This made test http/tests/download/literal-utf-8.html to start failing. The problem is that I removed the conversion made by String::fromUTF8WithLatin1Fallback that was added in r176930. I removed it because that made fast/dom/HTMLAnchorElement/anchor-file-blob-download-includes-unicode.html to timeout.
Created attachment 303055 [details] Patch
Comment on attachment 303055 [details] Patch I don't understand it. You convert the header from UTF-8 to the internal encoding (UTF-16)... but only if it's ASCII...?
(In reply to comment #2) > Comment on attachment 303055 [details] > Patch > > I don't understand it. You convert the header from UTF-8 to the internal > encoding (UTF-16)... but only if it's ASCII...? In r176930 ResourseResponse was changes to not apply any encoding conversion from message headers when adding them to the header map. So, we create the String with String constructor no String__fromaUTF8() so that we don't actually make any charset conversion. The header could be either utf8 or latin1, so we try to convert to utf8, only when it's not, falling back to latin1 if the conversion fails.
Committed r213226: <http://trac.webkit.org/changeset/213226>
Re-opened since this is blocked by bug 169051
Committed r213273: <http://trac.webkit.org/changeset/213273>