Double-struck unicode chars in \texorpdfstring

Question

Consider the following example (PDFLaTeX, TeXLive 2011):

\documentclass{article}
\usepackage{xspace,amssymb,mathtools,fixltx2e}
% This isn't *stricly* necessary, of course...
\ifdefined\TextOrMath
  \def\ensuremath#1{\TextOrMath{\(\relax #1\)}{#1}}
\fi
\usepackage[utf8]{inputenc}
\usepackage[unicode]{hyperref}
\def\defTeXorMathorUnicodePDFcmd#1#2#3{%
  \def#1{\texorpdfstring{\ensuremath{#2}}{#3}\xspace}}
% Common double-struck characters such as ℂ, ℍ, ℕ, ℙ, ℚ, ℝ and ℤ are
% set up to work by inputenc.sty.  They are in the range u+2102-2149
% along with some other characters that aren't in amsmath's blackboard
% bold font such as ⅆ and ⅈ.
\defTeXorMathorUnicodePDFcmd\CC{\mathbb C}{ℂ}
% 𝔽 is an example of a character that inputenc.sty doesn't make work. 
% It - along with the characters 𝔸, 𝔹, 𝔻, &c - are in the completely 
% different range u+1D538-1D7E1.
\defTeXorMathorUnicodePDFcmd\FF{\mathbb F}{𝔽}
\begin{document} \section{\CC and \FF, [\ldots]} \end{document}

The purpose of \CC is to get ℂ in both the text and the PDF TOC when used in for example a section header. What's needed to get \FF to do the same for 𝔽 (and similarly for the other "uncommon" double-struck characters)?

PS: I copied the characters from here: http://graphemica.com/search?q=double+struck

Answer 1

This is not an proper answer, but an analysis of the problem.

The problem seems pdf-viewer dependent. I see 𝔽 in the bookmarks with Adobe Reader, but not with Evince (a Linux pdf-viewer). The error message hints at the source of the error : it says that ' \xed\xa0\xb5\xed\xb4\xbd' (for 𝔽) is not a valid UTF8 sequence. Indeed, it is not a UTF-8 byte sequence, but a CESU-8 sequence. They are indistinguishable most of the common characters (on the basic multilingual plane), but 𝔽 is on another plane.

I don't know whether the pdf-format wants UTF-8 or CESU-8 encoded text, so I don't know if the bug is on the pdf generation side (in pdflatex) or on the viewer side.