1 /* $OpenBSD: utf8.c,v 1.3 2016/05/30 12:57:21 schwarze Exp $ */
3 * Copyright (c) 2016 Ingo Schwarze <schwarze@openbsd.org>
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.
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.
19 * Utility functions for multibyte-character handling,
20 * in particular to sanitize untrusted strings for terminal output.
25 #include <sys/types.h>
26 #ifdef HAVE_LANGINFO_H
27 # include <langinfo.h>
35 #if defined(HAVE_STRNVIS) && defined(HAVE_VIS_H) && !defined(BROKEN_STRNVIS)
44 static int dangerous_locale(void);
45 static int grow_dst(char **, size_t *, size_t, char **, size_t);
46 static int vasnmprintf(char **, size_t, int *, const char *, va_list);
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.
59 dangerous_locale(void) {
62 loc = nl_langinfo(CODESET);
63 return strcmp(loc, "US-ASCII") && strcmp(loc, "UTF-8");
67 grow_dst(char **dst, size_t *sz, size_t maxsz, char **dp, size_t need)
72 if (*dp + need < *dst + *sz)
77 if ((tp = realloc(*dst, tsz)) == NULL)
79 *dp = tp + (*dp - *dst);
86 * The following two functions limit the number of bytes written,
87 * including the terminating '\0', to sz. Unless wp is NULL,
88 * they limit the number of display columns occupied to *wp.
89 * Whichever is reached first terminates the output string.
90 * To stay close to the standard interfaces, they return the number of
91 * non-NUL bytes that would have been written if both were unlimited.
92 * If wp is NULL, newline, carriage return, and tab are allowed;
93 * otherwise, the actual number of columns occupied by what was
94 * written is returned in *wp.
98 vasnmprintf(char **str, size_t maxsz, int *wp, const char *fmt, va_list ap)
100 char *src; /* Source string returned from vasprintf. */
101 char *sp; /* Pointer into src. */
102 char *dst; /* Destination string to be returned. */
103 char *dp; /* Pointer into dst. */
104 char *tp; /* Temporary pointer for dst. */
105 size_t sz; /* Number of bytes allocated for dst. */
106 wchar_t wc; /* Wide character at sp. */
107 int len; /* Number of bytes in the character at sp. */
108 int ret; /* Number of bytes needed to format src. */
109 int width; /* Display width of the character wc. */
110 int total_width, max_width, print;
113 if ((ret = vasprintf(&src, fmt, ap)) <= 0)
116 sz = strlen(src) + 1;
117 if ((dst = malloc(sz)) == NULL) {
130 max_width = wp == NULL ? INT_MAX : *wp;
131 while (*sp != '\0') {
132 if ((len = mbtowc(&wc, sp, MB_CUR_MAX)) == -1) {
133 (void)mbtowc(NULL, NULL, MB_CUR_MAX);
134 if (dangerous_locale()) {
140 } else if (wp == NULL &&
141 (wc == L'\n' || wc == L'\r' || wc == L'\t')) {
143 * Don't use width uninitialized; the actual
144 * value doesn't matter because total_width
145 * is only returned for wp != NULL.
148 } else if ((width = wcwidth(wc)) == -1 &&
149 dangerous_locale()) {
154 /* Valid, printable character. */
157 if (print && (dp - dst >= (int)maxsz - len ||
158 total_width > max_width - width))
161 if (grow_dst(&dst, &sz, maxsz,
166 total_width += width;
176 /* Escaping required. */
179 if (print && (dp - dst >= (int)maxsz - 4 ||
180 total_width > max_width - 4))
183 if (grow_dst(&dst, &sz, maxsz,
188 tp = vis(dp, *sp, VIS_OCTAL | VIS_ALL, 0);
190 total_width += width;
209 * If the string was truncated by the width limit but
210 * would have fit into the size limit, the only sane way
211 * to report the problem is using the return value, such
212 * that the usual idiom "if (ret < 0 || ret >= sz) error"
216 if (ret < (int)maxsz && !print)
233 snmprintf(char *str, size_t sz, int *wp, const char *fmt, ...)
240 ret = vasnmprintf(&cp, sz, wp, fmt, ap);
243 (void)strlcpy(str, cp, sz);
251 * To stay close to the standard interfaces, the following functions
252 * return the number of non-NUL bytes written.
256 vfmprintf(FILE *stream, const char *fmt, va_list ap)
261 if ((ret = vasnmprintf(&str, INT_MAX, NULL, fmt, ap)) < 0)
263 if (fputs(str, stream) == EOF)
270 fmprintf(FILE *stream, const char *fmt, ...)
276 ret = vfmprintf(stream, fmt, ap);
282 mprintf(const char *fmt, ...)
288 ret = vfmprintf(stdout, fmt, ap);
294 * Set up libc for multibyte output in the user's chosen locale.
296 * XXX: we are known to have problems with Turkish (i/I confusion) so we
297 * deliberately fall back to the C locale for now. Longer term we should
298 * always prefer to select C.[encoding] if possible, but there's no
299 * standardisation in locales between systems, so we'll need to survey
300 * what's out there first.
305 const char *vars[] = { "LC_ALL", "LC_CTYPE", "LANG", NULL };
310 * We can't yet cope with dotless/dotted I in Turkish locales,
311 * so fall back to the C locale for these.
313 for (i = 0; vars[i] != NULL; i++) {
314 if ((cp = getenv(vars[i])) == NULL)
316 if (strncasecmp(cp, "TR", 2) != 0)
319 * If we're in a UTF-8 locale then prefer to use
320 * the C.UTF-8 locale (or equivalent) if it exists.
322 if ((strcasestr(cp, "UTF-8") != NULL ||
323 strcasestr(cp, "UTF8") != NULL) &&
324 (setlocale(LC_CTYPE, "C.UTF-8") != NULL ||
325 setlocale(LC_CTYPE, "POSIX.UTF-8") != NULL))
327 setlocale(LC_CTYPE, "C");
330 /* We can handle this locale */
331 setlocale(LC_CTYPE, "");