Improve XeTeX PDF outline support

[Masamichi HOSODA]

texinfo.tex ver. 2016-02-07.16 can not compile following attached files.

test-U201E.texi
test-set-value.texi

I've fixed it.
Here's the patch for texinfo.tex ver. 2016-02-07.16.

ChangeLog:

Improve XeTeX PDF outline support

2016-02-XX  Masamichi Hosoda  <address@hidden>

        * doc/texinfo.tex:
        Improve XeTeX PDF outline support.
        (\pdfmkdest): Add \indexnofonts and \makevalueexpandable,
        (\dopdfoutline): Add \turnoffactive,
        (\pdfmakeoutlines): Add some comments. Use \let instead of \def.

\input texinfo.tex @c -*- coding: utf-8 -*-

@documentencoding UTF-8

@contents

@node „
@chapter test „: U+201E

„: U+201E DOUBLE LOW-9 QUOTATION MARK

@bye

\input texinfo.tex

@documentencoding UTF-8

@contents

@set foo-bar_ test test test

@node nøùü @address@hidden
@chapter øùü @address@hidden test

øùü

@bye

--- texinfo.tex.org     2016-02-08 01:22:13.800799100 +0900
+++ texinfo.tex 2016-02-08 01:23:28.667073400 +0900
@@ -1455,26 +1455,38 @@
 \ifx\XeTeXrevision\thisisundefined
 \else
   \pdfmakepagedesttrue \relax
+  % Emulate the primitive of pdfTeX
   \def\pdfdest name#1 xyz{%
     \special{pdf:dest (name#1) address@hidden /XYZ @xpos @ypos]}%
   }
-  \def\pdfmkdest#1{%
-    \special{pdf:dest (name#1) address@hidden /XYZ @xpos @ypos]}%
-  }
+  \def\pdfmkdest#1{{%
+    % We have to set dummies so commands such as @code, and characters
+    % such as \, aren't expanded when present in a section title.
+    \indexnofonts
+    \makevalueexpandable
+    % In the case of XeTeX, xdvipdfmx converts strings to UTF-16.
+    % Therefore \txiescapepdf is not necessary.
+    \safewhatsit{\pdfdest name{#1} xyz}%
+  }}
   %
   \def\dopdfoutline#1#2#3#4{%
     \edef\pdfoutlinedest{#3}%
     \ifx\pdfoutlinedest\empty
       \def\pdfoutlinedest{#4}%
     \fi
-    %
-    \edef\pdfoutlinetext{#1}%
-    %
+    \turnoffactive
+    % In the case of XeTeX, xdvipdfmx converts strings to UTF-16.
+    % Therefore \txiescapepdf is not necessary.
     \special{pdf:out [-] #2 << /Title (#1) /A << /S /GoTo /D 
(name\pdfoutlinedest) >> >> }%
   }
   %
   \def\pdfmakeoutlines{%
     \begingroup
+      %
+      % In the case of XeTeX, counts of subentries is not necesary.
+      % Therefore, read toc only once.
+      %
+      % We use the node names as the destinations.
       \def\partentry##1##2##3##4{}% ignore parts in the outlines
       \def\numchapentry##1##2##3##4{%
         \dopdfoutline{##1}{1}{##3}{##4}}%
@@ -1485,24 +1497,33 @@
       \def\numsubsubsecentry##1##2##3##4{%
         \dopdfoutline{##1}{4}{##3}{##4}}%
       %
-      \def\appentry{\numchapentry}%
-      \def\appsecentry{\numsecentry}%
-      \def\appsubsecentry{\numsubsecentry}%
-      \def\appsubsubsecentry{\numsubsubsecentry}%
-      \def\unnchapentry{\numchapentry}%
-      \def\unnsecentry{\numsecentry}%
-      \def\unnsubsecentry{\numsubsecentry}%
-      \def\unnsubsubsecentry{\numsubsubsecentry}%
+      \let\appentry\numchapentry%
+      \let\appsecentry\numsecentry%
+      \let\appsubsecentry\numsubsecentry%
+      \let\appsubsubsecentry\numsubsubsecentry%
+      \let\unnchapentry\numchapentry%
+      \let\unnsecentry\numsecentry%
+      \let\unnsubsecentry\numsubsecentry%
+      \let\unnsubsubsecentry\numsubsubsecentry%
+      %
+      % In the case of XeTeX, xdvipdfmx converts strings to UTF-16.
+      % Therefore, the encoding and the language may not be considered.
       %
       \indexnofonts
       \setupdatafile
-      %
+      % We can have normal brace characters in the PDF outlines, unlike
+      % Texinfo index files.  So set that up.
       \def\{{\lbracecharliteral}%
       \def\}{\rbracecharliteral}%
       \catcode`\\=\active \otherbackslash
       \input \tocreadfilename
     \endgroup
   }
+  {\catcode`[=1 \catcode`]=2
+   \catcode`{=\other \catcode`}=\other
+   \gdef\lbracecharliteral[{]%
+   \gdef\rbracecharliteral[}]%
+  ]
 
   \special{pdf:docview << /PageMode /UseOutlines >> }
   \special{pdf:tounicode UTF8-UTF16 }