fast/css3-text/css3-text-decoration/text-decoration-thickness.html fails on GTK

[Martin Robinson]

It looks like the font size is insufficient to make the test pass on GTK+. We can fix it by adding a few more orders of magnitude to the font size.

[Martin Robinson]

Looks like this might actually be a bug in the Cairo GraphicsContext. Investigating now.

[Martin Robinson]

fast/css3-text/css3-text-decoration/text-decoration-thickness.html and fast/css3-text/css3-text-decoration/text-decoration-style-double-space-scales.html are also failing because of this issue.

[Martin Robinson]

*** Bug 126879 has been marked as a duplicate of this bug. ***

[Martin Robinson]

Created attachment 226217 [details]
Patch

[Martin Robinson]

I have ported mmaxfield's original CG patch to the Cairo GraphicsContext, which seems to fix the tests.

[Myles C. Maxfield]

Comment on attachment 226217 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=226217&action=review

> Source/WebCore/platform/graphics/cairo/GraphicsContextCairo.cpp:617
> +static FloatRect computeLineBoundsAndAntialiasingModeForText(GraphicsContext& initialContext, const FloatPoint& point, float width, bool printing, bool& shouldAntialias, Color& color)

Perhaps there is some cause for duplicate code reuse. Is there enough commonality here that the implementations could be unified?

> LayoutTests/fast/css3-text/css3-text-decoration/text-decoration-skip/text-decoration-skip-tall-underlines.html:-10
> -<div style="left: -160px; top: -320px; position: absolute; -webkit-transform: scale(20); font-family: helvetica; -webkit-transform-origin: left top; display: inline-block; text-decoration: underline; -webkit-text-decoration-skip: ink;">gy</div>

Glad you caught this. This is good.

[Martin Robinson]

(In reply to comment #6)
> (From update of attachment 226217 [details])
> View in context: https://bugs.webkit.org/attachment.cgi?id=226217&action=review
> 
> > Source/WebCore/platform/graphics/cairo/GraphicsContextCairo.cpp:617
> > +static FloatRect computeLineBoundsAndAntialiasingModeForText(GraphicsContext& initialContext, const FloatPoint& point, float width, bool printing, bool& shouldAntialias, Color& color)
> 
> Perhaps there is some cause for duplicate code reuse. Is there enough commonality here that the implementations could be unified?

I'd be happy to move the code to a shared GraphicsContext method if you think it makes sense. I've tried to make most of the code platform-independent. Would getting the transformation matrix via GraphicsContext::getCTM be sufficient for iOS and Mac?

[Myles C. Maxfield]

GraphicsContext::getCTM(DefinitelyIncludeDeviceScale) should do the trick.

[Martin Robinson]

(In reply to comment #8)
> GraphicsContext::getCTM(DefinitelyIncludeDeviceScale) should do the trick.

Thanks for the information. I'll try to have a unified patch tomorrow.

[Martin Robinson]

Created attachment 226460 [details]
Patch

[Myles C. Maxfield]

r=me but i'm not a reviewer

[Tim Horton]

Comment on attachment 226460 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=226460&action=review

> Source/WebCore/platform/graphics/GraphicsContext.cpp:1012
> +    return FloatRect(FloatPoint(origin.x(), origin.y()), FloatSize(width, thickness));

There is a FloatRect constructor that takes all four.

[Martin Robinson]

Committed r166902: <http://trac.webkit.org/changeset/166902>

[Michal Pakula vel Rutka]

*** Bug 129734 has been marked as a duplicate of this bug. ***

[Michal Pakula vel Rutka]

*** Bug 127282 has been marked as a duplicate of this bug. ***

[Michal Pakula vel Rutka]

*** Bug 128664 has been marked as a duplicate of this bug. ***