If I build current trunk on my workstation (Linux 64-bit, Ubuntu 13.10) then run

  % ./python -m test -u all test_curses

I get a traceback.  Here's the interesting part:

      File "/home/larry/src/python/buildtrunk/Lib/test/test_curses.py",
        line 118, in window_funcs

        win2.overlay(win, 1, 2, 3, 3, 2, 1)

      _curses.error: copywin() returned ERR

Zach: I just saw you had a patch specifically against test_curses.py, can you take a look?

I'm on Windows until this evening; I'll see what I can see when I'm on a machine that can actually build _curses.

I don't see any failures on any of the buildbots, so I'm not sure how likely it is I'll be able to reproduce your failure.

(And btw, that patch I have against test_curses is just a mechanical conversion to unittest, the only curses I'm an expert at are ones directed at problems...)

I can reproduce this (also on Ubuntu 13.10 64-bit). Maybe there's a bug
in the version of curses distributed with the latest Ubuntu release? It
looks like our only Ubuntu buildbot is using 8.04 (almost 6 years old!).

Also note that you won't be able to reproduce this with "make test" or
"make testall" (see issue 12669). "make buildbottest" does catch the bug,
though (which also rules out the possibility that the buildbots are just
skipping the test).

Ok, I've tracked down where the error is happening, but I don't know enough about curses to suggest where it should go from here.  It does appear to be the version of ncurses that Ubuntu 13.10 uses that's causing problems, version 5.9+20130608.  The issue stems from a change in ncurses/base/lib_overlay.c, here's the relevant part of the diff to copywin (left is vanilla ncurses-5.9 from http://ftp.gnu.org/pub/gnu/ncurses/, right is ncurses-5.9+20130608 obtained by 'sudo apt-get source ncurses'):

--- ncurses-5.9/ncurses/base/lib_overlay.c      2009-10-24 18:21:31.000000000 -0500
+++ ncurses-5.9+20130608/ncurses/base/lib_overlay.c     2014-01-25 17:13:08.461548012 -0600
@@ -151,7 +151,10 @@
        dminrow, dmincol,
        dmaxrow, dmaxcol, over));
 
-    if (src && dst) {
+    if (src != 0
+       && dst != 0
+       && dmaxrow >= dminrow
+       && dmaxcol >= dmincol) {
        _nc_lock_global(curses);
 
        bk = AttrOf(dst->_nc_bkgd);


The failing test passes 1, 2, 3, 3, 2, 1 for sminrow, smincol, dminrow, dmincol, dmaxrow, dmaxcol, which fails the dmaxrow >= dminrow check and copywin returns -1 (ERR).  Changing the test to call with 1, 1, 2, 2, 3, 3 prevents the failure, but does ugly things to the tty when the process ends.

New changeset 0ae768637a07 by Larry Hastings in branch 'default':
Issue #20358: Tests for curses.window.overlay and curses.window.overwrite
http://hg.python.org/cpython/rev/0ae768637a07

I changed the unit test so it doesn't specify a minium row/column greater than the maximum row/column.  The tests now pass.  So I'm calling this good.  Thanks for looking in to it, Zachary and Nadeem!