement = multiline_input(more_lines, ps1, ps2)
 File "C:\Users\ddini\PycharmProjects\cpython\Lib\_pyrepl\readline.py", line 389, in multiline_input
 return reader.readline()
 ~~~~~~~~~~~~~~~^^
 File "C:\Users\ddini\PycharmProjects\cpython\Lib\_pyrepl\reader.py", line 750, in readline
 self.handle1()
 ~~~~~~~~~~~~^^
 File "C:\Users\ddini\PycharmProjects\cpython\Lib\_pyrepl\reader.py", line 706, in handle1
 event = self.console.get_event(block=False)
 File "C:\Users\ddini\PycharmProjects\cpython\Lib\_pyrepl\windows_console.py", line 471, in get_event
 self.event_queue.push(rec.Event.KeyEvent.uChar.UnicodeChar)
 ~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 File "C:\Users\ddini\PycharmProjects\cpython\Lib\_pyrepl\base_eventqueue.py", line 88, in push
 assert len(self.buf) == 1
 ^^^^^^^^^^^^^^^^^^
AssertionError

In these cases, len(self.buf) is 3 instead of 1.

I'm running Python in Windows Terminal, chcp gives 850, changing it to 65001 (UTF-8) makes no difference.

CPython versions tested on:

CPython main branch

Operating systems tested on:

Windows

Output from running 'python -VV' on the command line:

Python 3.14.0a6+ (heads/main-dirty:7c3692fe275, Mar 24 2025, 19:20:37) [MSC v.1942 64 bit (AMD64)]

Linked PRs

• gh-131878: Fix input of unicode characters with two or more code points in new pyrepl on Windows #131901
• gh-131878: Handle top level exceptions in new pyrepl and prevent of closing it #131910
• gh-131878: Fix input of unicode characters with two or more code points in the REPL on Windows in vt mode #133030
• [3.13] gh-131878: Handle top level exceptions in new pyrepl and prevent of closing it (GH-131910) #133445
• [3.13] gh-131878: Fix input of unicode characters with two or more code points in new pyrepl on Windows (gh-131901) #133468

[StanFromIreland]

Related? #130273

This is a bug and not a crash.

[devdanzin]

This is a bug and not a crash.

Correct, sorry about that.

cc @DinoV, @y5c4l3, @encukou

[StanFromIreland]

This is os-windows specific. IIRC there was recently an issue about this, it was closed.

[StanFromIreland]

Yes looking at #130804 it seems it was not fully resolved, and it is presumably a side effect of the merged pr, do we want to reopen that?

[sergey-miryanov]

@StanFromIreland
It is in another execution path, so I think this should be a new issue (despite it looks similar).

[StanFromIreland]

@sergey-miryanov

cpython/Lib/_pyrepl/base_eventqueue.py

Lines 72 to 89 in fccf9ab

	def push(self, char: int | bytes | str) -> None:
	"""
	Processes a character by updating the buffer and handling special key mappings.
	"""
	ord_char = char if isinstance(char, int) else ord(char)
	if ord_char > 255:
	assert isinstance(char, str)
	char = bytes(char.encode(self.encoding, "replace"))
	self.buf.extend(char)
	else:
	char = bytes(bytearray((ord_char,)))
	self.buf.append(ord_char)
	if char in self.keymap:
	if self.keymap is self.compiled_keymap:
	# sanity check, buffer is empty when a special key comes
	assert len(self.buf) == 1
	k = self.keymap[char]

It is in the same path you modified, your changes are the last ones to modify the buffer before the check?

[sergey-miryanov]

@StanFromIreland Oh, I get what you mean. Yes, it is related, but I think it is related to keymap and compiled_keymap also, so it extends my previous patch a bit. I will test and fix. Can you assign it to me?