utf8.c

   1 #define DISABLE_SIGN_COMPARE_WARNINGS
   2
   3 #include "git-compat-util.h"
   4 #include "strbuf.h"
   5 #include "utf8.h"
   6
   7 /* This code is originally from https://www.cl.cam.ac.uk/~mgk25/ucs/ */
   8
   9 static const char utf16_be_bom[] = {'\xFE', '\xFF'};
  10 static const char utf16_le_bom[] = {'\xFF', '\xFE'};
  11 static const char utf32_be_bom[] = {'\0', '\0', '\xFE', '\xFF'};
  12 static const char utf32_le_bom[] = {'\xFF', '\xFE', '\0', '\0'};
  13
  14 struct interval {
  15         ucs_char_t first;
  16         ucs_char_t last;
  17 };
  18
  19 size_t display_mode_esc_sequence_len(const char *s)
  20 {
  21         const char *p = s;
  22         if (*p++ != '\033')
  23                 return 0;
  24         if (*p++ != '[')
  25                 return 0;
  26         while (isdigit(*p) || *p == ';')
  27                 p++;
  28         if (*p++ != 'm')
  29                 return 0;
  30         return p - s;
  31 }
  32
  33 /* auxiliary function for binary search in interval table */
  34 static int bisearch(ucs_char_t ucs, const struct interval *table, int max)
  35 {
  36         int min = 0;
  37         int mid;
  38
  39         if (ucs < table[0].first || ucs > table[max].last)
  40                 return 0;
  41         while (max >= min) {
  42                 mid = min + (max - min) / 2;
  43                 if (ucs > table[mid].last)
  44                         min = mid + 1;
  45                 else if (ucs < table[mid].first)
  46                         max = mid - 1;
  47                 else
  48                         return 1;
  49         }
  50
  51         return 0;
  52 }
  53
  54 /* The following two functions define the column width of an ISO 10646
  55  * character as follows:
  56  *
  57  *    - The null character (U+0000) has a column width of 0.
  58  *
  59  *    - Other C0/C1 control characters and DEL will lead to a return
  60  *      value of -1.
  61  *
  62  *    - Non-spacing and enclosing combining characters (general
  63  *      category code Mn or Me in the Unicode database) have a
  64  *      column width of 0.
  65  *
  66  *    - SOFT HYPHEN (U+00AD) has a column width of 1.
  67  *
  68  *    - Other format characters (general category code Cf in the Unicode
  69  *      database) and ZERO WIDTH SPACE (U+200B) have a column width of 0.
  70  *
  71  *    - Hangul Jamo medial vowels and final consonants (U+1160-U+11FF)
  72  *      have a column width of 0.
  73  *
  74  *    - Spacing characters in the East Asian Wide (W) or East Asian
  75  *      Full-width (F) category as defined in Unicode Technical
  76  *      Report #11 have a column width of 2.
  77  *
  78  *    - All remaining characters (including all printable
  79  *      ISO 8859-1 and WGL4 characters, Unicode control characters,
  80  *      etc.) have a column width of 1.
  81  *
  82  * This implementation assumes that ucs_char_t characters are encoded
  83  * in ISO 10646.
  84  */
  85
  86 static int git_wcwidth(ucs_char_t ch)
  87 {
  88         /*
  89          * Sorted list of non-overlapping intervals of non-spacing characters,
  90          */
  91 #include "unicode-width.h"
  92
  93         /* test for 8-bit control characters */
  94         if (ch == 0)
  95                 return 0;
  96         if (ch < 32 || (ch >= 0x7f && ch < 0xa0))
  97                 return -1;
  98
  99         /* binary search in table of non-spacing characters */
 100         if (bisearch(ch, zero_width, ARRAY_SIZE(zero_width) - 1))
 101                 return 0;
 102
 103         /* binary search in table of double width characters */
 104         if (bisearch(ch, double_width, ARRAY_SIZE(double_width) - 1))
 105                 return 2;
 106
 107         return 1;
 108 }
 109
 110 /*
 111  * Pick one ucs character starting from the location *start points at,
 112  * and return it, while updating the *start pointer to point at the
 113  * end of that character.  When remainder_p is not NULL, the location
 114  * holds the number of bytes remaining in the string that we are allowed
 115  * to pick from.  Otherwise we are allowed to pick up to the NUL that
 116  * would eventually appear in the string.  *remainder_p is also reduced
 117  * by the number of bytes we have consumed.
