src/basic/string-util.c

   1 /* SPDX-License-Identifier: LGPL-2.1+ */
   2 /***
   3   This file is part of systemd.
   4
   5   Copyright 2010 Lennart Poettering
   6 ***/
   7
   8 #include <errno.h>
   9 #include <stdarg.h>
  10 #include <stdint.h>
  11 #include <stdio.h>
  12 #include <stdio_ext.h>
  13 #include <stdlib.h>
  14 #include <string.h>
  15
  16 #include "alloc-util.h"
  17 #include "escape.h"
  18 #include "gunicode.h"
  19 #include "locale-util.h"
  20 #include "macro.h"
  21 #include "string-util.h"
  22 #include "terminal-util.h"
  23 #include "utf8.h"
  24 #include "util.h"
  25 #include "fileio.h"
  26
  27 int strcmp_ptr(const char *a, const char *b) {
  28
  29         /* Like strcmp(), but tries to make sense of NULL pointers */
  30         if (a && b)
  31                 return strcmp(a, b);
  32
  33         if (!a && b)
  34                 return -1;
  35
  36         if (a && !b)
  37                 return 1;
  38
  39         return 0;
  40 }
  41
  42 char* endswith(const char *s, const char *postfix) {
  43         size_t sl, pl;
  44
  45         assert(s);
  46         assert(postfix);
  47
  48         sl = strlen(s);
  49         pl = strlen(postfix);
  50
  51         if (pl == 0)
  52                 return (char*) s + sl;
  53
  54         if (sl < pl)
  55                 return NULL;
  56
  57         if (memcmp(s + sl - pl, postfix, pl) != 0)
  58                 return NULL;
  59
  60         return (char*) s + sl - pl;
  61 }
  62
  63 char* endswith_no_case(const char *s, const char *postfix) {
  64         size_t sl, pl;
  65
  66         assert(s);
  67         assert(postfix);
  68
  69         sl = strlen(s);
  70         pl = strlen(postfix);
  71
  72         if (pl == 0)
  73                 return (char*) s + sl;
  74
  75         if (sl < pl)
  76                 return NULL;
  77
  78         if (strcasecmp(s + sl - pl, postfix) != 0)
  79                 return NULL;
  80
  81         return (char*) s + sl - pl;
  82 }
  83
  84 char* first_word(const char *s, const char *word) {
  85         size_t sl, wl;
  86         const char *p;
  87
  88         assert(s);
  89         assert(word);
  90
  91         /* Checks if the string starts with the specified word, either
  92          * followed by NUL or by whitespace. Returns a pointer to the
  93          * NUL or the first character after the whitespace. */
  94
  95         sl = strlen(s);
  96         wl = strlen(word);
  97
  98         if (sl < wl)
  99                 return NULL;
 100
 101         if (wl == 0)
 102                 return (char*) s;
 103
 104         if (memcmp(s, word, wl) != 0)
 105                 return NULL;
 106
 107         p = s + wl;
 108         if (*p == 0)
 109                 return (char*) p;
 110
 111         if (!strchr(WHITESPACE, *p))
 112                 return NULL;
 113
 114         p += strspn(p, WHITESPACE);
 115         return (char*) p;
 116 }
 117
 118 static size_t strcspn_escaped(const char *s, const char *reject) {
 119         bool escaped = false;
 120         int n;
 121
 122         for (n=0; s[n]; n++) {
 123                 if (escaped)
 124                         escaped = false;
 125                 else if (s[n] == '\\')
 126                         escaped = true;
 127                 else if (strchr(reject, s[n]))
 128                         break;
 129         }
 130
 131         /* if s ends in \, return index of previous char */
 132         return n - escaped;
 133 }
 134
 135 /* Split a string into words. */
 136 const char* split(const char **state, size_t *l, const char *separator, bool quoted) {
 137         const char *current;
 138
 139         current = *state;
 140
 141         if (!*current) {
 142                 assert(**state == '\0');
 143                 return NULL;
