src/basic/string-util.c

   1 /* SPDX-License-Identifier: LGPL-2.1+ */
   2
   3 #include <errno.h>
   4 #include <stdarg.h>
   5 #include <stdint.h>
   6 #include <stdio.h>
   7 #include <stdio_ext.h>
   8 #include <stdlib.h>
   9 #include <string.h>
  10
  11 #include "alloc-util.h"
  12 #include "escape.h"
  13 #include "gunicode.h"
  14 #include "locale-util.h"
  15 #include "macro.h"
  16 #include "string-util.h"
  17 #include "terminal-util.h"
  18 #include "utf8.h"
  19 #include "util.h"
  20 #include "fileio.h"
  21
  22 int strcmp_ptr(const char *a, const char *b) {
  23
  24         /* Like strcmp(), but tries to make sense of NULL pointers */
  25         if (a && b)
  26                 return strcmp(a, b);
  27
  28         if (!a && b)
  29                 return -1;
  30
  31         if (a && !b)
  32                 return 1;
  33
  34         return 0;
  35 }
  36
  37 char* endswith(const char *s, const char *postfix) {
  38         size_t sl, pl;
  39
  40         assert(s);
  41         assert(postfix);
  42
  43         sl = strlen(s);
  44         pl = strlen(postfix);
  45
  46         if (pl == 0)
  47                 return (char*) s + sl;
  48
  49         if (sl < pl)
  50                 return NULL;
  51
  52         if (memcmp(s + sl - pl, postfix, pl) != 0)
  53                 return NULL;
  54
  55         return (char*) s + sl - pl;
  56 }
  57
  58 char* endswith_no_case(const char *s, const char *postfix) {
  59         size_t sl, pl;
  60
  61         assert(s);
  62         assert(postfix);
  63
  64         sl = strlen(s);
  65         pl = strlen(postfix);
  66
  67         if (pl == 0)
  68                 return (char*) s + sl;
  69
  70         if (sl < pl)
  71                 return NULL;
  72
  73         if (strcasecmp(s + sl - pl, postfix) != 0)
  74                 return NULL;
  75
  76         return (char*) s + sl - pl;
  77 }
  78
  79 char* first_word(const char *s, const char *word) {
  80         size_t sl, wl;
  81         const char *p;
  82
  83         assert(s);
  84         assert(word);
  85
  86         /* Checks if the string starts with the specified word, either
  87          * followed by NUL or by whitespace. Returns a pointer to the
  88          * NUL or the first character after the whitespace. */
  89
  90         sl = strlen(s);
  91         wl = strlen(word);
  92
  93         if (sl < wl)
  94                 return NULL;
  95
  96         if (wl == 0)
  97                 return (char*) s;
  98
  99         if (memcmp(s, word, wl) != 0)
 100                 return NULL;
 101
 102         p = s + wl;
 103         if (*p == 0)
 104                 return (char*) p;
 105
 106         if (!strchr(WHITESPACE, *p))
 107                 return NULL;
 108
 109         p += strspn(p, WHITESPACE);
 110         return (char*) p;
 111 }
 112
 113 static size_t strcspn_escaped(const char *s, const char *reject) {
 114         bool escaped = false;
 115         int n;
 116
 117         for (n=0; s[n]; n++) {
 118                 if (escaped)
 119                         escaped = false;
 120                 else if (s[n] == '\\')
 121                         escaped = true;
 122                 else if (strchr(reject, s[n]))
 123                         break;
 124         }
 125
 126         /* if s ends in \, return index of previous char */
 127         return n - escaped;
 128 }
 129
 130 /* Split a string into words. */
 131 const char* split(const char **state, size_t *l, const char *separator, SplitFlags flags) {
 132         const char *current;
 133
 134         current = *state;
 135
 136         if (!*current) {
 137                 assert(**state == '\0');
 138                 return NULL;
 139         }
 140
 141         current += strspn(current, separator);
 142         if (!*current) {
 143                 *state = current;
 144                 return NULL;
