/*
 * avrdude - A Downloader/Uploader for AVR device programmers
 * Copyright (C) 2023 Hans Eirik Bull
 * Copyright (C) 2023 Stefan Rueger <stefan.rueger@urclocks.com>
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program. If not, see <http://www.gnu.org/licenses/>.
 */

#include <ac_cfg.h>

#include <string.h>
#include <errno.h>
#include <stdio.h>
#include <stdarg.h>
#include <stdlib.h>
#include <ctype.h>
#include <math.h>

#include "avrdude.h"
#include "libavrdude.h"

// Return 1 if str starts with starts, 0 otherwise
int str_starts(const char *str, const char *starts) {
  return strncmp(str, starts, strlen(starts)) == 0;
}

// Return 1 if str1 and str2 are the same, 0 otherwise
int str_eq(const char *str1, const char *str2) {
  return strcmp(str1, str2) == 0;
}

// Return 1 if str contains substr as substring, 0 otherwise
int str_contains(const char *str, const char *substr) {
  return !!strstr(str, substr);
}

// Return 1 if str ends in ends, 0 otherwise
int str_ends(const char *str, const char *ends) {
  size_t str_len = strlen(str);
  size_t ends_len = strlen(ends);

  if(ends_len > str_len)
    return 0;

  return str_eq(str + str_len - ends_len, ends);
}

// Return 1 if str starts with starts irrespective of case, 0 otherwise
int str_casestarts(const char *str, const char *starts) {
  return strncasecmp(str, starts, strlen(starts)) == 0;
}

// Return 1 if str1 and str2 are the same irrespective of case, 0 otherwise
int str_caseeq(const char *str1, const char *str2) {
  return strcasecmp(str1, str2) == 0;
}

// Return 1 if str ends in ends irrespective of case, 0 otherwise
int str_caseends(const char *str, const char *ends) {
  size_t str_len = strlen(str);
  size_t ends_len = strlen(ends);

  if(ends_len > str_len)
    return 0;

  return str_caseeq(str + str_len - ends_len, ends);
}

/*
 * Match string against the partname pattern, returning 1 if it matches, 0 if
 * not. Note: str_match_core() is a modified old copy of !fnmatch() from the
 * GNU C Library (published under GLP v2), which uses shell wildcards for
 * constructing patterns, ie, *, ? and single character classes, eg, [^0-6].
 * Used for portability.
 */

inline static int fold(int c) {
  return (c >= 'A' && c <= 'Z')? c + ('a' - 'A'): c;
}

inline static int nofold(int c) {
  return c;
}

static int str_match_core(const char *pattern, const char *string, int (*fold)(int c)) {
  unsigned char c;
  const char *p = pattern, *n = string;

  if(!*n)                       // AVRDUDE specialty: empty string never matches
    return 0;

  while((c = fold(*p++))) {
    switch(c) {
    case '?':
      if(*n == 0)
        return 0;
      break;

    case '\\':
      c = fold(*p++);
      if(fold(*n) != c)
        return 0;
      break;

    case '*':
      for(c = *p++; c == '?' || c == '*'; c = *p++)
        if(c == '?' && *n++ == 0)
          return 0;

      if(c == 0)
        return 1;

      {
        unsigned char c1 = fold(c == '\\'? *p: c);    // This char

        for(--p; *n; ++n)       // Recursively check reminder of string for *
          if((c == '[' || fold(*n) == c1) && str_match_core(p, n, fold) == 1)
            return 1;
        return 0;
      }

    case '[':
      {
        int negate;

        if(*n == 0)
          return 0;

        negate = (*p == '!' || *p == '^');
        if(negate)
          ++p;

        c = *p++;
        for(;;) {
          unsigned char cstart = c, cend = c;

          if(c == '\\')
            cstart = cend = *p++;

          cstart = cend = fold(cstart);

          if(c == 0)            // [ (unterminated)
            return 0;

          c = *p++;
          c = fold(c);

          if(c == '-' && *p != ']') {
            cend = *p++;
            if(cend == '\\')
              cend = *p++;
            if(cend == 0)
              return 0;
            cend = fold(cend);

            c = *p++;
          }

          if(fold(*n) >= cstart && fold(*n) <= cend)
            goto matched;

          if(c == ']')
            break;
        }
        if(!negate)
          return 0;
        break;

      matched:;
        while(c != ']') {       // Skip the rest of the [...] that already matched

          if(c == 0)            // [... (unterminated)
            return 0;

          c = *p++;
          if(c == '\\')
            ++p;
        }
        if(negate)
          return 0;
      }
      break;

    default:
      if(c != fold(*n))
        return 0;
    }

    ++n;
  }

  return *n == 0;
}

int str_match(const char *pattern, const char *string) {
  return str_match_core(pattern, string, nofold);
}

int str_casematch(const char *pattern, const char *string) {
  return str_match_core(pattern, string, fold);
}

int str_matched_by(const char *string, const char *pattern) {
  return str_match_core(pattern, string, nofold);
}

int str_casematched_by(const char *string, const char *pattern) {
  return str_match_core(pattern, string, fold);
}

// Does the string contain wildcard pattern matching elements?
int str_is_pattern(const char *str) {
  for(;;)
    switch(*str++) {
    case 0:
      return 0;
    case '*':
    case '?':
    case '[':
    case '\\':
      return 1;
    }
}

// Is the string s in the list l of strings as matched by f(s, l[i])?
int str_is_in_list(const char *s, const char **l, size_t nl, int (*f)(const char *, const char *)) {
  for(size_t i = 0; i < nl; i++)
    if(f(s, l[i]))
      return 1;
  return 0;
}

// Is the string a decimal bus/device number that ends in a colon or nul?
static int is_busdev_num(const char *s) {
  const char *t = s;

  while(*t >= '0' && *t <= '9')
    t++;

  return t > s && (*t == ':' || *t == 0);
}

// Do the two strings represent the same decimal number or are they the same up to a limiting colon?
int str_busdev_eq(const char *s, const char *t) {
  if(is_busdev_num(s) && is_busdev_num(t))
    if(strtoull(s, NULL, 10) == strtoull(t, NULL, 10))
      return 1;

  int len = 0;                  // Length of strings up to optional colon
  while(s[len] && s[len] != ':' && t[len] && t[len] != ':')
    len++;

  // Different length means strings are not the same
  if((s[len] && s[len] != ':') || (t[len] && t[len] != ':'))
    return 0;

  return !strncmp(s, t, len);
}

// Return a mmt_malloc'd string with the sprintf() result
char *str_sprintf(const char *fmt, ...) {
  int size = 0;
  va_list ap;

  // Compute size
  va_start(ap, fmt);
  size = vsnprintf(NULL, 0, fmt, ap);
  va_end(ap);

  if(size < 0)
    return mmt_strdup("");

  size++;                       // For terminating '\0'
  char *p = mmt_malloc(size);

  va_start(ap, fmt);
  size = vsnprintf(p, size, fmt, ap);
  va_end(ap);

  if(size < 0)
    *p = 0;

  return p;
}

// Return a string in closed-circuit space with the sprintf() result
const char *str_ccprintf(const char *fmt, ...) {
  int size = 0, avail = sizeof cx->avr_space - AVR_SAFETY_MARGIN;
  va_list ap;