 144         }
 145
 146         current += strspn(current, separator);
 147         if (!*current) {
 148                 *state = current;
 149                 return NULL;
 150         }
 151
 152         if (quoted && strchr("\'\"", *current)) {
 153                 char quotechars[2] = {*current, '\0'};
 154
 155                 *l = strcspn_escaped(current + 1, quotechars);
 156                 if (current[*l + 1] == '\0' || current[*l + 1] != quotechars[0] ||
 157                     (current[*l + 2] && !strchr(separator, current[*l + 2]))) {
 158                         /* right quote missing or garbage at the end */
 159                         *state = current;
 160                         return NULL;
 161                 }
 162                 *state = current++ + *l + 2;
 163         } else if (quoted) {
 164                 *l = strcspn_escaped(current, separator);
 165                 if (current[*l] && !strchr(separator, current[*l])) {
 166                         /* unfinished escape */
 167                         *state = current;
 168                         return NULL;
 169                 }
 170                 *state = current + *l;
 171         } else {
 172                 *l = strcspn(current, separator);
 173                 *state = current + *l;
 174         }
 175
 176         return current;
 177 }
 178
 179 char *strnappend(const char *s, const char *suffix, size_t b) {
 180         size_t a;
 181         char *r;
 182
 183         if (!s && !suffix)
 184                 return strdup("");
 185
 186         if (!s)
 187                 return strndup(suffix, b);
 188
 189         if (!suffix)
 190                 return strdup(s);
 191
 192         assert(s);
 193         assert(suffix);
 194
 195         a = strlen(s);
 196         if (b > ((size_t) -1) - a)
 197                 return NULL;
 198
 199         r = new(char, a+b+1);
 200         if (!r)
 201                 return NULL;
 202
 203         memcpy(r, s, a);
 204         memcpy(r+a, suffix, b);
 205         r[a+b] = 0;
 206
 207         return r;
 208 }
 209
 210 char *strappend(const char *s, const char *suffix) {
 211         return strnappend(s, suffix, strlen_ptr(suffix));
 212 }
 213
 214 char *strjoin_real(const char *x, ...) {
 215         va_list ap;
 216         size_t l;
 217         char *r, *p;
 218
 219         va_start(ap, x);
 220
 221         if (x) {
 222                 l = strlen(x);
 223
 224                 for (;;) {
 225                         const char *t;
 226                         size_t n;
 227
 228                         t = va_arg(ap, const char *);
 229                         if (!t)
 230                                 break;
 231
 232                         n = strlen(t);
 233                         if (n > ((size_t) -1) - l) {
 234                                 va_end(ap);
 235                                 return NULL;
 236                         }
 237
 238                         l += n;
 239                 }
 240         } else
 241                 l = 0;
 242
 243         va_end(ap);
 244
 245         r = new(char, l+1);
 246         if (!r)
 247                 return NULL;
 248
 249         if (x) {
 250                 p = stpcpy(r, x);
 251
 252                 va_start(ap, x);
 253
 254                 for (;;) {
 255                         const char *t;
 256
 257                         t = va_arg(ap, const char *);
 258                         if (!t)
 259                                 break;
 260
 261                         p = stpcpy(p, t);
 262                 }
 263
 264                 va_end(ap);
 265         } else
 266                 r[0] = 0;
 267
 268         return r;
 269 }
 270
 271 char *strstrip(char *s) {
 272         if (!s)
 273                 return NULL;
 274
