schwarze@openbsd.org | b8ae02a | 2018-08-21 13:56:27 +0000 | [diff] [blame] | 1 | /* $OpenBSD: utf8.c,v 1.8 2018/08/21 13:56:27 schwarze Exp $ */ |
schwarze@openbsd.org | 0e059cd | 2016-05-25 23:48:45 +0000 | [diff] [blame] | 2 | /* |
| 3 | * Copyright (c) 2016 Ingo Schwarze <schwarze@openbsd.org> |
| 4 | * |
| 5 | * Permission to use, copy, modify, and distribute this software for any |
| 6 | * purpose with or without fee is hereby granted, provided that the above |
| 7 | * copyright notice and this permission notice appear in all copies. |
| 8 | * |
| 9 | * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES |
| 10 | * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF |
| 11 | * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR |
| 12 | * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES |
| 13 | * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN |
| 14 | * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF |
| 15 | * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. |
| 16 | */ |
| 17 | |
| 18 | /* |
| 19 | * Utility functions for multibyte-character handling, |
| 20 | * in particular to sanitize untrusted strings for terminal output. |
| 21 | */ |
| 22 | |
Darren Tucker | df82072 | 2016-06-06 11:36:13 +1000 | [diff] [blame] | 23 | #include "includes.h" |
| 24 | |
schwarze@openbsd.org | 0e059cd | 2016-05-25 23:48:45 +0000 | [diff] [blame] | 25 | #include <sys/types.h> |
Darren Tucker | f3f2cc8 | 2016-07-11 17:23:38 +1000 | [diff] [blame] | 26 | #ifdef HAVE_LANGINFO_H |
| 27 | # include <langinfo.h> |
| 28 | #endif |
schwarze@openbsd.org | 0e059cd | 2016-05-25 23:48:45 +0000 | [diff] [blame] | 29 | #include <limits.h> |
Damien Miller | dda78a0 | 2016-12-12 13:57:10 +1100 | [diff] [blame] | 30 | #include <locale.h> |
schwarze@openbsd.org | 0e059cd | 2016-05-25 23:48:45 +0000 | [diff] [blame] | 31 | #include <stdarg.h> |
| 32 | #include <stdio.h> |
| 33 | #include <stdlib.h> |
| 34 | #include <string.h> |
Darren Tucker | df82072 | 2016-06-06 11:36:13 +1000 | [diff] [blame] | 35 | #if defined(HAVE_STRNVIS) && defined(HAVE_VIS_H) && !defined(BROKEN_STRNVIS) |
| 36 | # include <vis.h> |
| 37 | #endif |
Darren Tucker | f3f2cc8 | 2016-07-11 17:23:38 +1000 | [diff] [blame] | 38 | #ifdef HAVE_WCHAR_H |
| 39 | # include <wchar.h> |
| 40 | #endif |
schwarze@openbsd.org | 0e059cd | 2016-05-25 23:48:45 +0000 | [diff] [blame] | 41 | |
| 42 | #include "utf8.h" |
| 43 | |
| 44 | static int dangerous_locale(void); |
schwarze@openbsd.org | cd9e1ea | 2016-05-30 12:57:21 +0000 | [diff] [blame] | 45 | static int grow_dst(char **, size_t *, size_t, char **, size_t); |
schwarze@openbsd.org | 0e059cd | 2016-05-25 23:48:45 +0000 | [diff] [blame] | 46 | static int vasnmprintf(char **, size_t, int *, const char *, va_list); |
| 47 | |
| 48 | |
| 49 | /* |
| 50 | * For US-ASCII and UTF-8 encodings, we can safely recover from |
| 51 | * encoding errors and from non-printable characters. For any |
| 52 | * other encodings, err to the side of caution and abort parsing: |
| 53 | * For state-dependent encodings, recovery is impossible. |
| 54 | * For arbitrary encodings, replacement of non-printable |
| 55 | * characters would be non-trivial and too fragile. |
schwarze@openbsd.org | b8ae02a | 2018-08-21 13:56:27 +0000 | [diff] [blame] | 56 | * The comments indicate what nl_langinfo(CODESET) |
| 57 | * returns for US-ASCII on various operating systems. |
schwarze@openbsd.org | 0e059cd | 2016-05-25 23:48:45 +0000 | [diff] [blame] | 58 | */ |
