Enhance auto hyphenation.

1. Bypass leading white spaces, since webkit does pass them down.
2. Return better hyphenation point, since previously it's more
aggressive and causes display issue in Google books.

issue: 2672163
Change-Id: I8ae47f7c553f533f752d6f7c697cf2fffd421e5b

1 files changed, 13 insertions, 5 deletions

diff --git a/WebCore/platform/text/android/HyphenationAndroid.cpp b/WebCore/platform/text/android/HyphenationAndroid.cpp
index e8ba5ef..cff0a66 100644
--- a/WebCore/platform/text/android/HyphenationAndroid.cpp
+++ b/WebCore/platform/text/android/HyphenationAndroid.cpp
@@ -75,6 +75,7 @@ size_t lastHyphenLocation(const UChar* characters, size_t length, size_t beforeI
         return 0;
 
     char word[maxWordLen];
+    int wordLength = 0;
     for (size_t i = 0; i < length; ++i) {
         const UChar ch = characters[i];
         // Only English for now.
@@ -82,17 +83,24 @@ size_t lastHyphenLocation(const UChar* characters, size_t length, size_t beforeI
         // detection or rely on the langAttr in the html element.  Though
         // seems right now the langAttr is not used or quite implemented in
         // webkit.
-        if (!isASCIIAlpha(ch))
+        if (!isASCIIAlpha(ch)) {
+            // Bypass leading spaces.
+            if (isASCIISpace(ch) && !wordLength)
+              continue;
             return 0;
-        word[i] = ch;
+        }
+        word[wordLength++] = ch;
     }
+    if (wordLength < minWordLen)
+        return 0;
 
     static const int extraBuffer = 5;
+    const int leadingSpacesCount = length - wordLength;
     char hyphens[maxWordLen + extraBuffer];
-    if (!hnj_hyphen_hyphenate(dict, word, length, hyphens)) {
-        for (size_t i = beforeIndex - 1; i > 0; --i) {
+    if (!hnj_hyphen_hyphenate(dict, word, wordLength, hyphens)) {
+        for (size_t i = beforeIndex - 2 - leadingSpacesCount; i > 0; --i) {
             if (hyphens[i] & 1)
-                return i + 1;
+                return i + 1 + leadingSpacesCount;
         }
     }