wayland: rework scale factor handling

[Matthias-Fauconneau]

wayland: rework scale factor handling

Fixes some hidpi use cases on wayland :

• initialize with primary scale factor (avoid flicker)
• also call Resize on scale factor changes
  apps usually resize their swap chains there and don't handle the explict scalefactorchanged.
  As discussed with kchibisov, calling this on any scale change even if it happens to yield same physical size (e.g move window from 4K laptop to 4K TV), as a defense in depth to try to trigger any application UI/font scale redraw (even when they don't explicitly handle scalefactorchanged).
  This might make applications do an identical swap chain resize as well if they don't check, but that's okay.
• minor fix: inner_size within Resize callback (not hdpi related)

Fixes Iced

[kchibisov]

It seems like this it is this bug #1451 , right? I completely forgot about this issue in winit, thx.

[Matthias-Fauconneau]

Indeed. Could have saved some time if this had gotten fixed right away. Can't see any reason why you'd want the old size :/

[kchibisov]

I don't see any, but it was opened as a followed up to some other bug, which was closed for some reason, and I forgot about this issue, since I never use inner_size.

[kchibisov]

I've memorized why it wasn't fixed right away, it's because you're not actually updating size, but just some variable, and your inner size will still be incorrect, so you're not fixing anything here. To fix it properly you should update size of the inner window, where this for each closure was invoked.

[Matthias-Fauconneau]

Okay, I'm not familiar with winit, already had a hard time enough trying to figure out why the whole code is so complicated and what is the expected behaviour across platforms, this works for me for now, so I'll let you take someone else take over if there is a more proper fix to be figured out :).

…

On Fri, Apr 17, 2020 at 1:27 PM Kirill Chibisov ***@***.***> wrote: ***@***.**** requested changes on this pull request. I've memorized why it wasn't fixed right away, it's because you're not actually updating size, but just some variable, and you inner size will still be incorrect, so you're not fixing anything here. To fix it properly you should update size of the inner window, where this for each closure was invoked. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#1538 (review)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAGGMFWQYLH26MYCX3V2WGLRNA4JDANCNFSM4MKUFTCQ> .

[kchibisov]

I'm not sure what is working here, but on your PR I get missmatch when running glutin window example, so it's not a fix.

INNER SIZE PhysicalSize { width: 1278, height: 1280 }
PHysical size PhysicalSize { width: 1280, height: 1280 }

[Matthias-Fauconneau]

You are right, it doesn't. I must have messed up my setup and testing a binary which still had the ScaleFactor handler by mistake ><

…

On Fri, Apr 17, 2020 at 1:38 PM Kirill Chibisov ***@***.***> wrote: I'm not sure what is working here, but on your PR I get missmatch when running glutin window example, so it's not a fix. INNER SIZE PhysicalSize { width: 1278, height: 1280 } PHysical size PhysicalSize { width: 1280, height: 1280 } — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.

[kchibisov]

It's fine from me.

cc @vberger if it fixes your anvil issue.

[elinorbgr]

Thanks!