/ src / bech32.cpp
bech32.cpp
  1  // Copyright (c) 2017, 2021 Pieter Wuille
  2  // Copyright (c) 2021-2022 The Bitcoin Core developers
  3  // Distributed under the MIT software license, see the accompanying
  4  // file COPYING or http://www.opensource.org/licenses/mit-license.php.
  5  
  6  #include <bech32.h>
  7  #include <util/vector.h>
  8  
  9  #include <array>
 10  #include <assert.h>
 11  #include <numeric>
 12  #include <optional>
 13  
 14  namespace bech32
 15  {
 16  
 17  namespace
 18  {
 19  
 20  typedef std::vector<uint8_t> data;
 21  
 22  /** The Bech32 and Bech32m character set for encoding. */
 23  const char* CHARSET = "qpzry9x8gf2tvdw0s3jn54khce6mua7l";
 24  
 25  /** The Bech32 and Bech32m character set for decoding. */
 26  const int8_t CHARSET_REV[128] = {
 27      -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
 28      -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
 29      -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
 30      15, -1, 10, 17, 21, 20, 26, 30,  7,  5, -1, -1, -1, -1, -1, -1,
 31      -1, 29, -1, 24, 13, 25,  9,  8, 23, -1, 18, 22, 31, 27, 19, -1,
 32       1,  0,  3, 16, 11, 28, 12, 14,  6,  4,  2, -1, -1, -1, -1, -1,
 33      -1, 29, -1, 24, 13, 25,  9,  8, 23, -1, 18, 22, 31, 27, 19, -1,
 34       1,  0,  3, 16, 11, 28, 12, 14,  6,  4,  2, -1, -1, -1, -1, -1
 35  };
 36  
 37  /** We work with the finite field GF(1024) defined as a degree 2 extension of the base field GF(32)
 38   * The defining polynomial of the extension is x^2 + 9x + 23.
 39   * Let (e) be a root of this defining polynomial. Then (e) is a primitive element of GF(1024),
 40   * that is, a generator of the field. Every non-zero element of the field can then be represented
 41   * as (e)^k for some power k.
 42   * The array GF1024_EXP contains all these powers of (e) - GF1024_EXP[k] = (e)^k in GF(1024).
 43   * Conversely, GF1024_LOG contains the discrete logarithms of these powers, so
 44   * GF1024_LOG[GF1024_EXP[k]] == k.
 45   * The following function generates the two tables GF1024_EXP and GF1024_LOG as constexprs. */
 46  constexpr std::pair<std::array<int16_t, 1023>, std::array<int16_t, 1024>> GenerateGFTables()
 47  {
 48      // Build table for GF(32).
 49      // We use these tables to perform arithmetic in GF(32) below, when constructing the
 50      // tables for GF(1024).
 51      std::array<int8_t, 31> GF32_EXP{};
 52      std::array<int8_t, 32> GF32_LOG{};
 53  
 54      // fmod encodes the defining polynomial of GF(32) over GF(2), x^5 + x^3 + 1.
 55      // Because coefficients in GF(2) are binary digits, the coefficients are packed as 101001.
 56      const int fmod = 41;
 57  
 58      // Elements of GF(32) are encoded as vectors of length 5 over GF(2), that is,
 59      // 5 binary digits. Each element (b_4, b_3, b_2, b_1, b_0) encodes a polynomial
 60      // b_4*x^4 + b_3*x^3 + b_2*x^2 + b_1*x^1 + b_0 (modulo fmod).
 61      // For example, 00001 = 1 is the multiplicative identity.
 62      GF32_EXP[0] = 1;
 63      GF32_LOG[0] = -1;
 64      GF32_LOG[1] = 0;
 65      int v = 1;
 66      for (int i = 1; i < 31; ++i) {
 67          // Multiplication by x is the same as shifting left by 1, as
 68          // every coefficient of the polynomial is moved up one place.
 69          v = v << 1;
 70          // If the polynomial now has an x^5 term, we subtract fmod from it
 71          // to remain working modulo fmod. Subtraction is the same as XOR in characteristic
 72          // 2 fields.