  // Compute size
  va_start(ap, fmt);
  size = vsnprintf(NULL, 0, fmt, ap);
  va_end(ap);

  if(size < 0)
    return "";

  size++;                       // For terminating '\0'
  if(size > avail)
    size = avail;
  char *p = avr_cc_buffer(size);

  va_start(ap, fmt);
  size = vsnprintf(p, size, fmt, ap);
  va_end(ap);

  if(size < 0)
    *p = 0;

  return p;
}

// Returns a temporary, possibly abbreviated copy of str in closed-circuit space
const char *str_ccstrdup(const char *str) {
  size_t size = strlen(str) + 1, avail = sizeof cx->avr_space - AVR_SAFETY_MARGIN;

  if(size > avail)
    size = avail;
  char *ret = avr_cc_buffer(size);

  strncpy(ret, str, size);
  ret[size - 1] = 0;

  return ret;
}

// Reads a potentially long line and returns it in a mmt_malloc'd buffer
char *str_fgets(FILE *fp, const char **errpp) {
  int bs = 1023;                // Must be 2^n - 1
  char *ret = (char *) mmt_malloc(bs);

  ret[bs - 2] = 0;
  if(!fgets(ret, bs, fp)) {
    mmt_free(ret);
    if(errpp)
      *errpp = ferror(fp) && !feof(fp)? "I/O error": NULL;
    return NULL;
  }

  while(ret[bs - 2] != 0 && ret[bs - 2] != '\n' && ret[bs - 2] != '\r') {
    if(bs >= INT_MAX/2) {
      mmt_free(ret);
      if(errpp)
        *errpp = "cannot cope with lines longer than INT_MAX/2 bytes";
      return NULL;
    }
    int was = bs;

    bs = 2*bs + 1;
    ret = mmt_realloc(ret, bs);
    ret[was - 1] = ret[bs - 2] = 0;
    if(!fgets(ret + was - 1, bs - (was - 1), fp)) {     // EOF? Error?
      if(ferror(fp)) {
        mmt_free(ret);
        if(errpp)
          *errpp = "I/O error";
        return NULL;
      }
      break;
    }
  }

  if(errpp)
    *errpp = NULL;
  return ret;
}

// Return the number of times a character c occurs in str
size_t str_numc(const char *str, char c) {
  size_t ret = 0;
  char is;

  while((is = *str++))
    if(is == c)
      ret++;
  return ret;
}

// Return a pointer to the first non-white-space character in a string (or the end)
const char *str_ltrim(const char *s) {
  while(*s && isascii(*s & 0xff) && isspace(*s & 0xff))
    s++;
  return s;
}

// Terminate at position n and remove white space before that
char *str_nrtrim(char *s, size_t n) {
  s[n] = 0;
  if(n)
    for(char *z = s + n - 1; z >= s && isascii(*z & 0xff) && isspace(*z & 0xff); z--)
      *z = 0;
  return s;
}

// Remove trailing white space
char *str_rtrim(char *s) {
  return str_nrtrim(s, strlen(s));
}

// Terminate at position n and remove leading and trailing white space
char *str_ntrim(char *s, size_t n) {
  return (char *) str_ltrim(str_nrtrim(s, n));
}

// Remove leading and trailing white space
char *str_trim(char *s) {
  return (char *) str_ltrim(str_nrtrim(s, strlen(s)));
}

// Changes string to be all lowercase and returns original pointer
char *str_lc(char *s) {
  for(char *t = s; *t; t++)
    *t = tolower(*t & 0xff);
  return s;
}

// Changes string to be all uppercase and returns original pointer
char *str_uc(char *s) {
  for(char *t = s; *t; t++)
    *t = toupper(*t & 0xff);
  return s;
}

// Changes first character in a string to be lowercase and returns original pointer
char *str_lcfirst(char *s) {
  *s = tolower(*s & 0xff);
  return s;
}

// Changes first character in a string to be uppercase and returns original pointer
char *str_ucfirst(char *s) {
  *s = toupper(*s & 0xff);
  return s;
}

// Convert to ASCII name leaving only letters, numbers, underscore, period and dash
char *str_asciiname(char *s) {
  for(char *t = s; *t; t++)
    switch(*t) {
    case '?':
      *t = 'Q';
      break;
    case '*':
      *t = 'X';
      break;
    case '|':
      *t = 'I';
      break;
    case '{':
      *t = 'l';
      break;
    case '}':
      *t = 'j';
      break;
    case '[':
      *t = 'L';
      break;
    case ']':
      *t = 'J';
      break;
    case '(':
      *t = 'L';
      break;
    case ')':
      *t = 'J';
      break;
    case '<':
      *t = 'l';
      break;
    case '>':
      *t = 'j';
      break;
    case '&':
      *t = '+';
      break;
    case '!':
      *t = 'I';
      break;
    case '"':
      *t = 'q';
      break;
    case '\'':
      *t = 'q';
      break;
    case '`':
      *t = 'q';
      break;
    case '.':
    case '-':
      break;
    default:
      if(!isascii(*t & 0xff) || !isalnum(*t & 0xff))
        *t = '_';
    }

  return s;
}

// Convert unsigned to ASCII string; caller needs to allocate enough space for buf
char *str_utoa(unsigned n, char *buf, int base) {
  unsigned q;
  char *cp;

  if(base == 'r') {
    const char *units = "IVXLCDMFTYHSNabcdefghijkl";
    const char *rep[10] = { "", "a", "aa", "aaa", "ab", "b", "ba", "baa", "baaa", "ac" };

    if(n == 0) {
      strcpy(buf, "0");
      return buf;
    }

    int i = 0;

    for(unsigned u = n; u; u /= 10)
      i++;
    for(*buf = 0; i > 0; i--) {
      unsigned u = n;

      for(int j = 1; j < i; j++)
        u /= 10;
      char *q = buf + strlen(buf);

      for(const char *p = rep[u%10], *d = units + (i - 1)*2; *p; p++)
        *q++ = d[*p - 'a'];
      *q = 0;
    }
    return buf;
  }

  if(base < 2 || base > 36) {
    *buf = 0;
    return buf;
  }

  cp = buf;
  // Divide by base until the number disappeared, but ensure at least one digit will be emitted
  do {
    q = n%base;
    n /= base;
    *cp++ = q < 10? q + '0': q + 'a' - 10;
  } while(n);

  // Terminate the string
  *cp-- = 0;

  // Revert the result string
  for(char *cp2 = buf; cp > cp2;) {
    char c = *cp;

