Re: init_color() problem

[Thomas Dickey]

On Wed, 15 Nov 2006, Simon Sasburg wrote:

> I recently started fiddling around with ncurses to see if i could use it to
> port an MSDOS app of mine to linux.
>
> After finding out I could use A_BOLD to access all 16 colors instead of just
> 8, i started looking how i could change the palette.

BOLD attribute may display as a brighter color, it may display as a 
thicker font, it may even be a _different_ color.  It depends on the
terminal implementation.  But curses doesn't believe that it's a color,
any more than any of the other video attributes.  (I'd ask what color is
BLINK, but that depends on the terminal implementation ;-)

> Now, obviously this is a bit of a hack, and init_color() uses its first
> parameter as an index into some array, so it may even cause buffer overflows
> or similar nastyness.

yes, that's the way I read the code.  It allocates an array within 
start_colors() and sets COLORS there based on the terminfo entry.
Since it's using COLORS thereafter for the array limit, bad things will 
happen if you increase COLORS.

> But, it does work, so the terminal does support changing the rgb values of
> the "bold" colors...
>
> I shouldn't really have to use such a hack to make ncurses let me do this, i
> think.

You don't - normally that's done in the terminfo description.

You could for instance design an alternate terminfo entry which tells 
ncurses that the terminal does not support BOLD+color, and alter the 
setaf/setab strings to use the BOLD attribute as part of their 
computation.

--
Thomas E. Dickey
http://invisible-island.net
ftp://invisible-island.net