 145         }
 146
 147         if (flags & SPLIT_QUOTES && strchr("\'\"", *current)) {
 148                 char quotechars[2] = {*current, '\0'};
 149
 150                 *l = strcspn_escaped(current + 1, quotechars);
 151                 if (current[*l + 1] == '\0' || current[*l + 1] != quotechars[0] ||
 152                     (current[*l + 2] && !strchr(separator, current[*l + 2]))) {
 153                         /* right quote missing or garbage at the end */
 154                         if (flags & SPLIT_RELAX) {
 155                                 *state = current + *l + 1 + (current[*l + 1] != '\0');
 156                                 return current + 1;
 157                         }
 158                         *state = current;
 159                         return NULL;
 160                 }
 161                 *state = current++ + *l + 2;
 162         } else if (flags & SPLIT_QUOTES) {
 163                 *l = strcspn_escaped(current, separator);
 164                 if (current[*l] && !strchr(separator, current[*l]) && !(flags & SPLIT_RELAX)) {
 165                         /* unfinished escape */
 166                         *state = current;
 167                         return NULL;
 168                 }
 169                 *state = current + *l;
 170         } else {
 171                 *l = strcspn(current, separator);
 172                 *state = current + *l;
 173         }
 174
 175         return current;
 176 }
 177
 178 char *strnappend(const char *s, const char *suffix, size_t b) {
 179         size_t a;
 180         char *r;
 181
 182         if (!s && !suffix)
 183                 return strdup("");
 184
 185         if (!s)
 186                 return strndup(suffix, b);
 187
 188         if (!suffix)
 189                 return strdup(s);
 190
 191         assert(s);
 192         assert(suffix);
 193
 194         a = strlen(s);
 195         if (b > ((size_t) -1) - a)
 196                 return NULL;
 197
 198         r = new(char, a+b+1);
 199         if (!r)
 200                 return NULL;
 201
 202         memcpy(r, s, a);
 203         memcpy(r+a, suffix, b);
 204         r[a+b] = 0;
 205
 206         return r;
 207 }
 208
 209 char *strappend(const char *s, const char *suffix) {
 210         return strnappend(s, suffix, strlen_ptr(suffix));
 211 }
 212
 213 char *strjoin_real(const char *x, ...) {
 214         va_list ap;
 215         size_t l;
 216         char *r, *p;
 217
 218         va_start(ap, x);
 219
 220         if (x) {
 221                 l = strlen(x);
 222
 223                 for (;;) {
 224                         const char *t;
 225                         size_t n;
 226
 227                         t = va_arg(ap, const char *);
 228                         if (!t)
 229                                 break;
 230
 231                         n = strlen(t);
 232                         if (n > ((size_t) -1) - l) {
 233                                 va_end(ap);
 234                                 return NULL;
 235                         }
 236
 237                         l += n;
 238                 }
 239         } else
 240                 l = 0;
 241
 242         va_end(ap);
 243
 244         r = new(char, l+1);
 245         if (!r)
 246                 return NULL;
 247
 248         if (x) {
 249                 p = stpcpy(r, x);
 250
 251                 va_start(ap, x);
 252
 253                 for (;;) {
 254                         const char *t;
 255
 256                         t = va_arg(ap, const char *);
 257                         if (!t)
 258                                 break;
 259
 260                         p = stpcpy(p, t);
 261                 }
 262
 263                 va_end(ap);
 264         } else
 265                 r[0] = 0;
 266
 267         return r;
 268 }
 269
 270 char *strstrip(char *s) {
 271         if (!s)
