Phone numbers written with Eastern Arabic numerals are garbled
Created attachment 228791 [details] Patch
Comment on attachment 228791 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=228791&action=review > Source/WebCore/ChangeLog:11 > + If we find some Other Neutral characters between two Eastern Arabic numbers, > + we used to label the Other Neutral characters as RTL no matter what. Instead, > + we can label them as more arabic numbers so that they don't get re-ordered > + or reversed. It would have been nice to have a reference to the section of the UBA that we are following here. > Source/WebCore/platform/text/BidiResolver.h:840 > // Begin an R run for the neutrals. This comment (which is sadly a “what” comment rather than a “why” comment) is not longer true. It went from useless to outright misleading! > LayoutTests/ChangeLog:13 > + * platform/mac/fast/text/international/bidi-neutral-run-expected.txt: Update expected results Are all of these changes expected? Did the UBA change or was the implementation in WebKit always wrong?
I don’t undertand why this patch is correct. The expected result doesn’t match what I get at <http://unicode.org/cldr/utility/bidi.jsp?a=%28٤%D9%A0٨%29+٢١٥-٨١٩٨&p=LTR>. WebKit’s current behavior does match that.