 275         /* Drops trailing whitespace. Modifies the string in place. Returns pointer to first non-space character */
 276
 277         return delete_trailing_chars(skip_leading_chars(s, WHITESPACE), WHITESPACE);
 278 }
 279
 280 char *delete_chars(char *s, const char *bad) {
 281         char *f, *t;
 282
 283         /* Drops all specified bad characters, regardless where in the string */
 284
 285         if (!s)
 286                 return NULL;
 287
 288         if (!bad)
 289                 bad = WHITESPACE;
 290
 291         for (f = s, t = s; *f; f++) {
 292                 if (strchr(bad, *f))
 293                         continue;
 294
 295                 *(t++) = *f;
 296         }
 297
 298         *t = 0;
 299
 300         return s;
 301 }
 302
 303 char *delete_trailing_chars(char *s, const char *bad) {
 304         char *p, *c = s;
 305
 306         /* Drops all specified bad characters, at the end of the string */
 307
 308         if (!s)
 309                 return NULL;
 310
 311         if (!bad)
 312                 bad = WHITESPACE;
 313
 314         for (p = s; *p; p++)
 315                 if (!strchr(bad, *p))
 316                         c = p + 1;
 317
 318         *c = 0;
 319
 320         return s;
 321 }
 322
 323 char *truncate_nl(char *s) {
 324         assert(s);
 325
 326         s[strcspn(s, NEWLINE)] = 0;
 327         return s;
 328 }
 329
 330 char ascii_tolower(char x) {
 331
 332         if (x >= 'A' && x <= 'Z')
 333                 return x - 'A' + 'a';
 334
 335         return x;
 336 }
 337
 338 char ascii_toupper(char x) {
 339
 340         if (x >= 'a' && x <= 'z')
 341                 return x - 'a' + 'A';
 342
 343         return x;
 344 }
 345
 346 char *ascii_strlower(char *t) {
 347         char *p;
 348
 349         assert(t);
 350
 351         for (p = t; *p; p++)
 352                 *p = ascii_tolower(*p);
 353
 354         return t;
 355 }
 356
 357 char *ascii_strupper(char *t) {
 358         char *p;
 359
 360         assert(t);
 361
 362         for (p = t; *p; p++)
 363                 *p = ascii_toupper(*p);
 364
 365         return t;
 366 }
 367
 368 char *ascii_strlower_n(char *t, size_t n) {
 369         size_t i;
 370
 371         if (n <= 0)
 372                 return t;
 373
 374         for (i = 0; i < n; i++)
 375                 t[i] = ascii_tolower(t[i]);
 376
 377         return t;
 378 }
 379
 380 int ascii_strcasecmp_n(const char *a, const char *b, size_t n) {
 381
 382         for (; n > 0; a++, b++, n--) {
 383                 int x, y;
 384
 385                 x = (int) (uint8_t) ascii_tolower(*a);
 386                 y = (int) (uint8_t) ascii_tolower(*b);
 387
 388                 if (x != y)
 389                         return x - y;
 390         }
 391
 392         return 0;
 393 }
 394
 395 int ascii_strcasecmp_nn(const char *a, size_t n, const char *b, size_t m) {
 396         int r;
 397
 398         r = ascii_strcasecmp_n(a, b, MIN(n, m));
 399         if (r != 0)
 400                 return r;
 401
 402         if (n < m)
 403                 return -1;
 404         else if (n > m)
 405                 return 1;
 406         else
 407                 return 0;
 408 }
 409
 410 bool chars_intersect(const char *a, const char *b) {
 411         const char *p;
 412
 413         /* Returns true if any of the chars in a are in b. */
 414         for (p = a; *p; p++)
 415                 if (strchr(b, *p))
 416                         return true;
 417
 418         return false;
 419 }
 420
 421 bool string_has_cc(const char *p, const char *ok) {
 422         const char *t;
 423
 424         assert(p);
 425
 426         /*
 427          * Check if a string contains control characters. If 'ok' is
 428          * non-NULL it may be a string containing additional CCs to be
 429          * considered OK.
