Merge remote-tracking branch 'origin/master' into custom

application/palemoon/components/nsBrowserGlue.js

@@ -1198,7 +1198,7 @@ BrowserGlue.prototype = {
   },
 
   _migrateUI: function BG__migrateUI() {
-    const UI_VERSION = 19;
+    const UI_VERSION = 20;
     const BROWSER_DOCURL = "chrome://browser/content/browser.xul#";
     let currentUIVersion = 0;
     try {
@@ -1433,6 +1433,11 @@ BrowserGlue.prototype = {
     }
 #endif
 
+    if (currentUIVersion < 20) {
+      // HPKP change of UI preference; reset enforcement level
+      Services.prefs.clearUserPref("security.cert_pinning.enforcement_level");
+    }
+
     // Update the migration version.
     Services.prefs.setIntPref("browser.migration.version", UI_VERSION);
   },

application/palemoon/components/preferences/security.js

@@ -18,7 +18,6 @@ var gSecurityPane = {
   {
     this._pane = document.getElementById("paneSecurity");
     this._initMasterPasswordUI();
-    this._initHPKPUI();
   },
 
   // ADD-ONS
@@ -233,31 +232,5 @@ var gSecurityPane = {
     document.documentElement.openWindow("Toolkit:PasswordManager",
                                         "chrome://passwordmgr/content/passwordManager.xul",
                                         "", null);
-  },
-  
-  _initHPKPUI: function() {
-    let checkbox = document.getElementById("enableHPKP");
-    let HPKPpref = document.getElementById("security.cert_pinning.enforcement_level");
-
-    if (HPKPpref.value == 0) {
-      checkbox.checked = false;
-    } else {
-      checkbox.checked = true;
-    }
-  },
-      
-  /**
-   * Updates the HPKP enforcement level to the proper value depending on checkbox
-   * state.
-   */
-  updateHPKPPref: function() {
-    let checkbox = document.getElementById("enableHPKP");
-    let HPKPpref = document.getElementById("security.cert_pinning.enforcement_level");
-    
-    if (checkbox.checked) {
-      HPKPpref.value = 2;
-    } else {
-      HPKPpref.value = 0;
-    }
   }
 };

application/palemoon/components/preferences/security.xul

@@ -46,9 +46,9 @@
       <preference id="network.stricttransportsecurity.enabled"
                   name="network.stricttransportsecurity.enabled"
                   type="bool"/>
-      <preference id="security.cert_pinning.enforcement_level"
-                  name="security.cert_pinning.enforcement_level"
-                  type="int"/>
+      <preference id="security.cert_pinning.hpkp.enabled"
+                  name="security.cert_pinning.hpkp.enabled"
+                  type="bool"/>
       
       <!-- Opportunistic Encryption -->
                   
@@ -150,7 +150,7 @@
         <checkbox id="enableHPKP"
                   label="&enableHPKP.label;"
                   accesskey="&enableHPKP.accesskey;"
-                  oncommand="gSecurityPane.updateHPKPPref();"/>
+                  preference="security.cert_pinning.hpkp.enabled"/>
       </vbox>
     </groupbox>
     

devtools/client/webconsole/test/browser_webconsole_hpkp_invalid-headers.js

@@ -14,14 +14,18 @@ const SJS_URL = "https://example.com/browser/devtools/client/webconsole/" +
                 "test/test_hpkp-invalid-headers.sjs";
 const LEARN_MORE_URI = "https://developer.mozilla.org/docs/Web/Security/" +
                        "Public_Key_Pinning" + DOCS_GA_PARAMS;
+const HPKP_ENABLED_PREF = "security.cert_pinning.hpkp.enabled";
 const NON_BUILTIN_ROOT_PREF = "security.cert_pinning.process_headers_from_" +
                               "non_builtin_roots";
 
 add_task(function* () {
   registerCleanupFunction(() => {
+    Services.prefs.clearUserPref(HPKP_ENABLED_PREF);
     Services.prefs.clearUserPref(NON_BUILTIN_ROOT_PREF);
   });
 
+  Services.prefs.setBoolPref(HPKP_ENABLED_PREF, true);
+  
   yield loadTab(TEST_URI);
 
   let hud = yield openConsole();

devtools/shared/webconsole/test/test_network_security-hpkp.html

@@ -17,7 +17,8 @@
 SimpleTest.waitForExplicitFinish();
 
 let gCurrentTestCase = -1;
-const HPKP_PREF = "security.cert_pinning.process_headers_from_non_builtin_roots";
+const HPKP_ENABLED_PREF = "security.cert_pinning.hpkp.enabled";
+const PROCESS_HPKP_FROM_NON_BUILTIN_ROOTS_PREF = "security.cert_pinning.process_headers_from_non_builtin_roots";
 
 // Static pins tested by unit/test_security-info-static-hpkp.js.
 const TEST_CASES = [
@@ -41,11 +42,11 @@ const TEST_CASES = [
 
 function startTest()
 {
-  // Need to enable this pref or pinning headers are rejected due test
-  // certificate.
-  Services.prefs.setBoolPref(HPKP_PREF, true);
+  Services.prefs.setBoolPref(HPKP_ENABLED_PREF, true);
+  Services.prefs.setBoolPref(PROCESS_HPKP_FROM_NON_BUILTIN_ROOTS_PREF, true);
   SimpleTest.registerCleanupFunction(() => {
-    Services.prefs.setBoolPref(HPKP_PREF, false);
+    Services.prefs.setBoolPref(HPKP_ENABLED_PREF, false);
+    Services.prefs.setBoolPref(PROCESS_HPKP_FROM_NON_BUILTIN_ROOTS_PREF, false);
 
     // Reset pinning state.
     let gSSService = Cc["@mozilla.org/ssservice;1"]

gfx/ots/README.mozilla

@@ -2,7 +2,7 @@ This is the Sanitiser for OpenType project, from http://code.google.com/p/ots/.
 
 Our reference repository is https://github.com/khaledhosny/ots/.
 
-Current revision: f87b4556191e4132ef5c47365762eb88ace97fc3 (6.0.0)
+Current revision: 8bba749d9d5401726a7d7609ab914fdb5e92bfbe (8.0.0)
 
 Upstream files included: LICENSE, src/, include/, tests/*.cc
 

gfx/ots/include/opentype-sanitiser.h

@@ -35,13 +35,17 @@ typedef int int32_t;
 typedef unsigned int uint32_t;
 typedef __int64 int64_t;
 typedef unsigned __int64 uint64_t;
-#define ntohl(x) _byteswap_ulong (x)
-#define ntohs(x) _byteswap_ushort (x)
-#define htonl(x) _byteswap_ulong (x)
-#define htons(x) _byteswap_ushort (x)
+#define ots_ntohl(x) _byteswap_ulong (x)
+#define ots_ntohs(x) _byteswap_ushort (x)
+#define ots_htonl(x) _byteswap_ulong (x)
+#define ots_htons(x) _byteswap_ushort (x)
 #else
 #include <arpa/inet.h>
 #include <stdint.h>
+#define ots_ntohl(x) ntohl (x)
+#define ots_ntohs(x) ntohs (x)
+#define ots_htonl(x) htonl (x)
+#define ots_htons(x) htons (x)
 #endif
 
 #include <sys/types.h>
@@ -80,7 +84,7 @@ class OTSStream {
       const size_t l = std::min(length, static_cast<size_t>(4) - chksum_offset);
       uint32_t tmp = 0;
       std::memcpy(reinterpret_cast<uint8_t *>(&tmp) + chksum_offset, data, l);
-      chksum_ += ntohl(tmp);
+      chksum_ += ots_ntohl(tmp);
       length -= l;
       offset += l;
     }
@@ -89,7 +93,7 @@ class OTSStream {
       uint32_t tmp;
       std::memcpy(&tmp, reinterpret_cast<const uint8_t *>(data) + offset,
         sizeof(uint32_t));
-      chksum_ += ntohl(tmp);
+      chksum_ += ots_ntohl(tmp);
       length -= 4;
       offset += 4;
     }
@@ -99,7 +103,7 @@ class OTSStream {
       uint32_t tmp = 0;
       std::memcpy(&tmp,
                   reinterpret_cast<const uint8_t*>(data) + offset, length);
-      chksum_ += ntohl(tmp);
+      chksum_ += ots_ntohl(tmp);
     }
 
     return WriteRaw(data, orig_length);
@@ -127,27 +131,27 @@ class OTSStream {
   }
 
   bool WriteU16(uint16_t v) {
-    v = htons(v);
+    v = ots_htons(v);
     return Write(&v, sizeof(v));
   }
 
   bool WriteS16(int16_t v) {
-    v = htons(v);
+    v = ots_htons(v);
     return Write(&v, sizeof(v));
   }
 
   bool WriteU24(uint32_t v) {
-    v = htonl(v);
+    v = ots_htonl(v);
     return Write(reinterpret_cast<uint8_t*>(&v)+1, 3);
   }
 
   bool WriteU32(uint32_t v) {
-    v = htonl(v);
+    v = ots_htonl(v);
     return Write(&v, sizeof(v));
   }
 
   bool WriteS32(int32_t v) {
-    v = htonl(v);
+    v = ots_htonl(v);
     return Write(&v, sizeof(v));
   }
 

gfx/ots/ots-lz4.patch

@@ -1,7 +1,8 @@
 diff --git a/gfx/ots/src/glat.cc b/gfx/ots/src/glat.cc
 --- a/gfx/ots/src/glat.cc
 +++ b/gfx/ots/src/glat.cc
-@@ -5,7 +5,7 @@
+@@ -4,9 +4,9 @@
+ 
  #include "glat.h"
  
  #include "gloc.h"
@@ -10,10 +11,12 @@ diff --git a/gfx/ots/src/glat.cc b/gfx/ots/src/glat.cc
  #include <list>
  
  namespace ots {
-@@ -201,14 +201,15 @@ bool OpenTypeGLAT_v3::Parse(const uint8_t* data, size_t length,
-         return DropGraphite("Illegal nested compression");
+ 
+@@ -212,16 +212,17 @@ bool OpenTypeGLAT_v3::Parse(const uint8_
+         return DropGraphite("Decompressed size exceeds 30MB: %gMB",
+                             decompressed_size / (1024.0 * 1024.0));
        }
-       std::vector<uint8_t> decompressed(this->compHead & FULL_SIZE);
+       std::vector<uint8_t> decompressed(decompressed_size);
 -      int ret = LZ4_decompress_safe_partial(
 +      size_t outputSize = 0;
 +      bool ret = mozilla::Compression::LZ4::decompressPartial(
@@ -23,7 +26,7 @@ diff --git a/gfx/ots/src/glat.cc b/gfx/ots/src/glat.cc
 +          reinterpret_cast<char*>(decompressed.data()),
            decompressed.size(),  // target output size
 -          decompressed.size());  // output buffer size
--      if (ret != decompressed.size()) {
+-      if (ret < 0 || unsigned(ret) != decompressed.size()) {
 -        return DropGraphite("Decompression failed with error code %d", ret);
 +          &outputSize);  // return output size
 +      if (!ret || outputSize != decompressed.size()) {
@@ -31,10 +34,12 @@ diff --git a/gfx/ots/src/glat.cc b/gfx/ots/src/glat.cc
        }
        return this->Parse(decompressed.data(), decompressed.size(), true);
      }
+     default:
 diff --git a/gfx/ots/src/silf.cc b/gfx/ots/src/silf.cc
 --- a/gfx/ots/src/silf.cc
 +++ b/gfx/ots/src/silf.cc
-@@ -5,7 +5,7 @@
+@@ -4,9 +4,9 @@
+ 
  #include "silf.h"
  
  #include "name.h"
@@ -43,10 +48,12 @@ diff --git a/gfx/ots/src/silf.cc b/gfx/ots/src/silf.cc
  #include <cmath>
  
  namespace ots {
-@@ -39,14 +39,15 @@ bool OpenTypeSILF::Parse(const uint8_t* data, size_t length,
-           return DropGraphite("Illegal nested compression");
+ 
+@@ -50,16 +50,17 @@ bool OpenTypeSILF::Parse(const uint8_t* 
+           return DropGraphite("Decompressed size exceeds 30MB: %gMB",
+                               decompressed_size / (1024.0 * 1024.0));
          }
-         std::vector<uint8_t> decompressed(this->compHead & FULL_SIZE);
+         std::vector<uint8_t> decompressed(decompressed_size);
 -        int ret = LZ4_decompress_safe_partial(
 +        size_t outputSize = 0;
 +        bool ret = mozilla::Compression::LZ4::decompressPartial(
@@ -56,7 +63,7 @@ diff --git a/gfx/ots/src/silf.cc b/gfx/ots/src/silf.cc
 +            reinterpret_cast<char*>(decompressed.data()),
              decompressed.size(),  // target output size
 -            decompressed.size());  // output buffer size
--        if (ret != decompressed.size()) {
+-        if (ret < 0 || unsigned(ret) != decompressed.size()) {
 -          return DropGraphite("Decompression failed with error code %d", ret);
 +            &outputSize);   // return output size
 +        if (!ret || outputSize != decompressed.size()) {
@@ -64,3 +71,4 @@ diff --git a/gfx/ots/src/silf.cc b/gfx/ots/src/silf.cc
          }
          return this->Parse(decompressed.data(), decompressed.size(), true);
        }
+       default:

gfx/ots/ots-visibility.patch

@@ -1,7 +1,8 @@
 diff --git a/gfx/ots/include/opentype-sanitiser.h b/gfx/ots/include/opentype-sanitiser.h
 --- a/gfx/ots/include/opentype-sanitiser.h
 +++ b/gfx/ots/include/opentype-sanitiser.h
-@@ -5,6 +5,26 @@
+@@ -4,8 +4,28 @@
+ 
  #ifndef OPENTYPE_SANITISER_H_
  #define OPENTYPE_SANITISER_H_
  
@@ -28,7 +29,9 @@ diff --git a/gfx/ots/include/opentype-sanitiser.h b/gfx/ots/include/opentype-san
  #if defined(_WIN32)
  #include <stdlib.h>
  typedef signed char int8_t;
-@@ -161,7 +181,7 @@ enum TableAction {
+ typedef unsigned char uint8_t;
+@@ -164,9 +184,9 @@ enum TableAction {
+   TABLE_ACTION_PASSTHRU, // Serialize the table unchanged
    TABLE_ACTION_DROP      // Drop the table
  };
  
@@ -37,3 +40,4 @@ diff --git a/gfx/ots/include/opentype-sanitiser.h b/gfx/ots/include/opentype-san
    public:
      OTSContext() {}
      virtual ~OTSContext() {}
+ 

gfx/ots/src/avar.cc

@@ -0,0 +1,109 @@
+// Copyright (c) 2018 The OTS Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "avar.h"
+
+#include "fvar.h"
+
+namespace ots {
+
+// -----------------------------------------------------------------------------
+// OpenTypeAVAR
+// -----------------------------------------------------------------------------
+
+bool OpenTypeAVAR::Parse(const uint8_t* data, size_t length) {
+  Buffer table(data, length);
+  if (!table.ReadU16(&this->majorVersion) ||
+      !table.ReadU16(&this->minorVersion) ||
+      !table.ReadU16(&this->reserved) ||
+      !table.ReadU16(&this->axisCount)) {
+    return Drop("Failed to read table header");
+  }
+  if (this->majorVersion != 1) {
+    return Drop("Unknown table version");
+  }
+  if (this->minorVersion > 0) {
+    // we only know how to serialize version 1.0
+    Warning("Downgrading minor version to 0");
+    this->minorVersion = 0;
+  }
+  if (this->reserved != 0) {
+    Warning("Expected reserved=0");
+    this->reserved = 0;
+  }
+
+  OpenTypeFVAR* fvar = static_cast<OpenTypeFVAR*>(
+      GetFont()->GetTypedTable(OTS_TAG_FVAR));
+  if (!fvar) {
+    return DropVariations("Required fvar table is missing");
+  }
+  if (axisCount != fvar->AxisCount()) {
+    return Drop("Axis count mismatch");
+  }
+
+  for (size_t i = 0; i < this->axisCount; i++) {
+    this->axisSegmentMaps.emplace_back();
+    uint16_t positionMapCount;
+    if (!table.ReadU16(&positionMapCount)) {
+      return Drop("Failed to read position map count");
+    }
+    int foundRequiredMappings = 0;
+    for (size_t j = 0; j < positionMapCount; j++) {
+      AxisValueMap map;
+      if (!table.ReadS16(&map.fromCoordinate) ||
+          !table.ReadS16(&map.toCoordinate)) {
+        return Drop("Failed to read axis value map");
+      }
+      if (map.fromCoordinate < -0x4000 ||
+          map.fromCoordinate > 0x4000 ||
+          map.toCoordinate < -0x4000 ||
+          map.toCoordinate > 0x4000) {
+        return Drop("Axis value map coordinate out of range");
+      }
+      if (j > 0) {
+        if (map.fromCoordinate <= this->axisSegmentMaps[i].back().fromCoordinate ||
+            map.toCoordinate < this->axisSegmentMaps[i].back().toCoordinate) {
+          return Drop("Axis value map out of order");
+        }
+      }
+      if ((map.fromCoordinate == -0x4000 && map.toCoordinate == -0x4000) ||
+          (map.fromCoordinate == 0 && map.toCoordinate == 0) ||
+          (map.fromCoordinate == 0x4000 && map.toCoordinate == 0x4000)) {
+        ++foundRequiredMappings;
+      }
+      this->axisSegmentMaps[i].push_back(map);
+    }
+    if (positionMapCount > 0 && foundRequiredMappings != 3) {
+      return Drop("A required mapping (for -1, 0 or 1) is missing");
+    }
+  }
+
+  return true;
+}
+
+bool OpenTypeAVAR::Serialize(OTSStream* out) {
+  if (!out->WriteU16(this->majorVersion) ||
+      !out->WriteU16(this->minorVersion) ||
+      !out->WriteU16(this->reserved) ||
+      !out->WriteU16(this->axisCount)) {
+    return Error("Failed to write table");
+  }
+
+  for (size_t i = 0; i < this->axisCount; i++) {
+    const auto& axisValueMap = this->axisSegmentMaps[i];
+    if (!out->WriteU16(axisValueMap.size())) {
+      return Error("Failed to write table");
+    }
+    for (size_t j = 0; j < axisValueMap.size(); j++) {
+      if (!out->WriteS16(axisValueMap[j].fromCoordinate) ||
+          !out->WriteS16(axisValueMap[j].toCoordinate)) {
+        return Error("Failed to write table");
+      }
+    }
+  }
+
+  return true;
+}
+
+}  // namespace ots

gfx/ots/src/avar.h

@@ -0,0 +1,42 @@
+// Copyright (c) 2018 The OTS Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef OTS_AVAR_H_
+#define OTS_AVAR_H_
+
+#include "ots.h"
+
+#include <vector>
+
+namespace ots {
+
+// -----------------------------------------------------------------------------
+// OpenTypeAVAR Interface
+// -----------------------------------------------------------------------------
+
+class OpenTypeAVAR : public Table {
+ public:
+  explicit OpenTypeAVAR(Font* font, uint32_t tag)
+      : Table(font, tag, tag) { }
+
+  bool Parse(const uint8_t* data, size_t length);
+  bool Serialize(OTSStream* out);
+
+ private:
+  uint16_t majorVersion;
+  uint16_t minorVersion;
+  uint16_t reserved;
+  uint16_t axisCount;
+
+  struct AxisValueMap {
+    int16_t fromCoordinate;
+    int16_t toCoordinate;
+  };
+
+  std::vector<std::vector<AxisValueMap>> axisSegmentMaps;
+};
+
+}  // namespace ots
+
+#endif  // OTS_AVAR_H_

gfx/ots/src/cff.h

@@ -11,22 +11,28 @@
 #include <string>
 #include <vector>
 
+#undef major // glibc defines major!
+
 namespace ots {
 
 struct CFFIndex {
   CFFIndex()
       : count(0), off_size(0), offset_to_next(0) {}
-  uint16_t count;
+  uint32_t count;
   uint8_t off_size;
   std::vector<uint32_t> offsets;
   uint32_t offset_to_next;
 };
 
+typedef std::map<uint32_t, uint16_t> CFFFDSelect;
+
 class OpenTypeCFF : public Table {
  public:
   explicit OpenTypeCFF(Font *font, uint32_t tag)
       : Table(font, tag, tag),
+        major(0),
         font_dict_length(0),
+        charstrings_index(NULL),
         local_subrs(NULL),
         m_data(NULL),
         m_length(0) {
@@ -37,21 +43,46 @@ class OpenTypeCFF : public Table {
   bool Parse(const uint8_t *data, size_t length);
   bool Serialize(OTSStream *out);
 
+  // Major version number.
+  uint8_t major;
+
   // Name INDEX. This name is used in name.cc as a postscript font name.
   std::string name;
 
   // The number of fonts the file has.
   size_t font_dict_length;
   // A map from glyph # to font #.
-  std::map<uint16_t, uint8_t> fd_select;
+  CFFFDSelect fd_select;
 
   // A list of char strings.
-  std::vector<CFFIndex *> char_strings_array;
+  CFFIndex* charstrings_index;
   // A list of Local Subrs associated with FDArrays. Can be empty.
   std::vector<CFFIndex *> local_subrs_per_font;
   // A Local Subrs associated with Top DICT. Can be NULL.
   CFFIndex *local_subrs;
 
+  // CFF2 VariationStore regionIndexCount.
+  std::vector<uint16_t> region_index_count;
+
+ protected:
+  bool ValidateFDSelect(uint16_t num_glyphs);
+
+ private:
+  const uint8_t *m_data;
+  size_t m_length;
+};
+
+class OpenTypeCFF2 : public OpenTypeCFF {
+ public:
+  explicit OpenTypeCFF2(Font *font, uint32_t tag)
+      : OpenTypeCFF(font, tag),
+        m_data(NULL),
+        m_length(0) {
+  }
+
+  bool Parse(const uint8_t *data, size_t length);
+  bool Serialize(OTSStream *out);
+
  private:
   const uint8_t *m_data;
   size_t m_length;

gfx/ots/src/cff_charstring.cc

@@ -5,7 +5,7 @@
 // A parser for the Type 2 Charstring Format.
 // http://www.adobe.com/devnet/font/pdfs/5177.Type2.pdf
 
-#include "cff_type2_charstring.h"
+#include "cff_charstring.h"
 
 #include <climits>
 #include <cstdio>
@@ -22,7 +22,6 @@ namespace {
 // Note #5177.
 const int32_t kMaxSubrsCount = 65536;
 const size_t kMaxCharStringLength = 65535;
-const size_t kMaxArgumentStack = 48;
 const size_t kMaxNumberOfStemHints = 96;
 const size_t kMaxSubrNesting = 10;
 
@@ -30,117 +29,130 @@ const size_t kMaxSubrNesting = 10;
 // will fail with the dummy value.
 const int32_t dummy_result = INT_MAX;
 
-bool ExecuteType2CharString(ots::Font *font,
-                            size_t call_depth,
-                            const ots::CFFIndex& global_subrs_index,
-                            const ots::CFFIndex& local_subrs_index,
-                            ots::Buffer *cff_table,
-                            ots::Buffer *char_string,
-                            std::stack<int32_t> *argument_stack,
-                            bool *out_found_endchar,
-                            bool *out_found_width,
-                            size_t *in_out_num_stems);
+bool ExecuteCharString(ots::OpenTypeCFF& cff,
+                       size_t call_depth,
+                       const ots::CFFIndex& global_subrs_index,
+                       const ots::CFFIndex& local_subrs_index,
+                       ots::Buffer *cff_table,
+                       ots::Buffer *char_string,
+                       std::stack<int32_t> *argument_stack,
+                       bool *out_found_endchar,
+                       bool *out_found_width,
+                       size_t *in_out_num_stems,
+                       bool cff2);
+
+bool ArgumentStackOverflows(std::stack<int32_t> *argument_stack, bool cff2) {
+  if ((cff2 && argument_stack->size() > ots::kMaxCFF2ArgumentStack) ||
+      (!cff2 && argument_stack->size() > ots::kMaxCFF1ArgumentStack)) {
+    return true;
+  }
+  return false;
+}
 
 #ifdef DUMP_T2CHARSTRING
 // Converts |op| to a string and returns it.
-const char *Type2CharStringOperatorToString(ots::Type2CharStringOperator op) {
+const char *CharStringOperatorToString(ots::CharStringOperator op) {
   switch (op) {
   case ots::kHStem:
-    return "HStem";
+    return "hstem";
   case ots::kVStem:
-    return "VStem";
+    return "vstem";
   case ots::kVMoveTo:
-    return "VMoveTo";
+    return "vmoveto";
   case ots::kRLineTo:
-    return "RLineTo";
+    return "rlineto";
   case ots::kHLineTo:
-    return "HLineTo";
+    return "hlineto";
   case ots::kVLineTo:
-    return "VLineTo";
+    return "vlineto";
   case ots::kRRCurveTo:
-    return "RRCurveTo";
+    return "rrcurveto";
   case ots::kCallSubr:
-    return "CallSubr";
+    return "callsubr";
   case ots::kReturn:
-    return "Return";
+    return "return";
   case ots::kEndChar:
-    return "EndChar";
+    return "endchar";
+  case ots::kVSIndex:
+    return "vsindex";
+  case ots::kBlend:
+    return "blend";
   case ots::kHStemHm:
-    return "HStemHm";
+    return "hstemhm";
   case ots::kHintMask:
-    return "HintMask";
+    return "hintmask";
   case ots::kCntrMask:
-    return "CntrMask";
+    return "cntrmask";
   case ots::kRMoveTo:
-    return "RMoveTo";
+    return "rmoveto";
   case ots::kHMoveTo:
-    return "HMoveTo";
+    return "hmoveto";
   case ots::kVStemHm:
-    return "VStemHm";
+    return "vstemhm";
   case ots::kRCurveLine:
-    return "RCurveLine";
+    return "rcurveline";
   case ots::kRLineCurve:
-    return "RLineCurve";
+    return "rlinecurve";
   case ots::kVVCurveTo:
     return "VVCurveTo";
   case ots::kHHCurveTo:
-    return "HHCurveTo";
+    return "hhcurveto";
   case ots::kCallGSubr:
-    return "CallGSubr";
+    return "callgsubr";
   case ots::kVHCurveTo:
-    return "VHCurveTo";
+    return "vhcurveto";
   case ots::kHVCurveTo:
     return "HVCurveTo";
   case ots::kDotSection:
-    return "DotSection";
+    return "dotsection";
   case ots::kAnd:
-    return "And";
+    return "and";
   case ots::kOr:
-    return "Or";
+    return "or";
   case ots::kNot:
-    return "Not";
+    return "not";
   case ots::kAbs:
-    return "Abs";
+    return "abs";
   case ots::kAdd:
-    return "Add";
+    return "add";
   case ots::kSub:
-    return "Sub";
+    return "sub";
   case ots::kDiv:
-    return "Div";
+    return "div";
   case ots::kNeg:
-    return "Neg";
+    return "neg";
   case ots::kEq:
-    return "Eq";
+    return "eq";
   case ots::kDrop:
-    return "Drop";
+    return "drop";
   case ots::kPut:
-    return "Put";
+    return "put";
   case ots::kGet:
-    return "Get";
+    return "get";
   case ots::kIfElse:
-    return "IfElse";
+    return "ifelse";
   case ots::kRandom:
-    return "Random";
+    return "random";
   case ots::kMul:
-    return "Mul";
+    return "mul";
   case ots::kSqrt:
-    return "Sqrt";
+    return "sqrt";
   case ots::kDup:
-    return "Dup";
+    return "dup";
   case ots::kExch:
-    return "Exch";
+    return "exch";
   case ots::kIndex:
-    return "Index";
+    return "index";
   case ots::kRoll:
-    return "Roll";
+    return "roll";
   case ots::kHFlex:
-    return "HFlex";
+    return "hflex";
   case ots::kFlex:
-    return "Flex";
+    return "flex";
   case ots::kHFlex1:
-    return "HFlex1";
+    return "hflex1";
   case ots::kFlex1:
-    return "Flex1";
+    return "flex1";
   }
 
   return "UNKNOWN";
@@ -150,9 +162,9 @@ const char *Type2CharStringOperatorToString(ots::Type2CharStringOperator op) {
 // Read one or more bytes from the |char_string| buffer and stores the number
 // read on |out_number|. If the number read is an operator (ex 'vstem'), sets
 // true on |out_is_operator|. Returns true if the function read a number.
-bool ReadNextNumberFromType2CharString(ots::Buffer *char_string,
-                                       int32_t *out_number,
-                                       bool *out_is_operator) {
+bool ReadNextNumberFromCharString(ots::Buffer *char_string,
+                                  int32_t *out_number,
+                                  bool *out_is_operator) {
   uint8_t v = 0;
   if (!char_string->ReadU8(&v)) {
     return OTS_FAILURE();
@@ -174,7 +186,7 @@ bool ReadNextNumberFromType2CharString(ots::Buffer *char_string,
     *out_is_operator = true;
   } else if (v <= 27) {
     // Special handling for v==19 and v==20 are implemented in
-    // ExecuteType2CharStringOperator().
+    // ExecuteCharStringOperator().
     *out_number = v;
     *out_is_operator = true;
   } else if (v == 28) {
@@ -221,23 +233,63 @@ bool ReadNextNumberFromType2CharString(ots::Buffer *char_string,
   return true;
 }
 
+bool ValidCFF2Operator(int32_t op) {
+  switch (op) {
+  case ots::kReturn:
+  case ots::kEndChar:
+  case ots::kAbs:
+  case ots::kAdd:
+  case ots::kSub:
+  case ots::kDiv:
+  case ots::kNeg:
+  case ots::kRandom:
+  case ots::kMul:
+  case ots::kSqrt:
+  case ots::kDrop:
+  case ots::kExch:
+  case ots::kIndex:
+  case ots::kRoll:
+  case ots::kDup:
+  case ots::kPut:
+  case ots::kGet:
+  case ots::kDotSection:
+  case ots::kAnd:
+  case ots::kOr:
+  case ots::kNot:
+  case ots::kEq:
+  case ots::kIfElse:
+    return false;
+  }
+
+  return true;
+}
+
 // Executes |op| and updates |argument_stack|. Returns true if the execution
 // succeeds. If the |op| is kCallSubr or kCallGSubr, the function recursively
-// calls ExecuteType2CharString() function. The arguments other than |op| and
+// calls ExecuteCharString() function. The arguments other than |op| and
 // |argument_stack| are passed for that reason.
-bool ExecuteType2CharStringOperator(ots::Font *font,
-                                    int32_t op,
-                                    size_t call_depth,
-                                    const ots::CFFIndex& global_subrs_index,
-                                    const ots::CFFIndex& local_subrs_index,
-                                    ots::Buffer *cff_table,
-                                    ots::Buffer *char_string,
-                                    std::stack<int32_t> *argument_stack,
-                                    bool *out_found_endchar,
-                                    bool *in_out_found_width,
-                                    size_t *in_out_num_stems) {
+bool ExecuteCharStringOperator(ots::OpenTypeCFF& cff,
+                               int32_t op,
+                               size_t call_depth,
+                               const ots::CFFIndex& global_subrs_index,
+                               const ots::CFFIndex& local_subrs_index,
+                               ots::Buffer *cff_table,
+                               ots::Buffer *char_string,
+                               std::stack<int32_t> *argument_stack,
+                               bool *out_found_endchar,
+                               bool *in_out_found_width,
+                               size_t *in_out_num_stems,
+                               bool *in_out_have_blend,
+                               bool *in_out_have_visindex,
+                               int32_t *in_out_vsindex,
+                               bool cff2) {
+  ots::Font* font = cff.GetFont();
   const size_t stack_size = argument_stack->size();
 
+  if (cff2 && !ValidCFF2Operator(op)) {
+    return OTS_FAILURE();
+  }
+
   switch (op) {
   case ots::kCallSubr:
   case ots::kCallGSubr: {
@@ -290,16 +342,17 @@ bool ExecuteType2CharStringOperator(ots::Font *font,
     }
     ots::Buffer char_string_to_jump(cff_table->buffer() + offset, length);
 
-    return ExecuteType2CharString(font,
-                                  call_depth + 1,
-                                  global_subrs_index,
-                                  local_subrs_index,
-                                  cff_table,
-                                  &char_string_to_jump,
-                                  argument_stack,
-                                  out_found_endchar,
-                                  in_out_found_width,
-                                  in_out_num_stems);
+    return ExecuteCharString(cff,
+                             call_depth + 1,
+                             global_subrs_index,
+                             local_subrs_index,
+                             cff_table,
+                             &char_string_to_jump,
+                             argument_stack,
+                             out_found_endchar,
+                             in_out_found_width,
+                             in_out_num_stems,
+                             cff2);
   }
 
   case ots::kReturn:
@@ -310,6 +363,51 @@ bool ExecuteType2CharStringOperator(ots::Font *font,
     *in_out_found_width = true;  // just in case.
     return true;
 
+  case ots::kVSIndex: {
+    if (!cff2) {
+      return OTS_FAILURE();
+    }
+    if (stack_size != 1) {
+      return OTS_FAILURE();
+    }
+    if (*in_out_have_blend || *in_out_have_visindex) {
+      return OTS_FAILURE();
+    }
+    if (argument_stack->top() >= cff.region_index_count.size()) {
+      return OTS_FAILURE();
+    }
+    *in_out_have_visindex = true;
+    *in_out_vsindex = argument_stack->top();
+    while (!argument_stack->empty())
+      argument_stack->pop();
+    return true;
+  }
+
+  case ots::kBlend: {
+    if (!cff2) {
+      return OTS_FAILURE();
+    }
+    if (stack_size < 1) {
+      return OTS_FAILURE();
+    }
+    if (*in_out_vsindex >= cff.region_index_count.size()) {
+      return OTS_FAILURE();
+    }
+    uint16_t k = cff.region_index_count.at(*in_out_vsindex);
+    uint16_t n = argument_stack->top();
+    if (stack_size < n * (k + 1) + 1) {
+      return OTS_FAILURE();
+    }
+
+    // Keep the 1st n operands on the stack for the next operator to use and
+    // pop the rest. There can be multiple consecutive blend operator, so this
+    // makes sure the operands of all of them are kept on the stack.
+    while (argument_stack->size() > stack_size - ((n * k) + 1))
+      argument_stack->pop();
+    *in_out_have_blend = true;
+    return true;
+  }
+
   case ots::kHStem:
   case ots::kVStem:
   case ots::kHStemHm:
@@ -649,7 +747,7 @@ bool ExecuteType2CharStringOperator(ots::Font *font,
     argument_stack->pop();
     argument_stack->push(dummy_result);
     argument_stack->push(dummy_result);
-    if (argument_stack->size() > kMaxArgumentStack) {
+    if (ArgumentStackOverflows(argument_stack, cff2)) {
       return OTS_FAILURE();
     }
     // TODO(yusukes): Implement this. We should push a real value for all
@@ -729,26 +827,29 @@ bool ExecuteType2CharStringOperator(ots::Font *font,
 // in_out_found_width: true is set if |char_string| contains 'width' byte (which
 //                     is 0 or 1 byte.)
 // in_out_num_stems: total number of hstems and vstems processed so far.
-bool ExecuteType2CharString(ots::Font *font,
-                            size_t call_depth,
-                            const ots::CFFIndex& global_subrs_index,
-                            const ots::CFFIndex& local_subrs_index,
-                            ots::Buffer *cff_table,
-                            ots::Buffer *char_string,
-                            std::stack<int32_t> *argument_stack,
-                            bool *out_found_endchar,
-                            bool *in_out_found_width,
-                            size_t *in_out_num_stems) {
+bool ExecuteCharString(ots::OpenTypeCFF& cff,
+                       size_t call_depth,
+                       const ots::CFFIndex& global_subrs_index,
+                       const ots::CFFIndex& local_subrs_index,
+                       ots::Buffer *cff_table,
+                       ots::Buffer *char_string,
+                       std::stack<int32_t> *argument_stack,
+                       bool *out_found_endchar,
+                       bool *in_out_found_width,
+                       size_t *in_out_num_stems,
+                       bool cff2) {
   if (call_depth > kMaxSubrNesting) {
     return OTS_FAILURE();
   }
   *out_found_endchar = false;
 
+  bool in_out_have_blend = false, in_out_have_visindex = false;
+  int32_t in_out_vsindex = 0;
   const size_t length = char_string->length();
   while (char_string->offset() < length) {
     int32_t operator_or_operand = 0;
     bool is_operator = false;
-    if (!ReadNextNumberFromType2CharString(char_string,
+    if (!ReadNextNumberFromCharString(char_string,
                                            &operator_or_operand,
                                            &is_operator)) {
       return OTS_FAILURE();
@@ -761,35 +862,39 @@ bool ExecuteType2CharString(ots::Font *font,
     */
 
       if (!is_operator) {
-        std::fprintf(stderr, "#%d# ", operator_or_operand);
+        std::fprintf(stderr, "%d ", operator_or_operand);
       } else {
-        std::fprintf(stderr, "#%s#\n",
-           Type2CharStringOperatorToString(
-               ots::Type2CharStringOperator(operator_or_operand))
+        std::fprintf(stderr, "%s\n",
+           CharStringOperatorToString(
+               ots::CharStringOperator(operator_or_operand))
            );
       }
 #endif
 
     if (!is_operator) {
       argument_stack->push(operator_or_operand);
-      if (argument_stack->size() > kMaxArgumentStack) {
+      if (ArgumentStackOverflows(argument_stack, cff2)) {
         return OTS_FAILURE();
       }
       continue;
     }
 
     // An operator is found. Execute it.
-    if (!ExecuteType2CharStringOperator(font,
-                                        operator_or_operand,
-                                        call_depth,
-                                        global_subrs_index,
-                                        local_subrs_index,
-                                        cff_table,
-                                        char_string,
-                                        argument_stack,
-                                        out_found_endchar,
-                                        in_out_found_width,
-                                        in_out_num_stems)) {
+    if (!ExecuteCharStringOperator(cff,
+                                   operator_or_operand,
+                                   call_depth,
+                                   global_subrs_index,
+                                   local_subrs_index,
+                                   cff_table,
+                                   char_string,
+                                   argument_stack,
+                                   out_found_endchar,
+                                   in_out_found_width,
+                                   in_out_num_stems,
+                                   &in_out_have_blend,
+                                   &in_out_have_visindex,
+                                   &in_out_vsindex,
+                                   cff2)) {
       return OTS_FAILURE();
     }
     if (*out_found_endchar) {
@@ -801,37 +906,39 @@ bool ExecuteType2CharString(ots::Font *font,
   }
 
   // No endchar operator is found.
+  if (cff2)
+    return true;
   return OTS_FAILURE();
 }
 
 // Selects a set of subroutings for |glyph_index| from |cff| and sets it on
 // |out_local_subrs_to_use|. Returns true on success.
-bool SelectLocalSubr(const std::map<uint16_t, uint8_t> &fd_select,
-                     const std::vector<ots::CFFIndex *> &local_subrs_per_font,
-                     const ots::CFFIndex *local_subrs,
+bool SelectLocalSubr(const ots::OpenTypeCFF& cff,
                      uint16_t glyph_index,  // 0-origin
                      const ots::CFFIndex **out_local_subrs_to_use) {
+  bool cff2 = (cff.major == 2);
   *out_local_subrs_to_use = NULL;
 
   // First, find local subrs from |local_subrs_per_font|.
-  if ((fd_select.size() > 0) &&
-      (!local_subrs_per_font.empty())) {
+  if ((cff.fd_select.size() > 0) &&
+      (!cff.local_subrs_per_font.empty())) {
     // Look up FDArray index for the glyph.
-    std::map<uint16_t, uint8_t>::const_iterator iter =
-        fd_select.find(glyph_index);
-    if (iter == fd_select.end()) {
+    const auto& iter = cff.fd_select.find(glyph_index);
+    if (iter == cff.fd_select.end()) {
       return OTS_FAILURE();
     }
-    const uint8_t fd_index = iter->second;
-    if (fd_index >= local_subrs_per_font.size()) {
+    const auto fd_index = iter->second;
+    if (fd_index >= cff.local_subrs_per_font.size()) {
       return OTS_FAILURE();
     }
-    *out_local_subrs_to_use = local_subrs_per_font.at(fd_index);
-  } else if (local_subrs) {
+    *out_local_subrs_to_use = cff.local_subrs_per_font.at(fd_index);
+  } else if (cff.local_subrs) {
     // Second, try to use |local_subrs|. Most Latin fonts don't have FDSelect
     // entries. If The font has a local subrs index associated with the Top
     // DICT (not FDArrays), use it.
-    *out_local_subrs_to_use = local_subrs;
+    *out_local_subrs_to_use = cff.local_subrs;
+  } else if (cff2 && cff.local_subrs_per_font.size() == 1) {
+    *out_local_subrs_to_use = cff.local_subrs_per_font.at(0);
   } else {
     // Just return NULL.
     *out_local_subrs_to_use = NULL;
@@ -844,18 +951,16 @@ bool SelectLocalSubr(const std::map<uint16_t, uint8_t> &fd_select,
 
 namespace ots {
 
-bool ValidateType2CharStringIndex(
-    ots::Font *font,
-    const CFFIndex& char_strings_index,
+bool ValidateCFFCharStrings(
+    ots::OpenTypeCFF& cff,
     const CFFIndex& global_subrs_index,
-    const std::map<uint16_t, uint8_t> &fd_select,
-    const std::vector<CFFIndex *> &local_subrs_per_font,
-    const CFFIndex *local_subrs,
     Buffer* cff_table) {
+  const CFFIndex& char_strings_index = *(cff.charstrings_index);
   if (char_strings_index.offsets.size() == 0) {
     return OTS_FAILURE();  // no charstring.
   }
 
+  bool cff2 = (cff.major == 2);
   // For each glyph, validate the corresponding charstring.
   for (unsigned i = 1; i < char_strings_index.offsets.size(); ++i) {
     // Prepare a Buffer object, |char_string|, which contains the charstring
@@ -875,9 +980,7 @@ bool ValidateType2CharStringIndex(
     // Get a local subrs for the glyph.
     const unsigned glyph_index = i - 1;  // index in the map is 0-origin.
     const CFFIndex *local_subrs_to_use = NULL;
-    if (!SelectLocalSubr(fd_select,
-                         local_subrs_per_font,
-                         local_subrs,
+    if (!SelectLocalSubr(cff,
                          glyph_index,
                          &local_subrs_to_use)) {
       return OTS_FAILURE();
@@ -891,16 +994,19 @@ bool ValidateType2CharStringIndex(
     // Check a charstring for the |i|-th glyph.
     std::stack<int32_t> argument_stack;
     bool found_endchar = false;
-    bool found_width = false;
+    // CFF2 CharString has no value for width, so we start with true here to
+    // error out if width is found.
+    bool found_width = cff2;
     size_t num_stems = 0;
-    if (!ExecuteType2CharString(font,
-                                0 /* initial call_depth is zero */,
-                                global_subrs_index, *local_subrs_to_use,
-                                cff_table, &char_string, &argument_stack,
-                                &found_endchar, &found_width, &num_stems)) {
+    if (!ExecuteCharString(cff,
+                           0 /* initial call_depth is zero */,
+                           global_subrs_index, *local_subrs_to_use,
+                           cff_table, &char_string, &argument_stack,
+                           &found_endchar, &found_width, &num_stems,
+                           cff2)) {
       return OTS_FAILURE();
     }
-    if (!found_endchar) {
+    if (!cff2 && !found_endchar) {
       return OTS_FAILURE();
     }
   }

gfx/ots/src/cff_charstring.h

@@ -13,6 +13,9 @@
 
 namespace ots {
 
+const size_t kMaxCFF1ArgumentStack = 48;
+const size_t kMaxCFF2ArgumentStack = 513;
+
 // Validates all charstrings in |char_strings_index|. Charstring is a small
 // language for font hinting defined in Adobe Technical Note #5177.
 // http://www.adobe.com/devnet/font/pdfs/5177.Type2.pdf
@@ -34,17 +37,14 @@ namespace ots {
 //  local_subrs: A Local Subrs associated with Top DICT. Can be NULL.
 //  cff_table: A buffer which contains actual byte code of charstring, global
 //             subroutines and local subroutines.
-bool ValidateType2CharStringIndex(
-    Font *font,
-    const CFFIndex &char_strings_index,
+bool ValidateCFFCharStrings(
+    OpenTypeCFF& cff,
     const CFFIndex &global_subrs_index,
-    const std::map<uint16_t, uint8_t> &fd_select,
-    const std::vector<CFFIndex *> &local_subrs_per_font,
-    const CFFIndex *local_subrs,
     Buffer *cff_table);
 
 // The list of Operators. See Appendix. A in Adobe Technical Note #5177.
-enum Type2CharStringOperator {
+// and https://docs.microsoft.com/en-us/typography/opentype/spec/cff2charstr
+enum CharStringOperator {
   kHStem = 1,
   kVStem = 3,
   kVMoveTo = 4,
@@ -55,6 +55,8 @@ enum Type2CharStringOperator {
   kCallSubr = 10,
   kReturn = 11,
   kEndChar = 14,
+  kVSIndex = 15,
+  kBlend = 16,
   kHStemHm = 18,
   kHintMask = 19,
   kCntrMask = 20,

gfx/ots/src/cmap.cc

@@ -238,7 +238,7 @@ bool OpenTypeCMAP::ParseFormat4(int platform, int encoding,
         }
         uint16_t glyph;
         std::memcpy(&glyph, data + glyph_id_offset, 2);
-        glyph = ntohs(glyph);
+        glyph = ots_ntohs(glyph);
         if (glyph >= num_glyphs) {
           return Error("Range glyph reference too high (%d > %d)", glyph, num_glyphs - 1);
         }
@@ -771,9 +771,10 @@ bool OpenTypeCMAP::Parse(const uint8_t *data, size_t length) {
                       subtable_headers[i].length, num_glyphs)) {
           return Error("Failed to parse format 4 cmap subtable %d", i);
         }
-      } else if ((subtable_headers[i].encoding == 3) &&
+      } else if ((subtable_headers[i].encoding == 3 ||
+                  subtable_headers[i].encoding == 4) &&
                  (subtable_headers[i].format == 12)) {
-        // parse and output the 0-3-12 table as 3-10-12 table.
+        // parse and output the 0-3-12 or 0-4-12 tables as 3-10-12 table.
         if (!Parse31012(data + subtable_headers[i].offset,
                         subtable_headers[i].length, num_glyphs)) {
           return Error("Failed to parse format 12 cmap subtable %d", i);

gfx/ots/src/cvar.cc

@@ -0,0 +1,56 @@
+// Copyright (c) 2018 The OTS Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "cvar.h"
+
+#include "fvar.h"
+#include "variations.h"
+
+namespace ots {
+
+// -----------------------------------------------------------------------------
+// OpenTypeCVAR
+// -----------------------------------------------------------------------------
+
+bool OpenTypeCVAR::Parse(const uint8_t* data, size_t length) {
+  Buffer table(data, length);
+
+  uint16_t majorVersion;
+  uint16_t minorVersion;
+
+  if (!table.ReadU16(&majorVersion) ||
+      !table.ReadU16(&minorVersion)) {
+    return Drop("Failed to read table header");
+  }
+
+  if (majorVersion != 1) {
+    return Drop("Unknown table version");
+  }
+
+  OpenTypeFVAR* fvar = static_cast<OpenTypeFVAR*>(
+      GetFont()->GetTypedTable(OTS_TAG_FVAR));
+  if (!fvar) {
+    return DropVariations("Required fvar table is missing");
+  }
+
+  if (!ParseVariationData(GetFont(), data + table.offset(), length - table.offset(),
+                          fvar->AxisCount(), 0)) {
+    return Drop("Failed to parse variation data");
+  }
+
+  this->m_data = data;
+  this->m_length = length;
+
+  return true;
+}
+
+bool OpenTypeCVAR::Serialize(OTSStream* out) {
+  if (!out->Write(this->m_data, this->m_length)) {
+    return Error("Failed to write cvar table");
+  }
+
+  return true;
+}
+
+}  // namespace ots

gfx/ots/src/cvar.h

@@ -0,0 +1,31 @@
+// Copyright (c) 2018 The OTS Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef OTS_CVAR_H_
+#define OTS_CVAR_H_
+
+#include "ots.h"
+
+namespace ots {
+
+// -----------------------------------------------------------------------------
+// OpenTypeCVAR Interface
+// -----------------------------------------------------------------------------
+
+class OpenTypeCVAR : public Table {
+ public:
+  explicit OpenTypeCVAR(Font* font, uint32_t tag)
+      : Table(font, tag, tag) { }
+
+  bool Parse(const uint8_t* data, size_t length);
+  bool Serialize(OTSStream* out);
+
+ private:
+  const uint8_t *m_data;
+  size_t m_length;
+};
+
+}  // namespace ots
+
+#endif  // OTS_CVAR_H_

gfx/ots/src/fvar.cc

@@ -0,0 +1,164 @@
+// Copyright (c) 2018 The OTS Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "fvar.h"
+
+namespace ots {
+
+// -----------------------------------------------------------------------------
+// OpenTypeFVAR
+// -----------------------------------------------------------------------------
+
+bool OpenTypeFVAR::Parse(const uint8_t* data, size_t length) {
+  Buffer table(data, length);
+  if (!table.ReadU16(&this->majorVersion) ||
+      !table.ReadU16(&this->minorVersion) ||
+      !table.ReadU16(&this->axesArrayOffset) ||
+      !table.ReadU16(&this->reserved) ||
+      !table.ReadU16(&this->axisCount) ||
+      !table.ReadU16(&this->axisSize) ||
+      !table.ReadU16(&this->instanceCount) ||
+      !table.ReadU16(&this->instanceSize)) {
+    return DropVariations("Failed to read table header");
+  }
+  if (this->majorVersion != 1) {
+    return DropVariations("Unknown table version");
+  }
+  if (this->minorVersion > 0) {
+    Warning("Downgrading minor version to 0");
+    this->minorVersion = 0;
+  }
+  if (this->axesArrayOffset > length || this->axesArrayOffset < table.offset()) {
+    return DropVariations("Bad axesArrayOffset");
+  }
+  if (this->reserved != 2) {
+    Warning("Expected reserved=2");
+    this->reserved = 2;
+  }
+  if (this->axisCount == 0) {
+    return DropVariations("No variation axes");
+  }
+  if (this->axisSize != 20) {
+    return DropVariations("Invalid axisSize");
+  }
+  // instanceCount is not validated
+  if (this->instanceSize == this->axisCount * sizeof(Fixed) + 6) {
+    this->instancesHavePostScriptNameID = true;
+  } else if (this->instanceSize == this->axisCount * sizeof(Fixed) + 4) {
+    this->instancesHavePostScriptNameID = false;
+  } else {
+    return DropVariations("Invalid instanceSize");
+  }
+
+  // When we serialize, the axes array will go here, even if it was
+  // originally at a different offset. So we update the axesArrayOffset
+  // field for the header.
+  uint32_t origAxesArrayOffset = this->axesArrayOffset;
+  this->axesArrayOffset = table.offset();
+
+  table.set_offset(origAxesArrayOffset);
+  for (unsigned i = 0; i < this->axisCount; i++) {
+    this->axes.emplace_back();
+    auto& axis = this->axes[i];
+    if (!table.ReadU32(&axis.axisTag) ||
+        !table.ReadS32(&axis.minValue) ||
+        !table.ReadS32(&axis.defaultValue) ||
+        !table.ReadS32(&axis.maxValue) ||
+        !table.ReadU16(&axis.flags) ||
+        !table.ReadU16(&axis.axisNameID)) {
+      return DropVariations("Failed to read axis record");
+    }
+    if (!CheckTag(axis.axisTag)) {
+      return DropVariations("Bad axis tag");
+    }
+    if (!(axis.minValue <= axis.defaultValue && axis.defaultValue <= axis.maxValue)) {
+      return DropVariations("Bad axis value range");
+    }
+    if ((axis.flags & 0xFFFEu) != 0) {
+      Warning("Discarding unknown axis flags");
+      axis.flags &= ~0xFFFEu;
+    }
+    if (axis.axisNameID <= 255 || axis.axisNameID >= 32768) {
+      Warning("Axis nameID out of range");
+      // We don't check that the name actually exists -- assume the client can handle
+      // a missing name when it tries to read the table.
+    }
+  }
+
+  for (unsigned i = 0; i < this->instanceCount; i++) {
+    this->instances.emplace_back();
+    auto& inst = this->instances[i];
+    if (!table.ReadU16(&inst.subfamilyNameID) ||
+        !table.ReadU16(&inst.flags)) {
+      return DropVariations("Failed to read instance record");
+    }
+    inst.coordinates.reserve(this->axisCount);
+    for (unsigned j = 0; j < this->axisCount; j++) {
+      inst.coordinates.emplace_back();
+      auto& coord = inst.coordinates[j];
+      if (!table.ReadS32(&coord)) {
+        return DropVariations("Failed to read instance coordinates");
+      }
+    }
+    if (this->instancesHavePostScriptNameID) {
+      if (!table.ReadU16(&inst.postScriptNameID)) {
+        return DropVariations("Failed to read instance psname ID");
+      }
+    }
+  }
+
+  if (table.remaining()) {
+    return Warning("%zu bytes unparsed", table.remaining());
+  }
+
+  return true;
+}
+
+bool OpenTypeFVAR::Serialize(OTSStream* out) {
+  if (!out->WriteU16(this->majorVersion) ||
+      !out->WriteU16(this->minorVersion) ||
+      !out->WriteU16(this->axesArrayOffset) ||
+      !out->WriteU16(this->reserved) ||
+      !out->WriteU16(this->axisCount) ||
+      !out->WriteU16(this->axisSize) ||
+      !out->WriteU16(this->instanceCount) ||
+      !out->WriteU16(this->instanceSize)) {
+    return Error("Failed to write table");
+  }
+
+  for (unsigned i = 0; i < this->axisCount; i++) {
+    const auto& axis = this->axes[i];
+    if (!out->WriteU32(axis.axisTag) ||
+        !out->WriteS32(axis.minValue) ||
+        !out->WriteS32(axis.defaultValue) ||
+        !out->WriteS32(axis.maxValue) ||
+        !out->WriteU16(axis.flags) ||
+        !out->WriteU16(axis.axisNameID)) {
+      return Error("Failed to write table");
+    }
+  }
+
+  for (unsigned i = 0; i < this->instanceCount; i++) {
+    const auto& inst = this->instances[i];
+    if (!out->WriteU16(inst.subfamilyNameID) ||
+        !out->WriteU16(inst.flags)) {
+      return Error("Failed to write table");
+    }
+    for (unsigned j = 0; j < this->axisCount; j++) {
+      const auto& coord = inst.coordinates[j];
+      if (!out->WriteS32(coord)) {
+        return Error("Failed to write table");
+      }
+    }
+    if (this->instancesHavePostScriptNameID) {
+      if (!out->WriteU16(inst.postScriptNameID)) {
+        return Error("Failed to write table");
+      }
+    }
+  }
+
+  return true;
+}
+
+}  // namespace ots

gfx/ots/src/fvar.h

@@ -0,0 +1,63 @@
+// Copyright (c) 2018 The OTS Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef OTS_FVAR_H_
+#define OTS_FVAR_H_
+
+#include <vector>
+
+#include "ots.h"
+
+namespace ots {
+
+// -----------------------------------------------------------------------------
+// OpenTypeFVAR Interface
+// -----------------------------------------------------------------------------
+
+class OpenTypeFVAR : public Table {
+ public:
+  explicit OpenTypeFVAR(Font* font, uint32_t tag)
+      : Table(font, tag, tag) { }
+
+  bool Parse(const uint8_t* data, size_t length);
+  bool Serialize(OTSStream* out);
+
+  uint16_t AxisCount() const { return axisCount; }
+
+ private:
+  uint16_t majorVersion;
+  uint16_t minorVersion;
+  uint16_t axesArrayOffset;
+  uint16_t reserved;
+  uint16_t axisCount;
+  uint16_t axisSize;
+  uint16_t instanceCount;
+  uint16_t instanceSize;
+
+  typedef int32_t Fixed; /* 16.16 fixed-point value */
+
+  struct VariationAxisRecord {
+    uint32_t axisTag;
+    Fixed    minValue;
+    Fixed    defaultValue;
+    Fixed    maxValue;
+    uint16_t flags;
+    uint16_t axisNameID;
+  };
+  std::vector<VariationAxisRecord> axes;
+
+  struct InstanceRecord {
+    uint16_t subfamilyNameID;
+    uint16_t flags;
+    std::vector<Fixed> coordinates;
+    uint16_t postScriptNameID; // optional
+  };
+  std::vector<InstanceRecord> instances;
+
+  bool instancesHavePostScriptNameID;
+};
+
+}  // namespace ots
+
+#endif  // OTS_FVAR_H_

gfx/ots/src/gdef.cc

@@ -11,20 +11,17 @@
 #include "gsub.h"
 #include "layout.h"
 #include "maxp.h"
+#include "variations.h"
 
