Quelle  TestUtf8.cpp

/* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 2 -*- */
/* vim: set ts=8 sts=2 et sw=2 tw=80: */
/* This Source Code Form is subject to the terms of the Mozilla Public
 * License, v. 2.0. If a copy of the MPL was not distributed with this
 * file, You can obtain one at http://mozilla.org/MPL/2.0/. */

#define MOZ_PRETEND_NO_JSRUST 1

#include "mozilla/Utf8.h"

#include "mozilla/ArrayUtils.h"
#include "mozilla/Assertions.h"
#include "mozilla/EnumSet.h"
#include "mozilla/IntegerRange.h"
#include "mozilla/Span.h"

using mozilla::AsChars;
using mozilla::DecodeOneUtf8CodePoint;
using mozilla::EnumSet;
using mozilla::IntegerRange;
using mozilla::IsAscii;
using mozilla::IsUtf8;
using mozilla::Span;
using mozilla::Utf8Unit;

static void TestUtf8Unit() {
  Utf8Unit c('A');
  MOZ_RELEASE_ASSERT(c.toChar() == 'A');
  MOZ_RELEASE_ASSERT(c == Utf8Unit('A'));
  MOZ_RELEASE_ASSERT(c != Utf8Unit('B'));
  MOZ_RELEASE_ASSERT(c.toUint8() == 0x41);

  unsigned char asUnsigned = 'A';
  MOZ_RELEASE_ASSERT(c.toUnsignedChar() == asUnsigned);
  MOZ_RELEASE_ASSERT(Utf8Unit('B').toUnsignedChar() != asUnsigned);

  Utf8Unit first('@');
  Utf8Unit second('#');

  MOZ_RELEASE_ASSERT(first != second);

  first = second;
  MOZ_RELEASE_ASSERT(first == second);
}

template <typename Char>
struct ToUtf8Units {
 public:
  explicit ToUtf8Units(const Char* aStart, const Char* aEnd)
      : lead(Utf8Unit(aStart[0])), iter(aStart + 1), end(aEnd) {
    MOZ_RELEASE_ASSERT(!IsAscii(aStart[0]));
  }

  const Utf8Unit lead;
  const Char* iter;
  const Char* const end;
};

class AssertIfCalled {
 public:
  template <typename... Args>
  void operator()(Args&&... aArgs) {
    MOZ_RELEASE_ASSERT(false, "AssertIfCalled instance was called");
  }
};

// NOTE: For simplicity in treating |aCharN| identically regardless whether it's
//       a string literal or a more-generalized array, we require |aCharN| be
//       null-terminated.

template <typename Char, size_t N>
static void ExpectValidCodePoint(const Char (&aCharN)[N],
                                 char32_t aExpectedCodePoint) {
  MOZ_RELEASE_ASSERT(aCharN[N - 1] == 0,
                     "array must be null-terminated for |aCharN + N - 1| to "
                     "compute the value of |aIter| as altered by "
                     "DecodeOneUtf8CodePoint");

  ToUtf8Units<Char> simpleUnit(aCharN, aCharN + N - 1);
  auto simple =
      DecodeOneUtf8CodePoint(simpleUnit.lead, &simpleUnit.iter, simpleUnit.end);
  MOZ_RELEASE_ASSERT(simple.isSome());
  MOZ_RELEASE_ASSERT(*simple == aExpectedCodePoint);
  MOZ_RELEASE_ASSERT(simpleUnit.iter == simpleUnit.end);

  ToUtf8Units<Char> complexUnit(aCharN, aCharN + N - 1);
  auto complex = DecodeOneUtf8CodePoint(
      complexUnit.lead, &complexUnit.iter, complexUnit.end, AssertIfCalled(),
      AssertIfCalled(), AssertIfCalled(), AssertIfCalled(), AssertIfCalled());
  MOZ_RELEASE_ASSERT(complex.isSome());
  MOZ_RELEASE_ASSERT(*complex == aExpectedCodePoint);
  MOZ_RELEASE_ASSERT(complexUnit.iter == complexUnit.end);
}

enum class InvalidUtf8Reason {
  BadLeadUnit,
  NotEnoughUnits,
  BadTrailingUnit,
  BadCodePoint,
  NotShortestForm,
};

template <typename Char, size_t N>
static void ExpectInvalidCodePointHelper(const Char (&aCharN)[N],
                                         InvalidUtf8Reason aExpectedReason,
                                         uint8_t aExpectedUnitsAvailable,
                                         uint8_t aExpectedUnitsNeeded,
                                         char32_t aExpectedBadCodePoint,
                                         uint8_t aExpectedUnitsObserved) {
  MOZ_RELEASE_ASSERT(aCharN[N - 1] == 0,
                     "array must be null-terminated for |aCharN + N - 1| to "
                     "compute the value of |aIter| as altered by "
                     "DecodeOneUtf8CodePoint");