 430          */
 431
 432         for (t = p; *t; t++) {
 433                 if (ok && strchr(ok, *t))
 434                         continue;
 435
 436                 if (*t > 0 && *t < ' ')
 437                         return true;
 438
 439                 if (*t == 127)
 440                         return true;
 441         }
 442
 443         return false;
 444 }
 445
 446 static int write_ellipsis(char *buf, bool unicode) {
 447         if (unicode || is_locale_utf8()) {
 448                 buf[0] = 0xe2; /* tri-dot ellipsis: … */
 449                 buf[1] = 0x80;
 450                 buf[2] = 0xa6;
 451         } else {
 452                 buf[0] = '.';
 453                 buf[1] = '.';
 454                 buf[2] = '.';
 455         }
 456
 457         return 3;
 458 }
 459
 460 static char *ascii_ellipsize_mem(const char *s, size_t old_length, size_t new_length, unsigned percent) {
 461         size_t x, need_space, suffix_len;
 462         char *t;
 463
 464         assert(s);
 465         assert(percent <= 100);
 466         assert(new_length != (size_t) -1);
 467
 468         if (old_length <= new_length)
 469                 return strndup(s, old_length);
 470
 471         /* Special case short ellipsations */
 472         switch (new_length) {
 473
 474         case 0:
 475                 return strdup("");
 476
 477         case 1:
 478                 if (is_locale_utf8())
 479                         return strdup("…");
 480                 else
 481                         return strdup(".");
 482
 483         case 2:
 484                 if (!is_locale_utf8())
 485                         return strdup("..");
 486
 487                 break;
 488
 489         default:
 490                 break;
 491         }
 492
 493         /* Calculate how much space the ellipsis will take up. If we are in UTF-8 mode we only need space for one
 494          * character ("…"), otherwise for three characters ("..."). Note that in both cases we need 3 bytes of storage,
 495          * either for the UTF-8 encoded character or for three ASCII characters. */
 496         need_space = is_locale_utf8() ? 1 : 3;
 497
 498         t = new(char, new_length+3);
 499         if (!t)
 500                 return NULL;
 501
 502         assert(new_length >= need_space);
 503
 504         x = ((new_length - need_space) * percent + 50) / 100;
 505         assert(x <= new_length - need_space);
 506
 507         memcpy(t, s, x);
 508         write_ellipsis(t + x, false);
 509         suffix_len = new_length - x - need_space;
 510         memcpy(t + x + 3, s + old_length - suffix_len, suffix_len);
 511         *(t + x + 3 + suffix_len) = '\0';
 512
 513         return t;
 514 }
 515
 516 char *ellipsize_mem(const char *s, size_t old_length, size_t new_length, unsigned percent) {
 517         size_t x, k, len, len2;
 518         const char *i, *j;
 519         char *e;
 520         int r;
 521
 522         /* Note that 'old_length' refers to bytes in the string, while 'new_length' refers to character cells taken up
 523          * on screen. This distinction doesn't matter for ASCII strings, but it does matter for non-ASCII UTF-8
 524          * strings.
 525          *
 526          * Ellipsation is done in a locale-dependent way:
 527          * 1. If the string passed in is fully ASCII and the current locale is not UTF-8, three dots are used ("...")
 528          * 2. Otherwise, a unicode ellipsis is used ("…")
 529          *
 530          * In other words: you'll get a unicode ellipsis as soon as either the string contains non-ASCII characters or
 531          * the current locale is UTF-8.
