Mercurial > hg > CbC > CbC_gcc
annotate libcpp/charset.c @ 158:494b0b89df80 default tip
...
author | Shinji KONO <kono@ie.u-ryukyu.ac.jp> |
---|---|
date | Mon, 25 May 2020 18:13:55 +0900 |
parents | 1830386684a0 |
children |
rev | line source |
---|---|
0 | 1 /* CPP Library - charsets |
145 | 2 Copyright (C) 1998-2020 Free Software Foundation, Inc. |
0 | 3 |
4 Broken out of c-lex.c Apr 2003, adding valid C99 UCN ranges. | |
5 | |
6 This program is free software; you can redistribute it and/or modify it | |
7 under the terms of the GNU General Public License as published by the | |
8 Free Software Foundation; either version 3, or (at your option) any | |
9 later version. | |
10 | |
11 This program is distributed in the hope that it will be useful, | |
12 but WITHOUT ANY WARRANTY; without even the implied warranty of | |
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
14 GNU General Public License for more details. | |
15 | |
16 You should have received a copy of the GNU General Public License | |
17 along with this program; see the file COPYING3. If not see | |
18 <http://www.gnu.org/licenses/>. */ | |
19 | |
20 #include "config.h" | |
21 #include "system.h" | |
22 #include "cpplib.h" | |
23 #include "internal.h" | |
24 | |
25 /* Character set handling for C-family languages. | |
26 | |
27 Terminological note: In what follows, "charset" or "character set" | |
28 will be taken to mean both an abstract set of characters and an | |
29 encoding for that set. | |
30 | |
31 The C99 standard discusses two character sets: source and execution. | |
32 The source character set is used for internal processing in translation | |
33 phases 1 through 4; the execution character set is used thereafter. | |
34 Both are required by 5.2.1.2p1 to be multibyte encodings, not wide | |
35 character encodings (see 3.7.2, 3.7.3 for the standardese meanings | |
36 of these terms). Furthermore, the "basic character set" (listed in | |
37 5.2.1p3) is to be encoded in each with values one byte wide, and is | |
38 to appear in the initial shift state. | |
39 | |
40 It is not explicitly mentioned, but there is also a "wide execution | |
41 character set" used to encode wide character constants and wide | |
42 string literals; this is supposed to be the result of applying the | |
43 standard library function mbstowcs() to an equivalent narrow string | |
44 (6.4.5p5). However, the behavior of hexadecimal and octal | |
45 \-escapes is at odds with this; they are supposed to be translated | |
46 directly to wchar_t values (6.4.4.4p5,6). | |
47 | |
48 The source character set is not necessarily the character set used | |
49 to encode physical source files on disk; translation phase 1 converts | |
50 from whatever that encoding is to the source character set. | |
51 | |
52 The presence of universal character names in C99 (6.4.3 et seq.) | |
53 forces the source character set to be isomorphic to ISO 10646, | |
54 that is, Unicode. There is no such constraint on the execution | |
55 character set; note also that the conversion from source to | |
56 execution character set does not occur for identifiers (5.1.1.2p1#5). | |
57 | |
58 For convenience of implementation, the source character set's | |
59 encoding of the basic character set should be identical to the | |
60 execution character set OF THE HOST SYSTEM's encoding of the basic | |
61 character set, and it should not be a state-dependent encoding. | |
62 | |
63 cpplib uses UTF-8 or UTF-EBCDIC for the source character set, | |
64 depending on whether the host is based on ASCII or EBCDIC (see | |
65 respectively Unicode section 2.3/ISO10646 Amendment 2, and Unicode | |
66 Technical Report #16). With limited exceptions, it relies on the | |
67 system library's iconv() primitive to do charset conversion | |
68 (specified in SUSv2). */ | |
69 | |
70 #if !HAVE_ICONV | |
71 /* Make certain that the uses of iconv(), iconv_open(), iconv_close() | |
72 below, which are guarded only by if statements with compile-time | |
73 constant conditions, do not cause link errors. */ | |
74 #define iconv_open(x, y) (errno = EINVAL, (iconv_t)-1) | |
75 #define iconv(a,b,c,d,e) (errno = EINVAL, (size_t)-1) | |
76 #define iconv_close(x) (void)0 | |
77 #define ICONV_CONST | |
78 #endif | |
79 | |
80 #if HOST_CHARSET == HOST_CHARSET_ASCII | |
81 #define SOURCE_CHARSET "UTF-8" | |
82 #define LAST_POSSIBLY_BASIC_SOURCE_CHAR 0x7e | |
83 #elif HOST_CHARSET == HOST_CHARSET_EBCDIC | |
84 #define SOURCE_CHARSET "UTF-EBCDIC" | |
85 #define LAST_POSSIBLY_BASIC_SOURCE_CHAR 0xFF | |
86 #else | |
87 #error "Unrecognized basic host character set" | |
88 #endif | |
89 | |
90 #ifndef EILSEQ | |
91 #define EILSEQ EINVAL | |
92 #endif | |
93 | |
94 /* This structure is used for a resizable string buffer throughout. */ | |
95 /* Don't call it strbuf, as that conflicts with unistd.h on systems | |
96 such as DYNIX/ptx where unistd.h includes stropts.h. */ | |
97 struct _cpp_strbuf | |
98 { | |
99 uchar *text; | |
100 size_t asize; | |
101 size_t len; | |
102 }; | |
103 | |
104 /* This is enough to hold any string that fits on a single 80-column | |
105 line, even if iconv quadruples its size (e.g. conversion from | |
106 ASCII to UTF-32) rounded up to a power of two. */ | |
107 #define OUTBUF_BLOCK_SIZE 256 | |
108 | |
109 /* Conversions between UTF-8 and UTF-16/32 are implemented by custom | |
110 logic. This is because a depressing number of systems lack iconv, | |
111 or have have iconv libraries that do not do these conversions, so | |
112 we need a fallback implementation for them. To ensure the fallback | |
113 doesn't break due to neglect, it is used on all systems. | |
114 | |
115 UTF-32 encoding is nice and simple: a four-byte binary number, | |
116 constrained to the range 00000000-7FFFFFFF to avoid questions of | |
117 signedness. We do have to cope with big- and little-endian | |
118 variants. | |
119 | |
120 UTF-16 encoding uses two-byte binary numbers, again in big- and | |
121 little-endian variants, for all values in the 00000000-0000FFFF | |
122 range. Values in the 00010000-0010FFFF range are encoded as pairs | |
123 of two-byte numbers, called "surrogate pairs": given a number S in | |
124 this range, it is mapped to a pair (H, L) as follows: | |
125 | |
126 H = (S - 0x10000) / 0x400 + 0xD800 | |
127 L = (S - 0x10000) % 0x400 + 0xDC00 | |
128 | |
129 Two-byte values in the D800...DFFF range are ill-formed except as a | |
130 component of a surrogate pair. Even if the encoding within a | |
131 two-byte value is little-endian, the H member of the surrogate pair | |
132 comes first. | |
133 | |
134 There is no way to encode values in the 00110000-7FFFFFFF range, | |
135 which is not currently a problem as there are no assigned code | |
136 points in that range; however, the author expects that it will | |
137 eventually become necessary to abandon UTF-16 due to this | |
138 limitation. Note also that, because of these pairs, UTF-16 does | |
139 not meet the requirements of the C standard for a wide character | |
140 encoding (see 3.7.3 and 6.4.4.4p11). | |
141 | |
