Enable per-monitor DPI for Win10 #8786

zcbenz · 2017-02-27T10:11:11Z

Close #5429.

groundwater · 2017-02-28T01:08:44Z

I will test this with my Surface Pro 4 and non-high-DPI external monitor.

kevinsawicki · 2017-04-03T17:26:22Z

I will test this with my Surface Pro 4 and non-high-DPI external monitor.

@groundwater have you had a chance to test this out yet?

miniak

Please consider keeping the manifest entry instead of adding back the code.

miniak · 2017-04-04T06:55:23Z

atom/browser/resources/win/atom.manifest

@@ -32,7 +32,6 @@

  <asmv3:application xmlns:asmv3="urn:schemas-microsoft-com:asm.v3">
    <asmv3:windowsSettings xmlns="http://schemas.microsoft.com/SMI/2005/WindowsSettings">
-      <dpiAware>true</dpiAware>


you can use <dpiAware>True/PM</dpiAware> instead of adding back the code
http://stackoverflow.com/questions/23551112/how-can-i-set-the-dpiaware-property-in-a-windows-application-manifest-to-per-mo

I also think setting this via manifest is preferrable. You can also combine the <dpiAware> element with the new <dpiAwareness> element for Windows 10 build 1607 and higher to enable additional per-monitor DPI features.
https://msdn.microsoft.com/en-us/library/windows/desktop/aa374191.aspx

You may also want to look at what we added in Windows 10 Creators update which is detailed here: https://blogs.windows.com/buildingapps/2017/04/04/high-dpi-scaling-improvements-desktop-applications-windows-10-creators-update/#HBK3Qfrm4qbzWAEV.97

Non-client area scaling is available via the following:

PerMonitor tag + a call to EnableNonClinetDpiScaling() on 1607

PerMonitorV2 on 1703 (no need for that additional API call)

You can also use a combination of tags that will be consumed differently on different versions of Windows:
true/pm (this will give you per-monitor (version 1) on >= 8.1 and system awareness on Vista to 8)
PerMonitorV2, PerMonitor will result in PerMonitorV2 on 1703 (Creators Update) and PerMonitor (version 1) on 1607 (Anniversary Update). When running on versions of Windows that consume this tag the tag will be ignored.

I can probably work on this and I also have a combination of High/Low DPI screens running.
I think EnableNonClientDpiScaling is really important (or permonitorv2), i wouldn't go for true/pm only

you can use True/PM instead of adding back the code

I didn't know this, it is much better!

PerMonitor tag + a call to EnableNonClinetDpiScaling() on 1607
PerMonitorV2 on 1703 (no need for that additional API call)

I agree it is important to have these, but I'm not sure whether they work well with Chromium's own UI framework. Can someone who has Windows running on multiple monitors test it?

I actually plan to have a look at this. Leave it to me then, I'll make it work across all the different scenarios.

@alespergl You maybe be interested in: https://chromium.googlesource.com/chromium/src/+/b297db871b54cd58c32c96897a7a59af295ba796

That's included in Chromium 57.

alespergl

Let's keep using the manifest and adjust the value to true/pm

poiru · 2017-04-12T00:21:18Z

What do we think of merging this? I agree that we should ideally also have

<dpiAwareness xmlns="http://schemas.microsoft.com/SMI/2016/WindowsSettings">permonitorv2, permonitor</dpiAwareness>

... but it would be nice to have at least <dpiAware>true/pm</dpiAware> until we can confirm that permonitorv2 works properly.

kevinsawicki · 2017-04-12T19:42:10Z

What do we think of merging this? I agree that we should ideally also have

Yeah, sounds good 🚢

We can follow this up with more changes as well as pulling in anything via the Chrome 58 upgrade.

alespergl · 2017-04-12T19:48:40Z

When is Chrome 58 coming in? The change @poiru mentioned would likely need to be backported if it's delayed.

kevinsawicki · 2017-04-12T19:52:05Z

When is Chrome 58 coming in?

I believe @zcbenz is working through some offscreen window compilation issues right now that are being updated in #9116, no definite timeline at this time.

The change @poiru mentioned would likely need to be backported if it's delayed.