 532          */
 533
 534         assert(s);
 535         assert(percent <= 100);
 536
 537         if (new_length == (size_t) -1)
 538                 return strndup(s, old_length);
 539
 540         if (new_length == 0)
 541                 return strdup("");
 542
 543         /* If no multibyte characters use ascii_ellipsize_mem for speed */
 544         if (ascii_is_valid(s))
 545                 return ascii_ellipsize_mem(s, old_length, new_length, percent);
 546
 547         x = ((new_length - 1) * percent) / 100;
 548         assert(x <= new_length - 1);
 549
 550         k = 0;
 551         for (i = s; i < s + old_length; i = utf8_next_char(i)) {
 552                 char32_t c;
 553                 int w;
 554
 555                 r = utf8_encoded_to_unichar(i, &c);
 556                 if (r < 0)
 557                         return NULL;
 558
 559                 w = unichar_iswide(c) ? 2 : 1;
 560                 if (k + w <= x)
 561                         k += w;
 562                 else
 563                         break;
 564         }
 565
 566         for (j = s + old_length; j > i; ) {
 567                 char32_t c;
 568                 int w;
 569                 const char *jj;
 570
 571                 jj = utf8_prev_char(j);
 572                 r = utf8_encoded_to_unichar(jj, &c);
 573                 if (r < 0)
 574                         return NULL;
 575
 576                 w = unichar_iswide(c) ? 2 : 1;
 577                 if (k + w <= new_length) {
 578                         k += w;
 579                         j = jj;
 580                 } else
 581                         break;
 582         }
 583         assert(i <= j);
 584
 585         /* we don't actually need to ellipsize */
 586         if (i == j)
 587                 return memdup_suffix0(s, old_length);
 588
 589         /* make space for ellipsis, if possible */
 590         if (j < s + old_length)
 591                 j = utf8_next_char(j);
 592         else if (i > s)
 593                 i = utf8_prev_char(i);
 594
 595         len = i - s;
 596         len2 = s + old_length - j;
 597         e = new(char, len + 3 + len2 + 1);
 598         if (!e)
 599                 return NULL;
 600
 601         /*
 602         printf("old_length=%zu new_length=%zu x=%zu len=%u len2=%u k=%u\n",
 603                old_length, new_length, x, len, len2, k);
 604         */
 605
 606         memcpy(e, s, len);
 607         write_ellipsis(e + len, true);
 608         memcpy(e + len + 3, j, len2);
 609         *(e + len + 3 + len2) = '\0';
 610
 611         return e;
 612 }
 613
 614 char *cellescape(char *buf, size_t len, const char *s) {
 615         /* Escape and ellipsize s into buffer buf of size len. Only non-control ASCII
 616          * characters are copied as they are, everything else is escaped. The result
 617          * is different then if escaping and ellipsization was performed in two
 618          * separate steps, because each sequence is either stored in full or skipped.
 619          *
 620          * This function should be used for logging about strings which expected to
 621          * be plain ASCII in a safe way.
 622          *
 623          * An ellipsis will be used if s is too long. It was always placed at the
 624          * very end.
 625          */
 626
 627         size_t i = 0, last_char_width[4] = {}, k = 0, j;
 628
 629         assert(len > 0); /* at least a terminating NUL */
 630
 631         for (;;) {
 632                 char four[4];
 633                 int w;
 634
 635                 if (*s == 0) /* terminating NUL detected? then we are done! */
 636                         goto done;
 637
 638                 w = cescape_char(*s, four);
 639                 if (i + w + 1 > len) /* This character doesn't fit into the buffer anymore? In that case let's
 640                                       * ellipsize at the previous location */
 641                         break;
 642
 643                 /* OK, there was space, let's add this escaped character to the buffer */
 644                 memcpy(buf + i, four, w);
 645                 i += w;
 646
 647                 /* And remember its width in the ring buffer */
 648                 last_char_width[k] = w;
 649                 k = (k + 1) % 4;
 650
 651                 s++;
 652         }
 653