 118  *
 119  * If the string was not a valid UTF-8, *start pointer is set to NULL
 120  * and the return value is undefined.
 121  */
 122 static ucs_char_t pick_one_utf8_char(const char **start, size_t *remainder_p)
 123 {
 124         unsigned char *s = (unsigned char *)*start;
 125         ucs_char_t ch;
 126         size_t remainder, incr;
 127
 128         /*
 129          * A caller that assumes NUL terminated text can choose
 130          * not to bother with the remainder length.  We will
 131          * stop at the first NUL.
 132          */
 133         remainder = (remainder_p ? *remainder_p : 999);
 134
 135         if (remainder < 1) {
 136                 goto invalid;
 137         } else if (*s < 0x80) {
 138                 /* 0xxxxxxx */
 139                 ch = *s;
 140                 incr = 1;
 141         } else if ((s[0] & 0xe0) == 0xc0) {
 142                 /* 110XXXXx 10xxxxxx */
 143                 if (remainder < 2 ||
 144                     (s[1] & 0xc0) != 0x80 ||
 145                     (s[0] & 0xfe) == 0xc0)
 146                         goto invalid;
 147                 ch = ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
 148                 incr = 2;
 149         } else if ((s[0] & 0xf0) == 0xe0) {
 150                 /* 1110XXXX 10Xxxxxx 10xxxxxx */
 151                 if (remainder < 3 ||
 152                     (s[1] & 0xc0) != 0x80 ||
 153                     (s[2] & 0xc0) != 0x80 ||
 154                     /* overlong? */
 155                     (s[0] == 0xe0 && (s[1] & 0xe0) == 0x80) ||
 156                     /* surrogate? */
 157                     (s[0] == 0xed && (s[1] & 0xe0) == 0xa0) ||
 158                     /* U+FFFE or U+FFFF? */
 159                     (s[0] == 0xef && s[1] == 0xbf &&
 160                      (s[2] & 0xfe) == 0xbe))
 161                         goto invalid;
 162                 ch = ((s[0] & 0x0f) << 12) |
 163                         ((s[1] & 0x3f) << 6) | (s[2] & 0x3f);
 164                 incr = 3;
 165         } else if ((s[0] & 0xf8) == 0xf0) {
 166                 /* 11110XXX 10XXxxxx 10xxxxxx 10xxxxxx */
 167                 if (remainder < 4 ||
 168                     (s[1] & 0xc0) != 0x80 ||
 169                     (s[2] & 0xc0) != 0x80 ||
 170                     (s[3] & 0xc0) != 0x80 ||
 171                     /* overlong? */
 172                     (s[0] == 0xf0 && (s[1] & 0xf0) == 0x80) ||
 173                     /* > U+10FFFF? */
 174                     (s[0] == 0xf4 && s[1] > 0x8f) || s[0] > 0xf4)
 175                         goto invalid;
 176                 ch = ((s[0] & 0x07) << 18) | ((s[1] & 0x3f) << 12) |
 177                         ((s[2] & 0x3f) << 6) | (s[3] & 0x3f);
 178                 incr = 4;
 179         } else {
 180 invalid:
 181                 *start = NULL;
 182                 return 0;
 183         }
 184
 185         *start += incr;
 186         if (remainder_p)
 187                 *remainder_p = remainder - incr;
 188         return ch;
 189 }
 190
 191 /*
 192  * This function returns the number of columns occupied by the character
 193  * pointed to by the variable start. The pointer is updated to point at
 194  * the next character. When remainder_p is not NULL, it points at the
 195  * location that stores the number of remaining bytes we can use to pick
 196  * a character (see pick_one_utf8_char() above).
