need API to discover hidpi scale factor

[Allison Karlitskaya]

Although Mir has mir_surface_get_dpi(), this is not useful information for a toolkit or application authors in terms of which scaling factor they should use.

The user may want to configure the scaling factor independent of DPI and just because the DPI of a screen changes doesn't mean we should necessarily change the pixel size of controls.

We need an API that tells us the scaling factor directly.

[Daniel van Vugt (vanvugt)]

If your screen was physically 200DPI and you didn't like squinting and wanted more like 100DPI then that would certainly mean a scaling factor of 2.0 in theory. You could implement that as a literal scaling factor of 2.0, or you could implement it as a desktop DPI setting of "100DPI" (which is how Windows has done it for the past 20+ years).

So there are multiple ways to do this...

[Daniel van Vugt (vanvugt)]

Just remember that "scaling factor" like DPI will vary between monitors. And the approach so far has been to report the DPI/scale of a surface according to the monitor it is "mostly on" (not fully implemented yet).

What this means is that a "scaling factor" would be per-surface in case we need it to vary across different monitors.

[Daniel van Vugt (vanvugt)]

Or rather, the source of the scaling factor would be the monitor structure itself. And reported to clients as an attribute of the surface according to placement.

[Allison Karlitskaya (desrt)]

I agree, except I would suggest that the scaling factor be whatever the highest scale is for any monitor that the window is partially on. You want to over-render and then downscale rather than the opposite just because I only happen to be 49% on the better quality screen.

[Daniel van Vugt (vanvugt)]

Yes. Indeed it would be the highest scale factor of any monitor that the surface overlaps. Just as the reported DPI of a surface would be the highest for all monitors that are touched.

[Daniel van Vugt (vanvugt)]

Actually this will become easier soon. I'm working on the glue between the display layout and independent surfaces right now for other WM work.

[Chris Halse Rogers (raof)]

Fixed in Mir 0.17 (See surface_output_event)