Error `Bad file descriptor` in `prompt_toolkit` on ctrl+C interrupt about xonsh

[anki-code]

Is there a way to repeat this?

[clstaudt]

I am not sure what changed for the crash to appear.

The crashing version is

➤ xonsh.version
'0.14.2'

[Qyriad]

Does it happen consistently for you though?

[clstaudt]

It happens often enough so that I had to switch my shell back to fish. But not reproducible, I don't know what exactly triggers the crash.

[anki-code]

Please add the xonfig command output.

[clstaudt]

➤ xonfig
+------------------+---------------------+
| xonsh            | 0.14.2              |
| Python           | 3.12.0              |
| PLY              | 3.11                |
| have readline    | True                |
| prompt toolkit   | 3.0.41              |
| shell type       | prompt_toolkit      |
| history backend  | json                |
| pygments         | 2.17.2              |
| on posix         | True                |
| on linux         | False               |
| on darwin        | True                |
| on windows       | False               |
| on cygwin        | False               |
| on msys2         | False               |
| is superuser     | False               |
| default encoding | utf-8               |
| xonsh encoding   | utf-8               |
| encoding errors  | surrogateescape     |
| xontrib          | []                  |
| RC file 1        | /Users/cls/.xonshrc |
+------------------+---------------------+

[mltucker]

I have also started getting this error after a recent update (installed via homebrew on an m1 macbook). In my case it's not the ctrl+c interrupt that seems to be the problem, but just now and then running a command (most recently git diff HEAD and even ls ~/Library/Logs/Homebrew) will just hang for 10 seconds before I cancel it with ctrl+c (though it will hang for hours if left alone). Then I'll get a similar stack trace like @clstaudt and then I'll restart xonsh.

Mostly this happened to me within tmux, but I was also able to reproduce it in a new Terminal.app window without tmux.

[anki-code]

Hi @mltucker! I'm also mac m1 user and xonsh installed from conda and everything is perfect for me. Could you please alsoo paste the xonfig output.

[mltucker]

@anki-code

+------------------+-------------------+
| xonsh            | 0.14.3            |
| Python           | 3.12.1            |
| PLY              | 3.11              |
| have readline    | True              |
| prompt toolkit   | 3.0.41            |
| shell type       | prompt_toolkit    |
| history backend  | json              |
| pygments         | 2.17.2            |
| on posix         | True              |
| on linux         | False             |
| on darwin        | True              |
| on windows       | False             |
| on cygwin        | False             |
| on msys2         | False             |
| is superuser     | False             |
| default encoding | utf-8             |
| xonsh encoding   | utf-8             |
| encoding errors  | surrogateescape   |
| xontrib          | []                |
| RC file 1        | /Users/m/.xonshrc |
+------------------+-------------------+

I use zoxide but otherwise my xonshrc is barebones.

[anki-code]

Thanks @mltucker ! I'm using py 3.11 but you and the reporter are on 3.12. Maybe this is an upstream issue around prompt-toolkit and asyncio on 3.12. It will be cool to have a try to repeat this issue on 3.11 to confirm this.

[mltucker]

@anki-code it's very strange because my python --version gives me 3.11.7 and brew edit xonsh says it depends on [email protected]. I see that I have both 3.11 and 3.12 installed. I can't uninstall [email protected] since "it is required by llvm and xonsh, which are currently installed". Is there a way I can tell xonsh which python to use?

[anki-code]

@mltucker please read the section about xonsh and package managers to understand how it works. Read as much as possible examples after this section.

[mltucker]

@anki-code thanks! I've uninstalled xonsh from homebrew and instead now run it from conda3. I believe I'm seeing the same issue, though now it looks a bit different. Sometimes commands (most recently git add -p) will just hang. Ctrl+C closes the tmux window instead of quitting the process and showing the error message that @clstaudt reported above (which I used to also see). I will report back.