142 UTF-8 encoding looks like this: | |
143 | |
144 value range encoded as | |
145 00000000-0000007F 0xxxxxxx | |
146 00000080-000007FF 110xxxxx 10xxxxxx | |
147 00000800-0000FFFF 1110xxxx 10xxxxxx 10xxxxxx | |
148 00010000-001FFFFF 11110xxx 10xxxxxx 10xxxxxx 10xxxxxx | |
149 00200000-03FFFFFF 111110xx 10xxxxxx 10xxxxxx 10xxxxxx 10xxxxxx | |
150 04000000-7FFFFFFF 1111110x 10xxxxxx 10xxxxxx 10xxxxxx 10xxxxxx 10xxxxxx | |
151 | |
152 Values in the 0000D800 ... 0000DFFF range (surrogates) are invalid, | |
153 which means that three-byte sequences ED xx yy, with A0 <= xx <= BF, | |
154 never occur. Note also that any value that can be encoded by a | |
155 given row of the table can also be encoded by all successive rows, | |
156 but this is not done; only the shortest possible encoding for any | |
157 given value is valid. For instance, the character 07C0 could be | |
158 encoded as any of DF 80, E0 9F 80, F0 80 9F 80, F8 80 80 9F 80, or | |
159 FC 80 80 80 9F 80. Only the first is valid. | |
160 | |
161 An implementation note: the transformation from UTF-16 to UTF-8, or | |
162 vice versa, is easiest done by using UTF-32 as an intermediary. */ | |
163 | |
164 /* Internal primitives which go from an UTF-8 byte stream to native-endian | |
165 UTF-32 in a cppchar_t, or vice versa; this avoids an extra marshal/unmarshal | |
166 operation in several places below. */ | |
167 static inline int | |
168 one_utf8_to_cppchar (const uchar **inbufp, size_t *inbytesleftp, | |
169 cppchar_t *cp) | |
170 { | |
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
171 static const uchar masks[6] = { 0x7F, 0x1F, 0x0F, 0x07, 0x03, 0x01 }; |
0 | 172 static const uchar patns[6] = { 0x00, 0xC0, 0xE0, 0xF0, 0xF8, 0xFC }; |
173 | |
174 cppchar_t c; | |
175 const uchar *inbuf = *inbufp; | |
176 size_t nbytes, i; | |
177 | |
178 if (*inbytesleftp < 1) | |
179 return EINVAL; | |
180 | |
181 c = *inbuf; | |
182 if (c < 0x80) | |
183 { | |
184 *cp = c; | |
185 *inbytesleftp -= 1; | |
186 *inbufp += 1; | |
187 return 0; | |
188 } | |
189 | |
190 /* The number of leading 1-bits in the first byte indicates how many | |
191 bytes follow. */ | |
192 for (nbytes = 2; nbytes < 7; nbytes++) | |
193 if ((c & ~masks[nbytes-1]) == patns[nbytes-1]) | |
194 goto found; | |
195 return EILSEQ; | |
196 found: | |
197 | |
198 if (*inbytesleftp < nbytes) | |
199 return EINVAL; | |
200 | |
201 c = (c & masks[nbytes-1]); | |
202 inbuf++; | |
203 for (i = 1; i < nbytes; i++) | |
204 { | |
205 cppchar_t n = *inbuf++; | |
206 if ((n & 0xC0) != 0x80) | |
207 return EILSEQ; | |
208 c = ((c << 6) + (n & 0x3F)); | |
209 } | |
210 | |
211 /* Make sure the shortest possible encoding was used. */ | |
212 if (c <= 0x7F && nbytes > 1) return EILSEQ; | |
213 if (c <= 0x7FF && nbytes > 2) return EILSEQ; | |
214 if (c <= 0xFFFF && nbytes > 3) return EILSEQ; | |
215 if (c <= 0x1FFFFF && nbytes > 4) return EILSEQ; | |
216 if (c <= 0x3FFFFFF && nbytes > 5) return EILSEQ; | |
217 | |
218 /* Make sure the character is valid. */ | |
219 if (c > 0x7FFFFFFF || (c >= 0xD800 && c <= 0xDFFF)) return EILSEQ; | |
220 | |
221 *cp = c; | |
222 *inbufp = inbuf; | |
223 *inbytesleftp -= nbytes; | |
224 return 0; | |
225 } | |
226 | |
227 static inline int | |
228 one_cppchar_to_utf8 (cppchar_t c, uchar **outbufp, size_t *outbytesleftp) | |
229 { | |
230 static const uchar masks[6] = { 0x00, 0xC0, 0xE0, 0xF0, 0xF8, 0xFC }; | |
231 static const uchar limits[6] = { 0x80, 0xE0, 0xF0, 0xF8, 0xFC, 0xFE }; | |
232 size_t nbytes; | |
233 uchar buf[6], *p = &buf[6]; | |
234 uchar *outbuf = *outbufp; | |
235 | |
236 nbytes = 1; | |
237 if (c < 0x80) | |
238 *--p = c; | |
239 else | |
240 { | |
241 do | |
242 { | |
243 *--p = ((c & 0x3F) | 0x80); | |
244 c >>= 6; | |
245 nbytes++; | |
246 } | |
247 while (c >= 0x3F || (c & limits[nbytes-1])); | |
248 *--p = (c | masks[nbytes-1]); | |
249 } | |
250 | |
251 if (*outbytesleftp < nbytes) | |
252 return E2BIG; | |
253 | |
254 while (p < &buf[6]) | |
255 *outbuf++ = *p++; | |
256 *outbytesleftp -= nbytes; | |
257 *outbufp = outbuf; | |
258 return 0; | |
259 } | |
260 | |
261 /* The following four functions transform one character between the two | |
262 encodings named in the function name. All have the signature | |
263 int (*)(iconv_t bigend, const uchar **inbufp, size_t *inbytesleftp, | |
264 uchar **outbufp, size_t *outbytesleftp) | |
265 | |
266 BIGEND must have the value 0 or 1, coerced to (iconv_t); it is | |
267 interpreted as a boolean indicating whether big-endian or | |
268 little-endian encoding is to be used for the member of the pair | |
269 that is not UTF-8. | |
270 | |
271 INBUFP, INBYTESLEFTP, OUTBUFP, OUTBYTESLEFTP work exactly as they | |
272 do for iconv. | |
273 | |
274 The return value is either 0 for success, or an errno value for | |
275 failure, which may be E2BIG (need more space), EILSEQ (ill-formed | |
276 input sequence), ir EINVAL (incomplete input sequence). */ | |
277 | |
278 static inline int | |
279 one_utf8_to_utf32 (iconv_t bigend, const uchar **inbufp, size_t *inbytesleftp, | |
280 uchar **outbufp, size_t *outbytesleftp) | |
281 { | |
282 uchar *outbuf; | |
283 cppchar_t s = 0; | |
284 int rval; | |
285 | |
286 /* Check for space first, since we know exactly how much we need. */ | |
287 if (*outbytesleftp < 4) | |
288 return E2BIG; | |
289 | |
290 rval = one_utf8_to_cppchar (inbufp, inbytesleftp, &s); | |
291 if (rval) | |
292 return rval; | |
293 | |
294 outbuf = *outbufp; | |
295 outbuf[bigend ? 3 : 0] = (s & 0x000000FF); | |
296 outbuf[bigend ? 2 : 1] = (s & 0x0000FF00) >> 8; | |
297 outbuf[bigend ? 1 : 2] = (s & 0x00FF0000) >> 16; | |
298 outbuf[bigend ? 0 : 3] = (s & 0xFF000000) >> 24; | |
299 | |
300 *outbufp += 4; | |
301 *outbytesleftp -= 4; | |
302 return 0; | |
303 } | |
304 | |
305 static inline int | |
306 one_utf32_to_utf8 (iconv_t bigend, const uchar **inbufp, size_t *inbytesleftp, | |
307 uchar **outbufp, size_t *outbytesleftp) | |
308 { | |
309 cppchar_t s; | |
310 int rval; | |
311 const uchar *inbuf; | |
312 | |
313 if (*inbytesleftp < 4) | |
314 return EINVAL; | |
315 | |
316 inbuf = *inbufp; | |
317 | |
318 s = inbuf[bigend ? 0 : 3] << 24; | |
319 s += inbuf[bigend ? 1 : 2] << 16; | |
320 s += inbuf[bigend ? 2 : 1] << 8; | |
321 s += inbuf[bigend ? 3 : 0]; | |
322 | |
323 if (s >= 0x7FFFFFFF || (s >= 0xD800 && s <= 0xDFFF)) | |
324 return EILSEQ; | |
325 | |
326 rval = one_cppchar_to_utf8 (s, outbufp, outbytesleftp); | |
327 if (rval) | |
328 return rval; | |
329 | |
330 *inbufp += 4; | |
331 *inbytesleftp -= 4; | |
332 return 0; | |
333 } | |
334 | |
335 static inline int | |
336 one_utf8_to_utf16 (iconv_t bigend, const uchar **inbufp, size_t *inbytesleftp, | |
337 uchar **outbufp, size_t *outbytesleftp) | |
338 { | |
339 int rval; | |
340 cppchar_t s = 0; | |
341 const uchar *save_inbuf = *inbufp; | |
342 size_t save_inbytesleft = *inbytesleftp; | |
343 uchar *outbuf = *outbufp; | |
344 | |
345 rval = one_utf8_to_cppchar (inbufp, inbytesleftp, &s); | |
346 if (rval) | |
347 return rval; | |
348 | |
349 if (s > 0x0010FFFF) | |
350 { | |
