Mercurial > hg > CbC > CbC_gcc
comparison libcpp/charset.c @ 0:a06113de4d67
first commit
author | kent <kent@cr.ie.u-ryukyu.ac.jp> |
---|---|
date | Fri, 17 Jul 2009 14:47:48 +0900 |
parents | |
children | 77e2b8dfacca |
comparison
equal
deleted
inserted
replaced
-1:000000000000 | 0:a06113de4d67 |
---|---|
1 /* CPP Library - charsets | |
2 Copyright (C) 1998, 1999, 2000, 2001, 2002, 2003, 2004, 2006, 2008, 2009 | |
3 Free Software Foundation, Inc. | |
4 | |
5 Broken out of c-lex.c Apr 2003, adding valid C99 UCN ranges. | |
6 | |
7 This program is free software; you can redistribute it and/or modify it | |
8 under the terms of the GNU General Public License as published by the | |
9 Free Software Foundation; either version 3, or (at your option) any | |
10 later version. | |
11 | |
12 This program is distributed in the hope that it will be useful, | |
13 but WITHOUT ANY WARRANTY; without even the implied warranty of | |
14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
15 GNU General Public License for more details. | |
16 | |
17 You should have received a copy of the GNU General Public License | |
18 along with this program; see the file COPYING3. If not see | |
19 <http://www.gnu.org/licenses/>. */ | |
20 | |
21 #include "config.h" | |
22 #include "system.h" | |
23 #include "cpplib.h" | |
24 #include "internal.h" | |
25 | |
26 /* Character set handling for C-family languages. | |
27 | |
28 Terminological note: In what follows, "charset" or "character set" | |
29 will be taken to mean both an abstract set of characters and an | |
30 encoding for that set. | |
31 | |
32 The C99 standard discusses two character sets: source and execution. | |
33 The source character set is used for internal processing in translation | |
34 phases 1 through 4; the execution character set is used thereafter. | |
35 Both are required by 5.2.1.2p1 to be multibyte encodings, not wide | |
36 character encodings (see 3.7.2, 3.7.3 for the standardese meanings | |
37 of these terms). Furthermore, the "basic character set" (listed in | |
38 5.2.1p3) is to be encoded in each with values one byte wide, and is | |
39 to appear in the initial shift state. | |
40 | |
41 It is not explicitly mentioned, but there is also a "wide execution | |
42 character set" used to encode wide character constants and wide | |
43 string literals; this is supposed to be the result of applying the | |
44 standard library function mbstowcs() to an equivalent narrow string | |
45 (6.4.5p5). However, the behavior of hexadecimal and octal | |
46 \-escapes is at odds with this; they are supposed to be translated | |
47 directly to wchar_t values (6.4.4.4p5,6). | |
48 | |
49 The source character set is not necessarily the character set used | |
50 to encode physical source files on disk; translation phase 1 converts | |
51 from whatever that encoding is to the source character set. | |
52 | |
53 The presence of universal character names in C99 (6.4.3 et seq.) | |
54 forces the source character set to be isomorphic to ISO 10646, | |
55 that is, Unicode. There is no such constraint on the execution | |
56 character set; note also that the conversion from source to | |
57 execution character set does not occur for identifiers (5.1.1.2p1#5). | |
58 | |
59 For convenience of implementation, the source character set's | |
60 encoding of the basic character set should be identical to the | |
61 execution character set OF THE HOST SYSTEM's encoding of the basic | |
62 character set, and it should not be a state-dependent encoding. | |
63 | |
64 cpplib uses UTF-8 or UTF-EBCDIC for the source character set, | |
65 depending on whether the host is based on ASCII or EBCDIC (see | |
66 respectively Unicode section 2.3/ISO10646 Amendment 2, and Unicode | |
67 Technical Report #16). With limited exceptions, it relies on the | |
68 system library's iconv() primitive to do charset conversion | |
69 (specified in SUSv2). */ | |
70 | |
71 #if !HAVE_ICONV | |
72 /* Make certain that the uses of iconv(), iconv_open(), iconv_close() | |
73 below, which are guarded only by if statements with compile-time | |
74 constant conditions, do not cause link errors. */ | |
75 #define iconv_open(x, y) (errno = EINVAL, (iconv_t)-1) | |
76 #define iconv(a,b,c,d,e) (errno = EINVAL, (size_t)-1) | |
77 #define iconv_close(x) (void)0 | |
78 #define ICONV_CONST | |
79 #endif | |
80 | |
81 #if HOST_CHARSET == HOST_CHARSET_ASCII | |
82 #define SOURCE_CHARSET "UTF-8" | |
83 #define LAST_POSSIBLY_BASIC_SOURCE_CHAR 0x7e | |
84 #elif HOST_CHARSET == HOST_CHARSET_EBCDIC | |
85 #define SOURCE_CHARSET "UTF-EBCDIC" | |
86 #define LAST_POSSIBLY_BASIC_SOURCE_CHAR 0xFF | |
87 #else | |
88 #error "Unrecognized basic host character set" | |
89 #endif | |
90 | |
91 #ifndef EILSEQ | |
92 #define EILSEQ EINVAL | |
93 #endif | |
94 | |
95 /* This structure is used for a resizable string buffer throughout. */ | |
96 /* Don't call it strbuf, as that conflicts with unistd.h on systems | |
97 such as DYNIX/ptx where unistd.h includes stropts.h. */ | |
98 struct _cpp_strbuf | |
99 { | |
100 uchar *text; | |
101 size_t asize; | |
102 size_t len; | |
103 }; | |
104 | |
105 /* This is enough to hold any string that fits on a single 80-column | |
106 line, even if iconv quadruples its size (e.g. conversion from | |
107 ASCII to UTF-32) rounded up to a power of two. */ | |
108 #define OUTBUF_BLOCK_SIZE 256 | |
109 | |
110 /* Conversions between UTF-8 and UTF-16/32 are implemented by custom | |
111 logic. This is because a depressing number of systems lack iconv, | |
112 or have have iconv libraries that do not do these conversions, so | |
113 we need a fallback implementation for them. To ensure the fallback | |
114 doesn't break due to neglect, it is used on all systems. | |
115 | |
116 UTF-32 encoding is nice and simple: a four-byte binary number, | |
117 constrained to the range 00000000-7FFFFFFF to avoid questions of | |
118 signedness. We do have to cope with big- and little-endian | |
119 variants. | |
120 | |
121 UTF-16 encoding uses two-byte binary numbers, again in big- and | |
122 little-endian variants, for all values in the 00000000-0000FFFF | |
123 range. Values in the 00010000-0010FFFF range are encoded as pairs | |
124 of two-byte numbers, called "surrogate pairs": given a number S in | |
125 this range, it is mapped to a pair (H, L) as follows: | |
126 | |
127 H = (S - 0x10000) / 0x400 + 0xD800 | |
128 L = (S - 0x10000) % 0x400 + 0xDC00 | |
129 | |
130 Two-byte values in the D800...DFFF range are ill-formed except as a | |
131 component of a surrogate pair. Even if the encoding within a | |
132 two-byte value is little-endian, the H member of the surrogate pair | |
133 comes first. | |
134 | |
135 There is no way to encode values in the 00110000-7FFFFFFF range, | |
136 which is not currently a problem as there are no assigned code | |
137 points in that range; however, the author expects that it will | |
138 eventually become necessary to abandon UTF-16 due to this | |
139 limitation. Note also that, because of these pairs, UTF-16 does | |
140 not meet the requirements of the C standard for a wide character | |
141 encoding (see 3.7.3 and 6.4.4.4p11). | |
142 | |
143 UTF-8 encoding looks like this: | |
144 | |
145 value range encoded as | |
146 00000000-0000007F 0xxxxxxx | |
147 00000080-000007FF 110xxxxx 10xxxxxx | |
148 00000800-0000FFFF 1110xxxx 10xxxxxx 10xxxxxx | |
149 00010000-001FFFFF 11110xxx 10xxxxxx 10xxxxxx 10xxxxxx | |
150 00200000-03FFFFFF 111110xx 10xxxxxx 10xxxxxx 10xxxxxx 10xxxxxx | |
151 04000000-7FFFFFFF 1111110x 10xxxxxx 10xxxxxx 10xxxxxx 10xxxxxx 10xxxxxx | |
152 | |
153 Values in the 0000D800 ... 0000DFFF range (surrogates) are invalid, | |
154 which means that three-byte sequences ED xx yy, with A0 <= xx <= BF, | |
155 never occur. Note also that any value that can be encoded by a | |
156 given row of the table can also be encoded by all successive rows, | |
157 but this is not done; only the shortest possible encoding for any | |
158 given value is valid. For instance, the character 07C0 could be | |