 // GDEF - The Glyph Definition Table
 // http://www.microsoft.com/typography/otspec/gdef.htm
 
 namespace {
 
-// The maximum class value in class definition tables.
-const uint16_t kMaxClassDefValue = 0xFFFF;
 // The maximum class value in the glyph class definision table.
 const uint16_t kMaxGlyphClassDefValue = 4;
 // The maximum format number of caret value tables.
-// We don't support format 3 for now. See the comment in
-// ParseLigCaretListTable() for the reason.
-const uint16_t kMaxCaretValueFormat = 2;
+const uint16_t kMaxCaretValueFormat = 3;
 
 }  // namespace
 
@@ -165,9 +162,6 @@ bool OpenTypeGDEF::ParseLigCaretListTable(const uint8_t *data, size_t length) {
       if (!subtable.ReadU16(&caret_format)) {
         return Error("Can't read caret values table %d in glyph %d", j, i);
       }
-      // TODO(bashi): We only support caret value format 1 and 2 for now
-      // because there are no fonts which contain caret value format 3
-      // as far as we investigated.
       if (caret_format == 0 || caret_format > kMaxCaretValueFormat) {
         return Error("bad caret value format: %u", caret_format);
       }
@@ -176,6 +170,24 @@ bool OpenTypeGDEF::ParseLigCaretListTable(const uint8_t *data, size_t length) {
       if (!subtable.Skip(2)) {
         return Error("Bad caret value table structure %d in glyph %d", j, i);
       }
+      if (caret_format == 3) {
+        uint16_t offset_device = 0;
+        if (!subtable.ReadU16(&offset_device)) {
+          return Error("Can't read device offset for caret value %d "
+                       "in glyph %d", j, i);
+        }
+        uint16_t absolute_offset = lig_glyphs[i] + caret_value_offsets[j]
+                                   + offset_device;
+        if (offset_device == 0 || absolute_offset >= length) {
+          return Error("Bad device offset for caret value %d in glyph %d: %d",
+                       j, i, offset_device);
+        }
+        if (!ots::ParseDeviceTable(GetFont(), data + absolute_offset,
+                                   length - absolute_offset)) {
+          return Error("Bad device table for caret value %d in glyph %d",
+                       j, i, offset_device);
+        }
+      }
     }
   }
   return true;
@@ -228,18 +240,15 @@ bool OpenTypeGDEF::Parse(const uint8_t *data, size_t length) {
 
   Buffer table(data, length);
 
-  uint32_t version = 0;
-  if (!table.ReadU32(&version)) {
+  uint16_t version_major = 0, version_minor = 0;
+  if (!table.ReadU16(&version_major) ||
+      !table.ReadU16(&version_minor)) {
     return Error("Incomplete table");
   }
-  if (version < 0x00010000 || version == 0x00010001) {
+  if (version_major != 1 || version_minor == 1) { // there is no v1.1
     return Error("Bad version");
   }
 
-  if (version >= 0x00010002) {
-    this->version_2 = true;
-  }
-
   uint16_t offset_glyph_class_def = 0;
   uint16_t offset_attach_list = 0;
   uint16_t offset_lig_caret_list = 0;
@@ -251,15 +260,23 @@ bool OpenTypeGDEF::Parse(const uint8_t *data, size_t length) {
     return Error("Incomplete table");
   }
   uint16_t offset_mark_glyph_sets_def = 0;
-  if (this->version_2) {
+  if (version_minor >= 2) {
     if (!table.ReadU16(&offset_mark_glyph_sets_def)) {
       return Error("Incomplete table");
     }
   }
+  uint32_t item_var_store_offset = 0;
+  if (version_minor >= 3) {
+    if (!table.ReadU32(&item_var_store_offset)) {
+      return Error("Incomplete table");
+    }
+  }
 
   unsigned gdef_header_end = 4 + 4 * 2;
-  if (this->version_2)
+  if (version_minor >= 2)
     gdef_header_end += 2;
+  if (version_minor >= 3)
+    gdef_header_end += 4;
 
   // Parse subtables
   if (offset_glyph_class_def) {
@@ -272,7 +289,6 @@ bool OpenTypeGDEF::Parse(const uint8_t *data, size_t length) {
                                  this->m_num_glyphs, kMaxGlyphClassDefValue)) {
       return Error("Invalid glyph classes");
     }
-    this->has_glyph_class_def = true;
   }
 
   if (offset_attach_list) {
@@ -308,7 +324,6 @@ bool OpenTypeGDEF::Parse(const uint8_t *data, size_t length) {
                                  this->m_num_glyphs, kMaxClassDefValue)) {
       return Error("Invalid mark attachment list");
     }
-    this->has_mark_attachment_class_def = true;
   }
 
   if (offset_mark_glyph_sets_def) {
@@ -320,8 +335,19 @@ bool OpenTypeGDEF::Parse(const uint8_t *data, size_t length) {
                                     length - offset_mark_glyph_sets_def)) {
       return Error("Invalid mark glyph sets");
     }
-    this->has_mark_glyph_sets_def = true;
   }
+
+  if (item_var_store_offset) {
+    if (item_var_store_offset >= length ||
+        item_var_store_offset < gdef_header_end) {
+      return Error("invalid offset to item variation store");
+    }
+    if (!ParseItemVariationStore(GetFont(), data + item_var_store_offset,
+                                 length - item_var_store_offset)) {
+      return Error("Invalid item variation store");
+    }
+  }
+
   this->m_data = data;
   this->m_length = length;
   return true;

gfx/ots/src/gdef.h

@@ -13,10 +13,6 @@ class OpenTypeGDEF : public Table {
  public:
   explicit OpenTypeGDEF(Font *font, uint32_t tag)
       : Table(font, tag, tag),
-        version_2(false),
-        has_glyph_class_def(false),
-        has_mark_attachment_class_def(false),
-        has_mark_glyph_sets_def(false),
         num_mark_glyph_sets(0),
         m_data(NULL),
         m_length(0),
@@ -26,10 +22,6 @@ class OpenTypeGDEF : public Table {
   bool Parse(const uint8_t *data, size_t length);
   bool Serialize(OTSStream *out);
 
-  bool version_2;
-  bool has_glyph_class_def;
-  bool has_mark_attachment_class_def;
-  bool has_mark_glyph_sets_def;
   uint16_t num_mark_glyph_sets;
 
  private:

gfx/ots/src/glat.cc

@@ -200,7 +200,19 @@ bool OpenTypeGLAT_v3::Parse(const uint8_t* data, size_t length,
       if (prevent_decompression) {
         return DropGraphite("Illegal nested compression");
       }
-      std::vector<uint8_t> decompressed(this->compHead & FULL_SIZE);
+      size_t decompressed_size = this->compHead & FULL_SIZE;
+      if (decompressed_size < length) {
+        return DropGraphite("Decompressed size is less than compressed size");
+      }
+      if (decompressed_size == 0) {
+        return DropGraphite("Decompressed size is set to 0");
+      }
+      // decompressed table must be <= 30MB
+      if (decompressed_size > 30 * 1024 * 1024) {
+        return DropGraphite("Decompressed size exceeds 30MB: %gMB",
+                            decompressed_size / (1024.0 * 1024.0));
+      }
+      std::vector<uint8_t> decompressed(decompressed_size);
       size_t outputSize = 0;
       bool ret = mozilla::Compression::LZ4::decompressPartial(
           reinterpret_cast<const char*>(data + table.offset()),
@@ -305,7 +317,7 @@ OctaboxMetrics::ParsePart(Buffer& table) {
 
   unsigned subboxes_len = 0;  // count of 1's in this->subbox_bitmap
   for (uint16_t i = this->subbox_bitmap; i; i >>= 1) {
-    if (i & 1) {
+    if (i & 0b1) {
       ++subboxes_len;
     }
   }

gfx/ots/src/gloc.cc

@@ -25,7 +25,7 @@ bool OpenTypeGLOC::Parse(const uint8_t* data, size_t length) {
   if (this->version >> 16 != 1) {
     return DropGraphite("Unsupported table version: %u", this->version >> 16);
   }
-  if (!table.ReadU16(&this->flags) || this->flags > 3) {
+  if (!table.ReadU16(&this->flags) || this->flags > 0b11) {
     return DropGraphite("Failed to read valid flags");
   }
   if (!table.ReadU16(&this->numAttribs)) {

gfx/ots/src/glyf.cc

@@ -38,6 +38,16 @@ bool OpenTypeGLYF::ParseFlagsForSimpleGlyph(Buffer &glyph,
     delta += 2;
   }
 
+  /* MS and Apple specs say this bit is reserved and must be set to zero, but
+   * Apple spec then contradicts itself and says it should be set on the first
+   * contour flag for simple glyphs with overlapping contours:
+   * https://developer.apple.com/fonts/TrueType-Reference-Manual/RM06/Chap6AATIntro.html
+   * (“Overlapping contours” section) */
+  if (flag & (1u << 6) && *flag_index != 0) {
+    return Error("Bad glyph flag (%d), "
+                 "bit 6 must be set to zero for flag %d", flag, *flag_index);
+  }
+
   if (flag & (1u << 3)) {  // repeat
     if (*flag_index + 1 >= num_flags) {
       return Error("Count too high (%d + 1 >= %d)", *flag_index, num_flags);
@@ -57,8 +67,8 @@ bool OpenTypeGLYF::ParseFlagsForSimpleGlyph(Buffer &glyph,
     }
   }
 
-  if ((flag & (1u << 6)) || (flag & (1u << 7))) {  // reserved flags
-    return Error("Bad glyph flag value (%d), reserved flags must be set to zero", flag);
+  if (flag & (1u << 7)) {  // reserved flag
+    return Error("Bad glyph flag (%d), reserved bit 7 must be set to zero", flag);
   }
 
   *coordinates_length += delta;
@@ -96,8 +106,9 @@ bool OpenTypeGLYF::ParseSimpleGlyph(Buffer &glyph,
 
   if (this->maxp->version_1 &&
       this->maxp->max_size_glyf_instructions < bytecode_length) {
-    return Error("Bytecode length is bigger than maxp.maxSizeOfInstructions "
-        "%d: %d", this->maxp->max_size_glyf_instructions, bytecode_length);
+    this->maxp->max_size_glyf_instructions = bytecode_length;
+    Warning("Bytecode length is bigger than maxp.maxSizeOfInstructions %d: %d",
+            this->maxp->max_size_glyf_instructions, bytecode_length);
   }
 
   if (!glyph.Skip(bytecode_length)) {
@@ -192,9 +203,10 @@ bool OpenTypeGLYF::ParseCompositeGlyph(Buffer &glyph) {
 
     if (this->maxp->version_1 &&
         this->maxp->max_size_glyf_instructions < bytecode_length) {
-      return Error("Bytecode length is bigger than maxp.maxSizeOfInstructions "
-                   "%d: %d",
-                   this->maxp->max_size_glyf_instructions, bytecode_length);
+      this->maxp->max_size_glyf_instructions = bytecode_length;
+      Warning("Bytecode length is bigger than maxp.maxSizeOfInstructions "
+              "%d: %d",
+              this->maxp->max_size_glyf_instructions, bytecode_length);
     }
 
     if (!glyph.Skip(bytecode_length)) {

gfx/ots/src/gpos.cc

@@ -30,12 +30,12 @@ enum GPOS_TYPE {
   GPOS_TYPE_RESERVED = 10
 };
 
-// The size of gpos header.
-const unsigned kGposHeaderSize = 10;
+// The size of gpos header, version 1.0.
+const unsigned kGposHeaderSize_1_0 = 10;
+// The size of gpos header, version 1.1.
+const unsigned kGposHeaderSize_1_1 = 14;
 // The maximum format number for anchor tables.
 const uint16_t kMaxAnchorFormat = 3;
-// The maximum number of class value.
-const uint16_t kMaxClassDefValue = 0xFFFF;
 
 // Lookup type parsers.
 bool ParseSingleAdjustment(const ots::Font *font,
@@ -393,12 +393,12 @@ bool ParsePairPosFormat2(const ots::Font *font,
   // Check class definition tables.
   if (!ots::ParseClassDefTable(font, data + offset_class_def1,
                                length - offset_class_def1,
-                               num_glyphs, kMaxClassDefValue)) {
+                               num_glyphs, ots::kMaxClassDefValue)) {
     return OTS_FAILURE_MSG("Failed to parse class definition table 1");
   }
   if (!ots::ParseClassDefTable(font, data + offset_class_def2,
                                length - offset_class_def2,
-                               num_glyphs, kMaxClassDefValue)) {
+                               num_glyphs, ots::kMaxClassDefValue)) {
     return OTS_FAILURE_MSG("Failed to parse class definition table 2");
   }
 
@@ -749,23 +749,34 @@ bool OpenTypeGPOS::Parse(const uint8_t *data, size_t length) {
   Font *font = GetFont();
   Buffer table(data, length);
 
-  uint32_t version = 0;
+  uint16_t version_major = 0, version_minor = 0;
   uint16_t offset_script_list = 0;
   uint16_t offset_feature_list = 0;
   uint16_t offset_lookup_list = 0;
-  if (!table.ReadU32(&version) ||
+  uint32_t offset_feature_variations = 0;
+  if (!table.ReadU16(&version_major) ||
+      !table.ReadU16(&version_minor) ||
       !table.ReadU16(&offset_script_list) ||
       !table.ReadU16(&offset_feature_list) ||
       !table.ReadU16(&offset_lookup_list)) {
     return Error("Incomplete table");
   }
 
-  if (version != 0x00010000) {
+  if (version_major != 1 || version_minor > 1) {
     return Error("Bad version");
   }
 
+  if (version_minor > 0) {
+    if (!table.ReadU32(&offset_feature_variations)) {
+      return Error("Incomplete table");
+    }
+  }
+
+  const size_t header_size =
+    (version_minor == 0) ? kGposHeaderSize_1_0 : kGposHeaderSize_1_1;
+
   if (offset_lookup_list) {
-    if (offset_lookup_list < kGposHeaderSize || offset_lookup_list >= length) {
+    if (offset_lookup_list < header_size || offset_lookup_list >= length) {
       return Error("Bad lookup list offset in table header");
     }
 
@@ -779,7 +790,7 @@ bool OpenTypeGPOS::Parse(const uint8_t *data, size_t length) {
 
   uint16_t num_features = 0;
   if (offset_feature_list) {
-    if (offset_feature_list < kGposHeaderSize || offset_feature_list >= length) {
+    if (offset_feature_list < header_size || offset_feature_list >= length) {
       return Error("Bad feature list offset in table header");
     }
 
@@ -791,7 +802,7 @@ bool OpenTypeGPOS::Parse(const uint8_t *data, size_t length) {
   }
 
   if (offset_script_list) {
-    if (offset_script_list < kGposHeaderSize || offset_script_list >= length) {
+    if (offset_script_list < header_size || offset_script_list >= length) {
       return Error("Bad script list offset in table header");
     }
 
@@ -801,6 +812,18 @@ bool OpenTypeGPOS::Parse(const uint8_t *data, size_t length) {
     }
   }
 
+  if (offset_feature_variations) {
+    if (offset_feature_variations < header_size || offset_feature_variations >= length) {
+      return Error("Bad feature variations offset in table header");
+    }
+
+    if (!ParseFeatureVariationsTable(font, data + offset_feature_variations,
+                                     length - offset_feature_variations,
+                                     this->num_lookups)) {
+      return Error("Failed to parse feature variations table");
+    }
+  }
+
   this->m_data = data;
   this->m_length = length;
   return true;

gfx/ots/src/graphite.h

@@ -14,6 +14,7 @@ template<typename ParentType>
 class TablePart {
  public:
   TablePart(ParentType* parent) : parent(parent) { }
+  virtual ~TablePart() { }
   virtual bool ParsePart(Buffer& table) = 0;
   virtual bool SerializePart(OTSStream* out) const = 0;
  protected:

gfx/ots/src/gsub.cc

@@ -17,8 +17,10 @@
 
 namespace {
 
-// The GSUB header size
-const size_t kGsubHeaderSize = 4 + 3 * 2;
+// The GSUB header size for table version 1.0
+const size_t kGsubHeaderSize_1_0 = 4 + 3 * 2;
+// GSUB header size v1.1
+const size_t kGsubHeaderSize_1_1 = 4 + 3 * 2 + 4;
 
 enum GSUB_TYPE {
   GSUB_TYPE_SINGLE = 1,
@@ -580,23 +582,34 @@ bool OpenTypeGSUB::Parse(const uint8_t *data, size_t length) {
   Font *font = GetFont();
   Buffer table(data, length);
 
-  uint32_t version = 0;
+  uint16_t version_major = 0, version_minor = 0;
   uint16_t offset_script_list = 0;
   uint16_t offset_feature_list = 0;
   uint16_t offset_lookup_list = 0;
-  if (!table.ReadU32(&version) ||
+  uint32_t offset_feature_variations = 0;
+  if (!table.ReadU16(&version_major) ||
+      !table.ReadU16(&version_minor) ||
       !table.ReadU16(&offset_script_list) ||
       !table.ReadU16(&offset_feature_list) ||
       !table.ReadU16(&offset_lookup_list)) {
     return Error("Incomplete table");
   }
 
-  if (version != 0x00010000) {
+  if (version_major != 1 || version_minor > 1) {
     return Error("Bad version");
   }
 
+  if (version_minor > 0) {
+    if (!table.ReadU32(&offset_feature_variations)) {
+      return Error("Incomplete table");
+    }
+  }
+
+  const size_t header_size =
+    (version_minor == 0) ? kGsubHeaderSize_1_0 : kGsubHeaderSize_1_1;
+
   if (offset_lookup_list) {
-    if (offset_lookup_list < kGsubHeaderSize || offset_lookup_list >= length) {
+    if (offset_lookup_list < header_size || offset_lookup_list >= length) {
       return Error("Bad lookup list offset in table header");
     }
 
@@ -610,7 +623,7 @@ bool OpenTypeGSUB::Parse(const uint8_t *data, size_t length) {
 
   uint16_t num_features = 0;
   if (offset_feature_list) {
-    if (offset_feature_list < kGsubHeaderSize || offset_feature_list >= length) {
+    if (offset_feature_list < header_size || offset_feature_list >= length) {
       return Error("Bad feature list offset in table header");
     }
 
@@ -622,7 +635,7 @@ bool OpenTypeGSUB::Parse(const uint8_t *data, size_t length) {
   }
 
   if (offset_script_list) {
-    if (offset_script_list < kGsubHeaderSize || offset_script_list >= length) {
+    if (offset_script_list < header_size || offset_script_list >= length) {
       return Error("Bad script list offset in table header");
     }
 
@@ -632,6 +645,18 @@ bool OpenTypeGSUB::Parse(const uint8_t *data, size_t length) {
     }
   }
 
+  if (offset_feature_variations) {
+    if (offset_feature_variations < header_size || offset_feature_variations >= length) {
+      return Error("Bad feature variations offset in table header");
+    }
+
+    if (!ParseFeatureVariationsTable(font, data + offset_feature_variations,
+                                     length - offset_feature_variations,
+                                     this->num_lookups)) {
+      return Error("Failed to parse feature variations table");
+    }
+  }
+
   this->m_data = data;
   this->m_length = length;
   return true;

gfx/ots/src/gvar.cc

@@ -0,0 +1,158 @@
+// Copyright (c) 2018 The OTS Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "gvar.h"
+
+#include "fvar.h"
+#include "maxp.h"
+#include "variations.h"
+
+#define TABLE_NAME "gvar"
+
+namespace ots {
+
+// -----------------------------------------------------------------------------
+// OpenTypeGVAR
+// -----------------------------------------------------------------------------
+
+static bool ParseSharedTuples(const Font* font, const uint8_t* data, size_t length,
+                              size_t sharedTupleCount, size_t axisCount) {
+  Buffer subtable(data, length);
+  for (unsigned i = 0; i < sharedTupleCount; i++) {
+    for (unsigned j = 0; j < axisCount; j++) {
+      int16_t coordinate;
+      if (!subtable.ReadS16(&coordinate)) {
+        return OTS_FAILURE_MSG("Failed to read shared tuple coordinate");
+      }
+    }
+  }
+  return true;
+}
+
+static bool ParseGlyphVariationDataArray(const Font* font, const uint8_t* data, size_t length,
+                                         uint16_t flags, size_t glyphCount, size_t axisCount,
+                                         size_t sharedTupleCount,
+                                         const uint8_t* glyphVariationData,
+                                         size_t glyphVariationDataLength) {
+  Buffer subtable(data, length);
+
+  bool glyphVariationDataOffsetsAreLong = (flags & 0x0001u);
+  uint32_t prevOffset = 0;
+  for (size_t i = 0; i < glyphCount + 1; i++) {
+    uint32_t offset;
+    if (glyphVariationDataOffsetsAreLong) {
+      if (!subtable.ReadU32(&offset)) {
+        return OTS_FAILURE_MSG("Failed to read GlyphVariationData offset");
+      }
+    } else {
+      uint16_t halfOffset;
+      if (!subtable.ReadU16(&halfOffset)) {
+        return OTS_FAILURE_MSG("Failed to read GlyphVariationData offset");
+      }
+      offset = halfOffset * 2;
+    }
+
+    if (i > 0 && offset > prevOffset) {
+      if (prevOffset > glyphVariationDataLength) {
+        return OTS_FAILURE_MSG("Invalid GlyphVariationData offset");
+      }
+      if (!ParseVariationData(font, glyphVariationData + prevOffset,
+                              glyphVariationDataLength - prevOffset,
+                              axisCount, sharedTupleCount)) {
+        return OTS_FAILURE_MSG("Failed to parse GlyphVariationData");
+      }
+    }
+    prevOffset = offset;
+  }
+
+  return true;
+}
+
+bool OpenTypeGVAR::Parse(const uint8_t* data, size_t length) {
+  Buffer table(data, length);
+
+  uint16_t majorVersion;
+  uint16_t minorVersion;
+  uint16_t axisCount;
+  uint16_t sharedTupleCount;
+  uint32_t sharedTuplesOffset;
+  uint16_t glyphCount;
+  uint16_t flags;
+  uint32_t glyphVariationDataArrayOffset;
+
+  if (!table.ReadU16(&majorVersion) ||
+      !table.ReadU16(&minorVersion) ||
+      !table.ReadU16(&axisCount) ||
+      !table.ReadU16(&sharedTupleCount) ||
+      !table.ReadU32(&sharedTuplesOffset) ||
+      !table.ReadU16(&glyphCount) ||
+      !table.ReadU16(&flags) ||
+      !table.ReadU32(&glyphVariationDataArrayOffset)) {
+    return DropVariations("Failed to read table header");
+  }
+  if (majorVersion != 1) {
+    return DropVariations("Unknown table version");
+  }
+
+  // check axisCount == fvar->axisCount
+  OpenTypeFVAR* fvar = static_cast<OpenTypeFVAR*>(
+      GetFont()->GetTypedTable(OTS_TAG_FVAR));
+  if (!fvar) {
+    return DropVariations("Required fvar table is missing");
+  }
+  if (axisCount != fvar->AxisCount()) {
+    return DropVariations("Axis count mismatch");
+  }
+
+  // check glyphCount == maxp->num_glyphs
+  OpenTypeMAXP* maxp = static_cast<OpenTypeMAXP*>(
+      GetFont()->GetTypedTable(OTS_TAG_MAXP));
+  if (!maxp) {
+    return DropVariations("Required maxp table is missing");
+  }
+  if (glyphCount != maxp->num_glyphs) {
+    return DropVariations("Glyph count mismatch");
+  }
+
+  if (sharedTupleCount > 0) {
+    if (sharedTuplesOffset < table.offset() || sharedTuplesOffset > length) {
+      return DropVariations("Invalid sharedTuplesOffset");
+    }
+    if (!ParseSharedTuples(GetFont(),
+                           data + sharedTuplesOffset, length - sharedTuplesOffset,
+                           sharedTupleCount, axisCount)) {
+      return DropVariations("Failed to parse shared tuples");
+    }
+  }
+
+  if (glyphVariationDataArrayOffset) {
+    if (glyphVariationDataArrayOffset > length) {
+      return DropVariations("Invalid glyphVariationDataArrayOffset");
+    }
+    if (!ParseGlyphVariationDataArray(GetFont(),
+                                      data + table.offset(), length - table.offset(),
+                                      flags, glyphCount, axisCount, sharedTupleCount,
+                                      data + glyphVariationDataArrayOffset,
+                                      length - glyphVariationDataArrayOffset)) {
+      return DropVariations("Failed to read glyph variation data array");
+    }
+  }
+
+  this->m_data = data;
+  this->m_length = length;
+
+  return true;
+}
+
+bool OpenTypeGVAR::Serialize(OTSStream* out) {
+  if (!out->Write(this->m_data, this->m_length)) {
+    return Error("Failed to write gvar table");
+  }
+
+  return true;
+}
+
+}  // namespace ots
+
+#undef TABLE_NAME

gfx/ots/src/gvar.h

@@ -0,0 +1,31 @@
+// Copyright (c) 2018 The OTS Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef OTS_GVAR_H_
+#define OTS_GVAR_H_
+
+#include "ots.h"
+
+namespace ots {
+
+// -----------------------------------------------------------------------------
+// OpenTypeGVAR Interface
+// -----------------------------------------------------------------------------
+
+class OpenTypeGVAR : public Table {
+ public:
+  explicit OpenTypeGVAR(Font* font, uint32_t tag)
+      : Table(font, tag, tag) { }
+
+  bool Parse(const uint8_t* data, size_t length);
+  bool Serialize(OTSStream* out);
+
+ private:
+  const uint8_t *m_data;
+  size_t m_length;
+};
+
+}  // namespace ots
+
+#endif  // OTS_GVAR_H_

gfx/ots/src/hvar.cc

@@ -0,0 +1,85 @@
+// Copyright (c) 2018 The OTS Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "hvar.h"
+
+#include "variations.h"
+
+namespace ots {
+
+// -----------------------------------------------------------------------------
+// OpenTypeHVAR
+// -----------------------------------------------------------------------------
+
+bool OpenTypeHVAR::Parse(const uint8_t* data, size_t length) {
+  Buffer table(data, length);
+
+  uint16_t majorVersion;
+  uint16_t minorVersion;
+  uint32_t itemVariationStoreOffset;
+  uint32_t advanceWidthMappingOffset;
+  uint32_t lsbMappingOffset;
+  uint32_t rsbMappingOffset;
+
+  if (!table.ReadU16(&majorVersion) ||
+      !table.ReadU16(&minorVersion) ||
+      !table.ReadU32(&itemVariationStoreOffset) ||
+      !table.ReadU32(&advanceWidthMappingOffset) ||
+      !table.ReadU32(&lsbMappingOffset) ||
+      !table.ReadU32(&rsbMappingOffset)) {
+    return DropVariations("Failed to read table header");
+  }
+
+  if (majorVersion != 1) {
+    return DropVariations("Unknown table version");
+  }
+
+  if (itemVariationStoreOffset > length ||
+      advanceWidthMappingOffset > length ||
+      lsbMappingOffset > length ||
+      rsbMappingOffset > length) {
+    return DropVariations("Invalid subtable offset");
+  }
+
+  if (!ParseItemVariationStore(GetFont(), data + itemVariationStoreOffset,
+                               length - itemVariationStoreOffset)) {
+    return DropVariations("Failed to parse item variation store");
+  }
+
+  if (advanceWidthMappingOffset) {
+    if (!ParseDeltaSetIndexMap(GetFont(), data + advanceWidthMappingOffset,
+                               length - advanceWidthMappingOffset)) {
+      return DropVariations("Failed to parse advance width mappings");
+    }
+  }
+
+  if (lsbMappingOffset) {
+    if (!ParseDeltaSetIndexMap(GetFont(), data + lsbMappingOffset,
+                               length - lsbMappingOffset)) {
+      return DropVariations("Failed to parse LSB mappings");
+    }
+  }
+
+  if (rsbMappingOffset) {
+    if (!ParseDeltaSetIndexMap(GetFont(), data + rsbMappingOffset,
+                               length - rsbMappingOffset)) {
+      return DropVariations("Failed to parse RSB mappings");
+    }
+  }
+
+  this->m_data = data;
+  this->m_length = length;
+
+  return true;
+}
+
+bool OpenTypeHVAR::Serialize(OTSStream* out) {
+  if (!out->Write(this->m_data, this->m_length)) {
+    return Error("Failed to write HVAR table");
+  }
+
+  return true;
+}
+
+}  // namespace ots

gfx/ots/src/hvar.h

@@ -0,0 +1,31 @@
+// Copyright (c) 2018 The OTS Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef OTS_HVAR_H_
+#define OTS_HVAR_H_
+
+#include "ots.h"
+
+namespace ots {
+
+// -----------------------------------------------------------------------------
+// OpenTypeHVAR Interface
+// -----------------------------------------------------------------------------
+
+class OpenTypeHVAR : public Table {
+ public:
+  explicit OpenTypeHVAR(Font* font, uint32_t tag)
+      : Table(font, tag, tag) { }
+
+  bool Parse(const uint8_t* data, size_t length);
+  bool Serialize(OTSStream* out);
+
+ private:
+  const uint8_t *m_data;
+  size_t m_length;
+};
+
+}  // namespace ots
+
+#endif  // OTS_HVAR_H_

gfx/ots/src/layout.cc

@@ -7,6 +7,7 @@
 #include <limits>
 #include <vector>
 
+#include "fvar.h"
 #include "gdef.h"
 
 // OpenType Layout Common Table Formats
@@ -22,14 +23,11 @@ const uint32_t kScriptTableTagDflt = 0x44464c54;
 const uint16_t kNoRequiredFeatureIndexDefined = 0xFFFF;
 // The lookup flag bit which indicates existence of MarkFilteringSet.
 const uint16_t kUseMarkFilteringSetBit = 0x0010;
-// The lookup flags which require GDEF table.
-const uint16_t kGdefRequiredFlags = 0x0002 | 0x0004 | 0x0008;
-// The mask for MarkAttachmentType.
-const uint16_t kMarkAttachmentTypeMask = 0xFF00;
 // The maximum type number of format for device tables.
 const uint16_t kMaxDeltaFormatType = 3;
-// The maximum number of class value.
-const uint16_t kMaxClassDefValue = 0xFFFF;
+// In variation fonts, Device Tables are replaced by VariationIndex tables,
+// indicated by this flag in the deltaFormat field.
+const uint16_t kVariationIndex = 0x8000;
 
 struct ScriptRecord {
   uint32_t tag;
@@ -194,30 +192,7 @@ bool ParseLookupTable(ots::Font *font, const uint8_t *data,
     return OTS_FAILURE_MSG("Bad lookup type %d", lookup_type);
   }
 
-  ots::OpenTypeGDEF *gdef = static_cast<ots::OpenTypeGDEF*>(
-      font->GetTypedTable(OTS_TAG_GDEF));
-
-  // Check lookup flags.
-  if ((lookup_flag & kGdefRequiredFlags) &&
-      (!gdef || !gdef->has_glyph_class_def)) {
-    return OTS_FAILURE_MSG("Lookup flags require GDEF table, "
-                           "but none was found: %d", lookup_flag);
-  }
-  if ((lookup_flag & kMarkAttachmentTypeMask) &&
-      (!gdef || !gdef->has_mark_attachment_class_def)) {
-    return OTS_FAILURE_MSG("Lookup flags ask for mark attachment, "
-                           "but there is no GDEF table or it has no "
-                           "mark attachment classes: %d", lookup_flag);
-  }
-  bool use_mark_filtering_set = false;
-  if (lookup_flag & kUseMarkFilteringSetBit) {
-    if (!gdef || !gdef->has_mark_glyph_sets_def) {
-      return OTS_FAILURE_MSG("Lookup flags ask for mark filtering, "
-                             "but there is no GDEF table or it has no "
-                             "mark filtering sets: %d", lookup_flag);
-    }
-    use_mark_filtering_set = true;
-  }
+  bool use_mark_filtering_set = lookup_flag & kUseMarkFilteringSetBit;
 
   std::vector<uint16_t> subtables;
   subtables.reserve(subtable_count);
@@ -248,8 +223,12 @@ bool ParseLookupTable(ots::Font *font, const uint8_t *data,
     if (!subtable.ReadU16(&mark_filtering_set)) {
       return OTS_FAILURE_MSG("Failed to read mark filtering set");
     }
-    if (gdef->num_mark_glyph_sets == 0 ||
-        mark_filtering_set >= gdef->num_mark_glyph_sets) {
+
+    ots::OpenTypeGDEF *gdef = static_cast<ots::OpenTypeGDEF*>(
+        font->GetTypedTable(OTS_TAG_GDEF));
+
+    if (gdef && (gdef->num_mark_glyph_sets == 0 ||
+        mark_filtering_set >= gdef->num_mark_glyph_sets)) {
       return OTS_FAILURE_MSG("Bad mark filtering set %d", mark_filtering_set);
     }
   }
@@ -664,7 +643,7 @@ bool ParseContextFormat2(const ots::Font *font,
   }
   if (!ots::ParseClassDefTable(font, data + offset_class_def,
                                length - offset_class_def,
-                               num_glyphs, kMaxClassDefValue)) {
+                               num_glyphs, ots::kMaxClassDefValue)) {
     return OTS_FAILURE_MSG("Failed to parse class definition table in context format 2");
   }
 
@@ -1017,7 +996,7 @@ bool ParseChainContextFormat2(const ots::Font *font,
     }
     if (!ots::ParseClassDefTable(font, data + offset_backtrack_class_def,
                                  length - offset_backtrack_class_def,
-                                 num_glyphs, kMaxClassDefValue)) {
+                                 num_glyphs, ots::kMaxClassDefValue)) {
       return OTS_FAILURE_MSG("Failed to parse backtrack class defn table in chain context format 2");
     }
   }
@@ -1028,7 +1007,7 @@ bool ParseChainContextFormat2(const ots::Font *font,
   }
   if (!ots::ParseClassDefTable(font, data + offset_input_class_def,
                                length - offset_input_class_def,
-                               num_glyphs, kMaxClassDefValue)) {
+                               num_glyphs, ots::kMaxClassDefValue)) {
     return OTS_FAILURE_MSG("Failed to parse input class defn in chain context format 2");
   }
 
@@ -1039,7 +1018,7 @@ bool ParseChainContextFormat2(const ots::Font *font,
     }
     if (!ots::ParseClassDefTable(font, data + offset_lookahead_class_def,
                                  length - offset_lookahead_class_def,
-                                 num_glyphs, kMaxClassDefValue)) {
+                                 num_glyphs, ots::kMaxClassDefValue)) {
       return OTS_FAILURE_MSG("Failed to parse lookahead class defn in chain context format 2");
     }
   }
@@ -1403,11 +1382,17 @@ bool ParseDeviceTable(const ots::Font *font,
       !subtable.ReadU16(&delta_format)) {
     return OTS_FAILURE_MSG("Failed to read device table header");
   }
+  if (delta_format == kVariationIndex) {
+    // start_size and end_size are replaced by deltaSetOuterIndex
+    // and deltaSetInnerIndex respectively, but we don't attempt to
+    // check them here, so nothing more to do.
+    return true;
+  }
   if (start_size > end_size) {
-    return OTS_FAILURE_MSG("bad size range: %u > %u", start_size, end_size);
+    return OTS_FAILURE_MSG("Bad device table size range: %u > %u", start_size, end_size);
   }
   if (delta_format == 0 || delta_format > kMaxDeltaFormatType) {
-    return OTS_FAILURE_MSG("bad delta format: %u", delta_format);
+    return OTS_FAILURE_MSG("Bad device table delta format: 0x%x", delta_format);
   }
   // The number of delta values per uint16. The device table should contain
   // at least |num_units| * 2 bytes compressed data.
@@ -1519,6 +1504,173 @@ bool ParseExtensionSubtable(const Font *font,
   return true;
 }
 
+bool ParseConditionTable(const Font *font,
+                         const uint8_t *data, const size_t length,
+                         const uint16_t axis_count) {
+  Buffer subtable(data, length);
+
+  uint16_t format = 0;
+  if (!subtable.ReadU16(&format)) {
+    return OTS_FAILURE_MSG("Failed to read condition table format");
+  }
+
+  if (format != 1) {
+    // An unknown format is not an error, but should be ignored per spec.
+    return true;
+  }
+
+  uint16_t axis_index = 0;
+  int16_t filter_range_min_value = 0;
+  int16_t filter_range_max_value = 0;
+  if (!subtable.ReadU16(&axis_index) ||
+      !subtable.ReadS16(&filter_range_min_value) ||
+      !subtable.ReadS16(&filter_range_max_value)) {
+    return OTS_FAILURE_MSG("Failed to read condition table (format 1)");
+  }
+
+  if (axis_index >= axis_count) {
+    return OTS_FAILURE_MSG("Axis index out of range in condition");
+  }
+
+  // Check min/max values are within range -1.0 .. 1.0 and properly ordered
+  if (filter_range_min_value < -0x4000 || // -1.0 in F2DOT14 format
+      filter_range_max_value > 0x4000 || // +1.0 in F2DOT14 format
+      filter_range_min_value > filter_range_max_value) {
+    return OTS_FAILURE_MSG("Invalid filter range in condition");
+  }
+
+  return true;
+}
+
+bool ParseConditionSetTable(const Font *font,
+                            const uint8_t *data, const size_t length,
+                            const uint16_t axis_count) {
+  Buffer subtable(data, length);
+
+  uint16_t condition_count = 0;
+  if (!subtable.ReadU16(&condition_count)) {
+    return OTS_FAILURE_MSG("Failed to read condition count");
+  }
+
+  for (uint16_t i = 0; i < condition_count; i++) {
+    uint32_t condition_offset = 0;
+    if (!subtable.ReadU32(&condition_offset)) {
+      return OTS_FAILURE_MSG("Failed to read condition offset");
+    }
+    if (condition_offset < subtable.offset() || condition_offset >= length) {
+      return OTS_FAILURE_MSG("Offset out of range");
+    }
+    if (!ParseConditionTable(font, data + condition_offset, length - condition_offset,
+                             axis_count)) {
+      return OTS_FAILURE_MSG("Failed to parse condition table");
+    }
+  }
+
+  return true;
+}
+
+bool ParseFeatureTableSubstitutionTable(const Font *font,
+                                        const uint8_t *data, const size_t length,
+                                        const uint16_t num_lookups) {
+  Buffer subtable(data, length);
+
+  uint16_t version_major = 0;
+  uint16_t version_minor = 0;
+  uint16_t substitution_count = 0;
+  const size_t kFeatureTableSubstitutionHeaderSize = 3 * sizeof(uint16_t);
+
+  if (!subtable.ReadU16(&version_major) ||
+      !subtable.ReadU16(&version_minor) ||
+      !subtable.ReadU16(&substitution_count)) {
+    return OTS_FAILURE_MSG("Failed to read feature table substitution table header");
+  }
+
+  for (uint16_t i = 0; i < substitution_count; i++) {
+    uint16_t feature_index = 0;
+    uint32_t alternate_feature_table_offset = 0;
+    const size_t kFeatureTableSubstitutionRecordSize = sizeof(uint16_t) + sizeof(uint32_t);
+
+    if (!subtable.ReadU16(&feature_index) ||
+        !subtable.ReadU32(&alternate_feature_table_offset)) {
+      return OTS_FAILURE_MSG("Failed to read feature table substitution record");
+    }
+
+    if (alternate_feature_table_offset < kFeatureTableSubstitutionHeaderSize +
+                                         kFeatureTableSubstitutionRecordSize * substitution_count ||
+        alternate_feature_table_offset >= length) {
+      return OTS_FAILURE_MSG("Invalid alternate feature table offset");
+    }
+
+    if (!ParseFeatureTable(font, data + alternate_feature_table_offset,
+                           length - alternate_feature_table_offset, num_lookups)) {
+      return OTS_FAILURE_MSG("Failed to parse alternate feature table");
+    }
+  }
+
+  return true;
+}
+
+bool ParseFeatureVariationsTable(const Font *font,
+                                 const uint8_t *data, const size_t length,
+                                 const uint16_t num_lookups) {
+  Buffer subtable(data, length);
+
+  uint16_t version_major = 0;
+  uint16_t version_minor = 0;
+  uint32_t feature_variation_record_count = 0;
+
+  if (!subtable.ReadU16(&version_major) ||
+      !subtable.ReadU16(&version_minor) ||
+      !subtable.ReadU32(&feature_variation_record_count)) {
+    return OTS_FAILURE_MSG("Failed to read feature variations table header");
+  }
+
+  OpenTypeFVAR* fvar = static_cast<OpenTypeFVAR*>(font->GetTypedTable(OTS_TAG_FVAR));
+  if (!fvar) {
+    return OTS_FAILURE_MSG("Not a variation font");
+  }
+  const uint16_t axis_count = fvar->AxisCount();
+
+  const size_t kEndOfFeatureVariationRecords =
+    2 * sizeof(uint16_t) + sizeof(uint32_t) +
+    feature_variation_record_count * 2 * sizeof(uint32_t);
+
+  for (uint32_t i = 0; i < feature_variation_record_count; i++) {
+    uint32_t condition_set_offset = 0;
+    uint32_t feature_table_substitution_offset = 0;
+    if (!subtable.ReadU32(&condition_set_offset) ||
+        !subtable.ReadU32(&feature_table_substitution_offset)) {
+      return OTS_FAILURE_MSG("Failed to read feature variation record");
+    }
+
+    if (condition_set_offset) {
+      if (condition_set_offset < kEndOfFeatureVariationRecords ||
+          condition_set_offset >= length) {
+        return OTS_FAILURE_MSG("Condition set offset out of range");
+      }
+      if (!ParseConditionSetTable(font, data + condition_set_offset,
+                                  length - condition_set_offset,
+                                  axis_count)) {
+        return OTS_FAILURE_MSG("Failed to parse condition set table");
+      }
+    }
+
+    if (feature_table_substitution_offset) {
+      if (feature_table_substitution_offset < kEndOfFeatureVariationRecords ||
+          feature_table_substitution_offset >= length) {
+        return OTS_FAILURE_MSG("Feature table substitution offset out of range");
+      }
+      if (!ParseFeatureTableSubstitutionTable(font, data + feature_table_substitution_offset,
+                                              length - feature_table_substitution_offset,
+                                              num_lookups)) {
+        return OTS_FAILURE_MSG("Failed to parse feature table substitution table");
+      }
+    }
+  }
+
+  return true;
+}
+
 }  // namespace ots
 
 #undef TABLE_NAME

gfx/ots/src/layout.h

@@ -12,6 +12,8 @@
 
 namespace ots {
 
+// The maximum number of class value.
+const uint16_t kMaxClassDefValue = 0xFFFF;
 
 struct LookupSubtableParser {
   struct TypeParser {
@@ -70,6 +72,23 @@ bool ParseExtensionSubtable(const Font *font,
                             const uint8_t *data, const size_t length,
                             const LookupSubtableParser* parser);
 
+// For feature variations table (in GSUB/GPOS v1.1)
+bool ParseConditionTable(const Font *font,
+                         const uint8_t *data, const size_t length,
+                         const uint16_t axis_count);
+
+bool ParseConditionSetTable(const Font *font,
+                            const uint8_t *data, const size_t length,
+                            const uint16_t axis_count);
+
+bool ParseFeatureTableSubstitutionTable(const Font *font,
+                                        const uint8_t *data, const size_t length,
+                                        const uint16_t num_lookups);
+
+bool ParseFeatureVariationsTable(const Font *font,
+                                 const uint8_t *data, const size_t length,
+                                 const uint16_t num_lookups);
+
 }  // namespace ots
 
 #endif  // OTS_LAYOUT_H_

gfx/ots/src/moz.build

@@ -9,28 +9,29 @@ EXPORTS += [
     '../include/ots-memory-stream.h',
 ]
 
-SOURCES += [
-    # needs to be separate because gpos.cc also defines kMaxClassDefValue
-    'gdef.cc',
-]
-
 UNIFIED_SOURCES += [
+    'avar.cc',
     'cff.cc',
-    'cff_type2_charstring.cc',
+    'cff_charstring.cc',
     'cmap.cc',
+    'cvar.cc',
     'cvt.cc',
     'feat.cc',
     'fpgm.cc',
+    'fvar.cc',
     'gasp.cc',
+    'gdef.cc',
     'glat.cc',
     'gloc.cc',
     'glyf.cc',
     'gpos.cc',
     'gsub.cc',
+    'gvar.cc',
     'hdmx.cc',
     'head.cc',
     'hhea.cc',
     'hmtx.cc',
+    'hvar.cc',
     'kern.cc',
     'layout.cc',
     'loca.cc',
@@ -38,6 +39,7 @@ UNIFIED_SOURCES += [
     'math.cc',
     'maxp.cc',
     'metrics.cc',
+    'mvar.cc',
     'name.cc',
     'os2.cc',
     'ots.cc',
@@ -46,10 +48,13 @@ UNIFIED_SOURCES += [
     'sile.cc',
     'silf.cc',
     'sill.cc',
+    'stat.cc',
+    'variations.cc',
     'vdmx.cc',
     'vhea.cc',
     'vmtx.cc',
     'vorg.cc',
+    'vvar.cc',
 ]
 
 # We allow warnings for third-party code that can be updated from upstream.
@@ -60,6 +65,7 @@ FINAL_LIBRARY = 'gkmedias'
 DEFINES['PACKAGE_VERSION'] = '"moz"'
 DEFINES['PACKAGE_BUGREPORT'] = '"http://bugzilla.mozilla.org/"'
 DEFINES['OTS_GRAPHITE'] = 1
+DEFINES['OTS_VARIATIONS'] = 1
 
