Linux: Do better font substitution for Chinese characters.

content/browser/renderer_host/render_sandbox_host_linux.cc

Index: content/browser/renderer_host/render_sandbox_host_linux.cc
===================================================================
--- content/browser/renderer_host/render_sandbox_host_linux.cc	(revision 258675)
+++ content/browser/renderer_host/render_sandbox_host_linux.cc	(working copy)
@@ -407,7 +407,7 @@
     }
 
     FcLangSet* langset = FcLangSetCreate();
-    MSCharSetToFontconfig(langset, charset);
+    bool is_english = MSCharSetToFontconfig(langset, charset);
 
     FcPattern* pattern = FcPatternCreate();
     // TODO(agl): FC_FAMILy needs to change
@@ -415,16 +415,18 @@
                        reinterpret_cast<const FcChar8*>(face.c_str()));
 
     std::string generic_font_name;
-    switch (fallback_family) {
-      case PP_BROWSERFONT_TRUSTED_FAMILY_SERIF:
-        generic_font_name = "Times New Roman";
-        break;
-      case PP_BROWSERFONT_TRUSTED_FAMILY_SANSSERIF:
-        generic_font_name = "Arial";
-        break;
-      case PP_BROWSERFONT_TRUSTED_FAMILY_MONOSPACE:
-        generic_font_name = "Courier New";
-        break;
+    if (is_english) {

[jungshik at Google, 2014/03/27 11:06:11]

Can we access Chrome's per-script font pref here? Then, fallback fonts can be
set to them per script/language. Perhaps, in the next round. 

[Lei Zhang, 2014/03/27 19:31:25]

I don't know. Added a TODO. Will investigate in the future.

+      switch (fallback_family) {
+        case PP_BROWSERFONT_TRUSTED_FAMILY_SERIF:
+          generic_font_name = "Times New Roman";
+          break;
+        case PP_BROWSERFONT_TRUSTED_FAMILY_SANSSERIF:
+          generic_font_name = "Arial";
+          break;
+        case PP_BROWSERFONT_TRUSTED_FAMILY_MONOSPACE:
+          generic_font_name = "Courier New";
+          break;
+      }
     }
     if (!generic_font_name.empty()) {
       const FcChar8* fc_generic_font_name =
@@ -547,7 +549,7 @@
 
   // MSCharSetToFontconfig translates a Microsoft charset identifier to a
   // fontconfig language set by appending to |langset|.
-  static void MSCharSetToFontconfig(FcLangSet* langset, unsigned fdwCharSet) {
+  static bool MSCharSetToFontconfig(FcLangSet* langset, unsigned fdwCharSet) {
     // We have need to translate raw fdwCharSet values into terms that
     // fontconfig can understand. (See the description of fdwCharSet in the MSDN
     // documentation for CreateFont:
@@ -566,6 +568,7 @@
     // So, for each of the documented fdwCharSet values I've had to take a
     // guess at the set of ISO 639-1 languages intended.
 
+    bool is_english = false;

[jungshik at Google, 2014/03/27 11:06:11]

Perhaps, it'd be better to change change |is_english| to |is_lgc|
(Latin-Greek-Cyrillic) and set it to true for Windows code pages covering LGC
script. 

Arial/TNR/Courier covers more than LGC, but most Linux distros have very old
versions of them and the coverage does not go beyond LGC. 

[Lei Zhang, 2014/03/27 19:31:25]

Done.
On my system, half the characters in the PDF rendered with the default font.
With the zh-cn language preference set, it selected a simplified Chinese font
which has much better coverage.

     switch (fdwCharSet) {
       case NPCharsetAnsi:
       // These values I don't really know what to do with, so I'm going to map
@@ -574,6 +577,7 @@
       case NPCharsetMac:
       case NPCharsetOEM:
       case NPCharsetSymbol:
+        is_english = true;
         FcLangSetAdd(langset, reinterpret_cast<const FcChar8*>("en"));
         break;
       case NPCharsetBaltic:
@@ -582,12 +586,11 @@
         FcLangSetAdd(langset, reinterpret_cast<const FcChar8*>("lv"));
         FcLangSetAdd(langset, reinterpret_cast<const FcChar8*>("lt"));
         break;
-      // TODO(jungshik): Would we be better off mapping Big5 to zh-tw
-      // and GB2312 to zh-cn? Fontconfig has 4 separate orthography
-      // files (zh-{cn,tw,hk,mo}.
       case NPCharsetChineseBIG5:
+        FcLangSetAdd(langset, reinterpret_cast<const FcChar8*>("zh-tw"));
+        break;
       case NPCharsetGB2312:
-        FcLangSetAdd(langset, reinterpret_cast<const FcChar8*>("zh"));
+        FcLangSetAdd(langset, reinterpret_cast<const FcChar8*>("zh-cn"));
         break;
       case NPCharsetEastEurope:
         // A scattering of eastern European languages.
@@ -631,6 +634,7 @@
       // Don't add any languages in that case that we don't recognise the
       // constant.
     }
+    return is_english;
   }
 
   void SendRendererReply(const std::vector<int>& fds, const Pickle& reply,