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