  ToUtf8Units<Char> simpleUnit(aCharN, aCharN + N - 1);
  auto simple =
      DecodeOneUtf8CodePoint(simpleUnit.lead, &simpleUnit.iter, simpleUnit.end);
  MOZ_RELEASE_ASSERT(simple.isNothing());
  MOZ_RELEASE_ASSERT(static_cast<const void*>(simpleUnit.iter) == aCharN);

  EnumSet<InvalidUtf8Reason> reasons;
  uint8_t unitsAvailable;
  uint8_t unitsNeeded;
  char32_t badCodePoint;
  uint8_t unitsObserved;

  struct OnNotShortestForm {
    EnumSet<InvalidUtf8Reason>& reasons;
    char32_t& badCodePoint;
    uint8_t& unitsObserved;

    void operator()(char32_t aBadCodePoint, uint8_t aUnitsObserved) {
      reasons += InvalidUtf8Reason::NotShortestForm;
      badCodePoint = aBadCodePoint;
      unitsObserved = aUnitsObserved;
    }
  };

  ToUtf8Units<Char> complexUnit(aCharN, aCharN + N - 1);
  auto complex = DecodeOneUtf8CodePoint(
      complexUnit.lead, &complexUnit.iter, complexUnit.end,
      [&reasons]() { reasons += InvalidUtf8Reason::BadLeadUnit; },
      [&reasons, &unitsAvailable, &unitsNeeded](uint8_t aUnitsAvailable,
                                                uint8_t aUnitsNeeded) {
        reasons += InvalidUtf8Reason::NotEnoughUnits;
        unitsAvailable = aUnitsAvailable;
        unitsNeeded = aUnitsNeeded;
      },
      [&reasons, &unitsObserved](uint8_t aUnitsObserved) {
        reasons += InvalidUtf8Reason::BadTrailingUnit;
        unitsObserved = aUnitsObserved;
      },
      [&reasons, &badCodePoint, &unitsObserved](char32_t aBadCodePoint,
                                                uint8_t aUnitsObserved) {
        reasons += InvalidUtf8Reason::BadCodePoint;
        badCodePoint = aBadCodePoint;
        unitsObserved = aUnitsObserved;
      },
      [&reasons, &badCodePoint, &unitsObserved](char32_t aBadCodePoint,
                                                uint8_t aUnitsObserved) {
        reasons += InvalidUtf8Reason::NotShortestForm;
        badCodePoint = aBadCodePoint;
        unitsObserved = aUnitsObserved;
      });
  MOZ_RELEASE_ASSERT(complex.isNothing());
  MOZ_RELEASE_ASSERT(static_cast<const void*>(complexUnit.iter) == aCharN);

  bool alreadyIterated = false;
  for (InvalidUtf8Reason reason : reasons) {
    MOZ_RELEASE_ASSERT(!alreadyIterated);
    alreadyIterated = true;

    switch (reason) {
      case InvalidUtf8Reason::BadLeadUnit:
        break;

      case InvalidUtf8Reason::NotEnoughUnits:
        MOZ_RELEASE_ASSERT(unitsAvailable == aExpectedUnitsAvailable);
        MOZ_RELEASE_ASSERT(unitsNeeded == aExpectedUnitsNeeded);
        break;

      case InvalidUtf8Reason::BadTrailingUnit:
        MOZ_RELEASE_ASSERT(unitsObserved == aExpectedUnitsObserved);
        break;

      case InvalidUtf8Reason::BadCodePoint:
        MOZ_RELEASE_ASSERT(badCodePoint == aExpectedBadCodePoint);
        MOZ_RELEASE_ASSERT(unitsObserved == aExpectedUnitsObserved);
        break;

      case InvalidUtf8Reason::NotShortestForm:
        MOZ_RELEASE_ASSERT(badCodePoint == aExpectedBadCodePoint);
        MOZ_RELEASE_ASSERT(unitsObserved == aExpectedUnitsObserved);
        break;
    }
  }
}