 654         /* Ellipsation is necessary. This means we might need to truncate the string again to make space for 4
 655          * characters ideally, but the buffer is shorter than that in the first place take what we can get */
 656         for (j = 0; j < ELEMENTSOF(last_char_width); j++) {
 657
 658                 if (i + 4 <= len) /* nice, we reached our space goal */
 659                         break;
 660
 661                 k = k == 0 ? 3 : k - 1;
 662                 if (last_char_width[k] == 0) /* bummer, we reached the beginning of the strings */
 663                         break;
 664
 665                 assert(i >= last_char_width[k]);
 666                 i -= last_char_width[k];
 667         }
 668
 669         if (i + 4 <= len) /* yay, enough space */
 670                 i += write_ellipsis(buf + i, false);
 671         else if (i + 3 <= len) { /* only space for ".." */
 672                 buf[i++] = '.';
 673                 buf[i++] = '.';
 674         } else if (i + 2 <= len) /* only space for a single "." */
 675                 buf[i++] = '.';
 676         else
 677                 assert(i + 1 <= len);
 678
 679  done:
 680         buf[i] = '\0';
 681         return buf;
 682 }
 683
 684 bool nulstr_contains(const char *nulstr, const char *needle) {
 685         const char *i;
 686
 687         if (!nulstr)
 688                 return false;
 689
 690         NULSTR_FOREACH(i, nulstr)
 691                 if (streq(i, needle))
 692                         return true;
 693
 694         return false;
 695 }
 696
 697 char* strshorten(char *s, size_t l) {
 698         assert(s);
 699
 700         if (strnlen(s, l+1) > l)
 701                 s[l] = 0;
 702
 703         return s;
 704 }
 705
 706 char *strreplace(const char *text, const char *old_string, const char *new_string) {
 707         size_t l, old_len, new_len, allocated = 0;
 708         char *t, *ret = NULL;
 709         const char *f;
 710
 711         assert(old_string);
 712         assert(new_string);
 713
 714         if (!text)
 715                 return NULL;
 716
 717         old_len = strlen(old_string);
 718         new_len = strlen(new_string);
 719
 720         l = strlen(text);
 721         if (!GREEDY_REALLOC(ret, allocated, l+1))
 722                 return NULL;
 723
 724         f = text;
 725         t = ret;
 726         while (*f) {
 727                 size_t d, nl;
 728
 729                 if (!startswith(f, old_string)) {
 730                         *(t++) = *(f++);
 731                         continue;
 732                 }
 733
 734                 d = t - ret;
 735                 nl = l - old_len + new_len;
 736
 737                 if (!GREEDY_REALLOC(ret, allocated, nl + 1))
 738                         return mfree(ret);
 739
 740                 l = nl;
 741                 t = ret + d;
 742
 743                 t = stpcpy(t, new_string);
 744                 f += old_len;
 745         }
 746
 747         *t = 0;
 748         return ret;
 749 }
 750
 751 static void advance_offsets(ssize_t diff, size_t offsets[2], size_t shift[2], size_t size) {
 752         if (!offsets)
 753                 return;
 754
 755         if ((size_t) diff < offsets[0])
 756                 shift[0] += size;
 757         if ((size_t) diff < offsets[1])
 758                 shift[1] += size;
 759 }
 760
 761 char *strip_tab_ansi(char **ibuf, size_t *_isz, size_t highlight[2]) {
 762         const char *i, *begin = NULL;
 763         enum {
 764                 STATE_OTHER,
 765                 STATE_ESCAPE,
 766                 STATE_CSI,
 767                 STATE_CSO,
 768         } state = STATE_OTHER;
 769         char *obuf = NULL;
 770         size_t osz = 0, isz, shift[2] = {};
 771         FILE *f;
 772
 773         assert(ibuf);
 774         assert(*ibuf);
 775
 776         /* This does three things:
 777          *
 778          * 1. Replaces TABs by 8 spaces
 779          * 2. Strips ANSI color sequences (a subset of CSI), i.e. ESC '[' … 'm' sequences
 780          * 3. Strips ANSI operating system sequences (CSO), i.e. ESC ']' … BEL sequences
 781          *
 782          * Everything else will be left as it is. In particular other ANSI sequences are left as they are, as are any
 783          * other special characters. Truncated ANSI sequences are left-as is too. This call is supposed to suppress the
 784          * most basic formatting noise, but nothing else.