| 59 | |
| 60 | static int |
| 61 | dangerous_locale(void) { |
| 62 | char *loc; |
| 63 | |
| 64 | loc = nl_langinfo(CODESET); |
schwarze@openbsd.org | b8ae02a | 2018-08-21 13:56:27 +0000 | [diff] [blame] | 65 | return strcmp(loc, "UTF-8") != 0 && |
| 66 | strcmp(loc, "US-ASCII") != 0 && /* OpenBSD */ |
| 67 | strcmp(loc, "ANSI_X3.4-1968") != 0 && /* Linux */ |
| 68 | strcmp(loc, "ISO8859-1") != 0 && /* AIX */ |
| 69 | strcmp(loc, "646") != 0 && /* Solaris, NetBSD */ |
| 70 | strcmp(loc, "") != 0; /* Solaris 6 */ |
schwarze@openbsd.org | 0e059cd | 2016-05-25 23:48:45 +0000 | [diff] [blame] | 71 | } |
| 72 | |
schwarze@openbsd.org | cd9e1ea | 2016-05-30 12:57:21 +0000 | [diff] [blame] | 73 | static int |
| 74 | grow_dst(char **dst, size_t *sz, size_t maxsz, char **dp, size_t need) |
| 75 | { |
| 76 | char *tp; |
| 77 | size_t tsz; |
| 78 | |
| 79 | if (*dp + need < *dst + *sz) |
| 80 | return 0; |
| 81 | tsz = *sz + 128; |
| 82 | if (tsz > maxsz) |
| 83 | tsz = maxsz; |
deraadt@openbsd.org | 9e509d4 | 2017-05-31 09:15:42 +0000 | [diff] [blame] | 84 | if ((tp = recallocarray(*dst, *sz, tsz, 1)) == NULL) |
schwarze@openbsd.org | cd9e1ea | 2016-05-30 12:57:21 +0000 | [diff] [blame] | 85 | return -1; |
| 86 | *dp = tp + (*dp - *dst); |
| 87 | *dst = tp; |
| 88 | *sz = tsz; |
| 89 | return 0; |
| 90 | } |
| 91 | |
schwarze@openbsd.org | 0e059cd | 2016-05-25 23:48:45 +0000 | [diff] [blame] | 92 | /* |
| 93 | * The following two functions limit the number of bytes written, |
| 94 | * including the terminating '\0', to sz. Unless wp is NULL, |
| 95 | * they limit the number of display columns occupied to *wp. |
| 96 | * Whichever is reached first terminates the output string. |
| 97 | * To stay close to the standard interfaces, they return the number of |
| 98 | * non-NUL bytes that would have been written if both were unlimited. |
| 99 | * If wp is NULL, newline, carriage return, and tab are allowed; |
| 100 | * otherwise, the actual number of columns occupied by what was |
| 101 | * written is returned in *wp. |
| 102 | */ |
| 103 | |
| 104 | static int |
| 105 | vasnmprintf(char **str, size_t maxsz, int *wp, const char *fmt, va_list ap) |
| 106 | { |
| 107 | char *src; /* Source string returned from vasprintf. */ |
| 108 | char *sp; /* Pointer into src. */ |
| 109 | char *dst; /* Destination string to be returned. */ |
| 110 | char *dp; /* Pointer into dst. */ |
| 111 | char *tp; /* Temporary pointer for dst. */ |
| 112 | size_t sz; /* Number of bytes allocated for dst. */ |
schwarze@openbsd.org | 0e059cd | 2016-05-25 23:48:45 +0000 | [diff] [blame] | 113 | wchar_t wc; /* Wide character at sp. */ |
| 114 | int len; /* Number of bytes in the character at sp. */ |
| 115 | int ret; /* Number of bytes needed to format src. */ |
| 116 | int width; /* Display width of the character wc. */ |
| 117 | int total_width, max_width, print; |
| 118 | |
schwarze@openbsd.org | ac284a3 | 2016-05-30 12:05:56 +0000 | [diff] [blame] | 119 | src = NULL; |
| 120 | if ((ret = vasprintf(&src, fmt, ap)) <= 0) |
schwarze@openbsd.org | 0e059cd | 2016-05-25 23:48:45 +0000 | [diff] [blame] | 121 | goto fail; |
| 122 | |
schwarze@openbsd.org | cd9e1ea | 2016-05-30 12:57:21 +0000 | [diff] [blame] | 123 | sz = strlen(src) + 1; |
schwarze@openbsd.org | ac284a3 | 2016-05-30 12:05:56 +0000 | [diff] [blame] | 124 | if ((dst = malloc(sz)) == NULL) { |
| 125 | free(src); |
jsg@openbsd.org | 3ec5fa4 | 2017-02-02 10:54:25 +0000 | [diff] [blame] | 126 | ret = -1; |