    *cp-- = *cp2;
    *cp2++ = c;
  }

  return buf;
}

// Returns a pointer to the start of a trailing number in the string or NULL if not there
char *str_endnumber(const char *str) {
  const char *ret = NULL;

  if(!str)
    return NULL;

  for(const char *end = str + strlen(str) - 1; end >= str; end--)
    if(isdigit((unsigned char) *end))
      ret = end;
    else
      break;

  return (char *) ret;
}

// Convenience functions for printing
const char *str_plural(int x) {
  return x == 1? "": "s";
}

static const char *str_filename(const char *fn, const char *stdname) {
  if(!fn)
    fn = "???";
  char *p1 = strrchr(fn, '/'), *p2 = strrchr(fn, '\\');

  return str_eq(fn, "-")? stdname: str_starts(fn, "/dev/")? fn: p1? p1 + 1: p2? p2 + 1: fn;
}

// Path name fn or <stdin> if fn is -
const char *str_inname(const char *fn) {
  return !fn? "???": str_eq(fn, "-")? "<stdin>": fn;
}

// File name of fn or <stdin> if fn is -
const char *str_infilename(const char *fn) {
  return str_filename(fn, "<stdin>");
}

// Path name fn or <stdout> if fn is -
const char *str_outname(const char *fn) {
  return !fn? "???": str_eq(fn, "-")? "<stdout>": fn;
}

// File name of fn or <stdout> if fn is -
const char *str_outfilename(const char *fn) {
  return str_filename(fn, "<stdout>");
}

// Return sth like "[0, 0x1ff]" in closed-circuit space
const char *str_ccinterval(int a, int b) {
  char *ret = avr_cc_buffer(45);        // Interval strings each max 45 bytes at 64-bit int

  sprintf(ret, a < 16? "[%d": "[0x%x", a);
  sprintf(ret + strlen(ret), b < 16? ", %d]": ", 0x%x]", b);

  return ret;
}

bool is_bigendian() {
  union {
    char a[2];
    int16_t i;
  } u = {.i = 1 };
  return u.a[1] == 1;
}

// Change data item p of size bytes from big endian to little endian and vice versa
void change_endian(void *p, int size) {
  uint8_t tmp, *w = p;

  for(int i = 0; i < size/2; i++)
    tmp = w[i], w[i] = w[size - i - 1], w[size - i - 1] = tmp;
}

// Looks like a double mantissa in hex or dec notation?
static int is_mantissa_only(char *p) {
  char *digs;

  if(*p == '+' || *p == '-')
    p++;

  if(*p == '0' && (p[1] == 'x' || p[1] == 'X')) {
    p += 2;
    digs = "0123456789abcdefABCDEF";
  } else
    digs = "0123456789";

  if(!*p)
    return 0;

  while(*p)
    if(!strchr(digs, *p++))
      return 0;

  return 1;
}

// Return 1 if all n bytes in memory pointed to by p are c, 0 otherwise
int is_memset(const void *p, char c, size_t n) {
  const char *q = (const char *) p;

  return n <= 0 || (*q == c && memcmp(q, q + 1, n - 1) == 0);
}

// https://en.wikipedia.org/wiki/Easter_egg_(media)#Software
unsigned long long int easteregg(const char *str, const char **endpp) {
  unsigned long long int ret = 0;
  struct {
    char chr[3];
    unsigned lim, nxt, val;
  } eet[] = {
    {"M", 3, 1, 1000},
    {"CM", 1, 4, 900},
    {"D", 1, 2, 500},
    {"CD", 1, 2, 400},
    {"C", 3, 1, 100},
    {"XC", 1, 4, 90},
    {"L", 1, 2, 50},
    {"XL", 1, 2, 40},
    {"X", 3, 1, 10},
    {"IX", 1, 4, 9},
    {"V", 1, 2, 5},
    {"IV", 1, 2, 4},
    {"I", 3, 1, 1},
  }, *dig;

  for(size_t i = 0; i < sizeof eet/sizeof *eet;) {
    dig = eet + i;
    unsigned lim = dig->lim;
    size_t ni = i + 1;

    for(unsigned j = 0; j < lim; j++) {
      if(!str_starts(str, dig->chr))
        break;
      ret += dig->val;
      if(ret < dig->val) {
        if(endpp)
          *endpp = str;
        return 0;
      }
      str += strlen(dig->chr);
      ni = i + dig->nxt;
    }
    if(!*str)
      break;
    i = ni;
  }
  if(endpp)
    *endpp = str;

  return ret;
}

// Like strtoull but knows binary, too
unsigned long long int str_ull(const char *str, char **endptr, int base) {
  const char *nptr = str, *ep;
  unsigned long long int ret = 0;
  int neg = 0;

  while(isspace(*nptr & 0xff))
    nptr++;

  // Check explicit sign for benefit of 0b...
  if(*nptr == '-' || *nptr == '+') {
    if(*nptr == '-')
      neg = 1;
    nptr++;
    // Don't allow double signs
    if(*nptr == '-' || *nptr == '+') {
      if(endptr)
        *endptr = (char *) nptr;
      return ret;
    }
  }

  if((base == 0 || base == 2) && *nptr == '0' && (nptr[1] == 'b' || nptr[1] == 'B'))
    base = 2, nptr += 2;
  else if((base == 0 || base == 16) && *nptr == '0' && (nptr[1] == 'x' || nptr[1] == 'X'))
    base = 16, nptr += 2;

  errno = 0;
  if((base == 0 || base == 'r') && (ret = easteregg(nptr, &ep)) && ep != nptr && !*ep) {
    if(endptr)
      *endptr = (char *) ep;
  } else {
    ret = strtoull(nptr, endptr, base);
    if(endptr && *endptr == nptr)
      *endptr = (char *) str;
  }

  if(neg && errno == 0)
    ret = ~ret + 1;             // Same as -ret but silences overzealous compiler warnings

  return ret;
}

// Returns whether or not the string str looks like a number
int looks_like_number(const char *str) {
  int base = 0;
  char *endptr;

  while(isspace(*str & 0xff))
    str++;

  // Skip sign but don't allow double sign
  if(*str == '-' || *str == '+') {
    str++;
    if(*str == '-' || *str == '+')
      return 0;
  }

  if(*str == '0' && (str[1] == 'b' || str[1] == 'B'))
    base = 2, str += 2;
  else if(*str == '0' && (str[1] == 'x' || str[1] == 'X'))
    base = 16, str += 2;

  errno = 0;
  (void) strtoull(str, &endptr, base);

  return endptr != str && !*endptr && !errno;
}

/*
 * str_todata() is the workhorse for generic string to data conversion for the
 * terminal write function, but is also used for generic string to integer
 * conversions in str_int() below. Both routines define the "character" of how
 * avrdude understands strings in (most) of its dealings. The granularity of
 * type is an bitwise-or combination of bits making up STR_INTEGER; STR_FLOAT;
 * STR_DOUBLE or STR_STRING. The arguments part and memstr are only needed for
 * input from files.
 */

#define Return(...) do { \
  sd->errstr = mmt_sprintf(__VA_ARGS__); \
  sd->type = 0; \
  mmt_free(str); \
  return sd; \
} while(0)