351 *inbufp = save_inbuf; | |
352 *inbytesleftp = save_inbytesleft; | |
353 return EILSEQ; | |
354 } | |
355 | |
111 | 356 if (s <= 0xFFFF) |
0 | 357 { |
358 if (*outbytesleftp < 2) | |
359 { | |
360 *inbufp = save_inbuf; | |
361 *inbytesleftp = save_inbytesleft; | |
362 return E2BIG; | |
363 } | |
364 outbuf[bigend ? 1 : 0] = (s & 0x00FF); | |
365 outbuf[bigend ? 0 : 1] = (s & 0xFF00) >> 8; | |
366 | |
367 *outbufp += 2; | |
368 *outbytesleftp -= 2; | |
369 return 0; | |
370 } | |
371 else | |
372 { | |
373 cppchar_t hi, lo; | |
374 | |
375 if (*outbytesleftp < 4) | |
376 { | |
377 *inbufp = save_inbuf; | |
378 *inbytesleftp = save_inbytesleft; | |
379 return E2BIG; | |
380 } | |
381 | |
382 hi = (s - 0x10000) / 0x400 + 0xD800; | |
383 lo = (s - 0x10000) % 0x400 + 0xDC00; | |
384 | |
385 /* Even if we are little-endian, put the high surrogate first. | |
386 ??? Matches practice? */ | |
387 outbuf[bigend ? 1 : 0] = (hi & 0x00FF); | |
388 outbuf[bigend ? 0 : 1] = (hi & 0xFF00) >> 8; | |
389 outbuf[bigend ? 3 : 2] = (lo & 0x00FF); | |
390 outbuf[bigend ? 2 : 3] = (lo & 0xFF00) >> 8; | |
391 | |
392 *outbufp += 4; | |
393 *outbytesleftp -= 4; | |
394 return 0; | |
395 } | |
396 } | |
397 | |
398 static inline int | |
399 one_utf16_to_utf8 (iconv_t bigend, const uchar **inbufp, size_t *inbytesleftp, | |
400 uchar **outbufp, size_t *outbytesleftp) | |
401 { | |
402 cppchar_t s; | |
403 const uchar *inbuf = *inbufp; | |
404 int rval; | |
405 | |
406 if (*inbytesleftp < 2) | |
407 return EINVAL; | |
408 s = inbuf[bigend ? 0 : 1] << 8; | |
409 s += inbuf[bigend ? 1 : 0]; | |
410 | |
411 /* Low surrogate without immediately preceding high surrogate is invalid. */ | |
412 if (s >= 0xDC00 && s <= 0xDFFF) | |
413 return EILSEQ; | |
414 /* High surrogate must have a following low surrogate. */ | |
415 else if (s >= 0xD800 && s <= 0xDBFF) | |
416 { | |
417 cppchar_t hi = s, lo; | |
418 if (*inbytesleftp < 4) | |
419 return EINVAL; | |
420 | |
421 lo = inbuf[bigend ? 2 : 3] << 8; | |
422 lo += inbuf[bigend ? 3 : 2]; | |
423 | |
424 if (lo < 0xDC00 || lo > 0xDFFF) | |
425 return EILSEQ; | |
426 | |
427 s = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000; | |
428 } | |
429 | |
430 rval = one_cppchar_to_utf8 (s, outbufp, outbytesleftp); | |
431 if (rval) | |
432 return rval; | |
433 | |
434 /* Success - update the input pointers (one_cppchar_to_utf8 has done | |
435 the output pointers for us). */ | |
436 if (s <= 0xFFFF) | |
437 { | |
438 *inbufp += 2; | |
439 *inbytesleftp -= 2; | |
440 } | |
441 else | |
442 { | |
443 *inbufp += 4; | |
444 *inbytesleftp -= 4; | |
445 } | |
446 return 0; | |
447 } | |
448 | |
449 /* Helper routine for the next few functions. The 'const' on | |
450 one_conversion means that we promise not to modify what function is | |
451 pointed to, which lets the inliner see through it. */ | |
452 | |
453 static inline bool | |
454 conversion_loop (int (*const one_conversion)(iconv_t, const uchar **, size_t *, | |
455 uchar **, size_t *), | |
456 iconv_t cd, const uchar *from, size_t flen, struct _cpp_strbuf *to) | |
457 { | |
458 const uchar *inbuf; | |
459 uchar *outbuf; | |
460 size_t inbytesleft, outbytesleft; | |
461 int rval; | |
462 | |
463 inbuf = from; | |
464 inbytesleft = flen; | |
465 outbuf = to->text + to->len; | |
466 outbytesleft = to->asize - to->len; | |
467 | |
468 for (;;) | |
469 { | |
470 do | |
471 rval = one_conversion (cd, &inbuf, &inbytesleft, | |
472 &outbuf, &outbytesleft); | |
473 while (inbytesleft && !rval); | |
474 | |
475 if (__builtin_expect (inbytesleft == 0, 1)) | |
476 { | |
477 to->len = to->asize - outbytesleft; | |
478 return true; | |
479 } | |
480 if (rval != E2BIG) | |
481 { | |
482 errno = rval; | |
483 return false; | |
484 } | |
485 | |
486 outbytesleft += OUTBUF_BLOCK_SIZE; | |
487 to->asize += OUTBUF_BLOCK_SIZE; | |
488 to->text = XRESIZEVEC (uchar, to->text, to->asize); | |
489 outbuf = to->text + to->asize - outbytesleft; | |
490 } | |
491 } | |
492 | |
493 | |
494 /* These functions convert entire strings between character sets. | |
495 They all have the signature | |
496 | |
497 bool (*)(iconv_t cd, const uchar *from, size_t flen, struct _cpp_strbuf *to); | |
498 | |
499 The input string FROM is converted as specified by the function | |
500 name plus the iconv descriptor CD (which may be fake), and the | |
501 result appended to TO. On any error, false is returned, otherwise true. */ | |
502 | |
503 /* These four use the custom conversion code above. */ | |
504 static bool | |
505 convert_utf8_utf16 (iconv_t cd, const uchar *from, size_t flen, | |
506 struct _cpp_strbuf *to) | |
507 { | |
508 return conversion_loop (one_utf8_to_utf16, cd, from, flen, to); | |
509 } | |
510 | |
511 static bool | |
512 convert_utf8_utf32 (iconv_t cd, const uchar *from, size_t flen, | |
513 struct _cpp_strbuf *to) | |
514 { | |
515 return conversion_loop (one_utf8_to_utf32, cd, from, flen, to); | |
516 } | |
517 | |
518 static bool | |
519 convert_utf16_utf8 (iconv_t cd, const uchar *from, size_t flen, | |
520 struct _cpp_strbuf *to) | |
521 { | |
522 return conversion_loop (one_utf16_to_utf8, cd, from, flen, to); | |
523 } | |
524 | |
525 static bool | |
526 convert_utf32_utf8 (iconv_t cd, const uchar *from, size_t flen, | |
527 struct _cpp_strbuf *to) | |
528 { | |
529 return conversion_loop (one_utf32_to_utf8, cd, from, flen, to); | |
530 } | |
531 | |
532 /* Identity conversion, used when we have no alternative. */ | |
533 static bool | |
534 convert_no_conversion (iconv_t cd ATTRIBUTE_UNUSED, | |
535 const uchar *from, size_t flen, struct _cpp_strbuf *to) | |
536 { | |
537 if (to->len + flen > to->asize) | |
538 { | |
539 to->asize = to->len + flen; | |
111 | 540 to->asize += to->asize / 4; |
0 | 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 | |
111 | 612 struct cpp_conversion |
0 | 613 { |
614 const char *pair; | |
615 convert_f func; | |
616 iconv_t fake_cd; | |
617 }; | |
111 | 618 static const struct cpp_conversion conversion_tab[] = { |
0 | 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); | |
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
724 pfile->utf8_cset_desc = init_iconv_desc (pfile, "UTF-8", SOURCE_CHARSET); |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
725 pfile->utf8_cset_desc.width = CPP_OPTION (pfile, char_precision); |
0 | 726 pfile->char16_cset_desc = init_iconv_desc (pfile, |
727 be ? "UTF-16BE" : "UTF-16LE", | |
728 SOURCE_CHARSET); | |
729 pfile->char16_cset_desc.width = 16; | |
730 pfile->char32_cset_desc = init_iconv_desc (pfile, | |
731 be ? "UTF-32BE" : "UTF-32LE", | |
732 SOURCE_CHARSET); | |
733 pfile->char32_cset_desc.width = 32; | |
734 pfile->wide_cset_desc = init_iconv_desc (pfile, wcset, SOURCE_CHARSET); | |
735 pfile->wide_cset_desc.width = CPP_OPTION (pfile, wchar_precision); | |
736 } | |
737 | |
738 /* Destroy iconv(3) descriptors set up by cpp_init_iconv, if necessary. */ | |
739 void | |
740 _cpp_destroy_iconv (cpp_reader *pfile) | |
741 { | |
742 if (HAVE_ICONV) | |
743 { | |
744 if (pfile->narrow_cset_desc.func == convert_using_iconv) | |