schwarze@openbsd.org | 0e059cd | 2016-05-25 23:48:45 +0000 | [diff] [blame] | 127 | goto fail; |
schwarze@openbsd.org | ac284a3 | 2016-05-30 12:05:56 +0000 | [diff] [blame] | 128 | } |
schwarze@openbsd.org | 0e059cd | 2016-05-25 23:48:45 +0000 | [diff] [blame] | 129 | |
| 130 | if (maxsz > INT_MAX) |
| 131 | maxsz = INT_MAX; |
| 132 | |
| 133 | sp = src; |
| 134 | dp = dst; |
| 135 | ret = 0; |
| 136 | print = 1; |
| 137 | total_width = 0; |
| 138 | max_width = wp == NULL ? INT_MAX : *wp; |
| 139 | while (*sp != '\0') { |
| 140 | if ((len = mbtowc(&wc, sp, MB_CUR_MAX)) == -1) { |
| 141 | (void)mbtowc(NULL, NULL, MB_CUR_MAX); |
| 142 | if (dangerous_locale()) { |
| 143 | ret = -1; |
| 144 | break; |
| 145 | } |
| 146 | len = 1; |
| 147 | width = -1; |
| 148 | } else if (wp == NULL && |
| 149 | (wc == L'\n' || wc == L'\r' || wc == L'\t')) { |
| 150 | /* |
| 151 | * Don't use width uninitialized; the actual |
| 152 | * value doesn't matter because total_width |
| 153 | * is only returned for wp != NULL. |
| 154 | */ |
| 155 | width = 0; |
| 156 | } else if ((width = wcwidth(wc)) == -1 && |
| 157 | dangerous_locale()) { |
| 158 | ret = -1; |
| 159 | break; |
| 160 | } |
| 161 | |
| 162 | /* Valid, printable character. */ |
| 163 | |
| 164 | if (width >= 0) { |
| 165 | if (print && (dp - dst >= (int)maxsz - len || |
| 166 | total_width > max_width - width)) |
| 167 | print = 0; |
| 168 | if (print) { |
schwarze@openbsd.org | cd9e1ea | 2016-05-30 12:57:21 +0000 | [diff] [blame] | 169 | if (grow_dst(&dst, &sz, maxsz, |
| 170 | &dp, len) == -1) { |
| 171 | ret = -1; |
| 172 | break; |
| 173 | } |
schwarze@openbsd.org | 0e059cd | 2016-05-25 23:48:45 +0000 | [diff] [blame] | 174 | total_width += width; |
| 175 | memcpy(dp, sp, len); |
| 176 | dp += len; |
| 177 | } |
| 178 | sp += len; |
| 179 | if (ret >= 0) |
| 180 | ret += len; |
| 181 | continue; |
| 182 | } |
| 183 | |
| 184 | /* Escaping required. */ |
| 185 | |
| 186 | while (len > 0) { |
| 187 | if (print && (dp - dst >= (int)maxsz - 4 || |
| 188 | total_width > max_width - 4)) |
| 189 | print = 0; |
| 190 | if (print) { |
schwarze@openbsd.org | cd9e1ea | 2016-05-30 12:57:21 +0000 | [diff] [blame] | 191 | if (grow_dst(&dst, &sz, maxsz, |
| 192 | &dp, 4) == -1) { |
| 193 | ret = -1; |
| 194 | break; |
schwarze@openbsd.org | 0e059cd | 2016-05-25 23:48:45 +0000 | [diff] [blame] | 195 | } |
| 196 | tp = vis(dp, *sp, VIS_OCTAL | VIS_ALL, 0); |
| 197 | width = tp - dp; |
| 198 | total_width += width; |
| 199 | dp = tp; |
| 200 | } else |
| 201 | width = 4; |
| 202 | len--; |
| 203 | sp++; |
| 204 | if (ret >= 0) |
| 205 | ret += width; |
| 206 | } |
| 207 | if (len > 0) |
| 208 | break; |
| 209 | } |
| 210 | free(src); |
| 211 | *dp = '\0'; |
| 212 | *str = dst; |
| 213 | if (wp != NULL) |
| 214 | *wp = total_width; |
| 215 | |
| 216 | /* |
| 217 | * If the string was truncated by the width limit but |
| 218 | * would have fit into the size limit, the only sane way |
| 219 | * to report the problem is using the return value, such |
| 220 | * that the usual idiom "if (ret < 0 || ret >= sz) error" |
| 221 | * works as expected. |
| 222 | */ |
| 223 | |
| 224 | if (ret < (int)maxsz && !print) |
| 225 | ret = -1; |
| 226 | return ret; |
| 227 | |
| 228 | fail: |
schwarze@openbsd.org | 0e059cd | 2016-05-25 23:48:45 +0000 | [diff] [blame] | 229 | if (wp != NULL) |
| 230 | *wp = 0; |