 73          if (v & 32) v ^= fmod;
 74          GF32_EXP[i] = v;
 75          GF32_LOG[v] = i;
 76      }
 77  
 78      // Build table for GF(1024)
 79      std::array<int16_t, 1023> GF1024_EXP{};
 80      std::array<int16_t, 1024> GF1024_LOG{};
 81  
 82      GF1024_EXP[0] = 1;
 83      GF1024_LOG[0] = -1;
 84      GF1024_LOG[1] = 0;
 85  
 86      // Each element v of GF(1024) is encoded as a 10 bit integer in the following way:
 87      // v = v1 || v0 where v0, v1 are 5-bit integers (elements of GF(32)).
 88      // The element (e) is encoded as 1 || 0, to represent 1*(e) + 0. Every other element
 89      // a*(e) + b is represented as a || b (a and b are both GF(32) elements). Given (v),
 90      // we compute (e)*(v) by multiplying in the following way:
 91      //
 92      // v0' = 23*v1
 93      // v1' = 9*v1 + v0
 94      // e*v = v1' || v0'
 95      //
 96      // Where 23, 9 are GF(32) elements encoded as described above. Multiplication in GF(32)
 97      // is done using the log/exp tables:
 98      // e^x * e^y = e^(x + y) so a * b = EXP[ LOG[a] + LOG [b] ]
 99      // for non-zero a and b.
100  
101      v = 1;
102      for (int i = 1; i < 1023; ++i) {
103          int v0 = v & 31;
104          int v1 = v >> 5;
105  
106          int v0n = v1 ? GF32_EXP.at((GF32_LOG.at(v1) + GF32_LOG.at(23)) % 31) : 0;
107          int v1n = (v1 ? GF32_EXP.at((GF32_LOG.at(v1) + GF32_LOG.at(9)) % 31) : 0) ^ v0;
108  
109          v = v1n << 5 | v0n;
110          GF1024_EXP[i] = v;
111          GF1024_LOG[v] = i;
112      }
113  
114      return std::make_pair(GF1024_EXP, GF1024_LOG);
115  }
116  
117  constexpr auto tables = GenerateGFTables();
118  constexpr const std::array<int16_t, 1023>& GF1024_EXP = tables.first;
119  constexpr const std::array<int16_t, 1024>& GF1024_LOG = tables.second;
120  
121  /* Determine the final constant to use for the specified encoding. */
122  uint32_t EncodingConstant(Encoding encoding) {
123      assert(encoding == Encoding::BECH32 || encoding == Encoding::BECH32M);
124      return encoding == Encoding::BECH32 ? 1 : 0x2bc830a3;
125  }
126  
127  /** This function will compute what 6 5-bit values to XOR into the last 6 input values, in order to
128   *  make the checksum 0. These 6 values are packed together in a single 30-bit integer. The higher
129   *  bits correspond to earlier values. */
130  uint32_t PolyMod(const data& v)
131  {
132      // The input is interpreted as a list of coefficients of a polynomial over F = GF(32), with an
133      // implicit 1 in front. If the input is [v0,v1,v2,v3,v4], that polynomial is v(x) =
134      // 1*x^5 + v0*x^4 + v1*x^3 + v2*x^2 + v3*x + v4. The implicit 1 guarantees that
135      // [v0,v1,v2,...] has a distinct checksum from [0,v0,v1,v2,...].
136  
137      // The output is a 30-bit integer whose 5-bit groups are the coefficients of the remainder of
138      // v(x) mod g(x), where g(x) is the Bech32 generator,
139      // x^6 + {29}x^5 + {22}x^4 + {20}x^3 + {21}x^2 + {29}x + {18}. g(x) is chosen in such a way
140      // that the resulting code is a BCH code, guaranteeing detection of up to 3 errors within a
141      // window of 1023 characters. Among the various possible BCH codes, one was selected to in
142      // fact guarantee detection of up to 4 errors within a window of 89 characters.