745 iconv_close (pfile->narrow_cset_desc.cd); | |
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
746 if (pfile->utf8_cset_desc.func == convert_using_iconv) |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
747 iconv_close (pfile->utf8_cset_desc.cd); |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
748 if (pfile->char16_cset_desc.func == convert_using_iconv) |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
749 iconv_close (pfile->char16_cset_desc.cd); |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
750 if (pfile->char32_cset_desc.func == convert_using_iconv) |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
751 iconv_close (pfile->char32_cset_desc.cd); |
0 | 752 if (pfile->wide_cset_desc.func == convert_using_iconv) |
753 iconv_close (pfile->wide_cset_desc.cd); | |
754 } | |
755 } | |
756 | |
757 /* Utility routine for use by a full compiler. C is a character taken | |
758 from the *basic* source character set, encoded in the host's | |
759 execution encoding. Convert it to (the target's) execution | |
760 encoding, and return that value. | |
761 | |
762 Issues an internal error if C's representation in the narrow | |
763 execution character set fails to be a single-byte value (C99 | |
764 5.2.1p3: "The representation of each member of the source and | |
765 execution character sets shall fit in a byte.") May also issue an | |
766 internal error if C fails to be a member of the basic source | |
767 character set (testing this exactly is too hard, especially when | |
768 the host character set is EBCDIC). */ | |
769 cppchar_t | |
770 cpp_host_to_exec_charset (cpp_reader *pfile, cppchar_t c) | |
771 { | |
772 uchar sbuf[1]; | |
773 struct _cpp_strbuf tbuf; | |
774 | |
775 /* This test is merely an approximation, but it suffices to catch | |
776 the most important thing, which is that we don't get handed a | |
777 character outside the unibyte range of the host character set. */ | |
778 if (c > LAST_POSSIBLY_BASIC_SOURCE_CHAR) | |
779 { | |
780 cpp_error (pfile, CPP_DL_ICE, | |
781 "character 0x%lx is not in the basic source character set\n", | |
782 (unsigned long)c); | |
783 return 0; | |
784 } | |
785 | |
786 /* Being a character in the unibyte range of the host character set, | |
787 we can safely splat it into a one-byte buffer and trust that that | |
788 is a well-formed string. */ | |
789 sbuf[0] = c; | |
790 | |
791 /* This should never need to reallocate, but just in case... */ | |
792 tbuf.asize = 1; | |
793 tbuf.text = XNEWVEC (uchar, tbuf.asize); | |
794 tbuf.len = 0; | |
795 | |
796 if (!APPLY_CONVERSION (pfile->narrow_cset_desc, sbuf, 1, &tbuf)) | |
797 { | |
798 cpp_errno (pfile, CPP_DL_ICE, "converting to execution character set"); | |
799 return 0; | |
800 } | |
801 if (tbuf.len != 1) | |
802 { | |
803 cpp_error (pfile, CPP_DL_ICE, | |
804 "character 0x%lx is not unibyte in execution character set", | |
805 (unsigned long)c); | |
806 return 0; | |
807 } | |
808 c = tbuf.text[0]; | |
809 free(tbuf.text); | |
810 return c; | |
811 } | |
812 | |
813 | |
814 | |
111 | 815 /* cpp_substring_ranges's constructor. */ |
816 | |
817 cpp_substring_ranges::cpp_substring_ranges () : | |
818 m_ranges (NULL), | |
819 m_num_ranges (0), | |
820 m_alloc_ranges (8) | |
821 { | |
822 m_ranges = XNEWVEC (source_range, m_alloc_ranges); | |
823 } | |
824 | |
825 /* cpp_substring_ranges's destructor. */ | |
826 | |
827 cpp_substring_ranges::~cpp_substring_ranges () | |
828 { | |
829 free (m_ranges); | |
830 } | |
831 | |
832 /* Add RANGE to the vector of source_range information. */ | |
833 | |
834 void | |
835 cpp_substring_ranges::add_range (source_range range) | |
836 { | |
837 if (m_num_ranges >= m_alloc_ranges) | |
838 { | |
839 m_alloc_ranges *= 2; | |
840 m_ranges | |
841 = (source_range *)xrealloc (m_ranges, | |
842 sizeof (source_range) * m_alloc_ranges); | |
843 } | |
844 m_ranges[m_num_ranges++] = range; | |
845 } | |
846 | |
847 /* Read NUM ranges from LOC_READER, adding them to the vector of source_range | |
848 information. */ | |
849 | |
850 void | |
851 cpp_substring_ranges::add_n_ranges (int num, | |
852 cpp_string_location_reader &loc_reader) | |
853 { | |
854 for (int i = 0; i < num; i++) | |
855 add_range (loc_reader.get_next ()); | |
856 } | |
857 | |
858 | |
859 | |
0 | 860 /* Utility routine that computes a mask of the form 0000...111... with |
861 WIDTH 1-bits. */ | |
862 static inline size_t | |
863 width_to_mask (size_t width) | |
864 { | |
865 width = MIN (width, BITS_PER_CPPCHAR_T); | |
866 if (width >= CHAR_BIT * sizeof (size_t)) | |
867 return ~(size_t) 0; | |
868 else | |
869 return ((size_t) 1 << width) - 1; | |
870 } | |
871 | |
872 /* A large table of unicode character information. */ | |
873 enum { | |
874 /* Valid in a C99 identifier? */ | |
875 C99 = 1, | |
876 /* Valid in a C99 identifier, but not as the first character? */ | |
111 | 877 N99 = 2, |
0 | 878 /* Valid in a C++ identifier? */ |
879 CXX = 4, | |
111 | 880 /* Valid in a C11/C++11 identifier? */ |
881 C11 = 8, | |
882 /* Valid in a C11/C++11 identifier, but not as the first character? */ | |
883 N11 = 16, | |
0 | 884 /* NFC representation is not valid in an identifier? */ |
111 | 885 CID = 32, |
0 | 886 /* Might be valid NFC form? */ |
111 | 887 NFC = 64, |
0 | 888 /* Might be valid NFKC form? */ |
111 | 889 NKC = 128, |
0 | 890 /* Certain preceding characters might make it not valid NFC/NKFC form? */ |
111 | 891 CTX = 256 |
0 | 892 }; |
893 | |
111 | 894 struct ucnrange { |
0 | 895 /* Bitmap of flags above. */ |
111 | 896 unsigned short flags; |
0 | 897 /* Combining class of the character. */ |
898 unsigned char combine; | |
899 /* Last character in the range described by this entry. */ | |
111 | 900 unsigned int end; |
901 }; | |
0 | 902 #include "ucnid.h" |
903 | |
145 | 904 /* ISO 10646 defines the UCS codespace as the range 0-0x10FFFF inclusive. */ |
905 #define UCS_LIMIT 0x10FFFF | |
906 | |
0 | 907 /* Returns 1 if C is valid in an identifier, 2 if C is valid except at |
908 the start of an identifier, and 0 if C is not valid in an | |
909 identifier. We assume C has already gone through the checks of | |
910 _cpp_valid_ucn. Also update NST for C if returning nonzero. The | |
911 algorithm is a simple binary search on the table defined in | |
912 ucnid.h. */ | |
913 | |
914 static int | |
915 ucn_valid_in_identifier (cpp_reader *pfile, cppchar_t c, | |
916 struct normalize_state *nst) | |
917 { | |
918 int mn, mx, md; | |
111 | 919 unsigned short valid_flags, invalid_start_flags; |
0 | 920 |
145 | 921 if (c > UCS_LIMIT) |
0 | 922 return 0; |
923 | |
924 mn = 0; | |
925 mx = ARRAY_SIZE (ucnranges) - 1; | |
926 while (mx != mn) | |
927 { | |
928 md = (mn + mx) / 2; | |
929 if (c <= ucnranges[md].end) | |
930 mx = md; | |
931 else | |
932 mn = md + 1; | |
933 } | |
934 | |
935 /* When -pedantic, we require the character to have been listed by | |
936 the standard for the current language. Otherwise, we accept the | |
111 | 937 union of the acceptable sets for all supported language versions. */ |
938 valid_flags = C99 | CXX | C11; | |