 USE_LIBS += [
     'brotli',

gfx/ots/src/mvar.cc

@@ -0,0 +1,105 @@
+// Copyright (c) 2018 The OTS Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "mvar.h"
+
+#include "variations.h"
+
+namespace ots {
+
+// -----------------------------------------------------------------------------
+// OpenTypeMVAR
+// -----------------------------------------------------------------------------
+
+bool OpenTypeMVAR::Parse(const uint8_t* data, size_t length) {
+  Buffer table(data, length);
+
+  uint16_t majorVersion;
+  uint16_t minorVersion;
+  uint16_t reserved;
+  uint16_t valueRecordSize;
+  uint16_t valueRecordCount;
+  uint16_t itemVariationStoreOffset;
+
+  if (!table.ReadU16(&majorVersion) ||
+      !table.ReadU16(&minorVersion) ||
+      !table.ReadU16(&reserved) ||
+      !table.ReadU16(&valueRecordSize) ||
+      !table.ReadU16(&valueRecordCount) ||
+      !table.ReadU16(&itemVariationStoreOffset)) {
+    return DropVariations("Failed to read table header");
+  }
+
+  if (majorVersion != 1) {
+    return DropVariations("Unknown table version");
+  }
+
+  if (reserved != 0) {
+    Warning("Expected reserved=0");
+  }
+
+  // The spec says that valueRecordSize "must be greater than zero",
+  // but we don't enforce this in the case where valueRecordCount
+  // is zero.
+  // The minimum size for a valueRecord to be valid is 8, for the
+  // three fields currently defined in the record (see below).
+  if (valueRecordSize < 8) {
+    if (valueRecordCount != 0) {
+      return DropVariations("Value record size too small");
+    }
+  }
+
+  if (valueRecordCount == 0) {
+    if (itemVariationStoreOffset != 0) {
+      // The spec says "if valueRecordCount is zero, set to zero",
+      // but having a variation store even when record count is zero
+      // should be harmless -- it just won't be useful for anything.
+      // But we don't need to reject altogether.
+      Warning("Unexpected item variation store");
+    }
+  } else {
+    if (itemVariationStoreOffset < table.offset() || itemVariationStoreOffset > length) {
+      return DropVariations("Invalid item variation store offset");
+    }
+    if (!ParseItemVariationStore(GetFont(), data + itemVariationStoreOffset,
+                                 length - itemVariationStoreOffset)) {
+      return DropVariations("Failed to parse item variation store");
+    }
+  }
+
+  uint32_t prevTag = 0;
+  size_t offset = table.offset();
+  for (unsigned i = 0; i < valueRecordCount; i++) {
+    uint32_t tag;
+    uint16_t deltaSetOuterIndex, deltaSetInnerIndex;
+    if (!table.ReadU32(&tag) ||
+        !table.ReadU16(&deltaSetOuterIndex) ||
+        !table.ReadU16(&deltaSetInnerIndex)) {
+      return DropVariations("Failed to read value record");
+    }
+    if (tag <= prevTag) {
+      return DropVariations("Invalid or out-of-order value tag");
+    }
+    prevTag = tag;
+    // Adjust offset in case additional fields have been added to the
+    // valueRecord by a new minor version (allowed by spec).
+    offset += valueRecordSize;
+    table.set_offset(offset);
+  }
+
+  this->m_data = data;
+  this->m_length = length;
+
+  return true;
+}
+
+bool OpenTypeMVAR::Serialize(OTSStream* out) {
+  if (!out->Write(this->m_data, this->m_length)) {
+    return Error("Failed to write MVAR table");
+  }
+
+  return true;
+}
+
+}  // namespace ots

gfx/ots/src/mvar.h

@@ -0,0 +1,31 @@
+// Copyright (c) 2018 The OTS Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef OTS_MVAR_H_
+#define OTS_MVAR_H_
+
+#include "ots.h"
+
+namespace ots {
+
+// -----------------------------------------------------------------------------
+// OpenTypeMVAR Interface
+// -----------------------------------------------------------------------------
+
+class OpenTypeMVAR : public Table {
+ public:
+  explicit OpenTypeMVAR(Font* font, uint32_t tag)
+      : Table(font, tag, tag) { }
+
+  bool Parse(const uint8_t* data, size_t length);
+  bool Serialize(OTSStream* out);
+
+ private:
+  const uint8_t *m_data;
+  size_t m_length;
+};
+
+}  // namespace ots
+
+#endif  // OTS_MVAR_H_

gfx/ots/src/name.cc

@@ -169,6 +169,14 @@ bool OpenTypeNAME::Parse(const uint8_t* data, size_t length) {
       if (tag_end > length) {
         return Error("bad end of tag %d > %ld for langTagRecord %d", tag_end, length, i);
       }
+      // Lang tag is BCP 47 tag per the spec, the recommonded BCP 47 max tag
+      // length is 35:
+      // https://tools.ietf.org/html/bcp47#section-4.4.1
+      // We are being too generous and allowing for 100 (multiplied by 2 since
+      // this is UTF-16 string).
+      if (tag_length > 100 * 2) {
+        return Error("Too long language tag for LangTagRecord %d: %d", i, tag_length);
+      }
       std::string tag(string_base + tag_offset, tag_length);
       this->lang_tags.push_back(tag);
     }
@@ -203,17 +211,16 @@ bool OpenTypeNAME::Parse(const uint8_t* data, size_t length) {
   // if not, we'll add our fixed versions here
   bool mac_name[kStdNameCount] = { 0 };
   bool win_name[kStdNameCount] = { 0 };
-  for (std::vector<NameRecord>::iterator name_iter = this->names.begin();
-       name_iter != this->names.end(); ++name_iter) {
-    const uint16_t id = name_iter->name_id;
+  for (const auto& name : this->names) {
+    const uint16_t id = name.name_id;
     if (id >= kStdNameCount || kStdNames[id] == NULL) {
       continue;
     }
-    if (name_iter->platform_id == 1) {
+    if (name.platform_id == 1) {
       mac_name[id] = true;
       continue;
     }
-    if (name_iter->platform_id == 3) {
+    if (name.platform_id == 3) {
       win_name[id] = true;
       continue;
     }
@@ -266,9 +273,7 @@ bool OpenTypeNAME::Serialize(OTSStream* out) {
   }
 
   std::string string_data;
-  for (std::vector<NameRecord>::const_iterator name_iter = this->names.begin();
-       name_iter != this->names.end(); ++name_iter) {
-    const NameRecord& rec = *name_iter;
+  for (const auto& rec : this->names) {
     if (string_data.size() + rec.text.size() >
             std::numeric_limits<uint16_t>::max() ||
         !out->WriteU16(rec.platform_id) ||
@@ -286,16 +291,14 @@ bool OpenTypeNAME::Serialize(OTSStream* out) {
     if (!out->WriteU16(lang_tag_count)) {
       return Error("Faile to write langTagCount");
     }
-    for (std::vector<std::string>::const_iterator tag_iter =
-             this->lang_tags.begin();
-         tag_iter != this->lang_tags.end(); ++tag_iter) {
-      if (string_data.size() + tag_iter->size() >
+    for (const auto& tag : this->lang_tags) {
+      if (string_data.size() + tag.size() >
               std::numeric_limits<uint16_t>::max() ||
-          !out->WriteU16(static_cast<uint16_t>(tag_iter->size())) ||
+          !out->WriteU16(static_cast<uint16_t>(tag.size())) ||
           !out->WriteU16(static_cast<uint16_t>(string_data.size()))) {
         return Error("Failed to write langTagRecord");
       }
-      string_data.append(*tag_iter);
+      string_data.append(tag);
     }
   }
 

gfx/ots/src/os2.cc

@@ -38,17 +38,14 @@ bool OpenTypeOS2::Parse(const uint8_t *data, size_t length) {
     return Error("Unsupported table version: %u", this->table.version);
   }
 
-  // Follow WPF Font Selection Model's advice.
-  if (1 <= this->table.weight_class && this->table.weight_class <= 9) {
-    Warning("Bad usWeightClass: %u, changing it to %u",
-            this->table.weight_class, this->table.weight_class * 100);
-    this->table.weight_class *= 100;
-  }
-  // Ditto.
-  if (this->table.weight_class > 999) {
+  if (this->table.weight_class < 1) {
     Warning("Bad usWeightClass: %u, changing it to %d",
-             this->table.weight_class, 999);
-    this->table.weight_class = 999;
+             this->table.weight_class, 1);
+    this->table.weight_class = 1;
+  } else if (this->table.weight_class > 1000) {
+    Warning("Bad usWeightClass: %u, changing it to %d",
+             this->table.weight_class, 1000);
+    this->table.weight_class = 1000;
   }
 
   if (this->table.width_class < 1) {
@@ -89,7 +86,7 @@ bool OpenTypeOS2::Parse(const uint8_t *data, size_t length) {
   SET_TO_ZERO("yStrikeoutSize", strikeout_size);
 #undef SET_TO_ZERO
 
-  static std::string panose_strings[10] = {
+  static const char* panose_strings[10] = {
     "bFamilyType",
     "bSerifStyle",
     "bWeight",
@@ -103,7 +100,7 @@ bool OpenTypeOS2::Parse(const uint8_t *data, size_t length) {
   };
   for (unsigned i = 0; i < 10; ++i) {
     if (!table.ReadU8(&this->table.panose[i])) {
-      return Error("Failed to read PANOSE %s", panose_strings[i].c_str());
+      return Error("Failed to read PANOSE %s", panose_strings[i]);
     }
   }
 
@@ -155,16 +152,17 @@ bool OpenTypeOS2::Parse(const uint8_t *data, size_t length) {
   if ((this->table.version < 4) &&
       (this->table.selection & 0x300)) {
     // bit 8 and 9 must be unset in OS/2 table versions less than 4.
-    return Error("fSelection bits 8 and 9 must be unset for table version %d",
-                 this->table.version);
+    Warning("fsSelection bits 8 and 9 must be unset for table version %d",
+            this->table.version);
   }
 
   // mask reserved bits. use only 0..9 bits.
   this->table.selection &= 0x3ff;
 
   if (this->table.first_char_index > this->table.last_char_index) {
-    return Error("usFirstCharIndex %d > usLastCharIndex %d",
-                 this->table.first_char_index, this->table.last_char_index);
+    Warning("usFirstCharIndex %d > usLastCharIndex %d",
+            this->table.first_char_index, this->table.last_char_index);
+    this->table.first_char_index = this->table.last_char_index;
   }
   if (this->table.typo_linegap < 0) {
     Warning("Bad sTypoLineGap, setting it to 0: %d", this->table.typo_linegap);

gfx/ots/src/ots.cc

@@ -14,38 +14,46 @@
 #include <map>
 #include <vector>
 
-#include "woff2_dec.h"
+#include <woff2/decode.h>
 
 // The OpenType Font File
-// http://www.microsoft.com/typography/otspec/cmap.htm
+// http://www.microsoft.com/typography/otspec/otff.htm
 
+#include "avar.h"
 #include "cff.h"
 #include "cmap.h"
+#include "cvar.h"
 #include "cvt.h"
 #include "fpgm.h"
+#include "fvar.h"
 #include "gasp.h"
 #include "gdef.h"
 #include "glyf.h"
 #include "gpos.h"
 #include "gsub.h"
+#include "gvar.h"
 #include "hdmx.h"
 #include "head.h"
 #include "hhea.h"
 #include "hmtx.h"
+#include "hvar.h"
 #include "kern.h"
 #include "loca.h"
 #include "ltsh.h"
 #include "math_.h"
 #include "maxp.h"
+#include "mvar.h"
 #include "name.h"
 #include "os2.h"
 #include "ots.h"
 #include "post.h"
 #include "prep.h"
+#include "stat.h"
 #include "vdmx.h"
 #include "vhea.h"
 #include "vmtx.h"
 #include "vorg.h"
+#include "vvar.h"
 
 // Graphite tables
 #ifdef OTS_GRAPHITE
@@ -62,9 +70,8 @@ namespace ots {
 struct Arena {
  public:
   ~Arena() {
-    for (std::vector<uint8_t*>::iterator
-         i = hunks_.begin(); i != hunks_.end(); ++i) {
-      delete[] *i;
+    for (auto& hunk : hunks_) {
+      delete[] hunk;
     }
   }
 
@@ -78,6 +85,17 @@ struct Arena {
   std::vector<uint8_t*> hunks_;
 };
 
+bool CheckTag(uint32_t tag_value) {
+  for (unsigned i = 0; i < 4; ++i) {
+    const uint32_t check = tag_value & 0xff;
+    if (check < 32 || check > 126) {
+      return false;  // non-ASCII character found.
+    }
+    tag_value >>= 8;
+  }
+  return true;
+}
+
 }; // namespace ots
 
 namespace {
@@ -91,17 +109,6 @@ namespace {
 #define OTS_WARNING_MSG_HDR(...)       OTS_WARNING_MSG_(header, __VA_ARGS__)
 
 
-bool CheckTag(uint32_t tag_value) {
-  for (unsigned i = 0; i < 4; ++i) {
-    const uint32_t check = tag_value & 0xff;
-    if (check < 32 || check > 126) {
-      return false;  // non-ASCII character found.
-    }
-    tag_value >>= 8;
-  }
-  return true;
-}
-
 const struct {
   uint32_t tag;
   bool required;
@@ -126,6 +133,17 @@ const struct {
   { OTS_TAG_LTSH, false },
   { OTS_TAG_VORG, false },
   { OTS_TAG_KERN, false },
+  // We need to parse fvar table before other tables that may need to know
+  // the number of variation axes (if any)
+  { OTS_TAG_FVAR, false },
+  { OTS_TAG_AVAR, false },
+  { OTS_TAG_CVAR, false },
+  { OTS_TAG_GVAR, false },
+  { OTS_TAG_HVAR, false },
+  { OTS_TAG_MVAR, false },
+  { OTS_TAG_STAT, false },
+  { OTS_TAG_VVAR, false },
+  { OTS_TAG_CFF2, false },
   // We need to parse GDEF table in advance of parsing GSUB/GPOS tables
   // because they could refer GDEF table.
   { OTS_TAG_GDEF, false },
@@ -580,7 +598,7 @@ bool ProcessGeneric(ots::FontFile *header,
     }
 
     // all tag names must be built from printable ASCII characters
-    if (!CheckTag(tables[i].tag)) {
+    if (!ots::CheckTag(tables[i].tag)) {
       OTS_WARNING_MSG_HDR("Invalid table tag: 0x%X", tables[i].tag);
     }
 
@@ -686,7 +704,7 @@ bool ProcessGeneric(ots::FontFile *header,
     }
   }
 
-  if (font->GetTable(OTS_TAG_CFF)) {
+  if (font->GetTable(OTS_TAG_CFF) || font->GetTable(OTS_TAG_CFF2)) {
     // font with PostScript glyph
     if (font->version != OTS_TAG('O','T','T','O')) {
       return OTS_FAILURE_MSG_HDR("wrong font version for PostScript glyph data");
@@ -862,32 +880,41 @@ bool Font::ParseTable(const TableEntry& table_entry, const uint8_t* data,
     table = new TablePassthru(this, tag);
   } else {
     switch (tag) {
+      case OTS_TAG_AVAR: table = new OpenTypeAVAR(this, tag); break;
       case OTS_TAG_CFF:  table = new OpenTypeCFF(this,  tag); break;
+      case OTS_TAG_CFF2: table = new OpenTypeCFF2(this, tag); break;
       case OTS_TAG_CMAP: table = new OpenTypeCMAP(this, tag); break;
+      case OTS_TAG_CVAR: table = new OpenTypeCVAR(this, tag); break;
       case OTS_TAG_CVT:  table = new OpenTypeCVT(this,  tag); break;
       case OTS_TAG_FPGM: table = new OpenTypeFPGM(this, tag); break;
+      case OTS_TAG_FVAR: table = new OpenTypeFVAR(this, tag); break;
       case OTS_TAG_GASP: table = new OpenTypeGASP(this, tag); break;
       case OTS_TAG_GDEF: table = new OpenTypeGDEF(this, tag); break;
       case OTS_TAG_GLYF: table = new OpenTypeGLYF(this, tag); break;
       case OTS_TAG_GPOS: table = new OpenTypeGPOS(this, tag); break;
       case OTS_TAG_GSUB: table = new OpenTypeGSUB(this, tag); break;
+      case OTS_TAG_GVAR: table = new OpenTypeGVAR(this, tag); break;
       case OTS_TAG_HDMX: table = new OpenTypeHDMX(this, tag); break;
       case OTS_TAG_HEAD: table = new OpenTypeHEAD(this, tag); break;
       case OTS_TAG_HHEA: table = new OpenTypeHHEA(this, tag); break;
       case OTS_TAG_HMTX: table = new OpenTypeHMTX(this, tag); break;
+      case OTS_TAG_HVAR: table = new OpenTypeHVAR(this, tag); break;
       case OTS_TAG_KERN: table = new OpenTypeKERN(this, tag); break;
       case OTS_TAG_LOCA: table = new OpenTypeLOCA(this, tag); break;
       case OTS_TAG_LTSH: table = new OpenTypeLTSH(this, tag); break;
       case OTS_TAG_MATH: table = new OpenTypeMATH(this, tag); break;
       case OTS_TAG_MAXP: table = new OpenTypeMAXP(this, tag); break;
+      case OTS_TAG_MVAR: table = new OpenTypeMVAR(this, tag); break;
       case OTS_TAG_NAME: table = new OpenTypeNAME(this, tag); break;
       case OTS_TAG_OS2:  table = new OpenTypeOS2(this,  tag); break;
       case OTS_TAG_POST: table = new OpenTypePOST(this, tag); break;
       case OTS_TAG_PREP: table = new OpenTypePREP(this, tag); break;
+      case OTS_TAG_STAT: table = new OpenTypeSTAT(this, tag); break;
       case OTS_TAG_VDMX: table = new OpenTypeVDMX(this, tag); break;
-      case OTS_TAG_VORG: table = new OpenTypeVORG(this, tag); break;
       case OTS_TAG_VHEA: table = new OpenTypeVHEA(this, tag); break;
       case OTS_TAG_VMTX: table = new OpenTypeVMTX(this, tag); break;
+      case OTS_TAG_VORG: table = new OpenTypeVORG(this, tag); break;
+      case OTS_TAG_VVAR: table = new OpenTypeVVAR(this, tag); break;
       // Graphite tables
 #ifdef OTS_GRAPHITE
       case OTS_TAG_FEAT: table = new OpenTypeFEAT(this, tag); break;
@@ -953,6 +980,23 @@ void Font::DropGraphite() {
   dropped_graphite = true;
 }
 
+void Font::DropVariations() {
+  file->context->Message(0, "Dropping all Variation tables");
+  for (const std::pair<uint32_t, Table*> entry : m_tables) {
+    if (entry.first == OTS_TAG_AVAR ||
+        entry.first == OTS_TAG_CVAR ||
+        entry.first == OTS_TAG_FVAR ||
+        entry.first == OTS_TAG_GVAR ||
+        entry.first == OTS_TAG_HVAR ||
+        entry.first == OTS_TAG_MVAR ||
+        entry.first == OTS_TAG_STAT ||
+        entry.first == OTS_TAG_VVAR) {
+      entry.second->Drop("Discarding Variations table");
+    }
+  }
+  dropped_variations = true;
+}
+
 bool Table::ShouldSerialize() {
   return m_shouldSerialize;
 }
@@ -960,7 +1004,6 @@ bool Table::ShouldSerialize() {
 void Table::Message(int level, const char *format, va_list va) {
   char msg[206] = { OTS_UNTAG(m_tag), ':', ' ' };
   std::vsnprintf(msg + 6, 200, format, va);
-// fprintf(stderr, format, va); // font debugging, see TenFourFox issue 317
   m_font->file->context->Message(level, msg);
 }
 
@@ -1004,6 +1047,16 @@ bool Table::DropGraphite(const char *format, ...) {
   return true;
 }
 
+bool Table::DropVariations(const char *format, ...) {
+  va_list va;
+  va_start(va, format);
+  Message(0, format, va);
+  va_end(va);
+
+  m_font->DropVariations();
+  return true;
+}
+
 bool TablePassthru::Parse(const uint8_t *data, size_t length) {
   m_data = data;
   m_length = length;

gfx/ots/src/ots.h

@@ -112,7 +112,7 @@ class Buffer {
       return OTS_FAILURE();
     }
     std::memcpy(value, buffer_ + offset_, sizeof(uint16_t));
-    *value = ntohs(*value);
+    *value = ots_ntohs(*value);
     offset_ += 2;
     return true;
   }
@@ -137,7 +137,7 @@ class Buffer {
       return OTS_FAILURE();
     }
     std::memcpy(value, buffer_ + offset_, sizeof(uint32_t));
-    *value = ntohl(*value);
+    *value = ots_ntohl(*value);
     offset_ += 4;
     return true;
   }
@@ -184,9 +184,13 @@ template<typename T> T Round2(T value) {
   return (value + 1) & ~1;
 }
 
+// Check that a tag consists entirely of printable ASCII characters
+bool CheckTag(uint32_t tag_value);
+
 bool IsValidVersionTag(uint32_t tag);
 
 #define OTS_TAG_CFF  OTS_TAG('C','F','F',' ')
+#define OTS_TAG_CFF2 OTS_TAG('C','F','F','2')
 #define OTS_TAG_CMAP OTS_TAG('c','m','a','p')
 #define OTS_TAG_CVT  OTS_TAG('c','v','t',' ')
 #define OTS_TAG_FEAT OTS_TAG('F','e','a','t')
@@ -219,6 +223,15 @@ bool IsValidVersionTag(uint32_t tag);
 #define OTS_TAG_VMTX OTS_TAG('v','m','t','x')
 #define OTS_TAG_VORG OTS_TAG('V','O','R','G')
 
+#define OTS_TAG_AVAR OTS_TAG('a','v','a','r')
+#define OTS_TAG_CVAR OTS_TAG('c','v','a','r')
+#define OTS_TAG_FVAR OTS_TAG('f','v','a','r')
+#define OTS_TAG_GVAR OTS_TAG('g','v','a','r')
+#define OTS_TAG_HVAR OTS_TAG('H','V','A','R')
+#define OTS_TAG_MVAR OTS_TAG('M','V','A','R')
+#define OTS_TAG_VVAR OTS_TAG('V','V','A','R')
+#define OTS_TAG_STAT OTS_TAG('S','T','A','T')
+
 struct Font;
 struct FontFile;
 struct TableEntry;
@@ -252,6 +265,7 @@ class Table {
   bool Warning(const char *format, ...);
   bool Drop(const char *format, ...);
   bool DropGraphite(const char *format, ...);
+  bool DropVariations(const char *format, ...);
 
  private:
   void Message(int level, const char *format, va_list va);
@@ -286,7 +300,8 @@ struct Font {
         search_range(0),
         entry_selector(0),
         range_shift(0),
-        dropped_graphite(false) {
+        dropped_graphite(false),
+        dropped_variations(false) {
   }
 
   bool ParseTable(const TableEntry& tableinfo, const uint8_t* data,
@@ -301,6 +316,9 @@ struct Font {
   // Drop all Graphite tables and don't parse new ones.
   void DropGraphite();
 
+  // Drop all Variations tables and don't parse new ones.
+  void DropVariations();
+
   FontFile *file;
 
   uint32_t version;
@@ -309,6 +327,7 @@ struct Font {
   uint16_t entry_selector;
   uint16_t range_shift;
   bool dropped_graphite;
+  bool dropped_variations;
 
  private:
   std::map<uint32_t, Table*> m_tables;

gfx/ots/src/post.cc

@@ -9,8 +9,6 @@
 // post - PostScript
 // http://www.microsoft.com/typography/otspec/post.htm
 
-#define TABLE_NAME "post"
-
 namespace ots {
 
 bool OpenTypePOST::Parse(const uint8_t *data, size_t length) {

gfx/ots/src/silf.cc

@@ -38,7 +38,19 @@ bool OpenTypeSILF::Parse(const uint8_t* data, size_t length,
         if (prevent_decompression) {
           return DropGraphite("Illegal nested compression");
         }
-        std::vector<uint8_t> decompressed(this->compHead & FULL_SIZE);
+        size_t decompressed_size = this->compHead & FULL_SIZE;
+        if (decompressed_size < length) {
+          return DropGraphite("Decompressed size is less than compressed size");
+        }
+        if (decompressed_size == 0) {
+          return DropGraphite("Decompressed size is set to 0");
+        }
+        // decompressed table must be <= 30MB
+        if (decompressed_size > 30 * 1024 * 1024) {
+          return DropGraphite("Decompressed size exceeds 30MB: %gMB",
+                              decompressed_size / (1024.0 * 1024.0));
+        }
+        std::vector<uint8_t> decompressed(decompressed_size);
         size_t outputSize = 0;
         bool ret = mozilla::Compression::LZ4::decompressPartial(
             reinterpret_cast<const char*>(data + table.offset()),
@@ -165,15 +177,9 @@ bool OpenTypeSILF::SILSub::ParsePart(Buffer& table) {
     if (!table.ReadU8(&this->attrMirroring)) {
       return parent->Error("SILSub: Failed to read attrMirroring");
     }
-    if (parent->version >> 16 < 4 && this->attrMirroring != 0) {
-      parent->Warning("SILSub: Nonzero attrMirroring (reserved before v4)");
-    }
     if (!table.ReadU8(&this->attrSkipPasses)) {
       return parent->Error("SILSub: Failed to read attrSkipPasses");
     }
-    if (parent->version >> 16 < 4 && this->attrSkipPasses != 0) {
-      parent->Warning("SILSub: Nonzero attrSkipPasses (reserved2 before v4)");
-    }
 
     if (!table.ReadU8(&this->numJLevels)) {
       return parent->Error("SILSub: Failed to read numJLevels");
@@ -642,9 +648,6 @@ SILPass::ParsePart(Buffer& table, const size_t SILSub_init_offset,
     if (!table.ReadU16(&this->fsmOffset)) {
       return parent->Error("SILPass: Failed to read fsmOffset");
     }
-    if (parent->version >> 16 == 2 && this->fsmOffset != 0) {
-      parent->Warning("SILPass: Nonzero fsmOffset (reserved in SILSub v2)");
-    }
     if (!table.ReadU32(&this->pcCode) ||
         (parent->version >= 3 && this->pcCode < this->fsmOffset)) {
       return parent->Error("SILPass: Failed to read pcCode");
@@ -770,10 +773,6 @@ SILPass::ParsePart(Buffer& table, const size_t SILSub_init_offset,
     if (!table.ReadU8(&this->collisionThreshold)) {
       return parent->Error("SILPass: Failed to read collisionThreshold");
     }
-    if (parent->version >> 16 < 5 && this->collisionThreshold != 0) {
-      parent->Warning("SILPass: Nonzero collisionThreshold"
-                      " (reserved before v5)");
-    }
     if (!table.ReadU16(&this->pConstraint)) {
       return parent->Error("SILPass: Failed to read pConstraint");
     }

gfx/ots/src/stat.cc

@@ -0,0 +1,347 @@
+// Copyright (c) 2018 The OTS Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "stat.h"
+#include "name.h"
+
+namespace ots {
+
+// -----------------------------------------------------------------------------
+// OpenTypeSTAT
+// -----------------------------------------------------------------------------
+
+bool OpenTypeSTAT::ValidateNameId(uint16_t nameid, bool allowPredefined) {
+  OpenTypeNAME* name = static_cast<OpenTypeNAME*>(
+      GetFont()->GetTypedTable(OTS_TAG_NAME));
+
+  if (!name || !name->IsValidNameId(nameid)) {
+    Drop("Invalid nameID: %d", nameid);
+    return false;
+  }
+
+  if (!allowPredefined && nameid < 26) {
+    Warning("nameID out of range: %d", nameid);
+    return true;
+  }
+
+  if ((nameid >= 26 && nameid <= 255) || nameid >= 32768) {
+    Warning("nameID out of range: %d", nameid);
+    return true;
+  }
+
+  return  true;
+}
+
+bool OpenTypeSTAT::Parse(const uint8_t* data, size_t length) {
+  Buffer table(data, length);
+  if (!table.ReadU16(&this->majorVersion) ||
+      !table.ReadU16(&this->minorVersion) ||
+      !table.ReadU16(&this->designAxisSize) ||
+      !table.ReadU16(&this->designAxisCount) ||
+      !table.ReadU32(&this->designAxesOffset) ||
+      !table.ReadU16(&this->axisValueCount) ||
+      !table.ReadU32(&this->offsetToAxisValueOffsets) ||
+      !(this->minorVersion < 1 || table.ReadU16(&this->elidedFallbackNameID))) {
+    return Drop("Failed to read table header");
+  }
+  if (this->majorVersion != 1) {
+    return Drop("Unknown table version");
+  }
+  if (this->minorVersion > 2) {
+    Warning("Unknown minor version, downgrading to 2");
+    this->minorVersion = 2;
+  }
+
+  if (this->designAxisSize < sizeof(AxisRecord)) {
+    return Drop("Invalid designAxisSize");
+  }
+
+  size_t headerEnd = table.offset();
+
+  if (this->designAxisCount == 0) {
+    if (this->designAxesOffset != 0) {
+      Warning("Unexpected non-zero designAxesOffset");
+      this->designAxesOffset = 0;
+    }
+  } else {
+    if (this->designAxesOffset < headerEnd ||
+        size_t(this->designAxesOffset) +
+          size_t(this->designAxisCount) * size_t(this->designAxisSize) > length) {
+      return Drop("Invalid designAxesOffset");
+    }
+  }
+
+  for (size_t i = 0; i < this->designAxisCount; i++) {
+    table.set_offset(this->designAxesOffset + i * this->designAxisSize);
+    this->designAxes.emplace_back();
+    auto& axis = this->designAxes[i];
+    if (!table.ReadU32(&axis.axisTag) ||
+        !table.ReadU16(&axis.axisNameID) ||
+        !table.ReadU16(&axis.axisOrdering)) {
+      return Drop("Failed to read design axis");
+    }
+    if (!CheckTag(axis.axisTag)) {
+      return Drop("Bad design axis tag");
+    }
+    if (!ValidateNameId(axis.axisNameID, false)) {
+      return true;
+    }
+  }
+
+  // TODO
+  // - check that all axes defined in fvar are covered by STAT
+  // - check that axisOrdering values are not duplicated (warn only)
+
+  if (this->axisValueCount == 0) {
+    if (this->offsetToAxisValueOffsets != 0) {
+      Warning("Unexpected non-zero offsetToAxisValueOffsets");
+      this->offsetToAxisValueOffsets = 0;
+    }
+  } else {
+    if (this->offsetToAxisValueOffsets < headerEnd ||
+        size_t(this->offsetToAxisValueOffsets) +
+          size_t(this->axisValueCount) * sizeof(uint16_t) > length) {
+      return Drop("Invalid offsetToAxisValueOffsets");
+    }
+  }
+
+  for (size_t i = 0; i < this->axisValueCount; i++) {
+    table.set_offset(this->offsetToAxisValueOffsets + i * sizeof(uint16_t));
+    uint16_t axisValueOffset;
+    if (!table.ReadU16(&axisValueOffset)) {
+      return Drop("Failed to read axis value offset");
+    }
+    if (this->offsetToAxisValueOffsets + axisValueOffset > length) {
+      return Drop("Invalid axis value offset");
+    }
+    table.set_offset(this->offsetToAxisValueOffsets + axisValueOffset);
+    uint16_t format;
+    if (!table.ReadU16(&format)) {
+      return Drop("Failed to read axis value format");
+    }
+    this->axisValues.emplace_back(format);
+    auto& axisValue = axisValues[i];
+    switch (format) {
+    case 1:
+      if (!table.ReadU16(&axisValue.format1.axisIndex) ||
+          !table.ReadU16(&axisValue.format1.flags) ||
+          !table.ReadU16(&axisValue.format1.valueNameID) ||
+          !table.ReadS32(&axisValue.format1.value)) {
+        return Drop("Failed to read axis value (format 1)");
+      }
+      if (axisValue.format1.axisIndex >= this->designAxisCount) {
+        return Drop("Axis index out of range");
+      }
+      if ((axisValue.format1.flags & 0xFFFCu) != 0) {
+        Warning("Unexpected axis value flags");
+        axisValue.format1.flags &= ~0xFFFCu;
+      }
+      if (!ValidateNameId(axisValue.format1.valueNameID)) {
+        return true;
+      }
+      break;
+    case 2:
+      if (!table.ReadU16(&axisValue.format2.axisIndex) ||
+          !table.ReadU16(&axisValue.format2.flags) ||
+          !table.ReadU16(&axisValue.format2.valueNameID) ||
+          !table.ReadS32(&axisValue.format2.nominalValue) ||
+          !table.ReadS32(&axisValue.format2.rangeMinValue) ||
+          !table.ReadS32(&axisValue.format2.rangeMaxValue)) {
+        return Drop("Failed to read axis value (format 2)");
+      }
+      if (axisValue.format2.axisIndex >= this->designAxisCount) {
+        return Drop("Axis index out of range");
+      }
+      if ((axisValue.format2.flags & 0xFFFCu) != 0) {
+        Warning("Unexpected axis value flags");
+        axisValue.format1.flags &= ~0xFFFCu;
+      }
+      if (!ValidateNameId(axisValue.format2.valueNameID)) {
+        return true;
+      }
+      if (!(axisValue.format2.rangeMinValue <= axisValue.format2.nominalValue &&
+            axisValue.format2.nominalValue <= axisValue.format2.rangeMaxValue)) {
+        Warning("Bad axis value range or nominal value");
+      }
+      break;
+    case 3:
+      if (!table.ReadU16(&axisValue.format3.axisIndex) ||
+          !table.ReadU16(&axisValue.format3.flags) ||
+          !table.ReadU16(&axisValue.format3.valueNameID) ||
+          !table.ReadS32(&axisValue.format3.value) ||
+          !table.ReadS32(&axisValue.format3.linkedValue)) {
+        return Drop("Failed to read axis value (format 3)");
+      }
+      if (axisValue.format3.axisIndex >= this->designAxisCount) {
+        return Drop("Axis index out of range");
+      }
+      if ((axisValue.format3.flags & 0xFFFCu) != 0) {
+        Warning("Unexpected axis value flags");
+        axisValue.format3.flags &= ~0xFFFCu;
+      }
+      if (!ValidateNameId(axisValue.format3.valueNameID)) {
+        return true;
+      }
+      break;
+    case 4:
+      if (this->minorVersion < 2) {
+        Warning("Invalid table version for format 4 axis values - updating");
+        this->minorVersion = 2;
+      }
+      if (!table.ReadU16(&axisValue.format4.axisCount) ||
+          !table.ReadU16(&axisValue.format4.flags) ||
+          !table.ReadU16(&axisValue.format4.valueNameID)) {
+        return Drop("Failed to read axis value (format 4)");
+      }
+      if (axisValue.format4.axisCount > this->designAxisCount) {
+        return Drop("Axis count out of range");
+      }
+      if ((axisValue.format4.flags & 0xFFFCu) != 0) {
+        Warning("Unexpected axis value flags");
+        axisValue.format4.flags &= ~0xFFFCu;
+      }
+      if (!ValidateNameId(axisValue.format4.valueNameID)) {
+        return true;
+      }
+      for (unsigned j = 0; j < axisValue.format4.axisCount; j++) {
+        axisValue.format4.axisValues.emplace_back();
+        auto& v = axisValue.format4.axisValues[j];
+        if (!table.ReadU16(&v.axisIndex) ||
+            !table.ReadS32(&v.value)) {
+          return Drop("Failed to read axis value");
+        }
+        if (v.axisIndex >= this->designAxisCount) {
+          return Drop("Axis index out of range");
+        }
+      }
+      break;
+    default:
+      return Drop("Unknown axis value format");
+    }
+  }
+
+  return true;
+}
+
+bool OpenTypeSTAT::Serialize(OTSStream* out) {
+  off_t tableStart = out->Tell();
+
+  size_t headerSize = 5 * sizeof(uint16_t) + 2 * sizeof(uint32_t);
+  if (this->minorVersion >= 1) {
+    headerSize += sizeof(uint16_t);
+  }
+
+  if (this->designAxisCount == 0) {
+    this->designAxesOffset = 0;
+  } else {
+    this->designAxesOffset = headerSize;
+  }
+
+  this->designAxisSize = sizeof(AxisRecord);
+
+  if (this->axisValueCount == 0) {
+    this->offsetToAxisValueOffsets = 0;
+  } else {
+    if (this->designAxesOffset == 0) {
+      this->offsetToAxisValueOffsets = headerSize;
+    } else {
+      this->offsetToAxisValueOffsets = this->designAxesOffset + this->designAxisCount * this->designAxisSize;
+    }
+  }
+
+  if (!out->WriteU16(this->majorVersion) ||
+      !out->WriteU16(this->minorVersion) ||
+      !out->WriteU16(this->designAxisSize) ||
+      !out->WriteU16(this->designAxisCount) ||
+      !out->WriteU32(this->designAxesOffset) ||
+      !out->WriteU16(this->axisValueCount) ||
+      !out->WriteU32(this->offsetToAxisValueOffsets) ||
+      !(this->minorVersion < 1 || out->WriteU16(this->elidedFallbackNameID))) {
+    return Error("Failed to write table header");
+  }
+
+  if (this->designAxisCount > 0) {
+    if (out->Tell() - tableStart != this->designAxesOffset) {
+      return Error("Error computing designAxesOffset");
+    }
+  }
+
+  for (unsigned i = 0; i < this->designAxisCount; i++) {
+    const auto& axis = this->designAxes[i];
+    if (!out->WriteU32(axis.axisTag) ||
+        !out->WriteU16(axis.axisNameID) ||
+        !out->WriteU16(axis.axisOrdering)) {
+      return Error("Failed to write design axis");
+    }
+  }
+
+  if (this->axisValueCount > 0) {
+    if (out->Tell() - tableStart != this->offsetToAxisValueOffsets) {
+      return Error("Error computing offsetToAxisValueOffsets");
+    }
+  }
+
+  uint32_t axisValueOffset = this->axisValueCount * sizeof(uint16_t);
+  for (unsigned i = 0; i < this->axisValueCount; i++) {
+    const auto& value = this->axisValues[i];
+    if (!out->WriteU16(axisValueOffset)) {
+      return Error("Failed to write axis value offset");
+    }
+    axisValueOffset += value.Length();
+  }
+  for (unsigned i = 0; i < this->axisValueCount; i++) {
+    const auto& value = this->axisValues[i];
+    if (!out->WriteU16(value.format)) {
+      return Error("Failed to write axis value");
+    }
+    switch (value.format) {
+    case 1:
+      if (!out->WriteU16(value.format1.axisIndex) ||
+          !out->WriteU16(value.format1.flags) ||
+          !out->WriteU16(value.format1.valueNameID) ||
+          !out->WriteS32(value.format1.value)) {
+        return Error("Failed to write axis value");
+      }
+      break;
+    case 2:
+      if (!out->WriteU16(value.format2.axisIndex) ||
+          !out->WriteU16(value.format2.flags) ||
+          !out->WriteU16(value.format2.valueNameID) ||
+          !out->WriteS32(value.format2.nominalValue) ||
+          !out->WriteS32(value.format2.rangeMinValue) ||
+          !out->WriteS32(value.format2.rangeMaxValue)) {
+        return Error("Failed to write axis value");
+      }
+      break;
+    case 3:
+      if (!out->WriteU16(value.format3.axisIndex) ||
+          !out->WriteU16(value.format3.flags) ||
+          !out->WriteU16(value.format3.valueNameID) ||
+          !out->WriteS32(value.format3.value) ||
+          !out->WriteS32(value.format3.linkedValue)) {
+        return Error("Failed to write axis value");
+      }
+      break;
+    case 4:
+      if (!out->WriteU16(value.format4.axisCount) ||
+          !out->WriteU16(value.format4.flags) ||
+          !out->WriteU16(value.format4.valueNameID)) {
+        return Error("Failed to write axis value");
+      }
+      for (unsigned j = 0; j < value.format4.axisValues.size(); j++) {
+        if (!out->WriteU16(value.format4.axisValues[j].axisIndex) ||
+            !out->WriteS32(value.format4.axisValues[j].value)) {
+          return Error("Failed to write axis value");
+        }
+      }
+      break;
+    default:
+      return Error("Bad value format");
+    }
+  }
+
+  return true;
+}
+
+}  // namespace ots

gfx/ots/src/stat.h

@@ -0,0 +1,155 @@
+// Copyright (c) 2018 The OTS Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef OTS_STAT_H_
+#define OTS_STAT_H_
+
+#include <vector>
+
+#include "ots.h"
+
+namespace ots {
+
+// -----------------------------------------------------------------------------
+// OpenTypeSTAT Interface
+// -----------------------------------------------------------------------------
+
+class OpenTypeSTAT : public Table {
+ public:
+  explicit OpenTypeSTAT(Font* font, uint32_t tag)
+      : Table(font, tag, tag) { }
+
+  bool Parse(const uint8_t* data, size_t length);
+  bool Serialize(OTSStream* out);
+
+ private:
+  bool ValidateNameId(uint16_t nameid, bool allowPredefined = true);
+
+  uint16_t majorVersion;
+  uint16_t minorVersion;
+  uint16_t designAxisSize;
+  uint16_t designAxisCount;
+  uint32_t designAxesOffset;
+  uint16_t axisValueCount;
+  uint32_t offsetToAxisValueOffsets;
+  uint16_t elidedFallbackNameID;
+
+  struct AxisRecord {
+    uint32_t axisTag;
+    uint16_t axisNameID;
+    uint16_t axisOrdering;
+  };
+  std::vector<AxisRecord> designAxes;
+
+  typedef int32_t Fixed; /* 16.16 fixed-point value */
+
+  struct AxisValueFormat1 {
+    uint16_t axisIndex;
+    uint16_t flags;
+    uint16_t valueNameID;
+    Fixed    value;
+    static size_t Length() {
+      return 3 * sizeof(uint16_t) + sizeof(Fixed);
+    }
+  };
+
+  struct AxisValueFormat2 {
+    uint16_t axisIndex;
+    uint16_t flags;
+    uint16_t valueNameID;
+    Fixed    nominalValue;
+    Fixed    rangeMinValue;
+    Fixed    rangeMaxValue;
+    static size_t Length() {
+      return 3 * sizeof(uint16_t) + 3 * sizeof(Fixed);
+    }
+  };
+
+  struct AxisValueFormat3 {
+    uint16_t axisIndex;
+    uint16_t flags;
+    uint16_t valueNameID;
+    Fixed    value;
+    Fixed    linkedValue;
+    static size_t Length() {
+      return 3 * sizeof(uint16_t) + 2 * sizeof(Fixed);
+    }
+  };
+
+  struct AxisValueFormat4 {
+    uint16_t axisCount;
+    uint16_t flags;
+    uint16_t valueNameID;
+    struct AxisValue {
+      uint16_t axisIndex;
+      Fixed    value;
+    };
+    std::vector<AxisValue> axisValues;
+    size_t Length() const {
+      return 3 * sizeof(uint16_t) + axisValues.size() * (sizeof(uint16_t) + sizeof(Fixed));
+    }
+  };
+
+  struct AxisValueRecord {
+    uint16_t format;
+    union {
+      AxisValueFormat1 format1;
+      AxisValueFormat2 format2;
+      AxisValueFormat3 format3;
+      AxisValueFormat4 format4;
+    };
+    explicit AxisValueRecord(uint16_t format_)
+      : format(format_)
+    {
+      if (format == 4) {
+        new (&this->format4) AxisValueFormat4();
+      }
+    }
+    AxisValueRecord(const AxisValueRecord& other_)
+      : format(other_.format)
+    {
+      switch (format) {
+      case 1:
+        format1 = other_.format1;
+        break;
+      case 2:
+        format2 = other_.format2;
+        break;
+      case 3:
+        format3 = other_.format3;
+        break;
+      case 4:
+        new (&this->format4) AxisValueFormat4();
+        format4 = other_.format4;
+        break;
+      }
+    }
+    ~AxisValueRecord() {
+      if (format == 4) {
+        this->format4.~AxisValueFormat4();
+      }
+    }
+    uint32_t Length() const {
+      switch (format) {
+      case 1:
+        return sizeof(uint16_t) + format1.Length();
+      case 2:
+        return sizeof(uint16_t) + format2.Length();
+      case 3:
+        return sizeof(uint16_t) + format3.Length();
+      case 4:
+        return sizeof(uint16_t) + format4.Length();
+      default:
+        // can't happen
+        return 0;
+      }
+    }
+  };
+
+  std::vector<AxisValueRecord> axisValues;
+};
+
+}  // namespace ots
+
+#endif  // OTS_STAT_H_

gfx/ots/src/variations.cc

@@ -0,0 +1,261 @@
+// Copyright (c) 2018 The OTS Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "layout.h"
+
+#include "fvar.h"
+
+// OpenType Variations Common Table Formats
+
+#define TABLE_NAME "Variations" // XXX: use individual table names
+
+namespace {
+
+bool ParseVariationRegionList(const ots::Font* font, const uint8_t* data, const size_t length,
+                              uint16_t* regionCount) {
+  ots::Buffer subtable(data, length);
+
+  uint16_t axisCount;
+
+  if (!subtable.ReadU16(&axisCount) ||
+      !subtable.ReadU16(regionCount)) {
+    return OTS_FAILURE_MSG("Failed to read variation region list header");
+  }
+
+  if (*regionCount == 0) {
+    return true;
+  }
+
+  const ots::OpenTypeFVAR* fvar =
+    static_cast<ots::OpenTypeFVAR*>(font->GetTypedTable(OTS_TAG_FVAR));
+  if (!fvar) {
+    return OTS_FAILURE_MSG("Required fvar table is missing");
+  }
+  if (axisCount != fvar->AxisCount()) {
+    return OTS_FAILURE_MSG("Axis count mismatch");
+  }
+
+  for (unsigned i = 0; i < *regionCount; i++) {
+    for (unsigned j = 0; j < axisCount; j++) {
+      int16_t startCoord, peakCoord, endCoord;
+      if (!subtable.ReadS16(&startCoord) ||
+          !subtable.ReadS16(&peakCoord) ||
+          !subtable.ReadS16(&endCoord)) {
+        return OTS_FAILURE_MSG("Failed to read region axis coordinates");
+      }
+      if (startCoord > peakCoord || peakCoord > endCoord) {
+        return OTS_FAILURE_MSG("Region axis coordinates out of order");
+      }
+      if (startCoord < -0x4000 || endCoord > 0x4000) {
+        return OTS_FAILURE_MSG("Region axis coordinate out of range");
+      }
+      if ((peakCoord < 0 && endCoord > 0) ||
+          (peakCoord > 0 && startCoord < 0)) {
+        return OTS_FAILURE_MSG("Invalid region axis coordinates");
+      }
+    }
+  }
+
+  return true;
+}
+
+bool
+ParseVariationDataSubtable(const ots::Font* font, const uint8_t* data, const size_t length,
+                           const uint16_t regionCount,
+                           uint16_t* regionIndexCount) {
+  ots::Buffer subtable(data, length);
+
+  uint16_t itemCount;
+  uint16_t shortDeltaCount;
+
+  if (!subtable.ReadU16(&itemCount) ||
+      !subtable.ReadU16(&shortDeltaCount) ||
+      !subtable.ReadU16(regionIndexCount)) {
+    return OTS_FAILURE_MSG("Failed to read variation data subtable header");
+  }
+
+  for (unsigned i = 0; i < *regionIndexCount; i++) {
+    uint16_t regionIndex;
+    if (!subtable.ReadU16(&regionIndex) || regionIndex >= regionCount) {
+      return OTS_FAILURE_MSG("Bad region index");
+    }
+  }
+
+  if (!subtable.Skip(size_t(itemCount) * (size_t(shortDeltaCount) + size_t(*regionIndexCount)))) {
+    return OTS_FAILURE_MSG("Failed to read delta data");
+  }
+
+  return true;
+}
+
+} // namespace
+
+namespace ots {
+
+bool
+ParseItemVariationStore(const Font* font,
+                        const uint8_t* data, const size_t length,
+                        std::vector<uint16_t>* regionIndexCounts) {
+  Buffer subtable(data, length);
+
+  uint16_t format;
+  uint32_t variationRegionListOffset;
+  uint16_t itemVariationDataCount;
+
+  if (!subtable.ReadU16(&format) ||
+      !subtable.ReadU32(&variationRegionListOffset) ||
+      !subtable.ReadU16(&itemVariationDataCount)) {
+    return OTS_FAILURE_MSG("Failed to read item variation store header");
+  }
+
+  if (format != 1) {
+    return OTS_FAILURE_MSG("Unknown item variation store format");
+  }
+
+  if (variationRegionListOffset < subtable.offset() + 4 * itemVariationDataCount ||
+      variationRegionListOffset > length) {
+    return OTS_FAILURE_MSG("Invalid variation region list offset");
+  }
+
+  uint16_t regionCount;
+  if (!ParseVariationRegionList(font,
+                                data + variationRegionListOffset,
+                                length - variationRegionListOffset,
+                                &regionCount)) {
+    return OTS_FAILURE_MSG("Failed to parse variation region list");
+  }
+
+  for (unsigned i = 0; i < itemVariationDataCount; i++) {
+    uint32_t offset;
+    if (!subtable.ReadU32(&offset)) {
+      return OTS_FAILURE_MSG("Failed to read variation data subtable offset");
+    }
+    if (offset >= length) {
+      return OTS_FAILURE_MSG("Bad offset to variation data subtable");
+    }
+    uint16_t regionIndexCount = 0;
+    if (!ParseVariationDataSubtable(font, data + offset, length - offset,
+                                    regionCount,
+                                    &regionIndexCount)) {
+      return OTS_FAILURE_MSG("Failed to parse variation data subtable");
+    }
+    if (regionIndexCounts) {
+      regionIndexCounts->push_back(regionIndexCount);
+    }
+  }
+
+  return true;
+}
+
+bool ParseDeltaSetIndexMap(const Font* font, const uint8_t* data, const size_t length) {
+  Buffer subtable(data, length);
+
+  uint16_t entryFormat;
+  uint16_t mapCount;
+
+  if (!subtable.ReadU16(&entryFormat) ||
+      !subtable.ReadU16(&mapCount)) {
+    return OTS_FAILURE_MSG("Failed to read delta set index map header");
+  }
+
+  const uint16_t MAP_ENTRY_SIZE_MASK = 0x0030;
+
+  const uint16_t entrySize = (((entryFormat & MAP_ENTRY_SIZE_MASK) >> 4) + 1);
+  if (!subtable.Skip(entrySize * mapCount)) {
+    return OTS_FAILURE_MSG("Failed to read delta set index map data");
+  }
+
+  return true;
+}
+
+bool ParseVariationData(const Font* font, const uint8_t* data, size_t length,
+                        size_t axisCount, size_t sharedTupleCount) {
+  Buffer subtable(data, length);
+
+  uint16_t tupleVariationCount;
+  uint16_t dataOffset;
+  if (!subtable.ReadU16(&tupleVariationCount) ||
+      !subtable.ReadU16(&dataOffset)) {
+    return OTS_FAILURE_MSG("Failed to read variation data header");
+  }
+
+  if (dataOffset > length) {
+    return OTS_FAILURE_MSG("Invalid serialized data offset");
+  }
+
+  tupleVariationCount &= 0x0FFF; // mask off flags
+
+  const uint16_t EMBEDDED_PEAK_TUPLE = 0x8000;
+  const uint16_t INTERMEDIATE_REGION = 0x4000;
+  const uint16_t TUPLE_INDEX_MASK    = 0x0FFF;
+
+  for (unsigned i = 0; i < tupleVariationCount; i++) {
+    uint16_t variationDataSize;
+    uint16_t tupleIndex;
+
+    if (!subtable.ReadU16(&variationDataSize) ||
+        !subtable.ReadU16(&tupleIndex)) {
+      return OTS_FAILURE_MSG("Failed to read tuple variation header");
+    }
+
+    if (tupleIndex & EMBEDDED_PEAK_TUPLE) {
+      for (unsigned axis = 0; axis < axisCount; axis++) {
+        int16_t coordinate;
+        if (!subtable.ReadS16(&coordinate)) {
+          return OTS_FAILURE_MSG("Failed to read tuple coordinate");
+        }
+        if (coordinate < -0x4000 || coordinate > 0x4000) {
+          return OTS_FAILURE_MSG("Invalid tuple coordinate");
+        }
+      }
+    }
+
+    if (tupleIndex & INTERMEDIATE_REGION) {
+      std::vector<int16_t> startTuple(axisCount);
+      for (unsigned axis = 0; axis < axisCount; axis++) {
+        int16_t coordinate;
+        if (!subtable.ReadS16(&coordinate)) {
+          return OTS_FAILURE_MSG("Failed to read tuple coordinate");
+        }
+        if (coordinate < -0x4000 || coordinate > 0x4000) {
+          return OTS_FAILURE_MSG("Invalid tuple coordinate");
+        }
+        startTuple.push_back(coordinate);
+      }
+
+      std::vector<int16_t> endTuple(axisCount);
+      for (unsigned axis = 0; axis < axisCount; axis++) {
+        int16_t coordinate;
+        if (!subtable.ReadS16(&coordinate)) {
+          return OTS_FAILURE_MSG("Failed to read tuple coordinate");
+        }
+        if (coordinate < -0x4000 || coordinate > 0x4000) {
+          return OTS_FAILURE_MSG("Invalid tuple coordinate");
+        }
+        endTuple.push_back(coordinate);
+      }
+
+      for (unsigned axis = 0; axis < axisCount; axis++) {
+        if (startTuple[axis] > endTuple[axis]) {
+          return OTS_FAILURE_MSG("Invalid intermediate range");
+        }
+      }
+    }
+
+    if (!(tupleIndex & EMBEDDED_PEAK_TUPLE)) {
+      tupleIndex &= TUPLE_INDEX_MASK;
+      if (tupleIndex >= sharedTupleCount) {
+        return OTS_FAILURE_MSG("Tuple index out of range");
+      }
+    }
+  }
+
+  // TODO: we don't attempt to interpret the serialized data block
+
+  return true;
+}
+
+} // namespace ots
+
+#undef TABLE_NAME

gfx/ots/src/variations.h

@@ -0,0 +1,26 @@
+// Copyright (c) 2018 The OTS Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef OTS_VARIATIONS_H_
+#define OTS_VARIATIONS_H_
+
+#include <vector>
+#include "ots.h"
+
+// Utility functions for OpenType variations common table formats.
+
+namespace ots {
+
+bool ParseItemVariationStore(const Font* font,
+                             const uint8_t* data, const size_t length,
+                             std::vector<uint16_t>* out_region_index_count = NULL);
+
+bool ParseDeltaSetIndexMap(const Font* font, const uint8_t* data, const size_t length);
+
+bool ParseVariationData(const Font* font, const uint8_t* data, size_t length,
+                        size_t axisCount, size_t sharedTupleCount);
+
+}  // namespace ots
+
+#endif  // OTS_VARIATIONS_H_

gfx/ots/src/vvar.cc

@@ -0,0 +1,95 @@
+// Copyright (c) 2018 The OTS Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "vvar.h"
+
+#include "variations.h"
+
+namespace ots {
+
+// -----------------------------------------------------------------------------
+// OpenTypeVVAR
+// -----------------------------------------------------------------------------
+
+bool OpenTypeVVAR::Parse(const uint8_t* data, size_t length) {
+  Buffer table(data, length);
+
+  uint16_t majorVersion;
+  uint16_t minorVersion;
+  uint32_t itemVariationStoreOffset;
+  uint32_t advanceHeightMappingOffset;
+  uint32_t tsbMappingOffset;
+  uint32_t bsbMappingOffset;
+  uint32_t vOrgMappingOffset;
+
+  if (!table.ReadU16(&majorVersion) ||
+      !table.ReadU16(&minorVersion) ||
+      !table.ReadU32(&itemVariationStoreOffset) ||
+      !table.ReadU32(&advanceHeightMappingOffset) ||
+      !table.ReadU32(&tsbMappingOffset) ||
+      !table.ReadU32(&bsbMappingOffset) ||
+      !table.ReadU32(&vOrgMappingOffset)) {
+    return DropVariations("Failed to read table header");
+  }
+
+  if (majorVersion != 1) {
+    return DropVariations("Unknown table version");
+  }
+
+  if (itemVariationStoreOffset > length ||
+      advanceHeightMappingOffset > length ||
+      tsbMappingOffset > length ||
+      bsbMappingOffset > length ||
+      vOrgMappingOffset > length) {
+    return DropVariations("Invalid subtable offset");
+  }
+
+  if (!ParseItemVariationStore(GetFont(), data + itemVariationStoreOffset,
+                               length - itemVariationStoreOffset)) {
+    return DropVariations("Failed to parse item variation store");
+  }
+
+  if (advanceHeightMappingOffset) {
+    if (!ParseDeltaSetIndexMap(GetFont(), data + advanceHeightMappingOffset,
+                               length - advanceHeightMappingOffset)) {
+      return DropVariations("Failed to parse advance height mappings");
+    }
+  }
+
+  if (tsbMappingOffset) {
+    if (!ParseDeltaSetIndexMap(GetFont(), data + tsbMappingOffset,
+                               length - tsbMappingOffset)) {
+      return DropVariations("Failed to parse TSB mappings");
+    }
+  }
+
+  if (bsbMappingOffset) {
+    if (!ParseDeltaSetIndexMap(GetFont(), data + bsbMappingOffset,
+                               length - bsbMappingOffset)) {
+      return DropVariations("Failed to parse BSB mappings");
+    }
+  }
+
+  if (vOrgMappingOffset) {
+    if (!ParseDeltaSetIndexMap(GetFont(), data + vOrgMappingOffset,
+                               length - vOrgMappingOffset)) {
+      return DropVariations("Failed to parse vOrg mappings");
+    }
+  }
+
+  this->m_data = data;
+  this->m_length = length;
+
+  return true;
+}
+
+bool OpenTypeVVAR::Serialize(OTSStream* out) {
+  if (!out->Write(this->m_data, this->m_length)) {
+    return Error("Failed to write VVAR table");
+  }
+
+  return true;
+}
+
+}  // namespace ots

gfx/ots/src/vvar.h

@@ -0,0 +1,31 @@
+// Copyright (c) 2018 The OTS Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef OTS_VVAR_H_
+#define OTS_VVAR_H_
+
+#include "ots.h"
+
+namespace ots {
+
+// -----------------------------------------------------------------------------
+// OpenTypeVVAR Interface
+// -----------------------------------------------------------------------------
+
+class OpenTypeVVAR : public Table {
+ public:
+  explicit OpenTypeVVAR(Font* font, uint32_t tag)
+      : Table(font, tag, tag) { }
+
+  bool Parse(const uint8_t* data, size_t length);
+  bool Serialize(OTSStream* out);
+
+ private:
+  const uint8_t *m_data;
+  size_t m_length;
+};
+
+}  // namespace ots
+
+#endif  // OTS_VVAR_H_

gfx/ots/tests/layout_common_table_test.cc

@@ -533,7 +533,7 @@ TEST_F(LookupListTableTest, TesBadLookupFlag) {
   // Set IgnoreBaseGlyphs(0x0002) to the lookup flag of LookupTable[0].
   out.Seek(6);
   out.WriteU16(0x0002);
-  EXPECT_FALSE(Parse());
+  EXPECT_TRUE(Parse());
 }
 
