Re: Why buffer_shared is a mistake

[Eli Zaretskii]

> Date: Mon, 10 Dec 2012 11:29:46 +0400
> From: Dmitry Antipov <address@hidden>
> CC: address@hidden
> 
> But now I'm thinking about maintaining per-buffer window counter,
> which value is independent from the redisplay actions.  This opens
> the way to a different optimizations (e.g. if we're modifying a buffer
> which isn't shown in a window, we can avoid redisplay at all).

The idea is fine with me.  Just keep in mind the TTY display, where
windows are not necessarily independent as far as redisplay is
concerned.

> (My view of the whole thing is: Emacs global state is huge, and
> redisplay is controlled by too much amount of global variables.
> This is error-prone and makes some valuable things like threading
> too complicated or even impossible. So it's desirable to deduce
> redisplay actions from per-buffer/window/frame variables rather
> than from the global ones.)

Perhaps a good starting point would be to find all those global
variables that control the display engine, and document, or at least
list them in one place.  Then those of them that become buffer- or
window- or frame-local could be removed from that list.