Created attachment 172739 [details] testcase Original report: http://code.google.com/p/chromium/issues/detail?id=158969 Test case attached (from the original report). Arabic-Indic digits (U+06f0-U+06f9) should appear in increasing order from left to right. HarfBuzzShaper renders these digits right to left in increasing order.
Created attachment 172741 [details] Patch
Hi Tony, Could you take a look?
Comment on attachment 172741 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=172741&action=review > Source/WebCore/platform/graphics/harfbuzz/ng/HarfBuzzShaper.cpp:235 > + // HarfBuzz when we are calculating widths (except for directionalOverride() is set). "except when directionOverride() is set" reads better in English.
Created attachment 173155 [details] Patch for landing
Comment on attachment 172741 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=172741&action=review Thank you for review! >> Source/WebCore/platform/graphics/harfbuzz/ng/HarfBuzzShaper.cpp:235 >> + // HarfBuzz when we are calculating widths (except for directionalOverride() is set). > > "except when directionOverride() is set" reads better in English. Done.
Comment on attachment 173155 [details] Patch for landing Clearing flags on attachment: 173155 Committed r133983: <http://trac.webkit.org/changeset/133983>
All reviewed patches have been landed. Closing bug.
Humm. In general webkit must always pass direction down to HarfBuzz. It would be wrong to rely on autodetection in HarfBuzz since Webkit would have its own idea of the run direction.