 TEST_F(LookupListTableTest, TesBadSubtableCount) {

gfx/thebes/gfxUserFontSet.cpp

@@ -180,18 +180,9 @@ public:
 
     virtual ots::TableAction GetTableAction(uint32_t aTag) override {
         // Preserve Graphite, color glyph and SVG tables
-        if (
-#ifdef RELEASE_OR_BETA // For Beta/Release, also allow OT Layout tables through
-                     // unchecked, and rely on harfbuzz to handle them safely.
-            aTag == TRUETYPE_TAG('G', 'D', 'E', 'F') ||
+        if (aTag == TRUETYPE_TAG('G', 'D', 'E', 'F') ||
             aTag == TRUETYPE_TAG('G', 'P', 'O', 'S') ||
             aTag == TRUETYPE_TAG('G', 'S', 'U', 'B') ||
-#endif
-            aTag == TRUETYPE_TAG('S', 'i', 'l', 'f') ||
-            aTag == TRUETYPE_TAG('S', 'i', 'l', 'l') ||
-            aTag == TRUETYPE_TAG('G', 'l', 'o', 'c') ||
-            aTag == TRUETYPE_TAG('G', 'l', 'a', 't') ||
-            aTag == TRUETYPE_TAG('F', 'e', 'a', 't') ||
             aTag == TRUETYPE_TAG('S', 'V', 'G', ' ') ||
             aTag == TRUETYPE_TAG('C', 'O', 'L', 'R') ||
             aTag == TRUETYPE_TAG('C', 'P', 'A', 'L')) {

js/src/builtin/Promise.cpp

@@ -2006,6 +2006,13 @@ CommonStaticResolveRejectImpl(JSContext* cx, HandleValue thisVal, HandleValue ar
     return promise;
 }
 
+MOZ_MUST_USE JSObject*
+js::PromiseResolve(JSContext* cx, HandleObject constructor, HandleValue value)
+{
+    RootedValue C(cx, ObjectValue(*constructor));
+    return CommonStaticResolveRejectImpl(cx, C, value, ResolveMode);
+}
+
 /**
  * ES2016, 25.4.4.4, Promise.reject.
  */
@@ -2739,6 +2746,7 @@ CreatePromisePrototype(JSContext* cx, JSProtoKey key)
 static const JSFunctionSpec promise_methods[] = {
     JS_SELF_HOSTED_FN("catch", "Promise_catch", 1, 0),
     JS_FN("then", Promise_then, 2, 0),
+    JS_SELF_HOSTED_FN("finally", "Promise_finally", 1, 0),
     JS_FS_END
 };
 

js/src/builtin/Promise.h

@@ -128,6 +128,14 @@ OriginalPromiseThen(JSContext* cx, Handle<PromiseObject*> promise,
                     HandleValue onFulfilled, HandleValue onRejected,
                     MutableHandleObject dependent, bool createDependent);
 
+/**
+ * PromiseResolve ( C, x )
+ *
+ * The abstract operation PromiseResolve, given a constructor and a value,
+ * returns a new promise resolved with that value.
+ */
+MOZ_MUST_USE JSObject*
+PromiseResolve(JSContext* cx, HandleObject constructor, HandleValue value);
 
 MOZ_MUST_USE PromiseObject*
 CreatePromiseObjectForAsync(JSContext* cx, HandleValue generatorVal);

js/src/builtin/Promise.js

@@ -14,3 +14,72 @@ function Promise_catch(onRejected) {
     // Steps 1-2.
     return callContentFunction(this.then, this, undefined, onRejected);
 }
+
+// Promise.prototype.finally(onFinally)
+// See https://tc39.es/proposal-promise-finally/
+function Promise_finally(onFinally) {
+    // Step 1.
+    var promise = this;
+
+    // Step 2.
+    if (!IsObject(promise))
+        ThrowTypeError(JSMSG_INCOMPATIBLE_PROTO, "Promise", "finally", "value");
+
+    // Step 3.
+    var C = SpeciesConstructor(promise, GetBuiltinConstructor("Promise"));
+
+    // Step 4.
+    assert(IsConstructor(C), "SpeciesConstructor returns a constructor function");
+
+    // Steps 5-6.
+    var thenFinally, catchFinally;
+    if (!IsCallable(onFinally)) {
+        thenFinally = onFinally;
+        catchFinally = onFinally;
+    } else {
+        // ThenFinally Function.
+        // The parentheses prevent the inferring of a function name.
+        (thenFinally) = function(value) {
+            // Steps 1-2 (implicit).
+
+            // Step 3.
+            var result = onFinally();
+
+            // Steps 4-5 (implicit).
+
+            // Step 6.
+            var promise = PromiseResolve(C, result);
+
+            // Step 7.
+            // FIXME: spec issue - "be equivalent to a function that" is not a defined spec term.
+            // https://github.com/tc39/ecma262/issues/933
+
+            // Step 8.
+            return callContentFunction(promise.then, promise, function() { return value; });
+        };
+
+        // CatchFinally Function.
+        // The parentheses prevent the inferring of a function name.
+        (catchFinally) = function(reason) {
+            // Steps 1-2 (implicit).
+
+            // Step 3.
+            var result = onFinally();
+
+            // Steps 4-5 (implicit).
+
+            // Step 6.
+            var promise = PromiseResolve(C, result);
+
+            // Step 7.
+            // FIXME: spec issue - "be equivalent to a function that" is not a defined spec term.
+            // https://github.com/tc39/ecma262/issues/933
+
+            // Step 8.
+            return callContentFunction(promise.then, promise, function() { throw reason; });
+        };
+    }
+
+    // Step 7.
+    return callContentFunction(promise.then, promise, thenFinally, catchFinally);
+}

js/src/builtin/TestingFunctions.cpp

@@ -3900,10 +3900,10 @@ ConvertRegExpTreeToObject(JSContext* cx, irregexp::RegExpTree* tree)
             return nullptr;
         return obj;
     }
-    if (tree->IsLookahead()) {
-        if (!StringProp(cx, obj, "type", "Lookahead"))
+    if (tree->IsLookaround()) {
+        if (!StringProp(cx, obj, "type", "Lookaround"))
             return nullptr;
-        irregexp::RegExpLookahead* t = tree->AsLookahead();
+        irregexp::RegExpLookaround* t = tree->AsLookaround();
         if (!BooleanProp(cx, obj, "is_positive", t->is_positive()))
             return nullptr;
         if (!TreeProp(cx, obj, "body", t->body()))

js/src/irregexp/NativeRegExpMacroAssembler.cpp

@@ -582,7 +582,7 @@ NativeRegExpMacroAssembler::CheckAtStart(Label* on_at_start)
 }
 
 void
-NativeRegExpMacroAssembler::CheckNotAtStart(Label* on_not_at_start)
+NativeRegExpMacroAssembler::CheckNotAtStart(int cp_offset, Label* on_not_at_start)
 {
     JitSpew(SPEW_PREFIX "CheckNotAtStart");
 
@@ -673,7 +673,7 @@ NativeRegExpMacroAssembler::CheckGreedyLoop(Label* on_tos_equals_current_positio
 }
 
 void
-NativeRegExpMacroAssembler::CheckNotBackReference(int start_reg, Label* on_no_match)
+NativeRegExpMacroAssembler::CheckNotBackReference(int start_reg, bool read_backward, Label* on_no_match)
 {
     JitSpew(SPEW_PREFIX "CheckNotBackReference(%d)", start_reg);
 
@@ -744,8 +744,8 @@ NativeRegExpMacroAssembler::CheckNotBackReference(int start_reg, Label* on_no_ma
 }
 
 void
-NativeRegExpMacroAssembler::CheckNotBackReferenceIgnoreCase(int start_reg, Label* on_no_match,
-                                                            bool unicode)
+NativeRegExpMacroAssembler::CheckNotBackReferenceIgnoreCase(int start_reg, bool read_backward,
+                                                            Label* on_no_match, bool unicode)
 {
     JitSpew(SPEW_PREFIX "CheckNotBackReferenceIgnoreCase(%d, %d)", start_reg, unicode);
 

js/src/irregexp/NativeRegExpMacroAssembler.h

@@ -105,9 +105,10 @@ class MOZ_STACK_CLASS NativeRegExpMacroAssembler final : public RegExpMacroAssem
     void CheckCharacterGT(char16_t limit, jit::Label* on_greater);
     void CheckCharacterLT(char16_t limit, jit::Label* on_less);
     void CheckGreedyLoop(jit::Label* on_tos_equals_current_position);
-    void CheckNotAtStart(jit::Label* on_not_at_start);
-    void CheckNotBackReference(int start_reg, jit::Label* on_no_match);
-    void CheckNotBackReferenceIgnoreCase(int start_reg, jit::Label* on_no_match, bool unicode);
+    void CheckNotAtStart(int cp_offset, jit::Label* on_not_at_start);
+    void CheckNotBackReference(int start_reg, bool read_backward, jit::Label* on_no_match);
+    void CheckNotBackReferenceIgnoreCase(int start_reg, bool read_backward,
+                                         jit::Label* on_no_match, bool unicode);
     void CheckNotCharacter(unsigned c, jit::Label* on_not_equal);
     void CheckNotCharacterAfterAnd(unsigned c, unsigned and_with, jit::Label* on_not_equal);
     void CheckNotCharacterAfterMinusAnd(char16_t c, char16_t minus, char16_t and_with,

js/src/irregexp/RegExpAST.cpp

@@ -250,16 +250,16 @@ RegExpCapture::CaptureRegisters()
 }
 
 // ----------------------------------------------------------------------------
-// RegExpLookahead
+// RegExpLookaround
 
 Interval
-RegExpLookahead::CaptureRegisters()
+RegExpLookaround::CaptureRegisters()
 {
     return body()->CaptureRegisters();
 }
 
 bool
-RegExpLookahead::IsAnchoredAtStart()
+RegExpLookaround::IsAnchoredAtStart()
 {
-    return is_positive() && body()->IsAnchoredAtStart();
+    return is_positive() && type() == LOOKAHEAD && body()->IsAnchoredAtStart();
 }

js/src/irregexp/RegExpAST.h

@@ -360,6 +360,7 @@ class RegExpCapture : public RegExpTree
     virtual int min_match() { return body_->min_match(); }
     virtual int max_match() { return body_->max_match(); }
     RegExpTree* body() { return body_; }
+    void set_body(RegExpTree* body) { body_ = body; }
     int index() { return index_; }
     static int StartRegister(int index) { return index * 2; }
     static int EndRegister(int index) { return index * 2 + 1; }
@@ -369,25 +370,29 @@ class RegExpCapture : public RegExpTree
     int index_;
 };
 
-class RegExpLookahead : public RegExpTree
+class RegExpLookaround : public RegExpTree
 {
   public:
-    RegExpLookahead(RegExpTree* body,
-                    bool is_positive,
-                    int capture_count,
-                    int capture_from)
+    enum Type { LOOKAHEAD, LOOKBEHIND };
+
+    RegExpLookaround(RegExpTree* body,
+                     bool is_positive,
+                     int capture_count,
+                     int capture_from,
+                     Type type)
       : body_(body),
         is_positive_(is_positive),
         capture_count_(capture_count),
-        capture_from_(capture_from)
+        capture_from_(capture_from),
+        type_(type)
     {}
 
     virtual void* Accept(RegExpVisitor* visitor, void* data);
     virtual RegExpNode* ToNode(RegExpCompiler* compiler,
                                RegExpNode* on_success);
-    virtual RegExpLookahead* AsLookahead();
+    virtual RegExpLookaround* AsLookaround();
     virtual Interval CaptureRegisters();
-    virtual bool IsLookahead();
+    virtual bool IsLookaround();
     virtual bool IsAnchoredAtStart();
     virtual int min_match() { return 0; }
     virtual int max_match() { return 0; }
@@ -395,12 +400,14 @@ class RegExpLookahead : public RegExpTree
     bool is_positive() { return is_positive_; }
     int capture_count() { return capture_count_; }
     int capture_from() { return capture_from_; }
+    Type type() { return type_; }
 
   private:
     RegExpTree* body_;
     bool is_positive_;
     int capture_count_;
     int capture_from_;
+    Type type_;
 };
 
 typedef InfallibleVector<RegExpCapture*, 1> RegExpCaptureVector;
@@ -417,8 +424,14 @@ class RegExpBackReference : public RegExpTree
                                RegExpNode* on_success);
     virtual RegExpBackReference* AsBackReference();
     virtual bool IsBackReference();
-    virtual int min_match() { return 0; }
-    virtual int max_match() { return capture_->max_match(); }
+    virtual int min_match() override { return 0; }
+    // The capture may not be completely parsed yet, if the reference occurs
+    // before the capture. In the ordinary case, nothing has been captured yet,
+    // so the back reference must have the length 0. If the back reference is
+    // inside a lookbehind, effectively making it a forward reference, we return
+    virtual int max_match() override {
+      return capture_->body() ? capture_->max_match() : 0;
+    }
     int index() { return capture_->index(); }
     RegExpCapture* capture() { return capture_; }
   private:

js/src/irregexp/RegExpBytecode.h

@@ -82,16 +82,19 @@ V(CHECK_LT,          35, 8)   /* bc8 pad8 uc16 addr32                       */ \
 V(CHECK_GT,          36, 8)   /* bc8 pad8 uc16 addr32                       */ \
 V(CHECK_NOT_BACK_REF, 37, 8)  /* bc8 reg_idx24 addr32                       */ \
 V(CHECK_NOT_BACK_REF_NO_CASE, 38, 8) /* bc8 reg_idx24 addr32                */ \
-V(CHECK_NOT_REGS_EQUAL, 39, 12) /* bc8 regidx24 reg_idx32 addr32            */ \
-V(CHECK_REGISTER_LT, 40, 12)  /* bc8 reg_idx24 value32 addr32               */ \
-V(CHECK_REGISTER_GE, 41, 12)  /* bc8 reg_idx24 value32 addr32               */ \
-V(CHECK_REGISTER_EQ_POS, 42, 8) /* bc8 reg_idx24 addr32                     */ \
-V(CHECK_AT_START,    43, 8)   /* bc8 pad24 addr32                           */ \
-V(CHECK_NOT_AT_START, 44, 8)  /* bc8 pad24 addr32                           */ \
-V(CHECK_GREEDY,      45, 8)   /* bc8 pad24 addr32                           */ \
-V(ADVANCE_CP_AND_GOTO, 46, 8) /* bc8 offset24 addr32                        */ \
-V(SET_CURRENT_POSITION_FROM_END, 47, 4) /* bc8 idx24                        */ \
-V(CHECK_NOT_BACK_REF_NO_CASE_UNICODE, 48, 8) /* bc8 reg_idx24 addr32        */
+V(CHECK_NOT_BACK_REF_BACKWARD, 39, 8)         /* bc8 reg_idx24 addr32       */ \
+V(CHECK_NOT_BACK_REF_NO_CASE_BACKWARD, 40, 8) /* bc8 reg_idx24 addr32       */ \
+V(CHECK_NOT_REGS_EQUAL, 41, 12) /* bc8 regidx24 reg_idx32 addr32            */ \
+V(CHECK_REGISTER_LT, 42, 12)  /* bc8 reg_idx24 value32 addr32               */ \
+V(CHECK_REGISTER_GE, 43, 12)  /* bc8 reg_idx24 value32 addr32               */ \
+V(CHECK_REGISTER_EQ_POS, 44, 8) /* bc8 reg_idx24 addr32                     */ \
+V(CHECK_AT_START,    45, 8)   /* bc8 pad24 addr32                           */ \
+V(CHECK_NOT_AT_START, 46, 8)  /* bc8 pad24 addr32                           */ \
+V(CHECK_GREEDY,      47, 8)   /* bc8 pad24 addr32                           */ \
+V(ADVANCE_CP_AND_GOTO, 48, 8) /* bc8 offset24 addr32                        */ \
+V(SET_CURRENT_POSITION_FROM_END, 49, 4) /* bc8 idx24                        */ \
+V(CHECK_NOT_BACK_REF_NO_CASE_UNICODE, 50, 8) /* bc8 reg_idx24 addr32        */ \
+V(CHECK_NOT_BACK_REF_NO_CASE_BACKWARD_UNICODE, 51, 8) /* bc8 reg_idx24 addr32 */
 
 #define DECLARE_BYTECODES(name, code, length) \
   static const int BC_##name = code;

js/src/irregexp/RegExpEngine.cpp

@@ -721,6 +721,8 @@ ActionNode::EmptyMatchCheck(int start_register,
 int
 TextNode::EatsAtLeast(int still_to_find, int budget, bool not_at_start)
 {
+    if (read_backward())
+        return 0;
     int answer = Length();
     if (answer >= still_to_find)
         return answer;
@@ -736,8 +738,7 @@ TextNode::EatsAtLeast(int still_to_find, int budget, bool not_at_start)
 int
 TextNode::GreedyLoopTextLength()
 {
-    TextElement elm = elements()[elements().length() - 1];
-    return elm.cp_offset() + elm.length();
+    return Length();
 }
 
 RegExpNode*
@@ -887,6 +888,8 @@ AssertionNode::FillInBMInfo(int offset, int budget, BoyerMooreLookahead* bm, boo
 int
 BackReferenceNode::EatsAtLeast(int still_to_find, int budget, bool not_at_start)
 {
+    if (read_backward())
+        return 0;
     if (budget <= 0)
         return 0;
     return on_success()->EatsAtLeast(still_to_find, budget - 1, not_at_start);
@@ -1578,6 +1581,9 @@ class irregexp::RegExpCompiler
         current_expansion_factor_ = value;
     }
 
+    bool read_backward() { return read_backward_; }
+    void set_read_backward(bool value) { read_backward_ = value; }
+
     JSContext* cx() const { return cx_; }
     LifoAlloc* alloc() const { return alloc_; }
 
@@ -1595,6 +1601,7 @@ class irregexp::RegExpCompiler
     bool unicode_;
     bool reg_exp_too_big_;
     int current_expansion_factor_;
+    bool read_backward_;
     FrequencyCollator frequency_collator_;
     JSContext* cx_;
     LifoAlloc* alloc_;
@@ -1624,6 +1631,7 @@ RegExpCompiler::RegExpCompiler(JSContext* cx, LifoAlloc* alloc, int capture_coun
     unicode_(unicode),
     reg_exp_too_big_(false),
     current_expansion_factor_(1),
+    read_backward_(false),
     frequency_collator_(),
     cx_(cx),
     alloc_(alloc)
@@ -1747,7 +1755,7 @@ irregexp::CompilePattern(JSContext* cx, RegExpShared* shared, RegExpCompileData*
             // at the start of input.
             ChoiceNode* first_step_node = alloc.newInfallible<ChoiceNode>(&alloc, 2);
             RegExpNode* char_class =
-                alloc.newInfallible<TextNode>(alloc.newInfallible<RegExpCharacterClass>('*'), loop_node);
+                alloc.newInfallible<TextNode>(alloc.newInfallible<RegExpCharacterClass>('*'), false, loop_node);
             first_step_node->AddAlternative(GuardedAlternative(captured_body));
             first_step_node->AddAlternative(GuardedAlternative(char_class));
             node = first_step_node;
@@ -1850,19 +1858,19 @@ RegExpAtom::ToNode(RegExpCompiler* compiler, RegExpNode* on_success)
     TextElementVector* elms =
         compiler->alloc()->newInfallible<TextElementVector>(*compiler->alloc());
     elms->append(TextElement::Atom(this));
-    return compiler->alloc()->newInfallible<TextNode>(elms, on_success);
+    return compiler->alloc()->newInfallible<TextNode>(elms, compiler->read_backward(), on_success);
 }
 
 RegExpNode*
 RegExpText::ToNode(RegExpCompiler* compiler, RegExpNode* on_success)
 {
-    return compiler->alloc()->newInfallible<TextNode>(&elements_, on_success);
+    return compiler->alloc()->newInfallible<TextNode>(&elements_, compiler->read_backward(), on_success);
 }
 
 RegExpNode*
 RegExpCharacterClass::ToNode(RegExpCompiler* compiler, RegExpNode* on_success)
 {
-    return compiler->alloc()->newInfallible<TextNode>(this, on_success);
+    return compiler->alloc()->newInfallible<TextNode>(this, compiler->read_backward(), on_success);
 }
 
 RegExpNode*
@@ -2003,7 +2011,8 @@ RegExpQuantifier::ToNode(int min,
                         alternation->AddAlternative(GuardedAlternative(body->ToNode(compiler, answer)));
                     }
                     answer = alternation;
-                    if (not_at_start) alternation->set_not_at_start();
+                    if (not_at_start && !compiler->read_backward())
+                        alternation->set_not_at_start();
                 }
                 return answer;
             }
@@ -2015,8 +2024,9 @@ RegExpQuantifier::ToNode(int min,
     int reg_ctr = needs_counter
         ? compiler->AllocateRegister()
         : RegExpCompiler::kNoRegister;
-    LoopChoiceNode* center = alloc->newInfallible<LoopChoiceNode>(alloc, body->min_match() == 0);
-    if (not_at_start)
+    LoopChoiceNode* center = alloc->newInfallible<LoopChoiceNode>(alloc, body->min_match() == 0,
+                                                                  compiler->read_backward());
+    if (not_at_start && !compiler->read_backward())
         center->set_not_at_start();
     RegExpNode* loop_return = needs_counter
         ? static_cast<RegExpNode*>(ActionNode::IncrementRegister(reg_ctr, center))
@@ -2092,7 +2102,7 @@ RegExpAssertion::ToNode(RegExpCompiler* compiler,
         CharacterRange::AddClassEscape(alloc, 'n', newline_ranges);
         RegExpCharacterClass* newline_atom = alloc->newInfallible<RegExpCharacterClass>('n');
         TextNode* newline_matcher =
-            alloc->newInfallible<TextNode>(newline_atom,
+            alloc->newInfallible<TextNode>(newline_atom, false,
                 ActionNode::PositiveSubmatchSuccess(stack_pointer_register,
                                                     position_register,
                                                     0,  // No captures inside.
@@ -2124,6 +2134,7 @@ RegExpBackReference::ToNode(RegExpCompiler* compiler, RegExpNode* on_success)
 {
     return compiler->alloc()->newInfallible<BackReferenceNode>(RegExpCapture::StartRegister(index()),
                                                                RegExpCapture::EndRegister(index()),
+                                                               compiler->read_backward(),
                                                                on_success);
 }
 
@@ -2134,7 +2145,7 @@ RegExpEmpty::ToNode(RegExpCompiler* compiler, RegExpNode* on_success)
 }
 
 RegExpNode*
-RegExpLookahead::ToNode(RegExpCompiler* compiler, RegExpNode* on_success)
+RegExpLookaround::ToNode(RegExpCompiler* compiler, RegExpNode* on_success)
 {
     int stack_pointer_register = compiler->AllocateRegister();
     int position_register = compiler->AllocateRegister();
@@ -2145,6 +2156,10 @@ RegExpLookahead::ToNode(RegExpCompiler* compiler, RegExpNode* on_success)
     int register_start =
         register_of_first_capture + capture_from_ * registers_per_capture;
 
+    RegExpNode* result;
+    bool was_reading_backward = compiler->read_backward();
+    compiler->set_read_backward(type() == LOOKBEHIND);
+
     if (is_positive()) {
         RegExpNode* bodyNode =
             body()->ToNode(compiler,
@@ -2153,37 +2168,39 @@ RegExpLookahead::ToNode(RegExpCompiler* compiler, RegExpNode* on_success)
                                                                register_count,
                                                                register_start,
                                                                on_success));
-        return ActionNode::BeginSubmatch(stack_pointer_register,
+        result = ActionNode::BeginSubmatch(stack_pointer_register,
+                                           position_register,
+                                           bodyNode);
+    } else {
+        // We use a ChoiceNode for a negative lookahead because it has most of
+        // the characteristics we need. It has the body of the lookahead as its
+        // first alternative and the expression after the lookahead of the second
+        // alternative. If the first alternative succeeds then the
+        // NegativeSubmatchSuccess will unwind the stack including everything the
+        // choice node set up and backtrack. If the first alternative fails then
+        // the second alternative is tried, which is exactly the desired result
+        // for a negative lookahead. The NegativeLookaheadChoiceNode is a special
+        // ChoiceNode that knows to ignore the first exit when calculating quick
+        // checks.
+        LifoAlloc* alloc = compiler->alloc();
+
+        RegExpNode* success =
+            alloc->newInfallible<NegativeSubmatchSuccess>(alloc,
+                                                          stack_pointer_register,
+                                                          position_register,
+                                                          register_count,
+                                                          register_start);
+        GuardedAlternative body_alt(body()->ToNode(compiler, success));
+
+        ChoiceNode* choice_node =
+            alloc->newInfallible<NegativeLookaheadChoiceNode>(alloc, body_alt, GuardedAlternative(on_success));
+
+        result = ActionNode::BeginSubmatch(stack_pointer_register,
                                          position_register,
-                                         bodyNode);
+                                         choice_node);
     }
-
-    // We use a ChoiceNode for a negative lookahead because it has most of
-    // the characteristics we need.  It has the body of the lookahead as its
-    // first alternative and the expression after the lookahead of the second
-    // alternative.  If the first alternative succeeds then the
-    // NegativeSubmatchSuccess will unwind the stack including everything the
-    // choice node set up and backtrack.  If the first alternative fails then
-    // the second alternative is tried, which is exactly the desired result
-    // for a negative lookahead.  The NegativeLookaheadChoiceNode is a special
-    // ChoiceNode that knows to ignore the first exit when calculating quick
-    // checks.
-    LifoAlloc* alloc = compiler->alloc();
-
-    RegExpNode* success =
-        alloc->newInfallible<NegativeSubmatchSuccess>(alloc,
-                                                      stack_pointer_register,
-                                                      position_register,
-                                                      register_count,
-                                                      register_start);
-    GuardedAlternative body_alt(body()->ToNode(compiler, success));
-
-    ChoiceNode* choice_node =
-        alloc->newInfallible<NegativeLookaheadChoiceNode>(alloc, body_alt, GuardedAlternative(on_success));
-
-    return ActionNode::BeginSubmatch(stack_pointer_register,
-                                     position_register,
-                                     choice_node);
+    compiler->set_read_backward(was_reading_backward);
+    return result;
 }
 
 RegExpNode*
@@ -2198,8 +2215,14 @@ RegExpCapture::ToNode(RegExpTree* body,
                       RegExpCompiler* compiler,
                       RegExpNode* on_success)
 {
+    MOZ_ASSERT(body);
     int start_reg = RegExpCapture::StartRegister(index);
     int end_reg = RegExpCapture::EndRegister(index);
+    if (compiler->read_backward()) {
+        // std::swap(start_reg, end_reg);
+        start_reg = RegExpCapture::EndRegister(index);
+        end_reg = RegExpCapture::StartRegister(index);
+    }
     RegExpNode* store_end = ActionNode::StorePosition(end_reg, true, on_success);
     RegExpNode* body_node = body->ToNode(compiler, store_end);
     return ActionNode::StorePosition(start_reg, true, body_node);
@@ -2210,8 +2233,15 @@ RegExpAlternative::ToNode(RegExpCompiler* compiler, RegExpNode* on_success)
 {
     const RegExpTreeVector& children = nodes();
     RegExpNode* current = on_success;
-    for (int i = children.length() - 1; i >= 0; i--)
-        current = children[i]->ToNode(compiler, current);
+    if (compiler->read_backward()) {
+        for (int i = 0; i < children.length(); i++) {
+            current = children[i]->ToNode(compiler, current);
+        }
+    } else {
+        for (int i = children.length() - 1; i >= 0; i--) {
+            current = children[i]->ToNode(compiler, current);
+        }
+    }
     return current;
 }
 
@@ -2764,7 +2794,6 @@ Trace::InvalidateCurrentCharacter()
 void
 Trace::AdvanceCurrentPositionInTrace(int by, RegExpCompiler* compiler)
 {
-    MOZ_ASSERT(by > 0);
     // We don't have an instruction for shifting the current character register
     // down or for using a shifted value for anything so lets just forget that
     // we preloaded any characters into it.
@@ -3109,9 +3138,9 @@ AssertionNode::Emit(RegExpCompiler* compiler, Trace* trace)
             return;
         }
         if (trace->at_start() == Trace::UNKNOWN) {
-            assembler->CheckNotAtStart(trace->backtrack());
+            assembler->CheckNotAtStart(trace->cp_offset(), trace->backtrack());
             Trace at_start_trace = *trace;
-            at_start_trace.set_at_start(true);
+            at_start_trace.set_at_start(Trace::TRUE_VALUE);
             on_success()->Emit(compiler, &at_start_trace);
             return;
         }
@@ -3814,9 +3843,10 @@ TextNode::TextEmitPass(RegExpCompiler* compiler,
     jit::Label* backtrack = trace->backtrack();
     QuickCheckDetails* quick_check = trace->quick_check_performed();
     int element_count = elements().length();
+    int backward_offset = read_backward() ? -Length() : 0;
     for (int i = preloaded ? 0 : element_count - 1; i >= 0; i--) {
         TextElement elm = elements()[i];
-        int cp_offset = trace->cp_offset() + elm.cp_offset();
+        int cp_offset = trace->cp_offset() + elm.cp_offset() + backward_offset;
         if (elm.text_type() == TextElement::ATOM) {
             const CharacterVector& quarks = elm.atom()->data();
             for (int j = preloaded ? 0 : quarks.length() - 1; j >= 0; j--) {
@@ -3844,11 +3874,12 @@ TextNode::TextEmitPass(RegExpCompiler* compiler,
                     break;
                 }
                 if (emit_function != nullptr) {
+                    bool bounds_check = *checked_up_to < cp_offset + j || read_backward();
                     bool bound_checked = emit_function(compiler,
                                                        quarks[j],
                                                        backtrack,
                                                        cp_offset + j,
-                                                       *checked_up_to < cp_offset + j,
+                                                       bounds_check,
                                                        preloaded);
                     if (bound_checked) UpdateBoundsCheck(cp_offset + j, checked_up_to);
                 }
@@ -3859,13 +3890,14 @@ TextNode::TextEmitPass(RegExpCompiler* compiler,
                 if (first_element_checked && i == 0) continue;
                 if (DeterminedAlready(quick_check, elm.cp_offset())) continue;
                 RegExpCharacterClass* cc = elm.char_class();
+                bool bounds_check = *checked_up_to < cp_offset || read_backward();
                 EmitCharClass(alloc(),
                               assembler,
                               cc,
                               ascii,
                               backtrack,
                               cp_offset,
-                              *checked_up_to < cp_offset,
+                              bounds_check,
                               preloaded);
                 UpdateBoundsCheck(cp_offset, checked_up_to);
             }
@@ -3945,8 +3977,11 @@ TextNode::Emit(RegExpCompiler* compiler, Trace* trace)
     }
 
     Trace successor_trace(*trace);
-    successor_trace.set_at_start(false);
-    successor_trace.AdvanceCurrentPositionInTrace(Length(), compiler);
+    // If we advance backward, we may end up at the start.
+    successor_trace.AdvanceCurrentPositionInTrace(
+        read_backward() ? -Length() : Length(), compiler);
+    successor_trace.set_at_start(read_backward() ? Trace::UNKNOWN
+                                                 : Trace::FALSE_VALUE);
     RecursionCheck rc(compiler);
     on_success()->Emit(compiler, &successor_trace);
 }
@@ -4118,6 +4153,8 @@ ChoiceNode::CalculatePreloadCharacters(RegExpCompiler* compiler, int eats_at_lea
 RegExpNode*
 TextNode::GetSuccessorOfOmnivorousTextNode(RegExpCompiler* compiler)
 {
+    if (read_backward()) return NULL;
+
     if (elements().length() != 1)
         return nullptr;
 
@@ -4165,7 +4202,7 @@ ChoiceNode::GreedyLoopTextLengthForAlternative(GuardedAlternative* alternative)
         SeqRegExpNode* seq_node = static_cast<SeqRegExpNode*>(node);
         node = seq_node->on_success();
     }
-    return length;
+    return read_backward() ? -length : length;
 }
 
 // Creates a list of AlternativeGenerations.  If the list has a reasonable
@@ -4240,7 +4277,7 @@ ChoiceNode::Emit(RegExpCompiler* compiler, Trace* trace)
     jit::Label greedy_loop_label;
     Trace counter_backtrack_trace;
     counter_backtrack_trace.set_backtrack(&greedy_loop_label);
-    if (not_at_start()) counter_backtrack_trace.set_at_start(false);
+    if (not_at_start()) counter_backtrack_trace.set_at_start(Trace::FALSE_VALUE);
 
     if (choice_count > 1 && text_length != kNodeIsTooComplexForGreedyLoops) {
         // Here we have special handling for greedy loops containing only text nodes
@@ -4256,7 +4293,7 @@ ChoiceNode::Emit(RegExpCompiler* compiler, Trace* trace)
         current_trace = &counter_backtrack_trace;
         jit::Label greedy_match_failed;
         Trace greedy_match_trace;
-        if (not_at_start()) greedy_match_trace.set_at_start(false);
+        if (not_at_start()) greedy_match_trace.set_at_start(Trace::FALSE_VALUE);
         greedy_match_trace.set_backtrack(&greedy_match_failed);
         jit::Label loop_label;
         macro_assembler->Bind(&loop_label);
@@ -4605,11 +4642,14 @@ BackReferenceNode::Emit(RegExpCompiler* compiler, Trace* trace)
     MOZ_ASSERT(start_reg_ + 1 == end_reg_);
     if (compiler->ignore_case()) {
         assembler->CheckNotBackReferenceIgnoreCase(start_reg_,
+                                                   read_backward(),
                                                    trace->backtrack(),
                                                    compiler->unicode());
     } else {
-        assembler->CheckNotBackReference(start_reg_, trace->backtrack());
+        assembler->CheckNotBackReference(start_reg_, read_backward(), trace->backtrack());
     }
+    // We are going to advance backward, so we may end up at the start.
+    if (read_backward()) trace->set_at_start(Trace::UNKNOWN);
     on_success()->Emit(compiler, trace);
 }
 
@@ -4977,7 +5017,6 @@ QuickCheckDetails::Clear()
 void
 QuickCheckDetails::Advance(int by, bool ascii)
 {
-    MOZ_ASSERT(by >= 0);
     if (by >= characters_) {
         Clear();
         return;

js/src/irregexp/RegExpEngine.h

@@ -119,7 +119,7 @@ InterpretCode(JSContext* cx, const uint8_t* byteCode, const CharT* chars, size_t
   VISIT(Atom)                                                        \
   VISIT(Quantifier)                                                  \
   VISIT(Capture)                                                     \
-  VISIT(Lookahead)                                                   \
+  VISIT(Lookaround)                                                  \
   VISIT(BackReference)                                               \
   VISIT(Empty)                                                       \
   VISIT(Text)
@@ -763,15 +763,19 @@ class TextNode : public SeqRegExpNode
 {
   public:
     TextNode(TextElementVector* elements,
+             bool read_backward,
              RegExpNode* on_success)
       : SeqRegExpNode(on_success),
-        elements_(elements)
+        elements_(elements),
+        read_backward_(read_backward)
     {}
 
     TextNode(RegExpCharacterClass* that,
+             bool read_backward,
              RegExpNode* on_success)
       : SeqRegExpNode(on_success),
-        elements_(alloc()->newInfallible<TextElementVector>(*alloc()))
+        elements_(alloc()->newInfallible<TextElementVector>(*alloc())),
+        read_backward_(read_backward)
     {
         elements_->append(TextElement::CharClass(that));
     }
@@ -784,6 +788,7 @@ class TextNode : public SeqRegExpNode
                                       int characters_filled_in,
                                       bool not_at_start);
     TextElementVector& elements() { return *elements_; }
+    bool read_backward() { return read_backward_; }
     void MakeCaseIndependent(bool is_ascii, bool unicode);
     virtual int GreedyLoopTextLength();
     virtual RegExpNode* GetSuccessorOfOmnivorousTextNode(
@@ -814,6 +819,7 @@ class TextNode : public SeqRegExpNode
                       int* checked_up_to);
     int Length();
     TextElementVector* elements_;
+    bool read_backward_;
 };
 
 class AssertionNode : public SeqRegExpNode
@@ -882,15 +888,18 @@ class BackReferenceNode : public SeqRegExpNode
   public:
     BackReferenceNode(int start_reg,
                       int end_reg,
+                      bool read_backward,
                       RegExpNode* on_success)
       : SeqRegExpNode(on_success),
         start_reg_(start_reg),
-        end_reg_(end_reg)
+        end_reg_(end_reg),
+        read_backward_(read_backward)
     {}
 
     virtual void Accept(NodeVisitor* visitor);
     int start_register() { return start_reg_; }
     int end_register() { return end_reg_; }
+    bool read_backward() { return read_backward_; }
     virtual void Emit(RegExpCompiler* compiler, Trace* trace);
     virtual int EatsAtLeast(int still_to_find,
                             int recursion_depth,
@@ -909,6 +918,7 @@ class BackReferenceNode : public SeqRegExpNode
   private:
     int start_reg_;
     int end_reg_;
+    bool read_backward_;
 };
 
 class EndNode : public RegExpNode
@@ -1053,6 +1063,7 @@ class ChoiceNode : public RegExpNode
     void set_being_calculated(bool b) { being_calculated_ = b; }
     virtual bool try_to_emit_quick_check_for_alternative(int i) { return true; }
     virtual RegExpNode* FilterASCII(int depth, bool ignore_case, bool unicode);
+    virtual bool read_backward() { return false; }
 
   protected:
     int GreedyLoopTextLengthForAlternative(GuardedAlternative* alternative);
@@ -1111,11 +1122,13 @@ class NegativeLookaheadChoiceNode : public ChoiceNode
 class LoopChoiceNode : public ChoiceNode
 {
   public:
-    explicit LoopChoiceNode(LifoAlloc* alloc, bool body_can_be_zero_length)
+    explicit LoopChoiceNode(LifoAlloc* alloc, bool body_can_be_zero_length,
+                            bool read_backward)
       : ChoiceNode(alloc, 2),
         loop_node_(nullptr),
         continue_node_(nullptr),
-        body_can_be_zero_length_(body_can_be_zero_length)
+        body_can_be_zero_length_(body_can_be_zero_length),
+        read_backward_(read_backward)
     {}
 
     void AddLoopAlternative(GuardedAlternative alt);
@@ -1133,6 +1146,7 @@ class LoopChoiceNode : public ChoiceNode
     RegExpNode* loop_node() { return loop_node_; }
     RegExpNode* continue_node() { return continue_node_; }
     bool body_can_be_zero_length() { return body_can_be_zero_length_; }
+    virtual bool read_backward() { return read_backward_; }
     virtual void Accept(NodeVisitor* visitor);
     virtual RegExpNode* FilterASCII(int depth, bool ignore_case, bool unicode);
 
@@ -1147,6 +1161,7 @@ class LoopChoiceNode : public ChoiceNode
     RegExpNode* loop_node_;
     RegExpNode* continue_node_;
     bool body_can_be_zero_length_;
+    bool read_backward_;
 };
 
 // Improve the speed that we scan for an initial point where a non-anchored
@@ -1422,8 +1437,8 @@ class Trace
     }
 
     TriBool at_start() { return at_start_; }
-    void set_at_start(bool at_start) {
-        at_start_ = at_start ? TRUE_VALUE : FALSE_VALUE;
+    void set_at_start(TriBool at_start) {
+        at_start_ = at_start;
     }
     jit::Label* backtrack() { return backtrack_; }
     jit::Label* loop_label() { return loop_label_; }

js/src/irregexp/RegExpInterpreter.cpp

@@ -222,8 +222,8 @@ irregexp::InterpretCode(JSContext* cx, const uint8_t* byteCode, const CharT* cha
             }
             break;
           BYTECODE(LOAD_CURRENT_CHAR) {
-            size_t pos = current + (insn >> BYTECODE_SHIFT);
-            if (pos >= length) {
+            int pos = current + (insn >> BYTECODE_SHIFT);
+            if (pos >= (int)length || pos < 0) {
                 pc = byteCode + Load32Aligned(pc + 4);
             } else {
                 current_char = chars[pos];
@@ -238,8 +238,8 @@ irregexp::InterpretCode(JSContext* cx, const uint8_t* byteCode, const CharT* cha
             break;
           }
           BYTECODE(LOAD_2_CURRENT_CHARS) {
-            size_t pos = current + (insn >> BYTECODE_SHIFT);
-            if (pos + 2 > length) {
+            int pos = current + (insn >> BYTECODE_SHIFT);
+            if (pos + 2 > (int)length || pos < 0) {
                 pc = byteCode + Load32Aligned(pc + 4);
             } else {
                 CharT next = chars[pos + 1];
@@ -425,6 +425,30 @@ irregexp::InterpretCode(JSContext* cx, const uint8_t* byteCode, const CharT* cha
             pc += BC_CHECK_NOT_BACK_REF_LENGTH;
             break;
           }
+          BYTECODE(CHECK_NOT_BACK_REF_BACKWARD) {
+            int from = registers[insn >> BYTECODE_SHIFT];
+            int len = registers[(insn >> BYTECODE_SHIFT) + 1] - from;
+            if (from < 0 || len <= 0) {
+                pc += BC_CHECK_NOT_BACK_REF_BACKWARD_LENGTH;
+                break;
+            }
+            if (int(current) - len < 0) {
+                pc = byteCode + Load32Aligned(pc + 4);
+                break;
+            } else {
+                int i;
+                for (i = 0; i < len; i++) {
+                    if (chars[from + i] != chars[int(current) - len + i]) {
+                        pc = byteCode + Load32Aligned(pc + 4);
+                        break;
+                    }
+                }
+                if (i < len) break;
+                current -= len;
+            }
+            pc += BC_CHECK_NOT_BACK_REF_BACKWARD_LENGTH;
+            break;
+          }
           BYTECODE(CHECK_NOT_BACK_REF_NO_CASE) {
             int from = registers[insn >> BYTECODE_SHIFT];
             int len = registers[(insn >> BYTECODE_SHIFT) + 1] - from;
@@ -465,6 +489,46 @@ irregexp::InterpretCode(JSContext* cx, const uint8_t* byteCode, const CharT* cha
             }
             break;
           }
+          BYTECODE(CHECK_NOT_BACK_REF_NO_CASE_BACKWARD) {
+            int from = registers[insn >> BYTECODE_SHIFT];
+            int len = registers[(insn >> BYTECODE_SHIFT) + 1] - from;
+            if (from < 0 || len <= 0) {
+                pc += BC_CHECK_NOT_BACK_REF_NO_CASE_BACKWARD_LENGTH;
+                break;
+            }
+            if (int(current) - len < 0) {
+                pc = byteCode + Load32Aligned(pc + 4);
+                break;
+            }
+            if (CaseInsensitiveCompareStrings(chars + from, chars + int(current) - len, len * sizeof(CharT))) {
+                current -= len;
+                pc += BC_CHECK_NOT_BACK_REF_NO_CASE_BACKWARD_LENGTH;
+            } else {
+                pc = byteCode + Load32Aligned(pc + 4);
+            }
+            break;
+
+          }
+          BYTECODE(CHECK_NOT_BACK_REF_NO_CASE_BACKWARD_UNICODE) {
+            int from = registers[insn >> BYTECODE_SHIFT];
+            int len = registers[(insn >> BYTECODE_SHIFT) + 1] - from;
+            if (from < 0 || len <= 0) {
+                pc += BC_CHECK_NOT_BACK_REF_NO_CASE_BACKWARD_LENGTH;
+                break;
+            }
+            if (int(current) - len < 0) {
+                pc = byteCode + Load32Aligned(pc + 4);
+                break;
+            }
+            if (CaseInsensitiveCompareUCStrings(chars + from, chars + int(current) - len, len * sizeof(CharT))) {
+                current -= len;
+                pc += BC_CHECK_NOT_BACK_REF_NO_CASE_BACKWARD_LENGTH;
+            } else {
+                pc = byteCode + Load32Aligned(pc + 4);
+            }
+            break;
+
+          }
           BYTECODE(CHECK_AT_START)
             if (current == 0)
                 pc = byteCode + Load32Aligned(pc + 4);
@@ -472,7 +536,7 @@ irregexp::InterpretCode(JSContext* cx, const uint8_t* byteCode, const CharT* cha
                 pc += BC_CHECK_AT_START_LENGTH;
             break;
           BYTECODE(CHECK_NOT_AT_START)
-            if (current == 0)
+            if (current + (insn >> BYTECODE_SHIFT) == 0)
                 pc += BC_CHECK_NOT_AT_START_LENGTH;
             else
                 pc = byteCode + Load32Aligned(pc + 4);

js/src/irregexp/RegExpMacroAssembler.cpp

@@ -226,32 +226,37 @@ InterpretedRegExpMacroAssembler::CheckGreedyLoop(jit::Label* on_tos_equals_curre
 }
 
 void
-InterpretedRegExpMacroAssembler::CheckNotAtStart(jit::Label* on_not_at_start)
+InterpretedRegExpMacroAssembler::CheckNotAtStart(int cp_offset, jit::Label* on_not_at_start)
 {
-    Emit(BC_CHECK_NOT_AT_START, 0);
+    Emit(BC_CHECK_NOT_AT_START, cp_offset);
     EmitOrLink(on_not_at_start);
 }
 
 void
-InterpretedRegExpMacroAssembler::CheckNotBackReference(int start_reg, jit::Label* on_no_match)
+InterpretedRegExpMacroAssembler::CheckNotBackReference(int start_reg, bool read_backward,
+                                                       jit::Label* on_no_match)
 {
     MOZ_ASSERT(start_reg >= 0);
     MOZ_ASSERT(start_reg <= kMaxRegister);
-    Emit(BC_CHECK_NOT_BACK_REF, start_reg);
+    Emit(read_backward ? BC_CHECK_NOT_BACK_REF_BACKWARD : BC_CHECK_NOT_BACK_REF,
+         start_reg);
     EmitOrLink(on_no_match);
 }
 
 void
 InterpretedRegExpMacroAssembler::CheckNotBackReferenceIgnoreCase(int start_reg,
+                                                                 bool read_backward,
                                                                  jit::Label* on_no_match,
                                                                  bool unicode)
 {
     MOZ_ASSERT(start_reg >= 0);
     MOZ_ASSERT(start_reg <= kMaxRegister);
     if (unicode)
-        Emit(BC_CHECK_NOT_BACK_REF_NO_CASE_UNICODE, start_reg);
+        Emit(read_backward ? BC_CHECK_NOT_BACK_REF_NO_CASE_BACKWARD_UNICODE : BC_CHECK_NOT_BACK_REF_NO_CASE_UNICODE,
+             start_reg);
     else
-        Emit(BC_CHECK_NOT_BACK_REF_NO_CASE, start_reg);
+        Emit(read_backward ? BC_CHECK_NOT_BACK_REF_NO_CASE_BACKWARD : BC_CHECK_NOT_BACK_REF_NO_CASE,
+             start_reg);
     EmitOrLink(on_no_match);
 }
 

js/src/irregexp/RegExpMacroAssembler.h

@@ -110,10 +110,10 @@ class MOZ_STACK_CLASS RegExpMacroAssembler
     virtual void CheckCharacterGT(char16_t limit, jit::Label* on_greater) = 0;
     virtual void CheckCharacterLT(char16_t limit, jit::Label* on_less) = 0;
     virtual void CheckGreedyLoop(jit::Label* on_tos_equals_current_position) = 0;
-    virtual void CheckNotAtStart(jit::Label* on_not_at_start) = 0;
-    virtual void CheckNotBackReference(int start_reg, jit::Label* on_no_match) = 0;
-    virtual void CheckNotBackReferenceIgnoreCase(int start_reg, jit::Label* on_no_match,
-                                                 bool unicode) = 0;
+    virtual void CheckNotAtStart(int cp_offset, jit::Label* on_not_at_start) = 0;
+    virtual void CheckNotBackReference(int start_reg, bool read_backward, jit::Label* on_no_match) = 0;
+    virtual void CheckNotBackReferenceIgnoreCase(int start_reg, bool read_backward,
+                                                 jit::Label* on_no_match, bool unicode) = 0;
 