 197  */
 198 int utf8_width(const char **start, size_t *remainder_p)
 199 {
 200         ucs_char_t ch = pick_one_utf8_char(start, remainder_p);
 201         if (!*start)
 202                 return 0;
 203         return git_wcwidth(ch);
 204 }
 205
 206 /*
 207  * Returns the total number of columns required by a null-terminated
 208  * string, assuming that the string is utf8.  Returns strlen() instead
 209  * if the string does not look like a valid utf8 string.
 210  */
 211 int utf8_strnwidth(const char *string, size_t len, int skip_ansi)
 212 {
 213         const char *orig = string;
 214         size_t width = 0;
 215
 216         while (string && string < orig + len) {
 217                 int glyph_width;
 218                 size_t skip;
 219
 220                 while (skip_ansi &&
 221                        (skip = display_mode_esc_sequence_len(string)) != 0)
 222                         string += skip;
 223
 224                 glyph_width = utf8_width(&string, NULL);
 225                 if (glyph_width > 0)
 226                         width += glyph_width;
 227         }
 228
 229         /*
 230          * TODO: fix the interface of this function and `utf8_strwidth()` to
 231          * return `size_t` instead of `int`.
 232          */
 233         return cast_size_t_to_int(string ? width : len);
 234 }
 235
 236 int utf8_strwidth(const char *string)
 237 {
 238         return utf8_strnwidth(string, strlen(string), 0);
 239 }
 240
 241 int is_utf8(const char *text)
 242 {
 243         while (*text) {
 244                 if (*text == '\n' || *text == '\t' || *text == '\r') {
 245                         text++;
 246                         continue;
 247                 }
 248                 utf8_width(&text, NULL);
 249                 if (!text)
 250                         return 0;
 251         }
 252         return 1;
 253 }
 254
 255 static void strbuf_add_indented_text(struct strbuf *buf, const char *text,
 256                                      int indent, int indent2)
 257 {
 258         if (indent < 0)
 259                 indent = 0;
 260         while (*text) {
 261                 const char *eol = strchrnul(text, '\n');
 262                 if (*eol == '\n')
 263                         eol++;
 264                 strbuf_addchars(buf, ' ', indent);
 265                 strbuf_add(buf, text, eol - text);
 266                 text = eol;
 267                 indent = indent2;
 268         }
 269 }
 270
 271 /*
 272  * Wrap the text, if necessary. The variable indent is the indent for the
 273  * first line, indent2 is the indent for all other lines.
 274  * If indent is negative, assume that already -indent columns have been
 275  * consumed (and no extra indent is necessary for the first line).