159 encoded as any of DF 80, E0 9F 80, F0 80 9F 80, F8 80 80 9F 80, or | |
160 FC 80 80 80 9F 80. Only the first is valid. | |
161 | |
162 An implementation note: the transformation from UTF-16 to UTF-8, or | |
163 vice versa, is easiest done by using UTF-32 as an intermediary. */ | |
164 | |
165 /* Internal primitives which go from an UTF-8 byte stream to native-endian | |
166 UTF-32 in a cppchar_t, or vice versa; this avoids an extra marshal/unmarshal | |
167 operation in several places below. */ | |
168 static inline int | |
169 one_utf8_to_cppchar (const uchar **inbufp, size_t *inbytesleftp, | |
170 cppchar_t *cp) | |
171 { | |
172 static const uchar masks[6] = { 0x7F, 0x1F, 0x0F, 0x07, 0x02, 0x01 }; | |
173 static const uchar patns[6] = { 0x00, 0xC0, 0xE0, 0xF0, 0xF8, 0xFC }; | |
174 | |
175 cppchar_t c; | |
176 const uchar *inbuf = *inbufp; | |
177 size_t nbytes, i; | |
178 | |
179 if (*inbytesleftp < 1) | |
180 return EINVAL; | |
181 | |
182 c = *inbuf; | |
183 if (c < 0x80) | |
184 { | |
185 *cp = c; | |
186 *inbytesleftp -= 1; | |
187 *inbufp += 1; | |
188 return 0; | |
189 } | |
190 | |
191 /* The number of leading 1-bits in the first byte indicates how many | |
192 bytes follow. */ | |
193 for (nbytes = 2; nbytes < 7; nbytes++) | |
194 if ((c & ~masks[nbytes-1]) == patns[nbytes-1]) | |
195 goto found; | |
196 return EILSEQ; | |
197 found: | |
198 | |
199 if (*inbytesleftp < nbytes) | |
200 return EINVAL; | |
201 | |
202 c = (c & masks[nbytes-1]); | |
203 inbuf++; | |
204 for (i = 1; i < nbytes; i++) | |
205 { | |
206 cppchar_t n = *inbuf++; | |
207 if ((n & 0xC0) != 0x80) | |
208 return EILSEQ; | |
209 c = ((c << 6) + (n & 0x3F)); | |
210 } | |
211 | |
212 /* Make sure the shortest possible encoding was used. */ | |
213 if (c <= 0x7F && nbytes > 1) return EILSEQ; | |
214 if (c <= 0x7FF && nbytes > 2) return EILSEQ; | |
215 if (c <= 0xFFFF && nbytes > 3) return EILSEQ; | |
216 if (c <= 0x1FFFFF && nbytes > 4) return EILSEQ; | |
217 if (c <= 0x3FFFFFF && nbytes > 5) return EILSEQ; | |
218 | |
219 /* Make sure the character is valid. */ | |
220 if (c > 0x7FFFFFFF || (c >= 0xD800 && c <= 0xDFFF)) return EILSEQ; | |
221 | |
222 *cp = c; | |
223 *inbufp = inbuf; | |
224 *inbytesleftp -= nbytes; | |
225 return 0; | |
226 } | |
227 | |
228 static inline int | |
229 one_cppchar_to_utf8 (cppchar_t c, uchar **outbufp, size_t *outbytesleftp) | |
230 { | |
231 static const uchar masks[6] = { 0x00, 0xC0, 0xE0, 0xF0, 0xF8, 0xFC }; | |
232 static const uchar limits[6] = { 0x80, 0xE0, 0xF0, 0xF8, 0xFC, 0xFE }; | |
233 size_t nbytes; | |
234 uchar buf[6], *p = &buf[6]; | |
235 uchar *outbuf = *outbufp; | |
236 | |
237 nbytes = 1; | |
238 if (c < 0x80) | |
239 *--p = c; | |
240 else | |
241 { | |
242 do | |
243 { | |
244 *--p = ((c & 0x3F) | 0x80); | |
245 c >>= 6; | |
246 nbytes++; | |
247 } | |
248 while (c >= 0x3F || (c & limits[nbytes-1])); | |
249 *--p = (c | masks[nbytes-1]); | |
250 } | |
251 | |
252 if (*outbytesleftp < nbytes) | |
253 return E2BIG; | |
254 | |
255 while (p < &buf[6]) | |
256 *outbuf++ = *p++; | |
257 *outbytesleftp -= nbytes; | |
258 *outbufp = outbuf; | |
259 return 0; | |
260 } | |
261 | |
262 /* The following four functions transform one character between the two | |
263 encodings named in the function name. All have the signature | |
264 int (*)(iconv_t bigend, const uchar **inbufp, size_t *inbytesleftp, | |
265 uchar **outbufp, size_t *outbytesleftp) | |
266 | |
267 BIGEND must have the value 0 or 1, coerced to (iconv_t); it is | |
268 interpreted as a boolean indicating whether big-endian or | |
269 little-endian encoding is to be used for the member of the pair | |
270 that is not UTF-8. | |
271 | |
272 INBUFP, INBYTESLEFTP, OUTBUFP, OUTBYTESLEFTP work exactly as they | |
273 do for iconv. | |
274 | |
275 The return value is either 0 for success, or an errno value for | |
276 failure, which may be E2BIG (need more space), EILSEQ (ill-formed | |
277 input sequence), ir EINVAL (incomplete input sequence). */ | |
278 | |
279 static inline int | |
280 one_utf8_to_utf32 (iconv_t bigend, const uchar **inbufp, size_t *inbytesleftp, | |
281 uchar **outbufp, size_t *outbytesleftp) | |
282 { | |
283 uchar *outbuf; | |
284 cppchar_t s = 0; | |
285 int rval; | |
286 | |
287 /* Check for space first, since we know exactly how much we need. */ | |
288 if (*outbytesleftp < 4) | |
289 return E2BIG; | |
290 | |
291 rval = one_utf8_to_cppchar (inbufp, inbytesleftp, &s); | |
292 if (rval) | |
293 return rval; | |
294 | |
295 outbuf = *outbufp; | |
296 outbuf[bigend ? 3 : 0] = (s & 0x000000FF); | |
297 outbuf[bigend ? 2 : 1] = (s & 0x0000FF00) >> 8; | |
298 outbuf[bigend ? 1 : 2] = (s & 0x00FF0000) >> 16; | |
299 outbuf[bigend ? 0 : 3] = (s & 0xFF000000) >> 24; | |
300 | |
301 *outbufp += 4; | |
302 *outbytesleftp -= 4; | |
303 return 0; | |
304 } | |
305 | |
306 static inline int | |
307 one_utf32_to_utf8 (iconv_t bigend, const uchar **inbufp, size_t *inbytesleftp, | |
308 uchar **outbufp, size_t *outbytesleftp) | |
309 { | |
310 cppchar_t s; | |
311 int rval; | |
312 const uchar *inbuf; | |
313 | |
314 if (*inbytesleftp < 4) | |
315 return EINVAL; | |
316 | |
317 inbuf = *inbufp; | |
318 | |
319 s = inbuf[bigend ? 0 : 3] << 24; | |
320 s += inbuf[bigend ? 1 : 2] << 16; | |
321 s += inbuf[bigend ? 2 : 1] << 8; | |
322 s += inbuf[bigend ? 3 : 0]; | |
323 | |
324 if (s >= 0x7FFFFFFF || (s >= 0xD800 && s <= 0xDFFF)) | |
325 return EILSEQ; | |
326 | |
327 rval = one_cppchar_to_utf8 (s, outbufp, outbytesleftp); | |
328 if (rval) | |
329 return rval; | |
330 | |
331 *inbufp += 4; | |
332 *inbytesleftp -= 4; | |
333 return 0; | |
334 } | |
335 | |
336 static inline int | |
337 one_utf8_to_utf16 (iconv_t bigend, const uchar **inbufp, size_t *inbytesleftp, | |
338 uchar **outbufp, size_t *outbytesleftp) | |
339 { | |
340 int rval; | |
341 cppchar_t s = 0; | |
342 const uchar *save_inbuf = *inbufp; | |
343 size_t save_inbytesleft = *inbytesleftp; | |
344 uchar *outbuf = *outbufp; | |
345 | |
346 rval = one_utf8_to_cppchar (inbufp, inbytesleftp, &s); | |
347 if (rval) | |
348 return rval; | |
349 | |
350 if (s > 0x0010FFFF) | |
351 { | |
352 *inbufp = save_inbuf; | |
353 *inbytesleftp = save_inbytesleft; | |
354 return EILSEQ; | |
355 } | |
356 | |
357 if (s < 0xFFFF) | |
358 { | |
359 if (*outbytesleftp < 2) | |
360 { | |
361 *inbufp = save_inbuf; | |
362 *inbytesleftp = save_inbytesleft; | |
363 return E2BIG; | |
364 } | |
365 outbuf[bigend ? 1 : 0] = (s & 0x00FF); | |
366 outbuf[bigend ? 0 : 1] = (s & 0xFF00) >> 8; | |
367 | |
368 *outbufp += 2; | |
369 *outbytesleftp -= 2; | |
370 return 0; | |
371 } | |
372 else | |
373 { | |
374 cppchar_t hi, lo; | |
375 | |
376 if (*outbytesleftp < 4) | |
377 { | |
378 *inbufp = save_inbuf; | |
379 *inbytesleftp = save_inbytesleft; | |
380 return E2BIG; | |
381 } | |
382 | |
383 hi = (s - 0x10000) / 0x400 + 0xD800; | |
384 lo = (s - 0x10000) % 0x400 + 0xDC00; | |
385 | |
386 /* Even if we are little-endian, put the high surrogate first. | |
387 ??? Matches practice? */ | |
388 outbuf[bigend ? 1 : 0] = (hi & 0x00FF); | |
389 outbuf[bigend ? 0 : 1] = (hi & 0xFF00) >> 8; | |
390 outbuf[bigend ? 3 : 2] = (lo & 0x00FF); | |
391 outbuf[bigend ? 2 : 3] = (lo & 0xFF00) >> 8; | |
392 | |
393 *outbufp += 4; | |
394 *outbytesleftp -= 4; | |
395 return 0; | |
396 } | |
397 } | |
398 | |
399 static inline int | |
400 one_utf16_to_utf8 (iconv_t bigend, const uchar **inbufp, size_t *inbytesleftp, | |
401 uchar **outbufp, size_t *outbytesleftp) | |
402 { | |
403 cppchar_t s; | |
404 const uchar *inbuf = *inbufp; | |
405 int rval; | |
406 | |
407 if (*inbytesleftp < 2) | |
408 return EINVAL; | |
409 s = inbuf[bigend ? 0 : 1] << 8; | |
410 s += inbuf[bigend ? 1 : 0]; | |
411 | |
412 /* Low surrogate without immediately preceding high surrogate is invalid. */ | |
413 if (s >= 0xDC00 && s <= 0xDFFF) | |
414 return EILSEQ; | |
415 /* High surrogate must have a following low surrogate. */ | |
416 else if (s >= 0xD800 && s <= 0xDBFF) | |
417 { | |
418 cppchar_t hi = s, lo; | |
419 if (*inbytesleftp < 4) | |
420 return EINVAL; | |