// NOTE: For simplicity in treating |aCharN| identically regardless whether it's
//       a string literal or a more-generalized array, we require |aCharN| be
//       null-terminated in all these functions.

template <typename Char, size_t N>
static void ExpectBadLeadUnit(const Char (&aCharN)[N]) {
  ExpectInvalidCodePointHelper(aCharN, InvalidUtf8Reason::BadLeadUnit, 0xFF,
                               0xFF, 0xFFFFFFFF, 0xFF);
}

template <typename Char, size_t N>
static void ExpectNotEnoughUnits(const Char (&aCharN)[N],
                                 uint8_t aExpectedUnitsAvailable,
                                 uint8_t aExpectedUnitsNeeded) {
  ExpectInvalidCodePointHelper(aCharN, InvalidUtf8Reason::NotEnoughUnits,
                               aExpectedUnitsAvailable, aExpectedUnitsNeeded,
                               0xFFFFFFFF, 0xFF);
}

template <typename Char, size_t N>
static void ExpectBadTrailingUnit(const Char (&aCharN)[N],
                                  uint8_t aExpectedUnitsObserved) {
  ExpectInvalidCodePointHelper(aCharN, InvalidUtf8Reason::BadTrailingUnit, 0xFF,
                               0xFF, 0xFFFFFFFF, aExpectedUnitsObserved);
}

template <typename Char, size_t N>
static void ExpectNotShortestForm(const Char (&aCharN)[N],
                                  char32_t aExpectedBadCodePoint,
                                  uint8_t aExpectedUnitsObserved) {
  ExpectInvalidCodePointHelper(aCharN, InvalidUtf8Reason::NotShortestForm, 0xFF,
                               0xFF, aExpectedBadCodePoint,
                               aExpectedUnitsObserved);
}

template <typename Char, size_t N>
static void ExpectBadCodePoint(const Char (&aCharN)[N],
                               char32_t aExpectedBadCodePoint,
                               uint8_t aExpectedUnitsObserved) {
  ExpectInvalidCodePointHelper(aCharN, InvalidUtf8Reason::BadCodePoint, 0xFF,
                               0xFF, aExpectedBadCodePoint,
                               aExpectedUnitsObserved);
}

static void TestIsUtf8() {
  // Note we include the U+0000 NULL in this one -- and that's fine.
  static const char asciiBytes[] = u8"How about a nice game of chess?";
  MOZ_RELEASE_ASSERT(IsUtf8(Span(asciiBytes, std::size(asciiBytes))));

  static const char endNonAsciiBytes[] = u8"Life is like a ��";
  MOZ_RELEASE_ASSERT(
      IsUtf8(Span(endNonAsciiBytes, std::size(endNonAsciiBytes) - 1)));

  static const unsigned char badLeading[] = {0x80};
  MOZ_RELEASE_ASSERT(!IsUtf8(AsChars(Span(badLeading, std::size(badLeading)))));

  // Byte-counts

  // 1
  static const char oneBytes[] = u8"A";  // U+0041 LATIN CAPITAL LETTER A
  constexpr size_t oneBytesLen = std::size(oneBytes);
  static_assert(oneBytesLen == 2, "U+0041 plus nul");
  MOZ_RELEASE_ASSERT(IsUtf8(Span(oneBytes, oneBytesLen)));

  // 2
  static const char twoBytes[] = u8"؆";  // U+0606 ARABIC-INDIC CUBE ROOT
  constexpr size_t twoBytesLen = std::size(twoBytes);
  static_assert(twoBytesLen == 3, "U+0606 in two bytes plus nul");
  MOZ_RELEASE_ASSERT(IsUtf8(Span(twoBytes, twoBytesLen)));

  ExpectValidCodePoint(twoBytes, 0x0606);

  // 3
  static const char threeBytes[] = u8"᨞";  // U+1A1E BUGINESE PALLAWA
  constexpr size_t threeBytesLen = std::size(threeBytes);
  static_assert(threeBytesLen == 4, "U+1A1E in three bytes plus nul");
  MOZ_RELEASE_ASSERT(IsUtf8(Span(threeBytes, threeBytesLen)));

  ExpectValidCodePoint(threeBytes, 0x1A1E);

  // 4
  static const char fourBytes[] =
      u8"��";  // U+1F061 DOMINO TILE HORIZONTAL-06-06
  constexpr size_t fourBytesLen = std::size(fourBytes);
  static_assert(fourBytesLen == 5, "U+1F061 in four bytes plus nul");
  MOZ_RELEASE_ASSERT(IsUtf8(Span(fourBytes, fourBytesLen)));