 276  */
 277 void strbuf_add_wrapped_text(struct strbuf *buf,
 278                 const char *text, int indent1, int indent2, int width)
 279 {
 280         int indent, w, assume_utf8 = 1;
 281         const char *bol, *space, *start = text;
 282         size_t orig_len = buf->len;
 283
 284         if (width <= 0) {
 285                 strbuf_add_indented_text(buf, text, indent1, indent2);
 286                 return;
 287         }
 288
 289 retry:
 290         bol = text;
 291         w = indent = indent1;
 292         space = NULL;
 293         if (indent < 0) {
 294                 w = -indent;
 295                 space = text;
 296         }
 297
 298         for (;;) {
 299                 char c;
 300                 size_t skip;
 301
 302                 while ((skip = display_mode_esc_sequence_len(text)))
 303                         text += skip;
 304
 305                 c = *text;
 306                 if (!c || isspace(c)) {
 307                         if (w <= width || !space) {
 308                                 const char *start = bol;
 309                                 if (!c && text == start)
 310                                         return;
 311                                 if (space)
 312                                         start = space;
 313                                 else
 314                                         strbuf_addchars(buf, ' ', indent);
 315                                 strbuf_add(buf, start, text - start);
 316                                 if (!c)
 317                                         return;
 318                                 space = text;
 319                                 if (c == '\t')
 320                                         w |= 0x07;
 321                                 else if (c == '\n') {
 322                                         space++;
 323                                         if (*space == '\n') {
 324                                                 strbuf_addch(buf, '\n');
 325                                                 goto new_line;
 326                                         }
 327                                         else if (!isalnum(*space))
 328                                                 goto new_line;
 329                                         else
 330                                                 strbuf_addch(buf, ' ');
 331                                 }
 332                                 w++;
 333                                 text++;
 334                         }
 335                         else {
 336 new_line:
 337                                 strbuf_addch(buf, '\n');
 338                                 text = bol = space + isspace(*space);
 339                                 space = NULL;
 340                                 w = indent = indent2;
 341                         }
 342                         continue;
 343                 }
 344                 if (assume_utf8) {
 345                         w += utf8_width(&text, NULL);
 346                         if (!text) {
 347                                 assume_utf8 = 0;
 348                                 text = start;
 349                                 strbuf_setlen(buf, orig_len);
 350                                 goto retry;
 351                         }
 352                 } else {
 353                         w++;
 354                         text++;
 355                 }
 356         }
 357 }
 358
 359 void strbuf_add_wrapped_bytes(struct strbuf *buf, const char *data, int len,
 360                              int indent, int indent2, int width)
 361 {
 362         char *tmp = xstrndup(data, len);
 363         strbuf_add_wrapped_text(buf, tmp, indent, indent2, width);
 364         free(tmp);
 365 }
 366
 367 void strbuf_utf8_replace(struct strbuf *sb_src, int pos, int width,
 368                          const char *subst)
 369 {
 370         const char *src = sb_src->buf, *end = sb_src->buf + sb_src->len;
 371         struct strbuf dst;
 372         int w = 0;
 373
 374         strbuf_init(&dst, sb_src->len);
 375
 376         while (src < end) {
 377                 const char *old;
 378                 int glyph_width;
 379                 size_t n;
 380
 381                 while ((n = display_mode_esc_sequence_len(src))) {
 382                         strbuf_add(&dst, src, n);
 383                         src += n;
 384                 }
 385
 386                 if (src >= end)
 387                         break;
 388
 389                 old = src;
 390                 glyph_width = utf8_width((const char**)&src, NULL);
 391                 if (!src) /* broken utf-8, do nothing */
 392                         goto out;
 393
 394                 /*
 395                  * In case we see a control character we copy it into the
 396                  * buffer, but don't add it to the width.
 397                  */
 398                 if (glyph_width < 0)
 399                         glyph_width = 0;
 400
 401                 if (glyph_width && w >= pos && w < pos + width) {
 402                         if (subst) {
 403                                 strbuf_addstr(&dst, subst);
 404                                 subst = NULL;
 405                         }
 406                 } else {
 407                         strbuf_add(&dst, old, src - old);
 408                 }
 409
 410                 w += glyph_width;
 411         }
 412
 413         strbuf_swap(sb_src, &dst);
 414 out:
 415         strbuf_release(&dst);
 416 }
 417
 418 /*
 419  * Returns true (1) if the src encoding name matches the dst encoding
 420  * name directly or one of its alternative names. E.g. UTF-16BE is the
 421  * same as UTF16BE.
