I tried this change with the script:

fromcursesimportwrapperimportctypesdefmain(window):window.nodelay(True)returnwindow.getch()print(wrapper(main))

Before (without this change), it displays-1.

After (with this change), it fails with the exception:_curses.error: wgetch() (called by getch()): no input.

So this change is backward incompatible. Is it deliberate?

Hum. No it's not meant to be broken!

Only the signals should be handled, but otherwise we shouldn't raise an exception.

Now I am bit conflicted. Ideally, we should return -1 for everythingor raise an exception as in the other functions. I don't know why it was historically chosen to return -1 forgetch but not forget_wch in no-delay mode.

However, since I don't want to break stuff, I don't know how we can make a smooth transition, except by raising a warning before returning -1.

For now, let's keep the same behavior (namely, -1 forgetch, and exception forget_wch andgetkey). The functions have been there at least since Python 3.4 and I prefer not to change anything unless a more pressing issue exists. At least, signal handlers will be properly called.

I could backport this as it's half a bugfix and half a feature. I'll only backport this to 3.14 as the layout of 3.13 is very different for curses (there was a conversion from static types to heap types and I don't think it made it to 3.13)

You might backport the change to 3.13 and 3.14 bugfix branches.

Thanks@picnixz for the PR 🌮🎉.. I'm working now to backport this PR to: 3.14.
🐍🍒⛏🤖

GH-134783 is a backport of this pull request to the3.14 branch.

GH-134784 is a backport of this pull request to the3.13 branch.