 272                 return NULL;
 273
 274         /* Drops trailing whitespace. Modifies the string in place. Returns pointer to first non-space character */
 275
 276         return delete_trailing_chars(skip_leading_chars(s, WHITESPACE), WHITESPACE);
 277 }
 278
 279 char *delete_chars(char *s, const char *bad) {
 280         char *f, *t;
 281
 282         /* Drops all specified bad characters, regardless where in the string */
 283
 284         if (!s)
 285                 return NULL;
 286
 287         if (!bad)
 288                 bad = WHITESPACE;
 289
 290         for (f = s, t = s; *f; f++) {
 291                 if (strchr(bad, *f))
 292                         continue;
 293
 294                 *(t++) = *f;
 295         }
 296
 297         *t = 0;
 298
 299         return s;
 300 }
 301
 302 char *delete_trailing_chars(char *s, const char *bad) {
 303         char *p, *c = s;
 304
 305         /* Drops all specified bad characters, at the end of the string */
 306
 307         if (!s)
 308                 return NULL;
 309
 310         if (!bad)
 311                 bad = WHITESPACE;
 312
 313         for (p = s; *p; p++)
 314                 if (!strchr(bad, *p))
 315                         c = p + 1;
 316
 317         *c = 0;
 318
 319         return s;
 320 }
 321
 322 char *truncate_nl(char *s) {
 323         assert(s);
 324
 325         s[strcspn(s, NEWLINE)] = 0;
 326         return s;
 327 }
 328
 329 char ascii_tolower(char x) {
 330
 331         if (x >= 'A' && x <= 'Z')
 332                 return x - 'A' + 'a';
 333
 334         return x;
 335 }
 336
 337 char ascii_toupper(char x) {
 338
 339         if (x >= 'a' && x <= 'z')
 340                 return x - 'a' + 'A';
 341
 342         return x;
 343 }
 344
 345 char *ascii_strlower(char *t) {
 346         char *p;
 347
 348         assert(t);
 349
 350         for (p = t; *p; p++)
 351                 *p = ascii_tolower(*p);
 352
 353         return t;
 354 }
 355
 356 char *ascii_strupper(char *t) {
 357         char *p;
 358
 359         assert(t);
 360
 361         for (p = t; *p; p++)
 362                 *p = ascii_toupper(*p);
 363
 364         return t;
 365 }
 366
 367 char *ascii_strlower_n(char *t, size_t n) {
 368         size_t i;
 369
 370         if (n <= 0)
 371                 return t;
 372
 373         for (i = 0; i < n; i++)
 374                 t[i] = ascii_tolower(t[i]);
 375
 376         return t;
 377 }
 378
 379 int ascii_strcasecmp_n(const char *a, const char *b, size_t n) {
 380
 381         for (; n > 0; a++, b++, n--) {
 382                 int x, y;
 383
 384                 x = (int) (uint8_t) ascii_tolower(*a);
 385                 y = (int) (uint8_t) ascii_tolower(*b);
 386
 387                 if (x != y)
 388                         return x - y;
 389         }
 390
 391         return 0;
 392 }
 393
 394 int ascii_strcasecmp_nn(const char *a, size_t n, const char *b, size_t m) {
 395         int r;
 396
 397         r = ascii_strcasecmp_n(a, b, MIN(n, m));
 398         if (r != 0)
 399                 return r;
 400
 401         if (n < m)
 402                 return -1;
 403         else if (n > m)
 404                 return 1;
 405         else
 406                 return 0;
 407 }
 408
 409 bool chars_intersect(const char *a, const char *b) {
 410         const char *p;
 411
 412         /* Returns true if any of the chars in a are in b. */
 413         for (p = a; *p; p++)
 414                 if (strchr(b, *p))
 415                         return true;
 416
 417         return false;
 418 }
 419
 420 bool string_has_cc(const char *p, const char *ok) {
 421         const char *t;
 422
 423         assert(p);
 424
 425         /*
 426          * Check if a string contains control characters. If 'ok' is
 427          * non-NULL it may be a string containing additional CCs to be
 428          * considered OK.