  ExpectValidCodePoint(fourBytes, 0x1F061);

  // Max code point
  static const char maxCodePoint[] = u8"��";  // U+10FFFF
  constexpr size_t maxCodePointLen = std::size(maxCodePoint);
  static_assert(maxCodePointLen == 5, "U+10FFFF in four bytes plus nul");
  MOZ_RELEASE_ASSERT(IsUtf8(Span(maxCodePoint, maxCodePointLen)));

  ExpectValidCodePoint(maxCodePoint, 0x10FFFF);

  // One past max code point
  static const unsigned char onePastMaxCodePoint[] = {0xF4, 0x90, 0x80, 0x80,
                                                      0x0};
  constexpr size_t onePastMaxCodePointLen = std::size(onePastMaxCodePoint);
  MOZ_RELEASE_ASSERT(
      !IsUtf8(AsChars(Span(onePastMaxCodePoint, onePastMaxCodePointLen))));

  ExpectBadCodePoint(onePastMaxCodePoint, 0x110000, 4);

  // Surrogate-related testing

  // (Note that the various code unit sequences here are null-terminated to
  // simplify life for ExpectValidCodePoint, which presumes null termination.)

  static const unsigned char justBeforeSurrogates[] = {0xED, 0x9F, 0xBF, 0x0};
  constexpr size_t justBeforeSurrogatesLen =
      std::size(justBeforeSurrogates) - 1;
  MOZ_RELEASE_ASSERT(
      IsUtf8(AsChars(Span(justBeforeSurrogates, justBeforeSurrogatesLen))));

  ExpectValidCodePoint(justBeforeSurrogates, 0xD7FF);

  static const unsigned char leastSurrogate[] = {0xED, 0xA0, 0x80, 0x0};
  constexpr size_t leastSurrogateLen = std::size(leastSurrogate) - 1;
  MOZ_RELEASE_ASSERT(!IsUtf8(AsChars(Span(leastSurrogate, leastSurrogateLen))));

  ExpectBadCodePoint(leastSurrogate, 0xD800, 3);

  static const unsigned char arbitraryHighSurrogate[] = {0xED, 0xA2, 0x87, 0x0};
  constexpr size_t arbitraryHighSurrogateLen =
      std::size(arbitraryHighSurrogate) - 1;
  MOZ_RELEASE_ASSERT(!IsUtf8(
      AsChars(Span(arbitraryHighSurrogate, arbitraryHighSurrogateLen))));

  ExpectBadCodePoint(arbitraryHighSurrogate, 0xD887, 3);

  static const unsigned char arbitraryLowSurrogate[] = {0xED, 0xB7, 0xAF, 0x0};
  constexpr size_t arbitraryLowSurrogateLen =
      std::size(arbitraryLowSurrogate) - 1;
  MOZ_RELEASE_ASSERT(
      !IsUtf8(AsChars(Span(arbitraryLowSurrogate, arbitraryLowSurrogateLen))));

  ExpectBadCodePoint(arbitraryLowSurrogate, 0xDDEF, 3);

  static const unsigned char greatestSurrogate[] = {0xED, 0xBF, 0xBF, 0x0};
  constexpr size_t greatestSurrogateLen = std::size(greatestSurrogate) - 1;
  MOZ_RELEASE_ASSERT(
      !IsUtf8(AsChars(Span(greatestSurrogate, greatestSurrogateLen))));

  ExpectBadCodePoint(greatestSurrogate, 0xDFFF, 3);

  static const unsigned char justAfterSurrogates[] = {0xEE, 0x80, 0x80, 0x0};
  constexpr size_t justAfterSurrogatesLen = std::size(justAfterSurrogates) - 1;
  MOZ_RELEASE_ASSERT(
      IsUtf8(AsChars(Span(justAfterSurrogates, justAfterSurrogatesLen))));

  ExpectValidCodePoint(justAfterSurrogates, 0xE000);
}

static void TestDecodeOneValidUtf8CodePoint() {
  // NOTE: DecodeOneUtf8CodePoint decodes only *non*-ASCII code points that
  //       consist of multiple code units, so there are no ASCII tests below.

  // Length two.

