Issue 893250: curses getkey() crash in raw mode (original) (raw)

Created on 2004-02-09 07:11 by bogjohnson, last changed 2022-04-11 14:56 by admin. This issue is now closed.

Messages (5)
msg19946 - (view)	Author: J. David Lee (bogjohnson)	Date: 2004-02-09 07:11
Using python 2.3.3 on gentoo I have the following problem: When calling stdscr.getkey() in raw mode, the program crashes on a terminal resize event. In cooked and cbreak modes, the proper string is returned - "KEY_RESIZE". This problem appeared after upgrading from python 2.3.2, I believe. Below is a quick program that exhibits this behavior on my machine. ####################################################### #!/usr/bin/env python import curses; stdscr = curses.initscr(); curses.raw(); curses.noecho(); stdscr.keypad(1); while(1): stdscr.clear(); stdscr.addstr("Enter key: "); c = stdscr.getkey(); if(c == 'q'): break; stdscr.clear(); stdscr.addstr('"' + c + '"\n\n'); stdscr.addstr("Press any key..."); stdscr.getkey(); curses.noraw(); curses.endwin(); ####################################################### A couple of other notes: 1) No exception is thrown (a try block doesn't catch it). 2) The behavior is the same using the interactive interpreter. 3) The traceback is: File "./keyname", line 19, in ? stdscr.getkey(); _curses.error: no input
msg19947 - (view)	Author: A.M. Kuchling (akuchling) *	Date: 2004-04-19 12:45
Logged In: YES user_id=11375 1) and 3) are contradictory; 3 is certainly an exception traceback. When I try the test program on a Debian system, I get the same error, but a try:except: around the getkey() on line 19 lets the program continue running. Nothing in the Python curses module changed between 2.3.3 and 2.3.2. Did gentoo switch to a new revision of the ncurses library, perhaps?
msg19948 - (view)	Author: J. David Lee (bogjohnson)	Date: 2004-04-20 05:21
Logged In: YES user_id=971407 That Gentoo updated the ncurses library is quite possible. Though the behavior is still incorrect as the documentation says that getkey will return "KEY_RESIZE" on a terminal resize event, and in fact it does in cooked and cbreak modes, but throws an exception when in raw mode (though a terminal resize isn't an exceptional circumstance). Also, a program can't assume that an exception on getkey() is a resize event because an actual exception might occur and need to be dealt with. For the present, though, the workaround is acceptable because a real exception on getkey is very unlikely, as far as I can see.
msg19949 - (view)	Author: A.M. Kuchling (akuchling) *	Date: 2006-11-07 14:32
Logged In: YES user_id=11375 After experimenting a bit with the attached test program, it doesn't seem as if this is actually a bug in the module; the ncurses docs themselves are vague on the interaction between signals and a KEY_RESIZE. So I'll close this bug, because it doesn't look like there's anything in Python to fix.
msg19950 - (view)	Author: A.M. Kuchling (akuchling) *	Date: 2006-11-07 14:36
Logged In: YES user_id=11375 Forgot to attach the script I was using.