Understood, maybe we wait until the end of next week to see where Chrome 58 is then?

alespergl · 2017-04-13T14:56:48Z

@kevinsawicki I created a PR for the backport, just in case.

poiru · 2017-04-17T10:14:35Z

On Win10 version 1607, this patch made the titlebar, scrollbars, etc. huge:

Maybe EnableNonClinetDpiScaling() will fix this, but that's not available on all Windows 10 versions. Should we revert this patch? cc @alespergl

peterfelts · 2017-04-17T15:39:37Z

So it looks like before this change the process ran as system aware <dpiAware>true</dpiAware> and the change made it run under the original per-monitor mode <dpiAware>true/pm</dpiAware>. This would result in what @poiru is seeing: non-client area won't DPI scale on a DPI change. Getting the non-client area to scale (if Windows is drawing the non-client area, that is) wasn't supported until 1607 (via a call to EnableNonClinetDpiSacling(HWND)). What I would recommend is:

<dpiAware>true</dpiAware>
<dpiAwareness>PerMonitor</dpiAwareness>

call EnableNonClientDpiScaling()

This will result in the following behavior:

Vista -> 1607 -> System Awareness (what I think was used before this change)
>= 1607 -> per-monitor awareness and the non-client area will scale

Again, the non-client area will only scale if Windows is drawing it. I don't know how Electron draws its non-client area though. If its custom drawn, then Windows won't DPI scale it automatically.

alespergl · 2017-04-17T16:06:37Z

@poiru Yes, EnableNonClientDpiScaling fixes it. There's no way to make this work universally, unless custom non-client drawing is implemented, or we don't do per-monitor DPI pre Win10 1607. That would result in blurred client area. Other per-monitor apps work the same. I think it's reasonable to expect most users upgrading their Win10 systems to the latest version, so I wouldn't optimize too much for earlier versions.
Btw dpiAwareness makes no difference here, it's simply a shortcoming of Windows.

This fixes non-client area DPI scaling on recent Windows 10 versions. See discussion in #8786.

bengotow added the in progress label Feb 27, 2017

zcbenz removed the in progress label Feb 27, 2017

groundwater self-requested a review February 28, 2017 01:08

groundwater assigned zcbenz Feb 28, 2017

zeke mentioned this pull request Mar 2, 2017

Blurry text on a windows electron app #8708

Closed

MarshallOfSound mentioned this pull request Mar 3, 2017

Electron doesn't respect win10 scaling factor. #8823

Closed

groundwater self-assigned this Mar 14, 2017

miniak requested changes Apr 4, 2017

View reviewed changes

alespergl suggested changes Apr 5, 2017

View reviewed changes

Enable per-monitor DPI

c59ef0e

zcbenz force-pushed the per-monitor-dpi-aware branch from 47e7e90 to c59ef0e Compare April 11, 2017 06:17

kevinsawicki assigned alespergl Apr 11, 2017

seanzer mentioned this pull request Apr 12, 2017

Provide an option to enable DirectWrite microsoft/vscode#1896

Closed

kevinsawicki merged commit 1c44bcf into master Apr 12, 2017

kevinsawicki deleted the per-monitor-dpi-aware branch April 12, 2017 19:27

poiru added a commit that referenced this pull request Apr 18, 2017

Update libcc for electron/libchromiumcontent#285

4a7125d

This fixes non-client area DPI scaling on recent Windows 10 versions. See discussion in #8786.

poiru mentioned this pull request Apr 18, 2017

Update libcc for electron/libchromiumcontent#285 #9219

Merged

This was referenced Jun 7, 2017

In windows multiple monitor setup with different DPIs, electron app starts out of bounds #9688

Closed

Reverting per monitor dpi aware causing issues with different dpis #9708

Closed

wryun mentioned this pull request Jul 6, 2017

VSCode is not per monitor DPI aware microsoft/vscode#6196

Closed

kevinvangelder mentioned this pull request Aug 23, 2017

No HiDPI screen support on Windows? infinitered/reactotron#503

Closed

Enable per-monitor DPI for Win10 #8786

Enable per-monitor DPI for Win10 #8786

Uh oh!

Conversation

zcbenz commented Feb 27, 2017

Uh oh!

groundwater commented Feb 28, 2017

Uh oh!

kevinsawicki commented Apr 3, 2017

Uh oh!

miniak left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

alespergl left a comment

Choose a reason for hiding this comment

Uh oh!

poiru commented Apr 12, 2017 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

kevinsawicki commented Apr 12, 2017

Uh oh!

alespergl commented Apr 12, 2017

Uh oh!

kevinsawicki commented Apr 12, 2017

Uh oh!

alespergl commented Apr 13, 2017

Uh oh!

poiru commented Apr 17, 2017 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

peterfelts commented Apr 17, 2017

Uh oh!

alespergl commented Apr 17, 2017 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

poiru commented Apr 12, 2017 •

edited

Loading

poiru commented Apr 17, 2017 •

edited

Loading

alespergl commented Apr 17, 2017 •

edited

Loading