bug#11810: 24.1.50; `vc-diff' shrinks pre-existing window

[Dmitry Gutov <dgutov@yandex.ru>]

On 30.06.2012 13:09, martin rudalics wrote:
> `even-window-heights' is one of these options pertaining to the classic
> two windows one-above-the-other frame layout.  It should be redesigned
> to handle side-by-side windows and is largely obsolete anyway because
> you can now customize `window-combination-resize' instead.

How is it obsolete? As far as I can see, neither of the two possible 
values of `window-combination-resize' says "don't resize my existing 
windows until specifically asked".

>  > I don't think I've ever used `temp-buffer-resize-mode', but if it's a
>  > minor mode that a user has to enable explicitly, it should be fine if it
>  > overrides the value of `resize-windows-for-display'.
>  > Please note that `temp-buffer-resize-mode' already does the sane thing:
>  > it only resizes the window if the window is new and not reused.
>
> `temp-buffer-resize-mode' adds `resize-temp-buffer-window' to
> `temp-buffer-window-show-hook' so it resizes the window even when it's
> reused.  Or do I miss something?

Actually yes, I haven't tested it properly the last time.
It can shrink the window to make it fit the buffer, but it restores the 
window's size when it's done, which is the important thing.

>  > To answer you question in the last message:
>  >
>  >  >> !       (let ((resize-windows-for-display nil))
>  >  >> !  (pop-to-buffer (current-buffer)))
>  >  >
>  >  > Here you explicitly override the user option - is that intentional?
>  >
>  > It's intentional, because `vc-diff-internal' calls `pop-to-buffer'
>  > before the diff command returns its full output, so the window height
>  > adjustment happens in `vc-diff-finish' which runs after the process
>  > returns. So you might want to account for this usage.
>
> I don't recall the details but we always had a chicken-and-egg problem
> here: When you fill a buffer before displaying it you can't make its
> line breaks suit the window used for it.  When you display it first, you
> can't make the window suit the line breaks of the buffer.  And the
> number of line breaks determines the height of the region to display.

Do you mean auto-filling, as in, "inserting hard newlines"? I don't 
think diff-mode does that. It does pop a window immediately, but 
probably just because it would be weird to see a window pop up only 
after the diff command finishes (if that happens after some time has 
passed).

>  > Regarding "two similar approaches", I think just having an off by
>  > default `even-window-heights` variable and `temp-buffer-resize-mode' may
>  > satisfy more or less everyone, except there'd at least need to be a way
>  > to make shrinking asynchronous, as per above.
>
> Since I don't use `vc-diff' I can't really comment on how it should
> behave (I practically always use side-by-side windows for watching
> diffs).  Is anyone interested in the resizing behavior at all?  Couldn't
> we use the `quit-restore' window parameter to restore the original size
> of the window?

We probably could, if solving only this particular case is okay. As far 
as VC package goes, we'd need to do that in `vc-diff-internal' and in 
all (?) callers of `vc-log-internal-common'.

-- Dmitry