GitLab

CachingWordShapeIterator splits runs in the logical order, but
ShapeResult::offsetForPosition() expected the input to be in the visual
order.

This patch fixes ShapeResult::offsetForPosition() to handle the list
of ShapeResult from CachingWordShapeIterator in the logical order.

This is a regression of "Change Font::offsetForPositionForComplexText
to use CachingWordShapeIterator"[1]. Unfortunately tests in the CL did
not catch this logic/visual inconsistency. The test was updated to
include it.

The test string has a diacritic mark at offset 1, and the test ensures
offsetForPosition() dose not return 1. All platforms pass this criteria,
however, only Mac test runner does not return offset 5 (Mac content
shell is fine) so Mac -expeted is added separately.

[1] https://codereview.chromium.org/1248453004/

BUG=518347
TEST=editing/selection/offset-from-point-complex-scripts.html

Review URL: https://codereview.chromium.org/1314543003

git-svn-id: svn://svn.chromium.org/blink/trunk@201181 bbb929c8-8fbe-4397-9dbb-9b2b20218538