bkgd() with character part set to 0

[Bill Gray]

   I just ran into this issue,  and am not quite sure how it really ought to be 
handled.

   In a perfect world,  I think people would call bkgd() with the character 
part set to a non-zero value,  such as

   bkgd( ' ' | COLOR_PAIR( 0));

   However,  looking through both the ncurses and PDCurses* tests,  I see a 
fair number of calls where the character bits are left blank.  This appears to 
work,  right up to the point where you read a background character using the 
inch() family of functions and then output it using addch(),  as shown in the 
code snippet below.

   As the boilerplate text inside it says,  on wide-char ncurses,  the 0 
background character is read in,  then output as ^@.  I noticed the problem 
when building 'ozdemo' (a PDCurses* test program) with ncurses.

   PDCurses* evades this by saying that if you call bkgd( ) (internally,  
wbkgdset( )) with a zero character,  it gets remapped to a space :

https://github.com/Bill-Gray/PDCursesMod/blob/master/pdcurses/bkgd.c#L153

   I'd tend to think that the behavior with a zero character is 
implementation-dependent,  and mildly risky.  I expect to set the background 
explicitly to a space in future.  But the conversion of 0 to a space 
(apparently done in 8-bit-char ncurses as well as in all PDCurses*) seems like 
good behavior to me.

-- Bill

#include <curses.h>

/* Compile with gcc -Wall -Wextra -pedantic -O3 -oz z.c -lncursesw
(or with -lncurses for narrow-char build) */

int main( void)
{
    chtype c;

    initscr( );
    bkgd( COLOR_PAIR( 0));
    c = mvinch( 1, 1);
    mvaddch( 1, 1, c);
    mvprintw( 2, 1, "The above line will show ^@ on wide ncurses builds.");
    mvprintw( 3, 1, "It should be blank on narrow-char builds and on all");
    mvprintw( 4, 1, "PDCurses and PDCursesMod builds.  Hit the any key.");
    getch( );
    endwin( );
    return(0);
}