| 1 | /* |
| 2 | SPDX-FileCopyrightText: 2025 Stefan BrĂ¼ns <stefan.bruens@rwth-aachen.de> |
| 3 | SPDX-License-Identifier: LGPL-2.1-or-later |
| 4 | */ |
| 5 | |
| 6 | #include "../lib/decompressor.h" |
| 7 | |
| 8 | #include <QBuffer> |
| 9 | #include <QTest> |
| 10 | #include <QVector> |
| 11 | #include <QtEndian> |
| 12 | #include <array> |
| 13 | |
| 14 | using namespace Mobipocket; |
| 15 | |
| 16 | class DecompressorTest : public QObject |
| 17 | { |
| 18 | Q_OBJECT |
| 19 | private Q_SLOTS: |
| 20 | void testNoop(); |
| 21 | void testNoop_data(); |
| 22 | void testRLE(); |
| 23 | void testRLE_data(); |
| 24 | void testHuffInit(); |
| 25 | void testHuffDecompress(); |
| 26 | void testFuzzHuff(); |
| 27 | void benchmarkHuffDecompress(); |
| 28 | }; |
| 29 | |
| 30 | namespace { |
| 31 | QVector<QByteArray> createHuffIdentityDict() |
| 32 | { |
| 33 | // Create a Huffman dictionary which maps each input byte to itself |
| 34 | static std::array<quint8, 256 * 4> hdict = []() { |
| 35 | std::array<quint8, 256 * 4> d; |
| 36 | for (size_t i = 0; i < d.size(); i += 4) { |
| 37 | // 1. Codelen is 8 bits |
| 38 | // 2. Only use the first tree dictionary, set the termination flag |
| 39 | d[i] = 8 | 0x80; |
| 40 | d[i + 1] = i / 2; |
| 41 | d[i + 2] = i / 512; |
| 42 | } |
| 43 | return d; |
| 44 | }(); |
| 45 | |
| 46 | QByteArray huff("HUFF" , 4); |
| 47 | huff.resize(size: 24); |
| 48 | qToBigEndian<quint32>(src: huff.size(), dest: huff.data() + 16); |
| 49 | #if QT_VERSION > QT_VERSION_CHECK(6, 0, 0) |
| 50 | huff.append(a: QByteArrayView(hdict)); |
| 51 | #else |
| 52 | huff.append(QByteArray::fromRawData(reinterpret_cast<char*>(hdict.data()), hdict.size())); |
| 53 | #endif |
| 54 | qToBigEndian<quint32>(src: huff.size(), dest: huff.data() + 20); |
| 55 | huff.append(n: 64 * 4, ch: '\0'); |
| 56 | |
| 57 | static std::array<quint8, 256 * (2 + 3)> entries = []() { |
| 58 | std::array<quint8, 256 * (2 + 3)> d; |
| 59 | for (size_t i = 0; i < 256; i++) { |
| 60 | quint16 off = 512 + 3 * i; |
| 61 | qToBigEndian<quint16>(src: off, dest: &d[2 * i]); |
| 62 | qToBigEndian<quint16>(src: 0x8001, dest: &d[off]); // len==1 | termination flag |
| 63 | d[off + 2] = i; |
| 64 | } |
| 65 | return d; |
| 66 | }(); |
| 67 | |
| 68 | QByteArray cdic("CDIC\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0" , 16); |
| 69 | qToBigEndian<quint32>(src: 32, dest: cdic.data() + 12); |
| 70 | #if QT_VERSION > QT_VERSION_CHECK(6, 0, 0) |
| 71 | cdic.append(a: QByteArrayView(entries)); |
| 72 | #else |
| 73 | cdic.append(QByteArray::fromRawData(reinterpret_cast<char*>(entries.data()), entries.size())); |
| 74 | #endif |
| 75 | |
| 76 | return {huff, cdic}; |
| 77 | } |
| 78 | } |
| 79 | |
| 80 | void DecompressorTest::testNoop() |
| 81 | { |
| 82 | QFETCH(QByteArray, data); |
| 83 | |
| 84 | auto decompressor = Decompressor::create(type: 1, auxData: {}); |
| 85 | |
| 86 | auto r = decompressor->decompress(data); |
| 87 | // NOOP -> input and output are identical |
| 88 | QCOMPARE(r, data); |
| 89 | } |
| 90 | |
| 91 | void DecompressorTest::testNoop_data() |
| 92 | { |