 785          *
 786          * Why care for CSO sequences? Well, to undo what terminal_urlify() and friends generate. */
 787
 788         isz = _isz ? *_isz : strlen(*ibuf);
 789
 790         f = open_memstream(&obuf, &osz);
 791         if (!f)
 792                 return NULL;
 793
 794         /* Note we turn off internal locking on f for performance reasons.  It's safe to do so since we created f here
 795          * and it doesn't leave our scope. */
 796
 797         (void) __fsetlocking(f, FSETLOCKING_BYCALLER);
 798
 799         for (i = *ibuf; i < *ibuf + isz + 1; i++) {
 800
 801                 switch (state) {
 802
 803                 case STATE_OTHER:
 804                         if (i >= *ibuf + isz) /* EOT */
 805                                 break;
 806                         else if (*i == '\x1B')
 807                                 state = STATE_ESCAPE;
 808                         else if (*i == '\t') {
 809                                 fputs("        ", f);
 810                                 advance_offsets(i - *ibuf, highlight, shift, 7);
 811                         } else
 812                                 fputc(*i, f);
 813
 814                         break;
 815
 816                 case STATE_ESCAPE:
 817                         if (i >= *ibuf + isz) { /* EOT */
 818                                 fputc('\x1B', f);
 819                                 advance_offsets(i - *ibuf, highlight, shift, 1);
 820                                 break;
 821                         } else if (*i == '[') { /* ANSI CSI */
 822                                 state = STATE_CSI;
 823                                 begin = i + 1;
 824                         } else if (*i == ']') { /* ANSI CSO */
 825                                 state = STATE_CSO;
 826                                 begin = i + 1;
 827                         } else {
 828                                 fputc('\x1B', f);
 829                                 fputc(*i, f);
 830                                 advance_offsets(i - *ibuf, highlight, shift, 1);
 831                                 state = STATE_OTHER;
 832                         }
 833
 834                         break;
 835
 836                 case STATE_CSI:
 837
 838                         if (i >= *ibuf + isz || /* EOT … */
 839                             !strchr("01234567890;m", *i)) { /* … or invalid chars in sequence */
 840                                 fputc('\x1B', f);
 841                                 fputc('[', f);
 842                                 advance_offsets(i - *ibuf, highlight, shift, 2);
 843                                 state = STATE_OTHER;
 844                                 i = begin-1;
 845                         } else if (*i == 'm')
 846                                 state = STATE_OTHER;
 847
 848                         break;
 849
 850                 case STATE_CSO:
 851
 852                         if (i >= *ibuf + isz || /* EOT … */
 853                             (*i != '\a' && (uint8_t) *i < 32U) || (uint8_t) *i > 126U) { /* … or invalid chars in sequence */
 854                                 fputc('\x1B', f);
 855                                 fputc(']', f);
 856                                 advance_offsets(i - *ibuf, highlight, shift, 2);
 857                                 state = STATE_OTHER;
 858                                 i = begin-1;
 859                         } else if (*i == '\a')
 860                                 state = STATE_OTHER;
 861
 862                         break;
 863                 }
 864         }
 865
 866         if (fflush_and_check(f) < 0) {
 867                 fclose(f);
 868                 return mfree(obuf);
 869         }
 870
 871         fclose(f);
 872
 873         free(*ibuf);
 874         *ibuf = obuf;
 875
 876         if (_isz)
 877                 *_isz = osz;
 878
 879         if (highlight) {
 880                 highlight[0] += shift[0];
 881                 highlight[1] += shift[1];
 882         }
 883
 884         return obuf;
 885 }
 886
 887 char *strextend_with_separator(char **x, const char *separator, ...) {
 888         bool need_separator;
 889         size_t f, l, l_separator;
 890         char *r, *p;
 891         va_list ap;
 892
 893         assert(x);
 894
 895         l = f = strlen_ptr(*x);
 896
 897         need_separator = !isempty(*x);
 898         l_separator = strlen_ptr(separator);
 899
 900         va_start(ap, separator);
