Bug 147242 – [Win] Support High DPI drawing with CACFLayers

• Status: RESOLVED
• Resolution: FIXED
• Component: Layout and Rendering

[Brent Fulgham]

This is a follow-up to Bug 146335, which added support for High DPI use in our non-accelerated drawing paths.

The Windows platform doesn't allow a single device scaling factor to be applied to individual windows. Instead, high DPI state is handled by working with larger pixel surfaces at higher DPIs, and performing scaling calculations to control widget size, positioning, etc. Yuck!

This patch applies the proper scaling at various phases of our CACFLayer sizing and positioning.

[Brent Fulgham]

Note: Just like with Bug 146335, these scaling calculations are all done at 1x, unless the WebKit client opts into a high DPI state.

Currently, testing is just to make sure no regressions to existing 1x drawing on Windows (and all drawing on other platforms) are not affected.

[Brent Fulgham]

Created attachment 265975 [details]
Patch

[Tim Horton]

Comment on attachment 265975 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=265975&action=review

> Source/WebCore/platform/graphics/ca/win/WKCACFViewLayerTreeHost.cpp:128
> +    CACFLayerSetContentsScale(layer->platformLayer(), scaleFactor);

Does the base layer actually have contents? Why do we have to set contentsScale?

[Simon Fraser (smfr)]

Comment on attachment 265975 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=265975&action=review

> Source/WebCore/ChangeLog:15
> +        operations are don properly.

don

> Source/WebCore/platform/graphics/ca/win/PlatformCALayerWinInternal.cpp:230
> +#if HAVE(CACFLAYER_SETCONTENTSSCALE)
> +    float scaleFactor = CACFLayerGetContentsScale(owner()->platformLayer());
> +    CACFLayerSetContentsScale(layer.platformLayer(), scaleFactor);
> +#endif

This is confusing; I'm not sure how owner()->platformLayer() and layer.platformLayer() are different.

Why isn't this handled via GraphicsLayerCA::deviceOrPageScaleFactorChanged() as it is for other platforms?

[Brent Fulgham]

Comment on attachment 265975 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=265975&action=review

>> Source/WebCore/platform/graphics/ca/win/PlatformCALayerWinInternal.cpp:230
>> +#endif
> 
> This is confusing; I'm not sure how owner()->platformLayer() and layer.platformLayer() are different.
> 
> Why isn't this handled via GraphicsLayerCA::deviceOrPageScaleFactorChanged() as it is for other platforms?

Hmmmmm. I'll investigate before moving forward.

[Brent Fulgham]

(In reply to comment #3)
> Comment on attachment 265975 [details]
> Patch
> 
> View in context:
> https://bugs.webkit.org/attachment.cgi?id=265975&action=review
> 
> > Source/WebCore/platform/graphics/ca/win/WKCACFViewLayerTreeHost.cpp:128
> > +    CACFLayerSetContentsScale(layer->platformLayer(), scaleFactor);
> 
> Does the base layer actually have contents? Why do we have to set
> contentsScale?

According to the CA engineer who did the Windows work, this is needed for some arcane Windows-related reason.

[Brent Fulgham]

Created attachment 265987 [details]
Patch v2 (Ignore)

[Brent Fulgham]

(In reply to comment #5)
> Comment on attachment 265975 [details]
> Patch
> 
> View in context:
> https://bugs.webkit.org/attachment.cgi?id=265975&action=review
> 
> >> Source/WebCore/platform/graphics/ca/win/PlatformCALayerWinInternal.cpp:230
> >> +#endif
> > 
> > This is confusing; I'm not sure how owner()->platformLayer() and layer.platformLayer() are different.
> > 
> > Why isn't this handled via GraphicsLayerCA::deviceOrPageScaleFactorChanged() as it is for other platforms?
> 
> Hmmmmm. I'll investigate before moving forward.

It looks like this isn't needed. In fact, I wasn't even being consistent, as the other places where CACFLayerInsertSublayer was being called did not have the scaling factor change included.

[Brent Fulgham]

Created attachment 265989 [details]
Patch

[Brent Fulgham]

Committed r192700: <http://trac.webkit.org/changeset/192700>