Bug 62478 – RenderText::absoluteRectsForRange() and absoluteQuadsForRange() have nearly duplicate code

• Component: Layout and Rendering
• Status: RESOLVED FIXED

[David Kilzer (:ddkilzer)]

* SUMMARY
In RenderText::absoluteRectsForRange(), the following code is used in an else clause:

            IntRect r = box->selectionRect(0, 0, start, realEnd);
            if (!r.isEmpty()) {
                if (!useSelectionHeight) {

In RenderText::absoluteQuadsForRange(), the same code uses a slightly different check:

            IntRect r = box->selectionRect(0, 0, start, realEnd);
            if (r.height()) {
                if (!useSelectionHeight) {

It seems like these methods should be using the same check, and it would be great if the duplicate code could be extracted into a local static method.

[Simon Fraser (smfr)]

From bug 34239

[Emil A Eklund]

Created attachment 101949 [details]
Patch

[Simon Fraser (smfr)]

Comment on attachment 101949 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=101949&action=review

> Source/WebCore/rendering/RenderText.cpp:281
> +            // change the height and y position because selectionRect uses selection-specific values

Please change this to use sentence case and end with a period.

> Source/WebCore/rendering/RenderText.cpp:324
> +            if (!rect.size().isZero())

Don't we have rect.isEmpty()?

> Source/WebCore/rendering/RenderText.cpp:405
> +            if (!rect.size().isZero())

Ditto.

[Emil A Eklund]

Created attachment 102019 [details]
Patch for landing

[Emil A Eklund]

Thanks for the quick review.

(In reply to comment #3)
> (From update of attachment 101949 [details])
> View in context: https://bugs.webkit.org/attachment.cgi?id=101949&action=review
> 
> > Source/WebCore/rendering/RenderText.cpp:281
> > +            // change the height and y position because selectionRect uses selection-specific values
> 
> Please change this to use sentence case and end with a period.

Done.

> 
> > Source/WebCore/rendering/RenderText.cpp:324
> > +            if (!rect.size().isZero())
> 
> Don't we have rect.isEmpty()?

We do but I want to check for zero as opposed to empty here as we have a couple of tests that expect getClientRects to return a rect with a zero width.

Added FloatRect::isZero that calls FloatSize::isZero and changed the code in RenderText to use that method.

[WebKit Review Bot]

Comment on attachment 102019 [details]
Patch for landing

Clearing flags on attachment: 102019

Committed r91763: <http://trac.webkit.org/changeset/91763>

[WebKit Review Bot]

All reviewed patches have been landed.  Closing bug.

[Ryosuke Niwa]

This test has been failing on GTK:
http://build.webkit.org/old-results/GTK%20Linux%2032-bit%20Release/r91765%20(15930)/fast/dom/Range/getClientRects-pretty-diff.html

[Simon Fraser (smfr)]

Comment on attachment 102019 [details]
Patch for landing

View in context: https://bugs.webkit.org/attachment.cgi?id=102019&action=review

> Source/WebCore/rendering/RenderText.cpp:407
> +            FloatRect rect = absoluteQuadForTextBox(box, start, end, useSelectionHeight);
> +            if (!rect.isZero())
> +                quads.append(localToAbsoluteQuad(rect));

This is wrong. It converts local to absolute twice.