421 | |
422 lo = inbuf[bigend ? 2 : 3] << 8; | |
423 lo += inbuf[bigend ? 3 : 2]; | |
424 | |
425 if (lo < 0xDC00 || lo > 0xDFFF) | |
426 return EILSEQ; | |
427 | |
428 s = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000; | |
429 } | |
430 | |
431 rval = one_cppchar_to_utf8 (s, outbufp, outbytesleftp); | |
432 if (rval) | |
433 return rval; | |
434 | |
435 /* Success - update the input pointers (one_cppchar_to_utf8 has done | |
436 the output pointers for us). */ | |
437 if (s <= 0xFFFF) | |
438 { | |
439 *inbufp += 2; | |
440 *inbytesleftp -= 2; | |
441 } | |
442 else | |
443 { | |
444 *inbufp += 4; | |
445 *inbytesleftp -= 4; | |
446 } | |
447 return 0; | |
448 } | |
449 | |
450 /* Helper routine for the next few functions. The 'const' on | |
451 one_conversion means that we promise not to modify what function is | |
452 pointed to, which lets the inliner see through it. */ | |
453 | |
454 static inline bool | |
455 conversion_loop (int (*const one_conversion)(iconv_t, const uchar **, size_t *, | |
456 uchar **, size_t *), | |
457 iconv_t cd, const uchar *from, size_t flen, struct _cpp_strbuf *to) | |
458 { | |
459 const uchar *inbuf; | |
460 uchar *outbuf; | |
461 size_t inbytesleft, outbytesleft; | |
462 int rval; | |
463 | |
464 inbuf = from; | |
465 inbytesleft = flen; | |
466 outbuf = to->text + to->len; | |
467 outbytesleft = to->asize - to->len; | |
468 | |
469 for (;;) | |
470 { | |
471 do | |
472 rval = one_conversion (cd, &inbuf, &inbytesleft, | |
473 &outbuf, &outbytesleft); | |
474 while (inbytesleft && !rval); | |
475 | |
476 if (__builtin_expect (inbytesleft == 0, 1)) | |
477 { | |
478 to->len = to->asize - outbytesleft; | |
479 return true; | |
480 } | |
481 if (rval != E2BIG) | |
482 { | |
483 errno = rval; | |
484 return false; | |
485 } | |
486 | |
487 outbytesleft += OUTBUF_BLOCK_SIZE; | |
488 to->asize += OUTBUF_BLOCK_SIZE; | |
489 to->text = XRESIZEVEC (uchar, to->text, to->asize); | |
490 outbuf = to->text + to->asize - outbytesleft; | |
491 } | |
492 } | |
493 | |
494 | |
495 /* These functions convert entire strings between character sets. | |
496 They all have the signature | |
497 | |
498 bool (*)(iconv_t cd, const uchar *from, size_t flen, struct _cpp_strbuf *to); | |
499 | |
500 The input string FROM is converted as specified by the function | |
501 name plus the iconv descriptor CD (which may be fake), and the | |
502 result appended to TO. On any error, false is returned, otherwise true. */ | |
503 | |
504 /* These four use the custom conversion code above. */ | |
505 static bool | |
506 convert_utf8_utf16 (iconv_t cd, const uchar *from, size_t flen, | |
507 struct _cpp_strbuf *to) | |
508 { | |
509 return conversion_loop (one_utf8_to_utf16, cd, from, flen, to); | |
510 } | |
511 | |
512 static bool | |
513 convert_utf8_utf32 (iconv_t cd, const uchar *from, size_t flen, | |
514 struct _cpp_strbuf *to) | |
515 { | |
516 return conversion_loop (one_utf8_to_utf32, cd, from, flen, to); | |
517 } | |
518 | |
519 static bool | |
520 convert_utf16_utf8 (iconv_t cd, const uchar *from, size_t flen, | |
521 struct _cpp_strbuf *to) | |
522 { | |
523 return conversion_loop (one_utf16_to_utf8, cd, from, flen, to); | |
524 } | |
525 | |
526 static bool | |
527 convert_utf32_utf8 (iconv_t cd, const uchar *from, size_t flen, | |
528 struct _cpp_strbuf *to) | |
529 { | |
530 return conversion_loop (one_utf32_to_utf8, cd, from, flen, to); | |
531 } | |
532 | |
533 /* Identity conversion, used when we have no alternative. */ | |
534 static bool | |
535 convert_no_conversion (iconv_t cd ATTRIBUTE_UNUSED, | |
536 const uchar *from, size_t flen, struct _cpp_strbuf *to) | |
537 { | |
538 if (to->len + flen > to->asize) | |
539 { | |
540 to->asize = to->len + flen; | |
541 to->text = XRESIZEVEC (uchar, to->text, to->asize); | |
542 } | |
543 memcpy (to->text + to->len, from, flen); | |
544 to->len += flen; | |
545 return true; | |
546 } | |
547 | |
548 /* And this one uses the system iconv primitive. It's a little | |
549 different, since iconv's interface is a little different. */ | |
550 #if HAVE_ICONV | |
551 | |
552 #define CONVERT_ICONV_GROW_BUFFER \ | |
553 do { \ | |
554 outbytesleft += OUTBUF_BLOCK_SIZE; \ | |
555 to->asize += OUTBUF_BLOCK_SIZE; \ | |
556 to->text = XRESIZEVEC (uchar, to->text, to->asize); \ | |
557 outbuf = (char *)to->text + to->asize - outbytesleft; \ | |
558 } while (0) | |
559 | |
560 static bool | |
561 convert_using_iconv (iconv_t cd, const uchar *from, size_t flen, | |
562 struct _cpp_strbuf *to) | |
563 { | |
564 ICONV_CONST char *inbuf; | |
565 char *outbuf; | |
566 size_t inbytesleft, outbytesleft; | |
567 | |
568 /* Reset conversion descriptor and check that it is valid. */ | |
569 if (iconv (cd, 0, 0, 0, 0) == (size_t)-1) | |
570 return false; | |
571 | |
572 inbuf = (ICONV_CONST char *)from; | |
573 inbytesleft = flen; | |
574 outbuf = (char *)to->text + to->len; | |
575 outbytesleft = to->asize - to->len; | |
576 | |
577 for (;;) | |
578 { | |
579 iconv (cd, &inbuf, &inbytesleft, &outbuf, &outbytesleft); | |
580 if (__builtin_expect (inbytesleft == 0, 1)) | |
581 { | |
582 /* Close out any shift states, returning to the initial state. */ | |
583 if (iconv (cd, 0, 0, &outbuf, &outbytesleft) == (size_t)-1) | |
584 { | |
585 if (errno != E2BIG) | |
586 return false; | |
587 | |
588 CONVERT_ICONV_GROW_BUFFER; | |
589 if (iconv (cd, 0, 0, &outbuf, &outbytesleft) == (size_t)-1) | |
590 return false; | |
591 } | |
592 | |
593 to->len = to->asize - outbytesleft; | |
594 return true; | |
595 } | |
596 if (errno != E2BIG) | |
597 return false; | |
598 | |
599 CONVERT_ICONV_GROW_BUFFER; | |
600 } | |
601 } | |
602 #else | |
603 #define convert_using_iconv 0 /* prevent undefined symbol error below */ | |
604 #endif | |
605 | |
606 /* Arrange for the above custom conversion logic to be used automatically | |
607 when conversion between a suitable pair of character sets is requested. */ | |
608 | |
609 #define APPLY_CONVERSION(CONVERTER, FROM, FLEN, TO) \ | |
610 CONVERTER.func (CONVERTER.cd, FROM, FLEN, TO) | |
611 | |
612 struct conversion | |
613 { | |
614 const char *pair; | |
615 convert_f func; | |
616 iconv_t fake_cd; | |
617 }; | |
618 static const struct conversion conversion_tab[] = { | |
619 { "UTF-8/UTF-32LE", convert_utf8_utf32, (iconv_t)0 }, | |
620 { "UTF-8/UTF-32BE", convert_utf8_utf32, (iconv_t)1 }, | |
621 { "UTF-8/UTF-16LE", convert_utf8_utf16, (iconv_t)0 }, | |
622 { "UTF-8/UTF-16BE", convert_utf8_utf16, (iconv_t)1 }, | |
623 { "UTF-32LE/UTF-8", convert_utf32_utf8, (iconv_t)0 }, | |
624 { "UTF-32BE/UTF-8", convert_utf32_utf8, (iconv_t)1 }, | |
625 { "UTF-16LE/UTF-8", convert_utf16_utf8, (iconv_t)0 }, | |
626 { "UTF-16BE/UTF-8", convert_utf16_utf8, (iconv_t)1 }, | |
627 }; | |
628 | |
629 /* Subroutine of cpp_init_iconv: initialize and return a | |
630 cset_converter structure for conversion from FROM to TO. If | |
631 iconv_open() fails, issue an error and return an identity | |
632 converter. Silently return an identity converter if FROM and TO | |
633 are identical. */ | |
634 static struct cset_converter | |
635 init_iconv_desc (cpp_reader *pfile, const char *to, const char *from) | |
636 { | |
637 struct cset_converter ret; | |
638 char *pair; | |
639 size_t i; | |
640 | |
641 if (!strcasecmp (to, from)) | |
642 { | |
643 ret.func = convert_no_conversion; | |
644 ret.cd = (iconv_t) -1; | |
645 ret.width = -1; | |
646 return ret; | |
647 } | |
648 | |
649 pair = (char *) alloca(strlen(to) + strlen(from) + 2); | |
650 | |
651 strcpy(pair, from); | |
652 strcat(pair, "/"); | |
653 strcat(pair, to); | |
654 for (i = 0; i < ARRAY_SIZE (conversion_tab); i++) | |
655 if (!strcasecmp (pair, conversion_tab[i].pair)) | |
656 { | |
657 ret.func = conversion_tab[i].func; | |
658 ret.cd = conversion_tab[i].fake_cd; | |
659 ret.width = -1; | |
660 return ret; | |
661 } | |
662 | |
663 /* No custom converter - try iconv. */ | |
664 if (HAVE_ICONV) | |
665 { | |
666 ret.func = convert_using_iconv; | |
667 ret.cd = iconv_open (to, from); | |
668 ret.width = -1; | |
669 | |
