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

[Aaron Jensen <aaronjensen@gmail.com>]

On Fri, Jun 24, 2022 at 5:20 AM martin rudalics <rudalics@gmx.at> wrote:
>
>  > That seems to work for me. I do wonder though if the the check for
>  > `only' should be first (i.e. if only is vertically, max-width is nil).
>  > Is there a reason that we should not ignore a specified max-width when
>  > only is set to vertically?
>
> The idea of 'fit-frame-to-buffer' was that an application should be able
> to call it (maybe implicitly via 'temp-buffer-resize-mode') and not care
> about where on the display the frame will be shown.  Hence, a major
> concern of its design was to constrain the frame to some specified area
> on the display, to avoid that parts of it move off screen.  That's why
> 'fit-frame-to-buffer-margins' and 'fit-frame-to-buffer-sizes' have been
> provided.
>
> Currently, we check for an explicit MAX-WIDTH first and then consult
> 'fit-frame-to-buffer-sizes' via SIZES as
>
>                      ((numberp max-width) (* max-width char-width))
>                      ((numberp (nth 2 sizes)) (* (nth 2 sizes) char-width))
>
> If we were to override MAX-WIDTH by setting ONLY to 'vertically', where
> and how would we check SIZES?

I'm definitely not familiar enough with the workings and the intended
use cases. My only thought about it was one of principle of least
surprise. If I say that I only want it to scale vertically, I would
expect no width changes to occur, which means I would expect that
max-width and anything else width related would not be relevant.
There's likely something I'm missing here, so please feel free to
disregard if this is necessary for some reason.

>  > I ask because in the package that I had
>  > this issue with I employed a work-around where I set the max-width to
>  > (frame-parameter frame 'width), which seems to work well enough, but
>  > probably not as good as your fix. We may not be able to remove that
>  > workaround for some time, so ignoring max-width if set would probably
>  > work better in our specific case.
>
> Are these issues really related?  If your workaround works, it should
> continue working regardless of whether we ignore MAX-WIDTH when ONLY is
> 'vertically' or not.  Or am I missing something?

My concern was this:

> Also, the sizes of frame and window decorations would hardly match.

Effectively, that there would be some subtle difference between using
(frame-parameter frame 'width) as MAX-WIDTH and passing nil to
window-text-pixel-size window.

If there's no difference to be concerned about then my workaround
should be fine with the code as it was in the patch.

Thanks,

Aaron