143  
144      // Note that the coefficients are elements of GF(32), here represented as decimal numbers
145      // between {}. In this finite field, addition is just XOR of the corresponding numbers. For
146      // example, {27} + {13} = {27 ^ 13} = {22}. Multiplication is more complicated, and requires
147      // treating the bits of values themselves as coefficients of a polynomial over a smaller field,
148      // GF(2), and multiplying those polynomials mod a^5 + a^3 + 1. For example, {5} * {26} =
149      // (a^2 + 1) * (a^4 + a^3 + a) = (a^4 + a^3 + a) * a^2 + (a^4 + a^3 + a) = a^6 + a^5 + a^4 + a
150      // = a^3 + 1 (mod a^5 + a^3 + 1) = {9}.
151  
152      // During the course of the loop below, `c` contains the bitpacked coefficients of the
153      // polynomial constructed from just the values of v that were processed so far, mod g(x). In
154      // the above example, `c` initially corresponds to 1 mod g(x), and after processing 2 inputs of
155      // v, it corresponds to x^2 + v0*x + v1 mod g(x). As 1 mod g(x) = 1, that is the starting value
156      // for `c`.
157  
158      // The following Sage code constructs the generator used:
159      //
160      // B = GF(2) # Binary field
161      // BP.<b> = B[] # Polynomials over the binary field
162      // F_mod = b**5 + b**3 + 1
163      // F.<f> = GF(32, modulus=F_mod, repr='int') # GF(32) definition
164      // FP.<x> = F[] # Polynomials over GF(32)
165      // E_mod = x**2 + F.fetch_int(9)*x + F.fetch_int(23)
166      // E.<e> = F.extension(E_mod) # GF(1024) extension field definition
167      // for p in divisors(E.order() - 1): # Verify e has order 1023.
168      //    assert((e**p == 1) == (p % 1023 == 0))
169      // G = lcm([(e**i).minpoly() for i in range(997,1000)])
170      // print(G) # Print out the generator
171      //
172      // It demonstrates that g(x) is the least common multiple of the minimal polynomials
173      // of 3 consecutive powers (997,998,999) of a primitive element (e) of GF(1024).
174      // That guarantees it is, in fact, the generator of a primitive BCH code with cycle
175      // length 1023 and distance 4. See https://en.wikipedia.org/wiki/BCH_code for more details.
176  
177      uint32_t c = 1;
178      for (const auto v_i : v) {
179          // We want to update `c` to correspond to a polynomial with one extra term. If the initial
180          // value of `c` consists of the coefficients of c(x) = f(x) mod g(x), we modify it to
181          // correspond to c'(x) = (f(x) * x + v_i) mod g(x), where v_i is the next input to
182          // process. Simplifying:
183          // c'(x) = (f(x) * x + v_i) mod g(x)
184          //         ((f(x) mod g(x)) * x + v_i) mod g(x)
185          //         (c(x) * x + v_i) mod g(x)
186          // If c(x) = c0*x^5 + c1*x^4 + c2*x^3 + c3*x^2 + c4*x + c5, we want to compute
187          // c'(x) = (c0*x^5 + c1*x^4 + c2*x^3 + c3*x^2 + c4*x + c5) * x + v_i mod g(x)
188          //       = c0*x^6 + c1*x^5 + c2*x^4 + c3*x^3 + c4*x^2 + c5*x + v_i mod g(x)
189          //       = c0*(x^6 mod g(x)) + c1*x^5 + c2*x^4 + c3*x^3 + c4*x^2 + c5*x + v_i
190          // If we call (x^6 mod g(x)) = k(x), this can be written as
191          // c'(x) = (c1*x^5 + c2*x^4 + c3*x^3 + c4*x^2 + c5*x + v_i) + c0*k(x)
192  
193          // First, determine the value of c0:
194          uint8_t c0 = c >> 25;
195  
196          // Then compute c1*x^5 + c2*x^4 + c3*x^3 + c4*x^2 + c5*x + v_i:
197          c = ((c & 0x1ffffff) << 5) ^ v_i;
198  
199          // Finally, for each set bit n in c0, conditionally add {2^n}k(x). These constants can be
200          // computed using the following Sage code (continuing the code above):
201          //
202          // for i in [1,2,4,8,16]: # Print out {1,2,4,8,16}*(g(x) mod x^6), packed in hex integers.