670 if (ret.cd == (iconv_t) -1) | |
671 { | |
672 if (errno == EINVAL) | |
673 cpp_error (pfile, CPP_DL_ERROR, /* FIXME should be DL_SORRY */ | |
674 "conversion from %s to %s not supported by iconv", | |
675 from, to); | |
676 else | |
677 cpp_errno (pfile, CPP_DL_ERROR, "iconv_open"); | |
678 | |
679 ret.func = convert_no_conversion; | |
680 } | |
681 } | |
682 else | |
683 { | |
684 cpp_error (pfile, CPP_DL_ERROR, /* FIXME: should be DL_SORRY */ | |
685 "no iconv implementation, cannot convert from %s to %s", | |
686 from, to); | |
687 ret.func = convert_no_conversion; | |
688 ret.cd = (iconv_t) -1; | |
689 ret.width = -1; | |
690 } | |
691 return ret; | |
692 } | |
693 | |
694 /* If charset conversion is requested, initialize iconv(3) descriptors | |
695 for conversion from the source character set to the execution | |
696 character sets. If iconv is not present in the C library, and | |
697 conversion is requested, issue an error. */ | |
698 | |
699 void | |
700 cpp_init_iconv (cpp_reader *pfile) | |
701 { | |
702 const char *ncset = CPP_OPTION (pfile, narrow_charset); | |
703 const char *wcset = CPP_OPTION (pfile, wide_charset); | |
704 const char *default_wcset; | |
705 | |
706 bool be = CPP_OPTION (pfile, bytes_big_endian); | |
707 | |
708 if (CPP_OPTION (pfile, wchar_precision) >= 32) | |
709 default_wcset = be ? "UTF-32BE" : "UTF-32LE"; | |
710 else if (CPP_OPTION (pfile, wchar_precision) >= 16) | |
711 default_wcset = be ? "UTF-16BE" : "UTF-16LE"; | |
712 else | |
713 /* This effectively means that wide strings are not supported, | |
714 so don't do any conversion at all. */ | |
715 default_wcset = SOURCE_CHARSET; | |
716 | |
717 if (!ncset) | |
718 ncset = SOURCE_CHARSET; | |
719 if (!wcset) | |
720 wcset = default_wcset; | |
721 | |
722 pfile->narrow_cset_desc = init_iconv_desc (pfile, ncset, SOURCE_CHARSET); | |
723 pfile->narrow_cset_desc.width = CPP_OPTION (pfile, char_precision); | |
724 pfile->char16_cset_desc = init_iconv_desc (pfile, | |
725 be ? "UTF-16BE" : "UTF-16LE", | |
726 SOURCE_CHARSET); | |
727 pfile->char16_cset_desc.width = 16; | |
728 pfile->char32_cset_desc = init_iconv_desc (pfile, | |
729 be ? "UTF-32BE" : "UTF-32LE", | |
730 SOURCE_CHARSET); | |
731 pfile->char32_cset_desc.width = 32; | |
732 pfile->wide_cset_desc = init_iconv_desc (pfile, wcset, SOURCE_CHARSET); | |
733 pfile->wide_cset_desc.width = CPP_OPTION (pfile, wchar_precision); | |
734 } | |
735 | |
736 /* Destroy iconv(3) descriptors set up by cpp_init_iconv, if necessary. */ | |
737 void | |
738 _cpp_destroy_iconv (cpp_reader *pfile) | |
739 { | |
740 if (HAVE_ICONV) | |
741 { | |
742 if (pfile->narrow_cset_desc.func == convert_using_iconv) | |
743 iconv_close (pfile->narrow_cset_desc.cd); | |
744 if (pfile->wide_cset_desc.func == convert_using_iconv) | |
745 iconv_close (pfile->wide_cset_desc.cd); | |
746 } | |
747 } | |
748 | |
749 /* Utility routine for use by a full compiler. C is a character taken | |
750 from the *basic* source character set, encoded in the host's | |
751 execution encoding. Convert it to (the target's) execution | |
752 encoding, and return that value. | |
753 | |
754 Issues an internal error if C's representation in the narrow | |
755 execution character set fails to be a single-byte value (C99 | |
756 5.2.1p3: "The representation of each member of the source and | |
757 execution character sets shall fit in a byte.") May also issue an | |
758 internal error if C fails to be a member of the basic source | |
759 character set (testing this exactly is too hard, especially when | |
760 the host character set is EBCDIC). */ | |
761 cppchar_t | |
762 cpp_host_to_exec_charset (cpp_reader *pfile, cppchar_t c) | |
763 { | |
764 uchar sbuf[1]; | |
765 struct _cpp_strbuf tbuf; | |
766 | |
767 /* This test is merely an approximation, but it suffices to catch | |
768 the most important thing, which is that we don't get handed a | |
769 character outside the unibyte range of the host character set. */ | |
770 if (c > LAST_POSSIBLY_BASIC_SOURCE_CHAR) | |
771 { | |
772 cpp_error (pfile, CPP_DL_ICE, | |
773 "character 0x%lx is not in the basic source character set\n", | |
774 (unsigned long)c); | |
775 return 0; | |
776 } | |
777 | |
778 /* Being a character in the unibyte range of the host character set, | |
779 we can safely splat it into a one-byte buffer and trust that that | |
780 is a well-formed string. */ | |
781 sbuf[0] = c; | |
782 | |
783 /* This should never need to reallocate, but just in case... */ | |
784 tbuf.asize = 1; | |
785 tbuf.text = XNEWVEC (uchar, tbuf.asize); | |
786 tbuf.len = 0; | |
787 | |
788 if (!APPLY_CONVERSION (pfile->narrow_cset_desc, sbuf, 1, &tbuf)) | |
789 { | |
790 cpp_errno (pfile, CPP_DL_ICE, "converting to execution character set"); | |
791 return 0; | |
792 } | |
793 if (tbuf.len != 1) | |
794 { | |
795 cpp_error (pfile, CPP_DL_ICE, | |
796 "character 0x%lx is not unibyte in execution character set", | |
797 (unsigned long)c); | |
798 return 0; | |
799 } | |
800 c = tbuf.text[0]; | |
801 free(tbuf.text); | |
802 return c; | |
803 } | |
804 | |
805 | |
806 | |
807 /* Utility routine that computes a mask of the form 0000...111... with | |
808 WIDTH 1-bits. */ | |
809 static inline size_t | |
810 width_to_mask (size_t width) | |
811 { | |
812 width = MIN (width, BITS_PER_CPPCHAR_T); | |
813 if (width >= CHAR_BIT * sizeof (size_t)) | |
814 return ~(size_t) 0; | |
815 else | |
816 return ((size_t) 1 << width) - 1; | |
817 } | |
818 | |
819 /* A large table of unicode character information. */ | |
820 enum { | |
821 /* Valid in a C99 identifier? */ | |
822 C99 = 1, | |
823 /* Valid in a C99 identifier, but not as the first character? */ | |
824 DIG = 2, | |
825 /* Valid in a C++ identifier? */ | |
826 CXX = 4, | |
827 /* NFC representation is not valid in an identifier? */ | |
828 CID = 8, | |
829 /* Might be valid NFC form? */ | |
830 NFC = 16, | |
831 /* Might be valid NFKC form? */ | |
832 NKC = 32, | |
833 /* Certain preceding characters might make it not valid NFC/NKFC form? */ | |
834 CTX = 64 | |
835 }; | |
836 | |
837 static const struct { | |
838 /* Bitmap of flags above. */ | |
839 unsigned char flags; | |
840 /* Combining class of the character. */ | |
841 unsigned char combine; | |
842 /* Last character in the range described by this entry. */ | |
843 unsigned short end; | |
844 } ucnranges[] = { | |
845 #include "ucnid.h" | |
846 }; | |
847 | |
848 /* Returns 1 if C is valid in an identifier, 2 if C is valid except at | |
849 the start of an identifier, and 0 if C is not valid in an | |
850 identifier. We assume C has already gone through the checks of | |
851 _cpp_valid_ucn. Also update NST for C if returning nonzero. The | |
852 algorithm is a simple binary search on the table defined in | |
853 ucnid.h. */ | |
854 | |
855 static int | |
856 ucn_valid_in_identifier (cpp_reader *pfile, cppchar_t c, | |
857 struct normalize_state *nst) | |
858 { | |
859 int mn, mx, md; | |
860 | |
861 if (c > 0xFFFF) | |
862 return 0; | |
863 | |
864 mn = 0; | |
865 mx = ARRAY_SIZE (ucnranges) - 1; | |
866 while (mx != mn) | |
867 { | |
868 md = (mn + mx) / 2; | |
869 if (c <= ucnranges[md].end) | |
870 mx = md; | |
871 else | |
872 mn = md + 1; | |
873 } | |
874 | |
875 /* When -pedantic, we require the character to have been listed by | |
876 the standard for the current language. Otherwise, we accept the | |
877 union of the acceptable sets for C++98 and C99. */ | |
878 if (! (ucnranges[mn].flags & (C99 | CXX))) | |
879 return 0; | |
880 | |
881 if (CPP_PEDANTIC (pfile) | |
882 && ((CPP_OPTION (pfile, c99) && !(ucnranges[mn].flags & C99)) | |
883 || (CPP_OPTION (pfile, cplusplus) | |
884 && !(ucnranges[mn].flags & CXX)))) | |
885 return 0; | |
886 | |
887 /* Update NST. */ | |
888 if (ucnranges[mn].combine != 0 && ucnranges[mn].combine < nst->prev_class) | |
889 nst->level = normalized_none; | |
890 else if (ucnranges[mn].flags & CTX) | |
891 { | |
892 bool safe; | |
893 cppchar_t p = nst->previous; | |
894 | |
895 /* Easy cases from Bengali, Oriya, Tamil, Jannada, and Malayalam. */ | |
896 if (c == 0x09BE) | |
