Following line inside "float InlineTextBox::positionForOffset(int offset) const" might be wrong: // FIXME: Do we need to add rightBearing here? return f.selectionRectForText(TextRun(text->text()->characters() + m_start, m_len, textRenderer()->allowTabs(), textPos(), m_expansion, trailingExpansionBehavior(), !isLeftToRightDirection(), m_dirOverride), IntPoint(logicalLeft(), 0), 0, from, to).maxX(); convert "logicalLeft()" to "int" in "IntPoint(logicalLeft(), 0)" should not be done. For example, <textarea id="textarea_rtl_no_wrap" style="padding:0; margin:0; white-space:nowrap; width: 100px; font: 8px" dir="rtl">דגלחכ גדכ לחידגכ יחעדד</textarea> the width of the text is 120, the logicalLeft() is -19.775, which is round to -19. Then, the positionForOffset(0) will be "120 - 19 = 101" which is greater than the textarea width: 100px.
Created attachment 88331 [details] patch w/ layout test Dan, Enrica, Could you please help on verifying the layout test result on editing/selection/mixed-editability-10.html and platform/mac/editing/input/caret-primary-bidi.html? I tried moving caret in caret-primary-bidi.html, and I did not see any where that the caret becomes invisible. Seems the result is acceptable. But I am not sure on mixed-editability-10.html. The result changed for the first <div> with one regression and one progression.
The patch is not completed. It breaks mixed-editability-10.html in the sense that when clicking left of the span, the caret overlaps with the span, the caret should be placed before the span. The code in the patch itself seems reasonable. But there are probably some other code pieces that do float/int conversion or float rounding or other float hacking that need to be taken care of.
Comment on attachment 88331 [details] patch w/ layout test View in context: https://bugs.webkit.org/attachment.cgi?id=88331&action=review > LayoutTests/platform/mac/editing/input/caret-primary-bidi-expected.txt:3 > -0: 124,508,0,28 > +0: 125,508,0,28 > 1: 21,564,0,28 > 2: 36,564,0,28 These output makes no sense to me. We should come up with a better output here.
@Alan - I looked into it and it seems that it is in IFC part and I am sure that you might have already fixed it and improved it. Anything to do it? https://searchfox.org/wubkat/source/Source/WebCore/layout/integration/inline/InlineIteratorTextBox.cpp#93