203          //     v = 0
204          //     for coef in reversed((F.fetch_int(i)*(G % x**6)).coefficients(sparse=True)):
205          //         v = v*32 + coef.integer_representation()
206          //     print("0x%x" % v)
207          //
208          if (c0 & 1)  c ^= 0x3b6a57b2; //     k(x) = {29}x^5 + {22}x^4 + {20}x^3 + {21}x^2 + {29}x + {18}
209          if (c0 & 2)  c ^= 0x26508e6d; //  {2}k(x) = {19}x^5 +  {5}x^4 +     x^3 +  {3}x^2 + {19}x + {13}
210          if (c0 & 4)  c ^= 0x1ea119fa; //  {4}k(x) = {15}x^5 + {10}x^4 +  {2}x^3 +  {6}x^2 + {15}x + {26}
211          if (c0 & 8)  c ^= 0x3d4233dd; //  {8}k(x) = {30}x^5 + {20}x^4 +  {4}x^3 + {12}x^2 + {30}x + {29}
212          if (c0 & 16) c ^= 0x2a1462b3; // {16}k(x) = {21}x^5 +     x^4 +  {8}x^3 + {24}x^2 + {21}x + {19}
213  
214      }
215      return c;
216  }
217  
218  /** Syndrome computes the values s_j = R(e^j) for j in [997, 998, 999]. As described above, the
219   * generator polynomial G is the LCM of the minimal polynomials of (e)^997, (e)^998, and (e)^999.
220   *
221   * Consider a codeword with errors, of the form R(x) = C(x) + E(x). The residue is the bit-packed
222   * result of computing R(x) mod G(X), where G is the generator of the code. Because C(x) is a valid
223   * codeword, it is a multiple of G(X), so the residue is in fact just E(x) mod G(x). Note that all
224   * of the (e)^j are roots of G(x) by definition, so R((e)^j) = E((e)^j).
225   *
226   * Let R(x) = r1*x^5 + r2*x^4 + r3*x^3 + r4*x^2 + r5*x + r6
227   *
228   * To compute R((e)^j), we are really computing:
229   * r1*(e)^(j*5) + r2*(e)^(j*4) + r3*(e)^(j*3) + r4*(e)^(j*2) + r5*(e)^j + r6
230   *
231   * Now note that all of the (e)^(j*i) for i in [5..0] are constants and can be precomputed.
232   * But even more than that, we can consider each coefficient as a bit-string.
233   * For example, take r5 = (b_5, b_4, b_3, b_2, b_1) written out as 5 bits. Then:
234   * r5*(e)^j = b_1*(e)^j + b_2*(2*(e)^j) + b_3*(4*(e)^j) + b_4*(8*(e)^j) + b_5*(16*(e)^j)
235   * where all the (2^i*(e)^j) are constants and can be precomputed.
236   *
237   * Then we just add each of these corresponding constants to our final value based on the
238   * bit values b_i. This is exactly what is done in the Syndrome function below.
239   */
240  constexpr std::array<uint32_t, 25> GenerateSyndromeConstants() {
241      std::array<uint32_t, 25> SYNDROME_CONSTS{};
242      for (int k = 1; k < 6; ++k) {
243          for (int shift = 0; shift < 5; ++shift) {
244              int16_t b = GF1024_LOG.at(size_t{1} << shift);
245              int16_t c0 = GF1024_EXP.at((997*k + b) % 1023);
246              int16_t c1 = GF1024_EXP.at((998*k + b) % 1023);
247              int16_t c2 = GF1024_EXP.at((999*k + b) % 1023);
248              uint32_t c = c2 << 20 | c1 << 10 | c0;
249              int ind = 5*(k-1) + shift;
250              SYNDROME_CONSTS[ind] = c;
251          }
252      }
253      return SYNDROME_CONSTS;
254  }
255  constexpr std::array<uint32_t, 25> SYNDROME_CONSTS = GenerateSyndromeConstants();
256  
257  /**
258   * Syndrome returns the three values s_997, s_998, and s_999 described above,
259   * packed into a 30-bit integer, where each group of 10 bits encodes one value.