     // Check the current character for a match with a literal character.  If we
     // fail to match then goto the on_failure label.  End of input always
@@ -245,9 +245,10 @@ class MOZ_STACK_CLASS InterpretedRegExpMacroAssembler final : public RegExpMacro
     void CheckCharacterGT(char16_t limit, jit::Label* on_greater);
     void CheckCharacterLT(char16_t limit, jit::Label* on_less);
     void CheckGreedyLoop(jit::Label* on_tos_equals_current_position);
-    void CheckNotAtStart(jit::Label* on_not_at_start);
-    void CheckNotBackReference(int start_reg, jit::Label* on_no_match);
-    void CheckNotBackReferenceIgnoreCase(int start_reg, jit::Label* on_no_match, bool unicode);
+    void CheckNotAtStart(int cp_offset, jit::Label* on_not_at_start);
+    void CheckNotBackReference(int start_reg, bool read_backward, jit::Label* on_no_match);
+    void CheckNotBackReferenceIgnoreCase(int start_reg, bool read_backward,
+                                         jit::Label* on_no_match, bool unicode);
     void CheckNotCharacter(unsigned c, jit::Label* on_not_equal);
     void CheckNotCharacterAfterAnd(unsigned c, unsigned and_with, jit::Label* on_not_equal);
     void CheckNotCharacterAfterMinusAnd(char16_t c, char16_t minus, char16_t and_with,

js/src/irregexp/RegExpParser.cpp

@@ -227,6 +227,7 @@ RegExpParser<CharT>::RegExpParser(frontend::TokenStream& ts, LifoAlloc* alloc,
     alloc(alloc),
     captures_(nullptr),
     next_pos_(chars),
+    captures_started_(0),
     end_(end),
     current_(kEndMarker),
     capture_count_(0),
@@ -418,7 +419,8 @@ RangeAtom(LifoAlloc* alloc, char16_t from, char16_t to)
 static inline RegExpTree*
 NegativeLookahead(LifoAlloc* alloc, char16_t from, char16_t to)
 {
-    return alloc->newInfallible<RegExpLookahead>(RangeAtom(alloc, from, to), false, 0, 0);
+    return alloc->newInfallible<RegExpLookaround>(RangeAtom(alloc, from, to), false,
+                                                  0, 0, RegExpLookaround::LOOKAHEAD);
 }
 
 static bool
@@ -1213,6 +1215,38 @@ RegExpParser<CharT>::ParseBackReferenceIndex(int* index_out)
     return true;
 }
 
+template <typename CharT>
+RegExpCapture*
+RegExpParser<CharT>::GetCapture(int index) {
+  // The index for the capture groups are one-based. Its index in the list is
+  // zero-based.
+  int known_captures =
+      is_scanned_for_captures_ ? capture_count_ : captures_started_;
+  MOZ_ASSERT(index <= known_captures);
+  if (captures_ == NULL) {
+    captures_ = alloc->newInfallible<RegExpCaptureVector>(*alloc);
+  }
+  while ((int)captures_->length() < known_captures) {
+    RegExpCapture* capture = alloc->newInfallible<RegExpCapture>(nullptr, captures_->length() + 1);
+    captures_->append(capture);
+  }
+  return (*captures_)[index - 1];
+}
+
+
+template <typename CharT>
+bool
+RegExpParser<CharT>::RegExpParserState::IsInsideCaptureGroup(int index) {
+  for (RegExpParserState* s = this; s != NULL; s = s->previous_state()) {
+    if (s->group_type() != CAPTURE) continue;
+    // Return true if we found the matching capture index.
+    if (index == s->capture_index()) return true;
+    // Abort if index is larger than what has been parsed up till this state.
+    if (index > s->capture_index()) return false;
+  }
+  return false;
+}
+
 // QuantifierPrefix ::
 //   { DecimalDigits }
 //   { DecimalDigits , }
@@ -1423,24 +1457,24 @@ RegExpTree*
 RegExpParser<CharT>::ParseDisjunction()
 {
     // Used to store current state while parsing subexpressions.
-    RegExpParserState initial_state(alloc, nullptr, INITIAL, 0);
-    RegExpParserState* stored_state = &initial_state;
+    RegExpParserState initial_state(alloc, nullptr, INITIAL, RegExpLookaround::LOOKAHEAD, 0);
+    RegExpParserState* state = &initial_state;
     // Cache the builder in a local variable for quick access.
     RegExpBuilder* builder = initial_state.builder();
     while (true) {
         switch (current()) {
           case kEndMarker:
-            if (stored_state->IsSubexpression()) {
+            if (state->IsSubexpression()) {
                 // Inside a parenthesized group when hitting end of input.
                 return ReportError(JSMSG_MISSING_PAREN);
             }
-            MOZ_ASSERT(INITIAL == stored_state->group_type());
+            MOZ_ASSERT(INITIAL == state->group_type());
             // Parsing completed successfully.
             return builder->ToRegExp();
           case ')': {
-            if (!stored_state->IsSubexpression())
+            if (!state->IsSubexpression())
                 return ReportError(JSMSG_UNMATCHED_RIGHT_PAREN);
-            MOZ_ASSERT(INITIAL != stored_state->group_type());
+            MOZ_ASSERT(INITIAL != state->group_type());
 
             Advance();
             // End disjunction parsing and convert builder content to new single
@@ -1449,29 +1483,30 @@ RegExpParser<CharT>::ParseDisjunction()
 
             int end_capture_index = captures_started();
 
-            int capture_index = stored_state->capture_index();
-            SubexpressionType group_type = stored_state->group_type();
-
-            // Restore previous state.
-            stored_state = stored_state->previous_state();
-            builder = stored_state->builder();
+            int capture_index = state->capture_index();
+            SubexpressionType group_type = state->group_type();
 
             // Build result of subexpression.
             if (group_type == CAPTURE) {
-                RegExpCapture* capture = alloc->newInfallible<RegExpCapture>(body, capture_index);
-                (*captures_)[capture_index - 1] = capture;
+                RegExpCapture* capture = GetCapture(capture_index);
+                capture->set_body(body);
                 body = capture;
             } else if (group_type != GROUPING) {
-                MOZ_ASSERT(group_type == POSITIVE_LOOKAHEAD ||
-                           group_type == NEGATIVE_LOOKAHEAD);
-                bool is_positive = (group_type == POSITIVE_LOOKAHEAD);
-                body = alloc->newInfallible<RegExpLookahead>(body,
+                MOZ_ASSERT(group_type == POSITIVE_LOOKAROUND ||
+                           group_type == NEGATIVE_LOOKAROUND);
+                bool is_positive = (group_type == POSITIVE_LOOKAROUND);
+                body = alloc->newInfallible<RegExpLookaround>(body,
                                                    is_positive,
                                                    end_capture_index - capture_index,
-                                                   capture_index);
+                                                   capture_index,
+                                                   state->lookaround_type());
             }
+
+            // Restore previous state.
+            state = state->previous_state();
+            builder = state->builder();
             builder->AddAtom(body);
-            if (unicode_ && (group_type == POSITIVE_LOOKAHEAD || group_type == NEGATIVE_LOOKAHEAD))
+            if (unicode_ && (group_type == POSITIVE_LOOKAROUND || group_type == NEGATIVE_LOOKAROUND))
                 continue;
             // For compatability with JSC and ES3, we allow quantifiers after
             // lookaheads, and break in all cases.
@@ -1519,6 +1554,7 @@ RegExpParser<CharT>::ParseDisjunction()
           }
           case '(': {
             SubexpressionType subexpr_type = CAPTURE;
+            RegExpLookaround::Type lookaround_type = state->lookaround_type();
             Advance();
             if (current() == '?') {
                 switch (Next()) {
@@ -1526,26 +1562,39 @@ RegExpParser<CharT>::ParseDisjunction()
                     subexpr_type = GROUPING;
                     break;
                   case '=':
-                    subexpr_type = POSITIVE_LOOKAHEAD;
+                    lookaround_type = RegExpLookaround::LOOKAHEAD;
+                    subexpr_type = POSITIVE_LOOKAROUND;
                     break;
                   case '!':
-                    subexpr_type = NEGATIVE_LOOKAHEAD;
+                    lookaround_type = RegExpLookaround::LOOKAHEAD;
+                    subexpr_type = NEGATIVE_LOOKAROUND;
                     break;
+                  case '<':
+                    Advance();
+                    lookaround_type = RegExpLookaround::LOOKBEHIND;
+                    if (Next() == '=') {
+                      subexpr_type = POSITIVE_LOOKAROUND;
+                      break;
+                    } else if (Next() == '!') {
+                      subexpr_type = NEGATIVE_LOOKAROUND;
+                      break;
+                    }
+                    // We didn't get a positive or negative after '<'.
+                    // That's an error.
+                    return ReportError(JSMSG_INVALID_GROUP);
                   default:
                     return ReportError(JSMSG_INVALID_GROUP);
                 }
                 Advance(2);
             } else {
-                if (captures_ == nullptr)
-                    captures_ = alloc->newInfallible<RegExpCaptureVector>(*alloc);
                 if (captures_started() >= kMaxCaptures)
                     return ReportError(JSMSG_TOO_MANY_PARENS);
-                captures_->append((RegExpCapture*) nullptr);
+                captures_started_++;
             }
             // Store current state and begin new disjunction parsing.
-            stored_state = alloc->newInfallible<RegExpParserState>(alloc, stored_state, subexpr_type,
-                                                                   captures_started());
-            builder = stored_state->builder();
+            state = alloc->newInfallible<RegExpParserState>(alloc, state, subexpr_type,
+                                                            lookaround_type, captures_started_);
+            builder = state->builder();
             continue;
           }
           case '[': {
@@ -1600,19 +1649,18 @@ RegExpParser<CharT>::ParseDisjunction()
               case '7': case '8': case '9': {
                 int index = 0;
                 if (ParseBackReferenceIndex(&index)) {
-                    RegExpCapture* capture = nullptr;
-                    if (captures_ != nullptr && index <= (int) captures_->length()) {
-                        capture = (*captures_)[index - 1];
+                    if (state->IsInsideCaptureGroup(index)) {
+                      // The backreference is inside the capture group it refers to.
+                      // Nothing can possibly have been captured yet.
+                      builder->AddEmpty();
+                    } else {
+                      RegExpCapture* capture = GetCapture(index);
+                      RegExpTree* atom = alloc->newInfallible<RegExpBackReference>(capture);
+                      if (unicode_)
+                          builder->AddAtom(UnicodeBackReferenceAtom(alloc, atom));
+                      else
+                          builder->AddAtom(atom);
                     }
-                    if (capture == nullptr) {
-                        builder->AddEmpty();
-                        break;
-                    }
-                    RegExpTree* atom = alloc->newInfallible<RegExpBackReference>(capture);
-                    if (unicode_)
-                        builder->AddAtom(UnicodeBackReferenceAtom(alloc, atom));
-                    else
-                        builder->AddAtom(atom);
                     break;
                 }
                 if (unicode_)

js/src/irregexp/RegExpParser.h

@@ -229,7 +229,7 @@ class RegExpParser
     bool simple() { return simple_; }
     bool contains_anchor() { return contains_anchor_; }
     void set_contains_anchor() { contains_anchor_ = true; }
-    int captures_started() { return captures_ == nullptr ? 0 : captures_->length(); }
+    int captures_started() { return captures_started_; }
     const CharT* position() { return next_pos_ - 1; }
 
     static const int kMaxCaptures = 1 << 16;
@@ -239,8 +239,8 @@ class RegExpParser
     enum SubexpressionType {
         INITIAL,
         CAPTURE,  // All positive values represent captures.
-        POSITIVE_LOOKAHEAD,
-        NEGATIVE_LOOKAHEAD,
+        POSITIVE_LOOKAROUND,
+        NEGATIVE_LOOKAROUND,
         GROUPING
     };
 
@@ -249,10 +249,12 @@ class RegExpParser
         RegExpParserState(LifoAlloc* alloc,
                           RegExpParserState* previous_state,
                           SubexpressionType group_type,
+                          RegExpLookaround::Type lookaround_type,
                           int disjunction_capture_index)
             : previous_state_(previous_state),
               builder_(alloc->newInfallible<RegExpBuilder>(alloc)),
               group_type_(group_type),
+              lookaround_type_(lookaround_type),
               disjunction_capture_index_(disjunction_capture_index)
         {}
         // Parser state of containing expression, if any.
@@ -262,11 +264,16 @@ class RegExpParser
         RegExpBuilder* builder() { return builder_; }
         // Type of regexp being parsed (parenthesized group or entire regexp).
         SubexpressionType group_type() { return group_type_; }
+        // Lookahead or Lookbehind.
+        RegExpLookaround::Type lookaround_type() { return lookaround_type_; }
         // Index in captures array of first capture in this sub-expression, if any.
         // Also the capture index of this sub-expression itself, if group_type
         // is CAPTURE.
         int capture_index() { return disjunction_capture_index_; }
 
+        // Check whether the parser is inside a capture group with the given index.
+        bool IsInsideCaptureGroup(int index);
+
       private:
         // Linked list implementation of stack of states.
         RegExpParserState* previous_state_;
@@ -274,10 +281,15 @@ class RegExpParser
         RegExpBuilder* builder_;
         // Stored disjunction type (capture, look-ahead or grouping), if any.
         SubexpressionType group_type_;
+        // Stored read direction.
+        RegExpLookaround::Type lookaround_type_;
         // Stored disjunction's capture index (if any).
         int disjunction_capture_index_;
     };
 
+    // Return the 1-indexed RegExpCapture object, allocate if necessary.
+    RegExpCapture* GetCapture(int index);
+
     widechar current() { return current_; }
     bool has_more() { return has_more_; }
     bool has_next() { return next_pos_ < end_; }
@@ -294,6 +306,7 @@ class RegExpParser
     const CharT* next_pos_;
     const CharT* end_;
     widechar current_;
+    int captures_started_;
     // The capture count is only valid after we have scanned for captures.
     int capture_count_;
     bool has_more_;

js/src/vm/SelfHosting.cpp

@@ -2102,6 +2102,21 @@ intrinsic_ModuleNamespaceExports(JSContext* cx, unsigned argc, Value* vp)
     return true;
 }
 
+static bool
+intrinsic_PromiseResolve(JSContext* cx, unsigned argc, Value* vp)
+{
+    CallArgs args = CallArgsFromVp(argc, vp);
+    MOZ_ASSERT(args.length() == 2);
+
+    RootedObject constructor(cx, &args[0].toObject());
+    JSObject* promise = js::PromiseResolve(cx, constructor, args[1]);
+    if (!promise)
+        return false;
+
+    args.rval().setObject(*promise);
+    return true;
+}
+
 // The self-hosting global isn't initialized with the normal set of builtins.
 // Instead, individual C++-implemented functions that're required by
 // self-hosted code are defined as global functions. Accessing these
@@ -2498,6 +2513,10 @@ static const JSFunctionSpec intrinsic_functions[] = {
     JS_FN("AddModuleNamespaceBinding", intrinsic_AddModuleNamespaceBinding, 4, 0),
     JS_FN("ModuleNamespaceExports", intrinsic_ModuleNamespaceExports, 1, 0),
 
+    JS_FN("IsPromiseObject", intrinsic_IsInstanceOfBuiltin<PromiseObject>, 1, 0),
+    JS_FN("CallPromiseMethodIfWrapped", CallNonGenericSelfhostedMethod<Is<PromiseObject>>, 2, 0),
+    JS_FN("PromiseResolve", intrinsic_PromiseResolve, 2, 0),
+
     JS_FS_END
 };
 

js/xpconnect/tests/chrome/test_xrayToJS.xul

@@ -243,7 +243,7 @@ https://bugzilla.mozilla.org/show_bug.cgi?id=933681
                       "$`", "$'", Symbol.species])
 
   gPrototypeProperties['Promise'] =
-    ["constructor", "catch", "then", Symbol.toStringTag];
+    ["constructor", "catch", "then", "finally", Symbol.toStringTag];
   gConstructorProperties['Promise'] =
     constructorProps(["resolve", "reject", "all", "race", Symbol.species]);
 

modules/brotli/README.mozilla

@@ -1,9 +1,6 @@
 This is the Brotli data compression library from
 https://github.com/google/brotli.
 
-Currently, we import only the Brotli decoder (the /dec/ subdirectory), not the
-encoder (/enc/ subdirectory).
-
 Upstream code can be viewed at
   https://github.com/google/brotli/tree/master/dec
 
@@ -14,4 +11,4 @@ The in-tree copy is updated by running
   sh update.sh
 from within the modules/brotli directory.
 
-Current version: [commit 5b4769990dc14a2bd466d2599c946c5652cba4b2].
+Current version: [commit d6d98957ca8ccb1ef45922e978bb10efca0ea541].

modules/brotli/common/constants.h

@@ -28,18 +28,25 @@
 /* "code length of 8 is repeated" */
 #define BROTLI_INITIAL_REPEATED_CODE_LENGTH 8
 
+/* "Large Window Brotli" */
+#define BROTLI_LARGE_MAX_DISTANCE_BITS 62U
+#define BROTLI_LARGE_MIN_WBITS 10
+#define BROTLI_LARGE_MAX_WBITS 30
+
 /* Specification: 4. Encoding of distances */
 #define BROTLI_NUM_DISTANCE_SHORT_CODES 16
 #define BROTLI_MAX_NPOSTFIX 3
 #define BROTLI_MAX_NDIRECT 120
 #define BROTLI_MAX_DISTANCE_BITS 24U
-/* BROTLI_NUM_DISTANCE_SYMBOLS == 520 */
-#define BROTLI_NUM_DISTANCE_SYMBOLS (BROTLI_NUM_DISTANCE_SHORT_CODES + \
-                                     BROTLI_MAX_NDIRECT +              \
-                                     (BROTLI_MAX_DISTANCE_BITS <<      \
-                                      (BROTLI_MAX_NPOSTFIX + 1)))
-/* Distance that is guaranteed to be representable in any stream. */
+#define BROTLI_DISTANCE_ALPHABET_SIZE(NPOSTFIX, NDIRECT, MAXNBITS) ( \
+    BROTLI_NUM_DISTANCE_SHORT_CODES + (NDIRECT) +                    \
+    ((MAXNBITS) << ((NPOSTFIX) + 1)))
+/* BROTLI_NUM_DISTANCE_SYMBOLS == 1128 */
+#define BROTLI_NUM_DISTANCE_SYMBOLS \
+    BROTLI_DISTANCE_ALPHABET_SIZE(  \
+        BROTLI_MAX_NDIRECT, BROTLI_MAX_NPOSTFIX, BROTLI_LARGE_MAX_DISTANCE_BITS)
 #define BROTLI_MAX_DISTANCE 0x3FFFFFC
+#define BROTLI_MAX_ALLOWED_DISTANCE 0x7FFFFFFC
 
 /* 7.1. Context modes and context ID lookup for literals */
 /* "context IDs for literals are in the range of 0..63" */

modules/brotli/common/context.h

@@ -6,110 +6,171 @@
 
 /* Lookup table to map the previous two bytes to a context id.
 
-   There are four different context modeling modes defined here:
-     CONTEXT_LSB6: context id is the least significant 6 bits of the last byte,
-     CONTEXT_MSB6: context id is the most significant 6 bits of the last byte,
-     CONTEXT_UTF8: second-order context model tuned for UTF8-encoded text,
-     CONTEXT_SIGNED: second-order context model tuned for signed integers.
+  There are four different context modeling modes defined here:
+    CONTEXT_LSB6: context id is the least significant 6 bits of the last byte,
+    CONTEXT_MSB6: context id is the most significant 6 bits of the last byte,
+    CONTEXT_UTF8: second-order context model tuned for UTF8-encoded text,
+    CONTEXT_SIGNED: second-order context model tuned for signed integers.
 
-   The context id for the UTF8 context model is calculated as follows. If p1
-   and p2 are the previous two bytes, we calculate the context as
+  If |p1| and |p2| are the previous two bytes, and |mode| is current context
+  mode, we calculate the context as:
 
-     context = kContextLookup[p1] | kContextLookup[p2 + 256].
+    context = ContextLut(mode)[p1] | ContextLut(mode)[p2 + 256].
 
-   If the previous two bytes are ASCII characters (i.e. < 128), this will be
-   equivalent to
+  For CONTEXT_UTF8 mode, if the previous two bytes are ASCII characters
+  (i.e. < 128), this will be equivalent to
 
-     context = 4 * context1(p1) + context2(p2),
+    context = 4 * context1(p1) + context2(p2),
 
-   where context1 is based on the previous byte in the following way:
+  where context1 is based on the previous byte in the following way:
 
-     0  : non-ASCII control
-     1  : \t, \n, \r
-     2  : space
-     3  : other punctuation
-     4  : " '
-     5  : %
-     6  : ( < [ {
-     7  : ) > ] }
-     8  : , ; :
-     9  : .
-     10 : =
-     11 : number
-     12 : upper-case vowel
-     13 : upper-case consonant
-     14 : lower-case vowel
-     15 : lower-case consonant
+    0  : non-ASCII control
+    1  : \t, \n, \r
+    2  : space
+    3  : other punctuation
+    4  : " '
+    5  : %
+    6  : ( < [ {
+    7  : ) > ] }
+    8  : , ; :
+    9  : .
+    10 : =
+    11 : number
+    12 : upper-case vowel
+    13 : upper-case consonant
+    14 : lower-case vowel
+    15 : lower-case consonant
 
-   and context2 is based on the second last byte:
+  and context2 is based on the second last byte:
 
-     0 : control, space
-     1 : punctuation
-     2 : upper-case letter, number
-     3 : lower-case letter
+    0 : control, space
+    1 : punctuation
+    2 : upper-case letter, number
+    3 : lower-case letter
 
-   If the last byte is ASCII, and the second last byte is not (in a valid UTF8
-   stream it will be a continuation byte, value between 128 and 191), the
-   context is the same as if the second last byte was an ASCII control or space.
+  If the last byte is ASCII, and the second last byte is not (in a valid UTF8
+  stream it will be a continuation byte, value between 128 and 191), the
+  context is the same as if the second last byte was an ASCII control or space.
 
-   If the last byte is a UTF8 lead byte (value >= 192), then the next byte will
-   be a continuation byte and the context id is 2 or 3 depending on the LSB of
-   the last byte and to a lesser extent on the second last byte if it is ASCII.
+  If the last byte is a UTF8 lead byte (value >= 192), then the next byte will
+  be a continuation byte and the context id is 2 or 3 depending on the LSB of
+  the last byte and to a lesser extent on the second last byte if it is ASCII.
 
-   If the last byte is a UTF8 continuation byte, the second last byte can be:
-     - continuation byte: the next byte is probably ASCII or lead byte (assuming
-       4-byte UTF8 characters are rare) and the context id is 0 or 1.
-     - lead byte (192 - 207): next byte is ASCII or lead byte, context is 0 or 1
-     - lead byte (208 - 255): next byte is continuation byte, context is 2 or 3
+  If the last byte is a UTF8 continuation byte, the second last byte can be:
+    - continuation byte: the next byte is probably ASCII or lead byte (assuming
+      4-byte UTF8 characters are rare) and the context id is 0 or 1.
+    - lead byte (192 - 207): next byte is ASCII or lead byte, context is 0 or 1
+    - lead byte (208 - 255): next byte is continuation byte, context is 2 or 3
 
-   The possible value combinations of the previous two bytes, the range of
-   context ids and the type of the next byte is summarized in the table below:
+  The possible value combinations of the previous two bytes, the range of
+  context ids and the type of the next byte is summarized in the table below:
 
-   |--------\-----------------------------------------------------------------|
-   |         \                         Last byte                              |
-   | Second   \---------------------------------------------------------------|
-   | last byte \    ASCII            |   cont. byte        |   lead byte      |
-   |            \   (0-127)          |   (128-191)         |   (192-)         |
-   |=============|===================|=====================|==================|
-   |  ASCII      | next: ASCII/lead  |  not valid          |  next: cont.     |
-   |  (0-127)    | context: 4 - 63   |                     |  context: 2 - 3  |
-   |-------------|-------------------|---------------------|------------------|
-   |  cont. byte | next: ASCII/lead  |  next: ASCII/lead   |  next: cont.     |
-   |  (128-191)  | context: 4 - 63   |  context: 0 - 1     |  context: 2 - 3  |
-   |-------------|-------------------|---------------------|------------------|
-   |  lead byte  | not valid         |  next: ASCII/lead   |  not valid       |
-   |  (192-207)  |                   |  context: 0 - 1     |                  |
-   |-------------|-------------------|---------------------|------------------|
-   |  lead byte  | not valid         |  next: cont.        |  not valid       |
-   |  (208-)     |                   |  context: 2 - 3     |                  |
-   |-------------|-------------------|---------------------|------------------|
-
-   The context id for the signed context mode is calculated as:
-
-     context = (kContextLookup[512 + p1] << 3) | kContextLookup[512 + p2].
-
-   For any context modeling modes, the context ids can be calculated by |-ing
-   together two lookups from one table using context model dependent offsets:
-
-     context = kContextLookup[offset1 + p1] | kContextLookup[offset2 + p2].
-
-   where offset1 and offset2 are dependent on the context mode.
+  |--------\-----------------------------------------------------------------|
+  |         \                         Last byte                              |
+  | Second   \---------------------------------------------------------------|
+  | last byte \    ASCII            |   cont. byte        |   lead byte      |
+  |            \   (0-127)          |   (128-191)         |   (192-)         |
+  |=============|===================|=====================|==================|
+  |  ASCII      | next: ASCII/lead  |  not valid          |  next: cont.     |
+  |  (0-127)    | context: 4 - 63   |                     |  context: 2 - 3  |
+  |-------------|-------------------|---------------------|------------------|
+  |  cont. byte | next: ASCII/lead  |  next: ASCII/lead   |  next: cont.     |
+  |  (128-191)  | context: 4 - 63   |  context: 0 - 1     |  context: 2 - 3  |
+  |-------------|-------------------|---------------------|------------------|
+  |  lead byte  | not valid         |  next: ASCII/lead   |  not valid       |
+  |  (192-207)  |                   |  context: 0 - 1     |                  |
+  |-------------|-------------------|---------------------|------------------|
+  |  lead byte  | not valid         |  next: cont.        |  not valid       |
+  |  (208-)     |                   |  context: 2 - 3     |                  |
+  |-------------|-------------------|---------------------|------------------|
 */
 
-#ifndef BROTLI_DEC_CONTEXT_H_
-#define BROTLI_DEC_CONTEXT_H_
+#ifndef BROTLI_COMMON_CONTEXT_H_
+#define BROTLI_COMMON_CONTEXT_H_
 
 #include <brotli/types.h>
 
-enum ContextType {
+typedef enum ContextType {
   CONTEXT_LSB6 = 0,
   CONTEXT_MSB6 = 1,
   CONTEXT_UTF8 = 2,
   CONTEXT_SIGNED = 3
-};
+} ContextType;
 
 /* Common context lookup table for all context modes. */
-static const uint8_t kContextLookup[1792] = {
+static const uint8_t kContextLookup[2048] = {
+  /* CONTEXT_LSB6, last byte. */
+   0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15,
+  16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31,
+  32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47,
+  48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63,
+   0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15,
+  16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31,
+  32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47,
+  48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63,
+   0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15,
+  16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31,
+  32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47,
+  48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63,
+   0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15,
+  16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31,
+  32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47,
+  48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63,
+
+  /* CONTEXT_LSB6, second last byte, */
+  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+
+  /* CONTEXT_MSB6, last byte. */
+   0,  0,  0,  0,  1,  1,  1,  1,  2,  2,  2,  2,  3,  3,  3,  3,
+   4,  4,  4,  4,  5,  5,  5,  5,  6,  6,  6,  6,  7,  7,  7,  7,
+   8,  8,  8,  8,  9,  9,  9,  9, 10, 10, 10, 10, 11, 11, 11, 11,
+  12, 12, 12, 12, 13, 13, 13, 13, 14, 14, 14, 14, 15, 15, 15, 15,
+  16, 16, 16, 16, 17, 17, 17, 17, 18, 18, 18, 18, 19, 19, 19, 19,
+  20, 20, 20, 20, 21, 21, 21, 21, 22, 22, 22, 22, 23, 23, 23, 23,
+  24, 24, 24, 24, 25, 25, 25, 25, 26, 26, 26, 26, 27, 27, 27, 27,
+  28, 28, 28, 28, 29, 29, 29, 29, 30, 30, 30, 30, 31, 31, 31, 31,
+  32, 32, 32, 32, 33, 33, 33, 33, 34, 34, 34, 34, 35, 35, 35, 35,
+  36, 36, 36, 36, 37, 37, 37, 37, 38, 38, 38, 38, 39, 39, 39, 39,
+  40, 40, 40, 40, 41, 41, 41, 41, 42, 42, 42, 42, 43, 43, 43, 43,
+  44, 44, 44, 44, 45, 45, 45, 45, 46, 46, 46, 46, 47, 47, 47, 47,
+  48, 48, 48, 48, 49, 49, 49, 49, 50, 50, 50, 50, 51, 51, 51, 51,
+  52, 52, 52, 52, 53, 53, 53, 53, 54, 54, 54, 54, 55, 55, 55, 55,
+  56, 56, 56, 56, 57, 57, 57, 57, 58, 58, 58, 58, 59, 59, 59, 59,
+  60, 60, 60, 60, 61, 61, 61, 61, 62, 62, 62, 62, 63, 63, 63, 63,
+
+  /* CONTEXT_MSB6, second last byte, */
+  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+
   /* CONTEXT_UTF8, last byte. */
   /* ASCII range. */
    0,  0,  0,  0,  0,  0,  0,  0,  0,  4,  4,  0,  0,  4,  0,  0,
@@ -130,6 +191,7 @@ static const uint8_t kContextLookup[1792] = {
   2, 3, 2, 3, 2, 3, 2, 3, 2, 3, 2, 3, 2, 3, 2, 3,
   2, 3, 2, 3, 2, 3, 2, 3, 2, 3, 2, 3, 2, 3, 2, 3,
   2, 3, 2, 3, 2, 3, 2, 3, 2, 3, 2, 3, 2, 3, 2, 3,
+
   /* CONTEXT_UTF8 second last byte. */
   /* ASCII range. */
   0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
@@ -150,23 +212,7 @@ static const uint8_t kContextLookup[1792] = {
   0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
   2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
   2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
-  /* CONTEXT_SIGNED, second last byte. */
-  0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
-  2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
-  2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
-  2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
-  3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
-  3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
-  3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
-  3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
-  4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
-  4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
-  4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
-  4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
-  5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
-  5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
-  5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
-  6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 7,
+
   /* CONTEXT_SIGNED, last byte, same as the above values shifted by 3 bits. */
    0, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
   16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
@@ -184,68 +230,32 @@ static const uint8_t kContextLookup[1792] = {
   40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40,
   40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40,
   48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 56,
-  /* CONTEXT_LSB6, last byte. */
-   0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15,
-  16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31,
-  32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47,
-  48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63,
-   0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15,
-  16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31,
-  32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47,
-  48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63,
-   0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15,
-  16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31,
-  32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47,
-  48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63,
-   0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15,
-  16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31,
-  32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47,
-  48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63,
-  /* CONTEXT_MSB6, last byte. */
-   0,  0,  0,  0,  1,  1,  1,  1,  2,  2,  2,  2,  3,  3,  3,  3,
-   4,  4,  4,  4,  5,  5,  5,  5,  6,  6,  6,  6,  7,  7,  7,  7,
-   8,  8,  8,  8,  9,  9,  9,  9, 10, 10, 10, 10, 11, 11, 11, 11,
-  12, 12, 12, 12, 13, 13, 13, 13, 14, 14, 14, 14, 15, 15, 15, 15,
-  16, 16, 16, 16, 17, 17, 17, 17, 18, 18, 18, 18, 19, 19, 19, 19,
-  20, 20, 20, 20, 21, 21, 21, 21, 22, 22, 22, 22, 23, 23, 23, 23,
-  24, 24, 24, 24, 25, 25, 25, 25, 26, 26, 26, 26, 27, 27, 27, 27,
-  28, 28, 28, 28, 29, 29, 29, 29, 30, 30, 30, 30, 31, 31, 31, 31,
-  32, 32, 32, 32, 33, 33, 33, 33, 34, 34, 34, 34, 35, 35, 35, 35,
-  36, 36, 36, 36, 37, 37, 37, 37, 38, 38, 38, 38, 39, 39, 39, 39,
-  40, 40, 40, 40, 41, 41, 41, 41, 42, 42, 42, 42, 43, 43, 43, 43,
-  44, 44, 44, 44, 45, 45, 45, 45, 46, 46, 46, 46, 47, 47, 47, 47,
-  48, 48, 48, 48, 49, 49, 49, 49, 50, 50, 50, 50, 51, 51, 51, 51,
-  52, 52, 52, 52, 53, 53, 53, 53, 54, 54, 54, 54, 55, 55, 55, 55,
-  56, 56, 56, 56, 57, 57, 57, 57, 58, 58, 58, 58, 59, 59, 59, 59,
-  60, 60, 60, 60, 61, 61, 61, 61, 62, 62, 62, 62, 63, 63, 63, 63,
-  /* CONTEXT_{M,L}SB6, second last byte, */
-  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+
+  /* CONTEXT_SIGNED, second last byte. */
+  0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
+  2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
+  2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
+  2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
+  3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
+  3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
+  3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
+  3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
+  4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
+  4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
+  4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
+  4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
+  5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+  5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+  5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+  6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 7,
 };
 
-static const int kContextLookupOffsets[8] = {
-  /* CONTEXT_LSB6 */
-  1024, 1536,
-  /* CONTEXT_MSB6 */
-  1280, 1536,
-  /* CONTEXT_UTF8 */
-  0, 256,
-  /* CONTEXT_SIGNED */
-  768, 512,
-};
+typedef const uint8_t* ContextLut;
 
-#endif  /* BROTLI_DEC_CONTEXT_H_ */
+/* typeof(MODE) == ContextType; returns ContextLut */
+#define BROTLI_CONTEXT_LUT(MODE) (&kContextLookup[(MODE) << 9])
+
+/* typeof(LUT) == ContextLut */
+#define BROTLI_CONTEXT(P1, P2, LUT) ((LUT)[P1] | ((LUT) + 256)[P2])
+
+#endif  /* BROTLI_COMMON_CONTEXT_H_ */

modules/brotli/common/dictionary.c

@@ -5883,7 +5883,7 @@ static BrotliDictionary kBrotliDictionary = {
   122784,
 
   /* data */
-#ifdef BROTLI_EXTERNAL_DICTIONARY_DATA
+#if defined(BROTLI_EXTERNAL_DICTIONARY_DATA)
   NULL
 #else
   kBrotliDictionaryData

modules/brotli/common/dictionary.h

@@ -27,13 +27,13 @@ typedef struct BrotliDictionary {
    * Dictionary consists of words with length of [4..24] bytes.
    * Values at [0..3] and [25..31] indices should not be addressed.
    */
-  const uint8_t size_bits_by_length[32];
+  uint8_t size_bits_by_length[32];
 
   /* assert(offset[i + 1] == offset[i] + (bits[i] ? (i << bits[i]) : 0)) */
-  const uint32_t offsets_by_length[32];
+  uint32_t offsets_by_length[32];
 
   /* assert(data_size == offsets_by_length[31]) */
-  const size_t data_size;
+  size_t data_size;
 
   /* Data array is not bound, and should obey to size_bits_by_length values.
      Specified size matches default (RFC 7932) dictionary. Its size is
@@ -41,7 +41,7 @@ typedef struct BrotliDictionary {
   const uint8_t* data;
 } BrotliDictionary;
 
-BROTLI_COMMON_API extern const BrotliDictionary* BrotliGetDictionary(void);
+BROTLI_COMMON_API const BrotliDictionary* BrotliGetDictionary(void);
 
 /**
  * Sets dictionary data.

