setupterm exit silently on Python 3.12

[irvinren]

after pip install windows-curses==2.3.2 setupterm failed, python process exit silently.

from _curses import setupterm
setupterm(term="xterm",fd=1)

[irvinren]

I am running on windows 10 with the Cpython 3.12 downloaded from python.org

[irvinren]

can you speed up to deal with this issue? It impacts my software development.

[irvinren]

in the curses module of cpython 3.12,

the function initscr() will call setupterm which quit silently on my machine.

cat d:\Python312\Lib\curses\init.py

`def initscr():
import _curses, curses
# we call setupterm() here because it raises an error
# instead of calling exit() in error cases.
setupterm(term=_os.environ.get("TERM", "unknown"),
fd=_sys.stdout.fileno())
stdscr = _curses.initscr()
for key, value in curses.dict.items():
if key[0:4] == 'ACS' or key in ('LINES', 'COLS'):
setattr(curses, key, value)

return stdscr`

[jasonbu]

same issue, work all right on my python3.11, exit silently on python3.12

[kartben]

@gschizas would you have any insights?

[irvinren]

@stephanosio isn't it a bug? Is this project still active?

[stephanosio]

@stephanosio isn't it a bug? Is this project still active?

@irvinren Hi. The original author of this project is no longer active and this project is more or less on life support. I currently lack the bandwidth to look into this issue at the moment -- any help from a third party would be greatly appreciated.

[irvinren]

@stephanosio isn't it a bug? Is this project still active?

@irvinren Hi. The original author of this project is no longer active and this project is more or less on life support. I currently lack the bandwidth to look into this issue at the moment -- any help from a third party would be greatly appreciated.

Is there anything I can do to help you? To be frank, I could support it to some degree if it doesn't cause too much effort.

[edreamleo]

Hello all. I am the author of Leo. Leo's cursesGui2 plugin suffers the same problem.

There is an easy workaround. The cursesGui2 plugin contains only one call to curses.initscr(). I replaced it as follows:

if 1:  # Call our own version of curses.initscr().
    import _curses
    # This crashes on Python 3.12.
        # setupterm(term=_os.environ.get("TERM", "unknown"),
            # fd=_sys.__stdout__.fileno())
    stdscr = _curses.initscr()
    for key, value in _curses.__dict__.items():
        if key[0:4] == 'ACS_' or key in ('LINES', 'COLS'):
            setattr(curses, key, value)
    # return stdscr
else:
    stdscr = curses.initscr()

If your program calls curses.initscr from multiple places, wrap the code above in a function.

I tested this hack on Python 3.9, 3.11, 3.12.

HTH.

[edreamleo]

The _curses module defines setupterm, so _curses.cp312-win_amd64.pyd seems the likely culprit.

Probably a bad build somehow.

[gotyaoi]

I don't know if this is for sure the culprit, but it's possible that the generated argument clinic code should be regenerated, since the error is a memory access violation in PyArg_UnpackKeywords. A test run of clinic.py produces a very different _cursesmodule.c.h than the one committed currently.

[gotyaoi]

Seems like I was right. With a small patch, I was able to build a new working 3.12 wheel after running clinic.py to regenerate the clinic headers.

diff --git a/py312/_cursesmodule.c b/py312/_cursesmodule.c
index 4437114..102bb41 100644
--- a/py312/_cursesmodule.c
+++ b/py312/_cursesmodule.c
@@ -1119,9 +1119,9 @@ _curses_window_border_impl(PyCursesWindowObject *self, PyObject *ls,
 _curses.window.box

     [
-    verch: object(c_default="_PyLong_GetZero()") = 0
+    verch: object(c_default="PyLong_FromLong(0)") = 0
         Left and right side.
-    horch: object(c_default="_PyLong_GetZero()") = 0
+    horch: object(c_default="PyLong_FromLong(0)") = 0
         Top and bottom side.
     ]
     /

The current clinic header for _cursesmodule.c was manually modified to include that change, so this just slips it into the source instead.

[gotyaoi]

@stephanosio I could put in a PR, if that's desired?

[stephanosio]

@stephanosio I could put in a PR, if that's desired?

That would be greatly appreciated.

[stephanosio]

Leaving this issue open until it is confirmed to be fixed by others.

[stephanosio]

v2.3.3a1 has been released with the proposed fix from #53. Please test it out and let me know if it fixes the issue.

[LewisNeal]

v2.3.3a1 worked for me running python 3.12.2

[stephanosio]

v2.3.3, which fixes this issue, has been released.