1 /*
2 * This file is part of mpv.
3 *
4 * Based on code taken from libass (ISC license), which was originally part
5 * of MPlayer (GPL).
6 * Copyright (C) 2006 Evgeniy Stepanov <eugeni.stepanov@gmail.com>
7 *
8 * mpv is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
11 * version 2.1 of the License, or (at your option) any later version.
12 *
13 * mpv is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 * GNU Lesser General Public License for more details.
17 *
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with mpv. If not, see <http://www.gnu.org/licenses/>.
20 */
21
22 #include <stdlib.h>
23 #include <errno.h>
24 #include <strings.h>
25 #include <assert.h>
26
27 #include "config.h"
28
29 #include "common/msg.h"
30
31 #if HAVE_UCHARDET
32 #include <uchardet.h>
33 #endif
34
35 #if HAVE_ICONV
36 #include <iconv.h>
37 #endif
38
39 #include "charset_conv.h"
40
mp_charset_is_utf8(const char * user_cp)41 bool mp_charset_is_utf8(const char *user_cp)
42 {
43 return user_cp && (strcasecmp(user_cp, "utf8") == 0 ||
44 strcasecmp(user_cp, "utf-8") == 0);
45 }
46
mp_charset_is_utf16(const char * user_cp)47 bool mp_charset_is_utf16(const char *user_cp)
48 {
49 bstr s = bstr0(user_cp);
50 return bstr_case_startswith(s, bstr0("utf16")) ||
51 bstr_case_startswith(s, bstr0("utf-16"));
52 }
53
54 static const char *const utf_bom[3] = {"\xEF\xBB\xBF", "\xFF\xFE", "\xFE\xFF"};
55 static const char *const utf_enc[3] = {"utf-8", "utf-16le", "utf-16be"};
56
ms_bom_guess(bstr buf)57 static const char *ms_bom_guess(bstr buf)
58 {
59 for (int n = 0; n < 3; n++) {
60 if (bstr_startswith0(buf, utf_bom[n]))
61 return utf_enc[n];
62 }
63 return NULL;
64 }
65
66 #if HAVE_UCHARDET
mp_uchardet(void * talloc_ctx,struct mp_log * log,bstr buf)67 static const char *mp_uchardet(void *talloc_ctx, struct mp_log *log, bstr buf)
68 {
69 uchardet_t det = uchardet_new();
70 if (!det)
71 return NULL;
72 if (uchardet_handle_data(det, buf.start, buf.len) != 0) {
73 uchardet_delete(det);
74 return NULL;
75 }
76 uchardet_data_end(det);
77 char *res = talloc_strdup(talloc_ctx, uchardet_get_charset(det));
78 if (res && !res[0])
79 res = NULL;
80 if (res) {
81 mp_verbose(log, "libuchardet detected charset as %s\n", res);
82 iconv_t icdsc = iconv_open("UTF-8", res);
83 if (icdsc == (iconv_t)(-1)) {
84 mp_warn(log, "Charset '%s' not supported by iconv.\n", res);
85 res = NULL;
86 } else {
87 iconv_close(icdsc);
88 }
89 }
90 uchardet_delete(det);
91 return res;
92 }
93 #endif
94
95 // Runs charset auto-detection on the input buffer, and returns the result.
96 // If auto-detection fails, NULL is returned.
97 // If user_cp doesn't refer to any known auto-detection (for example because
98 // it's a real iconv codepage), user_cp is returned without even looking at
99 // the buf data.
100 // The return value may (but doesn't have to) be allocated under talloc_ctx.
mp_charset_guess(void * talloc_ctx,struct mp_log * log,bstr buf,const char * user_cp,int flags)101 const char *mp_charset_guess(void *talloc_ctx, struct mp_log *log, bstr buf,
102 const char *user_cp, int flags)
103 {
104 if (strcasecmp(user_cp, "enca") == 0 || strcasecmp(user_cp, "guess") == 0 ||
105 strcasecmp(user_cp, "uchardet") == 0 || strchr(user_cp, ':'))
106 {
107 mp_err(log, "This syntax for the --sub-codepage option was deprecated "
108 "and has been removed.\n");
109 if (strncasecmp(user_cp, "utf8:", 5) == 0) {
110 user_cp = user_cp + 5;
111 } else {
112 user_cp = "";
113 }
114 }
115
116 if (user_cp[0] == '+') {
117 mp_verbose(log, "Forcing charset '%s'.\n", user_cp + 1);
118 return user_cp + 1;
119 }
120
121 const char *bom_cp = ms_bom_guess(buf);
122 if (bom_cp) {
123 mp_verbose(log, "Data has a BOM, assuming %s as charset.\n", bom_cp);
124 return bom_cp;
125 }
126
127 int r = bstr_validate_utf8(buf);
128 if (r >= 0 || (r > -8 && (flags & MP_ICONV_ALLOW_CUTOFF))) {
129 mp_verbose(log, "Data looks like UTF-8, ignoring user-provided charset.\n");
130 return "utf-8";
131 }
132
133 const char *res = NULL;
134 if (strcasecmp(user_cp, "auto") == 0) {
135 #if HAVE_UCHARDET
136 res = mp_uchardet(talloc_ctx, log, buf);
137 #endif
138 if (!res) {
139 mp_verbose(log, "Charset auto-detection failed.\n");
140 res = "UTF-8-BROKEN";
141 }
142 } else {
143 res = user_cp;
144 }
145
146 mp_verbose(log, "Using charset '%s'.\n", res);
147 return res;
148 }
149
150 // Use iconv to convert buf to UTF-8.