 429          */
 430
 431         for (t = p; *t; t++) {
 432                 if (ok && strchr(ok, *t))
 433                         continue;
 434
 435                 if (*t > 0 && *t < ' ')
 436                         return true;
 437
 438                 if (*t == 127)
 439                         return true;
 440         }
 441
 442         return false;
 443 }
 444
 445 static int write_ellipsis(char *buf, bool unicode) {
 446         if (unicode || is_locale_utf8()) {
 447                 buf[0] = 0xe2; /* tri-dot ellipsis: … */
 448                 buf[1] = 0x80;
 449                 buf[2] = 0xa6;
 450         } else {
 451                 buf[0] = '.';
 452                 buf[1] = '.';
 453                 buf[2] = '.';
 454         }
 455
 456         return 3;
 457 }
 458
 459 static char *ascii_ellipsize_mem(const char *s, size_t old_length, size_t new_length, unsigned percent) {
 460         size_t x, need_space, suffix_len;
 461         char *t;
 462
 463         assert(s);
 464         assert(percent <= 100);
 465         assert(new_length != (size_t) -1);
 466
 467         if (old_length <= new_length)
 468                 return strndup(s, old_length);
 469
 470         /* Special case short ellipsations */
 471         switch (new_length) {
 472
 473         case 0:
 474                 return strdup("");
 475
 476         case 1:
 477                 if (is_locale_utf8())
 478                         return strdup("…");
 479                 else
 480                         return strdup(".");
 481
 482         case 2:
 483                 if (!is_locale_utf8())
 484                         return strdup("..");
 485
 486                 break;
 487
 488         default:
 489                 break;
 490         }
 491
 492         /* Calculate how much space the ellipsis will take up. If we are in UTF-8 mode we only need space for one
 493          * character ("…"), otherwise for three characters ("..."). Note that in both cases we need 3 bytes of storage,
 494          * either for the UTF-8 encoded character or for three ASCII characters. */
 495         need_space = is_locale_utf8() ? 1 : 3;
 496
 497         t = new(char, new_length+3);
 498         if (!t)
 499                 return NULL;
 500
 501         assert(new_length >= need_space);
 502
 503         x = ((new_length - need_space) * percent + 50) / 100;
 504         assert(x <= new_length - need_space);
 505
 506         memcpy(t, s, x);
 507         write_ellipsis(t + x, false);
 508         suffix_len = new_length - x - need_space;
 509         memcpy(t + x + 3, s + old_length - suffix_len, suffix_len);
 510         *(t + x + 3 + suffix_len) = '\0';
 511
 512         return t;
 513 }
 514
 515 char *ellipsize_mem(const char *s, size_t old_length, size_t new_length, unsigned percent) {
 516         size_t x, k, len, len2;
 517         const char *i, *j;
 518         char *e;
 519         int r;
 520
 521         /* Note that 'old_length' refers to bytes in the string, while 'new_length' refers to character cells taken up
 522          * on screen. This distinction doesn't matter for ASCII strings, but it does matter for non-ASCII UTF-8
 523          * strings.
 524          *
 525          * Ellipsation is done in a locale-dependent way:
 526          * 1. If the string passed in is fully ASCII and the current locale is not UTF-8, three dots are used ("...")
 527          * 2. Otherwise, a unicode ellipsis is used ("…")
 528          *
 529          * In other words: you'll get a unicode ellipsis as soon as either the string contains non-ASCII characters or
 530          * the current locale is UTF-8.
 531          */
 532
 533         assert(s);
 534         assert(percent <= 100);
 535
 536         if (new_length == (size_t) -1)
 537                 return strndup(s, old_length);
 538
 539         if (new_length == 0)
 540                 return strdup("");
 541
 542         /* If no multibyte characters use ascii_ellipsize_mem for speed */
 543         if (ascii_is_valid_n(s, old_length))
 544                 return ascii_ellipsize_mem(s, old_length, new_length, percent);
 545