schwarze@openbsd.org | ac284a3 | 2016-05-30 12:05:56 +0000 | [diff] [blame] | 231 | if (ret == 0) { |
| 232 | *str = src; |
| 233 | return 0; |
| 234 | } else { |
| 235 | *str = NULL; |
| 236 | return -1; |
| 237 | } |
schwarze@openbsd.org | 0e059cd | 2016-05-25 23:48:45 +0000 | [diff] [blame] | 238 | } |
| 239 | |
| 240 | int |
| 241 | snmprintf(char *str, size_t sz, int *wp, const char *fmt, ...) |
| 242 | { |
| 243 | va_list ap; |
| 244 | char *cp; |
| 245 | int ret; |
| 246 | |
| 247 | va_start(ap, fmt); |
| 248 | ret = vasnmprintf(&cp, sz, wp, fmt, ap); |
| 249 | va_end(ap); |
schwarze@openbsd.org | ac284a3 | 2016-05-30 12:05:56 +0000 | [diff] [blame] | 250 | if (cp != NULL) { |
| 251 | (void)strlcpy(str, cp, sz); |
| 252 | free(cp); |
| 253 | } else |
| 254 | *str = '\0'; |
schwarze@openbsd.org | 0e059cd | 2016-05-25 23:48:45 +0000 | [diff] [blame] | 255 | return ret; |
| 256 | } |
| 257 | |
| 258 | /* |
| 259 | * To stay close to the standard interfaces, the following functions |
| 260 | * return the number of non-NUL bytes written. |
| 261 | */ |
| 262 | |
| 263 | int |
| 264 | vfmprintf(FILE *stream, const char *fmt, va_list ap) |
| 265 | { |
| 266 | char *str; |
| 267 | int ret; |
| 268 | |
| 269 | if ((ret = vasnmprintf(&str, INT_MAX, NULL, fmt, ap)) < 0) |
| 270 | return -1; |
| 271 | if (fputs(str, stream) == EOF) |
| 272 | ret = -1; |
| 273 | free(str); |
| 274 | return ret; |
| 275 | } |
| 276 | |
| 277 | int |
| 278 | fmprintf(FILE *stream, const char *fmt, ...) |
| 279 | { |
| 280 | va_list ap; |
| 281 | int ret; |
| 282 | |
| 283 | va_start(ap, fmt); |
| 284 | ret = vfmprintf(stream, fmt, ap); |
| 285 | va_end(ap); |
| 286 | return ret; |
| 287 | } |
| 288 | |
| 289 | int |
| 290 | mprintf(const char *fmt, ...) |
| 291 | { |
| 292 | va_list ap; |
| 293 | int ret; |
| 294 | |
| 295 | va_start(ap, fmt); |
| 296 | ret = vfmprintf(stdout, fmt, ap); |
| 297 | va_end(ap); |
| 298 | return ret; |
| 299 | } |
Damien Miller | dda78a0 | 2016-12-12 13:57:10 +1100 | [diff] [blame] | 300 | |
| 301 | /* |
| 302 | * Set up libc for multibyte output in the user's chosen locale. |
| 303 | * |
| 304 | * XXX: we are known to have problems with Turkish (i/I confusion) so we |
| 305 | * deliberately fall back to the C locale for now. Longer term we should |
| 306 | * always prefer to select C.[encoding] if possible, but there's no |
| 307 | * standardisation in locales between systems, so we'll need to survey |
| 308 | * what's out there first. |
| 309 | */ |
| 310 | void |
| 311 | msetlocale(void) |
| 312 | { |
| 313 | const char *vars[] = { "LC_ALL", "LC_CTYPE", "LANG", NULL }; |
| 314 | char *cp; |
| 315 | int i; |
| 316 | |
| 317 | /* |
| 318 | * We can't yet cope with dotless/dotted I in Turkish locales, |
| 319 | * so fall back to the C locale for these. |
| 320 | */ |
| 321 | for (i = 0; vars[i] != NULL; i++) { |
| 322 | if ((cp = getenv(vars[i])) == NULL) |
| 323 | continue; |
| 324 | if (strncasecmp(cp, "TR", 2) != 0) |
| 325 | break; |
| 326 | /* |
| 327 | * If we're in a UTF-8 locale then prefer to use |
| 328 | * the C.UTF-8 locale (or equivalent) if it exists. |
| 329 | */ |
| 330 | if ((strcasestr(cp, "UTF-8") != NULL || |
| 331 | strcasestr(cp, "UTF8") != NULL) && |
| 332 | (setlocale(LC_CTYPE, "C.UTF-8") != NULL || |
| 333 | setlocale(LC_CTYPE, "POSIX.UTF-8") != NULL)) |
| 334 | return; |
| 335 | setlocale(LC_CTYPE, "C"); |
| 336 | return; |
| 337 | } |
| 338 | /* We can handle this locale */ |
| 339 | setlocale(LC_CTYPE, ""); |
| 340 | } |