Bug 205563 – Precision of getClientRects(), getBoundingClientRect() differs depending whether simple line layout or line box layout is used

• Status: RESOLVED
• Resolution: FIXED
• Component: Text

[Daniel Bates]

While looking into bug #205527, I noticed that the rects returned by getClientRects(), getBoundingClientRect() differ in precision depending on whether they were computed using the simple line layout or line box layout code.

Steps to reproduce:

1. Open the test case attached to bug #205527: attachment #386273 [details].

Compare the rectangle returned under "getClientRects: Before selection:": <-- simple line layout

{"x":8,"y":26,"width":28.453125,"height":18,"top":26,"right":36.453125,"bottom":44,"left":8}

to the **second** rectangle returned under "getClientRects: After selection:": <-- line box layout

{"x":8,"y":26,"width":29,"height":18,"top":26,"right":37,"bottom":44,"left":8}

^^^ The width using simple line layout is 28.453125, but is 29 using line box layout.

[Radar WebKit Bug Importer]

<rdar://problem/58165528>

[Daniel Bates]

Created attachment 386358 [details]
Patch

[Daniel Bates]

This patch will fail to apply until bug #205527 is fixed.

[Daniel Bates]

Created attachment 386360 [details]
Patch

[Daniel Bates]

Thanks Alan!

[Daniel Bates]

Test failures are actually progressions: both the simple line layout and line box layout code paths now return the same rect.

The fast/dom/Range/getClientRects.html failure also reveals another issue: the fix for bug #172057 was not enough. More code changes are needed to solve bug #172057 if we still are interested in doing so.

[Daniel Bates]

Created attachment 386864 [details]
To land

[Daniel Bates]

Comment on attachment 386864 [details]
To land

Clearing flags on attachment: 386864

Committed r254091: <https://trac.webkit.org/changeset/254091>

[Daniel Bates]

All reviewed patches have been landed.  Closing bug.

[Truitt Savell]

Looks like the changes in https://trac.webkit.org/changeset/254091/webkit

broke fast/frames/iframe-translucent-background.html

tracking this in https://bugs.webkit.org/show_bug.cgi?id=205879

[Daniel Bates]

Comment on attachment 386864 [details]
To land

View in context: https://bugs.webkit.org/attachment.cgi?id=386864&action=review

> Source/WebCore/rendering/InlineTextBox.cpp:205
> +    auto logicalWidth = snappedSelectionRect.width();

This is wrong. Data type must be LayoutUnit or integer truncation can occur if second branch is taken. Fixed in <https://trac.webkit.org/changeset/254160/webkit>.

See bug #205889 for a way to fix this problem to prevent the same mistake in the future.

[Daniel Bates]

More layout test fixes landed in <https://trac.webkit.org/changeset/254167>

[Aakash Jain]

(In reply to Daniel Bates from comment #11)
> Fixed in  <https://trac.webkit.org/changeset/254160/webkit>.
This broke 6 API tests. Tracked in https://bugs.webkit.org/show_bug.cgi?id=205935