151 // Returns buf.start==NULL on error. Returns buf if cp is NULL, or if there is
152 // obviously no conversion required (e.g. if cp is "UTF-8").
153 // Returns a newly allocated buffer if conversion is done and succeeds. The
154 // buffer will be terminated with 0 for convenience (the terminating 0 is not
155 // included in the returned length).
156 // Free the returned buffer with talloc_free().
157 // buf: input data
158 // cp: iconv codepage (or NULL)
159 // flags: combination of MP_ICONV_* flags
160 // returns: buf (no conversion), .start==NULL (error), or allocated buffer
mp_iconv_to_utf8(struct mp_log * log,bstr buf,const char * cp,int flags)161 bstr mp_iconv_to_utf8(struct mp_log *log, bstr buf, const char *cp, int flags)
162 {
163 #if HAVE_ICONV
164 if (!cp || !cp[0] || mp_charset_is_utf8(cp))
165 return buf;
166
167 if (strcasecmp(cp, "ASCII") == 0)
168 return buf;
169
170 if (strcasecmp(cp, "UTF-8-BROKEN") == 0)
171 return bstr_sanitize_utf8_latin1(NULL, buf);
172
173 // Force CP949 over EUC-KR since iconv distinguishes them and
174 // EUC-KR causes error on CP949 encoded data
175 if (strcasecmp(cp, "EUC-KR") == 0)
176 cp = "CP949";
177
178 iconv_t icdsc;
179 if ((icdsc = iconv_open("UTF-8", cp)) == (iconv_t) (-1)) {
180 if (flags & MP_ICONV_VERBOSE)
181 mp_err(log, "Error opening iconv with codepage '%s'\n", cp);
182 goto failure;
183 }
184
185 size_t size = buf.len;
186 size_t osize = size;
187 size_t ileft = size;
188 size_t oleft = size - 1;
189
190 char *outbuf = talloc_size(NULL, osize);
191 char *ip = buf.start;
192 char *op = outbuf;
193
194 while (1) {
195 int clear = 0;
196 size_t rc;
197 if (ileft)
198 rc = iconv(icdsc, &ip, &ileft, &op, &oleft);
199 else {
200 clear = 1; // clear the conversion state and leave
201 rc = iconv(icdsc, NULL, NULL, &op, &oleft);
202 }
203 if (rc == (size_t) (-1)) {
204 if (errno == E2BIG) {
205 size_t offset = op - outbuf;
206 outbuf = talloc_realloc_size(NULL, outbuf, osize + size);
207 op = outbuf + offset;
208 osize += size;
209 oleft += size;
210 } else {
211 if (errno == EINVAL && (flags & MP_ICONV_ALLOW_CUTOFF)) {
212 // This is intended for cases where the input buffer is cut
213 // at a random byte position. If this happens in the middle
214 // of the buffer, it should still be an error. We say it's
215 // fine if the error is within 10 bytes of the end.
216 if (ileft <= 10)
217 break;
218 }
219 if (flags & MP_ICONV_VERBOSE) {
220 mp_err(log, "Error recoding text with codepage '%s'\n", cp);
221 }
222 talloc_free(outbuf);
223 iconv_close(icdsc);
224 goto failure;
225 }
226 } else if (clear)
227 break;
228 }
229
230 iconv_close(icdsc);
231
232 outbuf[osize - oleft - 1] = 0;
233 return (bstr){outbuf, osize - oleft - 1};
234 #endif
235
236 failure:
237 if (flags & MP_NO_LATIN1_FALLBACK) {
238 return buf;
239 } else {
240 return bstr_sanitize_utf8_latin1(NULL, buf);
241 }
242 }
243