939 if (CPP_PEDANTIC (pfile)) | |
940 { | |
941 if (CPP_OPTION (pfile, c11_identifiers)) | |
942 valid_flags = C11; | |
943 else if (CPP_OPTION (pfile, c99)) | |
944 valid_flags = C99; | |
945 else if (CPP_OPTION (pfile, cplusplus)) | |
946 valid_flags = CXX; | |
947 } | |
948 if (! (ucnranges[mn].flags & valid_flags)) | |
0 | 949 return 0; |
111 | 950 if (CPP_OPTION (pfile, c11_identifiers)) |
951 invalid_start_flags = N11; | |
952 else if (CPP_OPTION (pfile, c99)) | |
953 invalid_start_flags = N99; | |
954 else | |
955 invalid_start_flags = 0; | |
0 | 956 |
957 /* Update NST. */ | |
958 if (ucnranges[mn].combine != 0 && ucnranges[mn].combine < nst->prev_class) | |
959 nst->level = normalized_none; | |
960 else if (ucnranges[mn].flags & CTX) | |
961 { | |
962 bool safe; | |
963 cppchar_t p = nst->previous; | |
964 | |
965 /* For Hangul, characters in the range AC00-D7A3 are NFC/NFKC, | |
966 and are combined algorithmically from a sequence of the form | |
967 1100-1112 1161-1175 11A8-11C2 | |
968 (if the third is not present, it is treated as 11A7, which is not | |
969 really a valid character). | |
970 Unfortunately, C99 allows (only) the NFC form, but C++ allows | |
971 only the combining characters. */ | |
111 | 972 if (c >= 0x1161 && c <= 0x1175) |
0 | 973 safe = p < 0x1100 || p > 0x1112; |
974 else if (c >= 0x11A8 && c <= 0x11C2) | |
975 safe = (p < 0xAC00 || p > 0xD7A3 || (p - 0xAC00) % 28 != 0); | |
976 else | |
111 | 977 safe = check_nfc (pfile, c, p); |
978 if (!safe) | |
0 | 979 { |
111 | 980 if ((c >= 0x1161 && c <= 0x1175) || (c >= 0x11A8 && c <= 0x11C2)) |
981 nst->level = MAX (nst->level, normalized_identifier_C); | |
982 else | |
983 nst->level = normalized_none; | |
0 | 984 } |
985 } | |
986 else if (ucnranges[mn].flags & NKC) | |
987 ; | |
988 else if (ucnranges[mn].flags & NFC) | |
989 nst->level = MAX (nst->level, normalized_C); | |
990 else if (ucnranges[mn].flags & CID) | |
991 nst->level = MAX (nst->level, normalized_identifier_C); | |
992 else | |
993 nst->level = normalized_none; | |
111 | 994 if (ucnranges[mn].combine == 0) |
995 nst->previous = c; | |
0 | 996 nst->prev_class = ucnranges[mn].combine; |
997 | |
111 | 998 /* In C99, UCN digits may not begin identifiers. In C11 and C++11, |
999 UCN combining characters may not begin identifiers. */ | |
1000 if (ucnranges[mn].flags & invalid_start_flags) | |
0 | 1001 return 2; |
1002 | |
1003 return 1; | |
1004 } | |
1005 | |
1006 /* [lex.charset]: The character designated by the universal character | |
1007 name \UNNNNNNNN is that character whose character short name in | |
1008 ISO/IEC 10646 is NNNNNNNN; the character designated by the | |
1009 universal character name \uNNNN is that character whose character | |
1010 short name in ISO/IEC 10646 is 0000NNNN. If the hexadecimal value | |
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1011 for a universal character name corresponds to a surrogate code point |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1012 (in the range 0xD800-0xDFFF, inclusive), the program is ill-formed. |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1013 Additionally, if the hexadecimal value for a universal-character-name |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1014 outside a character or string literal corresponds to a control character |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1015 (in either of the ranges 0x00-0x1F or 0x7F-0x9F, both inclusive) or to a |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1016 character in the basic source character set, the program is ill-formed. |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1017 |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1018 C99 6.4.3: A universal character name shall not specify a character |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1019 whose short identifier is less than 00A0 other than 0024 ($), 0040 (@), |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1020 or 0060 (`), nor one in the range D800 through DFFF inclusive. |
0 | 1021 |
145 | 1022 If the hexadecimal value is larger than the upper bound of the UCS |
1023 codespace specified in ISO/IEC 10646, a pedantic warning is issued | |
1024 in all versions of C and in the C++2a or later versions of C++. | |
1025 | |
0 | 1026 *PSTR must be preceded by "\u" or "\U"; it is assumed that the |
111 | 1027 buffer end is delimited by a non-hex digit. Returns false if the |
1028 UCN has not been consumed, true otherwise. | |
0 | 1029 |
111 | 1030 The value of the UCN, whether valid or invalid, is returned in *CP. |
1031 Diagnostics are emitted for invalid values. PSTR is updated to point | |
1032 one beyond the UCN, or to the syntactically invalid character. | |
0 | 1033 |
1034 IDENTIFIER_POS is 0 when not in an identifier, 1 for the start of | |
111 | 1035 an identifier, or 2 otherwise. |
0 | 1036 |
111 | 1037 If LOC_READER is non-NULL, then position information is |
1038 read from *LOC_READER and CHAR_RANGE->m_finish is updated accordingly. */ | |
1039 | |
1040 bool | |
0 | 1041 _cpp_valid_ucn (cpp_reader *pfile, const uchar **pstr, |
1042 const uchar *limit, int identifier_pos, | |
111 | 1043 struct normalize_state *nst, cppchar_t *cp, |
1044 source_range *char_range, | |
1045 cpp_string_location_reader *loc_reader) | |
0 | 1046 { |
1047 cppchar_t result, c; | |
1048 unsigned int length; | |
1049 const uchar *str = *pstr; | |
1050 const uchar *base = str - 2; | |
1051 | |
1052 if (!CPP_OPTION (pfile, cplusplus) && !CPP_OPTION (pfile, c99)) | |
1053 cpp_error (pfile, CPP_DL_WARNING, | |
1054 "universal character names are only valid in C++ and C99"); | |
111 | 1055 else if (CPP_OPTION (pfile, cpp_warn_c90_c99_compat) > 0 |
1056 && !CPP_OPTION (pfile, cplusplus)) | |
1057 cpp_error (pfile, CPP_DL_WARNING, | |
1058 "C99's universal character names are incompatible with C90"); | |
0 | 1059 else if (CPP_WTRADITIONAL (pfile) && identifier_pos == 0) |
63
b7f97abdc517
update gcc from gcc-4.5.0 to gcc-4.6
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
55
diff
changeset
|
1060 cpp_warning (pfile, CPP_W_TRADITIONAL, |
b7f97abdc517
update gcc from gcc-4.5.0 to gcc-4.6
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
55
diff
changeset
|
1061 "the meaning of '\\%c' is different in traditional C", |
b7f97abdc517
update gcc from gcc-4.5.0 to gcc-4.6
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
55
diff
changeset
|
1062 (int) str[-1]); |
0 | 1063 |
1064 if (str[-1] == 'u') | |
1065 length = 4; | |
1066 else if (str[-1] == 'U') | |
1067 length = 8; | |
1068 else | |
1069 { | |
1070 cpp_error (pfile, CPP_DL_ICE, "In _cpp_valid_ucn but not a UCN"); | |
1071 length = 4; | |
1072 } | |
1073 | |
1074 result = 0; | |
1075 do | |
1076 { | |
1077 c = *str; | |
1078 if (!ISXDIGIT (c)) | |
1079 break; | |
1080 str++; | |
111 | 1081 if (loc_reader) |
1082 { | |
1083 gcc_assert (char_range); | |
1084 char_range->m_finish = loc_reader->get_next ().m_finish; | |
1085 } | |
0 | 1086 result = (result << 4) + hex_value (c); |
1087 } | |
1088 while (--length && str < limit); | |
1089 | |