modules/brotli/common/platform.h

@@ -0,0 +1,568 @@
+/* Copyright 2016 Google Inc. All Rights Reserved.
+
+   Distributed under MIT license.
+   See file LICENSE for detail or copy at https://opensource.org/licenses/MIT
+*/
+
+/* Macros for compiler / platform specific features and build options.
+
+   Build options are:
+    * BROTLI_BUILD_32_BIT disables 64-bit optimizations
+    * BROTLI_BUILD_64_BIT forces to use 64-bit optimizations
+    * BROTLI_BUILD_BIG_ENDIAN forces to use big-endian optimizations
+    * BROTLI_BUILD_ENDIAN_NEUTRAL disables endian-aware optimizations
+    * BROTLI_BUILD_LITTLE_ENDIAN forces to use little-endian optimizations
+    * BROTLI_BUILD_PORTABLE disables dangerous optimizations, like unaligned
+      read and overlapping memcpy; this reduces decompression speed by 5%
+    * BROTLI_BUILD_NO_RBIT disables "rbit" optimization for ARM CPUs
+    * BROTLI_DEBUG dumps file name and line number when decoder detects stream
+      or memory error
+    * BROTLI_ENABLE_LOG enables asserts and dumps various state information
+*/
+
+#ifndef BROTLI_COMMON_PLATFORM_H_
+#define BROTLI_COMMON_PLATFORM_H_
+
+#include <string.h>  /* memcpy */
+#include <stdlib.h>  /* malloc, free */
+
+#include <brotli/port.h>
+#include <brotli/types.h>
+
+#if defined(OS_LINUX) || defined(OS_CYGWIN)
+#include <endian.h>
+#elif defined(OS_FREEBSD)
+#include <machine/endian.h>
+#elif defined(OS_MACOSX)
+#include <machine/endian.h>
+/* Let's try and follow the Linux convention */
+#define BROTLI_X_BYTE_ORDER BYTE_ORDER
+#define BROTLI_X_LITTLE_ENDIAN LITTLE_ENDIAN
+#define BROTLI_X_BIG_ENDIAN BIG_ENDIAN
+#endif
+
+#if defined(BROTLI_ENABLE_LOG) || defined(BROTLI_DEBUG)
+#include <assert.h>
+#include <stdio.h>
+#endif
+
+/* The following macros were borrowed from https://github.com/nemequ/hedley
+ * with permission of original author - Evan Nemerson <evan@nemerson.com> */
+
+/* >>> >>> >>> hedley macros */
+
+/* Define "BROTLI_PREDICT_TRUE" and "BROTLI_PREDICT_FALSE" macros for capable
+   compilers.
+
+To apply compiler hint, enclose the branching condition into macros, like this:
+
+  if (BROTLI_PREDICT_TRUE(zero == 0)) {
+    // main execution path
+  } else {
+    // compiler should place this code outside of main execution path
+  }
+
+OR:
+
+  if (BROTLI_PREDICT_FALSE(something_rare_or_unexpected_happens)) {
+    // compiler should place this code outside of main execution path
+  }
+
+*/
+#if BROTLI_GNUC_HAS_BUILTIN(__builtin_expect, 3, 0, 0) || \
+    BROTLI_INTEL_VERSION_CHECK(16, 0, 0) ||               \
+    BROTLI_SUNPRO_VERSION_CHECK(5, 15, 0) ||              \
+    BROTLI_ARM_VERSION_CHECK(4, 1, 0) ||                  \
+    BROTLI_IBM_VERSION_CHECK(10, 1, 0) ||                 \
+    BROTLI_TI_VERSION_CHECK(7, 3, 0) ||                   \
+    BROTLI_TINYC_VERSION_CHECK(0, 9, 27)
+#define BROTLI_PREDICT_TRUE(x) (__builtin_expect(!!(x), 1))
+#define BROTLI_PREDICT_FALSE(x) (__builtin_expect(x, 0))
+#else
+#define BROTLI_PREDICT_FALSE(x) (x)
+#define BROTLI_PREDICT_TRUE(x) (x)
+#endif
+
+#if defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L) && \
+    !defined(__cplusplus)
+#define BROTLI_RESTRICT restrict
+#elif BROTLI_GNUC_VERSION_CHECK(3, 1, 0) ||                         \
+    BROTLI_MSVC_VERSION_CHECK(14, 0, 0) ||                          \
+    BROTLI_INTEL_VERSION_CHECK(16, 0, 0) ||                         \
+    BROTLI_ARM_VERSION_CHECK(4, 1, 0) ||                            \
+    BROTLI_IBM_VERSION_CHECK(10, 1, 0) ||                           \
+    BROTLI_PGI_VERSION_CHECK(17, 10, 0) ||                          \
+    BROTLI_TI_VERSION_CHECK(8, 0, 0) ||                             \
+    BROTLI_IAR_VERSION_CHECK(8, 0, 0) ||                            \
+    (BROTLI_SUNPRO_VERSION_CHECK(5, 14, 0) && defined(__cplusplus))
+#define BROTLI_RESTRICT __restrict
+#elif BROTLI_SUNPRO_VERSION_CHECK(5, 3, 0) && !defined(__cplusplus)
+#define BROTLI_RESTRICT _Restrict
+#else
+#define BROTLI_RESTRICT
+#endif
+
+#if (defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L)) || \
+    (defined(__cplusplus) && (__cplusplus >= 199711L))
+#define BROTLI_MAYBE_INLINE inline
+#elif defined(__GNUC_STDC_INLINE__) || defined(__GNUC_GNU_INLINE__) || \
+    BROTLI_ARM_VERSION_CHECK(6, 2, 0)
+#define BROTLI_MAYBE_INLINE __inline__
+#elif BROTLI_MSVC_VERSION_CHECK(12, 0, 0) || \
+    BROTLI_ARM_VERSION_CHECK(4, 1, 0) || BROTLI_TI_VERSION_CHECK(8, 0, 0)
+#define BROTLI_MAYBE_INLINE __inline
+#else
+#define BROTLI_MAYBE_INLINE
+#endif
+
+#if BROTLI_GNUC_HAS_ATTRIBUTE(always_inline, 4, 0, 0) ||                       \
+    BROTLI_INTEL_VERSION_CHECK(16, 0, 0) ||                                    \
+    BROTLI_SUNPRO_VERSION_CHECK(5, 11, 0) ||                                   \
+    BROTLI_ARM_VERSION_CHECK(4, 1, 0) ||                                       \
+    BROTLI_IBM_VERSION_CHECK(10, 1, 0) ||                                      \
+    BROTLI_TI_VERSION_CHECK(8, 0, 0) ||                                        \
+    (BROTLI_TI_VERSION_CHECK(7, 3, 0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__))
+#define BROTLI_INLINE BROTLI_MAYBE_INLINE __attribute__((__always_inline__))
+#elif BROTLI_MSVC_VERSION_CHECK(12, 0, 0)
+#define BROTLI_INLINE BROTLI_MAYBE_INLINE __forceinline
+#elif BROTLI_TI_VERSION_CHECK(7, 0, 0) && defined(__cplusplus)
+#define BROTLI_INLINE BROTLI_MAYBE_INLINE _Pragma("FUNC_ALWAYS_INLINE;")
+#elif BROTLI_IAR_VERSION_CHECK(8, 0, 0)
+#define BROTLI_INLINE BROTLI_MAYBE_INLINE _Pragma("inline=forced")
+#else
+#define BROTLI_INLINE BROTLI_MAYBE_INLINE
+#endif
+
+#if BROTLI_GNUC_HAS_ATTRIBUTE(noinline, 4, 0, 0) ||                            \
+    BROTLI_INTEL_VERSION_CHECK(16, 0, 0) ||                                    \
+    BROTLI_SUNPRO_VERSION_CHECK(5, 11, 0) ||                                   \
+    BROTLI_ARM_VERSION_CHECK(4, 1, 0) ||                                       \
+    BROTLI_IBM_VERSION_CHECK(10, 1, 0) ||                                      \
+    BROTLI_TI_VERSION_CHECK(8, 0, 0) ||                                        \
+    (BROTLI_TI_VERSION_CHECK(7, 3, 0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__))
+#define BROTLI_NOINLINE __attribute__((__noinline__))
+#elif BROTLI_MSVC_VERSION_CHECK(13, 10, 0)
+#define BROTLI_NOINLINE __declspec(noinline)
+#elif BROTLI_PGI_VERSION_CHECK(10, 2, 0)
+#define BROTLI_NOINLINE _Pragma("noinline")
+#elif BROTLI_TI_VERSION_CHECK(6, 0, 0) && defined(__cplusplus)
+#define BROTLI_NOINLINE _Pragma("FUNC_CANNOT_INLINE;")
+#elif BROTLI_IAR_VERSION_CHECK(8, 0, 0)
+#define BROTLI_NOINLINE _Pragma("inline=never")
+#else
+#define BROTLI_NOINLINE
+#endif
+
+/* BROTLI_INTERNAL could be defined to override visibility, e.g. for tests. */
+#if !defined(BROTLI_INTERNAL)
+#if defined(_WIN32) || defined(__CYGWIN__)
+#define BROTLI_INTERNAL
+#elif BROTLI_GNUC_VERSION_CHECK(3, 3, 0) ||                         \
+    BROTLI_TI_VERSION_CHECK(8, 0, 0) ||                             \
+    BROTLI_INTEL_VERSION_CHECK(16, 0, 0) ||                         \
+    BROTLI_ARM_VERSION_CHECK(4, 1, 0) ||                            \
+    BROTLI_IBM_VERSION_CHECK(13, 1, 0) ||                           \
+    BROTLI_SUNPRO_VERSION_CHECK(5, 11, 0) ||                        \
+    (BROTLI_TI_VERSION_CHECK(7, 3, 0) &&                            \
+     defined(__TI_GNU_ATTRIBUTE_SUPPORT__) && defined(__TI_EABI__))
+#define BROTLI_INTERNAL __attribute__ ((visibility ("hidden")))
+#else
+#define BROTLI_INTERNAL
+#endif
+#endif
+
+/* <<< <<< <<< end of hedley macros. */
+
+#if BROTLI_GNUC_HAS_ATTRIBUTE(unused, 2, 7, 0) || \
+    BROTLI_INTEL_VERSION_CHECK(16, 0, 0)
+#define BROTLI_UNUSED_FUNCTION static BROTLI_INLINE __attribute__ ((unused))
+#else
+#define BROTLI_UNUSED_FUNCTION static BROTLI_INLINE
+#endif
+
+#if BROTLI_GNUC_HAS_ATTRIBUTE(aligned, 2, 7, 0)
+#define BROTLI_ALIGNED(N) __attribute__((aligned(N)))
+#else
+#define BROTLI_ALIGNED(N)
+#endif
+
+#if (defined(__ARM_ARCH) && (__ARM_ARCH == 7)) || \
+    (defined(M_ARM) && (M_ARM == 7))
+#define BROTLI_TARGET_ARMV7
+#endif  /* ARMv7 */
+
+#if (defined(__ARM_ARCH) && (__ARM_ARCH == 8)) || \
+    defined(__aarch64__) || defined(__ARM64_ARCH_8__)
+#define BROTLI_TARGET_ARMV8_ANY
+
+#if defined(__ARM_32BIT_STATE)
+#define BROTLI_TARGET_ARMV8_32
+#elif defined(__ARM_64BIT_STATE)
+#define BROTLI_TARGET_ARMV8_64
+#endif
+
+#endif  /* ARMv8 */
+
+#if defined(__ARM_NEON__) || defined(__ARM_NEON)
+#define BROTLI_TARGET_NEON
+#endif
+
+#if defined(__i386) || defined(_M_IX86)
+#define BROTLI_TARGET_X86
+#endif
+
+#if defined(__x86_64__) || defined(_M_X64)
+#define BROTLI_TARGET_X64
+#endif
+
+#if defined(__PPC64__)
+#define BROTLI_TARGET_POWERPC64
+#endif
+
+#if defined(__riscv) && defined(__riscv_xlen) && __riscv_xlen == 64
+#define BROTLI_TARGET_RISCV64
+#endif
+
+#if defined(BROTLI_BUILD_64_BIT)
+#define BROTLI_64_BITS 1
+#elif defined(BROTLI_BUILD_32_BIT)
+#define BROTLI_64_BITS 0
+#elif defined(BROTLI_TARGET_X64) || defined(BROTLI_TARGET_ARMV8_64) || \
+    defined(BROTLI_TARGET_POWERPC64) || defined(BROTLI_TARGET_RISCV64)
+#define BROTLI_64_BITS 1
+#else
+#define BROTLI_64_BITS 0
+#endif
+
+#if (BROTLI_64_BITS)
+#define brotli_reg_t uint64_t
+#else
+#define brotli_reg_t uint32_t
+#endif
+
+#if defined(BROTLI_BUILD_BIG_ENDIAN)
+#define BROTLI_BIG_ENDIAN 1
+#elif defined(BROTLI_BUILD_LITTLE_ENDIAN)
+#define BROTLI_LITTLE_ENDIAN 1
+#elif defined(BROTLI_BUILD_ENDIAN_NEUTRAL)
+/* Just break elif chain. */
+#elif defined(__BYTE_ORDER__) && (__BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__)
+#define BROTLI_LITTLE_ENDIAN 1
+#elif defined(_WIN32) || defined(BROTLI_TARGET_X64)
+/* Win32 & x64 can currently always be assumed to be little endian */
+#define BROTLI_LITTLE_ENDIAN 1
+#elif defined(__BYTE_ORDER__) && (__BYTE_ORDER__ == __ORDER_BIG_ENDIAN__)
+#define BROTLI_BIG_ENDIAN 1
+#elif defined(BROTLI_X_BYTE_ORDER)
+#if BROTLI_X_BYTE_ORDER == BROTLI_X_LITTLE_ENDIAN
+#define BROTLI_LITTLE_ENDIAN 1
+#elif BROTLI_X_BYTE_ORDER == BROTLI_X_BIG_ENDIAN
+#define BROTLI_BIG_ENDIAN 1
+#endif
+#endif  /* BROTLI_X_BYTE_ORDER */
+
+#if !defined(BROTLI_LITTLE_ENDIAN)
+#define BROTLI_LITTLE_ENDIAN 0
+#endif
+
+#if !defined(BROTLI_BIG_ENDIAN)
+#define BROTLI_BIG_ENDIAN 0
+#endif
+
+#if defined(BROTLI_X_BYTE_ORDER)
+#undef BROTLI_X_BYTE_ORDER
+#undef BROTLI_X_LITTLE_ENDIAN
+#undef BROTLI_X_BIG_ENDIAN
+#endif
+
+#if defined(BROTLI_BUILD_PORTABLE)
+#define BROTLI_ALIGNED_READ (!!1)
+#elif defined(BROTLI_TARGET_X86) || defined(BROTLI_TARGET_X64) || \
+    defined(BROTLI_TARGET_ARMV7) || defined(BROTLI_TARGET_ARMV8_ANY) || \
+    defined(BROTLI_TARGET_RISCV64)
+/* Allow unaligned read only for white-listed CPUs. */
+#define BROTLI_ALIGNED_READ (!!0)
+#else
+#define BROTLI_ALIGNED_READ (!!1)
+#endif
+
+#if BROTLI_ALIGNED_READ
+/* Portable unaligned memory access: read / write values via memcpy. */
+static BROTLI_INLINE uint16_t BrotliUnalignedRead16(const void* p) {
+  uint16_t t;
+  memcpy(&t, p, sizeof t);
+  return t;
+}
+static BROTLI_INLINE uint32_t BrotliUnalignedRead32(const void* p) {
+  uint32_t t;
+  memcpy(&t, p, sizeof t);
+  return t;
+}
+static BROTLI_INLINE uint64_t BrotliUnalignedRead64(const void* p) {
+  uint64_t t;
+  memcpy(&t, p, sizeof t);
+  return t;
+}
+static BROTLI_INLINE void BrotliUnalignedWrite64(void* p, uint64_t v) {
+  memcpy(p, &v, sizeof v);
+}
+#else  /* BROTLI_ALIGNED_READ */
+/* Unaligned memory access is allowed: just cast pointer to requested type. */
+#if defined(ADDRESS_SANITIZER) || defined(THREAD_SANITIZER) || \
+    defined(MEMORY_SANITIZER)
+/* Consider we have an unaligned load/store of 4 bytes from address 0x...05.
+   AddressSanitizer will treat it as a 3-byte access to the range 05:07 and
+   will miss a bug if 08 is the first unaddressable byte.
+   ThreadSanitizer will also treat this as a 3-byte access to 05:07 and will
+   miss a race between this access and some other accesses to 08.
+   MemorySanitizer will correctly propagate the shadow on unaligned stores
+   and correctly report bugs on unaligned loads, but it may not properly
+   update and report the origin of the uninitialized memory.
+   For all three tools, replacing an unaligned access with a tool-specific
+   callback solves the problem. */
+#if defined(__cplusplus)
+extern "C" {
+#endif  /* __cplusplus */
+  uint16_t __sanitizer_unaligned_load16(const void* p);
+  uint32_t __sanitizer_unaligned_load32(const void* p);
+  uint64_t __sanitizer_unaligned_load64(const void* p);
+  void __sanitizer_unaligned_store64(void* p, uint64_t v);
+#if defined(__cplusplus)
+}  /* extern "C" */
+#endif  /* __cplusplus */
+#define BrotliUnalignedRead16 __sanitizer_unaligned_load16
+#define BrotliUnalignedRead32 __sanitizer_unaligned_load32
+#define BrotliUnalignedRead64 __sanitizer_unaligned_load64
+#define BrotliUnalignedWrite64 __sanitizer_unaligned_store64
+#else
+static BROTLI_INLINE uint16_t BrotliUnalignedRead16(const void* p) {
+  return *(const uint16_t*)p;
+}
+static BROTLI_INLINE uint32_t BrotliUnalignedRead32(const void* p) {
+  return *(const uint32_t*)p;
+}
+#if (BROTLI_64_BITS)
+static BROTLI_INLINE uint64_t BrotliUnalignedRead64(const void* p) {
+  return *(const uint64_t*)p;
+}
+static BROTLI_INLINE void BrotliUnalignedWrite64(void* p, uint64_t v) {
+  *(uint64_t*)p = v;
+}
+#else  /* BROTLI_64_BITS */
+/* Avoid emitting LDRD / STRD, which require properly aligned address. */
+/* If __attribute__(aligned) is available, use that. Otherwise, memcpy. */
+
+#if BROTLI_GNUC_HAS_ATTRIBUTE(aligned, 2, 7, 0)
+typedef BROTLI_ALIGNED(1) uint64_t brotli_unaligned_uint64_t;
+
+static BROTLI_INLINE uint64_t BrotliUnalignedRead64(const void* p) {
+  return (uint64_t) ((brotli_unaligned_uint64_t*) p)[0];
+}
+static BROTLI_INLINE void BrotliUnalignedWrite64(void* p, uint64_t v) {
+  brotli_unaligned_uint64_t* dwords = (brotli_unaligned_uint64_t*) p;
+  dwords[0] = (brotli_unaligned_uint64_t) v;
+}
+#else /* BROTLI_GNUC_HAS_ATTRIBUTE(aligned, 2, 7, 0) */
+static BROTLI_INLINE uint64_t BrotliUnalignedRead64(const void* p) {
+  uint64_t v;
+  memcpy(&v, p, sizeof(uint64_t));
+  return v;
+}
+
+static BROTLI_INLINE void BrotliUnalignedWrite64(void* p, uint64_t v) {
+  memcpy(p, &v, sizeof(uint64_t));
+}
+#endif  /* BROTLI_GNUC_HAS_ATTRIBUTE(aligned, 2, 7, 0) */
+#endif  /* BROTLI_64_BITS */
+#endif  /* ASAN / TSAN / MSAN */
+#endif  /* BROTLI_ALIGNED_READ */
+
+#if BROTLI_LITTLE_ENDIAN
+/* Straight endianness. Just read / write values. */
+#define BROTLI_UNALIGNED_LOAD16LE BrotliUnalignedRead16
+#define BROTLI_UNALIGNED_LOAD32LE BrotliUnalignedRead32
+#define BROTLI_UNALIGNED_LOAD64LE BrotliUnalignedRead64
+#define BROTLI_UNALIGNED_STORE64LE BrotliUnalignedWrite64
+#elif BROTLI_BIG_ENDIAN  /* BROTLI_LITTLE_ENDIAN */
+/* Explain compiler to byte-swap values. */
+#define BROTLI_BSWAP16_(V) ((uint16_t)( \
+  (((V) & 0xFFU) << 8) | \
+  (((V) >> 8) & 0xFFU)))
+static BROTLI_INLINE uint16_t BROTLI_UNALIGNED_LOAD16LE(const void* p) {
+  uint16_t value = BrotliUnalignedRead16(p);
+  return BROTLI_BSWAP16_(value);
+}
+#define BROTLI_BSWAP32_(V) ( \
+  (((V) & 0xFFU) << 24) | (((V) & 0xFF00U) << 8) | \
+  (((V) >> 8) & 0xFF00U) | (((V) >> 24) & 0xFFU))
+static BROTLI_INLINE uint32_t BROTLI_UNALIGNED_LOAD32LE(const void* p) {
+  uint32_t value = BrotliUnalignedRead32(p);
+  return BROTLI_BSWAP32_(value);
+}
+#define BROTLI_BSWAP64_(V) ( \
+  (((V) & 0xFFU) << 56) | (((V) & 0xFF00U) << 40) | \
+  (((V) & 0xFF0000U) << 24) | (((V) & 0xFF000000U) << 8) | \
+  (((V) >> 8) & 0xFF000000U) | (((V) >> 24) & 0xFF0000U) | \
+  (((V) >> 40) & 0xFF00U) | (((V) >> 56) & 0xFFU))
+static BROTLI_INLINE uint64_t BROTLI_UNALIGNED_LOAD64LE(const void* p) {
+  uint64_t value = BrotliUnalignedRead64(p);
+  return BROTLI_BSWAP64_(value);
+}
+static BROTLI_INLINE void BROTLI_UNALIGNED_STORE64LE(void* p, uint64_t v) {
+  uint64_t value = BROTLI_BSWAP64_(v);
+  BrotliUnalignedWrite64(p, value);
+}
+#else  /* BROTLI_LITTLE_ENDIAN */
+/* Read / store values byte-wise; hopefully compiler will understand. */
+static BROTLI_INLINE uint16_t BROTLI_UNALIGNED_LOAD16LE(const void* p) {
+  const uint8_t* in = (const uint8_t*)p;
+  return (uint16_t)(in[0] | (in[1] << 8));
+}
+static BROTLI_INLINE uint32_t BROTLI_UNALIGNED_LOAD32LE(const void* p) {
+  const uint8_t* in = (const uint8_t*)p;
+  uint32_t value = (uint32_t)(in[0]);
+  value |= (uint32_t)(in[1]) << 8;
+  value |= (uint32_t)(in[2]) << 16;
+  value |= (uint32_t)(in[3]) << 24;
+  return value;
+}
+static BROTLI_INLINE uint64_t BROTLI_UNALIGNED_LOAD64LE(const void* p) {
+  const uint8_t* in = (const uint8_t*)p;
+  uint64_t value = (uint64_t)(in[0]);
+  value |= (uint64_t)(in[1]) << 8;
+  value |= (uint64_t)(in[2]) << 16;
+  value |= (uint64_t)(in[3]) << 24;
+  value |= (uint64_t)(in[4]) << 32;
+  value |= (uint64_t)(in[5]) << 40;
+  value |= (uint64_t)(in[6]) << 48;
+  value |= (uint64_t)(in[7]) << 56;
+  return value;
+}
+static BROTLI_INLINE void BROTLI_UNALIGNED_STORE64LE(void* p, uint64_t v) {
+  uint8_t* out = (uint8_t*)p;
+  out[0] = (uint8_t)v;
+  out[1] = (uint8_t)(v >> 8);
+  out[2] = (uint8_t)(v >> 16);
+  out[3] = (uint8_t)(v >> 24);
+  out[4] = (uint8_t)(v >> 32);
+  out[5] = (uint8_t)(v >> 40);
+  out[6] = (uint8_t)(v >> 48);
+  out[7] = (uint8_t)(v >> 56);
+}
+#endif  /* BROTLI_LITTLE_ENDIAN */
+
+/* BROTLI_IS_CONSTANT macros returns true for compile-time constants. */
+#if BROTLI_GNUC_HAS_BUILTIN(__builtin_constant_p, 3, 0, 1) || \
+    BROTLI_INTEL_VERSION_CHECK(16, 0, 0)
+#define BROTLI_IS_CONSTANT(x) (!!__builtin_constant_p(x))
+#else
+#define BROTLI_IS_CONSTANT(x) (!!0)
+#endif
+
+#if defined(BROTLI_TARGET_ARMV7) || defined(BROTLI_TARGET_ARMV8_ANY)
+#define BROTLI_HAS_UBFX (!!1)
+#else
+#define BROTLI_HAS_UBFX (!!0)
+#endif
+
+#if defined(BROTLI_ENABLE_LOG)
+#define BROTLI_DCHECK(x) assert(x)
+#define BROTLI_LOG(x) printf x
+#else
+#define BROTLI_DCHECK(x)
+#define BROTLI_LOG(x)
+#endif
+
+#if defined(BROTLI_DEBUG) || defined(BROTLI_ENABLE_LOG)
+static BROTLI_INLINE void BrotliDump(const char* f, int l, const char* fn) {
+  fprintf(stderr, "%s:%d (%s)\n", f, l, fn);
+  fflush(stderr);
+}
+#define BROTLI_DUMP() BrotliDump(__FILE__, __LINE__, __FUNCTION__)
+#else
+#define BROTLI_DUMP() (void)(0)
+#endif
+
+/* TODO: add appropriate icc/sunpro/arm/ibm/ti checks. */
+#if (BROTLI_GNUC_VERSION_CHECK(3, 0, 0) || defined(__llvm__)) && \
+    !defined(BROTLI_BUILD_NO_RBIT)
+#if defined(BROTLI_TARGET_ARMV7) || defined(BROTLI_TARGET_ARMV8_ANY)
+/* TODO: detect ARMv6T2 and enable this code for it. */
+static BROTLI_INLINE brotli_reg_t BrotliRBit(brotli_reg_t input) {
+  brotli_reg_t output;
+  __asm__("rbit %0, %1\n" : "=r"(output) : "r"(input));
+  return output;
+}
+#define BROTLI_RBIT(x) BrotliRBit(x)
+#endif  /* armv7 / armv8 */
+#endif  /* gcc || clang */
+#if !defined(BROTLI_RBIT)
+static BROTLI_INLINE void BrotliRBit(void) { /* Should break build if used. */ }
+#endif  /* BROTLI_RBIT */
+
+#define BROTLI_REPEAT(N, X) {     \
+  if ((N & 1) != 0) {X;}          \
+  if ((N & 2) != 0) {X; X;}       \
+  if ((N & 4) != 0) {X; X; X; X;} \
+}
+
+#define BROTLI_UNUSED(X) (void)(X)
+
+#define BROTLI_MIN_MAX(T)                                                      \
+  static BROTLI_INLINE T brotli_min_ ## T (T a, T b) { return a < b ? a : b; } \
+  static BROTLI_INLINE T brotli_max_ ## T (T a, T b) { return a > b ? a : b; }
+BROTLI_MIN_MAX(double) BROTLI_MIN_MAX(float) BROTLI_MIN_MAX(int)
+BROTLI_MIN_MAX(size_t) BROTLI_MIN_MAX(uint32_t) BROTLI_MIN_MAX(uint8_t)
+#undef BROTLI_MIN_MAX
+#define BROTLI_MIN(T, A, B) (brotli_min_ ## T((A), (B)))
+#define BROTLI_MAX(T, A, B) (brotli_max_ ## T((A), (B)))
+
+#define BROTLI_SWAP(T, A, I, J) { \
+  T __brotli_swap_tmp = (A)[(I)]; \
+  (A)[(I)] = (A)[(J)];            \
+  (A)[(J)] = __brotli_swap_tmp;   \
+}
+
+/* Default brotli_alloc_func */
+static void* BrotliDefaultAllocFunc(void* opaque, size_t size) {
+  BROTLI_UNUSED(opaque);
+  return malloc(size);
+}
+
+/* Default brotli_free_func */
+static void BrotliDefaultFreeFunc(void* opaque, void* address) {
+  BROTLI_UNUSED(opaque);
+  free(address);
+}
+
+BROTLI_UNUSED_FUNCTION void BrotliSuppressUnusedFunctions(void) {
+  BROTLI_UNUSED(&BrotliSuppressUnusedFunctions);
+  BROTLI_UNUSED(&BrotliUnalignedRead16);
+  BROTLI_UNUSED(&BrotliUnalignedRead32);
+  BROTLI_UNUSED(&BrotliUnalignedRead64);
+  BROTLI_UNUSED(&BrotliUnalignedWrite64);
+  BROTLI_UNUSED(&BROTLI_UNALIGNED_LOAD16LE);
+  BROTLI_UNUSED(&BROTLI_UNALIGNED_LOAD32LE);
+  BROTLI_UNUSED(&BROTLI_UNALIGNED_LOAD64LE);
+  BROTLI_UNUSED(&BROTLI_UNALIGNED_STORE64LE);
+  BROTLI_UNUSED(&BrotliRBit);
+  BROTLI_UNUSED(&brotli_min_double);
+  BROTLI_UNUSED(&brotli_max_double);
+  BROTLI_UNUSED(&brotli_min_float);
+  BROTLI_UNUSED(&brotli_max_float);
+  BROTLI_UNUSED(&brotli_min_int);
+  BROTLI_UNUSED(&brotli_max_int);
+  BROTLI_UNUSED(&brotli_min_size_t);
+  BROTLI_UNUSED(&brotli_max_size_t);
+  BROTLI_UNUSED(&brotli_min_uint32_t);
+  BROTLI_UNUSED(&brotli_max_uint32_t);
+  BROTLI_UNUSED(&brotli_min_uint8_t);
+  BROTLI_UNUSED(&brotli_max_uint8_t);
+  BROTLI_UNUSED(&BrotliDefaultAllocFunc);
+  BROTLI_UNUSED(&BrotliDefaultFreeFunc);
+#if defined(BROTLI_DEBUG) || defined(BROTLI_ENABLE_LOG)
+  BROTLI_UNUSED(&BrotliDump);
+#endif
+}
+
+#endif  /* BROTLI_COMMON_PLATFORM_H_ */

modules/brotli/common/transform.c

@@ -0,0 +1,235 @@
+/* Copyright 2013 Google Inc. All Rights Reserved.
+
+   Distributed under MIT license.
+   See file LICENSE for detail or copy at https://opensource.org/licenses/MIT
+*/
+
+#include "./transform.h"
+
+#if defined(__cplusplus) || defined(c_plusplus)
+extern "C" {
+#endif
+
+/* RFC 7932 transforms string data */
+static const char kPrefixSuffix[217] =
+      "\1 \2, \10 of the \4 of \2s \1.\5 and \4 "
+/* 0x  _0 _2  __5        _E    _3  _6 _8     _E */
+      "in \1\"\4 to \2\">\1\n\2. \1]\5 for \3 a \6 "
+/* 2x     _3_ _5    _A_  _D_ _F  _2 _4     _A   _E */
+      "that \1\'\6 with \6 from \4 by \1(\6. T"
+/* 4x       _5_ _7      _E      _5    _A _C */
+      "he \4 on \4 as \4 is \4ing \2\n\t\1:\3ed "
+/* 6x     _3    _8    _D    _2    _7_ _ _A _C */
+      "\2=\"\4 at \3ly \1,\2=\'\5.com/\7. This \5"
+/* 8x  _0 _ _3    _8   _C _E _ _1     _7       _F */
+      " not \3er \3al \4ful \4ive \5less \4es"
+/* Ax       _5   _9   _D    _2    _7     _D */
+      "t \4ize \2\xc2\xa0\4ous \5 the \2e \0";
+/* Cx    _2    _7___ ___ _A    _F     _5  _8 */
+
+static const uint16_t kPrefixSuffixMap[50] = {
+  0x00, 0x02, 0x05, 0x0E, 0x13, 0x16, 0x18, 0x1E, 0x23, 0x25,
+  0x2A, 0x2D, 0x2F, 0x32, 0x34, 0x3A, 0x3E, 0x45, 0x47, 0x4E,
+  0x55, 0x5A, 0x5C, 0x63, 0x68, 0x6D, 0x72, 0x77, 0x7A, 0x7C,
+  0x80, 0x83, 0x88, 0x8C, 0x8E, 0x91, 0x97, 0x9F, 0xA5, 0xA9,
+  0xAD, 0xB2, 0xB7, 0xBD, 0xC2, 0xC7, 0xCA, 0xCF, 0xD5, 0xD8
+};
+
+/* RFC 7932 transforms */
+static const uint8_t kTransformsData[] = {
+  49, BROTLI_TRANSFORM_IDENTITY, 49,
+  49, BROTLI_TRANSFORM_IDENTITY, 0,
+   0, BROTLI_TRANSFORM_IDENTITY, 0,
+  49, BROTLI_TRANSFORM_OMIT_FIRST_1, 49,
+  49, BROTLI_TRANSFORM_UPPERCASE_FIRST, 0,
+  49, BROTLI_TRANSFORM_IDENTITY, 47,
+   0, BROTLI_TRANSFORM_IDENTITY, 49,
+   4, BROTLI_TRANSFORM_IDENTITY, 0,
+  49, BROTLI_TRANSFORM_IDENTITY, 3,
+  49, BROTLI_TRANSFORM_UPPERCASE_FIRST, 49,
+  49, BROTLI_TRANSFORM_IDENTITY, 6,
+  49, BROTLI_TRANSFORM_OMIT_FIRST_2, 49,
+  49, BROTLI_TRANSFORM_OMIT_LAST_1, 49,
+   1, BROTLI_TRANSFORM_IDENTITY, 0,
+  49, BROTLI_TRANSFORM_IDENTITY, 1,
+   0, BROTLI_TRANSFORM_UPPERCASE_FIRST, 0,
+  49, BROTLI_TRANSFORM_IDENTITY, 7,
+  49, BROTLI_TRANSFORM_IDENTITY, 9,
+  48, BROTLI_TRANSFORM_IDENTITY, 0,
+  49, BROTLI_TRANSFORM_IDENTITY, 8,
+  49, BROTLI_TRANSFORM_IDENTITY, 5,
+  49, BROTLI_TRANSFORM_IDENTITY, 10,
+  49, BROTLI_TRANSFORM_IDENTITY, 11,
+  49, BROTLI_TRANSFORM_OMIT_LAST_3, 49,
+  49, BROTLI_TRANSFORM_IDENTITY, 13,
+  49, BROTLI_TRANSFORM_IDENTITY, 14,
+  49, BROTLI_TRANSFORM_OMIT_FIRST_3, 49,
+  49, BROTLI_TRANSFORM_OMIT_LAST_2, 49,
+  49, BROTLI_TRANSFORM_IDENTITY, 15,
+  49, BROTLI_TRANSFORM_IDENTITY, 16,
+   0, BROTLI_TRANSFORM_UPPERCASE_FIRST, 49,
+  49, BROTLI_TRANSFORM_IDENTITY, 12,
+   5, BROTLI_TRANSFORM_IDENTITY, 49,
+   0, BROTLI_TRANSFORM_IDENTITY, 1,
+  49, BROTLI_TRANSFORM_OMIT_FIRST_4, 49,
+  49, BROTLI_TRANSFORM_IDENTITY, 18,
+  49, BROTLI_TRANSFORM_IDENTITY, 17,
+  49, BROTLI_TRANSFORM_IDENTITY, 19,
+  49, BROTLI_TRANSFORM_IDENTITY, 20,
+  49, BROTLI_TRANSFORM_OMIT_FIRST_5, 49,
+  49, BROTLI_TRANSFORM_OMIT_FIRST_6, 49,
+  47, BROTLI_TRANSFORM_IDENTITY, 49,
+  49, BROTLI_TRANSFORM_OMIT_LAST_4, 49,
+  49, BROTLI_TRANSFORM_IDENTITY, 22,
+  49, BROTLI_TRANSFORM_UPPERCASE_ALL, 49,
+  49, BROTLI_TRANSFORM_IDENTITY, 23,
+  49, BROTLI_TRANSFORM_IDENTITY, 24,
+  49, BROTLI_TRANSFORM_IDENTITY, 25,
+  49, BROTLI_TRANSFORM_OMIT_LAST_7, 49,
+  49, BROTLI_TRANSFORM_OMIT_LAST_1, 26,
+  49, BROTLI_TRANSFORM_IDENTITY, 27,
+  49, BROTLI_TRANSFORM_IDENTITY, 28,
+   0, BROTLI_TRANSFORM_IDENTITY, 12,
+  49, BROTLI_TRANSFORM_IDENTITY, 29,
+  49, BROTLI_TRANSFORM_OMIT_FIRST_9, 49,
+  49, BROTLI_TRANSFORM_OMIT_FIRST_7, 49,
+  49, BROTLI_TRANSFORM_OMIT_LAST_6, 49,
+  49, BROTLI_TRANSFORM_IDENTITY, 21,
+  49, BROTLI_TRANSFORM_UPPERCASE_FIRST, 1,
+  49, BROTLI_TRANSFORM_OMIT_LAST_8, 49,
+  49, BROTLI_TRANSFORM_IDENTITY, 31,
+  49, BROTLI_TRANSFORM_IDENTITY, 32,
+  47, BROTLI_TRANSFORM_IDENTITY, 3,
+  49, BROTLI_TRANSFORM_OMIT_LAST_5, 49,
+  49, BROTLI_TRANSFORM_OMIT_LAST_9, 49,
+   0, BROTLI_TRANSFORM_UPPERCASE_FIRST, 1,
+  49, BROTLI_TRANSFORM_UPPERCASE_FIRST, 8,
+   5, BROTLI_TRANSFORM_IDENTITY, 21,
+  49, BROTLI_TRANSFORM_UPPERCASE_ALL, 0,
+  49, BROTLI_TRANSFORM_UPPERCASE_FIRST, 10,
+  49, BROTLI_TRANSFORM_IDENTITY, 30,
+   0, BROTLI_TRANSFORM_IDENTITY, 5,
+  35, BROTLI_TRANSFORM_IDENTITY, 49,
+  47, BROTLI_TRANSFORM_IDENTITY, 2,
+  49, BROTLI_TRANSFORM_UPPERCASE_FIRST, 17,
+  49, BROTLI_TRANSFORM_IDENTITY, 36,
+  49, BROTLI_TRANSFORM_IDENTITY, 33,
+   5, BROTLI_TRANSFORM_IDENTITY, 0,
+  49, BROTLI_TRANSFORM_UPPERCASE_FIRST, 21,
+  49, BROTLI_TRANSFORM_UPPERCASE_FIRST, 5,
+  49, BROTLI_TRANSFORM_IDENTITY, 37,
+   0, BROTLI_TRANSFORM_IDENTITY, 30,
+  49, BROTLI_TRANSFORM_IDENTITY, 38,
+   0, BROTLI_TRANSFORM_UPPERCASE_ALL, 0,
+  49, BROTLI_TRANSFORM_IDENTITY, 39,
+   0, BROTLI_TRANSFORM_UPPERCASE_ALL, 49,
+  49, BROTLI_TRANSFORM_IDENTITY, 34,
+  49, BROTLI_TRANSFORM_UPPERCASE_ALL, 8,
+  49, BROTLI_TRANSFORM_UPPERCASE_FIRST, 12,
+   0, BROTLI_TRANSFORM_IDENTITY, 21,
+  49, BROTLI_TRANSFORM_IDENTITY, 40,
+   0, BROTLI_TRANSFORM_UPPERCASE_FIRST, 12,
+  49, BROTLI_TRANSFORM_IDENTITY, 41,
+  49, BROTLI_TRANSFORM_IDENTITY, 42,
+  49, BROTLI_TRANSFORM_UPPERCASE_ALL, 17,
+  49, BROTLI_TRANSFORM_IDENTITY, 43,
+   0, BROTLI_TRANSFORM_UPPERCASE_FIRST, 5,
+  49, BROTLI_TRANSFORM_UPPERCASE_ALL, 10,
+   0, BROTLI_TRANSFORM_IDENTITY, 34,
+  49, BROTLI_TRANSFORM_UPPERCASE_FIRST, 33,
+  49, BROTLI_TRANSFORM_IDENTITY, 44,
+  49, BROTLI_TRANSFORM_UPPERCASE_ALL, 5,
+  45, BROTLI_TRANSFORM_IDENTITY, 49,
+   0, BROTLI_TRANSFORM_IDENTITY, 33,
+  49, BROTLI_TRANSFORM_UPPERCASE_FIRST, 30,
+  49, BROTLI_TRANSFORM_UPPERCASE_ALL, 30,
+  49, BROTLI_TRANSFORM_IDENTITY, 46,
+  49, BROTLI_TRANSFORM_UPPERCASE_ALL, 1,
+  49, BROTLI_TRANSFORM_UPPERCASE_FIRST, 34,
+   0, BROTLI_TRANSFORM_UPPERCASE_FIRST, 33,
+   0, BROTLI_TRANSFORM_UPPERCASE_ALL, 30,
+   0, BROTLI_TRANSFORM_UPPERCASE_ALL, 1,
+  49, BROTLI_TRANSFORM_UPPERCASE_ALL, 33,
+  49, BROTLI_TRANSFORM_UPPERCASE_ALL, 21,
+  49, BROTLI_TRANSFORM_UPPERCASE_ALL, 12,
+   0, BROTLI_TRANSFORM_UPPERCASE_ALL, 5,
+  49, BROTLI_TRANSFORM_UPPERCASE_ALL, 34,
+   0, BROTLI_TRANSFORM_UPPERCASE_ALL, 12,
+   0, BROTLI_TRANSFORM_UPPERCASE_FIRST, 30,
+   0, BROTLI_TRANSFORM_UPPERCASE_ALL, 34,
+   0, BROTLI_TRANSFORM_UPPERCASE_FIRST, 34,
+};
+
+static BrotliTransforms kBrotliTransforms = {
+  sizeof(kPrefixSuffix),
+  (const uint8_t*)kPrefixSuffix,
+  kPrefixSuffixMap,
+  sizeof(kTransformsData) / (3 * sizeof(kTransformsData[0])),
+  kTransformsData,
+  {0, 12, 27, 23, 42, 63, 56, 48, 59, 64}
+};
+
+const BrotliTransforms* BrotliGetTransforms(void) {
+  return &kBrotliTransforms;
+}
+
+static int ToUpperCase(uint8_t* p) {
+  if (p[0] < 0xC0) {
+    if (p[0] >= 'a' && p[0] <= 'z') {
+      p[0] ^= 32;
+    }
+    return 1;
+  }
+  /* An overly simplified uppercasing model for UTF-8. */
+  if (p[0] < 0xE0) {
+    p[1] ^= 32;
+    return 2;
+  }
+  /* An arbitrary transform for three byte characters. */
+  p[2] ^= 5;
+  return 3;
+}
+
+int BrotliTransformDictionaryWord(uint8_t* dst, const uint8_t* word, int len,
+    const BrotliTransforms* transforms, int transform_idx) {
+  int idx = 0;
+  const uint8_t* prefix = BROTLI_TRANSFORM_PREFIX(transforms, transform_idx);
+  uint8_t type = BROTLI_TRANSFORM_TYPE(transforms, transform_idx);
+  const uint8_t* suffix = BROTLI_TRANSFORM_SUFFIX(transforms, transform_idx);
+  {
+    int prefix_len = *prefix++;
+    while (prefix_len--) { dst[idx++] = *prefix++; }
+  }
+  {
+    const int t = type;
+    int i = 0;
+    if (t <= BROTLI_TRANSFORM_OMIT_LAST_9) {
+      len -= t;
+    } else if (t >= BROTLI_TRANSFORM_OMIT_FIRST_1
+        && t <= BROTLI_TRANSFORM_OMIT_FIRST_9) {
+      int skip = t - (BROTLI_TRANSFORM_OMIT_FIRST_1 - 1);
+      word += skip;
+      len -= skip;
+    }
+    while (i < len) { dst[idx++] = word[i++]; }
+    if (t == BROTLI_TRANSFORM_UPPERCASE_FIRST) {
+      ToUpperCase(&dst[idx - len]);
+    } else if (t == BROTLI_TRANSFORM_UPPERCASE_ALL) {
+      uint8_t* uppercase = &dst[idx - len];
+      while (len > 0) {
+        int step = ToUpperCase(uppercase);
+        uppercase += step;
+        len -= step;
+      }
+    }
+  }
+  {
+    int suffix_len = *suffix++;
+    while (suffix_len--) { dst[idx++] = *suffix++; }
+    return idx;
+  }
+}
+
+#if defined(__cplusplus) || defined(c_plusplus)
+}  /* extern "C" */
+#endif

modules/brotli/common/transform.h

@@ -0,0 +1,80 @@
+/* transforms is a part of ABI, but not API.
+
+   It means that there are some functions that are supposed to be in "common"
+   library, but header itself is not placed into include/brotli. This way,
+   aforementioned functions will be available only to brotli internals.
+ */
+
+#ifndef BROTLI_COMMON_TRANSFORM_H_
+#define BROTLI_COMMON_TRANSFORM_H_
+
+#include <brotli/port.h>
+#include <brotli/types.h>
+
+#if defined(__cplusplus) || defined(c_plusplus)
+extern "C" {
+#endif
+
+enum BrotliWordTransformType {
+  BROTLI_TRANSFORM_IDENTITY = 0,
+  BROTLI_TRANSFORM_OMIT_LAST_1 = 1,
+  BROTLI_TRANSFORM_OMIT_LAST_2 = 2,
+  BROTLI_TRANSFORM_OMIT_LAST_3 = 3,
+  BROTLI_TRANSFORM_OMIT_LAST_4 = 4,
+  BROTLI_TRANSFORM_OMIT_LAST_5 = 5,
+  BROTLI_TRANSFORM_OMIT_LAST_6 = 6,
+  BROTLI_TRANSFORM_OMIT_LAST_7 = 7,
+  BROTLI_TRANSFORM_OMIT_LAST_8 = 8,
+  BROTLI_TRANSFORM_OMIT_LAST_9 = 9,
+  BROTLI_TRANSFORM_UPPERCASE_FIRST = 10,
+  BROTLI_TRANSFORM_UPPERCASE_ALL = 11,
+  BROTLI_TRANSFORM_OMIT_FIRST_1 = 12,
+  BROTLI_TRANSFORM_OMIT_FIRST_2 = 13,
+  BROTLI_TRANSFORM_OMIT_FIRST_3 = 14,
+  BROTLI_TRANSFORM_OMIT_FIRST_4 = 15,
+  BROTLI_TRANSFORM_OMIT_FIRST_5 = 16,
+  BROTLI_TRANSFORM_OMIT_FIRST_6 = 17,
+  BROTLI_TRANSFORM_OMIT_FIRST_7 = 18,
+  BROTLI_TRANSFORM_OMIT_FIRST_8 = 19,
+  BROTLI_TRANSFORM_OMIT_FIRST_9 = 20,
+  BROTLI_NUM_TRANSFORM_TYPES  /* Counts transforms, not a transform itself. */
+};
+
+#define BROTLI_TRANSFORMS_MAX_CUT_OFF BROTLI_TRANSFORM_OMIT_LAST_9
+
+typedef struct BrotliTransforms {
+  uint16_t prefix_suffix_size;
+  /* Last character must be null, so prefix_suffix_size must be at least 1. */
+  const uint8_t* prefix_suffix;
+  const uint16_t* prefix_suffix_map;
+  uint32_t num_transforms;
+  /* Each entry is a [prefix_id, transform, suffix_id] triplet. */
+  const uint8_t* transforms;
+  /* Indices of transforms like ["", BROTLI_TRANSFORM_OMIT_LAST_#, ""].
+     0-th element corresponds to ["", BROTLI_TRANSFORM_IDENTITY, ""].
+     -1, if cut-off transform does not exist. */
+  int16_t cutOffTransforms[BROTLI_TRANSFORMS_MAX_CUT_OFF + 1];
+} BrotliTransforms;
+
+/* T is BrotliTransforms*; result is uint8_t. */
+#define BROTLI_TRANSFORM_PREFIX_ID(T, I) ((T)->transforms[((I) * 3) + 0])
+#define BROTLI_TRANSFORM_TYPE(T, I)      ((T)->transforms[((I) * 3) + 1])
+#define BROTLI_TRANSFORM_SUFFIX_ID(T, I) ((T)->transforms[((I) * 3) + 2])
+
+/* T is BrotliTransforms*; result is const uint8_t*. */
+#define BROTLI_TRANSFORM_PREFIX(T, I) (&(T)->prefix_suffix[ \
+    (T)->prefix_suffix_map[BROTLI_TRANSFORM_PREFIX_ID(T, I)]])
+#define BROTLI_TRANSFORM_SUFFIX(T, I) (&(T)->prefix_suffix[ \
+    (T)->prefix_suffix_map[BROTLI_TRANSFORM_SUFFIX_ID(T, I)]])
+
+BROTLI_COMMON_API const BrotliTransforms* BrotliGetTransforms(void);
+
+BROTLI_COMMON_API int BrotliTransformDictionaryWord(
+    uint8_t* dst, const uint8_t* word, int len,
+    const BrotliTransforms* transforms, int transform_idx);
+
+#if defined(__cplusplus) || defined(c_plusplus)
+}  /* extern "C" */
+#endif
+
+#endif  /* BROTLI_COMMON_TRANSFORM_H_ */

modules/brotli/common/version.h

@@ -14,6 +14,13 @@
    BrotliEncoderVersion methods. */
 
 /* Semantic version, calculated as (MAJOR << 24) | (MINOR << 12) | PATCH */
-#define BROTLI_VERSION 0x1000001
+#define BROTLI_VERSION 0x1000007
+
+/* This macro is used by build system to produce Libtool-friendly soname. See
+   https://www.gnu.org/software/libtool/manual/html_node/Libtool-versioning.html
+ */
+
+/* ABI version, calculated as (CURRENT << 24) | (REVISION << 12) | AGE */
+#define BROTLI_ABI_VERSION 0x1007000
 
 #endif  /* BROTLI_COMMON_VERSION_H_ */

modules/brotli/dec/Makefile

@@ -1,12 +0,0 @@
-#brotli/dec
-
-include ../shared.mk
-
-CFLAGS += -Wall
-
-OBJS = bit_reader.o decode.o dictionary.o huffman.o state.o
-
-all : $(OBJS)
-
-clean :
-	rm -f $(OBJS)

modules/brotli/dec/bit_reader.c

@@ -8,8 +8,8 @@
 
 #include "./bit_reader.h"
 
+#include "../common/platform.h"
 #include <brotli/types.h>
-#include "./port.h"
 
 #if defined(__cplusplus) || defined(c_plusplus)
 extern "C" {

modules/brotli/dec/bit_reader.h

@@ -11,16 +11,16 @@
 
 #include <string.h>  /* memcpy */
 
+#include "../common/platform.h"
 #include <brotli/types.h>
-#include "./port.h"
 
 #if defined(__cplusplus) || defined(c_plusplus)
 extern "C" {
 #endif
 
-#define BROTLI_SHORT_FILL_BIT_WINDOW_READ (sizeof(reg_t) >> 1)
+#define BROTLI_SHORT_FILL_BIT_WINDOW_READ (sizeof(brotli_reg_t) >> 1)
 
-static const uint32_t kBitMask[33] = { 0x0000,
+static const uint32_t kBitMask[33] = {  0x00000000,
     0x00000001, 0x00000003, 0x00000007, 0x0000000F,
     0x0000001F, 0x0000003F, 0x0000007F, 0x000000FF,
     0x000001FF, 0x000003FF, 0x000007FF, 0x00000FFF,
@@ -32,24 +32,24 @@ static const uint32_t kBitMask[33] = { 0x0000,
 };
 
 static BROTLI_INLINE uint32_t BitMask(uint32_t n) {
-  if (IS_CONSTANT(n) || BROTLI_HAS_UBFX) {
+  if (BROTLI_IS_CONSTANT(n) || BROTLI_HAS_UBFX) {
     /* Masking with this expression turns to a single
        "Unsigned Bit Field Extract" UBFX instruction on ARM. */
-    return ~((0xffffffffU) << n);
+    return ~((0xFFFFFFFFu) << n);
   } else {
     return kBitMask[n];
   }
 }
 
 typedef struct {
-  reg_t val_;              /* pre-fetched bits */
+  brotli_reg_t val_;       /* pre-fetched bits */
   uint32_t bit_pos_;       /* current bit-reading position in val_ */
   const uint8_t* next_in;  /* the byte we're reading from */
   size_t avail_in;
 } BrotliBitReader;
 
 typedef struct {
-  reg_t val_;
+  brotli_reg_t val_;
   uint32_t bit_pos_;
   const uint8_t* next_in;
   size_t avail_in;
@@ -58,8 +58,9 @@ typedef struct {
 /* Initializes the BrotliBitReader fields. */
 BROTLI_INTERNAL void BrotliInitBitReader(BrotliBitReader* const br);
 
-/* Ensures that accumulator is not empty. May consume one byte of input.
-   Returns 0 if data is required but there is no input available.
+/* Ensures that accumulator is not empty.
+   May consume up to sizeof(brotli_reg_t) - 1 bytes of input.
+   Returns BROTLI_FALSE if data is required but there is no input available.
    For BROTLI_ALIGNED_READ this function also prepares bit reader for aligned
    reading. */
 BROTLI_INTERNAL BROTLI_BOOL BrotliWarmupBitReader(BrotliBitReader* const br);
@@ -98,82 +99,27 @@ static BROTLI_INLINE BROTLI_BOOL BrotliCheckInputAmount(
   return TO_BROTLI_BOOL(br->avail_in >= num);
 }
 
-static BROTLI_INLINE uint16_t BrotliLoad16LE(const uint8_t* in) {
-  if (BROTLI_LITTLE_ENDIAN) {
-    return *((const uint16_t*)in);
-  } else if (BROTLI_BIG_ENDIAN) {
-    uint16_t value = *((const uint16_t*)in);
-    return (uint16_t)(((value & 0xFFU) << 8) | ((value & 0xFF00U) >> 8));
-  } else {
-    return (uint16_t)(in[0] | (in[1] << 8));
-  }
-}
-
-static BROTLI_INLINE uint32_t BrotliLoad32LE(const uint8_t* in) {
-  if (BROTLI_LITTLE_ENDIAN) {
-    return *((const uint32_t*)in);
-  } else if (BROTLI_BIG_ENDIAN) {
-    uint32_t value = *((const uint32_t*)in);
-    return ((value & 0xFFU) << 24) | ((value & 0xFF00U) << 8) |
-        ((value & 0xFF0000U) >> 8) | ((value & 0xFF000000U) >> 24);
-  } else {
-    uint32_t value = (uint32_t)(*(in++));
-    value |= (uint32_t)(*(in++)) << 8;
-    value |= (uint32_t)(*(in++)) << 16;
-    value |= (uint32_t)(*(in++)) << 24;
-    return value;
-  }
-}
-
-#if (BROTLI_64_BITS)
-static BROTLI_INLINE uint64_t BrotliLoad64LE(const uint8_t* in) {
-  if (BROTLI_LITTLE_ENDIAN) {
-    return *((const uint64_t*)in);
-  } else if (BROTLI_BIG_ENDIAN) {
-    uint64_t value = *((const uint64_t*)in);
-    return
-        ((value & 0xFFU) << 56) |
-        ((value & 0xFF00U) << 40) |
-        ((value & 0xFF0000U) << 24) |
-        ((value & 0xFF000000U) << 8) |
-        ((value & 0xFF00000000U) >> 8) |
-        ((value & 0xFF0000000000U) >> 24) |
-        ((value & 0xFF000000000000U) >> 40) |
-        ((value & 0xFF00000000000000U) >> 56);
-  } else {
-    uint64_t value = (uint64_t)(*(in++));
-    value |= (uint64_t)(*(in++)) << 8;
-    value |= (uint64_t)(*(in++)) << 16;
-    value |= (uint64_t)(*(in++)) << 24;
-    value |= (uint64_t)(*(in++)) << 32;
-    value |= (uint64_t)(*(in++)) << 40;
-    value |= (uint64_t)(*(in++)) << 48;
-    value |= (uint64_t)(*(in++)) << 56;
-    return value;
-  }
-}
-#endif
-
-/* Guarantees that there are at least n_bits + 1 bits in accumulator.
+/* Guarantees that there are at least |n_bits| + 1 bits in accumulator.
    Precondition: accumulator contains at least 1 bit.
-   n_bits should be in the range [1..24] for regular build. For portable
+   |n_bits| should be in the range [1..24] for regular build. For portable
    non-64-bit little-endian build only 16 bits are safe to request. */
 static BROTLI_INLINE void BrotliFillBitWindow(
     BrotliBitReader* const br, uint32_t n_bits) {
 #if (BROTLI_64_BITS)
-  if (!BROTLI_ALIGNED_READ && IS_CONSTANT(n_bits) && (n_bits <= 8)) {
+  if (!BROTLI_ALIGNED_READ && BROTLI_IS_CONSTANT(n_bits) && (n_bits <= 8)) {
     if (br->bit_pos_ >= 56) {
       br->val_ >>= 56;
       br->bit_pos_ ^= 56;  /* here same as -= 56 because of the if condition */
-      br->val_ |= BrotliLoad64LE(br->next_in) << 8;
+      br->val_ |= BROTLI_UNALIGNED_LOAD64LE(br->next_in) << 8;
       br->avail_in -= 7;
       br->next_in += 7;
     }
-  } else if (!BROTLI_ALIGNED_READ && IS_CONSTANT(n_bits) && (n_bits <= 16)) {
+  } else if (
+      !BROTLI_ALIGNED_READ && BROTLI_IS_CONSTANT(n_bits) && (n_bits <= 16)) {
     if (br->bit_pos_ >= 48) {
       br->val_ >>= 48;
       br->bit_pos_ ^= 48;  /* here same as -= 48 because of the if condition */
-      br->val_ |= BrotliLoad64LE(br->next_in) << 16;
+      br->val_ |= BROTLI_UNALIGNED_LOAD64LE(br->next_in) << 16;
       br->avail_in -= 6;
       br->next_in += 6;
     }
@@ -181,17 +127,17 @@ static BROTLI_INLINE void BrotliFillBitWindow(
     if (br->bit_pos_ >= 32) {
       br->val_ >>= 32;
       br->bit_pos_ ^= 32;  /* here same as -= 32 because of the if condition */
-      br->val_ |= ((uint64_t)BrotliLoad32LE(br->next_in)) << 32;
+      br->val_ |= ((uint64_t)BROTLI_UNALIGNED_LOAD32LE(br->next_in)) << 32;
       br->avail_in -= BROTLI_SHORT_FILL_BIT_WINDOW_READ;
       br->next_in += BROTLI_SHORT_FILL_BIT_WINDOW_READ;
     }
   }
 #else
-  if (!BROTLI_ALIGNED_READ && IS_CONSTANT(n_bits) && (n_bits <= 8)) {
+  if (!BROTLI_ALIGNED_READ && BROTLI_IS_CONSTANT(n_bits) && (n_bits <= 8)) {
     if (br->bit_pos_ >= 24) {
       br->val_ >>= 24;
       br->bit_pos_ ^= 24;  /* here same as -= 24 because of the if condition */
-      br->val_ |= BrotliLoad32LE(br->next_in) << 8;
+      br->val_ |= BROTLI_UNALIGNED_LOAD32LE(br->next_in) << 8;
       br->avail_in -= 3;
       br->next_in += 3;
     }
@@ -199,7 +145,7 @@ static BROTLI_INLINE void BrotliFillBitWindow(
     if (br->bit_pos_ >= 16) {
       br->val_ >>= 16;
       br->bit_pos_ ^= 16;  /* here same as -= 16 because of the if condition */
-      br->val_ |= ((uint32_t)BrotliLoad16LE(br->next_in)) << 16;
+      br->val_ |= ((uint32_t)BROTLI_UNALIGNED_LOAD16LE(br->next_in)) << 16;
       br->avail_in -= BROTLI_SHORT_FILL_BIT_WINDOW_READ;
       br->next_in += BROTLI_SHORT_FILL_BIT_WINDOW_READ;
     }
@@ -213,7 +159,8 @@ static BROTLI_INLINE void BrotliFillBitWindow16(BrotliBitReader* const br) {
   BrotliFillBitWindow(br, 17);
 }
 
-/* Pulls one byte of input to accumulator. */
+/* Tries to pull one byte of input to accumulator.
+   Returns BROTLI_FALSE if there is no input available. */
 static BROTLI_INLINE BROTLI_BOOL BrotliPullByte(BrotliBitReader* const br) {
   if (br->avail_in == 0) {
     return BROTLI_FALSE;
@@ -232,7 +179,8 @@ static BROTLI_INLINE BROTLI_BOOL BrotliPullByte(BrotliBitReader* const br) {
 
 /* Returns currently available bits.
    The number of valid bits could be calculated by BrotliGetAvailableBits. */
-static BROTLI_INLINE reg_t BrotliGetBitsUnmasked(BrotliBitReader* const br) {
+static BROTLI_INLINE brotli_reg_t BrotliGetBitsUnmasked(
+    BrotliBitReader* const br) {
   return br->val_ >> br->bit_pos_;
 }
 
@@ -244,15 +192,16 @@ static BROTLI_INLINE uint32_t BrotliGet16BitsUnmasked(
   return (uint32_t)BrotliGetBitsUnmasked(br);
 }
 
-/* Returns the specified number of bits from |br| without advancing bit pos. */
+/* Returns the specified number of bits from |br| without advancing bit
+   position. */
 static BROTLI_INLINE uint32_t BrotliGetBits(
     BrotliBitReader* const br, uint32_t n_bits) {
   BrotliFillBitWindow(br, n_bits);
   return (uint32_t)BrotliGetBitsUnmasked(br) & BitMask(n_bits);
 }
 
-/* Tries to peek the specified amount of bits. Returns 0, if there is not
-   enough input. */
+/* Tries to peek the specified amount of bits. Returns BROTLI_FALSE, if there
+   is not enough input. */
 static BROTLI_INLINE BROTLI_BOOL BrotliSafeGetBits(
     BrotliBitReader* const br, uint32_t n_bits, uint32_t* val) {
   while (BrotliGetAvailableBits(br) < n_bits) {
@@ -264,7 +213,7 @@ static BROTLI_INLINE BROTLI_BOOL BrotliSafeGetBits(
   return BROTLI_TRUE;
 }
 
-/* Advances the bit pos by n_bits. */
+/* Advances the bit pos by |n_bits|. */
 static BROTLI_INLINE void BrotliDropBits(
     BrotliBitReader* const br, uint32_t n_bits) {
   br->bit_pos_ += n_bits;
@@ -284,12 +233,12 @@ static BROTLI_INLINE void BrotliBitReaderUnload(BrotliBitReader* br) {
 }
 
 /* Reads the specified number of bits from |br| and advances the bit pos.
-   Precondition: accumulator MUST contain at least n_bits. */
+   Precondition: accumulator MUST contain at least |n_bits|. */
 static BROTLI_INLINE void BrotliTakeBits(
   BrotliBitReader* const br, uint32_t n_bits, uint32_t* val) {
   *val = (uint32_t)BrotliGetBitsUnmasked(br) & BitMask(n_bits);
   BROTLI_LOG(("[BrotliReadBits]  %d %d %d val: %6x\n",
-      (int)br->avail_in, (int)br->bit_pos_, n_bits, (int)*val));
+      (int)br->avail_in, (int)br->bit_pos_, (int)n_bits, (int)*val));
   BrotliDropBits(br, n_bits);
 }
 
@@ -313,8 +262,8 @@ static BROTLI_INLINE uint32_t BrotliReadBits(
   }
 }
 
-/* Tries to read the specified amount of bits. Returns 0, if there is not
-   enough input. n_bits MUST be positive. */
+/* Tries to read the specified amount of bits. Returns BROTLI_FALSE, if there
+   is not enough input. |n_bits| MUST be positive. */
 static BROTLI_INLINE BROTLI_BOOL BrotliSafeReadBits(
     BrotliBitReader* const br, uint32_t n_bits, uint32_t* val) {
   while (BrotliGetAvailableBits(br) < n_bits) {
@@ -338,7 +287,7 @@ static BROTLI_INLINE BROTLI_BOOL BrotliJumpToByteBoundary(BrotliBitReader* br) {
 }
 