#define Warning(...) do { \
  if(sd->warnstr) \
    mmt_free(sd->warnstr); \
   sd->warnstr = mmt_sprintf(__VA_ARGS__); \
} while(0)

#define sizeforsigned(ll) ( \
  (ll) < INT32_MIN || (ll) > INT32_MAX? 8: \
  (ll) < INT16_MIN || (ll) > INT16_MAX? 4: \
  (ll) < INT8_MIN  || (ll) > INT8_MAX? 2: 1)

Str2data *str_todata(const char *s, int type, const AVRPART *part, const char *memstr) {
  char *end_ptr;
  Str2data *sd = mmt_malloc(sizeof *sd);
  char *str = mmt_strdup(s);

  size_t arglen = strlen(str);

  // Remove trailing comma to allow cut and paste of lists
  if(arglen > 0 && str[arglen - 1] == ',')
    str[--arglen] = 0;

  if(arglen == 0)
    Return("no data to convert");

  // Try integers and assign data size
  if(type & STR_INTEGER) {
    bool is_big_endian, is_signed = 0, is_outside_int64 = 0, is_out_of_range = 0;
    char *stri = str;

    while(isspace(*stri & 0xff))
      stri++;

    sd->ull = 1;
    if(sizeof(long long) != sizeof(int64_t) || (sd->a[0] ^ sd->a[7]) != 1)
      Return("assumption on data types not met? Check source and recompile");
    is_big_endian = sd->a[7];

    sd->sigsz = sd->size = 0;
    errno = 0;
    sd->ull = str_ull(stri, &end_ptr, 0);

    if(!(end_ptr == stri || errno)) {
      unsigned int nu = 0, nl = 0, nh = 0, ns = 0, nx = 0;

      // Parse suffixes: ULL, LL, UL, L ... UHH, HH
      for(char *p = end_ptr; *p; p++) {
        switch(toupper(*p & 0xff)) {
        case 'U':
          nu++;
          break;
        case 'L':
          nl++;
          break;
        case 'H':
          nh++;
          break;
        case 'S':
          ns++;
          break;
        default:
          nx++;
        }
      }

      if(nx == 0 && nu < 2 && nl < 3 && nh < 3 && ns < 2) {     // Could be valid integer suffix
        // If U, then must be at start or end
        if(nu == 0 || toupper(*end_ptr & 0xff) == 'U' || toupper(str[arglen - 1] & 0xff) == 'U') {
          bool is_hex, is_bin;
          int ndigits;

          is_hex = str_casestarts(stri, "0x");  // Ordinary hex without explicit +/- sign
          is_bin = str_casestarts(stri, "0b");  // Ordinary bin without explicit +/- sign
          ndigits = end_ptr - stri - 2; // Used for is_hex and is_bin
          is_signed = !(nu || is_hex || is_bin);        // Neither explicitly unsigned nor 0x/0b

          if(is_signed) {       // Is input in range for int64_t?
            if(*stri == '-' && (sd->ull == ~(~0ULL >> 1) || sd->ll > 0))
              is_outside_int64 = 1;
            if(*stri != '-' && sd->ll < 0)
              is_outside_int64 = 1;
          }
          // Set size
          if(nl == 0 && ns == 0 && nh == 0) {   // No explicit data size
            // Ordinary hex/bin get implicit size by number of digits, including leading zeros
            if(is_hex) {
              sd->size = ndigits > 8? 8: ndigits > 4? 4: ndigits > 2? 2: 1;
            } else if(is_bin) {
              sd->size = ndigits > 32? 8: ndigits > 16? 4: ndigits > 8? 2: 1;
            } else if(is_signed) {
              // Smallest size that fits signed or unsigned (asymmetric to meet user expectation)
              sd->size =
                is_outside_int64? 8:
                sd->ll < INT32_MIN || sd->ll > (long long) UINT32_MAX? 8:
                sd->ll < INT16_MIN || sd->ll > (long long) UINT16_MAX? 4:
                sd->ll < INT8_MIN || sd->ll > (long long) UINT8_MAX? 2: 1;
              if(sd->size < 8)  // sigsz is the one needed for signed int
                sd->sigsz = sizeforsigned(sd->ll);
            } else {
              // Smallest size that fits unsigned representation
              sd->size = sd->ull > UINT32_MAX? 8: sd->ull > UINT16_MAX? 4: sd->ull > UINT8_MAX? 2: 1;
            }
          } else if(nl == 0 && nh == 2 && ns == 0) {    // HH
            sd->size = 1;
            if(is_signed && (sd->ll < INT8_MIN || sd->ll > INT8_MAX))
              is_out_of_range = 1;      // Out of range if uint64 and -uint64 are
            else if(!is_signed && sd->ull > UINT8_MAX && ~sd->ull + 1 > UINT8_MAX)
              is_out_of_range = 1;
            if(is_signed)
              sd->sigsz = sizeforsigned(sd->ll);
          } else if(nl == 0 && ((nh == 1 && ns == 0) || (nh == 0 && ns == 1))) {        // H or S
            sd->size = 2;
            if(is_signed && (sd->ll < INT16_MIN || sd->ll > INT16_MAX))
              is_out_of_range = 1;
            else if(!is_signed && sd->ull > UINT16_MAX && ~sd->ull + 1 > UINT16_MAX)
              is_out_of_range = 1;
            if(is_signed)
              sd->sigsz = sizeforsigned(sd->ll);
          } else if(nl == 1 && nh == 0 && ns == 0) {    // L
            sd->size = 4;
            if(is_signed && (sd->ll < INT32_MIN || sd->ll > INT32_MAX))
              is_out_of_range = 1;
            else if(!is_signed && sd->ull > UINT32_MAX && ~sd->ull + 1 > UINT32_MAX)
              is_out_of_range = 1;
            if(is_signed)
              sd->sigsz = sizeforsigned(sd->ll);
          } else if(nl == 2 && nh == 0 && ns == 0) {    // LL
            sd->size = 8;
          }
        }
      }
    }
    if(sd->size) {
      if(sd->sigsz < sd->size)
        sd->sigsz = sd->size;
      if(sd->sigsz < 8) {       // Curtail and sign extend the number
        if(is_big_endian && sd->size > 1)
          change_endian(sd->a, sd->size);
        memset(sd->a + sd->sigsz, is_signed && (sd->a[sd->sigsz - 1] & 0x80)? 0xff: 0, 8 - sd->sigsz);
        if(is_big_endian)
          change_endian(sd->a, sizeof sd->a);
      }

      if(is_signed && is_out_of_range)
        Warning("%s out of int%d range, interpreted as %d-byte %lld%sU",
          stri, sd->size*8, sd->size, (long long int) sd->ll, sd->size == 4? "L": sd->size == 2? "H": "HH");
      else if(is_out_of_range)
        Warning("%s out of uint%d range, interpreted as %d-byte %llu",
          stri, sd->size*8, sd->size, (long long unsigned int) sd->ull);
      else if(is_outside_int64)
        Warning("%s out of int64 range (consider U suffix)", stri);