1090 /* Partial UCNs are not valid in strings, but decompose into | |
1091 multiple tokens in identifiers, so we can't give a helpful | |
1092 error message in that case. */ | |
1093 if (length && identifier_pos) | |
111 | 1094 { |
1095 *cp = 0; | |
1096 return false; | |
1097 } | |
1098 | |
0 | 1099 *pstr = str; |
1100 if (length) | |
1101 { | |
1102 cpp_error (pfile, CPP_DL_ERROR, | |
1103 "incomplete universal character name %.*s", | |
1104 (int) (str - base), base); | |
1105 result = 1; | |
1106 } | |
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1107 /* The C99 standard permits $, @ and ` to be specified as UCNs. We use |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1108 hex escapes so that this also works with EBCDIC hosts. |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1109 C++0x permits everything below 0xa0 within literals; |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1110 ucn_valid_in_identifier will complain about identifiers. */ |
0 | 1111 else if ((result < 0xa0 |
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1112 && !CPP_OPTION (pfile, cplusplus) |
0 | 1113 && (result != 0x24 && result != 0x40 && result != 0x60)) |
1114 || (result & 0x80000000) | |
1115 || (result >= 0xD800 && result <= 0xDFFF)) | |
1116 { | |
1117 cpp_error (pfile, CPP_DL_ERROR, | |
1118 "%.*s is not a valid universal character", | |
1119 (int) (str - base), base); | |
1120 result = 1; | |
1121 } | |
1122 else if (identifier_pos && result == 0x24 | |
1123 && CPP_OPTION (pfile, dollars_in_ident)) | |
1124 { | |
1125 if (CPP_OPTION (pfile, warn_dollars) && !pfile->state.skipping) | |
1126 { | |
1127 CPP_OPTION (pfile, warn_dollars) = 0; | |
1128 cpp_error (pfile, CPP_DL_PEDWARN, "'$' in identifier or number"); | |
1129 } | |
111 | 1130 NORMALIZE_STATE_UPDATE_IDNUM (nst, result); |
0 | 1131 } |
1132 else if (identifier_pos) | |
1133 { | |
1134 int validity = ucn_valid_in_identifier (pfile, result, nst); | |
1135 | |
1136 if (validity == 0) | |
1137 cpp_error (pfile, CPP_DL_ERROR, | |
1138 "universal character %.*s is not valid in an identifier", | |
1139 (int) (str - base), base); | |
1140 else if (validity == 2 && identifier_pos == 1) | |
1141 cpp_error (pfile, CPP_DL_ERROR, | |
1142 "universal character %.*s is not valid at the start of an identifier", | |
1143 (int) (str - base), base); | |
1144 } | |
145 | 1145 else if (result > UCS_LIMIT |
1146 && (!CPP_OPTION (pfile, cplusplus) | |
1147 || CPP_OPTION (pfile, lang) > CLK_CXX17)) | |
1148 cpp_error (pfile, CPP_DL_PEDWARN, | |
1149 "%.*s is outside the UCS codespace", | |
1150 (int) (str - base), base); | |
0 | 1151 |
111 | 1152 *cp = result; |
1153 return true; | |
0 | 1154 } |
1155 | |
1156 /* Convert an UCN, pointed to by FROM, to UTF-8 encoding, then translate | |
111 | 1157 it to the execution character set and write the result into TBUF, |
1158 if TBUF is non-NULL. | |
1159 An advanced pointer is returned. Issues all relevant diagnostics. | |
1160 If LOC_READER is non-NULL, then RANGES must be non-NULL and CHAR_RANGE | |
1161 contains the location of the character so far: location information | |
1162 is read from *LOC_READER, and *RANGES is updated accordingly. */ | |
0 | 1163 static const uchar * |
1164 convert_ucn (cpp_reader *pfile, const uchar *from, const uchar *limit, | |
111 | 1165 struct _cpp_strbuf *tbuf, struct cset_converter cvt, |
1166 source_range char_range, | |
1167 cpp_string_location_reader *loc_reader, | |
1168 cpp_substring_ranges *ranges) | |
0 | 1169 { |
1170 cppchar_t ucn; | |
1171 uchar buf[6]; | |
1172 uchar *bufp = buf; | |
1173 size_t bytesleft = 6; | |
1174 int rval; | |
1175 struct normalize_state nst = INITIAL_NORMALIZE_STATE; | |
1176 | |
111 | 1177 /* loc_reader and ranges must either be both NULL, or both be non-NULL. */ |
1178 gcc_assert ((loc_reader != NULL) == (ranges != NULL)); | |
1179 | |
0 | 1180 from++; /* Skip u/U. */ |
111 | 1181 |
1182 if (loc_reader) | |
1183 /* The u/U is part of the spelling of this character. */ | |
1184 char_range.m_finish = loc_reader->get_next ().m_finish; | |
1185 | |
1186 _cpp_valid_ucn (pfile, &from, limit, 0, &nst, | |
1187 &ucn, &char_range, loc_reader); | |
0 | 1188 |
1189 rval = one_cppchar_to_utf8 (ucn, &bufp, &bytesleft); | |
1190 if (rval) | |
1191 { | |
1192 errno = rval; | |
1193 cpp_errno (pfile, CPP_DL_ERROR, | |
1194 "converting UCN to source character set"); | |
1195 } | |
111 | 1196 else |
1197 { | |
1198 if (tbuf) | |
1199 if (!APPLY_CONVERSION (cvt, buf, 6 - bytesleft, tbuf)) | |
1200 cpp_errno (pfile, CPP_DL_ERROR, | |
1201 "converting UCN to execution character set"); | |
1202 | |
1203 if (loc_reader) | |
1204 { | |
1205 int num_encoded_bytes = 6 - bytesleft; | |
1206 for (int i = 0; i < num_encoded_bytes; i++) | |
1207 ranges->add_range (char_range); | |
1208 } | |
1209 } | |
0 | 1210 |
1211 return from; | |
1212 } | |
1213 | |
145 | 1214 /* Performs a similar task as _cpp_valid_ucn, but parses UTF-8-encoded |
1215 extended characters rather than UCNs. If the return value is TRUE, then a | |
1216 character was successfully decoded and stored in *CP; *PSTR has been | |
1217 updated to point one past the valid UTF-8 sequence. Diagnostics may have | |
1218 been emitted if the character parsed is not allowed in the current context. | |
1219 If the return value is FALSE, then *PSTR has not been modified and *CP may | |
1220 equal 0, to indicate that *PSTR does not form a valid UTF-8 sequence, or it | |
1221 may, when processing an identifier in C mode, equal a codepoint that was | |
1222 validly encoded but is not allowed to appear in an identifier. In either | |
1223 case, no diagnostic is emitted, and the return value of FALSE should cause | |
1224 a new token to be formed. | |
1225 | |
1226 Unlike _cpp_valid_ucn, this will never be called when lexing a string; only | |
1227 a potential identifier, or a CPP_OTHER token. NST is unused in the latter | |
1228 case. | |
1229 | |
1230 As in _cpp_valid_ucn, IDENTIFIER_POS is 0 when not in an identifier, 1 for | |
1231 the start of an identifier, or 2 otherwise. */ | |
1232 | |
1233 extern bool | |
1234 _cpp_valid_utf8 (cpp_reader *pfile, | |
1235 const uchar **pstr, | |
1236 const uchar *limit, | |
1237 int identifier_pos, | |
1238 struct normalize_state *nst, | |
1239 cppchar_t *cp) | |
1240 { | |
1241 const uchar *base = *pstr; | |
1242 size_t inbytesleft = limit - base; | |
1243 if (one_utf8_to_cppchar (pstr, &inbytesleft, cp)) | |
1244 { | |
1245 /* No diagnostic here as this byte will rather become a | |
1246 new token. */ | |
1247 *cp = 0; | |
1248 return false; | |
1249 } | |
1250 | |
1251 if (identifier_pos) | |
1252 { | |
1253 switch (ucn_valid_in_identifier (pfile, *cp, nst)) | |
1254 { | |
1255 | |
1256 case 0: | |
1257 /* In C++, this is an error for invalid character in an identifier | |
1258 because logically, the UTF-8 was converted to a UCN during | |
1259 translation phase 1 (even though we don't physically do it that | |
1260 way). In C, this byte rather becomes grammatically a separate | |
1261 token. */ | |
1262 | |
1263 if (CPP_OPTION (pfile, cplusplus)) | |
