multiple displays and ttys

multiple displays and ttys

[Stefan Monnier]

Now that I'm using emacsclient even more, I more often bump into situations
where Emacs displays a buffer in some display that I cannot see.

I suggest the patch below which makes that only the frames on
the `current terminal' are considered as visible.

This way, when you're working on /dev/pts/2, all the frames on ":0.0" are
considered as invisible, and similarly between different X displays.

I find this invaluable.  Any objection?


        Stefan


diff -u -b /home/monnier/src/emacs/trunk/src/window.c /home/monnier/src/emacs/work/src/window.c
--- /home/monnier/src/emacs/trunk/src/window.c	2007-09-21 03:55:52.000000000 -0400
+++ /home/monnier/src/emacs/work/src/window.c	2007-09-21 04:03:11.000000000 -0400
@@ -1794,7 +1798,8 @@
   else if (EQ (all_frames, Qvisible))
     {
       FRAME_SAMPLE_VISIBILITY (f);
-      candidate_p = FRAME_VISIBLE_P (f);
+      candidate_p = FRAME_VISIBLE_P (f)
+	&& (FRAME_TERMINAL (f) == FRAME_TERMINAL (XFRAME (selected_frame)));
     }
   else if (INTEGERP (all_frames) && XINT (all_frames) == 0)
     {
@@ -1810,7 +1815,8 @@
 		     || (FRAME_X_P (f) && f->output_data.x->asked_for_visible
 			 && !f->output_data.x->has_been_visible)
 #endif
-		     );
+		     )
+	&& (FRAME_TERMINAL (f) == FRAME_TERMINAL (XFRAME (selected_frame)));
     }
   else if (WINDOWP (all_frames))
     candidate_p = (EQ (FRAME_MINIBUF_WINDOW (f), all_frames)

Re: multiple displays and ttys

[Richard Stallman]

    I suggest the patch below which makes that only the frames on
    the `current terminal' are considered as visible.

It seems logical, for most purposes.  But I think it would be a good
idea to check the calls to the principal primitives that loop over
windows, to see which calls specify visible windows only, and check if
this is right for all of them.  It is probably right for most of them;
what I am not sure of is whether it is right for all of them.

We might want to make the arguments which traditionally mean "visible
frames" now mean "visible on this terminal", and create some other way
to say "visible on any terminal" -- if the latter is really needed.
But if it isn't needed, we are better off keeping things simple
and not creating it.

Re: multiple displays and ttys

[Stefan Monnier]

>     I suggest the patch below which makes that only the frames on
>     the `current terminal' are considered as visible.

> It seems logical, for most purposes.  But I think it would be a good
> idea to check the calls to the principal primitives that loop over
> windows, to see which calls specify visible windows only, and check if
> this is right for all of them.  It is probably right for most of them;
> what I am not sure of is whether it is right for all of them.

> We might want to make the arguments which traditionally mean "visible
> frames" now mean "visible on this terminal", and create some other way
> to say "visible on any terminal" -- if the latter is really needed.
> But if it isn't needed, we are better off keeping things simple
> and not creating it.

It's very difficult to check each and every use of this since it's called
from several different functions, so I cannot claim that I've done an
exhaustive search.  But I've been using this change for a several years and
grepped several times during this interval to try and find relevant pieces
of code, mostly because they all tend to either forget about multiple-frames
situations (and pass an implicit nil argument to only consider the selected
frame) or use the t argument thinking "if nil is not good enough, then
t ought to do it".

I have never encountered any use that needed to know about "visible, even on
other screens", so I think we can install the change as is, and in case we
ever come across the need for the old meaning of `visible' (and 0), we can
add it at that point.


        Stefan

Re: multiple displays and ttys

[Richard Stallman]

      But I've been using this change for a several years and
    grepped several times during this interval to try and find relevant pieces
    of code,

Given that evidence, I am convinced.