[webkit-reviews] review requested: [Bug 14626] Make bidiReorderCharacters independent of RenderBlock : [Attachment 15562] Take bidiReorderCharacters out of RenderBlock

[bugzilla-daemon at webkit.org]

mitz at webkit.org has asked  for review:
Bug 14626: Make bidiReorderCharacters independent of RenderBlock
http://bugs.webkit.org/show_bug.cgi?id=14626

Attachment 15562: Take bidiReorderCharacters out of RenderBlock
http://bugs.webkit.org/attachment.cgi?id=15562&action=edit

------- Additional Comments from mitz at webkit.org
This is more or less what I would like to do as a first (and biggest) step.
After this separation of the render-tree case and the character buffer case is
done, I think you could add bidi-aware string drawing functionality to the
graphics layer (and expose it as WebKit SPI as necessary), and -- quite
independently -- continue cleaning up bidi.cpp (getting rid of the remaining
file statics).

No layout test regressions.

In terms of performance, I used both attachment 15549 and a PLT suite
consisting of all fast/text tests save one, and I have reached a point where
the numbers are better than TOT, and any further tweaking that I do does not
change the numbers significantly.