[PDF] Fix bug in ToUnicode table generation for Type 3 fonts.

src/pdf/SkPDFFont.cpp

 /*
  * Copyright 2011 Google Inc.
  *
  * Use of this source code is governed by a BSD-style license that can be
  * found in the LICENSE file.
  */
 
 #include <ctype.h>
 
 #include "SkData.h"
@@ skipping 442 matching lines @@
 // following case because the specification is not clear.
 //
 // 4 beginbfchar          1 beginbfchar
 // <0003> <0013>          <0020> <0014>
 // <0005> <0015>    to    endbfchar
 // <0007> <0017>          1 beginbfrange
 // <0020> <0014>          <0003> <0007> <0013>
 // endbfchar              endbfrange
 //
 // Adobe Technote 5014 said: "Code mappings (unlike codespace ranges) may
-// overlap, but succeeding maps superceded preceding maps."
+// overlap, but succeeding maps supersede preceding maps."
 //
 // In case of searching text in PDF, bfrange will have higher precedence so
 // typing char id 0x0014 in search box will get glyph id 0x0004 first.  However,
 // the spec does not mention how will this kind of conflict being resolved.
 //
 // For the worst case (having 65536 continuous unicode and we use every other
 // one of them), the possible savings by aggressive optimization is 416KB
 // pre-compressed and does not provide enough motivation for implementation.
 
 // FIXME: this should be in a header so that it is separately testable
 // ( see caller in tests/ToUnicode.cpp )
 void append_cmap_sections(const SkTDArray<SkUnichar>& glyphToUnicode,
                           const SkPDFGlyphSet* subset,
                           SkDynamicMemoryWStream* cmap,
+                          bool multiByteGlyphs,
                           uint16_t firstGlyphID,
                           uint16_t lastGlyphID);
 
 void append_cmap_sections(const SkTDArray<SkUnichar>& glyphToUnicode,
                           const SkPDFGlyphSet* subset,
                           SkDynamicMemoryWStream* cmap,
+                          bool multiByteGlyphs,
                           uint16_t firstGlyphID,
                           uint16_t lastGlyphID) {
     if (glyphToUnicode.isEmpty()) {
         return;
     }
+    int glyphOffset = 0;
+    if (!multiByteGlyphs) {
+        glyphOffset = firstGlyphID - 1;
+    }
 
     SkTDArray<BFChar> bfcharEntries;
     SkTDArray<BFRange> bfrangeEntries;
 
     BFRange currentRangeEntry = {0, 0, 0};
     bool rangeEmpty = true;
-    const int limit = SkMin32(lastGlyphID + 1, glyphToUnicode.count());
+    const int limit =
+            SkMin32(lastGlyphID + 1, glyphToUnicode.count()) - glyphOffset;
 
-    for (int i = firstGlyphID; i < limit + 1; ++i) {
+    for (int i = firstGlyphID - glyphOffset; i < limit + 1; ++i) {
         bool inSubset = i < limit &&
-                        (subset == NULL || subset->has(i));
+                        (subset == NULL || subset->has(i + glyphOffset));
         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
             bool inRange =
                 i == currentRangeEntry.fEnd + 1 &&
                 i >> 8 == currentRangeEntry.fStart >> 8 &&
                 i < limit &&
-                glyphToUnicode[i] == currentRangeEntry.fUnicode + i -
-                                         currentRangeEntry.fStart;
+                glyphToUnicode[i + glyphOffset] ==
+                    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;
                 }
                 rangeEmpty = true;
             }
         }
         if (inSubset) {
             currentRangeEntry.fEnd = i;
             if (rangeEmpty) {
               currentRangeEntry.fStart = i;
-              currentRangeEntry.fUnicode = glyphToUnicode[i];
+              currentRangeEntry.fUnicode = glyphToUnicode[i + glyphOffset];
               rangeEmpty = false;
             }
         }
     }
 
     // The spec requires all bfchar entries for a font must come before bfrange
     // entries.
     append_bfchar_section(bfcharEntries, cmap);
     append_bfrange_section(bfrangeEntries, cmap);
 }
 
 static SkPDFStream* generate_tounicode_cmap(
         const SkTDArray<SkUnichar>& glyphToUnicode,
         const SkPDFGlyphSet* subset,
+        bool multiByteGlyphs,
         uint16_t firstGlyphID,
         uint16_t lastGlyphID) {
     SkDynamicMemoryWStream cmap;
-    append_tounicode_header(&cmap, firstGlyphID, lastGlyphID);
-    append_cmap_sections(glyphToUnicode, subset, &cmap,
+    if (multiByteGlyphs) {
+        append_tounicode_header(&cmap, firstGlyphID, lastGlyphID);
+    } else {
+        append_tounicode_header(&cmap, 1, lastGlyphID - firstGlyphID + 1);
+    }
+    append_cmap_sections(glyphToUnicode, subset, &cmap, multiByteGlyphs,
                          firstGlyphID, lastGlyphID);
     append_cmap_footer(&cmap);
     SkAutoTUnref<SkMemoryStream> cmapStream(new SkMemoryStream());
     cmapStream->setData(cmap.copyToData())->unref();
     return new SkPDFStream(cmapStream.get());
 }
 
 #if defined (SK_SFNTLY_SUBSETTER)
 static void sk_delete_array(const void* ptr, size_t, void*) {
     // Use C-style cast to cast away const and cast type simultaneously.
@@ skipping 482 matching lines @@
       fFontID(fontID),
       fGlyphID(glyphID) {
 }
 
 void SkPDFFont::populateToUnicodeTable(const SkPDFGlyphSet* subset) {
     if (fFontInfo == NULL || fFontInfo->fGlyphToUnicode.begin() == NULL) {
         return;
     }
     SkAutoTUnref<SkPDFStream> pdfCmap(
         generate_tounicode_cmap(fFontInfo->fGlyphToUnicode, subset,
-                                firstGlyphID(), lastGlyphID()));
+                                multiByteGlyphs(), firstGlyphID(),
+                                lastGlyphID()));
     addResource(pdfCmap.get());
     insert("ToUnicode", new SkPDFObjRef(pdfCmap.get()))->unref();
 }
 
 ///////////////////////////////////////////////////////////////////////////////
 // class SkPDFType0Font
 ///////////////////////////////////////////////////////////////////////////////
 
 SkPDFType0Font::SkPDFType0Font(SkAdvancedTypefaceMetrics* info,
                                SkTypeface* typeface)
@@ skipping 394 matching lines @@
         glyphStream->setData(content.copyToData())->unref();
 
         SkAutoTUnref<SkPDFStream> glyphDescription(
             new SkPDFStream(glyphStream.get()));
         addResource(glyphDescription.get());
         charProcs->insert(characterName.c_str(),
                           new SkPDFObjRef(glyphDescription.get()))->unref();
     }
 
     insert("FontBBox", makeFontBBox(bbox, 1000))->unref();
-    insertInt("FirstChar", firstGlyphID());
-    insertInt("LastChar", lastGlyphID());
+    insertInt("FirstChar", 1);
+    insertInt("LastChar", lastGlyphID() - firstGlyphID() + 1);
     insert("Widths", widthArray.get());
     insertName("CIDToGIDMap", "Identity");
 
     populateToUnicodeTable(NULL);
     return true;
 }