 /* Copies remaining input bytes stored in the bit reader to the output. Value
-   num may not be larger than BrotliGetRemainingBytes. The bit reader must be
+   |num| may not be larger than BrotliGetRemainingBytes. The bit reader must be
    warmed up again after this. */
 static BROTLI_INLINE void BrotliCopyBytes(uint8_t* dest,
                                           BrotliBitReader* br, size_t num) {

modules/brotli/dec/decode.h

@@ -1,162 +0,0 @@
-/* Copyright 2013 Google Inc. All Rights Reserved.
-
-   Distributed under MIT license.
-   See file LICENSE for detail or copy at https://opensource.org/licenses/MIT
-*/
-
-/* API for Brotli decompression */
-
-#ifndef BROTLI_DEC_DECODE_H_
-#define BROTLI_DEC_DECODE_H_
-
-#include "./types.h"
-
-#if defined(__cplusplus) || defined(c_plusplus)
-extern "C" {
-#endif
-
-typedef struct BrotliStateStruct BrotliState;
-
-typedef enum {
-  /* Decoding error, e.g. corrupt input or memory allocation problem */
-  BROTLI_RESULT_ERROR = 0,
-  /* Decoding successfully completed */
-  BROTLI_RESULT_SUCCESS = 1,
-  /* Partially done; should be called again with more input */
-  BROTLI_RESULT_NEEDS_MORE_INPUT = 2,
-  /* Partially done; should be called again with more output */
-  BROTLI_RESULT_NEEDS_MORE_OUTPUT = 3
-} BrotliResult;
-
-#define BROTLI_ERROR_CODES_LIST(BROTLI_ERROR_CODE, SEPARATOR)              \
-  BROTLI_ERROR_CODE(_, NO_ERROR, 0) SEPARATOR                              \
-  /* Same as BrotliResult values */                                        \
-  BROTLI_ERROR_CODE(_, SUCCESS, 1) SEPARATOR                               \
-  BROTLI_ERROR_CODE(_, NEEDS_MORE_INPUT, 2) SEPARATOR                      \
-  BROTLI_ERROR_CODE(_, NEEDS_MORE_OUTPUT, 3) SEPARATOR                     \
-                                                                           \
-  /* Errors caused by invalid input */                                     \
-  BROTLI_ERROR_CODE(_ERROR_FORMAT_, EXUBERANT_NIBBLE, -1) SEPARATOR        \
-  BROTLI_ERROR_CODE(_ERROR_FORMAT_, RESERVED, -2) SEPARATOR                \
-  BROTLI_ERROR_CODE(_ERROR_FORMAT_, EXUBERANT_META_NIBBLE, -3) SEPARATOR   \
-  BROTLI_ERROR_CODE(_ERROR_FORMAT_, SIMPLE_HUFFMAN_ALPHABET, -4) SEPARATOR \
-  BROTLI_ERROR_CODE(_ERROR_FORMAT_, SIMPLE_HUFFMAN_SAME, -5) SEPARATOR     \
-  BROTLI_ERROR_CODE(_ERROR_FORMAT_, CL_SPACE, -6) SEPARATOR                \
-  BROTLI_ERROR_CODE(_ERROR_FORMAT_, HUFFMAN_SPACE, -7) SEPARATOR           \
-  BROTLI_ERROR_CODE(_ERROR_FORMAT_, CONTEXT_MAP_REPEAT, -8) SEPARATOR      \
-  BROTLI_ERROR_CODE(_ERROR_FORMAT_, BLOCK_LENGTH_1, -9) SEPARATOR          \
-  BROTLI_ERROR_CODE(_ERROR_FORMAT_, BLOCK_LENGTH_2, -10) SEPARATOR         \
-  BROTLI_ERROR_CODE(_ERROR_FORMAT_, TRANSFORM, -11) SEPARATOR              \
-  BROTLI_ERROR_CODE(_ERROR_FORMAT_, DICTIONARY, -12) SEPARATOR             \
-  BROTLI_ERROR_CODE(_ERROR_FORMAT_, WINDOW_BITS, -13) SEPARATOR            \
-  BROTLI_ERROR_CODE(_ERROR_FORMAT_, PADDING_1, -14) SEPARATOR              \
-  BROTLI_ERROR_CODE(_ERROR_FORMAT_, PADDING_2, -15) SEPARATOR              \
-                                                                           \
-  /* -16..-20 codes are reserved */                                        \
-                                                                           \
-  /* Memory allocation problems */                                         \
-  BROTLI_ERROR_CODE(_ERROR_ALLOC_, CONTEXT_MODES, -21) SEPARATOR           \
-  /* Literal, insert and distance trees together */                        \
-  BROTLI_ERROR_CODE(_ERROR_ALLOC_, TREE_GROUPS, -22) SEPARATOR             \
-  /* -23..-24 codes are reserved for distinct tree groups */               \
-  BROTLI_ERROR_CODE(_ERROR_ALLOC_, CONTEXT_MAP, -25) SEPARATOR             \
-  BROTLI_ERROR_CODE(_ERROR_ALLOC_, RING_BUFFER_1, -26) SEPARATOR           \
-  BROTLI_ERROR_CODE(_ERROR_ALLOC_, RING_BUFFER_2, -27) SEPARATOR           \
-  /* -28..-29 codes are reserved for dynamic ringbuffer allocation */      \
-  BROTLI_ERROR_CODE(_ERROR_ALLOC_, BLOCK_TYPE_TREES, -30) SEPARATOR        \
-                                                                           \
-  /* "Impossible" states */                                                \
-  BROTLI_ERROR_CODE(_ERROR_, UNREACHABLE, -31)
-
-typedef enum {
-#define _BROTLI_COMMA ,
-#define _BROTLI_ERROR_CODE_ENUM_ITEM(PREFIX, NAME, CODE) \
-    BROTLI ## PREFIX ## NAME = CODE
-  BROTLI_ERROR_CODES_LIST(_BROTLI_ERROR_CODE_ENUM_ITEM, _BROTLI_COMMA)
-#undef _BROTLI_ERROR_CODE_ENUM_ITEM
-#undef _BROTLI_COMMA
-} BrotliErrorCode;
-
-#define BROTLI_LAST_ERROR_CODE BROTLI_ERROR_UNREACHABLE
-
-/* Creates the instance of BrotliState and initializes it. |alloc_func| and
-   |free_func| MUST be both zero or both non-zero. In the case they are both
-   zero, default memory allocators are used. |opaque| is passed to |alloc_func|
-   and |free_func| when they are called. */
-BrotliState* BrotliCreateState(
-    brotli_alloc_func alloc_func, brotli_free_func free_func, void* opaque);
-
-/* Deinitializes and frees BrotliState instance. */
-void BrotliDestroyState(BrotliState* state);
-
-/* Sets |*decoded_size| to the decompressed size of the given encoded stream.
-   This function only works if the encoded buffer has a single meta block,
-   or if it has two meta-blocks, where the first is uncompressed and the
-   second is empty.
-   Returns 1 on success, 0 on failure. */
-int BrotliDecompressedSize(size_t encoded_size,
-                           const uint8_t* encoded_buffer,
-                           size_t* decoded_size);
-
-/* Decompresses the data in |encoded_buffer| into |decoded_buffer|, and sets
-   |*decoded_size| to the decompressed length. */
-BrotliResult BrotliDecompressBuffer(size_t encoded_size,
-                                    const uint8_t* encoded_buffer,
-                                    size_t* decoded_size,
-                                    uint8_t* decoded_buffer);
-
-/* Decompresses the data. Supports partial input and output.
-
-   Must be called with an allocated input buffer in |*next_in| and an allocated
-   output buffer in |*next_out|. The values |*available_in| and |*available_out|
-   must specify the allocated size in |*next_in| and |*next_out| respectively.
-
-   After each call, |*available_in| will be decremented by the amount of input
-   bytes consumed, and the |*next_in| pointer will be incremented by that
-   amount. Similarly, |*available_out| will be decremented by the amount of
-   output bytes written, and the |*next_out| pointer will be incremented by that
-   amount. |total_out|, if it is not a null-pointer, will be set to the number
-   of bytes decompressed since the last state initialization.
-
-   Input is never overconsumed, so |next_in| and |available_in| could be passed
-   to the next consumer after decoding is complete. */
-BrotliResult BrotliDecompressStream(size_t* available_in,
-                                    const uint8_t** next_in,
-                                    size_t* available_out,
-                                    uint8_t** next_out,
-                                    size_t* total_out,
-                                    BrotliState* s);
-
-/* Fills the new state with a dictionary for LZ77, warming up the ringbuffer,
-   e.g. for custom static dictionaries for data formats.
-   Not to be confused with the built-in transformable dictionary of Brotli.
-   |size| should be less or equal to 2^24 (16MiB), otherwise the dictionary will
-   be ignored. The dictionary must exist in memory until decoding is done and
-   is owned by the caller. To use:
-    1) Allocate and initialize state with BrotliCreateState
-    2) Use BrotliSetCustomDictionary
-    3) Use BrotliDecompressStream
-    4) Clean up and free state with BrotliDestroyState
-*/
-void BrotliSetCustomDictionary(
-    size_t size, const uint8_t* dict, BrotliState* s);
-
-/* Returns 1, if s is in a state where we have not read any input bytes yet,
-   and 0 otherwise */
-int BrotliStateIsStreamStart(const BrotliState* s);
-
-/* Returns 1, if s is in a state where we reached the end of the input and
-   produced all of the output, and 0 otherwise. */
-int BrotliStateIsStreamEnd(const BrotliState* s);
-
-/* Returns detailed error code after BrotliDecompressStream returns
-   BROTLI_RESULT_ERROR. */
-BrotliErrorCode BrotliGetErrorCode(const BrotliState* s);
-
-const char* BrotliErrorString(BrotliErrorCode c);
-
-#if defined(__cplusplus) || defined(c_plusplus)
-} /* extern "C" */
-#endif
-
-#endif  /* BROTLI_DEC_DECODE_H_ */

modules/brotli/dec/dictionary.h

@@ -1,38 +0,0 @@
-/* Copyright 2013 Google Inc. All Rights Reserved.
-
-   Distributed under MIT license.
-   See file LICENSE for detail or copy at https://opensource.org/licenses/MIT
-*/
-
-/* Collection of static dictionary words. */
-
-#ifndef BROTLI_DEC_DICTIONARY_H_
-#define BROTLI_DEC_DICTIONARY_H_
-
-#include "./types.h"
-
-#if defined(__cplusplus) || defined(c_plusplus)
-extern "C" {
-#endif
-
-extern const uint8_t kBrotliDictionary[122784];
-
-static const uint32_t kBrotliDictionaryOffsetsByLength[] = {
-  0, 0, 0, 0, 0, 4096, 9216, 21504, 35840, 44032, 53248, 63488, 74752, 87040,
-  93696, 100864, 104704, 106752, 108928, 113536, 115968, 118528, 119872, 121280,
-  122016
-};
-
-static const uint8_t kBrotliDictionarySizeBitsByLength[] = {
-  0, 0, 0, 0, 10, 10, 11, 11, 10, 10, 10, 10, 10,
-  9, 9, 8, 7, 7,  8,  7,  7,  6,  6,  5,  5,
-};
-
-static const int kBrotliMinDictionaryWordLength = 4;
-static const int kBrotliMaxDictionaryWordLength = 24;
-
-#if defined(__cplusplus) || defined(c_plusplus)
-}  /* extern "C" */
-#endif
-
-#endif  /* BROTLI_DEC_DICTIONARY_H_ */

modules/brotli/dec/huffman.c

@@ -11,8 +11,8 @@
 #include <string.h>  /* memcpy, memset */
 
 #include "../common/constants.h"
+#include "../common/platform.h"
 #include <brotli/types.h>
-#include "./port.h"
 
 #if defined(__cplusplus) || defined(c_plusplus)
 extern "C" {
@@ -20,9 +20,9 @@ extern "C" {
 
 #define BROTLI_REVERSE_BITS_MAX 8
 
-#ifdef BROTLI_RBIT
+#if defined(BROTLI_RBIT)
 #define BROTLI_REVERSE_BITS_BASE \
-  ((sizeof(reg_t) << 3) - BROTLI_REVERSE_BITS_MAX)
+  ((sizeof(brotli_reg_t) << 3) - BROTLI_REVERSE_BITS_MAX)
 #else
 #define BROTLI_REVERSE_BITS_BASE 0
 static uint8_t kReverseBits[1 << BROTLI_REVERSE_BITS_MAX] = {
@@ -62,13 +62,13 @@ static uint8_t kReverseBits[1 << BROTLI_REVERSE_BITS_MAX] = {
 #endif  /* BROTLI_RBIT */
 
 #define BROTLI_REVERSE_BITS_LOWEST \
-  ((reg_t)1 << (BROTLI_REVERSE_BITS_MAX - 1 + BROTLI_REVERSE_BITS_BASE))
+  ((brotli_reg_t)1 << (BROTLI_REVERSE_BITS_MAX - 1 + BROTLI_REVERSE_BITS_BASE))
 
 /* Returns reverse(num >> BROTLI_REVERSE_BITS_BASE, BROTLI_REVERSE_BITS_MAX),
    where reverse(value, len) is the bit-wise reversal of the len least
    significant bits of value. */
-static BROTLI_INLINE reg_t BrotliReverseBits(reg_t num) {
-#ifdef BROTLI_RBIT
+static BROTLI_INLINE brotli_reg_t BrotliReverseBits(brotli_reg_t num) {
+#if defined(BROTLI_RBIT)
   return BROTLI_RBIT(num);
 #else
   return kReverseBits[num];
@@ -86,9 +86,9 @@ static BROTLI_INLINE void ReplicateValue(HuffmanCode* table,
   } while (end > 0);
 }
 
-/* Returns the table width of the next 2nd level table. count is the histogram
-   of bit lengths for the remaining symbols, len is the code length of the next
-   processed symbol */
+/* Returns the table width of the next 2nd level table. |count| is the histogram
+   of bit lengths for the remaining symbols, |len| is the code length of the
+   next processed symbol. */
 static BROTLI_INLINE int NextTableBitSize(const uint16_t* const count,
                                           int len, int root_bits) {
   int left = 1 << (len - root_bits);
@@ -104,12 +104,12 @@ static BROTLI_INLINE int NextTableBitSize(const uint16_t* const count,
 void BrotliBuildCodeLengthsHuffmanTable(HuffmanCode* table,
                                         const uint8_t* const code_lengths,
                                         uint16_t* count) {
-  HuffmanCode code; /* current table entry */
-  int symbol;       /* symbol index in original or sorted table */
-  reg_t key;        /* prefix code */
-  reg_t key_step;   /* prefix code addend */
-  int step;         /* step size to replicate values in current table */
-  int table_size;   /* size of current table */
+  HuffmanCode code;       /* current table entry */
+  int symbol;             /* symbol index in original or sorted table */
+  brotli_reg_t key;       /* prefix code */
+  brotli_reg_t key_step;  /* prefix code addend */
+  int step;               /* step size to replicate values in current table */
+  int table_size;         /* size of current table */
   int sorted[BROTLI_CODE_LENGTH_CODES];  /* symbols sorted by code length */
   /* offsets in sorted table for each length */
   int offset[BROTLI_HUFFMAN_MAX_CODE_LENGTH_CODE_LENGTH + 1];
@@ -118,7 +118,7 @@ void BrotliBuildCodeLengthsHuffmanTable(HuffmanCode* table,
   BROTLI_DCHECK(BROTLI_HUFFMAN_MAX_CODE_LENGTH_CODE_LENGTH <=
                 BROTLI_REVERSE_BITS_MAX);
 
-  /* generate offsets into sorted symbol table by code length */
+  /* Generate offsets into sorted symbol table by code length. */
   symbol = -1;
   bits = 1;
   BROTLI_REPEAT(BROTLI_HUFFMAN_MAX_CODE_LENGTH_CODE_LENGTH, {
@@ -129,7 +129,7 @@ void BrotliBuildCodeLengthsHuffmanTable(HuffmanCode* table,
   /* Symbols with code length 0 are placed after all other symbols. */
   offset[0] = BROTLI_CODE_LENGTH_CODES - 1;
 
-  /* sort symbols by length, by symbol order within each length */
+  /* Sort symbols by length, by symbol order within each length. */
   symbol = BROTLI_CODE_LENGTH_CODES;
   do {
     BROTLI_REPEAT(6, {
@@ -142,24 +142,22 @@ void BrotliBuildCodeLengthsHuffmanTable(HuffmanCode* table,
 
   /* Special case: all symbols but one have 0 code length. */
   if (offset[0] == 0) {
-    code.bits = 0;
-    code.value = (uint16_t)sorted[0];
-    for (key = 0; key < (reg_t)table_size; ++key) {
+    code = ConstructHuffmanCode(0, (uint16_t)sorted[0]);
+    for (key = 0; key < (brotli_reg_t)table_size; ++key) {
       table[key] = code;
     }
     return;
   }
 
-  /* fill in table */
+  /* Fill in table. */
   key = 0;
   key_step = BROTLI_REVERSE_BITS_LOWEST;
   symbol = 0;
   bits = 1;
   step = 2;
   do {
-    code.bits = (uint8_t)bits;
     for (bits_count = count[bits]; bits_count != 0; --bits_count) {
-      code.value = (uint16_t)sorted[symbol++];
+      code = ConstructHuffmanCode((uint8_t)bits, (uint16_t)sorted[symbol++]);
       ReplicateValue(&table[BrotliReverseBits(key)], step, table_size, code);
       key += key_step;
     }
@@ -172,18 +170,18 @@ uint32_t BrotliBuildHuffmanTable(HuffmanCode* root_table,
                                  int root_bits,
                                  const uint16_t* const symbol_lists,
                                  uint16_t* count) {
-  HuffmanCode code;    /* current table entry */
-  HuffmanCode* table;  /* next available space in table */
-  int len;             /* current code length */
-  int symbol;          /* symbol index in original or sorted table */
-  reg_t key;           /* prefix code */
-  reg_t key_step;      /* prefix code addend */
-  reg_t sub_key;       /* 2nd level table prefix code */
-  reg_t sub_key_step;  /* 2nd level table prefix code addend */
-  int step;            /* step size to replicate values in current table */
-  int table_bits;      /* key length of current table */
-  int table_size;      /* size of current table */
-  int total_size;      /* sum of root table size and 2nd level table sizes */
+  HuffmanCode code;       /* current table entry */
+  HuffmanCode* table;     /* next available space in table */
+  int len;                /* current code length */
+  int symbol;             /* symbol index in original or sorted table */
+  brotli_reg_t key;       /* prefix code */
+  brotli_reg_t key_step;  /* prefix code addend */
+  brotli_reg_t sub_key;   /* 2nd level table prefix code */
+  brotli_reg_t sub_key_step;  /* 2nd level table prefix code addend */
+  int step;               /* step size to replicate values in current table */
+  int table_bits;         /* key length of current table */
+  int table_size;         /* size of current table */
+  int total_size;         /* sum of root table size and 2nd level table sizes */
   int max_length = -1;
   int bits;
   int bits_count;
@@ -200,9 +198,8 @@ uint32_t BrotliBuildHuffmanTable(HuffmanCode* root_table,
   table_size = 1 << table_bits;
   total_size = table_size;
 
-  /* fill in root table */
-  /* let's reduce the table size to a smaller size if possible, and */
-  /* create the repetitions by memcpy if possible in the coming loop */
+  /* Fill in the root table. Reduce the table size to if possible,
+     and create the repetitions by memcpy. */
   if (table_bits > max_length) {
     table_bits = max_length;
     table_size = 1 << table_bits;
@@ -212,11 +209,10 @@ uint32_t BrotliBuildHuffmanTable(HuffmanCode* root_table,
   bits = 1;
   step = 2;
   do {
-    code.bits = (uint8_t)bits;
     symbol = bits - (BROTLI_HUFFMAN_MAX_CODE_LENGTH + 1);
     for (bits_count = count[bits]; bits_count != 0; --bits_count) {
       symbol = symbol_lists[symbol];
-      code.value = (uint16_t)symbol;
+      code = ConstructHuffmanCode((uint8_t)bits, (uint16_t)symbol);
       ReplicateValue(&table[BrotliReverseBits(key)], step, table_size, code);
       key += key_step;
     }
@@ -224,15 +220,14 @@ uint32_t BrotliBuildHuffmanTable(HuffmanCode* root_table,
     key_step >>= 1;
   } while (++bits <= table_bits);
 
-  /* if root_bits != table_bits we only created one fraction of the */
-  /* table, and we need to replicate it now. */
+  /* If root_bits != table_bits then replicate to fill the remaining slots. */
   while (total_size != table_size) {
     memcpy(&table[table_size], &table[0],
            (size_t)table_size * sizeof(table[0]));
     table_size <<= 1;
   }
 
-  /* fill in 2nd level tables and add pointers to root table */
+  /* Fill in 2nd level tables and add pointers to root table. */
   key_step = BROTLI_REVERSE_BITS_LOWEST >> (root_bits - 1);
   sub_key = (BROTLI_REVERSE_BITS_LOWEST << 1);
   sub_key_step = BROTLI_REVERSE_BITS_LOWEST;
@@ -246,14 +241,13 @@ uint32_t BrotliBuildHuffmanTable(HuffmanCode* root_table,
         total_size += table_size;
         sub_key = BrotliReverseBits(key);
         key += key_step;
-        root_table[sub_key].bits = (uint8_t)(table_bits + root_bits);
-        root_table[sub_key].value =
-            (uint16_t)(((size_t)(table - root_table)) - sub_key);
+        root_table[sub_key] = ConstructHuffmanCode(
+            (uint8_t)(table_bits + root_bits),
+            (uint16_t)(((size_t)(table - root_table)) - sub_key));
         sub_key = 0;
       }
-      code.bits = (uint8_t)(len - root_bits);
       symbol = symbol_lists[symbol];
-      code.value = (uint16_t)symbol;
+      code = ConstructHuffmanCode((uint8_t)(len - root_bits), (uint16_t)symbol);
       ReplicateValue(
           &table[BrotliReverseBits(sub_key)], step, table_size, code);
       sub_key += sub_key_step;
@@ -272,35 +266,28 @@ uint32_t BrotliBuildSimpleHuffmanTable(HuffmanCode* table,
   const uint32_t goal_size = 1U << root_bits;
   switch (num_symbols) {
     case 0:
-      table[0].bits = 0;
-      table[0].value = val[0];
+      table[0] = ConstructHuffmanCode(0, val[0]);
       break;
     case 1:
-      table[0].bits = 1;
-      table[1].bits = 1;
       if (val[1] > val[0]) {
-        table[0].value = val[0];
-        table[1].value = val[1];
+        table[0] = ConstructHuffmanCode(1, val[0]);
+        table[1] = ConstructHuffmanCode(1, val[1]);
       } else {
-        table[0].value = val[1];
-        table[1].value = val[0];
+        table[0] = ConstructHuffmanCode(1, val[1]);
+        table[1] = ConstructHuffmanCode(1, val[0]);
       }
       table_size = 2;
       break;
     case 2:
-      table[0].bits = 1;
-      table[0].value = val[0];
-      table[2].bits = 1;
-      table[2].value = val[0];
+      table[0] = ConstructHuffmanCode(1, val[0]);
+      table[2] = ConstructHuffmanCode(1, val[0]);
       if (val[2] > val[1]) {
-        table[1].value = val[1];
-        table[3].value = val[2];
+        table[1] = ConstructHuffmanCode(2, val[1]);
+        table[3] = ConstructHuffmanCode(2, val[2]);
       } else {
-        table[1].value = val[2];
-        table[3].value = val[1];
+        table[1] = ConstructHuffmanCode(2, val[2]);
+        table[3] = ConstructHuffmanCode(2, val[1]);
       }
-      table[1].bits = 2;
-      table[3].bits = 2;
       table_size = 4;
       break;
     case 3: {
@@ -314,33 +301,27 @@ uint32_t BrotliBuildSimpleHuffmanTable(HuffmanCode* table,
           }
         }
       }
-      for (i = 0; i < 4; ++i) {
-        table[i].bits = 2;
-      }
-      table[0].value = val[0];
-      table[2].value = val[1];
-      table[1].value = val[2];
-      table[3].value = val[3];
+      table[0] = ConstructHuffmanCode(2, val[0]);
+      table[2] = ConstructHuffmanCode(2, val[1]);
+      table[1] = ConstructHuffmanCode(2, val[2]);
+      table[3] = ConstructHuffmanCode(2, val[3]);
       table_size = 4;
       break;
     }
     case 4: {
-      int i;
       if (val[3] < val[2]) {
         uint16_t t = val[3];
         val[3] = val[2];
         val[2] = t;
       }
-      for (i = 0; i < 7; ++i) {
-        table[i].value = val[0];
-        table[i].bits = (uint8_t)(1 + (i & 1));
-      }
-      table[1].value = val[1];
-      table[3].value = val[2];
-      table[5].value = val[1];
-      table[7].value = val[3];
-      table[3].bits = 3;
-      table[7].bits = 3;
+      table[0] = ConstructHuffmanCode(1, val[0]);
+      table[1] = ConstructHuffmanCode(2, val[1]);
+      table[2] = ConstructHuffmanCode(1, val[0]);
+      table[3] = ConstructHuffmanCode(3, val[2]);
+      table[4] = ConstructHuffmanCode(1, val[0]);
+      table[5] = ConstructHuffmanCode(2, val[1]);
+      table[6] = ConstructHuffmanCode(1, val[0]);
+      table[7] = ConstructHuffmanCode(3, val[3]);
       table_size = 8;
       break;
     }

modules/brotli/dec/huffman.h

@@ -9,8 +9,8 @@
 #ifndef BROTLI_DEC_HUFFMAN_H_
 #define BROTLI_DEC_HUFFMAN_H_
 
+#include "../common/platform.h"
 #include <brotli/types.h>
-#include "./port.h"
 
 #if defined(__cplusplus) || defined(c_plusplus)
 extern "C" {
@@ -19,10 +19,11 @@ extern "C" {
 #define BROTLI_HUFFMAN_MAX_CODE_LENGTH 15
 
 /* Maximum possible Huffman table size for an alphabet size of (index * 32),
- * max code length 15 and root table bits 8. */
+   max code length 15 and root table bits 8. */
 static const uint16_t kMaxHuffmanTableSize[] = {
   256, 402, 436, 468, 500, 534, 566, 598, 630, 662, 694, 726, 758, 790, 822,
-  854, 886, 920, 952, 984, 1016, 1048, 1080};
+  854, 886, 920, 952, 984, 1016, 1048, 1080, 1112, 1144, 1176, 1208, 1240, 1272,
+  1304, 1336, 1368, 1400, 1432, 1464, 1496, 1528};
 /* BROTLI_NUM_BLOCK_LEN_SYMBOLS == 26 */
 #define BROTLI_HUFFMAN_MAX_SIZE_26 396
 /* BROTLI_MAX_BLOCK_TYPE_SYMBOLS == 258 */
@@ -32,32 +33,90 @@ static const uint16_t kMaxHuffmanTableSize[] = {
 
 #define BROTLI_HUFFMAN_MAX_CODE_LENGTH_CODE_LENGTH 5
 
+#if ((defined(BROTLI_TARGET_ARMV7) || defined(BROTLI_TARGET_ARMV8_32)) && \
+  BROTLI_GNUC_HAS_ATTRIBUTE(aligned, 2, 7, 0))
+#define BROTLI_HUFFMAN_CODE_FAST_LOAD
+#endif
+
+#if !defined(BROTLI_HUFFMAN_CODE_FAST_LOAD)
+/* Do not create this struct directly - use the ConstructHuffmanCode
+ * constructor below! */
 typedef struct {
   uint8_t bits;    /* number of bits used for this symbol */
   uint16_t value;  /* symbol value or table offset */
 } HuffmanCode;
 
+static BROTLI_INLINE HuffmanCode ConstructHuffmanCode(const uint8_t bits,
+    const uint16_t value) {
+  HuffmanCode h;
+  h.bits = bits;
+  h.value = value;
+  return h;
+}
+
+/* Please use the following macros to optimize HuffmanCode accesses in hot
+ * paths.
+ *
+ * For example, assuming |table| contains a HuffmanCode pointer:
+ *
+ *   BROTLI_HC_MARK_TABLE_FOR_FAST_LOAD(table);
+ *   BROTLI_HC_ADJUST_TABLE_INDEX(table, index_into_table);
+ *   *bits = BROTLI_HC_GET_BITS(table);
+ *   *value = BROTLI_HC_GET_VALUE(table);
+ *   BROTLI_HC_ADJUST_TABLE_INDEX(table, offset);
+ *   *bits2 = BROTLI_HC_GET_BITS(table);
+ *   *value2 = BROTLI_HC_GET_VALUE(table);
+ *
+ */
+
+#define BROTLI_HC_MARK_TABLE_FOR_FAST_LOAD(H)
+#define BROTLI_HC_ADJUST_TABLE_INDEX(H, V) H += (V)
+
+/* These must be given a HuffmanCode pointer! */
+#define BROTLI_HC_FAST_LOAD_BITS(H) (H->bits)
+#define BROTLI_HC_FAST_LOAD_VALUE(H) (H->value)
+
+#else /* BROTLI_HUFFMAN_CODE_FAST_LOAD */
+
+typedef BROTLI_ALIGNED(4) uint32_t HuffmanCode;
+
+static BROTLI_INLINE HuffmanCode ConstructHuffmanCode(const uint8_t bits,
+    const uint16_t value) {
+  return ((value & 0xFFFF) << 16) | (bits & 0xFF);
+}
+
+#define BROTLI_HC_MARK_TABLE_FOR_FAST_LOAD(H) uint32_t __fastload_##H = (*H)
+#define BROTLI_HC_ADJUST_TABLE_INDEX(H, V) H += (V); __fastload_##H = (*H)
+
+/* These must be given a HuffmanCode pointer! */
+#define BROTLI_HC_FAST_LOAD_BITS(H) ((__fastload_##H) & 0xFF)
+#define BROTLI_HC_FAST_LOAD_VALUE(H) ((__fastload_##H) >> 16)
+#endif /* BROTLI_HUFFMAN_CODE_FAST_LOAD */
+
 /* Builds Huffman lookup table assuming code lengths are in symbol order. */
 BROTLI_INTERNAL void BrotliBuildCodeLengthsHuffmanTable(HuffmanCode* root_table,
     const uint8_t* const code_lengths, uint16_t* count);
 
-/* Builds Huffman lookup table assuming code lengths are in symbol order. */
-/* Returns size of resulting table. */
+/* Builds Huffman lookup table assuming code lengths are in symbol order.
+   Returns size of resulting table. */
 BROTLI_INTERNAL uint32_t BrotliBuildHuffmanTable(HuffmanCode* root_table,
     int root_bits, const uint16_t* const symbol_lists, uint16_t* count_arg);
 
-/* Builds a simple Huffman table. The num_symbols parameter is to be */
-/* interpreted as follows: 0 means 1 symbol, 1 means 2 symbols, 2 means 3 */
-/* symbols, 3 means 4 symbols with lengths 2,2,2,2, 4 means 4 symbols with */
-/* lengths 1,2,3,3. */
+/* Builds a simple Huffman table. The |num_symbols| parameter is to be
+   interpreted as follows: 0 means 1 symbol, 1 means 2 symbols,
+   2 means 3 symbols, 3 means 4 symbols with lengths [2, 2, 2, 2],
+   4 means 4 symbols with lengths [1, 2, 3, 3]. */
 BROTLI_INTERNAL uint32_t BrotliBuildSimpleHuffmanTable(HuffmanCode* table,
     int root_bits, uint16_t* symbols, uint32_t num_symbols);
 
 /* Contains a collection of Huffman trees with the same alphabet size. */
+/* max_symbol is needed due to simple codes since log2(alphabet_size) could be
+   greater than log2(max_symbol). */
 typedef struct {
   HuffmanCode** htrees;
   HuffmanCode* codes;
   uint16_t alphabet_size;
+  uint16_t max_symbol;
   uint16_t num_htrees;
 } HuffmanTreeGroup;
 

modules/brotli/dec/port.h

@@ -1,168 +0,0 @@
-/* Copyright 2015 Google Inc. All Rights Reserved.
-
-   Distributed under MIT license.
-   See file LICENSE for detail or copy at https://opensource.org/licenses/MIT
-*/
-
-/* Macros for compiler / platform specific features and build options.
-
-   Build options are:
-    * BROTLI_BUILD_32_BIT disables 64-bit optimizations
-    * BROTLI_BUILD_64_BIT forces to use 64-bit optimizations
-    * BROTLI_BUILD_BIG_ENDIAN forces to use big-endian optimizations
-    * BROTLI_BUILD_ENDIAN_NEUTRAL disables endian-aware optimizations
-    * BROTLI_BUILD_LITTLE_ENDIAN forces to use little-endian optimizations
-    * BROTLI_BUILD_MODERN_COMPILER forces to use modern compilers built-ins,
-      features and attributes
-    * BROTLI_BUILD_PORTABLE disables dangerous optimizations, like unaligned
-      read and overlapping memcpy; this reduces decompression speed by 5%
-    * BROTLI_BUILD_NO_RBIT disables "rbit" optimization for ARM CPUs
-    * BROTLI_DEBUG dumps file name and line number when decoder detects stream
-      or memory error
-    * BROTLI_ENABLE_LOG enables asserts and dumps various state information
- */
-
-#ifndef BROTLI_DEC_PORT_H_
-#define BROTLI_DEC_PORT_H_
-
-#if defined(BROTLI_ENABLE_LOG) || defined(BROTLI_DEBUG)
-#include <assert.h>
-#include <stdio.h>
-#endif
-
-#include <brotli/port.h>
-
-#if defined(__arm__) || defined(__thumb__) || \
-    defined(_M_ARM) || defined(_M_ARMT) || defined(__ARM64_ARCH_8__)
-#define BROTLI_TARGET_ARM
-#if (defined(__ARM_ARCH) && (__ARM_ARCH == 7)) || \
-    (defined(M_ARM) && (M_ARM == 7))
-#define BROTLI_TARGET_ARMV7
-#endif  /* ARMv7 */
-#if defined(__aarch64__) || defined(__ARM64_ARCH_8__)
-#define BROTLI_TARGET_ARMV8
-#endif  /* ARMv8 */
-#endif  /* ARM */
-
-#if defined(__i386) || defined(_M_IX86)
-#define BROTLI_TARGET_X86
-#endif
-
-#if defined(__x86_64__) || defined(_M_X64)
-#define BROTLI_TARGET_X64
-#endif
-
-#if defined(__PPC64__)
-#define BROTLI_TARGET_POWERPC64
-#endif
-
-#ifdef BROTLI_BUILD_PORTABLE
-#define BROTLI_ALIGNED_READ (!!1)
-#elif defined(BROTLI_TARGET_X86) || defined(BROTLI_TARGET_X64) || \
-     defined(BROTLI_TARGET_ARMV7) || defined(BROTLI_TARGET_ARMV8)
-/* Allow unaligned read only for white-listed CPUs. */
-#define BROTLI_ALIGNED_READ (!!0)
-#else
-#define BROTLI_ALIGNED_READ (!!1)
-#endif
-
-/* IS_CONSTANT macros returns true for compile-time constant expressions. */
-#if BROTLI_MODERN_COMPILER || __has_builtin(__builtin_constant_p)
-#define IS_CONSTANT(x) (!!__builtin_constant_p(x))
-#else
-#define IS_CONSTANT(x) (!!0)
-#endif
-
-#ifdef BROTLI_ENABLE_LOG
-#define BROTLI_DCHECK(x) assert(x)
-#define BROTLI_LOG(x) printf x
-#else
-#define BROTLI_DCHECK(x)
-#define BROTLI_LOG(x)
-#endif
-
-#if defined(BROTLI_DEBUG) || defined(BROTLI_ENABLE_LOG)
-static BROTLI_INLINE void BrotliDump(const char* f, int l, const char* fn) {
-  fprintf(stderr, "%s:%d (%s)\n", f, l, fn);
-  fflush(stderr);
-}
-#define BROTLI_DUMP() BrotliDump(__FILE__, __LINE__, __FUNCTION__)
-#else
-#define BROTLI_DUMP() (void)(0)
-#endif
-
-#if defined(BROTLI_BUILD_64_BIT)
-#define BROTLI_64_BITS 1
-#elif defined(BROTLI_BUILD_32_BIT)
-#define BROTLI_64_BITS 0
-#elif defined(BROTLI_TARGET_X64) || defined(BROTLI_TARGET_ARMV8) || \
-    defined(BROTLI_TARGET_POWERPC64)
-#define BROTLI_64_BITS 1
-#else
-#define BROTLI_64_BITS 0
-#endif
-
-#if (BROTLI_64_BITS)
-#define reg_t uint64_t
-#else
-#define reg_t uint32_t
-#endif
-
-#if defined(BROTLI_BUILD_BIG_ENDIAN)
-#define BROTLI_LITTLE_ENDIAN 0
-#define BROTLI_BIG_ENDIAN 1
-#elif defined(BROTLI_BUILD_LITTLE_ENDIAN)
-#define BROTLI_LITTLE_ENDIAN 1
-#define BROTLI_BIG_ENDIAN 0
-#elif defined(BROTLI_BUILD_ENDIAN_NEUTRAL)
-#define BROTLI_LITTLE_ENDIAN 0
-#define BROTLI_BIG_ENDIAN 0
-#elif defined(__BYTE_ORDER__) && (__BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__)
-#define BROTLI_LITTLE_ENDIAN 1
-#define BROTLI_BIG_ENDIAN 0
-#elif defined(_WIN32)
-/* Win32 can currently always be assumed to be little endian */
-#define BROTLI_LITTLE_ENDIAN 1
-#define BROTLI_BIG_ENDIAN 0
-#else
-#if (defined(__BYTE_ORDER__) && (__BYTE_ORDER__ == __ORDER_BIG_ENDIAN__))
-#define BROTLI_BIG_ENDIAN 1
-#else
-#define BROTLI_BIG_ENDIAN 0
-#endif
-#define BROTLI_LITTLE_ENDIAN 0
-#endif
-
-#define BROTLI_REPEAT(N, X) {     \
-  if ((N & 1) != 0) {X;}          \
-  if ((N & 2) != 0) {X; X;}       \
-  if ((N & 4) != 0) {X; X; X; X;} \
-}
-
-#if (BROTLI_MODERN_COMPILER || defined(__llvm__)) && \
-    !defined(BROTLI_BUILD_NO_RBIT)
-#if defined(BROTLI_TARGET_ARMV7) || defined(BROTLI_TARGET_ARMV8)
-/* TODO: detect ARMv6T2 and enable this code for it. */
-static BROTLI_INLINE reg_t BrotliRBit(reg_t input) {
-  reg_t output;
-  __asm__("rbit %0, %1\n" : "=r"(output) : "r"(input));
-  return output;
-}
-#define BROTLI_RBIT(x) BrotliRBit(x)
-#endif  /* armv7 */
-#endif  /* gcc || clang */
-
-#if defined(BROTLI_TARGET_ARM)
-#define BROTLI_HAS_UBFX (!!1)
-#else
-#define BROTLI_HAS_UBFX (!!0)
-#endif
-
-#define BROTLI_ALLOC(S, L) S->alloc_func(S->memory_manager_opaque, L)
-
-#define BROTLI_FREE(S, X) {                  \
-  S->free_func(S->memory_manager_opaque, X); \
-  X = NULL;                                  \
-}
-
-#endif  /* BROTLI_DEC_PORT_H_ */

modules/brotli/dec/prefix.h

@@ -5,8 +5,7 @@
 */
 
 /* Lookup tables to map prefix codes to value ranges. This is used during
-   decoding of the block lengths, literal insertion lengths and copy lengths.
-*/
+   decoding of the block lengths, literal insertion lengths and copy lengths. */
 
 #ifndef BROTLI_DEC_PREFIX_H_
 #define BROTLI_DEC_PREFIX_H_
@@ -14,8 +13,8 @@
 #include "../common/constants.h"
 #include <brotli/types.h>
 
-/* Represents the range of values belonging to a prefix code: */
-/* [offset, offset + 2^nbits) */
+/* Represents the range of values belonging to a prefix code:
+   [offset, offset + 2^nbits) */
 struct PrefixCodeRange {
   uint16_t offset;
   uint8_t nbits;

modules/brotli/dec/state.c

@@ -15,25 +15,11 @@
 extern "C" {
 #endif
 
-static void* DefaultAllocFunc(void* opaque, size_t size) {
-  BROTLI_UNUSED(opaque);
-  return malloc(size);
-}
-
-static void DefaultFreeFunc(void* opaque, void* address) {
-  BROTLI_UNUSED(opaque);
-  free(address);
-}
-
-void BrotliDecoderStateInit(BrotliDecoderState* s) {
-  BrotliDecoderStateInitWithCustomAllocators(s, 0, 0, 0);
-}
-
-void BrotliDecoderStateInitWithCustomAllocators(BrotliDecoderState* s,
+BROTLI_BOOL BrotliDecoderStateInit(BrotliDecoderState* s,
     brotli_alloc_func alloc_func, brotli_free_func free_func, void* opaque) {
   if (!alloc_func) {
-    s->alloc_func = DefaultAllocFunc;
-    s->free_func = DefaultFreeFunc;
+    s->alloc_func = BrotliDefaultAllocFunc;
+    s->free_func = BrotliDefaultFreeFunc;
     s->memory_manager_opaque = 0;
   } else {
     s->alloc_func = alloc_func;
@@ -45,6 +31,7 @@ void BrotliDecoderStateInitWithCustomAllocators(BrotliDecoderState* s,
 
   BrotliInitBitReader(&s->br);
   s->state = BROTLI_STATE_UNINITED;
+  s->large_window = 0;
   s->substate_metablock_header = BROTLI_STATE_METABLOCK_HEADER_NONE;
   s->substate_tree_group = BROTLI_STATE_TREE_GROUP_NONE;
   s->substate_context_map = BROTLI_STATE_CONTEXT_MAP_NONE;
@@ -53,8 +40,6 @@ void BrotliDecoderStateInitWithCustomAllocators(BrotliDecoderState* s,
   s->substate_decode_uint8 = BROTLI_STATE_DECODE_UINT8_NONE;
   s->substate_read_block_length = BROTLI_STATE_READ_BLOCK_LENGTH_NONE;
 
-  s->dictionary = BrotliGetDictionary();
-
   s->buffer_length = 0;
   s->loop_counter = 0;
   s->pos = 0;
@@ -103,13 +88,18 @@ void BrotliDecoderStateInitWithCustomAllocators(BrotliDecoderState* s,
   s->symbol_lists = &s->symbols_lists_array[BROTLI_HUFFMAN_MAX_CODE_LENGTH + 1];
 
   s->mtf_upper_bound = 63;
+
+  s->dictionary = BrotliGetDictionary();
+  s->transforms = BrotliGetTransforms();
+
+  return BROTLI_TRUE;
 }
 
 void BrotliDecoderStateMetablockBegin(BrotliDecoderState* s) {
   s->meta_block_remaining_len = 0;
-  s->block_length[0] = 1U << 28;
-  s->block_length[1] = 1U << 28;
-  s->block_length[2] = 1U << 28;
+  s->block_length[0] = 1U << 24;
+  s->block_length[1] = 1U << 24;
+  s->block_length[2] = 1U << 24;
   s->num_block_types[0] = 1;
   s->num_block_types[1] = 1;
   s->num_block_types[2] = 1;
@@ -126,8 +116,7 @@ void BrotliDecoderStateMetablockBegin(BrotliDecoderState* s) {
   s->literal_htree = NULL;
   s->dist_context_map_slice = NULL;
   s->dist_htree_index = 0;
-  s->context_lookup1 = NULL;
-  s->context_lookup2 = NULL;
+  s->context_lookup = NULL;
   s->literal_hgroup.codes = NULL;
   s->literal_hgroup.htrees = NULL;
   s->insert_copy_hgroup.codes = NULL;
@@ -137,30 +126,33 @@ void BrotliDecoderStateMetablockBegin(BrotliDecoderState* s) {
 }
 
 void BrotliDecoderStateCleanupAfterMetablock(BrotliDecoderState* s) {
-  BROTLI_FREE(s, s->context_modes);
-  BROTLI_FREE(s, s->context_map);
-  BROTLI_FREE(s, s->dist_context_map);
-  BROTLI_FREE(s, s->literal_hgroup.htrees);
-  BROTLI_FREE(s, s->insert_copy_hgroup.htrees);
-  BROTLI_FREE(s, s->distance_hgroup.htrees);
+  BROTLI_DECODER_FREE(s, s->context_modes);
+  BROTLI_DECODER_FREE(s, s->context_map);
+  BROTLI_DECODER_FREE(s, s->dist_context_map);
+  BROTLI_DECODER_FREE(s, s->literal_hgroup.htrees);
+  BROTLI_DECODER_FREE(s, s->insert_copy_hgroup.htrees);
+  BROTLI_DECODER_FREE(s, s->distance_hgroup.htrees);
 }
 
 void BrotliDecoderStateCleanup(BrotliDecoderState* s) {
   BrotliDecoderStateCleanupAfterMetablock(s);
 
-  BROTLI_FREE(s, s->ringbuffer);
-  BROTLI_FREE(s, s->block_type_trees);
+  BROTLI_DECODER_FREE(s, s->ringbuffer);
+  BROTLI_DECODER_FREE(s, s->block_type_trees);
 }
 
 BROTLI_BOOL BrotliDecoderHuffmanTreeGroupInit(BrotliDecoderState* s,
-    HuffmanTreeGroup* group, uint32_t alphabet_size, uint32_t ntrees) {
+    HuffmanTreeGroup* group, uint32_t alphabet_size, uint32_t max_symbol,
+    uint32_t ntrees) {
   /* Pack two allocations into one */
   const size_t max_table_size = kMaxHuffmanTableSize[(alphabet_size + 31) >> 5];
   const size_t code_size = sizeof(HuffmanCode) * ntrees * max_table_size;
   const size_t htree_size = sizeof(HuffmanCode*) * ntrees;
   /* Pointer alignment is, hopefully, wider than sizeof(HuffmanCode). */
-  HuffmanCode** p = (HuffmanCode**)BROTLI_ALLOC(s, code_size + htree_size);
+  HuffmanCode** p = (HuffmanCode**)BROTLI_DECODER_ALLOC(s,
+      code_size + htree_size);
   group->alphabet_size = (uint16_t)alphabet_size;
+  group->max_symbol = (uint16_t)max_symbol;
   group->num_htrees = (uint16_t)ntrees;
   group->htrees = p;
   group->codes = (HuffmanCode*)(&p[ntrees]);

modules/brotli/dec/state.h

@@ -11,10 +11,11 @@
 
 #include "../common/constants.h"
 #include "../common/dictionary.h"
+#include "../common/platform.h"
+#include "../common/transform.h"
 #include <brotli/types.h>
 #include "./bit_reader.h"
 #include "./huffman.h"
-#include "./port.h"
 
 #if defined(__cplusplus) || defined(c_plusplus)
 extern "C" {
@@ -22,6 +23,8 @@ extern "C" {
 
 typedef enum {
   BROTLI_STATE_UNINITED,
+  BROTLI_STATE_LARGE_WINDOW_BITS,
+  BROTLI_STATE_INITIALIZE,
   BROTLI_STATE_METABLOCK_BEGIN,
   BROTLI_STATE_METABLOCK_HEADER,
   BROTLI_STATE_METABLOCK_HEADER_2,
@@ -126,23 +129,22 @@ struct BrotliDecoderStateStruct {
   uint8_t* ringbuffer;
   uint8_t* ringbuffer_end;
   HuffmanCode* htree_command;
-  const uint8_t* context_lookup1;
-  const uint8_t* context_lookup2;
+  const uint8_t* context_lookup;
   uint8_t* context_map_slice;
   uint8_t* dist_context_map_slice;
 
-  /* This ring buffer holds a few past copy distances that will be used by */
-  /* some special distance codes. */
+  /* This ring buffer holds a few past copy distances that will be used by
+     some special distance codes. */
   HuffmanTreeGroup literal_hgroup;
   HuffmanTreeGroup insert_copy_hgroup;
   HuffmanTreeGroup distance_hgroup;
   HuffmanCode* block_type_trees;
   HuffmanCode* block_len_trees;
   /* This is true if the literal context map histogram type always matches the
-  block type. It is then not needed to keep the context (faster decoding). */
+     block type. It is then not needed to keep the context (faster decoding). */
   int trivial_literal_context;
-  /* Distance context is actual after command is decoded and before distance
-  is computed. After distance computation it is used as a temporary variable. */
+  /* Distance context is actual after command is decoded and before distance is
+     computed. After distance computation it is used as a temporary variable. */
   int distance_context;
   int meta_block_remaining_len;
   uint32_t block_length_index;
@@ -162,11 +164,11 @@ struct BrotliDecoderStateStruct {
   int copy_length;
   int distance_code;
 
-  /* For partial write operations */
-  size_t rb_roundtrips;  /* How many times we went around the ring-buffer */
-  size_t partial_pos_out;  /* How much output to the user in total */
+  /* For partial write operations. */
+  size_t rb_roundtrips;  /* how many times we went around the ring-buffer */
+  size_t partial_pos_out;  /* how much output to the user in total */
 
-  /* For ReadHuffmanCode */
+  /* For ReadHuffmanCode. */
   uint32_t symbol;
   uint32_t repeat;
   uint32_t space;
@@ -180,25 +182,26 @@ struct BrotliDecoderStateStruct {
   /* Tails of symbol chains. */
   int next_symbol[32];
   uint8_t code_length_code_lengths[BROTLI_CODE_LENGTH_CODES];
-  /* Population counts for the code lengths */
+  /* Population counts for the code lengths. */
   uint16_t code_length_histo[16];
 
-  /* For HuffmanTreeGroupDecode */
+  /* For HuffmanTreeGroupDecode. */
   int htree_index;
   HuffmanCode* next;
 
-  /* For DecodeContextMap */
+  /* For DecodeContextMap. */
   uint32_t context_index;
   uint32_t max_run_length_prefix;
   uint32_t code;
   HuffmanCode context_map_table[BROTLI_HUFFMAN_MAX_SIZE_272];
 
-  /* For InverseMoveToFrontTransform */
+  /* For InverseMoveToFrontTransform. */
   uint32_t mtf_upper_bound;
   uint32_t mtf[64 + 1];
 
-  /* less used attributes are in the end of this struct */
-  /* States inside function calls */
+  /* Less used attributes are at the end of this struct. */
+
+  /* States inside function calls. */
   BrotliRunningMetablockHeaderState substate_metablock_header;
   BrotliRunningTreeGroupState substate_tree_group;
   BrotliRunningContextMapState substate_context_map;
@@ -212,6 +215,7 @@ struct BrotliDecoderStateStruct {
   unsigned int is_metadata : 1;
   unsigned int should_wrap_ringbuffer : 1;
   unsigned int canny_ringbuffer_allocation : 1;
+  unsigned int large_window : 1;
   unsigned int size_nibbles : 8;
   uint32_t window_bits;
 
@@ -220,7 +224,9 @@ struct BrotliDecoderStateStruct {
   uint32_t num_literal_htrees;
   uint8_t* context_map;
   uint8_t* context_modes;
+
   const BrotliDictionary* dictionary;
+  const BrotliTransforms* transforms;
 
   uint32_t trivial_literal_contexts[8];  /* 256 bits */
 };
@@ -228,17 +234,22 @@ struct BrotliDecoderStateStruct {
 typedef struct BrotliDecoderStateStruct BrotliDecoderStateInternal;
 #define BrotliDecoderState BrotliDecoderStateInternal
 
-BROTLI_INTERNAL void BrotliDecoderStateInit(BrotliDecoderState* s);
-BROTLI_INTERNAL void BrotliDecoderStateInitWithCustomAllocators(
-    BrotliDecoderState* s, brotli_alloc_func alloc_func,
-    brotli_free_func free_func, void* opaque);
+BROTLI_INTERNAL BROTLI_BOOL BrotliDecoderStateInit(BrotliDecoderState* s,
+    brotli_alloc_func alloc_func, brotli_free_func free_func, void* opaque);
 BROTLI_INTERNAL void BrotliDecoderStateCleanup(BrotliDecoderState* s);
 BROTLI_INTERNAL void BrotliDecoderStateMetablockBegin(BrotliDecoderState* s);
 BROTLI_INTERNAL void BrotliDecoderStateCleanupAfterMetablock(
     BrotliDecoderState* s);
 BROTLI_INTERNAL BROTLI_BOOL BrotliDecoderHuffmanTreeGroupInit(
     BrotliDecoderState* s, HuffmanTreeGroup* group, uint32_t alphabet_size,
-    uint32_t ntrees);
+    uint32_t max_symbol, uint32_t ntrees);
+
+#define BROTLI_DECODER_ALLOC(S, L) S->alloc_func(S->memory_manager_opaque, L)
+
+#define BROTLI_DECODER_FREE(S, X) {          \
+  S->free_func(S->memory_manager_opaque, X); \
+  X = NULL;                                  \
+}
 