897 safe = p != 0x09C7; /* Use 09CB instead of 09C7 09BE. */ | |
898 else if (c == 0x0B3E) | |
899 safe = p != 0x0B47; /* Use 0B4B instead of 0B47 0B3E. */ | |
900 else if (c == 0x0BBE) | |
901 safe = p != 0x0BC6 && p != 0x0BC7; /* Use 0BCA/0BCB instead. */ | |
902 else if (c == 0x0CC2) | |
903 safe = p != 0x0CC6; /* Use 0CCA instead of 0CC6 0CC2. */ | |
904 else if (c == 0x0D3E) | |
905 safe = p != 0x0D46 && p != 0x0D47; /* Use 0D4A/0D4B instead. */ | |
906 /* For Hangul, characters in the range AC00-D7A3 are NFC/NFKC, | |
907 and are combined algorithmically from a sequence of the form | |
908 1100-1112 1161-1175 11A8-11C2 | |
909 (if the third is not present, it is treated as 11A7, which is not | |
910 really a valid character). | |
911 Unfortunately, C99 allows (only) the NFC form, but C++ allows | |
912 only the combining characters. */ | |
913 else if (c >= 0x1161 && c <= 0x1175) | |
914 safe = p < 0x1100 || p > 0x1112; | |
915 else if (c >= 0x11A8 && c <= 0x11C2) | |
916 safe = (p < 0xAC00 || p > 0xD7A3 || (p - 0xAC00) % 28 != 0); | |
917 else | |
918 { | |
919 /* Uh-oh, someone updated ucnid.h without updating this code. */ | |
920 cpp_error (pfile, CPP_DL_ICE, "Character %x might not be NFKC", c); | |
921 safe = true; | |
922 } | |
923 if (!safe && c < 0x1161) | |
924 nst->level = normalized_none; | |
925 else if (!safe) | |
926 nst->level = MAX (nst->level, normalized_identifier_C); | |
927 } | |
928 else if (ucnranges[mn].flags & NKC) | |
929 ; | |
930 else if (ucnranges[mn].flags & NFC) | |
931 nst->level = MAX (nst->level, normalized_C); | |
932 else if (ucnranges[mn].flags & CID) | |
933 nst->level = MAX (nst->level, normalized_identifier_C); | |
934 else | |
935 nst->level = normalized_none; | |
936 nst->previous = c; | |
937 nst->prev_class = ucnranges[mn].combine; | |
938 | |
939 /* In C99, UCN digits may not begin identifiers. */ | |
940 if (CPP_OPTION (pfile, c99) && (ucnranges[mn].flags & DIG)) | |
941 return 2; | |
942 | |
943 return 1; | |
944 } | |
945 | |
946 /* [lex.charset]: The character designated by the universal character | |
947 name \UNNNNNNNN is that character whose character short name in | |
948 ISO/IEC 10646 is NNNNNNNN; the character designated by the | |
949 universal character name \uNNNN is that character whose character | |
950 short name in ISO/IEC 10646 is 0000NNNN. If the hexadecimal value | |
951 for a universal character name is less than 0x20 or in the range | |
952 0x7F-0x9F (inclusive), or if the universal character name | |
953 designates a character in the basic source character set, then the | |
954 program is ill-formed. | |
955 | |
956 *PSTR must be preceded by "\u" or "\U"; it is assumed that the | |
957 buffer end is delimited by a non-hex digit. Returns zero if the | |
958 UCN has not been consumed. | |
959 | |
960 Otherwise the nonzero value of the UCN, whether valid or invalid, | |
961 is returned. Diagnostics are emitted for invalid values. PSTR | |
962 is updated to point one beyond the UCN, or to the syntactically | |
963 invalid character. | |
964 | |
965 IDENTIFIER_POS is 0 when not in an identifier, 1 for the start of | |
966 an identifier, or 2 otherwise. */ | |
967 | |
968 cppchar_t | |
969 _cpp_valid_ucn (cpp_reader *pfile, const uchar **pstr, | |
970 const uchar *limit, int identifier_pos, | |
971 struct normalize_state *nst) | |
972 { | |
973 cppchar_t result, c; | |
974 unsigned int length; | |
975 const uchar *str = *pstr; | |
976 const uchar *base = str - 2; | |
977 | |
978 if (!CPP_OPTION (pfile, cplusplus) && !CPP_OPTION (pfile, c99)) | |
979 cpp_error (pfile, CPP_DL_WARNING, | |
980 "universal character names are only valid in C++ and C99"); | |
981 else if (CPP_WTRADITIONAL (pfile) && identifier_pos == 0) | |
982 cpp_error (pfile, CPP_DL_WARNING, | |
983 "the meaning of '\\%c' is different in traditional C", | |
984 (int) str[-1]); | |
985 | |
986 if (str[-1] == 'u') | |
987 length = 4; | |
988 else if (str[-1] == 'U') | |
989 length = 8; | |
990 else | |
991 { | |
992 cpp_error (pfile, CPP_DL_ICE, "In _cpp_valid_ucn but not a UCN"); | |
993 length = 4; | |
994 } | |
995 | |
996 result = 0; | |
997 do | |
998 { | |
999 c = *str; | |
1000 if (!ISXDIGIT (c)) | |
1001 break; | |
1002 str++; | |
1003 result = (result << 4) + hex_value (c); | |
1004 } | |
1005 while (--length && str < limit); | |
1006 | |
1007 /* Partial UCNs are not valid in strings, but decompose into | |
1008 multiple tokens in identifiers, so we can't give a helpful | |
1009 error message in that case. */ | |
1010 if (length && identifier_pos) | |
1011 return 0; | |
1012 | |
1013 *pstr = str; | |
1014 if (length) | |
1015 { | |
1016 cpp_error (pfile, CPP_DL_ERROR, | |
1017 "incomplete universal character name %.*s", | |
1018 (int) (str - base), base); | |
1019 result = 1; | |
1020 } | |
1021 /* The standard permits $, @ and ` to be specified as UCNs. We use | |
1022 hex escapes so that this also works with EBCDIC hosts. */ | |
1023 else if ((result < 0xa0 | |
1024 && (result != 0x24 && result != 0x40 && result != 0x60)) | |
1025 || (result & 0x80000000) | |
1026 || (result >= 0xD800 && result <= 0xDFFF)) | |
1027 { | |
1028 cpp_error (pfile, CPP_DL_ERROR, | |
1029 "%.*s is not a valid universal character", | |
1030 (int) (str - base), base); | |
1031 result = 1; | |
1032 } | |
1033 else if (identifier_pos && result == 0x24 | |
1034 && CPP_OPTION (pfile, dollars_in_ident)) | |
1035 { | |
1036 if (CPP_OPTION (pfile, warn_dollars) && !pfile->state.skipping) | |
1037 { | |
1038 CPP_OPTION (pfile, warn_dollars) = 0; | |
1039 cpp_error (pfile, CPP_DL_PEDWARN, "'$' in identifier or number"); | |
1040 } | |
1041 NORMALIZE_STATE_UPDATE_IDNUM (nst); | |
1042 } | |
1043 else if (identifier_pos) | |
1044 { | |
1045 int validity = ucn_valid_in_identifier (pfile, result, nst); | |
1046 | |
1047 if (validity == 0) | |
1048 cpp_error (pfile, CPP_DL_ERROR, | |
1049 "universal character %.*s is not valid in an identifier", | |
1050 (int) (str - base), base); | |
1051 else if (validity == 2 && identifier_pos == 1) | |
1052 cpp_error (pfile, CPP_DL_ERROR, | |
1053 "universal character %.*s is not valid at the start of an identifier", | |
1054 (int) (str - base), base); | |
1055 } | |
1056 | |
1057 if (result == 0) | |
1058 result = 1; | |
1059 | |
1060 return result; | |
1061 } | |
1062 | |
1063 /* Convert an UCN, pointed to by FROM, to UTF-8 encoding, then translate | |
1064 it to the execution character set and write the result into TBUF. | |
1065 An advanced pointer is returned. Issues all relevant diagnostics. */ | |
1066 static const uchar * | |
1067 convert_ucn (cpp_reader *pfile, const uchar *from, const uchar *limit, | |
1068 struct _cpp_strbuf *tbuf, struct cset_converter cvt) | |
1069 { | |
1070 cppchar_t ucn; | |
1071 uchar buf[6]; | |
1072 uchar *bufp = buf; | |
1073 size_t bytesleft = 6; | |
1074 int rval; | |
1075 struct normalize_state nst = INITIAL_NORMALIZE_STATE; | |
1076 | |
1077 from++; /* Skip u/U. */ | |
1078 ucn = _cpp_valid_ucn (pfile, &from, limit, 0, &nst); | |
1079 | |
1080 rval = one_cppchar_to_utf8 (ucn, &bufp, &bytesleft); | |
1081 if (rval) | |
1082 { | |
1083 errno = rval; | |
1084 cpp_errno (pfile, CPP_DL_ERROR, | |
1085 "converting UCN to source character set"); | |
1086 } | |
1087 else if (!APPLY_CONVERSION (cvt, buf, 6 - bytesleft, tbuf)) | |
1088 cpp_errno (pfile, CPP_DL_ERROR, | |
1089 "converting UCN to execution character set"); | |
1090 | |
1091 return from; | |
1092 } | |
1093 | |
1094 /* Subroutine of convert_hex and convert_oct. N is the representation | |
1095 in the execution character set of a numeric escape; write it into the | |
1096 string buffer TBUF and update the end-of-string pointer therein. WIDE | |
1097 is true if it's a wide string that's being assembled in TBUF. This | |
1098 function issues no diagnostics and never fails. */ | |
1099 static void | |
1100 emit_numeric_escape (cpp_reader *pfile, cppchar_t n, | |
1101 struct _cpp_strbuf *tbuf, struct cset_converter cvt) | |
1102 { | |
1103 size_t width = cvt.width; | |
1104 | |