1264 cpp_error (pfile, CPP_DL_ERROR, | |
1265 "extended character %.*s is not valid in an identifier", | |
1266 (int) (*pstr - base), base); | |
1267 else | |
1268 { | |
1269 *pstr = base; | |
1270 return false; | |
1271 } | |
1272 | |
1273 break; | |
1274 | |
1275 case 2: | |
1276 if (identifier_pos == 1) | |
1277 { | |
1278 /* This is treated the same way in C++ or C99 -- lexed as an | |
1279 identifier which is then invalid because an identifier is | |
1280 not allowed to start with this character. */ | |
1281 cpp_error (pfile, CPP_DL_ERROR, | |
1282 "extended character %.*s is not valid at the start of an identifier", | |
1283 (int) (*pstr - base), base); | |
1284 } | |
1285 break; | |
1286 } | |
1287 } | |
1288 | |
1289 return true; | |
1290 } | |
1291 | |
0 | 1292 /* Subroutine of convert_hex and convert_oct. N is the representation |
1293 in the execution character set of a numeric escape; write it into the | |
1294 string buffer TBUF and update the end-of-string pointer therein. WIDE | |
1295 is true if it's a wide string that's being assembled in TBUF. This | |
1296 function issues no diagnostics and never fails. */ | |
1297 static void | |
1298 emit_numeric_escape (cpp_reader *pfile, cppchar_t n, | |
1299 struct _cpp_strbuf *tbuf, struct cset_converter cvt) | |
1300 { | |
1301 size_t width = cvt.width; | |
1302 | |
1303 if (width != CPP_OPTION (pfile, char_precision)) | |
1304 { | |
1305 /* We have to render this into the target byte order, which may not | |
1306 be our byte order. */ | |
1307 bool bigend = CPP_OPTION (pfile, bytes_big_endian); | |
1308 size_t cwidth = CPP_OPTION (pfile, char_precision); | |
1309 size_t cmask = width_to_mask (cwidth); | |
1310 size_t nbwc = width / cwidth; | |
1311 size_t i; | |
1312 size_t off = tbuf->len; | |
1313 cppchar_t c; | |
1314 | |
1315 if (tbuf->len + nbwc > tbuf->asize) | |
1316 { | |
1317 tbuf->asize += OUTBUF_BLOCK_SIZE; | |
1318 tbuf->text = XRESIZEVEC (uchar, tbuf->text, tbuf->asize); | |
1319 } | |
1320 | |
1321 for (i = 0; i < nbwc; i++) | |
1322 { | |
1323 c = n & cmask; | |
1324 n >>= cwidth; | |
1325 tbuf->text[off + (bigend ? nbwc - i - 1 : i)] = c; | |
1326 } | |
1327 tbuf->len += nbwc; | |
1328 } | |
1329 else | |
1330 { | |
1331 /* Note: this code does not handle the case where the target | |
1332 and host have a different number of bits in a byte. */ | |
1333 if (tbuf->len + 1 > tbuf->asize) | |
1334 { | |
1335 tbuf->asize += OUTBUF_BLOCK_SIZE; | |
1336 tbuf->text = XRESIZEVEC (uchar, tbuf->text, tbuf->asize); | |
1337 } | |
1338 tbuf->text[tbuf->len++] = n; | |
1339 } | |
1340 } | |
1341 | |
1342 /* Convert a hexadecimal escape, pointed to by FROM, to the execution | |
111 | 1343 character set and write it into the string buffer TBUF (if non-NULL). |
1344 Returns an advanced pointer, and issues diagnostics as necessary. | |
0 | 1345 No character set translation occurs; this routine always produces the |
1346 execution-set character with numeric value equal to the given hex | |
111 | 1347 number. You can, e.g. generate surrogate pairs this way. |
1348 If LOC_READER is non-NULL, then RANGES must be non-NULL and CHAR_RANGE | |
1349 contains the location of the character so far: location information | |
1350 is read from *LOC_READER, and *RANGES is updated accordingly. */ | |
0 | 1351 static const uchar * |
1352 convert_hex (cpp_reader *pfile, const uchar *from, const uchar *limit, | |
111 | 1353 struct _cpp_strbuf *tbuf, struct cset_converter cvt, |
1354 source_range char_range, | |
1355 cpp_string_location_reader *loc_reader, | |
1356 cpp_substring_ranges *ranges) | |
0 | 1357 { |
1358 cppchar_t c, n = 0, overflow = 0; | |
1359 int digits_found = 0; | |
1360 size_t width = cvt.width; | |
1361 size_t mask = width_to_mask (width); | |
1362 | |
111 | 1363 /* loc_reader and ranges must either be both NULL, or both be non-NULL. */ |
1364 gcc_assert ((loc_reader != NULL) == (ranges != NULL)); | |
1365 | |
0 | 1366 if (CPP_WTRADITIONAL (pfile)) |
63
b7f97abdc517
update gcc from gcc-4.5.0 to gcc-4.6
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
55
diff
changeset
|
1367 cpp_warning (pfile, CPP_W_TRADITIONAL, |
b7f97abdc517
update gcc from gcc-4.5.0 to gcc-4.6
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
55
diff
changeset
|
1368 "the meaning of '\\x' is different in traditional C"); |
0 | 1369 |
111 | 1370 /* Skip 'x'. */ |
1371 from++; | |
1372 | |
1373 /* The 'x' is part of the spelling of this character. */ | |
1374 if (loc_reader) | |
1375 char_range.m_finish = loc_reader->get_next ().m_finish; | |
1376 | |
0 | 1377 while (from < limit) |
1378 { | |
1379 c = *from; | |
1380 if (! hex_p (c)) | |
1381 break; | |
1382 from++; | |
111 | 1383 if (loc_reader) |
1384 char_range.m_finish = loc_reader->get_next ().m_finish; | |
0 | 1385 overflow |= n ^ (n << 4 >> 4); |
1386 n = (n << 4) + hex_value (c); | |
1387 digits_found = 1; | |
1388 } | |
1389 | |
1390 if (!digits_found) | |
1391 { | |
1392 cpp_error (pfile, CPP_DL_ERROR, | |
1393 "\\x used with no following hex digits"); | |
1394 return from; | |
1395 } | |
1396 | |
1397 if (overflow | (n != (n & mask))) | |
1398 { | |
1399 cpp_error (pfile, CPP_DL_PEDWARN, | |
1400 "hex escape sequence out of range"); | |
1401 n &= mask; | |
1402 } | |
1403 | |
111 | 1404 if (tbuf) |
1405 emit_numeric_escape (pfile, n, tbuf, cvt); | |
1406 if (ranges) | |
1407 ranges->add_range (char_range); | |
0 | 1408 |
1409 return from; | |
1410 } | |
1411 | |
1412 /* Convert an octal escape, pointed to by FROM, to the execution | |
1413 character set and write it into the string buffer TBUF. Returns an | |
1414 advanced pointer, and issues diagnostics as necessary. | |
1415 No character set translation occurs; this routine always produces the | |
1416 execution-set character with numeric value equal to the given octal | |
111 | 1417 number. |
1418 If LOC_READER is non-NULL, then RANGES must be non-NULL and CHAR_RANGE | |
1419 contains the location of the character so far: location information | |
1420 is read from *LOC_READER, and *RANGES is updated accordingly. */ | |
0 | 1421 static const uchar * |
1422 convert_oct (cpp_reader *pfile, const uchar *from, const uchar *limit, | |
111 | 1423 struct _cpp_strbuf *tbuf, struct cset_converter cvt, |
1424 source_range char_range, | |
1425 cpp_string_location_reader *loc_reader, | |
1426 cpp_substring_ranges *ranges) | |
0 | 1427 { |
1428 size_t count = 0; | |
1429 cppchar_t c, n = 0; | |
1430 size_t width = cvt.width; | |
1431 size_t mask = width_to_mask (width); | |
1432 bool overflow = false; | |
1433 | |
111 | 1434 /* loc_reader and ranges must either be both NULL, or both be non-NULL. */ |
1435 gcc_assert ((loc_reader != NULL) == (ranges != NULL)); | |
1436 | |
0 | 1437 while (from < limit && count++ < 3) |
1438 { | |
1439 c = *from; | |
1440 if (c < '0' || c > '7') | |
1441 break; | |
1442 from++; | |
111 | 1443 if (loc_reader) |
1444 char_range.m_finish = loc_reader->get_next ().m_finish; | |
0 | 1445 overflow |= n ^ (n << 3 >> 3); |
1446 n = (n << 3) + c - '0'; | |
1447 } | |
1448 | |
