bug#17768: 24.3; problem with two ruler-mode windows

[martin rudalics]

>    First patch applies the changes that we supposed could
> solve the problem of the wrong offset.

Thanks.  I'll apply it as soon as I can build again here.

>   I tested patch1 somewhat randomly, opening 4/5 windows horizontally,
> 1/2 windows vertically, and did also some attempt splitting windows
> in two frames. Played also a bit on the text-only Emacs (using MSYS shell).
>
>    Everything worked fine with patch1, except for an unexpected small
> and non-blocking issue (let's call it "diagonal-drag" issue) that
> can be reproduced as follows:
>
> - C-x 3 M-x ruler-mode (you should see two windows with ruler)
> - click S-<mouse-1> on the ruler of the rightmost window, this will set
>    a left margin, choose a width of around 10~20 chars
> - drag "diagonally" a column, for example moving down-left the
>    fill-column, until the cursor goes a few pixels below the ruler area

How do I drag a column?  What is "moving down-left"?  Can you explain in
terms of keystrokes or mouse clicks?

> - you will see that the dragged column will "jump" innaturally to the
>    left of around 10~20 chars
>
>    Tentative diagnosys: function `ruler-mode-window-col' returns a
> different column value depending if you are inside or outside the
> ruler area. The actual problem is probably inherited from `posn-col-row',
> but I could not clarify any better.

It seems so (from reading the doc-string of `ruler-mode-window-col').
But I would like to reproduce the problem first.  Please advice.

>    And finally: I noticed another independent bug ("text-scale issue"),
> i.e. most ruler's mouse operation have wrong offset if you adjust
> text scale (C-x C-+).

This should be definitively corrected (as well as other text scale
related issues happening, for example, when splitting windows).

>    If anyone reading is interested to help, here's a proposed action plan:
>
> - may you please try to play with patch1, and give advice if it could be
>    acceptable for a commit

I only need advice from Stefan whether to install it in the branch or on
trunk.

> - I guess that patch2 is unacceptable (BTW it changes the definition of
>    ruler-mode-window-col, I assumed that ruler-mode.el is the only
>    program calling it) but it's the best I could squeeze out of my head...
>    any help or constructive criticism is welcome :)

The premise of `ruler-mode-window-col' seems wrong so we have to do
something about it.

> - be aware of the "text-scale" issue but don't test for it; it seems
>    a totally independent issue so I'd rather send another bug report for it

Please do that.

Thanks, martin