[webkit-reviews] review granted: [Bug 12448] small text which is scaled to be large renders pixelated : [Attachment 80741] Patch v2

[bugzilla-daemon at webkit.org]

Dirk Schulze <krit at webkit.org> has granted Nikolas Zimmermann
<zimmermann at kde.org>'s request for review:
Bug 12448: small text which is scaled to be large renders pixelated
https://bugs.webkit.org/show_bug.cgi?id=12448

Attachment 80741: Patch v2
https://bugs.webkit.org/attachment.cgi?id=80741&action=review

------- Additional Comments from Dirk Schulze <krit at webkit.org>
View in context: https://bugs.webkit.org/attachment.cgi?id=80741&action=review

> LayoutTests/ChangeLog:18
> +	     Since we switched to use floating-point precision metrics, the
actual drawn selection may exceed the red rectangle to up to 1px.

a 'to' too much

> LayoutTests/ChangeLog:21
> +	     In theory the red re

sentence is incomplete.

> LayoutTests/ChangeLog:22
> +	     The red rectangle represents the results of the text query, using
getExtentOfChar(), which gives the precise text metrics using floating-point
accurancy.
> +	     In theory the red re
> +	     The selection rects are aligned on integer boundaries.

remove these lines completely

> LayoutTests/ChangeLog:27
> +	   - svg/text/selection-(doubleclick|tripleclick).svg:
> +	     These tests use DRTs dumpSelectionRect() which draws a red
rectangle, highlighting the SelectionController::bounds() obtained through
RenderView.
> +	     In theory this red rectangle, should be equal to the drawn blue
selection rect. Again it isn't, as the bounds() methods uses
selectionRectForRepaint()
> +	     of RenderText, which returns an enclosingIntRect of the actual
selection. The drawn selection starts at y=0.93 in the first testcase, the
selectionRectForRepaint()

You should concentrate on the facts, you already wrote about the red rects
before :-/

> LayoutTests/ChangeLog:33
> +	     the integer variants ascent/descent/height - hopefully we'll see
consistent results across platforms for these tests now, if not please comment
in the bug report.

There are some minor misspellings, talked on IRC about it. Won't mention these
mistakes here.

You're talking about a difference of <1px, but we can see more pixel
differences in some cases. We talked about the reason on IRC. But please add a
short note about the scaling of hinting that is responsible for this as well.