Re: Which should the display-pixel-width function return, physical pixel width or logical pixel width?

[Yuan Fu <casouri@gmail.com>]

> On Jan 1, 2021, at 9:44 AM, Yuuki Harano <masm+emacs@masm11.me> wrote:
> 
> Hi, I'm pgtk developer.
> 
> Currently, display-pixel-width of vanilla emacs returns physical pixel
> width.  i.e. it returns 3840 even if GDK_SCALE=2 when using 3840x2160.
> 
> Some emacs lisps (like preview in auctex) use display-pixel-width and
> display-mm-width to calculate physical dpi.
> 
> I think it is strange. On multi-monitor environment,
> display-pixel-width and display-mm-width contains all the monitors,
> that may be different dpi.  The result dpi is between the two.
> Those emacs lisps should use per-monitor information.
> 
> Currently, pgtk emacs returns logical pixel width, i.e. 1920, because
> Gdk returns it.  On multi-monitor environment, Gdk returns 1920 +
> something.
> 
> I'm going to add scale-factor in per-monitor information to support
> scaling.  dpi-sensitive emacs lisps can extract logical pixel width,
> mm width, and scale-factor from it, and calculate dpi.
> 
> "Monitor" is a recent concenpt.
> If pgtk emacs returns logical one, then compatibility may be broken.
> If pgtk emacs returns physical one, then those emacs lisps continue to
> do strange calculation.
> 
> What should the display-pixel-width function (and
> display-monitor-attributes-list) return, physical pixel width, logical
> pixel width, or implementation-dependent?  The documentation of
> display-pixel-width seems to say nothing about that.
> -- 
> Yuuki Harano
> 

As hidpi screens getting more common, we should separate pixel sizes reported from frontends into logical and physical pixels. This problem also makes images on Mac fuzzy. Currently ns frontend can distinct between logical and physical pixel size, from Yuuki’s post it seems pgtk can, too. Does other frontends have such ability?

Yuan