  ExpectValidCodePoint(u8"€", 0x80);  // <control>
  ExpectValidCodePoint(u8"©", 0xA9);   // COPYRIGHT SIGN
  ExpectValidCodePoint(u8"¶", 0xB6);   // PILCROW SIGN
  ExpectValidCodePoint(u8"¾", 0xBE);   // VULGAR FRACTION THREE QUARTERS
  ExpectValidCodePoint(u8"÷", 0xF7);   // DIVISION SIGN
  ExpectValidCodePoint(u8"ÿ", 0xFF);   // LATIN SMALL LETTER Y WITH DIAERESIS
  ExpectValidCodePoint(u8"Ā", 0x100);  // LATIN CAPITAL LETTER A WITH MACRON
  ExpectValidCodePoint(u8"IJ", 0x132);  // LATIN CAPITAL LETTER LIGATURE IJ
  ExpectValidCodePoint(u8"ͼ", 0x37C);  // GREEK SMALL DOTTED LUNATE SIGMA SYMBOL
  ExpectValidCodePoint(u8"Ӝ",
                       0x4DC);  // CYRILLIC CAPITAL LETTER ZHE WITTH DIAERESIS
  ExpectValidCodePoint(u8"۩", 0x6E9);  // ARABIC PLACE OF SAJDAH
  ExpectValidCodePoint(u8"߿", 0x7FF);  // <not assigned>

  // Length three.

  ExpectValidCodePoint(u8"ࠀ", 0x800);  // SAMARITAN LETTER ALAF
  ExpectValidCodePoint(u8"ࡁ", 0x841);  // MANDAIC LETTER AB
  ExpectValidCodePoint(u8"ࣿ", 0x8FF);   // ARABIC MARK SIDEWAYS NOON GHUNNA
  ExpectValidCodePoint(u8"ஆ", 0xB86);  // TAMIL LETTER AA
  ExpectValidCodePoint(u8"༃",
                       0xF03);  // TIBETAN MARK GTER YIG MGO -UM GTER TSHEG MA
  ExpectValidCodePoint(
      u8"࿉",
      0xFC9);  // TIBETAN SYMBOL NOR BU (but on my system it really looks like
               // SOFT-SERVE ICE CREAM FROM ABOVE THE PLANE if you ask me)
  ExpectValidCodePoint(u8"ဪ", 0x102A);           // MYANMAR LETTER AU
  ExpectValidCodePoint(u8"ᚏ", 0x168F);           // OGHAM LETTER RUIS
  ExpectValidCodePoint("\xE2\x80\xA8", 0x2028);  // (the hated) LINE SEPARATOR
  ExpectValidCodePoint("\xE2\x80\xA9",
                       0x2029);           // (the hated) PARAGRAPH SEPARATOR
  ExpectValidCodePoint(u8"☬", 0x262C);    // ADI SHAKTI
  ExpectValidCodePoint(u8"㊮", 0x32AE);   // CIRCLED IDEOGRAPH RESOURCE
  ExpectValidCodePoint(u8"㏖", 0x33D6);   // SQUARE MOL
  ExpectValidCodePoint(u8"ꔄ", 0xA504);    // VAI SYLLABLE WEEN
  ExpectValidCodePoint(u8"ퟕ", 0xD7D5);    // HANGUL JONGSEONG RIEUL-SSANGKIYEOK
  ExpectValidCodePoint(u8"퟿", 0xD7FF);  // <not assigned>
  ExpectValidCodePoint(u8"", 0xE000);  // <Private Use>
  ExpectValidCodePoint(u8"鱗", 0xF9F2);   // CJK COMPATIBILITY IDEOGRAPH-F9F
  ExpectValidCodePoint(
      u8"﷽", 0xFDFD);  // ARABIC LIGATURE BISMILLAH AR-RAHMAN AR-RAHHHEEEEM
  ExpectValidCodePoint(u8"￿", 0xFFFF);  // <not assigned>

