[webkit-reviews] review denied: [Bug 3739] Unreproducible - Assertion failure in isEqualIgnoringAffinity on double-click : [Attachment 4191] Force DOWNSTREAM at end of paragraph

[bugzilla-request-daemon at opendarwin.org]

David Harrison <harrison at apple.com> has denied Alexey Proskuryakov
<ap at nypop.com>'s request for review:
Bug 3739: Unreproducible - Assertion failure in isEqualIgnoringAffinity on
double-click
http://bugzilla.opendarwin.org/show_bug.cgi?id=3739

Attachment 4191: Force DOWNSTREAM at end of paragraph
http://bugzilla.opendarwin.org/attachment.cgi?id=4191&action=edit

------- Additional Comments from David Harrison <harrison at apple.com>
This patch does not address the root cause of the problem, which is that we
need to define how affinity applies to newline characters in text nodes with
whitespace:PRE style.	The assert fires because of an inconsistency:
RenderText::inlineBox() acts as if affinity applies to newlines in preformatted
text, but endOfParagraph() does not (it always produces DOWNSTREAM affinity).

The better fix is to decide a policy and make it consistent.  My gut says to
have affinity NOT apply in this case, because affinity is a "line wrap" concept
not a paragraph break concept (even if that break is within a text node, it is
not the same as "line wrap").  This would mean modifying
RenderText::inlineBox() to heed newlines in whitespace:PRE.  It probably means
other changes, as well, because this nuance was AFAIK not considererd when
writing the affinity-related code.  For example,  I locally fixed
RenderText::inlineBox(), and single-clicking at the end of the line leaves the
cursor blinking at the beginning of the second line.