Re: Fill column indicator functionality

[Eli Zaretskii <eliz@gnu.org>]

> From: Mattias Engdegård <mattiase@acm.org>
> Date: Fri, 15 Mar 2019 14:30:55 +0100
> Cc: emacs-devel <emacs-devel@gnu.org>
> 
> Terminal emulators often treat box-drawing characters specially, either drawing the lines themselves or extending parts of the glyphs in case there would be gaps. Even XTerm does this, mainly to supplement missing glyphs.
> 
> Being its own terminal emulator, Emacs could do the same to avoid gaps in horizontal and vertical lines that should connect.

Unfortunately, this is not currently possible, not without significant
changes in how the Emacs display engine works internally.  The display
routines are currently called in a way that doesn't allow them to have
a window-global view of the display, not even a way of knowing what's
on the previous and the next screen lines.  In fact, the display
routines need to do their job even if the information about the layout
of adjacent screen lines is not available at all, and redisplay
optimizations frequently make use of this property, by starting a new
redisplay cycle on a specific line, doing part of the layout, then
throwing out the results.  About the only thing the display routines
can rely on is the window-start point and the window pixel
dimensions.  Even the vertical coordinate of a line on which the
display routine was called is frequently unknown. 

> The details are fairly window-system-specific but do not have to be implemented everywhere at once.

I don't think I understand what you are saying here.  My point was
that we have xfns.c/xterm.c for X, w32fns.c/w32term.c for w32, and
nsfns.m/nsterm.m for NS.  Any line drawing that doesn't end up as part
f some glyph must be implemented in each of these back-ends, because
the APIs used for such drawing are different.