260   */
261  uint32_t Syndrome(const uint32_t residue) {
262      // low is the first 5 bits, corresponding to the r6 in the residue
263      // (the constant term of the polynomial).
264      uint32_t low = residue & 0x1f;
265  
266      // We begin by setting s_j = low = r6 for all three values of j, because these are unconditional.
267      uint32_t result = low ^ (low << 10) ^ (low << 20);
268  
269      // Then for each following bit, we add the corresponding precomputed constant if the bit is 1.
270      // For example, 0x31edd3c4 is 1100011110 1101110100 1111000100 when unpacked in groups of 10
271      // bits, corresponding exactly to a^999 || a^998 || a^997 (matching the corresponding values in
272      // GF1024_EXP above). In this way, we compute all three values of s_j for j in (997, 998, 999)
273      // simultaneously. Recall that XOR corresponds to addition in a characteristic 2 field.
274      for (int i = 0; i < 25; ++i) {
275          result ^= ((residue >> (5+i)) & 1 ? SYNDROME_CONSTS.at(i) : 0);
276      }
277      return result;
278  }
279  
280  /** Convert to lower case. */
281  inline unsigned char LowerCase(unsigned char c)
282  {
283      return (c >= 'A' && c <= 'Z') ? (c - 'A') + 'a' : c;
284  }
285  
286  /** Return indices of invalid characters in a Bech32 string. */
287  bool CheckCharacters(const std::string& str, std::vector<int>& errors)
288  {
289      bool lower = false, upper = false;
290      for (size_t i = 0; i < str.size(); ++i) {
291          unsigned char c{(unsigned char)(str[i])};
292          if (c >= 'a' && c <= 'z') {
293              if (upper) {
294                  errors.push_back(i);
295              } else {
296                  lower = true;
297              }
298          } else if (c >= 'A' && c <= 'Z') {
299              if (lower) {
300                  errors.push_back(i);
301              } else {
302                  upper = true;
303              }
304          } else if (c < 33 || c > 126) {
305              errors.push_back(i);
306          }
307      }
308      return errors.empty();
309  }
310  
311  /** Expand a HRP for use in checksum computation. */
312  data ExpandHRP(const std::string& hrp)
313  {
314      data ret;
315      ret.reserve(hrp.size() + 90);
316      ret.resize(hrp.size() * 2 + 1);
317      for (size_t i = 0; i < hrp.size(); ++i) {
318          unsigned char c = hrp[i];
319          ret[i] = c >> 5;
320          ret[i + hrp.size() + 1] = c & 0x1f;
321      }
322      ret[hrp.size()] = 0;
323      return ret;
324  }
325  
326  /** Verify a checksum. */
327  Encoding VerifyChecksum(const std::string& hrp, const data& values)
328  {
329      // PolyMod computes what value to xor into the final values to make the checksum 0. However,
330      // if we required that the checksum was 0, it would be the case that appending a 0 to a valid
331      // list of values would result in a new valid list. For that reason, Bech32 requires the
332      // resulting checksum to be 1 instead. In Bech32m, this constant was amended. See
333      // https://gist.github.com/sipa/14c248c288c3880a3b191f978a34508e for details.
334      const uint32_t check = PolyMod(Cat(ExpandHRP(hrp), values));
335      if (check == EncodingConstant(Encoding::BECH32)) return Encoding::BECH32;
336      if (check == EncodingConstant(Encoding::BECH32M)) return Encoding::BECH32M;
337      return Encoding::INVALID;
338  }
339  
340  /** Create a checksum. */
341  data CreateChecksum(Encoding encoding, const std::string& hrp, const data& values)
342  {
343      data enc = Cat(ExpandHRP(hrp), values);
344      enc.resize(enc.size() + 6); // Append 6 zeroes
345      uint32_t mod = PolyMod(enc) ^ EncodingConstant(encoding); // Determine what to XOR into those 6 zeroes.
