* Add new procedure `window-override-message' that returns the
override message, or #F if none. This is used to implement
`current-message', which operates on the current typein window.
`clear-message!' renamed to `clear-current-message!'.
`set-message!', renamed to `set-current-message!', now accepts #F as
an argument, in which case it acts like `clear-current-message!'.
* Split `select-buffer-in-new-screen' into two parts: a procedure
`make-screen' that generates a new screen, and a call to
`select-screen'. The new procedure `make-screen' replaces the
procedure of the same name which is now considered internal to the
screen abstraction.
* Change `select-screen' not to abort to top level, to run the
`select-buffer-hook', and to transfer the typein override message
from the previously-selected screen to the newly-selected one. The
X screen event handler now does the abort to top level, since it is
still needed in that case.
* Eliminate the typein bufferset -- typein buffers are shared by all
screens. Display of the typein buffers is suppressed in
non-selected screens by a blank override message. This has the
drawback that direct update of the typein window does not work if
there are multiple screens, which makes typein feel sluggish.
* Implement procedures `screen0', `screen1+', `screen-1+', and
`screen+' for moving around the screen list.
* `buffer-list' no longer copies its result -- don't clobber it!
* New procedure `change-selected-buffer' makes the handling of buffer
selection more uniform.
* Eliminate cacheing of `editor-input-port' from "input.scm". Just
extract the input port from `current-editor' every time. Change the
keyboard reader to use `input-port/read-char' instead of
`read-char', since the former is faster.
* Redesign the `button' abstraction to make it cleaner and more
general. Rename the `buttonN-down' and `buttonN-up' variables to
`x-buttonN-down' and `x-buttonN-up'. Change button-handling code so
that up buttons don't beep if they aren't bound to commands.
* Updating of an X screen is now terminated if a non-keypress event
arrives while checking for update preemption. This is done by
throwing out of the update. I believe this test only happens in
places where it is safe to do this.
* Make screen highlight control independent of the screen type.
Change screen abstractions so that screens support two operations,
one to turn on highlighting, and the other to turn it off.