| 93 | QTest::addColumn<QByteArray>(name: "data" ); |
| 94 | |
| 95 | QTest::addRow(format: "empty" ) << QByteArray(); |
| 96 | QTest::addRow(format: "0x00 * 10" ) << QByteArray(10, '\x00'); |
| 97 | QTest::addRow(format: "0xaa * 10" ) << QByteArray(10, '\xaa'); |
| 98 | } |
| 99 | |
| 100 | void DecompressorTest::testRLE() |
| 101 | { |
| 102 | QFETCH(QByteArray, data); |
| 103 | QFETCH(QByteArray, expected); |
| 104 | |
| 105 | auto decompressor = Decompressor::create(type: 2, auxData: {}); |
| 106 | |
| 107 | auto r = decompressor->decompress(data); |
| 108 | QCOMPARE(r, expected); |
| 109 | } |
| 110 | |
| 111 | void DecompressorTest::testRLE_data() |
| 112 | { |
| 113 | QTest::addColumn<QByteArray>(name: "data" ); |
| 114 | QTest::addColumn<QByteArray>(name: "expected" ); |
| 115 | |
| 116 | QTest::addRow(format: "empty" ) << QByteArray() << QByteArray(); |
| 117 | // Token '0x00' is passed verbatim |
| 118 | QTest::addRow(format: "0x00 * 10" ) << QByteArray(10, '\x00') << QByteArray(10, '\x00'); |
| 119 | |
| 120 | // Tokens in the range 0x01..0x08 denotes the length of raw copied data |
| 121 | QTest::addRow(format: "raw 0x01..." ) << QByteArray("\x01\xff" , 2) << QByteArray("\xff" , 1); |
| 122 | QTest::addRow(format: "raw .0x01..." ) << QByteArray("d\x01\xc0kj" , 5) << QByteArray("d\xc0kj" , 4); |
| 123 | QTest::addRow(format: "raw .0x02..." ) << QByteArray("d\x02\xc0kj" , 5) << QByteArray("d\xc0kj" , 4); |
| 124 | QTest::addRow(format: "raw .0x03..." ) << QByteArray("d\x03\xc0kj" , 5) << QByteArray("d\xc0kj" , 4); |
| 125 | // Short data |
| 126 | QTest::addRow(format: "short 0x03..." ) << QByteArray("d\x03\xc0k" , 4) << QByteArray("d" , 1); |
| 127 | |
| 128 | // Tokens in the range 0x09..0x7f are passed verbatim |
| 129 | QTest::addRow(format: "0x20 * 20" ) << QByteArray(20, '\x20') << QByteArray(20, '\x20'); |
| 130 | QTest::addRow(format: "0x7f * 20" ) << QByteArray(20, '\x7f') << QByteArray(20, '\x7f'); |
| 131 | |
| 132 | // Tokens in the range 0xc0..0xff are expanded to " \x40".." \x7f" |
| 133 | QTest::addRow(format: "0xc0 * 64" ) << QByteArray(64, '\xc0') << QByteArray(" \x40" , 2).repeated(times: 64); |
| 134 | QTest::addRow(format: "0xf0 * 64" ) << QByteArray(64, '\xf0') << QByteArray(" \x70" , 2).repeated(times: 64); |
| 135 | |
| 136 | QTest::addRow(format: "repeat" ) << QByteArray("\x32\x80\x0a" , 3) << QByteArray(6, '2'); |
| 137 | QTest::addRow(format: "repeat 2" ) << QByteArray("\x31\x65\x80\x13" , 4) << QByteArray("1e" ).repeated(times: 4); |
| 138 | } |
| 139 | |
| 140 | void DecompressorTest::testHuffInit() |
| 141 | { |
| 142 | { |
| 143 | auto decompressor = Decompressor::create(type: 'H', auxData: {}); |
| 144 | QVERIFY(!decompressor->isValid()); |
| 145 | } |
| 146 | { |
| 147 | auto decompressor = Decompressor::create(type: 'H', auxData: QVector<QByteArray>(2)); |
| 148 | QVERIFY(!decompressor->isValid()); |
| 149 | } |
| 150 | { |
| 151 | QByteArray HDic(512, '\0'); |
| 152 | QByteArray CDic(512, '\0'); |
| 153 | |
| 154 | auto decompressor = Decompressor::create(type: 'H', auxData: {HDic, CDic}); |
| 155 | QVERIFY(!decompressor->isValid()); |
| 156 | } |