 546         x = ((new_length - 1) * percent) / 100;
 547         assert(x <= new_length - 1);
 548
 549         k = 0;
 550         for (i = s; i < s + old_length; i = utf8_next_char(i)) {
 551                 char32_t c;
 552                 int w;
 553
 554                 r = utf8_encoded_to_unichar(i, &c);
 555                 if (r < 0)
 556                         return NULL;
 557
 558                 w = unichar_iswide(c) ? 2 : 1;
 559                 if (k + w <= x)
 560                         k += w;
 561                 else
 562                         break;
 563         }
 564
 565         for (j = s + old_length; j > i; ) {
 566                 char32_t c;
 567                 int w;
 568                 const char *jj;
 569
 570                 jj = utf8_prev_char(j);
 571                 r = utf8_encoded_to_unichar(jj, &c);
 572                 if (r < 0)
 573                         return NULL;
 574
 575                 w = unichar_iswide(c) ? 2 : 1;
 576                 if (k + w <= new_length) {
 577                         k += w;
 578                         j = jj;
 579                 } else
 580                         break;
 581         }
 582         assert(i <= j);
 583
 584         /* we don't actually need to ellipsize */
 585         if (i == j)
 586                 return memdup_suffix0(s, old_length);
 587
 588         /* make space for ellipsis, if possible */
 589         if (j < s + old_length)
 590                 j = utf8_next_char(j);
 591         else if (i > s)
 592                 i = utf8_prev_char(i);
 593
 594         len = i - s;
 595         len2 = s + old_length - j;
 596         e = new(char, len + 3 + len2 + 1);
 597         if (!e)
 598                 return NULL;
 599
 600         /*
 601         printf("old_length=%zu new_length=%zu x=%zu len=%u len2=%u k=%u\n",
 602                old_length, new_length, x, len, len2, k);
 603         */
 604
 605         memcpy(e, s, len);
 606         write_ellipsis(e + len, true);
 607         memcpy(e + len + 3, j, len2);
 608         *(e + len + 3 + len2) = '\0';
 609
 610         return e;
 611 }
 612
 613 char *cellescape(char *buf, size_t len, const char *s) {
 614         /* Escape and ellipsize s into buffer buf of size len. Only non-control ASCII
 615          * characters are copied as they are, everything else is escaped. The result
 616          * is different then if escaping and ellipsization was performed in two
 617          * separate steps, because each sequence is either stored in full or skipped.
 618          *
 619          * This function should be used for logging about strings which expected to
 620          * be plain ASCII in a safe way.
 621          *
 622          * An ellipsis will be used if s is too long. It was always placed at the
 623          * very end.
 624          */
 625
 626         size_t i = 0, last_char_width[4] = {}, k = 0, j;
 627
 628         assert(len > 0); /* at least a terminating NUL */
 629
 630         for (;;) {
 631                 char four[4];
 632                 int w;
 633
 634                 if (*s == 0) /* terminating NUL detected? then we are done! */
 635                         goto done;
 636
 637                 w = cescape_char(*s, four);
 638                 if (i + w + 1 > len) /* This character doesn't fit into the buffer anymore? In that case let's
 639                                       * ellipsize at the previous location */
 640                         break;
 641
 642                 /* OK, there was space, let's add this escaped character to the buffer */
 643                 memcpy(buf + i, four, w);
 644                 i += w;
 645
 646                 /* And remember its width in the ring buffer */
 647                 last_char_width[k] = w;
 648                 k = (k + 1) % 4;
 649
 650                 s++;
 651         }
 652
 653         /* Ellipsation is necessary. This means we might need to truncate the string again to make space for 4
 654          * characters ideally, but the buffer is shorter than that in the first place take what we can get */
 655         for (j = 0; j < ELEMENTSOF(last_char_width); j++) {
