bug#56102: 29.0.50; fit-frame-to-buffer's window-text-pixel-size calculation can be incorrect when only is set to vertically

[martin rudalics]

>> With a non-standard setup, say
>> after doing
>>
>> (set-window-margins nil 10 10)
>>
>> in the root window, your fix will fail with the scenario you provided
>> earlier.
>
> Ah, I can confirm this. Is there a reasonable way for me to calculate
> a max-width that would be based on the root window that would work?

Here

(set-window-margins nil 10 10)
(fit-frame-to-buffer nil nil nil (window-body-width) nil 'vertically)

seems to work.

> There's other math that happens within fit-frame-to-buffer I don't
> fully have my head wrapped around yet. I'm not super worried about
> this personally as I don't set window margins on this window.

Similar problems should happen when the fringe or scroll bar sizes of
the root window or its buffer differ from that of the frame.

> The patch works for me and seems good. When you say if you come up
> with a reasonable fix, could I ask what is unreasonable about the
> patch you attached?

The current code interprets all its -HEIGHT and -WIDTH arguments as if
they were body sizes and not total sizes.  This at least contradicts the
doc-strings of 'fit-frame-to-buffer-sizes' and of 'fit-frame-to-buffer'
itself.  If I can fix the problem by rewriting the doc-strings to match
what the code does, I'll probably do that.  In general, you can't get
these right anyway when you allow specifications in terms of character
sizes and windows/frames may have non-integral sizes wrt these.

> Regardless, if you do end up updating the fix to respect a supplied
> max-width even if only vertically is supplied, I could always make an
> Emacs version based decision on whether or not to pass the work-around
> max-width in, and Emacs 29 is as good a version as any to stop passing
> it in (people already on master will have to recompile, but that's
> fine imo). So, I would be fine with the new patch going straight to
> master, though as I mentioned I don't know what it would do
> differently or why.

I don't know yet either.  I'll send you a patch as soon as I have one.

martin