  // Length four.
  ExpectValidCodePoint(u8"��", 0x10000);      // LINEAR B SYLLABLE B008 A
  ExpectValidCodePoint(u8"��", 0x14440);      // ANATOLIAN HIEROGLYPH A058
  ExpectValidCodePoint(u8"��", 0x1D6D7);      // MATHEMATICAL BOLD SMALL PHI
  ExpectValidCodePoint(u8"��", 0x1F4A9);     // PILE OF POO
  ExpectValidCodePoint(u8"��", 0x1F52B);     // PISTOL
  ExpectValidCodePoint(u8"��", 0x1F94C);     // CURLING STONE
  ExpectValidCodePoint(u8"��", 0x1F94F);     // FLYING DISC
  ExpectValidCodePoint(u8"��", 0x20346);     // CJK UNIFIED IDEOGRAPH-20346
  ExpectValidCodePoint(u8"��", 0x2183A);     // CJK UNIFIED IDEOGRAPH-2183A
  ExpectValidCodePoint(u8"��", 0x417F6);   // <not assigned>
  ExpectValidCodePoint(u8"��", 0x7E836);   // <not assigned>
  ExpectValidCodePoint(u8"��", 0xFEF67);   // <Plane 15 Private Use>
  ExpectValidCodePoint(u8"��", 0x10FFFF);  //
}

static void TestDecodeBadLeadUnit() {
  // These tests are actually exhaustive.

  unsigned char badLead[] = {'\0', '\0'};

  for (uint8_t lead : IntegerRange(0b1000'0000, 0b1100'0000)) {
    badLead[0] = lead;
    ExpectBadLeadUnit(badLead);
  }

  {
    uint8_t lead = 0b1111'1000;
    do {
      badLead[0] = lead;
      ExpectBadLeadUnit(badLead);
      if (lead == 0b1111'1111) {
        break;
      }

      lead++;
    } while (true);
  }
}

static void TestTooFewOrBadTrailingUnits() {
  // Lead unit indicates a two-byte code point.

  char truncatedTwo[] = {'\0', '\0'};
  char badTrailTwo[] = {'\0', '\0', '\0'};

  for (uint8_t lead : IntegerRange(0b1100'0000, 0b1110'0000)) {
    truncatedTwo[0] = lead;
    ExpectNotEnoughUnits(truncatedTwo, 1, 2);

    badTrailTwo[0] = lead;
    for (uint8_t trail : IntegerRange(0b0000'0000, 0b1000'0000)) {
      badTrailTwo[1] = trail;
      ExpectBadTrailingUnit(badTrailTwo, 2);
    }

    for (uint8_t trail : IntegerRange(0b1100'0000, 0b1111'1111)) {
      badTrailTwo[1] = trail;
      ExpectBadTrailingUnit(badTrailTwo, 2);
    }
  }

  // Lead unit indicates a three-byte code point.

  char truncatedThreeOne[] = {'\0', '\0'};
  char truncatedThreeTwo[] = {'\0', '\0', '\0'};
  unsigned char badTrailThree[] = {'\0', '\0', '\0', '\0'};

  for (uint8_t lead : IntegerRange(0b1110'0000, 0b1111'0000)) {
    truncatedThreeOne[0] = lead;
    ExpectNotEnoughUnits(truncatedThreeOne, 1, 3);

    truncatedThreeTwo[0] = lead;
    ExpectNotEnoughUnits(truncatedThreeTwo, 2, 3);

    badTrailThree[0] = lead;
    badTrailThree[2] = 0b1011'1111;  // make valid to test overreads
    for (uint8_t mid : IntegerRange(0b0000'0000, 0b1000'0000)) {
      badTrailThree[1] = mid;
      ExpectBadTrailingUnit(badTrailThree, 2);
    }
    {
      uint8_t mid = 0b1100'0000;
      do {
        badTrailThree[1] = mid;
        ExpectBadTrailingUnit(badTrailThree, 2);
        if (mid == 0b1111'1111) {
          break;
        }

        mid++;
      } while (true);
    }

    badTrailThree[1] = 0b1011'1111;
    for (uint8_t last : IntegerRange(0b0000'0000, 0b1000'0000)) {
      badTrailThree[2] = last;
      ExpectBadTrailingUnit(badTrailThree, 3);
    }
    {
      uint8_t last = 0b1100'0000;
      do {
        badTrailThree[2] = last;
        ExpectBadTrailingUnit(badTrailThree, 3);
        if (last == 0b1111'1111) {
          break;
        }

        last++;
      } while (true);
    }
  }

  // Lead unit indicates a four-byte code point.