 656
 657                 if (i + 4 <= len) /* nice, we reached our space goal */
 658                         break;
 659
 660                 k = k == 0 ? 3 : k - 1;
 661                 if (last_char_width[k] == 0) /* bummer, we reached the beginning of the strings */
 662                         break;
 663
 664                 assert(i >= last_char_width[k]);
 665                 i -= last_char_width[k];
 666         }
 667
 668         if (i + 4 <= len) /* yay, enough space */
 669                 i += write_ellipsis(buf + i, false);
 670         else if (i + 3 <= len) { /* only space for ".." */
 671                 buf[i++] = '.';
 672                 buf[i++] = '.';
 673         } else if (i + 2 <= len) /* only space for a single "." */
 674                 buf[i++] = '.';
 675         else
 676                 assert(i + 1 <= len);
 677
 678  done:
 679         buf[i] = '\0';
 680         return buf;
 681 }
 682
 683 bool nulstr_contains(const char *nulstr, const char *needle) {
 684         const char *i;
 685
 686         if (!nulstr)
 687                 return false;
 688
 689         NULSTR_FOREACH(i, nulstr)
 690                 if (streq(i, needle))
 691                         return true;
 692
 693         return false;
 694 }
 695
 696 char* strshorten(char *s, size_t l) {
 697         assert(s);
 698
 699         if (strnlen(s, l+1) > l)
 700                 s[l] = 0;
 701
 702         return s;
 703 }
 704
 705 char *strreplace(const char *text, const char *old_string, const char *new_string) {
 706         size_t l, old_len, new_len, allocated = 0;
 707         char *t, *ret = NULL;
 708         const char *f;
 709
 710         assert(old_string);
 711         assert(new_string);
 712
 713         if (!text)
 714                 return NULL;
 715
 716         old_len = strlen(old_string);
 717         new_len = strlen(new_string);
 718
 719         l = strlen(text);
 720         if (!GREEDY_REALLOC(ret, allocated, l+1))
 721                 return NULL;
 722
 723         f = text;
 724         t = ret;
 725         while (*f) {
 726                 size_t d, nl;
 727
 728                 if (!startswith(f, old_string)) {
 729                         *(t++) = *(f++);
 730                         continue;
 731                 }
 732
 733                 d = t - ret;
 734                 nl = l - old_len + new_len;
 735
 736                 if (!GREEDY_REALLOC(ret, allocated, nl + 1))
 737                         return mfree(ret);
 738
 739                 l = nl;
 740                 t = ret + d;
 741
 742                 t = stpcpy(t, new_string);
 743                 f += old_len;
 744         }
 745
 746         *t = 0;
 747         return ret;
 748 }
 749
 750 static void advance_offsets(ssize_t diff, size_t offsets[2], size_t shift[2], size_t size) {
 751         if (!offsets)
 752                 return;
 753
 754         if ((size_t) diff < offsets[0])
 755                 shift[0] += size;
 756         if ((size_t) diff < offsets[1])
 757                 shift[1] += size;
 758 }
 759
 760 char *strip_tab_ansi(char **ibuf, size_t *_isz, size_t highlight[2]) {
 761         const char *i, *begin = NULL;
 762         enum {
 763                 STATE_OTHER,
 764                 STATE_ESCAPE,
 765                 STATE_CSI,
 766                 STATE_CSO,
 767         } state = STATE_OTHER;
 768         char *obuf = NULL;
 769         size_t osz = 0, isz, shift[2] = {};
 770         FILE *f;
 771
 772         assert(ibuf);
 773         assert(*ibuf);
 774
 775         /* This does three things:
 776          *
 777          * 1. Replaces TABs by 8 spaces
 778          * 2. Strips ANSI color sequences (a subset of CSI), i.e. ESC '[' … 'm' sequences
 779          * 3. Strips ANSI operating system sequences (CSO), i.e. ESC ']' … BEL sequences
 780          *
 781          * Everything else will be left as it is. In particular other ANSI sequences are left as they are, as are any
 782          * other special characters. Truncated ANSI sequences are left-as is too. This call is supposed to suppress the
 783          * most basic formatting noise, but nothing else.
