299063 - skipped border edges on split (broken) inlines on wrong side for RTL

Status: RESOLVED FIXED

(Core :: Layout: Text and Fonts, defect)

Description

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

When an inline element with a border is split across multiple lines, the borders
are supposed to be open at the continuation.  For RTL inlines, the border is
open at the wrong end.

Steps to reproduce: load attached testcase

Expected results: the two halves are mirror images of each other

Actual results: the RTL case has the borders open at the wrong end (in addition
to some other bugs)

Comment 1

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Attachment: testcase

Comment 2

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Attachment: patch that should work once bug 299065 is fixed

I plan to check a modified version of this (since more changes will be needed
there) into REFLOW_20050429_BRANCH.

Comment 3

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

(But it won't do anything there either since embedding levels don't seem to be
set on inlines; fixing that is bug 299065.)

Comment 4

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Actually, according to http://www.w3.org/TR/CSS21/box.html#q11 , that's the
wrong way to fix it.  We should actually depend on the 'direction' property of
the frame.  (But we should still split the frame as described in bug 299065.)

Comment 5

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

...but getting the leftmost piece on the first line might be a bit tricky.

Comment 6

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Although I actually think that would be confusing to read, and it makes more
sense to have the padding, border, and margin at the start of the element (and
potentially have left padding twice and right not at all, or vice-versa).

Comment 7

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

...except there would be no way to make the *-start and *-end versions work that
way, since that computation is based on 'direction' rather than bidi reordering.

Comment 8

[Uri Bernstein]

=> All/All

Comment 9

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Also need to fix nsInlineFrame::ReflowFrames:

    // Compute final width
    aMetrics.width = size.width;
    if (nsnull == mPrevInFlow) {
      aMetrics.width += aReflowState.mComputedBorderPadding.left;
    }
    if (NS_FRAME_IS_COMPLETE(aStatus)) {
      aMetrics.width += aReflowState.mComputedBorderPadding.right;
    }

Comment 10

[Uri Bernstein]

Bug 328168 describes a more general case of this bug.

Comment 11

[Uri Bernstein]

Fixed by Haamed's patch for bug 328168.

Comment 12

[(no longer active)]

Mass-assigning the new rtl keyword to RTL-related (see bug 349193).