 422  */
 423 static int same_utf_encoding(const char *src, const char *dst)
 424 {
 425         if (skip_iprefix(src, "utf", &src) && skip_iprefix(dst, "utf", &dst)) {
 426                 skip_prefix(src, "-", &src);
 427                 skip_prefix(dst, "-", &dst);
 428                 return !strcasecmp(src, dst);
 429         }
 430         return 0;
 431 }
 432
 433 int is_encoding_utf8(const char *name)
 434 {
 435         if (!name)
 436                 return 1;
 437         if (same_utf_encoding("utf-8", name))
 438                 return 1;
 439         return 0;
 440 }
 441
 442 int same_encoding(const char *src, const char *dst)
 443 {
 444         static const char utf8[] = "UTF-8";
 445
 446         if (!src)
 447                 src = utf8;
 448         if (!dst)
 449                 dst = utf8;
 450         if (same_utf_encoding(src, dst))
 451                 return 1;
 452         return !strcasecmp(src, dst);
 453 }
 454
 455 /*
 456  * Wrapper for fprintf and returns the total number of columns required
 457  * for the printed string, assuming that the string is utf8.
 458  */
 459 int utf8_fprintf(FILE *stream, const char *format, ...)
 460 {
 461         struct strbuf buf = STRBUF_INIT;
 462         va_list arg;
 463         int columns;
 464
 465         va_start(arg, format);
 466         strbuf_vaddf(&buf, format, arg);
 467         va_end(arg);
 468
 469         columns = fputs(buf.buf, stream);
 470         if (0 <= columns) /* keep the error from the I/O */
 471                 columns = utf8_strwidth(buf.buf);
 472         strbuf_release(&buf);
 473         return columns;
 474 }
 475
 476 /*
 477  * Given a buffer and its encoding, return it re-encoded
 478  * with iconv.  If the conversion fails, returns NULL.
 479  */
 480 #ifndef NO_ICONV
 481 #if defined(OLD_ICONV) || (defined(__sun__) && !defined(_XPG6))
 482         typedef const char * iconv_ibp;
 483 #else
 484         typedef char * iconv_ibp;
 485 #endif
 486 char *reencode_string_iconv(const char *in, size_t insz, iconv_t conv,
 487                             size_t bom_len, size_t *outsz_p)
 488 {
 489         size_t outsz, outalloc;
 490         char *out, *outpos;
 491         iconv_ibp cp;
 492
 493         outsz = insz;
 494         outalloc = st_add(outsz, 1 + bom_len); /* for terminating NUL */
 495         out = xmalloc(outalloc);
 496         outpos = out + bom_len;
 497         cp = (iconv_ibp)in;
 498
 499         while (1) {
 500                 size_t cnt = iconv(conv, &cp, &insz, &outpos, &outsz);
 501
 502                 if (cnt == (size_t) -1) {
 503                         size_t sofar;
 504                         if (errno != E2BIG) {
 505                                 free(out);
 506                                 return NULL;
 507                         }
 508                         /* insz has remaining number of bytes.
 509                          * since we started outsz the same as insz,
 510                          * it is likely that insz is not enough for
 511                          * converting the rest.
 512                          */
 513                         sofar = outpos - out;
 514                         outalloc = st_add3(sofar, st_mult(insz, 2), 32);
 515                         out = xrealloc(out, outalloc);
 516                         outpos = out + sofar;
 517                         outsz = outalloc - sofar - 1;
 518                 }
 519                 else {
 520                         *outpos = '\0';
 521                         if (outsz_p)
 522                                 *outsz_p = outpos - out;
 523                         break;
 524                 }
 525         }
 526         return out;
 527 }
 528
 529 static const char *fallback_encoding(const char *name)
 530 {
 531         /*
 532          * Some platforms do not have the variously spelled variants of
 533          * UTF-8, so let's fall back to trying the most official
 534          * spelling. We do so only as a fallback in case the platform
 535          * does understand the user's spelling, but not our official
 536          * one.
 537          */
 538         if (is_encoding_utf8(name))
 539                 return "UTF-8";
 540
 541         /*
 542          * Even though latin-1 is still seen in e-mail
 543          * headers, some platforms only install ISO-8859-1.
