pdm run does not catch ctrl-c on 2nd run (Win11)

[srfoster65]

• [ x] I have searched the issue tracker and believe that this is not a duplicate.

Steps to reproduce

Given a project that is setup to support mkdocs for generating documentation.
From a windows command shell (cmd), cd to the project root and enter the following commands

pdm run mkdocs serve

[ctrl-c] - To terminate serve process - this works correctly.

pdm run mkdocs serve

[ctrl-c] - Does not interrupt/terminate process

Note:
I have tested the same project with poetry, and this can repeatedly start and stop mkdocs serve with the command "poetry run mkdocs serve".

Actual behaviour

mkdocs serve process continues to run.

Expected behaviour

mkdocs serve process should be interrupted

Environment Information

pdm info
PDM version:
2.9.3
Python Interpreter:
C:\Users\Steve\Development\python\arg_init.venv\Scripts\python.exe (3.11)
Project Root:
C:/Users/Steve/Development/python/arg_init
Local Packages:

{
"implementation_name": "cpython",
"implementation_version": "3.11.3",
"os_name": "nt",
"platform_machine": "AMD64",
"platform_release": "10",
"platform_system": "Windows",
"platform_version": "10.0.22621",
"python_full_version": "3.11.3",
"platform_python_implementation": "CPython",
"python_version": "3.11",
"sys_platform": "win32"
}

[srfoster65]

Something very odd is happening here. Almost certainly windows specific.

The issue is in the file pdm\cli\commands\run.py and function _run_process().

If I simplify the code in _run_process() (This is not a solution, but an attempt to understand program flow) and remove the signal handler function and replace with a try/except block.

   try:
        process = subprocess.Popen(expanded_args, cwd=cwd, env=process_env, shell=shell, bufsize=0)
        process.wait()
    except KeyboardInterrupt:
        process.send_signal(signal.SIGINT) 

This works everytime. Note: this is sending signal.SIGINT to the subporocess.

If I modify the existing forward_signal() function

    def forward_signal(signum: int, frame: FrameType | None) -> None:
        process.send_signal(signum)

This will raise
[ValueError]: Unsupported signal: 2

everytime, but will abort the script.

But why should this raise a ValueError when calling send_signal() in the exception handler with the same value (signal.SIGINT = 2) does not?

[srfoster65]

https://docs.python.org/3/library/signal.html#signal.CTRL_C_EVENT

signal.CTRL_C_EVENT
The signal corresponding to the Ctrl+C keystroke event. This signal can only be used with os.kill().

So the code in forward signal should not work as intended, and may partially explain the observed behaviour.

[srfoster65]

A solution I found on stackoverflow that was used for a similar issue on postgress was to use signum=signal.SIGTERM for windows.

    def forward_signal(signum: int, frame: FrameType | None) -> None:
        if sys.platform == "win32" and signum == signal.SIGINT:
            signum = signal.SIGTERM
        process.send_signal(signum)

I have tested this and this seems to reliably terminate the subprocess.

[frostming]

A solution I found on stackoverflow that was used for a similar issue on postgress was to use signum=signal.SIGTERM for windows.

    def forward_signal(signum: int, frame: FrameType | None) -> None:
        if sys.platform == "win32" and signum == signal.SIGINT:
            signum = signal.SIGTERM
        process.send_signal(signum)

I have tested this and this seems to reliably terminate the subprocess.

This seems to fix the issue without causing any other problems though TBH, I was unable to reproduce it. Would you or @xzmeng like to send a PR?

[srfoster65]

Sure, I will create a PR.
This will be my first PR, so I hope I get things right.