  char truncatedFourOne[] = {'\0', '\0'};
  char truncatedFourTwo[] = {'\0', '\0', '\0'};
  char truncatedFourThree[] = {'\0', '\0', '\0', '\0'};

  unsigned char badTrailFour[] = {'\0', '\0', '\0', '\0', '\0'};

  for (uint8_t lead : IntegerRange(0b1111'0000, 0b1111'1000)) {
    truncatedFourOne[0] = lead;
    ExpectNotEnoughUnits(truncatedFourOne, 1, 4);

    truncatedFourTwo[0] = lead;
    ExpectNotEnoughUnits(truncatedFourTwo, 2, 4);

    truncatedFourThree[0] = lead;
    ExpectNotEnoughUnits(truncatedFourThree, 3, 4);

    badTrailFour[0] = lead;
    badTrailFour[2] = badTrailFour[3] = 0b1011'1111;  // test for overreads
    for (uint8_t second : IntegerRange(0b0000'0000, 0b1000'0000)) {
      badTrailFour[1] = second;
      ExpectBadTrailingUnit(badTrailFour, 2);
    }
    {
      uint8_t second = 0b1100'0000;
      do {
        badTrailFour[1] = second;
        ExpectBadTrailingUnit(badTrailFour, 2);
        if (second == 0b1111'1111) {
          break;
        }

        second++;
      } while (true);
    }

    badTrailFour[1] = badTrailFour[3] = 0b1011'1111;  // test for overreads
    for (uint8_t third : IntegerRange(0b0000'0000, 0b1000'0000)) {
      badTrailFour[2] = third;
      ExpectBadTrailingUnit(badTrailFour, 3);
    }
    {
      uint8_t third = 0b1100'0000;
      do {
        badTrailFour[2] = third;
        ExpectBadTrailingUnit(badTrailFour, 3);
        if (third == 0b1111'1111) {
          break;
        }

        third++;
      } while (true);
    }

    badTrailFour[2] = 0b1011'1111;
    for (uint8_t fourth : IntegerRange(0b0000'0000, 0b1000'0000)) {
      badTrailFour[3] = fourth;
      ExpectBadTrailingUnit(badTrailFour, 4);
    }
    {
      uint8_t fourth = 0b1100'0000;
      do {
        badTrailFour[3] = fourth;
        ExpectBadTrailingUnit(badTrailFour, 4);
        if (fourth == 0b1111'1111) {
          break;
        }

        fourth++;
      } while (true);
    }
  }
}

static void TestBadSurrogate() {
  // These tests are actually exhaustive.

  ExpectValidCodePoint("\xED\x9F\xBF", 0xD7FF);  // last before surrogates
  ExpectValidCodePoint("\xEE\x80\x80", 0xE000);  // first after surrogates

  // First invalid surrogate encoding is { 0xED, 0xA0, 0x80 }.  Last invalid
  // surrogate encoding is { 0xED, 0xBF, 0xBF }.

  char badSurrogate[] = {'\xED', '\0', '\0', '\0'};

  for (char32_t c = 0xD800; c < 0xE000; c++) {
    badSurrogate[1] = 0b1000'0000 ^ ((c & 0b1111'1100'0000) >> 6);
    badSurrogate[2] = 0b1000'0000 ^ ((c & 0b0000'0011'1111));

    ExpectBadCodePoint(badSurrogate, c, 3);
  }
}

static void TestBadTooBig() {
  // These tests are actually exhaustive.

  ExpectValidCodePoint("\xF4\x8F\xBF\xBF", 0x10'FFFF);  // last code point

  // Four-byte code points are
  //
  //   0b1111'0xxx 0b10xx'xxxx 0b10xx'xxxx 0b10xx'xxxx
  //
  // with 3 + 6 + 6 + 6 == 21 unconstrained bytes, so the structurally
  // representable limit (exclusive) is 2**21 - 1 == 2097152.

  char tooLargeCodePoint[] = {'\0', '\0', '\0', '\0', '\0'};

  for (char32_t c = 0x11'0000; c < (1 << 21); c++) {
    tooLargeCodePoint[0] =
        0b1111'0000 ^ ((c & 0b1'1100'0000'0000'0000'0000) >> 18);
    tooLargeCodePoint[1] =
        0b1000'0000 ^ ((c & 0b0'0011'1111'0000'0000'0000) >> 12);
    tooLargeCodePoint[2] =
        0b1000'0000 ^ ((c & 0b0'0000'0000'1111'1100'0000) >> 6);
    tooLargeCodePoint[3] = 0b1000'0000 ^ ((c & 0b0'0000'0000'0000'0011'1111));

    ExpectBadCodePoint(tooLargeCodePoint, c, 4);
  }
}

static void TestBadCodePoint() {
  TestBadSurrogate();
  TestBadTooBig();
}

static void TestNotShortestForm() {
  {
    // One-byte in two-byte.

