Re: VC check-in and window configuration

[Stefan Monnier]

> A user suggested that C-x v v, when used for check-in, should remember
> the current window configuration and restore it after the user typed
> C-c C-c in the *VC-log* buffer.  This is not the first time this
> suggestion comes up, but I remember vaguely that we have discussed
> this before.  Back then, the consensus seemed to be that it's just not
> possible to decide what is the "right thing" to do in this situation,
> and that things might as well be left as they are.
> 
> But that's only a vague memory.  Other than that, I think the idea is
> perfectly reasonable, so if I hear no objections, I'm going to install
> such a change.

I personally don't like such an approach because I often pop up
the *VC-Log* buffer long before doing C-c C-c and it would be
annoying to be put back into the window-state I was in "a few
hours ago".

Also, restoring the window configuration doesn't work so well when
the various buffers are displayed in different frames.

My current "best" idea is to extend the notion of dedicated window
to allow a "soft" for of dedication: a soft-dedicated window still
allows things like `switch-to-buffer' (at which point it goes back
to non-dedicated).  But just like dedicated windows, a soft-dedicated
window gets deleted when the buffer it displays is killed.

This would then be used by pop-to-buffer: if the buffer is
displayed in an existing window, then do things as before (i.e.
the window is non-dedicated), but if a new window is created
to display the buffer, then make it soft-dedicated.
This way

        (progn
          (pop-to-buffer buf)
          (bury-buffer))

should not change the window configuration.


        Stefan