      sd->type = STR_INTEGER;
      mmt_free(str);
      return sd;
    }
  }

  if(type & STR_DOUBLE) {       // Try double next, must have D suffix
    sd->d = strtod(str, &end_ptr);
    if(end_ptr != str && toupper(*end_ptr & 0xff) == 'D' && end_ptr[1] == 0) {
      sd->size = 8;
      sd->type = STR_DOUBLE;
      mmt_free(str);
      return sd;
    }
  }

  if(type & STR_FLOAT) {        // Try float next
    sd->size = 0;
    sd->f = strtof(str, &end_ptr);
    if(end_ptr != str && toupper(*end_ptr & 0xff) == 'F' && end_ptr[1] == 0)
      sd->size = 4;
    // Do not accept valid mantissa-only floats that are integer rejects (eg, 078 or ULL overflows)
    if(end_ptr != str && *end_ptr == 0 && !is_mantissa_only(str))
      sd->size = 4;
    if(sd->size) {
      sd->type = STR_FLOAT;
      mmt_free(str);
      return sd;
    }
  }

  if(type & STR_STRING && arglen > 1) { // Try C-style string or single character
    if((*str == '\'' && str[arglen - 1] == '\'') || (*str == '\"' && str[arglen - 1] == '\"')) {
      char *s = mmt_malloc(arglen - 1);

      // Strip start and end quotes, and unescape C string
      strncpy(s, str + 1, arglen - 2);
      cfg_unescape(s, s);
      if(*str == '\'') {        // Single C-style character
        if(*s && s[1])
          Warning("only using first character of %s", str);
        sd->a[0] = *s;
        memset(sd->a + 1, 0, 7);
        sd->sigsz = sd->size = 1;
        sd->type = STR_INTEGER;
        mmt_free(s);
      } else {                  // C-style string
        sd->str_ptr = s;
        sd->type = STR_STRING;
      }
      mmt_free(str);
      return sd;
    }
  }

  if(type & STR_FILE && part && memstr) { // File name containing data to be loaded
    int format = FMT_AUTO;
    FILE *f;
    char fmtstr[4] = { 0 };

    if(arglen > 2 && str[arglen - 2] == ':') {
      fmtstr[0] = ' ';
      strcpy(fmtstr + 1, str + arglen - 2);
      format = fileio_format(str[arglen - 1]);
      if(format == FMT_ERROR)
        Return("unknown format%s suffix of file name", fmtstr);
      str[arglen -= 2] = 0;
    }
    if(format == FMT_AUTO && is_generated_fname(str))
      format = FMT_IHEX;
    if(format == FMT_AUTO) {
      f = fileio_fopenr(str);
      if(f == NULL)
        Return("unable to read the%s file: %s", fmtstr, strerror(errno));
      format = fileio_fmt_autodetect_fp(f);
      fclose(f);

      if(format < 0)
        Return("cannot determine format for the file, specify explicitly");
    }
    // Obtain a copy of the part incl all memories
    AVRPART *dp = avr_dup_part(part);
    AVRMEM *mem = avr_locate_mem(dp, memstr);

    if(!mem) {
      avr_free_part(dp);
      Return("memory %s not configured for device %s", memstr, part->desc);
    }
    int rc = fileio(FIO_READ_FOR_VERIFY, str, format, dp, memstr, -1);

    if(rc < 0) {
      avr_free_part(dp);
      Return("unable to read the%s %s file", fmtstr, fileio_fmtstr(format));
    }
    sd->mem = avr_dup_mem(mem);
    sd->size = rc;
    avr_free_part(dp);
    sd->type = STR_FILE;
    mmt_free(str);
    return sd;
  }

  Return("cannot parse");
}

// Free the data structure returned by str_todata()
void str_freedata(Str2data *sd) {
  if(sd) {
    if(sd->warnstr)
      mmt_free(sd->warnstr);
    if(sd->errstr)
      mmt_free(sd->errstr);
    if(sd->mem)
      avr_free_mem(sd->mem);
    mmt_free(sd);
  }
}

/*
 * Generic string to integer routine that conforms to avrdude terminal syntax
 *
 * Str points to a string that contains an integer terminal data item. Type can
 * be STR_INTEGER or a non-zero bitwise-or combination of integer size
 * designators STR_1, STR_2, STR_4 and STR_8 and sign type STR_SIGNED or,
 * independently, STR_UNSIGNED. A corresponding range check will be done for
 * the numbers that are encoded in the string. If neither or both of
 * STR_UNSIGNED and STR_SIGNED was given then the admitted integer can be in
 * either the signed or the unsigned range of the given size, otherwise only
 * numbers of the requested signedness range will be admitted. Either way, if
 * the string itself restricts the size through a size suffix (see below) then
 * any overflow in there will trigger a range error, too. As in C, a
 * sign-changed unsigned number will yield another unsigned number greater than
 * or equal to zero. As such, the numbers 0, -255U, -254U, ..., -1U are all
 * valid unsigned representations of one-byte integers 0, 1, ... 255. At the
 * same time -256U and 256U are not in the one-byte unsigned range [0, 255].
 * Finally, in the case of  success str_int() will set the character pointer
 * pointed to by errptr to NULL and return an integer in the range of the
 * requested type as unsigned long long. In the case of a conversion error, the
 * pointer pointed to by errptr will be set to a human-readable error message
 * whilst the returned value has no meaning.
 *
 * Integer terminal data items are either a literal C-like character such as
 * '\t' or an integer string with optional leading white space; optional + or -
 * sign; a binary (leading 0b), octal (leading 0), decimal or hexadecimal
 * number (leading 0x); optional size suffix LL/L/S/H/HH; optional unsigned U
 * suffix. All terminal data items can have an optional trailing comma to allow
 * cutting and pasting lists, and will undergo automated data size and base
 * detection. The known integer sizes are either 1 (suffix HH), 2 (suffix H or
 * S), 4 (suffix L) or 8 bytes (suffix LL).
 *
 * Usage example:
 *
 * #include "libavrdude.h"
 *
 * const char *errptr;
 * int addr = str_int(string, STR_INT32, &errptr);
 * if(errptr) {
 *   pmsg_error("(dump) address %s: %s\n", string, errptr);
 *   return -1;
 * }
 */

unsigned long long int str_int(const char *str, int type, const char **errpp) {
  const char *err = NULL;
  Str2data *sd = NULL;
  unsigned long long int ret = 0ULL;

  type &= STR_INTEGER;
  if(type == 0) {
    err = "no integral type requested in str_int()";
    goto finished;
  }

  sd = str_todata(str, type | STR_STRING, NULL, NULL);
  // 1<<lds is number of expected bytes
  int lds = type & STR_8? 3: type & STR_4? 2: type & STR_2? 1: type & STR_1? 0: 3;

