New REPL: `python -q` outputs unnecessary blank space on empty terminal

[jfly]

Bug report

Bug description:

If I start the new REPL on a clear terminal, I get an unnecessary blank space.

clear; python -q

Whereas this does not happen if I run the old ("basic") REPL:

clear; PYTHON_BASIC_REPL=1 python -q

Analysis

I've tracked this down to these 2 lines of code: https://github.com/python/cpython/blob/3.13/Lib/_pyrepl/unix_console.py#L241-L242

self.__move(0, len(self.screen) - 1)
self.__write("\n")

I don't totally understand this code, but it certainly smells like there's a bug here when dealing with an empty screen (for example, if len(self.screen) == 0, are we trying to move the cursor to index -1?)

I'm not sure what the fix is here, though. I don't really understand this code: why are we moving up to a line just to move the cursor down a line afterwards? why can't we jump straight to where we want to be? I tried digging through the git history for answers, but it looks like this incantation dates back all the way to this initial commit to pypi/pyrepl. I'm not sure how to dig any further back in time.

CPython versions tested on:

3.13, 3.14, CPython main branch

Operating systems tested on:

Linux

I'm using Alacritty, in case that's relevant.

Linked PRs

• gh-131743: Add screen length check to console.refresh in _pyrepl #131907

[skirpichev]

I'm using Alacritty, in case that's relevant.

I think so. I can't reproduce this in gterm.

[jfly]

@skirpichev, what is gterm? Is it gnome-terminal? I am able to reproduce the issue with gnome-terminal.

[skirpichev]

I am able to reproduce the issue with gnome-terminal.

Hmm, my fault. It's important to run commands as one-liner's, i.e. clear;./python -q.

[donBarbos]

this is also reproduced on my system. i use kitty

It would be great to know what the behavior is on Windows

[donBarbos]

it seems that these actions are not needed in some cases, so if we do this inside block if len(self.screen): the problem is not reproduced and the tests still pass (at least for me)

we also have same code in WondowsConsole.refresh method, i think we should fix that too

if no one minds i'll send PR, and it would be nice to make backport for 3.13

[jfly]

That sounds like a safe change to me, @donBarbos.

I'm curious though: can you explain why we ever need to do this? Why are we jumping 1 line above where we need to go just to go one 1 down right after? Why are we even jumping the cursor around at all? (I don't think we're redrawing the entire screen, are we?)

[donBarbos]

for example when we want to move to the next line after entering : for some construction

but I'm not an expert :-) it's better to ask the author or some of core team

[jfly]

for example when we want to move to the next line after entering : for some construction

I still don't understand why we wouldn't jump straight to where we want to go. Why one above and then down?

[donBarbos]

cc @pablogsal as author of new python repl

sorry for ping, do you mind?

[StanFromIreland]

Possibly also causing this behavior:

$ ./python
Python 3.14.0a7+ (heads/main-dirty:1e5798e3725, Apr 13 2025, 08:52:57) [GCC 14.2.1 20250110 (Red Hat 14.2.1-7)] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> 
[1]+  Stopped                 ./python
$ fg
./python
test
>>> test
Traceback (most recent call last):
  File "<python-input-0>", line 1, in <module>
    test
NameError: name 'test' is not defined
>>> 

[jfly]

Sorry, what are you showing here, @StanFromIreland?