346      data ret(6);
347      for (size_t i = 0; i < 6; ++i) {
348          // Convert the 5-bit groups in mod to checksum values.
349          ret[i] = (mod >> (5 * (5 - i))) & 31;
350      }
351      return ret;
352  }
353  
354  } // namespace
355  
356  /** Encode a Bech32 or Bech32m string. */
357  std::string Encode(Encoding encoding, const std::string& hrp, const data& values) {
358      // First ensure that the HRP is all lowercase. BIP-173 and BIP350 require an encoder
359      // to return a lowercase Bech32/Bech32m string, but if given an uppercase HRP, the
360      // result will always be invalid.
361      for (const char& c : hrp) assert(c < 'A' || c > 'Z');
362      data checksum = CreateChecksum(encoding, hrp, values);
363      data combined = Cat(values, checksum);
364      std::string ret = hrp + '1';
365      ret.reserve(ret.size() + combined.size());
366      for (const auto c : combined) {
367          ret += CHARSET[c];
368      }
369      return ret;
370  }
371  
372  /** Decode a Bech32 or Bech32m string. */
373  DecodeResult Decode(const std::string& str) {
374      std::vector<int> errors;
375      if (!CheckCharacters(str, errors)) return {};
376      size_t pos = str.rfind('1');
377      if (str.size() > 90 || pos == str.npos || pos == 0 || pos + 7 > str.size()) {
378          return {};
379      }
380      data values(str.size() - 1 - pos);
381      for (size_t i = 0; i < str.size() - 1 - pos; ++i) {
382          unsigned char c = str[i + pos + 1];
383          int8_t rev = CHARSET_REV[c];
384  
385          if (rev == -1) {
386              return {};
387          }
388          values[i] = rev;
389      }
390      std::string hrp;
391      for (size_t i = 0; i < pos; ++i) {
392          hrp += LowerCase(str[i]);
393      }
394      Encoding result = VerifyChecksum(hrp, values);
395      if (result == Encoding::INVALID) return {};
396      return {result, std::move(hrp), data(values.begin(), values.end() - 6)};
397  }
398  
399  /** Find index of an incorrect character in a Bech32 string. */
400  std::pair<std::string, std::vector<int>> LocateErrors(const std::string& str) {
401      std::vector<int> error_locations{};
402  
403      if (str.size() > 90) {
404          error_locations.resize(str.size() - 90);
405          std::iota(error_locations.begin(), error_locations.end(), 90);
406          return std::make_pair("Bech32 string too long", std::move(error_locations));
407      }
408  
409      if (!CheckCharacters(str, error_locations)){
410          return std::make_pair("Invalid character or mixed case", std::move(error_locations));
411      }
412  
413      size_t pos = str.rfind('1');
414      if (pos == str.npos) {
415          return std::make_pair("Missing separator", std::vector<int>{});
416      }
417      if (pos == 0 || pos + 7 > str.size()) {
418          error_locations.push_back(pos);
419          return std::make_pair("Invalid separator position", std::move(error_locations));
420      }
421  
422      std::string hrp;
423      for (size_t i = 0; i < pos; ++i) {
424          hrp += LowerCase(str[i]);
425      }
426  
427      size_t length = str.size() - 1 - pos; // length of data part
428      data values(length);
429      for (size_t i = pos + 1; i < str.size(); ++i) {
430          unsigned char c = str[i];
431          int8_t rev = CHARSET_REV[c];
432          if (rev == -1) {
433              error_locations.push_back(i);
434              return std::make_pair("Invalid Base 32 character", std::move(error_locations));
435          }
436          values[i - pos - 1] = rev;
437      }
438  
439      // We attempt error detection with both bech32 and bech32m, and choose the one with the fewest errors
440      // We can't simply use the segwit version, because that may be one of the errors
441      std::optional<Encoding> error_encoding;
442      for (Encoding encoding : {Encoding::BECH32, Encoding::BECH32M}) {
443          std::vector<int> possible_errors;
444          // Recall that (ExpandHRP(hrp) ++ values) is interpreted as a list of coefficients of a polynomial
445          // over GF(32). PolyMod computes the "remainder" of this polynomial modulo the generator G(x).