  if(sd->type != STR_INTEGER || sd->errstr) {
    err = sd->errstr? cache_string(sd->errstr): "not an integral type";
    goto finished;
  }

  if(sd->warnstr && strstr(sd->warnstr, " out of ")) {  // Convert out of range warning into error
    char *p = strstr(sd->warnstr, "out of ");

    if(p) {
      p = mmt_strdup(p);
      if(strchr(p, ','))
        *strchr(p, ',') = 0;
      err = cache_string(p);
      mmt_free(p);
    } else {
      err = "out of range";
    }
    goto finished;
  }

  if(sd->sigsz > (1 << lds)) {  // Check for range if returned size bigger than requested
    int signd = type & (STR_SIGNED | STR_UNSIGNED);
    long long int smin[4] = { INT8_MIN, INT16_MIN, INT32_MIN, INT64_MIN };
    long long int smax[4] = { INT8_MAX, INT16_MAX, INT32_MAX, INT64_MAX };
    unsigned long long int umax[4] = { UINT8_MAX, UINT16_MAX, UINT32_MAX, UINT64_MAX };

    if(signd == STR_SIGNED) {   // Strictly signed
      if(sd->ll < smin[lds] || sd->ll > smax[lds]) {
        err = cache_string(str_ccprintf("out of int%d range", 1 << (3 + lds)));
        goto finished;
      }
    } else if(signd == STR_UNSIGNED) {  // Strictly unsigned are out of range if u and -u are
      if(sd->ull > umax[lds] && ~sd->ull + 1 > umax[lds]) {
        err = cache_string(str_ccprintf("out of uint%d range", 1 << (3 + lds)));
        goto finished;
      }
    } else {                    // Neither strictly signed or unsigned
      if((sd->ll < smin[lds] || sd->ll > smax[lds]) && sd->ull > umax[lds] && ~sd->ull + 1 > umax[lds]) {
        err = cache_string(str_ccprintf("out of int%d and uint%d range", 1 << (3 + lds), 1 << (3 + lds)));
        goto finished;
      }
    }
  }

  ret = sd->ull;

finished:
  if(errpp)
    *errpp = err;
  str_freedata(sd);

  return ret;
}

// Convert a data string (except STR_FILE) to a memory buffer suitable for AVRMEM use
int str_membuf(const char *str, int type, unsigned char *buf, int size, const char **errpp) {
  int n = 0;
  const char *err = NULL;
  Str2data *sd = NULL;

  type &= ~STR_FILE;
  if(type == 0)                 // Nothing requested, nothing gained
    goto finished;

  sd = str_todata(str, type, NULL, NULL);
  if(!sd->type || sd->errstr) {
    err = cache_string(sd->errstr);
    n = -1;
    goto finished;
  }

  if(sd->type == STR_STRING && sd->str_ptr) {
    size_t len = strlen(sd->str_ptr);

    for(size_t j = 0; j < len && n < size; j++)
      buf[n++] = (uint8_t) sd->str_ptr[j];
    if(n < size)                // Terminating nul
      buf[n++] = 0;
  } else if(sd->size > 0 && (sd->type & STR_NUMBER)) {
    // Always write little endian to AVR memory
    if(is_bigendian() && sd->size > 0 && (sd->type & STR_NUMBER))
      change_endian(sd->a, sd->size);
    for(int k = 0; k < sd->size && n < size; k++)
      buf[n++] = sd->a[k];
  }

finished:
  if(errpp)
    *errpp = err;
  str_freedata(sd);

  return n;
}

/*
 * The syntax of the -P port string is -P usb[:<vid>:<pid>]:[<serno>]
 *
 * Set *vidp and *pidp iff <vid> and <pid> are hex strings in [0, 0xffff]. If
 * there is no valid :<vid>:<pid> pair the full string after usb: is treated as
 * serial number, as is any trailing string after a valid usb:<vid>:<pid>:
 * sequence. Serial numbers, if detected in port, are stripped of colons, while
 * a maximum of n-1 serial-number characters are stored in the serno buffer
 * followed by a terminating nul character. If the serno buffer is too small
 * serno remains unset and the port argument is considered invalid.
 *
 * Returns
 *  -1 on failed parsing (-P string does not start with usb[:], n too small)
 *   0 if no paramters was set (-P usb)
 *   1 if only serno was set or serno could have been set but was NULL
 *   2 if only vid and pid were or could have been set
 *   3 if vid, pid and serno were or could have been set
 *
 * *vidp and *pidp remain unchanged for return values smaller than 2. The serno
 * buffer remains unchanged on return values other than 1 or 3.
 */

int str_set_vid_pid_serno(const char *port, unsigned short *vidp, unsigned short *pidp,
  char *serno, const size_t n) {

  unsigned long vid, pid;
  const char *sn = port + 4;
  int ret = 0;
  char *pidstr, *endhex;

  if(!str_casestarts(port, "usb"))
    return -1;

  if(port[3] != ':')            // -P usb returns 0, wrong character after -p usb returns -1
    return port[3] == 0? 0: -1;

  errno = 0;                    // Read hex vid and assert it is followed by a colon
  vid = strtoul(port + 4, &endhex, 16);
  if(endhex == port + 4 || *endhex != ':' || errno || vid > 0xffff)
    goto setserno;

  errno = 0;                    // Read hex pid and assert it is followed by nul or colon
  pidstr = endhex + 1;
  pid = strtoul(pidstr, &endhex, 16);
  if(endhex == pidstr || (*endhex != 0 && *endhex != ':') || errno || pid > 0xffff)
    goto setserno;

  if(vidp)
    *vidp = vid;
  if(pidp)
    *pidp = pid;

  if(*endhex == 0)              // No :<serno> after <pid>? Finished
    return 2;

  sn = endhex + 1;
  ret = 2;
setserno:

  if(!n)
    return -1;

  char *serbuf = mmt_malloc(n), *bp;
  // Copy the serial number into given buffer but without colons
  for(bp = serbuf; *sn; sn++)
    if(*sn != ':') {
      if(bp >= serbuf + n-1)
        break;
      *bp++ = *sn;
    }
  *bp = 0;

  if(serno && !*sn)
    memcpy(serno, serbuf, n);
  mmt_free(serbuf);

  return *sn? -1: ret+1;
}

/*
 * Returns the next space separated token in buf (terminating it) and places
 * start of next token into pointer pointed to by next. Keeps single or double
 * quoted strings together and changes backslash-space sequences to space
 * whilst keeping other backslashed characters. Used for terminal line parsing
 * and reading files with ASCII numbers.
 */
char *str_nexttok(char *buf, const char *delim, char **next) {
  unsigned char *q, *r, *w, inquote;

  q = (unsigned char *) buf;
  while(*q && strchr(delim, *q))
    q++;