 901         for (;;) {
 902                 const char *t;
 903                 size_t n;
 904
 905                 t = va_arg(ap, const char *);
 906                 if (!t)
 907                         break;
 908
 909                 n = strlen(t);
 910
 911                 if (need_separator)
 912                         n += l_separator;
 913
 914                 if (n > ((size_t) -1) - l) {
 915                         va_end(ap);
 916                         return NULL;
 917                 }
 918
 919                 l += n;
 920                 need_separator = true;
 921         }
 922         va_end(ap);
 923
 924         need_separator = !isempty(*x);
 925
 926         r = realloc(*x, l+1);
 927         if (!r)
 928                 return NULL;
 929
 930         p = r + f;
 931
 932         va_start(ap, separator);
 933         for (;;) {
 934                 const char *t;
 935
 936                 t = va_arg(ap, const char *);
 937                 if (!t)
 938                         break;
 939
 940                 if (need_separator && separator)
 941                         p = stpcpy(p, separator);
 942
 943                 p = stpcpy(p, t);
 944
 945                 need_separator = true;
 946         }
 947         va_end(ap);
 948
 949         assert(p == r + l);
 950
 951         *p = 0;
 952         *x = r;
 953
 954         return r + l;
 955 }
 956
 957 char *strrep(const char *s, unsigned n) {
 958         size_t l;
 959         char *r, *p;
 960         unsigned i;
 961
 962         assert(s);
 963
 964         l = strlen(s);
 965         p = r = malloc(l * n + 1);
 966         if (!r)
 967                 return NULL;
 968
 969         for (i = 0; i < n; i++)
 970                 p = stpcpy(p, s);
 971
 972         *p = 0;
 973         return r;
 974 }
 975
 976 int split_pair(const char *s, const char *sep, char **l, char **r) {
 977         char *x, *a, *b;
 978
 979         assert(s);
 980         assert(sep);
 981         assert(l);
 982         assert(r);
 983
 984         if (isempty(sep))
 985                 return -EINVAL;
 986
 987         x = strstr(s, sep);
 988         if (!x)
 989                 return -EINVAL;
 990
 991         a = strndup(s, x - s);
 992         if (!a)
 993                 return -ENOMEM;
 994
 995         b = strdup(x + strlen(sep));
 996         if (!b) {
 997                 free(a);
 998                 return -ENOMEM;
 999         }
1000
1001         *l = a;
1002         *r = b;
1003
1004         return 0;
1005 }
1006
1007 int free_and_strdup(char **p, const char *s) {
1008         char *t;
1009
1010         assert(p);
1011
1012         /* Replaces a string pointer with an strdup()ed new string,
1013          * possibly freeing the old one. */
1014
1015         if (streq_ptr(*p, s))
1016                 return 0;
1017
1018         if (s) {
1019                 t = strdup(s);
1020                 if (!t)
1021                         return -ENOMEM;
1022         } else
1023                 t = NULL;
1024
1025         free(*p);
1026         *p = t;
1027
1028         return 1;
1029 }
1030
1031 #if !HAVE_EXPLICIT_BZERO
1032 /*
1033  * Pointer to memset is volatile so that compiler must de-reference
1034  * the pointer and can't assume that it points to any function in
1035  * particular (such as memset, which it then might further "optimize")
1036  * This approach is inspired by openssl's crypto/mem_clr.c.
1037  */
1038 typedef void *(*memset_t)(void *,int,size_t);
1039
1040 static volatile memset_t memset_func = memset;
1041
1042 void explicit_bzero(void *p, size_t l) {
1043         memset_func(p, '\0', l);
1044 }
1045 #endif
1046
1047 char* string_erase(char *x) {
1048         if (!x)
1049                 return NULL;
1050
1051         /* A delicious drop of snake-oil! To be called on memory where
1052          * we stored passphrases or so, after we used them. */
1053         explicit_bzero(x, strlen(x));
1054         return x;
1055 }
1056
1057 char *string_free_erase(char *s) {
1058         return mfree(string_erase(s));
1059 }
1060
1061 bool string_is_safe(const char *p) {
1062         const char *t;
1063
1064         if (!p)
1065                 return false;
1066
1067         for (t = p; *t; t++) {
1068                 if (*t > 0 && *t < ' ') /* no control characters */
1069                         return false;
1070
1071                 if (strchr(QUOTES "\\\x7f", *t))
1072                         return false;
1073         }
1074
1075         return true;
1076 }