1105 if (width != CPP_OPTION (pfile, char_precision)) | |
1106 { | |
1107 /* We have to render this into the target byte order, which may not | |
1108 be our byte order. */ | |
1109 bool bigend = CPP_OPTION (pfile, bytes_big_endian); | |
1110 size_t cwidth = CPP_OPTION (pfile, char_precision); | |
1111 size_t cmask = width_to_mask (cwidth); | |
1112 size_t nbwc = width / cwidth; | |
1113 size_t i; | |
1114 size_t off = tbuf->len; | |
1115 cppchar_t c; | |
1116 | |
1117 if (tbuf->len + nbwc > tbuf->asize) | |
1118 { | |
1119 tbuf->asize += OUTBUF_BLOCK_SIZE; | |
1120 tbuf->text = XRESIZEVEC (uchar, tbuf->text, tbuf->asize); | |
1121 } | |
1122 | |
1123 for (i = 0; i < nbwc; i++) | |
1124 { | |
1125 c = n & cmask; | |
1126 n >>= cwidth; | |
1127 tbuf->text[off + (bigend ? nbwc - i - 1 : i)] = c; | |
1128 } | |
1129 tbuf->len += nbwc; | |
1130 } | |
1131 else | |
1132 { | |
1133 /* Note: this code does not handle the case where the target | |
1134 and host have a different number of bits in a byte. */ | |
1135 if (tbuf->len + 1 > tbuf->asize) | |
1136 { | |
1137 tbuf->asize += OUTBUF_BLOCK_SIZE; | |
1138 tbuf->text = XRESIZEVEC (uchar, tbuf->text, tbuf->asize); | |
1139 } | |
1140 tbuf->text[tbuf->len++] = n; | |
1141 } | |
1142 } | |
1143 | |
1144 /* Convert a hexadecimal escape, pointed to by FROM, to the execution | |
1145 character set and write it into the string buffer TBUF. Returns an | |
1146 advanced pointer, and issues diagnostics as necessary. | |
1147 No character set translation occurs; this routine always produces the | |
1148 execution-set character with numeric value equal to the given hex | |
1149 number. You can, e.g. generate surrogate pairs this way. */ | |
1150 static const uchar * | |
1151 convert_hex (cpp_reader *pfile, const uchar *from, const uchar *limit, | |
1152 struct _cpp_strbuf *tbuf, struct cset_converter cvt) | |
1153 { | |
1154 cppchar_t c, n = 0, overflow = 0; | |
1155 int digits_found = 0; | |
1156 size_t width = cvt.width; | |
1157 size_t mask = width_to_mask (width); | |
1158 | |
1159 if (CPP_WTRADITIONAL (pfile)) | |
1160 cpp_error (pfile, CPP_DL_WARNING, | |
1161 "the meaning of '\\x' is different in traditional C"); | |
1162 | |
1163 from++; /* Skip 'x'. */ | |
1164 while (from < limit) | |
1165 { | |
1166 c = *from; | |
1167 if (! hex_p (c)) | |
1168 break; | |
1169 from++; | |
1170 overflow |= n ^ (n << 4 >> 4); | |
1171 n = (n << 4) + hex_value (c); | |
1172 digits_found = 1; | |
1173 } | |
1174 | |
1175 if (!digits_found) | |
1176 { | |
1177 cpp_error (pfile, CPP_DL_ERROR, | |
1178 "\\x used with no following hex digits"); | |
1179 return from; | |
1180 } | |
1181 | |
1182 if (overflow | (n != (n & mask))) | |
1183 { | |
1184 cpp_error (pfile, CPP_DL_PEDWARN, | |
1185 "hex escape sequence out of range"); | |
1186 n &= mask; | |
1187 } | |
1188 | |
1189 emit_numeric_escape (pfile, n, tbuf, cvt); | |
1190 | |
1191 return from; | |
1192 } | |
1193 | |
1194 /* Convert an octal escape, pointed to by FROM, to the execution | |
1195 character set and write it into the string buffer TBUF. Returns an | |
1196 advanced pointer, and issues diagnostics as necessary. | |
1197 No character set translation occurs; this routine always produces the | |
1198 execution-set character with numeric value equal to the given octal | |
1199 number. */ | |
1200 static const uchar * | |
1201 convert_oct (cpp_reader *pfile, const uchar *from, const uchar *limit, | |
1202 struct _cpp_strbuf *tbuf, struct cset_converter cvt) | |
1203 { | |
1204 size_t count = 0; | |
1205 cppchar_t c, n = 0; | |
1206 size_t width = cvt.width; | |
1207 size_t mask = width_to_mask (width); | |
1208 bool overflow = false; | |
1209 | |
1210 while (from < limit && count++ < 3) | |
1211 { | |
1212 c = *from; | |
1213 if (c < '0' || c > '7') | |
1214 break; | |
1215 from++; | |
1216 overflow |= n ^ (n << 3 >> 3); | |
1217 n = (n << 3) + c - '0'; | |
1218 } | |
1219 | |
1220 if (n != (n & mask)) | |
1221 { | |
1222 cpp_error (pfile, CPP_DL_PEDWARN, | |
1223 "octal escape sequence out of range"); | |
1224 n &= mask; | |
1225 } | |
1226 | |
1227 emit_numeric_escape (pfile, n, tbuf, cvt); | |
1228 | |
1229 return from; | |
1230 } | |
1231 | |
1232 /* Convert an escape sequence (pointed to by FROM) to its value on | |
1233 the target, and to the execution character set. Do not scan past | |
1234 LIMIT. Write the converted value into TBUF. Returns an advanced | |
1235 pointer. Handles all relevant diagnostics. */ | |
1236 static const uchar * | |
1237 convert_escape (cpp_reader *pfile, const uchar *from, const uchar *limit, | |
1238 struct _cpp_strbuf *tbuf, struct cset_converter cvt) | |
1239 { | |
1240 /* Values of \a \b \e \f \n \r \t \v respectively. */ | |
1241 #if HOST_CHARSET == HOST_CHARSET_ASCII | |
1242 static const uchar charconsts[] = { 7, 8, 27, 12, 10, 13, 9, 11 }; | |
1243 #elif HOST_CHARSET == HOST_CHARSET_EBCDIC | |
1244 static const uchar charconsts[] = { 47, 22, 39, 12, 21, 13, 5, 11 }; | |
1245 #else | |
1246 #error "unknown host character set" | |
1247 #endif | |
1248 | |
1249 uchar c; | |
1250 | |
1251 c = *from; | |
1252 switch (c) | |
1253 { | |
1254 /* UCNs, hex escapes, and octal escapes are processed separately. */ | |
1255 case 'u': case 'U': | |
1256 return convert_ucn (pfile, from, limit, tbuf, cvt); | |
1257 | |
1258 case 'x': | |
1259 return convert_hex (pfile, from, limit, tbuf, cvt); | |
1260 break; | |
1261 | |
1262 case '0': case '1': case '2': case '3': | |
1263 case '4': case '5': case '6': case '7': | |
1264 return convert_oct (pfile, from, limit, tbuf, cvt); | |
1265 | |
1266 /* Various letter escapes. Get the appropriate host-charset | |
1267 value into C. */ | |
1268 case '\\': case '\'': case '"': case '?': break; | |
1269 | |
1270 case '(': case '{': case '[': case '%': | |
1271 /* '\(', etc, can be used at the beginning of a line in a long | |
1272 string split onto multiple lines with \-newline, to prevent | |
1273 Emacs or other text editors from getting confused. '\%' can | |
1274 be used to prevent SCCS from mangling printf format strings. */ | |
1275 if (CPP_PEDANTIC (pfile)) | |
1276 goto unknown; | |
1277 break; | |
1278 | |
1279 case 'b': c = charconsts[1]; break; | |
1280 case 'f': c = charconsts[3]; break; | |
1281 case 'n': c = charconsts[4]; break; | |
1282 case 'r': c = charconsts[5]; break; | |
1283 case 't': c = charconsts[6]; break; | |
1284 case 'v': c = charconsts[7]; break; | |
1285 | |
1286 case 'a': | |
1287 if (CPP_WTRADITIONAL (pfile)) | |
1288 cpp_error (pfile, CPP_DL_WARNING, | |
1289 "the meaning of '\\a' is different in traditional C"); | |
1290 c = charconsts[0]; | |
1291 break; | |
1292 | |
1293 case 'e': case 'E': | |
1294 if (CPP_PEDANTIC (pfile)) | |
1295 cpp_error (pfile, CPP_DL_PEDWARN, | |
1296 "non-ISO-standard escape sequence, '\\%c'", (int) c); | |
1297 c = charconsts[2]; | |
1298 break; | |
1299 | |
1300 default: | |
1301 unknown: | |
1302 if (ISGRAPH (c)) | |
1303 cpp_error (pfile, CPP_DL_PEDWARN, | |
1304 "unknown escape sequence '\\%c'", (int) c); | |
1305 else | |
1306 { | |
1307 /* diagnostic.c does not support "%03o". When it does, this | |
1308 code can use %03o directly in the diagnostic again. */ | |
1309 char buf[32]; | |
1310 sprintf(buf, "%03o", (int) c); | |
1311 cpp_error (pfile, CPP_DL_PEDWARN, | |
1312 "unknown escape sequence: '\\%s'", buf); | |
1313 } | |
1314 } | |
1315 | |
1316 /* Now convert what we have to the execution character set. */ | |
1317 if (!APPLY_CONVERSION (cvt, &c, 1, tbuf)) | |
1318 cpp_errno (pfile, CPP_DL_ERROR, | |
1319 "converting escape sequence to execution character set"); | |
1320 | |
1321 return from + 1; | |
1322 } | |
1323 | |
1324 /* TYPE is a token type. The return value is the conversion needed to | |
1325 convert from source to execution character set for the given type. */ | |
1326 static struct cset_converter | |
1327 converter_for_type (cpp_reader *pfile, enum cpp_ttype type) | |
1328 { | |
1329 switch (type) | |
1330 { | |
1331 default: | |
1332 return pfile->narrow_cset_desc; | |
1333 case CPP_CHAR16: | |
1334 case CPP_STRING16: | |
