Fix ToUnicode generation bug.

Code from arthurhsu@chromium.org Original CL: http://codereview.appspot.com/5492061/

BUG=Chromium:104062

Review URL: http://codereview.appspot.com/5498064

git-svn-id: http://skia.googlecode.com/svn/trunk@2944 2bbb7eff-a529-9590-31e7-b0007b416f81

2 files changed, 63 insertions, 32 deletions

diff --git a/src/pdf/SkPDFFont.cpp b/src/pdf/SkPDFFont.cpp
index 0da5698bc9..404c762457 100644
--- a/src/pdf/SkPDFFont.cpp
+++ b/src/pdf/SkPDFFont.cpp
@@ -474,48 +474,38 @@ void append_cmap_sections(const SkTDArray<SkUnichar>& glyphToUnicode,
     SkTDArray<BFRange> bfrangeEntries;
 
     BFRange currentRangeEntry;
-    bool haveBase = false;
-    int continuousEntries = 0;
+    bool rangeEmpty = true;
+    const int count = glyphToUnicode.count();
 
-    for (int i = 0; i < glyphToUnicode.count(); ++i) {
-        if (glyphToUnicode[i] && (subset == NULL || subset->has(i))) {
+    for (int i = 0; i < count + 1; ++i) {
+        bool inSubset = i < count && (subset == NULL || subset->has(i));
+        if (!rangeEmpty) {
             // PDF spec requires bfrange not changing the higher byte,
             // e.g. <1035> <10FF> <2222> is ok, but
             //      <1035> <1100> <2222> is no good
-            if (haveBase) {
-                ++continuousEntries;
-                if (i == currentRangeEntry.fStart + continuousEntries &&
-                            (i >> 8) == (currentRangeEntry.fStart >> 8) &&
-                            glyphToUnicode[i] == (currentRangeEntry.fUnicode +
-                                                  continuousEntries)) {
-                    currentRangeEntry.fEnd = i;
-                    if (i == glyphToUnicode.count() - 1) {
-                        // Last entry is in a range.
-                        bfrangeEntries.push(currentRangeEntry);
-                    }
-                    continue;
-                }
-
-                // Need to have at least 2 entries to form a bfrange.
-                if (continuousEntries >= 2) {
+            bool inRange =
+                i == currentRangeEntry.fEnd + 1 &&
+                i >> 8 == currentRangeEntry.fStart >> 8 &&
+                i < count &&
+                glyphToUnicode[i] == currentRangeEntry.fUnicode + i -
+                                         currentRangeEntry.fStart;
+            if (!inSubset || !inRange) {
+                if (currentRangeEntry.fEnd > currentRangeEntry.fStart) {
                     bfrangeEntries.push(currentRangeEntry);
                 } else {
                     BFChar* entry = bfcharEntries.append();
                     entry->fGlyphId = currentRangeEntry.fStart;
                     entry->fUnicode = currentRangeEntry.fUnicode;
                 }
-                continuousEntries = 0;
+                rangeEmpty = true;
             }
-
-            if (i != glyphToUnicode.count() - 1) {
-                currentRangeEntry.fStart = i;
-                currentRangeEntry.fEnd = i;
-                currentRangeEntry.fUnicode = glyphToUnicode[i];
-                haveBase = true;
-            } else {
-                BFChar* entry = bfcharEntries.append();
-                entry->fGlyphId = i;
-                entry->fUnicode = glyphToUnicode[i];
+        }
+        if (inSubset) {
+            currentRangeEntry.fEnd = i;
+            if (rangeEmpty) {
+              currentRangeEntry.fStart = i;
+              currentRangeEntry.fUnicode = glyphToUnicode[i];
+              rangeEmpty = false;
             }
         }
     }
diff --git a/tests/ToUnicode.cpp b/tests/ToUnicode.cpp
index ad1e0a990b..c0a945a5e7 100644
--- a/tests/ToUnicode.cpp
+++ b/tests/ToUnicode.cpp
@@ -16,17 +16,21 @@
 #include "SkStream.h"
 
 static bool stream_equals(const SkDynamicMemoryWStream& stream, size_t offset,
-                          const void* buffer, size_t len) {
+                          const char* buffer, size_t len) {
     SkAutoDataUnref data(stream.copyToData());
     if (offset + len > data.size()) {
         return false;
     }
+    if (len != strlen(buffer)) {
+        return false;
+    }
     return memcmp(data.bytes() + offset, buffer, len) == 0;
 }
 
 void append_cmap_sections(const SkTDArray<SkUnichar>& glyphToUnicode,
                           const SkPDFGlyphSet* subset,
                           SkDynamicMemoryWStream* cmap);
+
 static void TestToUnicode(skiatest::Reporter* reporter) {
     SkTDArray<SkUnichar> glyphToUnicode;
     SkTDArray<uint16_t> glyphsInSubset;
@@ -86,6 +90,43 @@ endbfrange\n";
 
     REPORTER_ASSERT(reporter, stream_equals(buffer, 0, expectedResult,
                                             buffer.getOffset()));
+
+    glyphToUnicode.reset();
+    glyphsInSubset.reset();
+    SkPDFGlyphSet subset2;
+
+    // Test mapping:
+    //           I  n  s  t  a  l
+    // Glyph id 2c 51 56 57 44 4f
+    // Unicode  49 6e 73 74 61 6c
+    for (size_t i = 0; i < 100; ++i) {
+      glyphToUnicode.push(i + 29);
+    }
+
+    glyphsInSubset.push(0x2C);
+    glyphsInSubset.push(0x44);
+    glyphsInSubset.push(0x4F);
+    glyphsInSubset.push(0x51);
+    glyphsInSubset.push(0x56);
+    glyphsInSubset.push(0x57);
+
+    SkDynamicMemoryWStream buffer2;
+    subset2.set(glyphsInSubset.begin(), glyphsInSubset.count());
+    append_cmap_sections(glyphToUnicode, &subset2, &buffer2);
+
+    char expectedResult2[] =
+"4 beginbfchar\n\
+<002C> <0049>\n\
+<0044> <0061>\n\
+<004F> <006C>\n\
+<0051> <006E>\n\
+endbfchar\n\
+1 beginbfrange\n\
+<0056> <0057> <0073>\n\
+endbfrange\n";
+
+    REPORTER_ASSERT(reporter, stream_equals(buffer2, 0, expectedResult2,
+                                            buffer2.getOffset()));
 }
 
 #include "TestClassDef.h"