 #if defined(__cplusplus) || defined(c_plusplus)
 }  /* extern "C" */

modules/brotli/dec/transform.h

@@ -1,300 +0,0 @@
-/* Copyright 2013 Google Inc. All Rights Reserved.
-
-   Distributed under MIT license.
-   See file LICENSE for detail or copy at https://opensource.org/licenses/MIT
-*/
-
-/* Transformations on dictionary words. */
-
-#ifndef BROTLI_DEC_TRANSFORM_H_
-#define BROTLI_DEC_TRANSFORM_H_
-
-#include <brotli/types.h>
-#include "./port.h"
-
-#if defined(__cplusplus) || defined(c_plusplus)
-extern "C" {
-#endif
-
-enum WordTransformType {
-  kIdentity = 0,
-  kOmitLast1 = 1,
-  kOmitLast2 = 2,
-  kOmitLast3 = 3,
-  kOmitLast4 = 4,
-  kOmitLast5 = 5,
-  kOmitLast6 = 6,
-  kOmitLast7 = 7,
-  kOmitLast8 = 8,
-  kOmitLast9 = 9,
-  kUppercaseFirst = 10,
-  kUppercaseAll = 11,
-  kOmitFirst1 = 12,
-  kOmitFirst2 = 13,
-  kOmitFirst3 = 14,
-  kOmitFirst4 = 15,
-  kOmitFirst5 = 16,
-  kOmitFirst6 = 17,
-  kOmitFirst7 = 18,
-  kOmitFirst8 = 19,
-  kOmitFirst9 = 20
-};
-
-typedef struct {
-  const uint8_t prefix_id;
-  const uint8_t transform;
-  const uint8_t suffix_id;
-} Transform;
-
-static const char kPrefixSuffix[208] =
-    "\0 \0, \0 of the \0 of \0s \0.\0 and \0 in \0\"\0 to \0\">\0\n\0. \0]\0"
-    " for \0 a \0 that \0\'\0 with \0 from \0 by \0(\0. The \0 on \0 as \0"
-    " is \0ing \0\n\t\0:\0ed \0=\"\0 at \0ly \0,\0=\'\0.com/\0. This \0"
-    " not \0er \0al \0ful \0ive \0less \0est \0ize \0\xc2\xa0\0ous ";
-
-enum {
-  /* EMPTY = ""
-     SP = " "
-     DQUOT = "\""
-     SQUOT = "'"
-     CLOSEBR = "]"
-     OPEN = "("
-     SLASH = "/"
-     NBSP = non-breaking space "\0xc2\xa0"
-  */
-  kPFix_EMPTY = 0,
-  kPFix_SP = 1,
-  kPFix_COMMASP = 3,
-  kPFix_SPofSPtheSP = 6,
-  kPFix_SPtheSP = 9,
-  kPFix_eSP = 12,
-  kPFix_SPofSP = 15,
-  kPFix_sSP = 20,
-  kPFix_DOT = 23,
-  kPFix_SPandSP = 25,
-  kPFix_SPinSP = 31,
-  kPFix_DQUOT = 36,
-  kPFix_SPtoSP = 38,
-  kPFix_DQUOTGT = 43,
-  kPFix_NEWLINE = 46,
-  kPFix_DOTSP = 48,
-  kPFix_CLOSEBR = 51,
-  kPFix_SPforSP = 53,
-  kPFix_SPaSP = 59,
-  kPFix_SPthatSP = 63,
-  kPFix_SQUOT = 70,
-  kPFix_SPwithSP = 72,
-  kPFix_SPfromSP = 79,
-  kPFix_SPbySP = 86,
-  kPFix_OPEN = 91,
-  kPFix_DOTSPTheSP = 93,
-  kPFix_SPonSP = 100,
-  kPFix_SPasSP = 105,
-  kPFix_SPisSP = 110,
-  kPFix_ingSP = 115,
-  kPFix_NEWLINETAB = 120,
-  kPFix_COLON = 123,
-  kPFix_edSP = 125,
-  kPFix_EQDQUOT = 129,
-  kPFix_SPatSP = 132,
-  kPFix_lySP = 137,
-  kPFix_COMMA = 141,
-  kPFix_EQSQUOT = 143,
-  kPFix_DOTcomSLASH = 146,
-  kPFix_DOTSPThisSP = 152,
-  kPFix_SPnotSP = 160,
-  kPFix_erSP = 166,
-  kPFix_alSP = 170,
-  kPFix_fulSP = 174,
-  kPFix_iveSP = 179,
-  kPFix_lessSP = 184,
-  kPFix_estSP = 190,
-  kPFix_izeSP = 195,
-  kPFix_NBSP = 200,
-  kPFix_ousSP = 203
-};
-
-static const Transform kTransforms[] = {
-  { kPFix_EMPTY, kIdentity, kPFix_EMPTY },
-  { kPFix_EMPTY, kIdentity, kPFix_SP },
-  { kPFix_SP, kIdentity, kPFix_SP },
-  { kPFix_EMPTY, kOmitFirst1, kPFix_EMPTY },
-  { kPFix_EMPTY, kUppercaseFirst, kPFix_SP },
-  { kPFix_EMPTY, kIdentity, kPFix_SPtheSP },
-  { kPFix_SP, kIdentity, kPFix_EMPTY },
-  { kPFix_sSP, kIdentity, kPFix_SP },
-  { kPFix_EMPTY, kIdentity, kPFix_SPofSP },
-  { kPFix_EMPTY, kUppercaseFirst, kPFix_EMPTY },
-  { kPFix_EMPTY, kIdentity, kPFix_SPandSP },
-  { kPFix_EMPTY, kOmitFirst2, kPFix_EMPTY },
-  { kPFix_EMPTY, kOmitLast1, kPFix_EMPTY },
-  { kPFix_COMMASP, kIdentity, kPFix_SP },
-  { kPFix_EMPTY, kIdentity, kPFix_COMMASP },
-  { kPFix_SP, kUppercaseFirst, kPFix_SP },
-  { kPFix_EMPTY, kIdentity, kPFix_SPinSP },
-  { kPFix_EMPTY, kIdentity, kPFix_SPtoSP },
-  { kPFix_eSP, kIdentity, kPFix_SP },
-  { kPFix_EMPTY, kIdentity, kPFix_DQUOT },
-  { kPFix_EMPTY, kIdentity, kPFix_DOT },
-  { kPFix_EMPTY, kIdentity, kPFix_DQUOTGT },
-  { kPFix_EMPTY, kIdentity, kPFix_NEWLINE },
-  { kPFix_EMPTY, kOmitLast3, kPFix_EMPTY },
-  { kPFix_EMPTY, kIdentity, kPFix_CLOSEBR },
-  { kPFix_EMPTY, kIdentity, kPFix_SPforSP },
-  { kPFix_EMPTY, kOmitFirst3, kPFix_EMPTY },
-  { kPFix_EMPTY, kOmitLast2, kPFix_EMPTY },
-  { kPFix_EMPTY, kIdentity, kPFix_SPaSP },
-  { kPFix_EMPTY, kIdentity, kPFix_SPthatSP },
-  { kPFix_SP, kUppercaseFirst, kPFix_EMPTY },
-  { kPFix_EMPTY, kIdentity, kPFix_DOTSP },
-  { kPFix_DOT, kIdentity, kPFix_EMPTY },
-  { kPFix_SP, kIdentity, kPFix_COMMASP },
-  { kPFix_EMPTY, kOmitFirst4, kPFix_EMPTY },
-  { kPFix_EMPTY, kIdentity, kPFix_SPwithSP },
-  { kPFix_EMPTY, kIdentity, kPFix_SQUOT },
-  { kPFix_EMPTY, kIdentity, kPFix_SPfromSP },
-  { kPFix_EMPTY, kIdentity, kPFix_SPbySP },
-  { kPFix_EMPTY, kOmitFirst5, kPFix_EMPTY },
-  { kPFix_EMPTY, kOmitFirst6, kPFix_EMPTY },
-  { kPFix_SPtheSP, kIdentity, kPFix_EMPTY },
-  { kPFix_EMPTY, kOmitLast4, kPFix_EMPTY },
-  { kPFix_EMPTY, kIdentity, kPFix_DOTSPTheSP },
-  { kPFix_EMPTY, kUppercaseAll, kPFix_EMPTY },
-  { kPFix_EMPTY, kIdentity, kPFix_SPonSP },
-  { kPFix_EMPTY, kIdentity, kPFix_SPasSP },
-  { kPFix_EMPTY, kIdentity, kPFix_SPisSP },
-  { kPFix_EMPTY, kOmitLast7, kPFix_EMPTY },
-  { kPFix_EMPTY, kOmitLast1, kPFix_ingSP },
-  { kPFix_EMPTY, kIdentity, kPFix_NEWLINETAB },
-  { kPFix_EMPTY, kIdentity, kPFix_COLON },
-  { kPFix_SP, kIdentity, kPFix_DOTSP },
-  { kPFix_EMPTY, kIdentity, kPFix_edSP },
-  { kPFix_EMPTY, kOmitFirst9, kPFix_EMPTY },
-  { kPFix_EMPTY, kOmitFirst7, kPFix_EMPTY },
-  { kPFix_EMPTY, kOmitLast6, kPFix_EMPTY },
-  { kPFix_EMPTY, kIdentity, kPFix_OPEN },
-  { kPFix_EMPTY, kUppercaseFirst, kPFix_COMMASP },
-  { kPFix_EMPTY, kOmitLast8, kPFix_EMPTY },
-  { kPFix_EMPTY, kIdentity, kPFix_SPatSP },
-  { kPFix_EMPTY, kIdentity, kPFix_lySP },
-  { kPFix_SPtheSP, kIdentity, kPFix_SPofSP },
-  { kPFix_EMPTY, kOmitLast5, kPFix_EMPTY },
-  { kPFix_EMPTY, kOmitLast9, kPFix_EMPTY },
-  { kPFix_SP, kUppercaseFirst, kPFix_COMMASP },
-  { kPFix_EMPTY, kUppercaseFirst, kPFix_DQUOT },
-  { kPFix_DOT, kIdentity, kPFix_OPEN },
-  { kPFix_EMPTY, kUppercaseAll, kPFix_SP },
-  { kPFix_EMPTY, kUppercaseFirst, kPFix_DQUOTGT },
-  { kPFix_EMPTY, kIdentity, kPFix_EQDQUOT },
-  { kPFix_SP, kIdentity, kPFix_DOT },
-  { kPFix_DOTcomSLASH, kIdentity, kPFix_EMPTY },
-  { kPFix_SPtheSP, kIdentity, kPFix_SPofSPtheSP },
-  { kPFix_EMPTY, kUppercaseFirst, kPFix_SQUOT },
-  { kPFix_EMPTY, kIdentity, kPFix_DOTSPThisSP },
-  { kPFix_EMPTY, kIdentity, kPFix_COMMA },
-  { kPFix_DOT, kIdentity, kPFix_SP },
-  { kPFix_EMPTY, kUppercaseFirst, kPFix_OPEN },
-  { kPFix_EMPTY, kUppercaseFirst, kPFix_DOT },
-  { kPFix_EMPTY, kIdentity, kPFix_SPnotSP },
-  { kPFix_SP, kIdentity, kPFix_EQDQUOT },
-  { kPFix_EMPTY, kIdentity, kPFix_erSP },
-  { kPFix_SP, kUppercaseAll, kPFix_SP },
-  { kPFix_EMPTY, kIdentity, kPFix_alSP },
-  { kPFix_SP, kUppercaseAll, kPFix_EMPTY },
-  { kPFix_EMPTY, kIdentity, kPFix_EQSQUOT },
-  { kPFix_EMPTY, kUppercaseAll, kPFix_DQUOT },
-  { kPFix_EMPTY, kUppercaseFirst, kPFix_DOTSP },
-  { kPFix_SP, kIdentity, kPFix_OPEN },
-  { kPFix_EMPTY, kIdentity, kPFix_fulSP },
-  { kPFix_SP, kUppercaseFirst, kPFix_DOTSP },
-  { kPFix_EMPTY, kIdentity, kPFix_iveSP },
-  { kPFix_EMPTY, kIdentity, kPFix_lessSP },
-  { kPFix_EMPTY, kUppercaseAll, kPFix_SQUOT },
-  { kPFix_EMPTY, kIdentity, kPFix_estSP },
-  { kPFix_SP, kUppercaseFirst, kPFix_DOT },
-  { kPFix_EMPTY, kUppercaseAll, kPFix_DQUOTGT },
-  { kPFix_SP, kIdentity, kPFix_EQSQUOT },
-  { kPFix_EMPTY, kUppercaseFirst, kPFix_COMMA },
-  { kPFix_EMPTY, kIdentity, kPFix_izeSP },
-  { kPFix_EMPTY, kUppercaseAll, kPFix_DOT },
-  { kPFix_NBSP, kIdentity, kPFix_EMPTY },
-  { kPFix_SP, kIdentity, kPFix_COMMA },
-  { kPFix_EMPTY, kUppercaseFirst, kPFix_EQDQUOT },
-  { kPFix_EMPTY, kUppercaseAll, kPFix_EQDQUOT },
-  { kPFix_EMPTY, kIdentity, kPFix_ousSP },
-  { kPFix_EMPTY, kUppercaseAll, kPFix_COMMASP },
-  { kPFix_EMPTY, kUppercaseFirst, kPFix_EQSQUOT },
-  { kPFix_SP, kUppercaseFirst, kPFix_COMMA },
-  { kPFix_SP, kUppercaseAll, kPFix_EQDQUOT },
-  { kPFix_SP, kUppercaseAll, kPFix_COMMASP },
-  { kPFix_EMPTY, kUppercaseAll, kPFix_COMMA },
-  { kPFix_EMPTY, kUppercaseAll, kPFix_OPEN },
-  { kPFix_EMPTY, kUppercaseAll, kPFix_DOTSP },
-  { kPFix_SP, kUppercaseAll, kPFix_DOT },
-  { kPFix_EMPTY, kUppercaseAll, kPFix_EQSQUOT },
-  { kPFix_SP, kUppercaseAll, kPFix_DOTSP },
-  { kPFix_SP, kUppercaseFirst, kPFix_EQDQUOT },
-  { kPFix_SP, kUppercaseAll, kPFix_EQSQUOT },
-  { kPFix_SP, kUppercaseFirst, kPFix_EQSQUOT },
-};
-
-static const int kNumTransforms = sizeof(kTransforms) / sizeof(kTransforms[0]);
-
-static int ToUpperCase(uint8_t* p) {
-  if (p[0] < 0xc0) {
-    if (p[0] >= 'a' && p[0] <= 'z') {
-      p[0] ^= 32;
-    }
-    return 1;
-  }
-  /* An overly simplified uppercasing model for UTF-8. */
-  if (p[0] < 0xe0) {
-    p[1] ^= 32;
-    return 2;
-  }
-  /* An arbitrary transform for three byte characters. */
-  p[2] ^= 5;
-  return 3;
-}
-
-static BROTLI_NOINLINE int TransformDictionaryWord(
-    uint8_t* dst, const uint8_t* word, int len, int transform) {
-  int idx = 0;
-  {
-    const char* prefix = &kPrefixSuffix[kTransforms[transform].prefix_id];
-    while (*prefix) { dst[idx++] = (uint8_t)*prefix++; }
-  }
-  {
-    const int t = kTransforms[transform].transform;
-    int i = 0;
-    int skip = t - (kOmitFirst1 - 1);
-    if (skip > 0) {
-      word += skip;
-      len -= skip;
-    } else if (t <= kOmitLast9) {
-      len -= t;
-    }
-    while (i < len) { dst[idx++] = word[i++]; }
-    if (t == kUppercaseFirst) {
-      ToUpperCase(&dst[idx - len]);
-    } else if (t == kUppercaseAll) {
-      uint8_t* uppercase = &dst[idx - len];
-      while (len > 0) {
-        int step = ToUpperCase(uppercase);
-        uppercase += step;
-        len -= step;
-      }
-    }
-  }
-  {
-    const char* suffix = &kPrefixSuffix[kTransforms[transform].suffix_id];
-    while (*suffix) { dst[idx++] = (uint8_t)*suffix++; }
-    return idx;
-  }
-}
-
-#if defined(__cplusplus) || defined(c_plusplus)
-}  /* extern "C" */
-#endif
-
-#endif  /* BROTLI_DEC_TRANSFORM_H_ */

modules/brotli/dec/types.h

@@ -1,38 +0,0 @@
-/* Copyright 2013 Google Inc. All Rights Reserved.
-
-   Distributed under MIT license.
-   See file LICENSE for detail or copy at https://opensource.org/licenses/MIT
-*/
-
-/* Common types */
-
-#ifndef BROTLI_DEC_TYPES_H_
-#define BROTLI_DEC_TYPES_H_
-
-#include <stddef.h>  /* for size_t */
-
-#if defined(_MSC_VER) && (_MSC_VER < 1600)
-typedef __int8 int8_t;
-typedef unsigned __int8 uint8_t;
-typedef __int16 int16_t;
-typedef unsigned __int16 uint16_t;
-typedef __int32 int32_t;
-typedef unsigned __int32 uint32_t;
-typedef unsigned __int64 uint64_t;
-typedef __int64 int64_t;
-#else
-#include <stdint.h>
-#endif  /* defined(_MSC_VER) && (_MSC_VER < 1600) */
-
-/* Allocating function pointer. Function MUST return 0 in the case of failure.
-   Otherwise it MUST return a valid pointer to a memory region of at least
-   size length. Neither items nor size are allowed to be 0.
-   opaque argument is a pointer provided by client and could be used to bind
-   function to specific object (memory pool). */
-typedef void* (*brotli_alloc_func)(void* opaque, size_t size);
-
-/* Deallocating function pointer. Function SHOULD be no-op in the case the
-   address is 0. */
-typedef void (*brotli_free_func)(void* opaque, void* address);
-
-#endif  /* BROTLI_DEC_TYPES_H_ */

modules/brotli/enc/backward_references.c

@@ -10,11 +10,11 @@
 
 #include "../common/constants.h"
 #include "../common/dictionary.h"
+#include "../common/platform.h"
 #include <brotli/types.h>
 #include "./command.h"
 #include "./dictionary_hash.h"
 #include "./memory.h"
-#include "./port.h"
 #include "./quality.h"
 
 #if defined(__cplusplus) || defined(c_plusplus)
@@ -49,6 +49,7 @@ static BROTLI_INLINE size_t ComputeDistanceCode(size_t distance,
 #define CAT(a, b) a ## b
 #define FN(X) EXPAND_CAT(X, HASHER())
 #define EXPORT_FN(X) EXPAND_CAT(X, EXPAND_CAT(PREFIX(), HASHER()))
+
 #define PREFIX() N
 
 #define HASHER() H2
@@ -96,29 +97,38 @@ static BROTLI_INLINE size_t ComputeDistanceCode(size_t distance,
 #include "./backward_references_inc.h"
 #undef HASHER
 
+#define HASHER() H35
+/* NOLINTNEXTLINE(build/include) */
+#include "./backward_references_inc.h"
+#undef HASHER
+
+#define HASHER() H55
+/* NOLINTNEXTLINE(build/include) */
+#include "./backward_references_inc.h"
+#undef HASHER
+
+#define HASHER() H65
+/* NOLINTNEXTLINE(build/include) */
+#include "./backward_references_inc.h"
+#undef HASHER
+
 #undef PREFIX
+
 #undef EXPORT_FN
 #undef FN
 #undef CAT
 #undef EXPAND_CAT
 
-void BrotliCreateBackwardReferences(const BrotliDictionary* dictionary,
-                                    size_t num_bytes,
-                                    size_t position,
-                                    const uint8_t* ringbuffer,
-                                    size_t ringbuffer_mask,
-                                    const BrotliEncoderParams* params,
-                                    HasherHandle hasher,
-                                    int* dist_cache,
-                                    size_t* last_insert_len,
-                                    Command* commands,
-                                    size_t* num_commands,
-                                    size_t* num_literals) {
+void BrotliCreateBackwardReferences(
+    size_t num_bytes, size_t position, const uint8_t* ringbuffer,
+    size_t ringbuffer_mask, const BrotliEncoderParams* params,
+    HasherHandle hasher, int* dist_cache, size_t* last_insert_len,
+    Command* commands, size_t* num_commands, size_t* num_literals) {
   switch (params->hasher.type) {
 #define CASE_(N)                                                  \
     case N:                                                       \
-      CreateBackwardReferencesNH ## N(dictionary,                 \
-          kStaticDictionaryHash, num_bytes, position, ringbuffer, \
+      CreateBackwardReferencesNH ## N(                            \
+          num_bytes, position, ringbuffer,                        \
           ringbuffer_mask, params, hasher, dist_cache,            \
           last_insert_len, commands, num_commands, num_literals); \
       return;

modules/brotli/enc/backward_references.h

@@ -11,10 +11,10 @@
 
 #include "../common/constants.h"
 #include "../common/dictionary.h"
+#include "../common/platform.h"
 #include <brotli/types.h>
 #include "./command.h"
 #include "./hash.h"
-#include "./port.h"
 #include "./quality.h"
 
 #if defined(__cplusplus) || defined(c_plusplus)
@@ -26,11 +26,10 @@ extern "C" {
    CreateBackwardReferences calls, and must be incremented by the amount written
    by this call. */
 BROTLI_INTERNAL void BrotliCreateBackwardReferences(
-    const BrotliDictionary* dictionary, size_t num_bytes, size_t position,
-    const uint8_t* ringbuffer, size_t ringbuffer_mask,
-    const BrotliEncoderParams* params, HasherHandle hasher, int* dist_cache,
-    size_t* last_insert_len, Command* commands, size_t* num_commands,
-    size_t* num_literals);
+    size_t num_bytes, size_t position, const uint8_t* ringbuffer,
+    size_t ringbuffer_mask, const BrotliEncoderParams* params,
+    HasherHandle hasher, int* dist_cache, size_t* last_insert_len,
+    Command* commands, size_t* num_commands, size_t* num_literals);
 
 #if defined(__cplusplus) || defined(c_plusplus)
 }  /* extern "C" */

modules/brotli/enc/backward_references_hq.c

@@ -11,13 +11,14 @@
 #include <string.h>  /* memcpy, memset */
 
 #include "../common/constants.h"
+#include "../common/platform.h"
 #include <brotli/types.h>
 #include "./command.h"
 #include "./fast_log.h"
 #include "./find_match_length.h"
 #include "./literal_cost.h"
 #include "./memory.h"
-#include "./port.h"
+#include "./params.h"
 #include "./prefix.h"
 #include "./quality.h"
 
@@ -25,6 +26,8 @@
 extern "C" {
 #endif
 
+#define BROTLI_MAX_EFFECTIVE_DISTANCE_ALPHABET_SIZE 544
+
 static const float kInfinity = 1.7e38f;  /* ~= 2 ^ 127 */
 
 static const uint32_t kDistanceCacheIndex[] = {
@@ -39,40 +42,41 @@ void BrotliInitZopfliNodes(ZopfliNode* array, size_t length) {
   size_t i;
   stub.length = 1;
   stub.distance = 0;
-  stub.insert_length = 0;
+  stub.dcode_insert_length = 0;
   stub.u.cost = kInfinity;
   for (i = 0; i < length; ++i) array[i] = stub;
 }
 
 static BROTLI_INLINE uint32_t ZopfliNodeCopyLength(const ZopfliNode* self) {
-  return self->length & 0xffffff;
+  return self->length & 0x1FFFFFF;
 }
 
 static BROTLI_INLINE uint32_t ZopfliNodeLengthCode(const ZopfliNode* self) {
-  const uint32_t modifier = self->length >> 24;
+  const uint32_t modifier = self->length >> 25;
   return ZopfliNodeCopyLength(self) + 9u - modifier;
 }
 
 static BROTLI_INLINE uint32_t ZopfliNodeCopyDistance(const ZopfliNode* self) {
-  return self->distance & 0x1ffffff;
+  return self->distance;
 }
 
 static BROTLI_INLINE uint32_t ZopfliNodeDistanceCode(const ZopfliNode* self) {
-  const uint32_t short_code = self->distance >> 25;
+  const uint32_t short_code = self->dcode_insert_length >> 27;
   return short_code == 0 ?
       ZopfliNodeCopyDistance(self) + BROTLI_NUM_DISTANCE_SHORT_CODES - 1 :
       short_code - 1;
 }
 
 static BROTLI_INLINE uint32_t ZopfliNodeCommandLength(const ZopfliNode* self) {
-  return ZopfliNodeCopyLength(self) + self->insert_length;
+  return ZopfliNodeCopyLength(self) + (self->dcode_insert_length & 0x7FFFFFF);
 }
 
 /* Histogram based cost model for zopflification. */
 typedef struct ZopfliCostModel {
   /* The insert and copy length symbols. */
   float cost_cmd_[BROTLI_NUM_COMMAND_SYMBOLS];
-  float cost_dist_[BROTLI_NUM_DISTANCE_SYMBOLS];
+  float* cost_dist_;
+  uint32_t distance_histogram_size;
   /* Cumulative costs of literals per position in the stream. */
   float* literal_costs_;
   float min_cost_cmd_;
@@ -80,28 +84,45 @@ typedef struct ZopfliCostModel {
 } ZopfliCostModel;
 
 static void InitZopfliCostModel(
-    MemoryManager* m, ZopfliCostModel* self, size_t num_bytes) {
+    MemoryManager* m, ZopfliCostModel* self, const BrotliDistanceParams* dist,
+    size_t num_bytes) {
+  uint32_t distance_histogram_size = dist->alphabet_size;
+  if (distance_histogram_size > BROTLI_MAX_EFFECTIVE_DISTANCE_ALPHABET_SIZE) {
+    distance_histogram_size = BROTLI_MAX_EFFECTIVE_DISTANCE_ALPHABET_SIZE;
+  }
   self->num_bytes_ = num_bytes;
   self->literal_costs_ = BROTLI_ALLOC(m, float, num_bytes + 2);
+  self->cost_dist_ = BROTLI_ALLOC(m, float, dist->alphabet_size);
+  self->distance_histogram_size = distance_histogram_size;
   if (BROTLI_IS_OOM(m)) return;
 }
 
 static void CleanupZopfliCostModel(MemoryManager* m, ZopfliCostModel* self) {
   BROTLI_FREE(m, self->literal_costs_);
+  BROTLI_FREE(m, self->cost_dist_);
 }
 
 static void SetCost(const uint32_t* histogram, size_t histogram_size,
-                    float* cost) {
+                    BROTLI_BOOL literal_histogram, float* cost) {
   size_t sum = 0;
+  size_t missing_symbol_sum;
   float log2sum;
+  float missing_symbol_cost;
   size_t i;
   for (i = 0; i < histogram_size; i++) {
     sum += histogram[i];
   }
   log2sum = (float)FastLog2(sum);
+  missing_symbol_sum = sum;
+  if (!literal_histogram) {
+    for (i = 0; i < histogram_size; i++) {
+      if (histogram[i] == 0) missing_symbol_sum++;
+    }
+  }
+  missing_symbol_cost = (float)FastLog2(missing_symbol_sum) + 2;
   for (i = 0; i < histogram_size; i++) {
     if (histogram[i] == 0) {
-      cost[i] = log2sum + 2;
+      cost[i] = missing_symbol_cost;
       continue;
     }
 
@@ -122,7 +143,7 @@ static void ZopfliCostModelSetFromCommands(ZopfliCostModel* self,
                                            size_t last_insert_len) {
   uint32_t histogram_literal[BROTLI_NUM_LITERAL_SYMBOLS];
   uint32_t histogram_cmd[BROTLI_NUM_COMMAND_SYMBOLS];
-  uint32_t histogram_dist[BROTLI_NUM_DISTANCE_SYMBOLS];
+  uint32_t histogram_dist[BROTLI_MAX_EFFECTIVE_DISTANCE_ALPHABET_SIZE];
   float cost_literal[BROTLI_NUM_LITERAL_SYMBOLS];
   size_t pos = position - last_insert_len;
   float min_cost_cmd = kInfinity;
@@ -136,7 +157,7 @@ static void ZopfliCostModelSetFromCommands(ZopfliCostModel* self,
   for (i = 0; i < num_commands; i++) {
     size_t inslength = commands[i].insert_len_;
     size_t copylength = CommandCopyLen(&commands[i]);
-    size_t distcode = commands[i].dist_prefix_;
+    size_t distcode = commands[i].dist_prefix_ & 0x3FF;
     size_t cmdcode = commands[i].cmd_prefix_;
     size_t j;
 
@@ -150,9 +171,12 @@ static void ZopfliCostModelSetFromCommands(ZopfliCostModel* self,
     pos += inslength + copylength;
   }
 
-  SetCost(histogram_literal, BROTLI_NUM_LITERAL_SYMBOLS, cost_literal);
-  SetCost(histogram_cmd, BROTLI_NUM_COMMAND_SYMBOLS, cost_cmd);
-  SetCost(histogram_dist, BROTLI_NUM_DISTANCE_SYMBOLS, self->cost_dist_);
+  SetCost(histogram_literal, BROTLI_NUM_LITERAL_SYMBOLS, BROTLI_TRUE,
+          cost_literal);
+  SetCost(histogram_cmd, BROTLI_NUM_COMMAND_SYMBOLS, BROTLI_FALSE,
+          cost_cmd);
+  SetCost(histogram_dist, self->distance_histogram_size, BROTLI_FALSE,
+          self->cost_dist_);
 
   for (i = 0; i < BROTLI_NUM_COMMAND_SYMBOLS; ++i) {
     min_cost_cmd = BROTLI_MIN(float, min_cost_cmd, cost_cmd[i]);
@@ -161,11 +185,14 @@ static void ZopfliCostModelSetFromCommands(ZopfliCostModel* self,
 
   {
     float* literal_costs = self->literal_costs_;
+    float literal_carry = 0.0;
     size_t num_bytes = self->num_bytes_;
     literal_costs[0] = 0.0;
     for (i = 0; i < num_bytes; ++i) {
-      literal_costs[i + 1] = literal_costs[i] +
+      literal_carry +=
           cost_literal[ringbuffer[(position + i) & ringbuffer_mask]];
+      literal_costs[i + 1] = literal_costs[i] + literal_carry;
+      literal_carry -= literal_costs[i + 1] - literal_costs[i];
     }
   }
 }
@@ -175,6 +202,7 @@ static void ZopfliCostModelSetFromLiteralCosts(ZopfliCostModel* self,
                                                const uint8_t* ringbuffer,
                                                size_t ringbuffer_mask) {
   float* literal_costs = self->literal_costs_;
+  float literal_carry = 0.0;
   float* cost_dist = self->cost_dist_;
   float* cost_cmd = self->cost_cmd_;
   size_t num_bytes = self->num_bytes_;
@@ -183,12 +211,14 @@ static void ZopfliCostModelSetFromLiteralCosts(ZopfliCostModel* self,
                                     ringbuffer, &literal_costs[1]);
   literal_costs[0] = 0.0;
   for (i = 0; i < num_bytes; ++i) {
-    literal_costs[i + 1] += literal_costs[i];
+    literal_carry += literal_costs[i + 1];
+    literal_costs[i + 1] = literal_costs[i] + literal_carry;
+    literal_carry -= literal_costs[i + 1] - literal_costs[i];
   }
   for (i = 0; i < BROTLI_NUM_COMMAND_SYMBOLS; ++i) {
     cost_cmd[i] = (float)FastLog2(11 + (uint32_t)i);
   }
-  for (i = 0; i < BROTLI_NUM_DISTANCE_SYMBOLS; ++i) {
+  for (i = 0; i < self->distance_histogram_size; ++i) {
     cost_dist[i] = (float)FastLog2(20 + (uint32_t)i);
   }
   self->min_cost_cmd_ = (float)FastLog2(11);
@@ -221,9 +251,10 @@ static BROTLI_INLINE void UpdateZopfliNode(ZopfliNode* nodes, size_t pos,
     size_t start_pos, size_t len, size_t len_code, size_t dist,
     size_t short_code, float cost) {
   ZopfliNode* next = &nodes[pos + len];
-  next->length = (uint32_t)(len | ((len + 9u - len_code) << 24));
-  next->distance = (uint32_t)(dist | (short_code << 25));
-  next->insert_length = (uint32_t)(pos - start_pos);
+  next->length = (uint32_t)(len | ((len + 9u - len_code) << 25));
+  next->distance = (uint32_t)dist;
+  next->dcode_insert_length = (uint32_t)(
+      (short_code << 27) | (pos - start_pos));
   next->u.cost = cost;
 }
 
@@ -299,21 +330,21 @@ static size_t ComputeMinimumCopyLength(const float start_cost,
    REQUIRES: nodes[0..pos] satisfies that "ZopfliNode array invariant". */
 static uint32_t ComputeDistanceShortcut(const size_t block_start,
                                         const size_t pos,
-                                        const size_t max_backward,
+                                        const size_t max_backward_limit,
                                         const size_t gap,
                                         const ZopfliNode* nodes) {
   const size_t clen = ZopfliNodeCopyLength(&nodes[pos]);
-  const size_t ilen = nodes[pos].insert_length;
+  const size_t ilen = nodes[pos].dcode_insert_length & 0x7FFFFFF;
   const size_t dist = ZopfliNodeCopyDistance(&nodes[pos]);
   /* Since |block_start + pos| is the end position of the command, the copy part
      starts from |block_start + pos - clen|. Distances that are greater than
-     this or greater than |max_backward| are static dictionary references, and
-     do not update the last distances. Also distance code 0 (last distance)
-     does not update the last distances. */
+     this or greater than |max_backward_limit| + |gap| are static dictionary
+     references, and do not update the last distances.
+     Also distance code 0 (last distance) does not update the last distances. */
   if (pos == 0) {
     return 0;
   } else if (dist + clen <= block_start + pos + gap &&
-             dist <= max_backward + gap &&
+             dist <= max_backward_limit + gap &&
              ZopfliNodeDistanceCode(&nodes[pos]) > 0) {
     return (uint32_t)pos;
   } else {
@@ -335,7 +366,7 @@ static void ComputeDistanceCache(const size_t pos,
   int idx = 0;
   size_t p = nodes[pos].u.shortcut;
   while (idx < 4 && p > 0) {
-    const size_t ilen = nodes[p].insert_length;
+    const size_t ilen = nodes[p].dcode_insert_length & 0x7FFFFFF;
     const size_t clen = ZopfliNodeCopyLength(&nodes[p]);
     const size_t dist = ZopfliNodeCopyDistance(&nodes[p]);
     dist_cache[idx++] = (int)dist;
@@ -423,9 +454,11 @@ static size_t UpdateNodes(
         break;
       }
       if (BROTLI_PREDICT_FALSE(backward > max_distance + gap)) {
+        /* Word dictionary -> ignore. */
         continue;
       }
       if (backward <= max_distance) {
+        /* Regular backward reference. */
         if (prev_ix >= cur_ix) {
           continue;
         }
@@ -482,10 +515,12 @@ static size_t UpdateNodes(
         uint32_t distnumextra;
         float dist_cost;
         size_t max_match_len;
-        PrefixEncodeCopyDistance(dist_code, 0, 0, &dist_symbol, &distextra);
-        distnumextra = distextra >> 24;
+        PrefixEncodeCopyDistance(
+            dist_code, params->dist.num_direct_distance_codes,
+            params->dist.distance_postfix_bits, &dist_symbol, &distextra);
+        distnumextra = dist_symbol >> 10;
         dist_cost = base_cost + (float)distnumextra +
-            ZopfliCostModelGetDistanceCost(model, dist_symbol);
+            ZopfliCostModelGetDistanceCost(model, dist_symbol & 0x3FF);
 
         /* Try all copy lengths up until the maximum copy length corresponding
            to this distance. If the distance refers to the static dictionary, or
@@ -517,7 +552,8 @@ static size_t ComputeShortestPathFromNodes(size_t num_bytes,
     ZopfliNode* nodes) {
   size_t index = num_bytes;
   size_t num_commands = 0;
-  while (nodes[index].insert_length == 0 && nodes[index].length == 1) --index;
+  while ((nodes[index].dcode_insert_length & 0x7FFFFFF) == 0 &&
+      nodes[index].length == 1) --index;
   nodes[index].u.next = BROTLI_UINT32_MAX;
   while (index != 0) {
     size_t len = ZopfliNodeCommandLength(&nodes[index]);
@@ -530,23 +566,18 @@ static size_t ComputeShortestPathFromNodes(size_t num_bytes,
 
 /* REQUIRES: nodes != NULL and len(nodes) >= num_bytes + 1 */
 void BrotliZopfliCreateCommands(const size_t num_bytes,
-                                const size_t block_start,
-                                const size_t max_backward_limit,
-                                const ZopfliNode* nodes,
-                                int* dist_cache,
-                                size_t* last_insert_len,
-                                const BrotliEncoderParams* params,
-                                Command* commands,
-                                size_t* num_literals) {
+    const size_t block_start, const ZopfliNode* nodes, int* dist_cache,
+    size_t* last_insert_len, const BrotliEncoderParams* params,
+    Command* commands, size_t* num_literals) {
+  const size_t max_backward_limit = BROTLI_MAX_BACKWARD_LIMIT(params->lgwin);
   size_t pos = 0;
   uint32_t offset = nodes[0].u.next;
   size_t i;
   size_t gap = 0;
-  BROTLI_UNUSED(params);
   for (i = 0; offset != BROTLI_UINT32_MAX; i++) {
     const ZopfliNode* next = &nodes[pos + offset];
     size_t copy_length = ZopfliNodeCopyLength(next);
-    size_t insert_length = next->insert_length;
+    size_t insert_length = next->dcode_insert_length & 0x7FFFFFF;
     pos += insert_length;
     offset = next->u.next;
     if (i == 0) {
@@ -560,8 +591,7 @@ void BrotliZopfliCreateCommands(const size_t num_bytes,
           BROTLI_MIN(size_t, block_start + pos, max_backward_limit);
       BROTLI_BOOL is_dictionary = TO_BROTLI_BOOL(distance > max_distance + gap);
       size_t dist_code = ZopfliNodeDistanceCode(next);
-
-      InitCommand(&commands[i], insert_length,
+      InitCommand(&commands[i], &params->dist, insert_length,
           copy_length, (int)len_code - (int)copy_length, dist_code);
 
       if (!is_dictionary && dist_code > 0) {
@@ -578,18 +608,12 @@ void BrotliZopfliCreateCommands(const size_t num_bytes,
   *last_insert_len += num_bytes - pos;
 }
 
-static size_t ZopfliIterate(size_t num_bytes,
-                            size_t position,
-                            const uint8_t* ringbuffer,
-                            size_t ringbuffer_mask,
-                            const BrotliEncoderParams* params,
-                            const size_t max_backward_limit,
-                            const size_t gap,
-                            const int* dist_cache,
-                            const ZopfliCostModel* model,
-                            const uint32_t* num_matches,
-                            const BackwardMatch* matches,
-                            ZopfliNode* nodes) {
+static size_t ZopfliIterate(size_t num_bytes, size_t position,
+    const uint8_t* ringbuffer, size_t ringbuffer_mask,
+    const BrotliEncoderParams* params, const size_t gap, const int* dist_cache,
+    const ZopfliCostModel* model, const uint32_t* num_matches,
+    const BackwardMatch* matches, ZopfliNode* nodes) {
+  const size_t max_backward_limit = BROTLI_MAX_BACKWARD_LIMIT(params->lgwin);
   const size_t max_zopfli_len = MaxZopfliLen(params);
   StartPosQueue queue;
   size_t cur_match_pos = 0;
@@ -613,8 +637,8 @@ static size_t ZopfliIterate(size_t num_bytes,
       while (skip) {
         i++;
         if (i + 3 >= num_bytes) break;
-        EvaluateNode(position, i, max_backward_limit, gap, dist_cache, model,
-            &queue, nodes);
+        EvaluateNode(position, i, max_backward_limit, gap,
+            dist_cache, model, &queue, nodes);
         cur_match_pos += num_matches[i];
         skip--;
       }
@@ -624,28 +648,23 @@ static size_t ZopfliIterate(size_t num_bytes,
 }
 
 /* REQUIRES: nodes != NULL and len(nodes) >= num_bytes + 1 */
-size_t BrotliZopfliComputeShortestPath(MemoryManager* m,
-                                       const BrotliDictionary* dictionary,
-                                       size_t num_bytes,
-                                       size_t position,
-                                       const uint8_t* ringbuffer,
-                                       size_t ringbuffer_mask,
-                                       const BrotliEncoderParams* params,
-                                       const size_t max_backward_limit,
-                                       const int* dist_cache,
-                                       HasherHandle hasher,
-                                       ZopfliNode* nodes) {
+size_t BrotliZopfliComputeShortestPath(MemoryManager* m, size_t num_bytes,
+    size_t position, const uint8_t* ringbuffer, size_t ringbuffer_mask,
+    const BrotliEncoderParams* params,
+    const int* dist_cache, HasherHandle hasher, ZopfliNode* nodes) {
+  const size_t max_backward_limit = BROTLI_MAX_BACKWARD_LIMIT(params->lgwin);
   const size_t max_zopfli_len = MaxZopfliLen(params);
   ZopfliCostModel model;
   StartPosQueue queue;
-  BackwardMatch matches[MAX_NUM_MATCHES_H10];
+  BackwardMatch matches[2 * (MAX_NUM_MATCHES_H10 + 64)];
   const size_t store_end = num_bytes >= StoreLookaheadH10() ?
       position + num_bytes - StoreLookaheadH10() + 1 : position;
   size_t i;
   size_t gap = 0;
+  size_t lz_matches_offset = 0;
   nodes[0].length = 0;
   nodes[0].u.cost = 0;
-  InitZopfliCostModel(m, &model, num_bytes);
+  InitZopfliCostModel(m, &model, &params->dist, num_bytes);
   if (BROTLI_IS_OOM(m)) return 0;
   ZopfliCostModelSetFromLiteralCosts(
       &model, position, ringbuffer, ringbuffer_mask);
@@ -653,10 +672,12 @@ size_t BrotliZopfliComputeShortestPath(MemoryManager* m,
   for (i = 0; i + HashTypeLengthH10() - 1 < num_bytes; i++) {
     const size_t pos = position + i;
     const size_t max_distance = BROTLI_MIN(size_t, pos, max_backward_limit);
-    size_t num_matches = FindAllMatchesH10(hasher, dictionary, ringbuffer,
-        ringbuffer_mask, pos, num_bytes - i, max_distance, gap, params,
-        matches);
     size_t skip;
+    size_t num_matches;
+    num_matches = FindAllMatchesH10(hasher,
+        &params->dictionary,
+        ringbuffer, ringbuffer_mask, pos, num_bytes - i, max_distance,
+        gap, params, &matches[lz_matches_offset]);
     if (num_matches > 0 &&
         BackwardMatchLength(&matches[num_matches - 1]) > max_zopfli_len) {
       matches[0] = matches[num_matches - 1];
@@ -677,8 +698,8 @@ size_t BrotliZopfliComputeShortestPath(MemoryManager* m,
       while (skip) {
         i++;
         if (i + HashTypeLengthH10() - 1 >= num_bytes) break;
-        EvaluateNode(position, i, max_backward_limit, gap, dist_cache, &model,
-            &queue, nodes);
+        EvaluateNode(position, i, max_backward_limit, gap,
+            dist_cache, &model, &queue, nodes);
         skip--;
       }
     }
@@ -687,32 +708,29 @@ size_t BrotliZopfliComputeShortestPath(MemoryManager* m,
   return ComputeShortestPathFromNodes(num_bytes, nodes);
 }
 
-void BrotliCreateZopfliBackwardReferences(
-    MemoryManager* m, const BrotliDictionary* dictionary, size_t num_bytes,
+void BrotliCreateZopfliBackwardReferences(MemoryManager* m, size_t num_bytes,
     size_t position, const uint8_t* ringbuffer, size_t ringbuffer_mask,
-    const BrotliEncoderParams* params, HasherHandle hasher, int* dist_cache,
-    size_t* last_insert_len, Command* commands, size_t* num_commands,
-    size_t* num_literals) {
-  const size_t max_backward_limit = BROTLI_MAX_BACKWARD_LIMIT(params->lgwin);
+    const BrotliEncoderParams* params,
+    HasherHandle hasher, int* dist_cache, size_t* last_insert_len,
+    Command* commands, size_t* num_commands, size_t* num_literals) {
   ZopfliNode* nodes;
   nodes = BROTLI_ALLOC(m, ZopfliNode, num_bytes + 1);
   if (BROTLI_IS_OOM(m)) return;
   BrotliInitZopfliNodes(nodes, num_bytes + 1);
-  *num_commands += BrotliZopfliComputeShortestPath(m, dictionary, num_bytes,
-      position, ringbuffer, ringbuffer_mask, params, max_backward_limit,
+  *num_commands += BrotliZopfliComputeShortestPath(m, num_bytes,
+      position, ringbuffer, ringbuffer_mask, params,
       dist_cache, hasher, nodes);
   if (BROTLI_IS_OOM(m)) return;
-  BrotliZopfliCreateCommands(num_bytes, position, max_backward_limit, nodes,
-      dist_cache, last_insert_len, params, commands, num_literals);
+  BrotliZopfliCreateCommands(num_bytes, position, nodes, dist_cache,
+      last_insert_len, params, commands, num_literals);
   BROTLI_FREE(m, nodes);
 }
 
-void BrotliCreateHqZopfliBackwardReferences(
-    MemoryManager* m, const BrotliDictionary* dictionary, size_t num_bytes,
+void BrotliCreateHqZopfliBackwardReferences(MemoryManager* m, size_t num_bytes,
     size_t position, const uint8_t* ringbuffer, size_t ringbuffer_mask,
-    const BrotliEncoderParams* params, HasherHandle hasher, int* dist_cache,
-    size_t* last_insert_len, Command* commands, size_t* num_commands,
-    size_t* num_literals) {
+    const BrotliEncoderParams* params,
+    HasherHandle hasher, int* dist_cache, size_t* last_insert_len,
+    Command* commands, size_t* num_commands, size_t* num_literals) {
   const size_t max_backward_limit = BROTLI_MAX_BACKWARD_LIMIT(params->lgwin);
   uint32_t* num_matches = BROTLI_ALLOC(m, uint32_t, num_bytes);
   size_t matches_size = 4 * num_bytes;
@@ -728,6 +746,7 @@ void BrotliCreateHqZopfliBackwardReferences(
   ZopfliNode* nodes;
   BackwardMatch* matches = BROTLI_ALLOC(m, BackwardMatch, matches_size);
   size_t gap = 0;
+  size_t shadow_matches = 0;
   if (BROTLI_IS_OOM(m)) return;
   for (i = 0; i + HashTypeLengthH10() - 1 < num_bytes; ++i) {
     const size_t pos = position + i;
@@ -738,14 +757,16 @@ void BrotliCreateHqZopfliBackwardReferences(
     size_t j;
     /* Ensure that we have enough free slots. */
     BROTLI_ENSURE_CAPACITY(m, BackwardMatch, matches, matches_size,
-        cur_match_pos + MAX_NUM_MATCHES_H10);
+        cur_match_pos + MAX_NUM_MATCHES_H10 + shadow_matches);
     if (BROTLI_IS_OOM(m)) return;
-    num_found_matches = FindAllMatchesH10(hasher, dictionary, ringbuffer,
-        ringbuffer_mask, pos, max_length, max_distance, gap, params,
-        &matches[cur_match_pos]);
+    num_found_matches = FindAllMatchesH10(hasher,
+        &params->dictionary,
+        ringbuffer, ringbuffer_mask, pos, max_length,
+        max_distance, gap, params,
+        &matches[cur_match_pos + shadow_matches]);
     cur_match_end = cur_match_pos + num_found_matches;
     for (j = cur_match_pos; j + 1 < cur_match_end; ++j) {
-      assert(BackwardMatchLength(&matches[j]) <=
+      BROTLI_DCHECK(BackwardMatchLength(&matches[j]) <=
           BackwardMatchLength(&matches[j + 1]));
     }
     num_matches[i] = (uint32_t)num_found_matches;
@@ -771,7 +792,7 @@ void BrotliCreateHqZopfliBackwardReferences(
   orig_num_commands = *num_commands;
   nodes = BROTLI_ALLOC(m, ZopfliNode, num_bytes + 1);
   if (BROTLI_IS_OOM(m)) return;
-  InitZopfliCostModel(m, &model, num_bytes);
+  InitZopfliCostModel(m, &model, &params->dist, num_bytes);
   if (BROTLI_IS_OOM(m)) return;
   for (i = 0; i < 2; i++) {
     BrotliInitZopfliNodes(nodes, num_bytes + 1);
@@ -788,10 +809,10 @@ void BrotliCreateHqZopfliBackwardReferences(
     *last_insert_len = orig_last_insert_len;
     memcpy(dist_cache, orig_dist_cache, 4 * sizeof(dist_cache[0]));
     *num_commands += ZopfliIterate(num_bytes, position, ringbuffer,
-        ringbuffer_mask, params, max_backward_limit, gap, dist_cache,
-        &model, num_matches, matches, nodes);
-    BrotliZopfliCreateCommands(num_bytes, position, max_backward_limit,
-        nodes, dist_cache, last_insert_len, params, commands, num_literals);
+        ringbuffer_mask, params, gap, dist_cache, &model, num_matches, matches,
+        nodes);
+    BrotliZopfliCreateCommands(num_bytes, position, nodes, dist_cache,
+        last_insert_len, params, commands, num_literals);
   }
   CleanupZopfliCostModel(m, &model);
   BROTLI_FREE(m, nodes);

modules/brotli/enc/backward_references_hq.h

@@ -11,41 +11,38 @@
 
 #include "../common/constants.h"
 #include "../common/dictionary.h"
+#include "../common/platform.h"
 #include <brotli/types.h>
 #include "./command.h"
 #include "./hash.h"
 #include "./memory.h"
-#include "./port.h"
 #include "./quality.h"
 
 #if defined(__cplusplus) || defined(c_plusplus)
 extern "C" {
 #endif
 
-BROTLI_INTERNAL void BrotliCreateZopfliBackwardReferences(
-    MemoryManager* m, const BrotliDictionary* dictionary, size_t num_bytes,
-    size_t position, const uint8_t* ringbuffer, size_t ringbuffer_mask,
-    const BrotliEncoderParams* params, HasherHandle hasher, int* dist_cache,
-    size_t* last_insert_len, Command* commands, size_t* num_commands,
-    size_t* num_literals);
+BROTLI_INTERNAL void BrotliCreateZopfliBackwardReferences(MemoryManager* m,
+    size_t num_bytes, size_t position, const uint8_t* ringbuffer,
+    size_t ringbuffer_mask, const BrotliEncoderParams* params,
+    HasherHandle hasher, int* dist_cache, size_t* last_insert_len,
+    Command* commands, size_t* num_commands, size_t* num_literals);
 
-BROTLI_INTERNAL void BrotliCreateHqZopfliBackwardReferences(
-    MemoryManager* m, const BrotliDictionary* dictionary, size_t num_bytes,
-    size_t position, const uint8_t* ringbuffer, size_t ringbuffer_mask,
-    const BrotliEncoderParams* params, HasherHandle hasher, int* dist_cache,
-    size_t* last_insert_len, Command* commands, size_t* num_commands,
-    size_t* num_literals);
+BROTLI_INTERNAL void BrotliCreateHqZopfliBackwardReferences(MemoryManager* m,
+    size_t num_bytes, size_t position, const uint8_t* ringbuffer,
+    size_t ringbuffer_mask, const BrotliEncoderParams* params,
+    HasherHandle hasher, int* dist_cache, size_t* last_insert_len,
+    Command* commands, size_t* num_commands, size_t* num_literals);
 
 typedef struct ZopfliNode {
-  /* best length to get up to this byte (not including this byte itself)
-     highest 8 bit is used to reconstruct the length code */
+  /* Best length to get up to this byte (not including this byte itself)
+     highest 7 bit is used to reconstruct the length code. */
   uint32_t length;
-  /* distance associated with the length
-     highest 7 bit contains distance short code + 1 (or zero if no short code)
-  */
+  /* Distance associated with the length. */
   uint32_t distance;
-  /* number of literal inserts before this copy */
-  uint32_t insert_length;
+  /* Number of literal inserts before this copy; highest 5 bits contain
+     distance short code + 1 (or zero if no short code). */
+  uint32_t dcode_insert_length;
 
   /* This union holds information used by dynamic-programming. During forward
      pass |cost| it used to store the goal function. When node is processed its
@@ -78,14 +75,13 @@ BROTLI_INTERNAL void BrotliInitZopfliNodes(ZopfliNode* array, size_t length);
      (2) nodes[i].command_length() <= i and
      (3) nodes[i - nodes[i].command_length()].cost < kInfinity */
 BROTLI_INTERNAL size_t BrotliZopfliComputeShortestPath(
-    MemoryManager* m, const BrotliDictionary* dictionary, size_t num_bytes,
+    MemoryManager* m, size_t num_bytes,
     size_t position, const uint8_t* ringbuffer, size_t ringbuffer_mask,
-    const BrotliEncoderParams* params, const size_t max_backward_limit,
+    const BrotliEncoderParams* params,
     const int* dist_cache, HasherHandle hasher, ZopfliNode* nodes);
 
 BROTLI_INTERNAL void BrotliZopfliCreateCommands(
-    const size_t num_bytes, const size_t block_start,
-    const size_t max_backward_limit, const ZopfliNode* nodes,
+    const size_t num_bytes, const size_t block_start, const ZopfliNode* nodes,
     int* dist_cache, size_t* last_insert_len, const BrotliEncoderParams* params,
     Command* commands, size_t* num_literals);