curses cell text is silently truncated at an embedded NUL

[fedonman]

Bug description:

curses.complexchar() keeps only the text before an embedded NUL, so complexchar('a\0\u0301') silently drops the combining acute, and complexchar('\0') builds a cell with no text at all, whose repr() is not a valid constructor call.

import curses

print(repr(curses.complexchar('a\0\u0301')))
print(repr(curses.complexchar('a\u0301')))
r = repr(curses.complexchar('\0'))
print(r)
eval(r, {'curses': curses})

curses.complexchar('a')
curses.complexchar('á')
curses.complexchar('')
Traceback (most recent call last):
  File "/tmp/bug.py", line 7, in <module>
    eval(r, {'curses': curses})
    ~~~~^^^^^^^^^^^^^^^^^^^^^^^
  File "<string>", line 1, in <module>
TypeError: expect a string of 1 to 5 characters, got a str of length 0

The empty cell is worse than the truncation: window.addstr(curses.complexstr([curses.complexchar('\0'), curses.complexchar('B')])) returns None having written nothing. The same text reaches the same converter through window.addch(), insch(), echochar(), hline(), vline(), bkgd(), bkgdset(), border() and box(), and curses.complexstr('a\0b') drops the NUL and returns two cells.

Expected: ValueError: embedded null character, which window.addstr('a\0b') already raises.

CPython versions tested on:

CPython main branch

Operating systems tested on:

Linux

Linked PRs

• gh-155877: Reject an embedded NUL in curses cell text #155878

[serhiy-storchaka]

addch('\0') was always accepted (it is equivalent to addch(0)), so I think addch(complexchar('\0')) should work the same way.

The repr for empty cell is complexchar('\0') in the narrow build, it should be the same in the wide build.

Silent truncation after embedded NUL is not good, we try to detect such errors.