[mltucker]

@anki-code OK, I reproduced it again and this time recorded my screen. Just for a frame or two the error message popped up. Here's a screenshot:

Possibly the reason the message briefly flashed and closed my terminal window is that instead of fixing my path issues, I made a script /usr/local/bin/xonsh with:

#!/bin/bash

/opt/miniconda3/bin/xonsh $@

Here's my xonfig, this time with python 3.11:

+------------------+-------------------+
| xonsh            | 0.14.4            |
| Python           | 3.11.7            |
| PLY              | 3.11              |
| have readline    | True              |
| prompt toolkit   | 3.0.42            |
| shell type       | prompt_toolkit    |
| history backend  | json              |
| pygments         | 2.17.2            |
| on posix         | True              |
| on linux         | False             |
| on darwin        | True              |
| on windows       | False             |
| on cygwin        | False             |
| on msys2         | False             |
| is superuser     | False             |
| default encoding | utf-8             |
| xonsh encoding   | utf-8             |
| encoding errors  | surrogateescape   |
| xontrib          | []                |
| RC file 1        | /Users/m/.xonshrc |
+------------------+-------------------+

[anki-code]

Weird error with asyncio.

I made a script /usr/local/bin/xonsh

It's not needed if you're doing right. Basically if you want to just have a shell you can install xonsh from brew's python via pip and have it in PATH:

# zsh
cat ~/.zshrc | grep brew
# eval "$(/opt/homebrew/bin/brew shellenv)"

which xonsh 
# /opt/homebrew/bin/xonsh

which python
# /opt/homebrew/bin/python

python -V
# Python 3.11.6

xonsh
[p for p in $PATH if 'brew/bin' in p]
# ['/opt/homebrew/bin']

This allows to run xonsh from zsh.

[anki-code]

OR if you're using conda:

conda activate base

[p for p in $PATH if 'conda' in p]
# /Users/user/miniconda3/bin

wooooooooooow
during I prepared the example to you I found that after conda activate base the Ctrl+C command hangs the terminal.
It's soooooo weird :)

Yeah, I can repeat this with this stack:

Traceback (most recent call last):
  File "/opt/homebrew/lib/python3.11/site-packages/xonsh/procs/posix.py", line 310, in _signal_int
    signal.pthread_kill(threading.get_ident(), signal.SIGINT)
  File "/opt/homebrew/lib/python3.11/site-packages/xonsh/procs/posix.py", line 310, in _signal_int
    signal.pthread_kill(threading.get_ident(), signal.SIGINT)
  File "/opt/homebrew/lib/python3.11/site-packages/xonsh/procs/posix.py", line 310, in _signal_int
    signal.pthread_kill(threading.get_ident(), signal.SIGINT)
  [Previous line repeated 994 more times]
  File "/opt/homebrew/lib/python3.11/site-packages/xonsh/procs/posix.py", line 309, in _signal_int
    if xt.on_main_thread() and not xp.ON_WINDOWS:
       ^^^^^^^^^^^^^^^^^^^
  File "/opt/homebrew/lib/python3.11/site-packages/xonsh/tools.py", line 1234, in on_main_thread
    return threading.current_thread() is threading.main_thread()
           ^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/opt/homebrew/Cellar/[email protected]/3.11.6/Frameworks/Python.framework/Versions/3.11/lib/python3.11/threading.py", line 1461, in current_thread
    return _active[get_ident()]
                   ^^^^^^^^^^^
RecursionError: maximum recursion depth exceeded while calling a Python object

I can't repeat this on xonsh --no-rc

[anki-code]

I found that issue appeared starting from prompt-toolkit 3.0.41 where asyncio was introduced (prompt-toolkit/python-prompt-toolkit@8c60193).

To fix this just downgrade the prompt-toolkit version:

xpip uninstall -y prompt-toolkit
xpip install prompt-toolkit==3.0.40

The issue was reported to upstream - prompt-toolkit/python-prompt-toolkit#1853