446          uint32_t residue = PolyMod(Cat(ExpandHRP(hrp), values)) ^ EncodingConstant(encoding);
447  
448          // All valid codewords should be multiples of G(x), so this remainder (after XORing with the encoding
449          // constant) should be 0 - hence 0 indicates there are no errors present.
450          if (residue != 0) {
451              // If errors are present, our polynomial must be of the form C(x) + E(x) where C is the valid
452              // codeword (a multiple of G(x)), and E encodes the errors.
453              uint32_t syn = Syndrome(residue);
454  
455              // Unpack the three 10-bit syndrome values
456              int s0 = syn & 0x3FF;
457              int s1 = (syn >> 10) & 0x3FF;
458              int s2 = syn >> 20;
459  
460              // Get the discrete logs of these values in GF1024 for more efficient computation
461              int l_s0 = GF1024_LOG.at(s0);
462              int l_s1 = GF1024_LOG.at(s1);
463              int l_s2 = GF1024_LOG.at(s2);
464  
465              // First, suppose there is only a single error. Then E(x) = e1*x^p1 for some position p1
466              // Then s0 = E((e)^997) = e1*(e)^(997*p1) and s1 = E((e)^998) = e1*(e)^(998*p1)
467              // Therefore s1/s0 = (e)^p1, and by the same logic, s2/s1 = (e)^p1 too.
468              // Hence, s1^2 == s0*s2, which is exactly the condition we check first:
469              if (l_s0 != -1 && l_s1 != -1 && l_s2 != -1 && (2 * l_s1 - l_s2 - l_s0 + 2046) % 1023 == 0) {
470                  // Compute the error position p1 as l_s1 - l_s0 = p1 (mod 1023)
471                  size_t p1 = (l_s1 - l_s0 + 1023) % 1023; // the +1023 ensures it is positive
472                  // Now because s0 = e1*(e)^(997*p1), we get e1 = s0/((e)^(997*p1)). Remember that (e)^1023 = 1,
473                  // so 1/((e)^997) = (e)^(1023-997).
474                  int l_e1 = l_s0 + (1023 - 997) * p1;
475                  // Finally, some sanity checks on the result:
476                  // - The error position should be within the length of the data
477                  // - e1 should be in GF(32), which implies that e1 = (e)^(33k) for some k (the 31 non-zero elements
478                  // of GF(32) form an index 33 subgroup of the 1023 non-zero elements of GF(1024)).
479                  if (p1 < length && !(l_e1 % 33)) {
480                      // Polynomials run from highest power to lowest, so the index p1 is from the right.
481                      // We don't return e1 because it is dangerous to suggest corrections to the user,
482                      // the user should check the address themselves.
483                      possible_errors.push_back(str.size() - p1 - 1);
484                  }
485              // Otherwise, suppose there are two errors. Then E(x) = e1*x^p1 + e2*x^p2.
486              } else {
487                  // For all possible first error positions p1
488                  for (size_t p1 = 0; p1 < length; ++p1) {
489                      // We have guessed p1, and want to solve for p2. Recall that E(x) = e1*x^p1 + e2*x^p2, so
490                      // s0 = E((e)^997) = e1*(e)^(997^p1) + e2*(e)^(997*p2), and similar for s1 and s2.
491                      //
492                      // Consider s2 + s1*(e)^p1
493                      //          = 2e1*(e)^(999^p1) + e2*(e)^(999*p2) + e2*(e)^(998*p2)*(e)^p1
494                      //          = e2*(e)^(999*p2) + e2*(e)^(998*p2)*(e)^p1
495                      //    (Because we are working in characteristic 2.)