 784          *
 785          * Why care for CSO sequences? Well, to undo what terminal_urlify() and friends generate. */
 786
 787         isz = _isz ? *_isz : strlen(*ibuf);
 788
 789         f = open_memstream(&obuf, &osz);
 790         if (!f)
 791                 return NULL;
 792
 793         /* Note we turn off internal locking on f for performance reasons.  It's safe to do so since we created f here
 794          * and it doesn't leave our scope. */
 795
 796         (void) __fsetlocking(f, FSETLOCKING_BYCALLER);
 797
 798         for (i = *ibuf; i < *ibuf + isz + 1; i++) {
 799
 800                 switch (state) {
 801
 802                 case STATE_OTHER:
 803                         if (i >= *ibuf + isz) /* EOT */
 804                                 break;
 805                         else if (*i == '\x1B')
 806                                 state = STATE_ESCAPE;
 807                         else if (*i == '\t') {
 808                                 fputs("        ", f);
 809                                 advance_offsets(i - *ibuf, highlight, shift, 7);
 810                         } else
 811                                 fputc(*i, f);
 812
 813                         break;
 814
 815                 case STATE_ESCAPE:
 816                         if (i >= *ibuf + isz) { /* EOT */
 817                                 fputc('\x1B', f);
 818                                 advance_offsets(i - *ibuf, highlight, shift, 1);
 819                                 break;
 820                         } else if (*i == '[') { /* ANSI CSI */
 821                                 state = STATE_CSI;
 822                                 begin = i + 1;
 823                         } else if (*i == ']') { /* ANSI CSO */
 824                                 state = STATE_CSO;
 825                                 begin = i + 1;
 826                         } else {
 827                                 fputc('\x1B', f);
 828                                 fputc(*i, f);
 829                                 advance_offsets(i - *ibuf, highlight, shift, 1);
 830                                 state = STATE_OTHER;
 831                         }
 832
 833                         break;
 834
 835                 case STATE_CSI:
 836
 837                         if (i >= *ibuf + isz || /* EOT … */
 838                             !strchr("01234567890;m", *i)) { /* … or invalid chars in sequence */
 839                                 fputc('\x1B', f);
 840                                 fputc('[', f);
 841                                 advance_offsets(i - *ibuf, highlight, shift, 2);
 842                                 state = STATE_OTHER;
 843                                 i = begin-1;
 844                         } else if (*i == 'm')
 845                                 state = STATE_OTHER;
 846
 847                         break;
 848
 849                 case STATE_CSO:
 850
 851                         if (i >= *ibuf + isz || /* EOT … */
 852                             (*i != '\a' && (uint8_t) *i < 32U) || (uint8_t) *i > 126U) { /* … or invalid chars in sequence */
 853                                 fputc('\x1B', f);
 854                                 fputc(']', f);
 855                                 advance_offsets(i - *ibuf, highlight, shift, 2);
 856                                 state = STATE_OTHER;
 857                                 i = begin-1;
 858                         } else if (*i == '\a')
 859                                 state = STATE_OTHER;
 860
 861                         break;
 862                 }
 863         }
 864
 865         if (fflush_and_check(f) < 0) {
 866                 fclose(f);
 867                 return mfree(obuf);
 868         }
 869
 870         fclose(f);
 871
 872         free(*ibuf);
 873         *ibuf = obuf;
 874
 875         if (_isz)
 876                 *_isz = osz;
 877
 878         if (highlight) {
 879                 highlight[0] += shift[0];
 880                 highlight[1] += shift[1];
 881         }
 882
 883         return obuf;
 884 }
 885
 886 char *strextend_with_separator(char **x, const char *separator, ...) {
 887         bool need_separator;
 888         size_t f, l, l_separator;
 889         char *r, *p;
 890         va_list ap;
 891
 892         assert(x);
 893
 894         l = f = strlen_ptr(*x);
 895
 896         need_separator = !isempty(*x);
 897         l_separator = strlen_ptr(separator);
 898
 899         va_start(ap, separator);
 900         for (;;) {
 901                 const char *t;
 902                 size_t n;
 903
 904                 t = va_arg(ap, const char *);
 905                 if (!t)
 906                         break;
 907
 908                 n = strlen(t);
 909
 910                 if (need_separator)
 911                         n += l_separator;
 912