    char oneInTwo[] = {'\0', '\0', '\0'};

    for (char32_t c = '\0'; c < 0x80; c++) {
      oneInTwo[0] = 0b1100'0000 ^ ((c & 0b0111'1100'0000) >> 6);
      oneInTwo[1] = 0b1000'0000 ^ ((c & 0b0000'0011'1111));

      ExpectNotShortestForm(oneInTwo, c, 2);
    }

    // One-byte in three-byte.

    char oneInThree[] = {'\0', '\0', '\0', '\0'};

    for (char32_t c = '\0'; c < 0x80; c++) {
      oneInThree[0] = 0b1110'0000 ^ ((c & 0b1111'0000'0000'0000) >> 12);
      oneInThree[1] = 0b1000'0000 ^ ((c & 0b0000'1111'1100'0000) >> 6);
      oneInThree[2] = 0b1000'0000 ^ ((c & 0b0000'0000'0011'1111));

      ExpectNotShortestForm(oneInThree, c, 3);
    }

    // One-byte in four-byte.

    char oneInFour[] = {'\0', '\0', '\0', '\0', '\0'};

    for (char32_t c = '\0'; c < 0x80; c++) {
      oneInFour[0] = 0b1111'0000 ^ ((c & 0b1'1100'0000'0000'0000'0000) >> 18);
      oneInFour[1] = 0b1000'0000 ^ ((c & 0b0'0011'1111'0000'0000'0000) >> 12);
      oneInFour[2] = 0b1000'0000 ^ ((c & 0b0'0000'0000'1111'1100'0000) >> 6);
      oneInFour[3] = 0b1000'0000 ^ ((c & 0b0'0000'0000'0000'0011'1111));

      ExpectNotShortestForm(oneInFour, c, 4);
    }
  }

  {
    // Two-byte in three-byte.

    char twoInThree[] = {'\0', '\0', '\0', '\0'};

    for (char32_t c = 0x80; c < 0x800; c++) {
      twoInThree[0] = 0b1110'0000 ^ ((c & 0b1111'0000'0000'0000) >> 12);
      twoInThree[1] = 0b1000'0000 ^ ((c & 0b0000'1111'1100'0000) >> 6);
      twoInThree[2] = 0b1000'0000 ^ ((c & 0b0000'0000'0011'1111));

      ExpectNotShortestForm(twoInThree, c, 3);
    }

    // Two-byte in four-byte.

    char twoInFour[] = {'\0', '\0', '\0', '\0', '\0'};

    for (char32_t c = 0x80; c < 0x800; c++) {
      twoInFour[0] = 0b1111'0000 ^ ((c & 0b1'1100'0000'0000'0000'0000) >> 18);
      twoInFour[1] = 0b1000'0000 ^ ((c & 0b0'0011'1111'0000'0000'0000) >> 12);
      twoInFour[2] = 0b1000'0000 ^ ((c & 0b0'0000'0000'1111'1100'0000) >> 6);
      twoInFour[3] = 0b1000'0000 ^ ((c & 0b0'0000'0000'0000'0011'1111));

      ExpectNotShortestForm(twoInFour, c, 4);
    }
  }

  {
    // Three-byte in four-byte.

    char threeInFour[] = {'\0', '\0', '\0', '\0', '\0'};

    for (char32_t c = 0x800; c < 0x1'0000; c++) {
      threeInFour[0] = 0b1111'0000 ^ ((c & 0b1'1100'0000'0000'0000'0000) >> 18);
      threeInFour[1] = 0b1000'0000 ^ ((c & 0b0'0011'1111'0000'0000'0000) >> 12);
      threeInFour[2] = 0b1000'0000 ^ ((c & 0b0'0000'0000'1111'1100'0000) >> 6);
      threeInFour[3] = 0b1000'0000 ^ ((c & 0b0'0000'0000'0000'0011'1111));

      ExpectNotShortestForm(threeInFour, c, 4);
    }
  }
}

static void TestDecodeOneInvalidUtf8CodePoint() {
  TestDecodeBadLeadUnit();
  TestTooFewOrBadTrailingUnits();
  TestBadCodePoint();
  TestNotShortestForm();
}

static void TestDecodeOneUtf8CodePoint() {
  TestDecodeOneValidUtf8CodePoint();
  TestDecodeOneInvalidUtf8CodePoint();
}

int main() {
  TestUtf8Unit();
  TestIsUtf8();
  TestDecodeOneUtf8CodePoint();
  return 0;
}