  // Isolate first token
  for(inquote = 0, w = r = q; *r && !(strchr(delim, *r) && !inquote); *w++ = *r++) {
    // Poor man's quote and escape processing
    if(*r == '"' || *r == '\'')
      inquote = inquote && *r == inquote? 0: inquote? inquote: *r;
    else if(*r == '\\' && r[1] && strchr(delim, r[1]))  // Remove \ before space for file names
      r++;
    else if(*r == '\\' && r[1]) // Leave other \ to keep C-style, eg, '\n'
      *w++ = *r++;
  }
  if(*r)
    r++;
  *w = 0;

  // Find start of next token
  while(*r && strchr(delim, *r))
    r++;

  if(next)
    *next = (char *) r;

  return (char *) q;
}

// Return string for frequency with n significant digits and xHz unit in closed-circuit space
const char *str_ccfrq(double f, int n) {
  struct {
    double fq;
    const char *pre;
  } prefix[] = { {1e9, "G"}, {1e6, "M"}, {1e3, "k"}, };

  for(size_t i = 0; i < sizeof prefix/sizeof *prefix; i++)
    if(f >= prefix[i].fq)
      return str_ccprintf("%.*g %sHz", n, f/prefix[i].fq, prefix[i].pre);
  return str_ccprintf("%.*g Hz", n, f);
}

// Return an uppercase hex string of max 64 len bytes from binary buffer buf
const char *str_cchex(const void *buf, size_t len, int add_space) {
  if(len > 64)                  // Sanity
    len = 64;
  int wd = 2 + !!add_space;
  char *ret = avr_cc_buffer(wd*len + 1);

  for(size_t i = 0; i < len; i++)
    sprintf(ret + i*wd, "%s%02X", &" "[3 - wd], ((unsigned char *) buf)[i]);
  return ret;
}

/*
 * From https://github.com/git/git/blob/master/levenshtein.c
 *
 * This function implements the Damerau-Levenshtein algorithm to calculate a
 * distance between strings.
 *
 * Basically, it says how many letters need to be swapped, substituted, deleted
 * from, or added to str1, at least, to get str2.
 *
 * The idea is to build a distance matrix for the substrings of both strings.
 * To avoid a large space complexity, only the last three rows are kept in
 * memory (if swaps had the same or higher cost as one deletion plus one
 * insertion, only two rows would be needed).
 *
 * At any stage, "i + 1" denotes the length of the current substring of str1
 * that the distance is calculated for.
 *
 * row2 holds the current row, row1 the previous row (i.e. for the substring of
 * str1 of length "i"), and row0 the row before that.
 *
 * In other words, at the start of the big loop, row2[j + 1] contains the
 * Damerau-Levenshtein distance between the substring of str1 of length "i" and
 * the substring of str2 of length "j + 1".
 *
 * All the big loop does is determine the partial minimum-cost paths.
 *
 * It does so by calculating the costs of the path ending in characters i (in
 * str1) and j (in str2), respectively, given that the last operation is a
 * substitution, a swap, a deletion, or an insertion.
 *
 * This implementation allows the costs to be weighted:
 *  - swap
 *  - subst (as in "Substitution")
 *  - add (for insertion, AKA "Add")
 *  - del (as in "Deletion")
 *
 * Note that this algorithm calculates a distance _iff_ del == add.
 */

int str_levenshtein(const char *str1, const char *str2, int swap, int subst, int add, int del) {

  int i, j, len1 = strlen(str1), len2 = strlen(str2);
  int *row0 = mmt_malloc((len2 + 1)*sizeof *row0);
  int *row1 = mmt_malloc((len2 + 1)*sizeof *row1);
  int *row2 = mmt_malloc((len2 + 1)*sizeof *row2);

  for(j = 0; j <= len2; j++)
    row1[j] = j*add;
  for(i = 0; i < len1; i++) {
    row2[0] = (i + 1)*del;
    for(j = 0; j < len2; j++) {
      // Substitution
      row2[j + 1] = row1[j] + subst*(str1[i] != str2[j]);
      // Swap
      if(i > 0 && j > 0 && str1[i - 1] == str2[j] && str1[i] == str2[j - 1] && row2[j + 1] > row0[j - 1] + swap)
        row2[j + 1] = row0[j - 1] + swap;
      // Deletion
      if(row2[j + 1] > row1[j + 1] + del)
        row2[j + 1] = row1[j + 1] + del;
      // Insertion
      if(row2[j + 1] > row2[j] + add)
        row2[j + 1] = row2[j] + add;
    }
    int *temp = row0;

    row0 = row1;
    row1 = row2;
    row2 = temp;
  }
  i = row1[len2];
  mmt_free(row0);
  mmt_free(row1);
  mmt_free(row2);
  return i;
}

// Alphanumeric chars get the full weight, all others such as hyphen or underscore get less
static size_t wchr(size_t w, unsigned char c) {
  return isascii(c) && isalnum(c)? w: w >= 8? w/8: 1;
}

// Index of character in string or -1 of not found
static int chridx(char *str, char c) {
  char *e = strchr(str, c);

  return e? e - str: -1;
}

// (x, y) position of key on keyboard 1 being the centre
static void xypos(char c, double *x, double *y) {
  int num = chridx("1234567890", c);
  int upp = chridx("qwertyuiop", c);
  int mid = chridx("asdfghjkl", c);
  int low = chridx("zxcvbnm", c);

  // My laptop's keyboard layout: your mileage may vary (smr)
  *x = num >= 0? num: upp >= 0? upp + 0.5: mid >= 0? mid + 0.75: low >= 0? low + 1.25: -3.0;
  *y = num >= 0? 0.0: upp >= 0? 1.0: mid >= 0? 2.0: low >= 0? 3.0: -3.0;
}

// Weight by keyboard distance
static size_t qwertydist(size_t w, unsigned char c1, unsigned char c2) {
  if(c1 == c2)
    return 0;

  double x1, y1, x2, y2;

  xypos(tolower(c1), &x1, &y1);
  xypos(tolower(c2), &x2, &y2);

  if(x1 == x2 && y1 == y2)
    return w;

  size_t ret = isalpha(c1) && isalpha(c2) && isupper(c1) != isupper(c2)? w/8: 0;

  ret += sqrt((x1 - x2)*(x1 - x2) + (y1 - y2)*(y1 - y2))/2.5*w;

  return ret > w? w: ret > 0? ret: 1;
}

// Substitution cost considering qwerty keyboard typos and case
static size_t csubs(size_t w, unsigned char c1, unsigned char c2) {
  if(c1 >= 128 || c2 >= 128)
    return c1 != c2? w: 0;

  if(w < 8)
    w = 8;

  static size_t wmat[128][128]; // Compute once, read-only cache

  if(!wmat[0][1])               // Initialize weight matrix
    for(size_t k1 = 0; k1 < 128; k1++)
      for(size_t k2 = 0; k2 < 128; k2++)
        wmat[k1][k2] =
          k1 == k2? 0:
          !isalnum(k1) && !isalnum(k2)? w/8:
          !isalnum(k1) || !isalnum(k2)? w:
          isalpha(k1) && isalpha(k2) && tolower(k1) == tolower(k2)? w/8: qwertydist(w, k1, k2);