 913                 if (n > ((size_t) -1) - l) {
 914                         va_end(ap);
 915                         return NULL;
 916                 }
 917
 918                 l += n;
 919                 need_separator = true;
 920         }
 921         va_end(ap);
 922
 923         need_separator = !isempty(*x);
 924
 925         r = realloc(*x, l+1);
 926         if (!r)
 927                 return NULL;
 928
 929         p = r + f;
 930
 931         va_start(ap, separator);
 932         for (;;) {
 933                 const char *t;
 934
 935                 t = va_arg(ap, const char *);
 936                 if (!t)
 937                         break;
 938
 939                 if (need_separator && separator)
 940                         p = stpcpy(p, separator);
 941
 942                 p = stpcpy(p, t);
 943
 944                 need_separator = true;
 945         }
 946         va_end(ap);
 947
 948         assert(p == r + l);
 949
 950         *p = 0;
 951         *x = r;
 952
 953         return r + l;
 954 }
 955
 956 char *strrep(const char *s, unsigned n) {
 957         size_t l;
 958         char *r, *p;
 959         unsigned i;
 960
 961         assert(s);
 962
 963         l = strlen(s);
 964         p = r = malloc(l * n + 1);
 965         if (!r)
 966                 return NULL;
 967
 968         for (i = 0; i < n; i++)
 969                 p = stpcpy(p, s);
 970
 971         *p = 0;
 972         return r;
 973 }
 974
 975 int split_pair(const char *s, const char *sep, char **l, char **r) {
 976         char *x, *a, *b;
 977
 978         assert(s);
 979         assert(sep);
 980         assert(l);
 981         assert(r);
 982
 983         if (isempty(sep))
 984                 return -EINVAL;
 985
 986         x = strstr(s, sep);
 987         if (!x)
 988                 return -EINVAL;
 989
 990         a = strndup(s, x - s);
 991         if (!a)
 992                 return -ENOMEM;
 993
 994         b = strdup(x + strlen(sep));
 995         if (!b) {
 996                 free(a);
 997                 return -ENOMEM;
 998         }
 999
1000         *l = a;
1001         *r = b;
1002
1003         return 0;
1004 }
1005
1006 int free_and_strdup(char **p, const char *s) {
1007         char *t;
1008
1009         assert(p);
1010
1011         /* Replaces a string pointer with a strdup()ed new string,
1012          * possibly freeing the old one. */
1013
1014         if (streq_ptr(*p, s))
1015                 return 0;
1016
1017         if (s) {
1018                 t = strdup(s);
1019                 if (!t)
1020                         return -ENOMEM;
1021         } else
1022                 t = NULL;
1023
1024         free(*p);
1025         *p = t;
1026
1027         return 1;
1028 }
1029
1030 int free_and_strndup(char **p, const char *s, size_t l) {
1031         char *t;
1032
1033         assert(p);
1034         assert(s || l == 0);
1035
1036         /* Replaces a string pointer with a strndup()ed new string,
1037          * freeing the old one. */
1038
1039         if (!*p && !s)
1040                 return 0;
1041
1042         if (*p && s && strneq(*p, s, l) && (l > strlen(*p) || (*p)[l] == '\0'))
1043                 return 0;
1044
1045         if (s) {
1046                 t = strndup(s, l);
1047                 if (!t)
1048                         return -ENOMEM;
1049         } else
1050                 t = NULL;
1051
1052         free_and_replace(*p, t);
1053         return 1;
1054 }
1055
1056 #if !HAVE_EXPLICIT_BZERO
1057 /*
1058  * Pointer to memset is volatile so that compiler must de-reference
1059  * the pointer and can't assume that it points to any function in
1060  * particular (such as memset, which it then might further "optimize")
1061  * This approach is inspired by openssl's crypto/mem_clr.c.
1062  */
1063 typedef void *(*memset_t)(void *,int,size_t);
1064
1065 static volatile memset_t memset_func = memset;
1066
1067 void explicit_bzero(void *p, size_t l) {
1068         memset_func(p, '\0', l);
1069 }
1070 #endif
1071
1072 char* string_erase(char *x) {
1073         if (!x)
1074                 return NULL;
1075
1076         /* A delicious drop of snake-oil! To be called on memory where
1077          * we stored passphrases or so, after we used them. */
1078         explicit_bzero(x, strlen(x));
1079         return x;
1080 }
1081
1082 char *string_free_erase(char *s) {
1083         return mfree(string_erase(s));
1084 }
1085
1086 bool string_is_safe(const char *p) {
1087         const char *t;
1088
1089         if (!p)
1090                 return false;
1091
1092         for (t = p; *t; t++) {
1093                 if (*t > 0 && *t < ' ') /* no control characters */
1094                         return false;
1095
1096                 if (strchr(QUOTES "\\\x7f", *t))
1097                         return false;
1098         }
1099
1100         return true;
1101 }