| 157 | { |
| 158 | QByteArray HDic("HUFF" , 4); |
| 159 | QByteArray CDic("CDIC" , 4); |
| 160 | QByteArray fill(60, '\0'); |
| 161 | |
| 162 | auto decompressor = Decompressor::create(type: 'H', auxData: {HDic + fill, CDic + fill}); |
| 163 | QVERIFY(!decompressor->isValid()); |
| 164 | } |
| 165 | { |
| 166 | auto decompressor = Decompressor::create(type: 'H', auxData: createHuffIdentityDict()); |
| 167 | QVERIFY(decompressor->isValid()); |
| 168 | } |
| 169 | } |
| 170 | |
| 171 | void DecompressorTest::testHuffDecompress() |
| 172 | { |
| 173 | auto decompressor = Decompressor::create(type: 'H', auxData: createHuffIdentityDict()); |
| 174 | QVERIFY(decompressor->isValid()); |
| 175 | |
| 176 | { |
| 177 | auto r = decompressor->decompress(data: QByteArray("\0" , 1)); |
| 178 | QCOMPARE(r, QByteArray("\0" , 1)); |
| 179 | } |
| 180 | { |
| 181 | auto r = decompressor->decompress(data: QByteArray("\1\xcc" , 2)); |
| 182 | QCOMPARE(r, QByteArray("\1\xcc" , 2)); |
| 183 | } |
| 184 | { |
| 185 | QByteArray d(256, '\0'); |
| 186 | for (int i = 0; i < d.size(); i++) { |
| 187 | d[i] = i; |
| 188 | } |
| 189 | auto r = decompressor->decompress(data: d); |
| 190 | QCOMPARE(r, d); |
| 191 | } |
| 192 | } |
| 193 | |
| 194 | void DecompressorTest::benchmarkHuffDecompress() |
| 195 | { |
| 196 | auto decompressor = Decompressor::create(type: 'H', auxData: createHuffIdentityDict()); |
| 197 | QVERIFY(decompressor->isValid()); |
| 198 | |
| 199 | QByteArray data(1024, '\x01'); |
| 200 | |
| 201 | QBENCHMARK { |
| 202 | auto r = decompressor->decompress(data); |
| 203 | QCOMPARE(r, data); |
| 204 | } |
| 205 | } |
| 206 | |
| 207 | void DecompressorTest::testFuzzHuff() |
| 208 | { |
| 209 | auto verify = [](const auto &decompressor) { |
| 210 | QByteArray d(256, '\0'); |
| 211 | for (int i = 0; i < d.size(); i++) { |
| 212 | d[i] = i; |
| 213 | } |
| 214 | // The output does not matter and is likely |
| 215 | // just garbage, but it should not crash |
| 216 | auto r = decompressor->decompress(d); |
| 217 | }; |
| 218 | |
| 219 | auto dict = createHuffIdentityDict(); |
| 220 | |
| 221 | for (auto i = dict.at(i: 0).size() - 1; i >= 0; i--) { |
| 222 | unsigned char originalValue = dict.at(i: 0).at(i); |
| 223 | |
| 224 | { |
| 225 | dict[0][i] = originalValue ^ 0xff; |
| 226 | auto decompressor = Decompressor::create(type: 'H', auxData: dict); |
| 227 | verify(decompressor); |
| 228 | } |
| 229 | |
| 230 | if ((originalValue == 0) || (originalValue == 0xff)) { |
| 231 | dict[0][i] = originalValue; |
| 232 | continue; |
| 233 | } |
| 234 | |
| 235 | { |
| 236 | dict[0][i] = 0; |
| 237 | auto decompressor = Decompressor::create(type: 'H', auxData: dict); |
| 238 | verify(decompressor); |
| 239 | } |
| 240 | |
| 241 | { |
| 242 | dict[0][i] = static_cast<unsigned char>(0xff); |
| 243 | auto decompressor = Decompressor::create(type: 'H', auxData: dict); |
| 244 | verify(decompressor); |
| 245 | } |
| 246 | |
| 247 | dict[0][i] = originalValue; |
| 248 | } |
| 249 | } |
| 250 | |
| 251 | QTEST_GUILESS_MAIN(DecompressorTest) |
| 252 | |
| 253 | #include "decompressortest.moc" |
| 254 | |