 544          */
 545         if (!strcasecmp(name, "latin-1"))
 546                 return "ISO-8859-1";
 547
 548         return name;
 549 }
 550
 551 char *reencode_string_len(const char *in, size_t insz,
 552                           const char *out_encoding, const char *in_encoding,
 553                           size_t *outsz)
 554 {
 555         iconv_t conv;
 556         char *out;
 557         const char *bom_str = NULL;
 558         size_t bom_len = 0;
 559
 560         if (!in_encoding)
 561                 return NULL;
 562
 563         /* UTF-16LE-BOM is the same as UTF-16 for reading */
 564         if (same_utf_encoding("UTF-16LE-BOM", in_encoding))
 565                 in_encoding = "UTF-16";
 566
 567         /*
 568          * For writing, UTF-16 iconv typically creates "UTF-16BE-BOM"
 569          * Some users under Windows want the little endian version
 570          *
 571          * We handle UTF-16 and UTF-32 ourselves only if the platform does not
 572          * provide a BOM (which we require), since we want to match the behavior
 573          * of the system tools and libc as much as possible.
 574          */
 575         if (same_utf_encoding("UTF-16LE-BOM", out_encoding)) {
 576                 bom_str = utf16_le_bom;
 577                 bom_len = sizeof(utf16_le_bom);
 578                 out_encoding = "UTF-16LE";
 579         } else if (same_utf_encoding("UTF-16BE-BOM", out_encoding)) {
 580                 bom_str = utf16_be_bom;
 581                 bom_len = sizeof(utf16_be_bom);
 582                 out_encoding = "UTF-16BE";
 583 #ifdef ICONV_OMITS_BOM
 584         } else if (same_utf_encoding("UTF-16", out_encoding)) {
 585                 bom_str = utf16_be_bom;
 586                 bom_len = sizeof(utf16_be_bom);
 587                 out_encoding = "UTF-16BE";
 588         } else if (same_utf_encoding("UTF-32", out_encoding)) {
 589                 bom_str = utf32_be_bom;
 590                 bom_len = sizeof(utf32_be_bom);
 591                 out_encoding = "UTF-32BE";
 592 #endif
 593         }
 594
 595         conv = iconv_open(out_encoding, in_encoding);
 596         if (conv == (iconv_t) -1) {
 597                 in_encoding = fallback_encoding(in_encoding);
 598                 out_encoding = fallback_encoding(out_encoding);
 599
 600                 conv = iconv_open(out_encoding, in_encoding);
 601                 if (conv == (iconv_t) -1)
 602                         return NULL;
 603         }
 604         out = reencode_string_iconv(in, insz, conv, bom_len, outsz);
 605         iconv_close(conv);
 606         if (out && bom_str && bom_len)
 607                 memcpy(out, bom_str, bom_len);
 608         return out;
 609 }
 610 #endif
 611
 612 static int has_bom_prefix(const char *data, size_t len,
 613                           const char *bom, size_t bom_len)
 614 {
 615         return data && bom && (len >= bom_len) && !memcmp(data, bom, bom_len);
 616 }
 617
 618 int has_prohibited_utf_bom(const char *enc, const char *data, size_t len)
 619 {
 620         return (
 621           (same_utf_encoding("UTF-16BE", enc) ||
 622            same_utf_encoding("UTF-16LE", enc)) &&
 623           (has_bom_prefix(data, len, utf16_be_bom, sizeof(utf16_be_bom)) ||
 624            has_bom_prefix(data, len, utf16_le_bom, sizeof(utf16_le_bom)))
 625         ) || (
 626           (same_utf_encoding("UTF-32BE",  enc) ||
 627            same_utf_encoding("UTF-32LE", enc)) &&
 628           (has_bom_prefix(data, len, utf32_be_bom, sizeof(utf32_be_bom)) ||
 629            has_bom_prefix(data, len, utf32_le_bom, sizeof(utf32_le_bom)))
 630         );
 631 }
 632
 633 int is_missing_required_utf_bom(const char *enc, const char *data, size_t len)
 634 {
 635         return (
 636            (same_utf_encoding(enc, "UTF-16")) &&
 637            !(has_bom_prefix(data, len, utf16_be_bom, sizeof(utf16_be_bom)) ||
 638              has_bom_prefix(data, len, utf16_le_bom, sizeof(utf16_le_bom)))
 639         ) || (
 640            (same_utf_encoding(enc, "UTF-32")) &&
 641            !(has_bom_prefix(data, len, utf32_be_bom, sizeof(utf32_be_bom)) ||
 642              has_bom_prefix(data, len, utf32_le_bom, sizeof(utf32_le_bom)))
 643         );
 644 }
 645
 646 /*
 647  * Returns first character length in bytes for multi-byte `text` according to
 648  * `encoding`.