496                      //          = e2*(e)^(998*p2) ((e)^p2 + (e)^p1)
497                      //
498                      int s2_s1p1 = s2 ^ (s1 == 0 ? 0 : GF1024_EXP.at((l_s1 + p1) % 1023));
499                      if (s2_s1p1 == 0) continue;
500                      int l_s2_s1p1 = GF1024_LOG.at(s2_s1p1);
501  
502                      // Similarly, s1 + s0*(e)^p1
503                      //          = e2*(e)^(997*p2) ((e)^p2 + (e)^p1)
504                      int s1_s0p1 = s1 ^ (s0 == 0 ? 0 : GF1024_EXP.at((l_s0 + p1) % 1023));
505                      if (s1_s0p1 == 0) continue;
506                      int l_s1_s0p1 = GF1024_LOG.at(s1_s0p1);
507  
508                      // So, putting these together, we can compute the second error position as
509                      // (e)^p2 = (s2 + s1^p1)/(s1 + s0^p1)
510                      // p2 = log((e)^p2)
511                      size_t p2 = (l_s2_s1p1 - l_s1_s0p1 + 1023) % 1023;
512  
513                      // Sanity checks that p2 is a valid position and not the same as p1
514                      if (p2 >= length || p1 == p2) continue;
515  
516                      // Now we want to compute the error values e1 and e2.
517                      // Similar to above, we compute s1 + s0*(e)^p2
518                      //          = e1*(e)^(997*p1) ((e)^p1 + (e)^p2)
519                      int s1_s0p2 = s1 ^ (s0 == 0 ? 0 : GF1024_EXP.at((l_s0 + p2) % 1023));
520                      if (s1_s0p2 == 0) continue;
521                      int l_s1_s0p2 = GF1024_LOG.at(s1_s0p2);
522  
523                      // And compute (the log of) 1/((e)^p1 + (e)^p2))
524                      int inv_p1_p2 = 1023 - GF1024_LOG.at(GF1024_EXP.at(p1) ^ GF1024_EXP.at(p2));
525  
526                      // Then (s1 + s0*(e)^p1) * (1/((e)^p1 + (e)^p2)))
527                      //         = e2*(e)^(997*p2)
528                      // Then recover e2 by dividing by (e)^(997*p2)
529                      int l_e2 = l_s1_s0p1 + inv_p1_p2 + (1023 - 997) * p2;
530                      // Check that e2 is in GF(32)
531                      if (l_e2 % 33) continue;
532  
533                      // In the same way, (s1 + s0*(e)^p2) * (1/((e)^p1 + (e)^p2)))
534                      //         = e1*(e)^(997*p1)
535                      // So recover e1 by dividing by (e)^(997*p1)
536                      int l_e1 = l_s1_s0p2 + inv_p1_p2 + (1023 - 997) * p1;
537                      // Check that e1 is in GF(32)
538                      if (l_e1 % 33) continue;
539  
540                      // Again, we do not return e1 or e2 for safety.
541                      // Order the error positions from the left of the string and return them
542                      if (p1 > p2) {
543                          possible_errors.push_back(str.size() - p1 - 1);
544                          possible_errors.push_back(str.size() - p2 - 1);
545                      } else {
546                          possible_errors.push_back(str.size() - p2 - 1);
547                          possible_errors.push_back(str.size() - p1 - 1);
548                      }
549                      break;
550                  }
551              }
552          } else {
553              // No errors
554              return std::make_pair("", std::vector<int>{});
555          }
556  
557          if (error_locations.empty() || (!possible_errors.empty() && possible_errors.size() < error_locations.size())) {
558              error_locations = std::move(possible_errors);
559              if (!error_locations.empty()) error_encoding = encoding;
560          }
561      }
562      std::string error_message = error_encoding == Encoding::BECH32M ? "Invalid Bech32m checksum"
563                                : error_encoding == Encoding::BECH32 ? "Invalid Bech32 checksum"
564                                : "Invalid checksum";
565  
566      return std::make_pair(error_message, std::move(error_locations));
567  }
568  
569  } // namespace bech32