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

[srfoster65]

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

[srfoster65]

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

[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?