1335 return pfile->char16_cset_desc; | |
1336 case CPP_CHAR32: | |
1337 case CPP_STRING32: | |
1338 return pfile->char32_cset_desc; | |
1339 case CPP_WCHAR: | |
1340 case CPP_WSTRING: | |
1341 return pfile->wide_cset_desc; | |
1342 } | |
1343 } | |
1344 | |
1345 /* FROM is an array of cpp_string structures of length COUNT. These | |
1346 are to be converted from the source to the execution character set, | |
1347 escape sequences translated, and finally all are to be | |
1348 concatenated. WIDE indicates whether or not to produce a wide | |
1349 string. The result is written into TO. Returns true for success, | |
1350 false for failure. */ | |
1351 bool | |
1352 cpp_interpret_string (cpp_reader *pfile, const cpp_string *from, size_t count, | |
1353 cpp_string *to, enum cpp_ttype type) | |
1354 { | |
1355 struct _cpp_strbuf tbuf; | |
1356 const uchar *p, *base, *limit; | |
1357 size_t i; | |
1358 struct cset_converter cvt = converter_for_type (pfile, type); | |
1359 | |
1360 tbuf.asize = MAX (OUTBUF_BLOCK_SIZE, from->len); | |
1361 tbuf.text = XNEWVEC (uchar, tbuf.asize); | |
1362 tbuf.len = 0; | |
1363 | |
1364 for (i = 0; i < count; i++) | |
1365 { | |
1366 p = from[i].text; | |
1367 if (*p == 'L' || *p == 'u' || *p == 'U') p++; | |
1368 p++; /* Skip leading quote. */ | |
1369 limit = from[i].text + from[i].len - 1; /* Skip trailing quote. */ | |
1370 | |
1371 for (;;) | |
1372 { | |
1373 base = p; | |
1374 while (p < limit && *p != '\\') | |
1375 p++; | |
1376 if (p > base) | |
1377 { | |
1378 /* We have a run of normal characters; these can be fed | |
1379 directly to convert_cset. */ | |
1380 if (!APPLY_CONVERSION (cvt, base, p - base, &tbuf)) | |
1381 goto fail; | |
1382 } | |
1383 if (p == limit) | |
1384 break; | |
1385 | |
1386 p = convert_escape (pfile, p + 1, limit, &tbuf, cvt); | |
1387 } | |
1388 } | |
1389 /* NUL-terminate the 'to' buffer and translate it to a cpp_string | |
1390 structure. */ | |
1391 emit_numeric_escape (pfile, 0, &tbuf, cvt); | |
1392 tbuf.text = XRESIZEVEC (uchar, tbuf.text, tbuf.len); | |
1393 to->text = tbuf.text; | |
1394 to->len = tbuf.len; | |
1395 return true; | |
1396 | |
1397 fail: | |
1398 cpp_errno (pfile, CPP_DL_ERROR, "converting to execution character set"); | |
1399 free (tbuf.text); | |
1400 return false; | |
1401 } | |
1402 | |
1403 /* Subroutine of do_line and do_linemarker. Convert escape sequences | |
1404 in a string, but do not perform character set conversion. */ | |
1405 bool | |
1406 cpp_interpret_string_notranslate (cpp_reader *pfile, const cpp_string *from, | |
1407 size_t count, cpp_string *to, | |
1408 enum cpp_ttype type ATTRIBUTE_UNUSED) | |
1409 { | |
1410 struct cset_converter save_narrow_cset_desc = pfile->narrow_cset_desc; | |
1411 bool retval; | |
1412 | |
1413 pfile->narrow_cset_desc.func = convert_no_conversion; | |
1414 pfile->narrow_cset_desc.cd = (iconv_t) -1; | |
1415 pfile->narrow_cset_desc.width = CPP_OPTION (pfile, char_precision); | |
1416 | |
1417 retval = cpp_interpret_string (pfile, from, count, to, CPP_STRING); | |
1418 | |
1419 pfile->narrow_cset_desc = save_narrow_cset_desc; | |
1420 return retval; | |
1421 } | |
1422 | |
1423 | |
1424 /* Subroutine of cpp_interpret_charconst which performs the conversion | |
1425 to a number, for narrow strings. STR is the string structure returned | |
1426 by cpp_interpret_string. PCHARS_SEEN and UNSIGNEDP are as for | |
1427 cpp_interpret_charconst. */ | |
1428 static cppchar_t | |
1429 narrow_str_to_charconst (cpp_reader *pfile, cpp_string str, | |
1430 unsigned int *pchars_seen, int *unsignedp) | |
1431 { | |
1432 size_t width = CPP_OPTION (pfile, char_precision); | |
1433 size_t max_chars = CPP_OPTION (pfile, int_precision) / width; | |
1434 size_t mask = width_to_mask (width); | |
1435 size_t i; | |
1436 cppchar_t result, c; | |
1437 bool unsigned_p; | |
1438 | |
1439 /* The value of a multi-character character constant, or a | |
1440 single-character character constant whose representation in the | |
1441 execution character set is more than one byte long, is | |
1442 implementation defined. This implementation defines it to be the | |
1443 number formed by interpreting the byte sequence in memory as a | |
1444 big-endian binary number. If overflow occurs, the high bytes are | |
1445 lost, and a warning is issued. | |
1446 | |
1447 We don't want to process the NUL terminator handed back by | |
1448 cpp_interpret_string. */ | |
1449 result = 0; | |
1450 for (i = 0; i < str.len - 1; i++) | |
1451 { | |
1452 c = str.text[i] & mask; | |
1453 if (width < BITS_PER_CPPCHAR_T) | |
1454 result = (result << width) | c; | |
1455 else | |
1456 result = c; | |
1457 } | |
1458 | |
1459 if (i > max_chars) | |
1460 { | |
1461 i = max_chars; | |
1462 cpp_error (pfile, CPP_DL_WARNING, | |
1463 "character constant too long for its type"); | |
1464 } | |
1465 else if (i > 1 && CPP_OPTION (pfile, warn_multichar)) | |
1466 cpp_error (pfile, CPP_DL_WARNING, "multi-character character constant"); | |
1467 | |
1468 /* Multichar constants are of type int and therefore signed. */ | |
1469 if (i > 1) | |
1470 unsigned_p = 0; | |
1471 else | |
1472 unsigned_p = CPP_OPTION (pfile, unsigned_char); | |
1473 | |
1474 /* Truncate the constant to its natural width, and simultaneously | |
1475 sign- or zero-extend to the full width of cppchar_t. | |
1476 For single-character constants, the value is WIDTH bits wide. | |
1477 For multi-character constants, the value is INT_PRECISION bits wide. */ | |
1478 if (i > 1) | |
1479 width = CPP_OPTION (pfile, int_precision); | |
1480 if (width < BITS_PER_CPPCHAR_T) | |
1481 { | |
1482 mask = ((cppchar_t) 1 << width) - 1; | |
1483 if (unsigned_p || !(result & (1 << (width - 1)))) | |
1484 result &= mask; | |
1485 else | |
1486 result |= ~mask; | |
1487 } | |
1488 *pchars_seen = i; | |
1489 *unsignedp = unsigned_p; | |
1490 return result; | |
1491 } | |
1492 | |
1493 /* Subroutine of cpp_interpret_charconst which performs the conversion | |
1494 to a number, for wide strings. STR is the string structure returned | |
1495 by cpp_interpret_string. PCHARS_SEEN and UNSIGNEDP are as for | |
1496 cpp_interpret_charconst. TYPE is the token type. */ | |
1497 static cppchar_t | |
1498 wide_str_to_charconst (cpp_reader *pfile, cpp_string str, | |
1499 unsigned int *pchars_seen, int *unsignedp, | |
1500 enum cpp_ttype type) | |
1501 { | |
1502 bool bigend = CPP_OPTION (pfile, bytes_big_endian); | |
1503 size_t width = converter_for_type (pfile, type).width; | |
1504 size_t cwidth = CPP_OPTION (pfile, char_precision); | |
1505 size_t mask = width_to_mask (width); | |
1506 size_t cmask = width_to_mask (cwidth); | |
1507 size_t nbwc = width / cwidth; | |
1508 size_t off, i; | |
1509 cppchar_t result = 0, c; | |
1510 | |
1511 /* This is finicky because the string is in the target's byte order, | |
1512 which may not be our byte order. Only the last character, ignoring | |
1513 the NUL terminator, is relevant. */ | |
1514 off = str.len - (nbwc * 2); | |
1515 result = 0; | |
1516 for (i = 0; i < nbwc; i++) | |
1517 { | |
1518 c = bigend ? str.text[off + i] : str.text[off + nbwc - i - 1]; | |
1519 result = (result << cwidth) | (c & cmask); | |
1520 } | |
1521 | |
1522 /* Wide character constants have type wchar_t, and a single | |
1523 character exactly fills a wchar_t, so a multi-character wide | |
1524 character constant is guaranteed to overflow. */ | |
1525 if (str.len > nbwc * 2) | |
1526 cpp_error (pfile, CPP_DL_WARNING, | |
1527 "character constant too long for its type"); | |
1528 | |
1529 /* Truncate the constant to its natural width, and simultaneously | |
1530 sign- or zero-extend to the full width of cppchar_t. */ | |
1531 if (width < BITS_PER_CPPCHAR_T) | |
1532 { | |
1533 if (type == CPP_CHAR16 || type == CPP_CHAR32 | |
1534 || CPP_OPTION (pfile, unsigned_wchar) | |
1535 || !(result & (1 << (width - 1)))) | |
1536 result &= mask; | |
1537 else | |
1538 result |= ~mask; | |
1539 } | |
1540 | |
1541 if (type == CPP_CHAR16 || type == CPP_CHAR32 | |
1542 || CPP_OPTION (pfile, unsigned_wchar)) | |