 649  *
 650  * - The `text` pointer is updated to point at the next character.
 651  * - When `remainder_p` is not NULL, on entry `*remainder_p` is how much bytes
 652  *   we can consume from text, and on exit `*remainder_p` is reduced by returned
 653  *   character length. Otherwise `text` is treated as limited by NUL.
 654  */
 655 int mbs_chrlen(const char **text, size_t *remainder_p, const char *encoding)
 656 {
 657         int chrlen;
 658         const char *p = *text;
 659         size_t r = (remainder_p ? *remainder_p : SIZE_MAX);
 660
 661         if (r < 1)
 662                 return 0;
 663
 664         if (is_encoding_utf8(encoding)) {
 665                 pick_one_utf8_char(&p, &r);
 666
 667                 chrlen = p ? (p - *text)
 668                            : 1 /* not valid UTF-8 -> raw byte sequence */;
 669         }
 670         else {
 671                 /*
 672                  * TODO use iconv to decode one char and obtain its chrlen
 673                  * for now, let's treat encodings != UTF-8 as one-byte
 674                  */
 675                 chrlen = 1;
 676         }
 677
 678         *text += chrlen;
 679         if (remainder_p)
 680                 *remainder_p -= chrlen;
 681
 682         return chrlen;
 683 }
 684
 685 /*
 686  * Pick the next char from the stream, ignoring codepoints an HFS+ would.
 687  * Note that this is _not_ complete by any means. It's just enough
 688  * to make is_hfs_dotgit() work, and should not be used otherwise.
 689  */
 690 static ucs_char_t next_hfs_char(const char **in)
 691 {
 692         while (1) {
 693                 ucs_char_t out = pick_one_utf8_char(in, NULL);
 694                 /*
 695                  * check for malformed utf8. Technically this
 696                  * gets converted to a percent-sequence, but
 697                  * returning 0 is good enough for is_hfs_dotgit
 698                  * to realize it cannot be .git
 699                  */
 700                 if (!*in)
 701                         return 0;
 702
 703                 /* these code points are ignored completely */
 704                 switch (out) {
 705                 case 0x200c: /* ZERO WIDTH NON-JOINER */
 706                 case 0x200d: /* ZERO WIDTH JOINER */
 707                 case 0x200e: /* LEFT-TO-RIGHT MARK */
 708                 case 0x200f: /* RIGHT-TO-LEFT MARK */
 709                 case 0x202a: /* LEFT-TO-RIGHT EMBEDDING */
 710                 case 0x202b: /* RIGHT-TO-LEFT EMBEDDING */
 711                 case 0x202c: /* POP DIRECTIONAL FORMATTING */
 712                 case 0x202d: /* LEFT-TO-RIGHT OVERRIDE */
 713                 case 0x202e: /* RIGHT-TO-LEFT OVERRIDE */
 714                 case 0x206a: /* INHIBIT SYMMETRIC SWAPPING */
 715                 case 0x206b: /* ACTIVATE SYMMETRIC SWAPPING */
 716                 case 0x206c: /* INHIBIT ARABIC FORM SHAPING */
 717                 case 0x206d: /* ACTIVATE ARABIC FORM SHAPING */
 718                 case 0x206e: /* NATIONAL DIGIT SHAPES */
 719                 case 0x206f: /* NOMINAL DIGIT SHAPES */
 720                 case 0xfeff: /* ZERO WIDTH NO-BREAK SPACE */
 721                         continue;
 722                 }
 723
 724                 return out;
 725         }
 726 }
 727
 728 static int is_hfs_dot_generic(const char *path,
 729                               const char *needle, size_t needle_len)
 730 {
 731         ucs_char_t c;
 732
 733         c = next_hfs_char(&path);
 734         if (c != '.')