  return wmat[c1][c2];
}

// Cost of morphing s1 to s2 modelling typos and mix-up of non-alphanumeric letters
size_t str_weighted_damerau_levenshtein(const char *s1, const char *s2) {
  const size_t swap = 3;        // Transposing neighbouring letters is an easy mistake to make
  const size_t subst = 32, add = 32, del = 32;  // Must be multiples of 8
  size_t i, j, len1 = strlen(s1), len2 = strlen(s2);
  size_t *row0 = mmt_malloc((len2 + 1)*sizeof *row0);
  size_t *row1 = mmt_malloc((len2 + 1)*sizeof *row1);
  size_t *row2 = mmt_malloc((len2 + 1)*sizeof *row2);
  unsigned char *str1 = (unsigned char *) s1, *str2 = (unsigned char *) s2;

  for(j = 0; j < len2; j++)
    row1[j + 1] = row1[j] + wchr(add, str2[j]);
  for(i = 0; i < len1; i++) {
    row2[0] = 0;
    for(size_t k = 0; k <= i; k++)
      row2[0] += wchr(del, str1[k]);
    for(j = 0; j < len2; j++) {
      // Substitution of str1[i] with str2[j]
      row2[j + 1] = row1[j] + (str1[i] != str2[j]? csubs(subst, str1[i], str2[j]): 0);
      // Swap: str1[i-1]str1[i] is same as str2[j]str2[j-1]
      if(i > 0 && j > 0 && str1[i - 1] == str2[j] && str1[i] == str2[j - 1] && row2[j + 1] > row0[j - 1] + swap)
        row2[j + 1] = row0[j - 1] + swap;
      // Deletion of str1[i]
      size_t wdel = wchr(del, str1[i]);

      if(row2[j + 1] > row1[j + 1] + wdel)
        row2[j + 1] = row1[j + 1] + wdel;
      // Insertion of str2[j]
      size_t wadd = wchr(add, str2[j]);

      if(row2[j + 1] > row2[j] + wadd)
        row2[j + 1] = row2[j] + wadd;
      // Todo: fat finger, eg, typing test as tesdt or tedst
    }
    size_t *temp = row0;

    row0 = row1;
    row1 = row2;
    row2 = temp;
  }
  i = row1[len2];               // Last row2[len2]
  mmt_free(row0);
  mmt_free(row1);
  mmt_free(row2);
  return i;
}

// Puts a comma-separated list of matching MCU names into array p with n chars space
int str_mcunames_signature(const unsigned char *sigs, int pm, char *p, size_t n) {
  const char *matches[100];
  int matching = 0, k, N = sizeof matches/sizeof *matches;

  if(!pm || (pm & PM_ALL) == PM_ALL)    // Look up uP table when unrestricted by prog modes
    for(size_t i = 0; i < sizeof uP_table/sizeof *uP_table; i++)
      if(!is_memset(uP_table[i].sigs, 0xff, 3) && !is_memset(uP_table[i].sigs, 0, 3))
        if(0 == memcmp(sigs, uP_table[i].sigs, sizeof uP_table->sigs) && matching < N)
          matches[matching++] = uP_table[i].name;

  for(LNODEID lp = lfirst(part_list); lp; lp = lnext(lp)) {
    AVRPART *pp = ldata(lp);

    if(!*pp->id || *pp->id == '.')      // Skip invalid entries
      continue;
    if(is_memset(pp->signature, 0xff, 3) || is_memset(pp->signature, 0, 3))
      continue;
    if(!memcmp(sigs, pp->signature, 3) && (!pm || (pp->prog_modes & pm))) {
      for(k = 0; k < matching; k++)
        if(str_eq(matches[k], pp->desc))
          break;
      if(k == matching && matching < N)
        matches[matching++] = pp->desc;
    }
  }

  if(n && p) {
    *p = 0;

    for(int i = 0; i < matching; i++) {
      size_t len = strlen(matches[i]);

      if(n > len + 2) {
        if(i) {
          strcpy(p, ", ");
          n -= 2, p += 2;
        }
        strcpy(p, matches[i]);
        n -= len, p += len;
      }
    }
  }

  return matching;
}

// Returns a comma-separated list of matching MCU names in closed-circuit space
const char *str_ccmcunames_signature(const unsigned char *sigs, int pm) {
  char names[1024];

  // If no match is found, given required prog_modes, relax the match to any prog mode
  if(!str_mcunames_signature(sigs, pm, names, sizeof names) && pm && (pm & PM_ALL) != PM_ALL)
    (void) str_mcunames_signature(sigs, 0, names, sizeof names);

  return str_ccprintf("%s", names);
}

// Returns a comma-separated list of pgm->id names
const char *str_ccpgmids(LISTID pgm_id) {
  char ids[1024], *idp = ids;

  for(LNODEID idn = lfirst(pgm_id); idn; idn = lnext(idn)) {
    char *id = ldata(idn);

    if((idp - ids) + 3 + strlen(id) <= sizeof ids) {
      if(idp > ids)
        strcpy(idp, ", "), idp += 2;
      strcpy(idp, id), idp += strlen(id);
    }
  }
  return str_ccprintf("%s", ids);
}

// Return a string in closed-circuit space of the address in a memory of size size
const char *str_ccaddress(int addr, int size) {
   return size <= 16?
     str_ccprintf("%d", addr):
     str_ccprintf("0x%0*x", intlog2(size-1)/4 + 1, addr);
}

// Return a malloc'd string quoted for a shell argument
char *str_quote_bash(const char *s) {
  size_t n = strlen(s);
  char *ret = mmt_malloc(4*n + 3), *r = ret;

  // Put string in single quotes and replace each ' with '\''
  *r++ = '\'';
  while(*s) {
    *r++ = *s++;
    if(r[-1] == '\'')
      *r++ = '\\', *r++ = '\'', *r++ = '\'';
  }
  *r++ = '\'';

  return ret;
}

// Return the string in closed-circuit space so it can be used as shell argument
const char *str_ccsharg(const char *str) {
  const char *special="\"'` $\\#[]<>|;{}()*?~&!", *s;

  for(s = str; *s; s++)
    if(strchr(special, *s))
      break;

  if(*s) {                      // str contains special bash char: must quote string
    char *r = str_quote_bash(str);
    str = str_ccprintf("%s", r);
    mmt_free(r);
  }

  return str;
}

// Return malloc'd ISR vector name without _ (given the vector number)
char *str_vectorname(const Avrintel *up, int vn) {
  if(!up->isrtable || vn < -1 || vn > up->ninterrupts)
    return mmt_strdup("unknown");

  char *ret = mmt_strdup((unsigned) vn >= up->ninterrupts? "ADDITIONAL_VECTOR": up->isrtable[vn]);

  // Remove all _ in vectorstr
  char *p = str_lc(ret), *q = p;
  do {
    while(*p == '_')
      p++;
    *q++ = *p;
  } while(*p++);

  return ret;
}