1543 *unsignedp = 1; | |
1544 else | |
1545 *unsignedp = 0; | |
1546 | |
1547 *pchars_seen = 1; | |
1548 return result; | |
1549 } | |
1550 | |
1551 /* Interpret a (possibly wide) character constant in TOKEN. | |
1552 PCHARS_SEEN points to a variable that is filled in with the number | |
1553 of characters seen, and UNSIGNEDP to a variable that indicates | |
1554 whether the result has signed type. */ | |
1555 cppchar_t | |
1556 cpp_interpret_charconst (cpp_reader *pfile, const cpp_token *token, | |
1557 unsigned int *pchars_seen, int *unsignedp) | |
1558 { | |
1559 cpp_string str = { 0, 0 }; | |
1560 bool wide = (token->type != CPP_CHAR); | |
1561 cppchar_t result; | |
1562 | |
1563 /* an empty constant will appear as L'', u'', U'' or '' */ | |
1564 if (token->val.str.len == (size_t) (2 + wide)) | |
1565 { | |
1566 cpp_error (pfile, CPP_DL_ERROR, "empty character constant"); | |
1567 return 0; | |
1568 } | |
1569 else if (!cpp_interpret_string (pfile, &token->val.str, 1, &str, token->type)) | |
1570 return 0; | |
1571 | |
1572 if (wide) | |
1573 result = wide_str_to_charconst (pfile, str, pchars_seen, unsignedp, | |
1574 token->type); | |
1575 else | |
1576 result = narrow_str_to_charconst (pfile, str, pchars_seen, unsignedp); | |
1577 | |
1578 if (str.text != token->val.str.text) | |
1579 free ((void *)str.text); | |
1580 | |
1581 return result; | |
1582 } | |
1583 | |
1584 /* Convert an identifier denoted by ID and LEN, which might contain | |
1585 UCN escapes, to the source character set, either UTF-8 or | |
1586 UTF-EBCDIC. Assumes that the identifier is actually a valid identifier. */ | |
1587 cpp_hashnode * | |
1588 _cpp_interpret_identifier (cpp_reader *pfile, const uchar *id, size_t len) | |
1589 { | |
1590 /* It turns out that a UCN escape always turns into fewer characters | |
1591 than the escape itself, so we can allocate a temporary in advance. */ | |
1592 uchar * buf = (uchar *) alloca (len + 1); | |
1593 uchar * bufp = buf; | |
1594 size_t idp; | |
1595 | |
1596 for (idp = 0; idp < len; idp++) | |
1597 if (id[idp] != '\\') | |
1598 *bufp++ = id[idp]; | |
1599 else | |
1600 { | |
1601 unsigned length = id[idp+1] == 'u' ? 4 : 8; | |
1602 cppchar_t value = 0; | |
1603 size_t bufleft = len - (bufp - buf); | |
1604 int rval; | |
1605 | |
1606 idp += 2; | |
1607 while (length && idp < len && ISXDIGIT (id[idp])) | |
1608 { | |
1609 value = (value << 4) + hex_value (id[idp]); | |
1610 idp++; | |
1611 length--; | |
1612 } | |
1613 idp--; | |
1614 | |
1615 /* Special case for EBCDIC: if the identifier contains | |
1616 a '$' specified using a UCN, translate it to EBCDIC. */ | |
1617 if (value == 0x24) | |
1618 { | |
1619 *bufp++ = '$'; | |
1620 continue; | |
1621 } | |
1622 | |
1623 rval = one_cppchar_to_utf8 (value, &bufp, &bufleft); | |
1624 if (rval) | |
1625 { | |
1626 errno = rval; | |
1627 cpp_errno (pfile, CPP_DL_ERROR, | |
1628 "converting UCN to source character set"); | |
1629 break; | |
1630 } | |
1631 } | |
1632 | |
1633 return CPP_HASHNODE (ht_lookup (pfile->hash_table, | |
1634 buf, bufp - buf, HT_ALLOC)); | |
1635 } | |
1636 | |
1637 /* Convert an input buffer (containing the complete contents of one | |
1638 source file) from INPUT_CHARSET to the source character set. INPUT | |
1639 points to the input buffer, SIZE is its allocated size, and LEN is | |
1640 the length of the meaningful data within the buffer. The | |
1641 translated buffer is returned, *ST_SIZE is set to the length of | |
1642 the meaningful data within the translated buffer, and *BUFFER_START | |
1643 is set to the start of the returned buffer. *BUFFER_START may | |
1644 differ from the return value in the case of a BOM or other ignored | |
1645 marker information. | |
1646 | |
1647 INPUT is expected to have been allocated with xmalloc. This | |
1648 function will either set *BUFFER_START to INPUT, or free it and set | |
1649 *BUFFER_START to a pointer to another xmalloc-allocated block of | |
1650 memory. */ | |
1651 uchar * | |
1652 _cpp_convert_input (cpp_reader *pfile, const char *input_charset, | |
1653 uchar *input, size_t size, size_t len, | |
1654 const unsigned char **buffer_start, off_t *st_size) | |
1655 { | |
1656 struct cset_converter input_cset; | |
1657 struct _cpp_strbuf to; | |
1658 unsigned char *buffer; | |
1659 | |
1660 input_cset = init_iconv_desc (pfile, SOURCE_CHARSET, input_charset); | |
1661 if (input_cset.func == convert_no_conversion) | |
1662 { | |
1663 to.text = input; | |
1664 to.asize = size; | |
1665 to.len = len; | |
1666 } | |
1667 else | |
1668 { | |
1669 to.asize = MAX (65536, len); | |
1670 to.text = XNEWVEC (uchar, to.asize); | |
1671 to.len = 0; | |
1672 | |
1673 if (!APPLY_CONVERSION (input_cset, input, len, &to)) | |
1674 cpp_error (pfile, CPP_DL_ERROR, | |
1675 "failure to convert %s to %s", | |
1676 CPP_OPTION (pfile, input_charset), SOURCE_CHARSET); | |
1677 | |
1678 free (input); | |
1679 } | |
1680 | |
1681 /* Clean up the mess. */ | |
1682 if (input_cset.func == convert_using_iconv) | |
1683 iconv_close (input_cset.cd); | |
1684 | |
1685 /* Resize buffer if we allocated substantially too much, or if we | |
1686 haven't enough space for the \n-terminator. */ | |
1687 if (to.len + 4096 < to.asize || to.len >= to.asize) | |
1688 to.text = XRESIZEVEC (uchar, to.text, to.len + 1); | |
1689 | |
1690 /* If the file is using old-school Mac line endings (\r only), | |
1691 terminate with another \r, not an \n, so that we do not mistake | |
1692 the \r\n sequence for a single DOS line ending and erroneously | |
1693 issue the "No newline at end of file" diagnostic. */ | |
1694 if (to.len && to.text[to.len - 1] == '\r') | |
1695 to.text[to.len] = '\r'; | |
1696 else | |
1697 to.text[to.len] = '\n'; | |
1698 | |
1699 buffer = to.text; | |
1700 *st_size = to.len; | |
1701 #if HOST_CHARSET == HOST_CHARSET_ASCII | |
1702 /* The HOST_CHARSET test just above ensures that the source charset | |
1703 is UTF-8. So, ignore a UTF-8 BOM if we see one. Note that | |
1704 glib'c UTF-8 iconv() provider (as of glibc 2.7) does not ignore a | |
1705 BOM -- however, even if it did, we would still need this code due | |
1706 to the 'convert_no_conversion' case. */ | |
1707 if (to.len >= 3 && to.text[0] == 0xef && to.text[1] == 0xbb | |
1708 && to.text[2] == 0xbf) | |
1709 { | |
1710 *st_size -= 3; | |
1711 buffer += 3; | |
1712 } | |
1713 #endif | |
1714 | |
1715 *buffer_start = to.text; | |
1716 return buffer; | |
1717 } | |
1718 | |
1719 /* Decide on the default encoding to assume for input files. */ | |
1720 const char * | |
1721 _cpp_default_encoding (void) | |
1722 { | |
1723 const char *current_encoding = NULL; | |
1724 | |
1725 /* We disable this because the default codeset is 7-bit ASCII on | |
1726 most platforms, and this causes conversion failures on every | |
1727 file in GCC that happens to have one of the upper 128 characters | |
1728 in it -- most likely, as part of the name of a contributor. | |
1729 We should definitely recognize in-band markers of file encoding, | |
1730 like: | |
1731 - the appropriate Unicode byte-order mark (FE FF) to recognize | |
1732 UTF16 and UCS4 (in both big-endian and little-endian flavors) | |
1733 and UTF8 | |
1734 - a "#i", "#d", "/ *", "//", " #p" or "#p" (for #pragma) to | |
1735 distinguish ASCII and EBCDIC. | |
1736 - now we can parse something like "#pragma GCC encoding <xyz> | |
1737 on the first line, or even Emacs/VIM's mode line tags (there's | |
1738 a problem here in that VIM uses the last line, and Emacs has | |
1739 its more elaborate "local variables" convention). | |
1740 - investigate whether Java has another common convention, which | |
1741 would be friendly to support. | |
1742 (Zack Weinberg and Paolo Bonzini, May 20th 2004) */ | |
1743 #if defined (HAVE_LOCALE_H) && defined (HAVE_LANGINFO_CODESET) && 0 | |
1744 setlocale (LC_CTYPE, ""); | |
1745 current_encoding = nl_langinfo (CODESET); | |
1746 #endif | |
1747 if (current_encoding == NULL || *current_encoding == '\0') | |
1748 current_encoding = SOURCE_CHARSET; | |
1749 | |
1750 return current_encoding; | |
1751 } |