Re: Colors in copywin()/overlay()

[Thomas Dickey]

On Wed, 7 Dec 2005, William McBrine wrote:

> I noticed a difference in the way copywin() and its derivatives are handled 
> by ncurses vs. PDCurses. When overlaying text, ncurses takes attributes like 
> A_BOLD from the original window, but it takes the colors from the destination 
> window. PDCurses takes all attributes, including colors, from the source 
> window. (You can see this in the "testcurs" demo program.)
>
> ncurses' behavior matches that of (at least) Solaris' and NetBSD's curses. 
> However, the Single UNIX Specification, Version 2, seems to me to specify the 
> behavior seen in PDCurses. It says:

I recall that PDCurses color model was supposed to be SVr3 - so there
might be some differences just from that.

> The overwrite() function copies characters as though a sequence of
> win_wch() and wadd_wch() were performed with the destination window's
> attributes and background attributes cleared.
>
> The overlay() function does the same thing, except that, whenever a
> character to be copied is the background character of the source window,
> overlay() does not copy the character but merely moves the destination
> cursor the width of the source background character.
>
> So, which is wrong: the spec, the implementations, or my understanding? :-)

The spec is flawed in several places since it was reworded from manpages 
by people who apparently had no specific expertise in the area.  So I tend 
to look at existing implementations (more than one, of course), and weight 
the results with the spec.  NetBSD is further downstream than ncurses - I 
don't know that it's been tested independently of ncurses.  Solaris is 
(usually) better but none of the work was done by Sun - it was all 
inherited from AT&T.  It's probably best to bring in a test-case that's 
run on Tru64 (which appears to have been independently developed by people 
who understood the spec ;-)

I don't have a test-program at hand.  When we were thrashing out the 
behavior of background character, I wrote a test program to do addch's and 
read back the data.  But I didn't do one for overlay(), myself.

short: the spec might be ambiguous but make sense in the light of other 
code.

(If no one submits a test program and results, this'll go on my to-do 
list)

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