 735                 return 0;
 736
 737         /*
 738          * there's a great deal of other case-folding that occurs
 739          * in HFS+, but this is enough to catch our fairly vanilla
 740          * hard-coded needles.
 741          */
 742         for (; needle_len > 0; needle++, needle_len--) {
 743                 c = next_hfs_char(&path);
 744
 745                 /*
 746                  * We know our needles contain only ASCII, so we clamp here to
 747                  * make the results of tolower() sane.
 748                  */
 749                 if (c > 127)
 750                         return 0;
 751                 if (tolower(c) != *needle)
 752                         return 0;
 753         }
 754
 755         c = next_hfs_char(&path);
 756         if (c && !is_dir_sep(c))
 757                 return 0;
 758
 759         return 1;
 760 }
 761
 762 /*
 763  * Inline wrapper to make sure the compiler resolves strlen() on literals at
 764  * compile time.
 765  */
 766 static inline int is_hfs_dot_str(const char *path, const char *needle)
 767 {
 768         return is_hfs_dot_generic(path, needle, strlen(needle));
 769 }
 770
 771 int is_hfs_dotgit(const char *path)
 772 {
 773         return is_hfs_dot_str(path, "git");
 774 }
 775
 776 int is_hfs_dotgitmodules(const char *path)
 777 {
 778         return is_hfs_dot_str(path, "gitmodules");
 779 }
 780
 781 int is_hfs_dotgitignore(const char *path)
 782 {
 783         return is_hfs_dot_str(path, "gitignore");
 784 }
 785
 786 int is_hfs_dotgitattributes(const char *path)
 787 {
 788         return is_hfs_dot_str(path, "gitattributes");
 789 }
 790
 791 int is_hfs_dotmailmap(const char *path)
 792 {
 793         return is_hfs_dot_str(path, "mailmap");
 794 }
 795
 796 const char utf8_bom[] = "\357\273\277";
 797
 798 int skip_utf8_bom(char **text, size_t len)
 799 {
 800         if (len < strlen(utf8_bom) ||
 801             memcmp(*text, utf8_bom, strlen(utf8_bom)))
 802                 return 0;
 803         *text += strlen(utf8_bom);
 804         return 1;
 805 }
 806
 807 void strbuf_utf8_align(struct strbuf *buf, align_type position, unsigned int width,
 808                        const char *s)
 809 {
 810         size_t slen = strlen(s);
 811         int display_len = utf8_strnwidth(s, slen, 0);
 812         int utf8_compensation = slen - display_len;
 813
 814         if (display_len >= width) {
 815                 strbuf_addstr(buf, s);
 816                 return;
 817         }
 818
 819         if (position == ALIGN_LEFT)
 820                 strbuf_addf(buf, "%-*s", width + utf8_compensation, s);
 821         else if (position == ALIGN_MIDDLE) {
 822                 int left = (width - display_len) / 2;
 823                 strbuf_addf(buf, "%*s%-*s", left, "", width - left + utf8_compensation, s);
 824         } else if (position == ALIGN_RIGHT)
 825                 strbuf_addf(buf, "%*s", width + utf8_compensation, s);
 826 }