1449 if (n != (n & mask)) | |
1450 { | |
1451 cpp_error (pfile, CPP_DL_PEDWARN, | |
1452 "octal escape sequence out of range"); | |
1453 n &= mask; | |
1454 } | |
1455 | |
111 | 1456 if (tbuf) |
1457 emit_numeric_escape (pfile, n, tbuf, cvt); | |
1458 if (ranges) | |
1459 ranges->add_range (char_range); | |
0 | 1460 |
1461 return from; | |
1462 } | |
1463 | |
1464 /* Convert an escape sequence (pointed to by FROM) to its value on | |
1465 the target, and to the execution character set. Do not scan past | |
111 | 1466 LIMIT. Write the converted value into TBUF, if TBUF is non-NULL. |
1467 Returns an advanced pointer. Handles all relevant diagnostics. | |
1468 If LOC_READER is non-NULL, then RANGES must be non-NULL: location | |
1469 information is read from *LOC_READER, and *RANGES is updated | |
1470 accordingly. */ | |
0 | 1471 static const uchar * |
1472 convert_escape (cpp_reader *pfile, const uchar *from, const uchar *limit, | |
111 | 1473 struct _cpp_strbuf *tbuf, struct cset_converter cvt, |
1474 cpp_string_location_reader *loc_reader, | |
1475 cpp_substring_ranges *ranges) | |
0 | 1476 { |
1477 /* Values of \a \b \e \f \n \r \t \v respectively. */ | |
1478 #if HOST_CHARSET == HOST_CHARSET_ASCII | |
1479 static const uchar charconsts[] = { 7, 8, 27, 12, 10, 13, 9, 11 }; | |
1480 #elif HOST_CHARSET == HOST_CHARSET_EBCDIC | |
1481 static const uchar charconsts[] = { 47, 22, 39, 12, 21, 13, 5, 11 }; | |
1482 #else | |
1483 #error "unknown host character set" | |
1484 #endif | |
1485 | |
1486 uchar c; | |
1487 | |
111 | 1488 /* Record the location of the backslash. */ |
1489 source_range char_range; | |
1490 if (loc_reader) | |
1491 char_range = loc_reader->get_next (); | |
1492 | |
0 | 1493 c = *from; |
1494 switch (c) | |
1495 { | |
1496 /* UCNs, hex escapes, and octal escapes are processed separately. */ | |
1497 case 'u': case 'U': | |
111 | 1498 return convert_ucn (pfile, from, limit, tbuf, cvt, |
1499 char_range, loc_reader, ranges); | |
0 | 1500 |
1501 case 'x': | |
111 | 1502 return convert_hex (pfile, from, limit, tbuf, cvt, |
1503 char_range, loc_reader, ranges); | |
0 | 1504 break; |
1505 | |
1506 case '0': case '1': case '2': case '3': | |
1507 case '4': case '5': case '6': case '7': | |
111 | 1508 return convert_oct (pfile, from, limit, tbuf, cvt, |
1509 char_range, loc_reader, ranges); | |
0 | 1510 |
1511 /* Various letter escapes. Get the appropriate host-charset | |
1512 value into C. */ | |
1513 case '\\': case '\'': case '"': case '?': break; | |
1514 | |
1515 case '(': case '{': case '[': case '%': | |
1516 /* '\(', etc, can be used at the beginning of a line in a long | |
1517 string split onto multiple lines with \-newline, to prevent | |
1518 Emacs or other text editors from getting confused. '\%' can | |
1519 be used to prevent SCCS from mangling printf format strings. */ | |
1520 if (CPP_PEDANTIC (pfile)) | |
1521 goto unknown; | |
1522 break; | |
1523 | |
1524 case 'b': c = charconsts[1]; break; | |
1525 case 'f': c = charconsts[3]; break; | |
1526 case 'n': c = charconsts[4]; break; | |
1527 case 'r': c = charconsts[5]; break; | |
1528 case 't': c = charconsts[6]; break; | |
1529 case 'v': c = charconsts[7]; break; | |
1530 | |
1531 case 'a': | |
1532 if (CPP_WTRADITIONAL (pfile)) | |
63
b7f97abdc517
update gcc from gcc-4.5.0 to gcc-4.6
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
55
diff
changeset
|
1533 cpp_warning (pfile, CPP_W_TRADITIONAL, |
b7f97abdc517
update gcc from gcc-4.5.0 to gcc-4.6
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
55
diff
changeset
|
1534 "the meaning of '\\a' is different in traditional C"); |
0 | 1535 c = charconsts[0]; |
1536 break; | |
1537 | |
1538 case 'e': case 'E': | |
1539 if (CPP_PEDANTIC (pfile)) | |
1540 cpp_error (pfile, CPP_DL_PEDWARN, | |
1541 "non-ISO-standard escape sequence, '\\%c'", (int) c); | |
1542 c = charconsts[2]; | |
1543 break; | |
1544 | |
1545 default: | |
1546 unknown: | |
1547 if (ISGRAPH (c)) | |
1548 cpp_error (pfile, CPP_DL_PEDWARN, | |
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1549 "unknown escape sequence: '\\%c'", (int) c); |
0 | 1550 else |
1551 { | |
1552 /* diagnostic.c does not support "%03o". When it does, this | |
1553 code can use %03o directly in the diagnostic again. */ | |
1554 char buf[32]; | |
1555 sprintf(buf, "%03o", (int) c); | |
1556 cpp_error (pfile, CPP_DL_PEDWARN, | |
1557 "unknown escape sequence: '\\%s'", buf); | |
1558 } | |
1559 } | |
1560 | |
111 | 1561 if (tbuf) |
1562 /* Now convert what we have to the execution character set. */ | |
1563 if (!APPLY_CONVERSION (cvt, &c, 1, tbuf)) | |
1564 cpp_errno (pfile, CPP_DL_ERROR, | |
1565 "converting escape sequence to execution character set"); | |
1566 | |
1567 if (loc_reader) | |
1568 { | |
1569 char_range.m_finish = loc_reader->get_next ().m_finish; | |
1570 ranges->add_range (char_range); | |
1571 } | |
0 | 1572 |
1573 return from + 1; | |
1574 } | |
1575 | |
1576 /* TYPE is a token type. The return value is the conversion needed to | |
1577 convert from source to execution character set for the given type. */ | |
1578 static struct cset_converter | |
1579 converter_for_type (cpp_reader *pfile, enum cpp_ttype type) | |
1580 { | |
1581 switch (type) | |
1582 { | |
1583 default: | |
1584 return pfile->narrow_cset_desc; | |
111 | 1585 case CPP_UTF8CHAR: |
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1586 case CPP_UTF8STRING: |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1587 return pfile->utf8_cset_desc; |
0 | 1588 case CPP_CHAR16: |
1589 case CPP_STRING16: | |
1590 return pfile->char16_cset_desc; | |
1591 case CPP_CHAR32: | |
1592 case CPP_STRING32: | |
1593 return pfile->char32_cset_desc; | |
1594 case CPP_WCHAR: | |
1595 case CPP_WSTRING: | |
1596 return pfile->wide_cset_desc; | |
1597 } | |
1598 } | |
1599 | |
1600 /* FROM is an array of cpp_string structures of length COUNT. These | |
1601 are to be converted from the source to the execution character set, | |
1602 escape sequences translated, and finally all are to be | |
1603 concatenated. WIDE indicates whether or not to produce a wide | |
111 | 1604 string. If TO is non-NULL, the result is written into TO. |
1605 If LOC_READERS and OUT are non-NULL, then location information | |
1606 is read from LOC_READERS (which must be an array of length COUNT), | |
1607 and location information is written to *RANGES. | |
1608 | |
1609 Returns true for success, false for failure. */ | |
1610 | |
1611 static bool | |
1612 cpp_interpret_string_1 (cpp_reader *pfile, const cpp_string *from, size_t count, | |
1613 cpp_string *to, enum cpp_ttype type, | |
1614 cpp_string_location_reader *loc_readers, | |
1615 cpp_substring_ranges *out) | |
0 | 1616 { |
1617 struct _cpp_strbuf tbuf; | |
1618 const uchar *p, *base, *limit; | |
1619 size_t i; | |
1620 struct cset_converter cvt = converter_for_type (pfile, type); | |
1621 | |
111 | 1622 /* loc_readers and out must either be both NULL, or both be non-NULL. */ |
1623 gcc_assert ((loc_readers != NULL) == (out != NULL)); | |
0 | 1624 |
111 | 1625 if (to) |
1626 { | |
1627 tbuf.asize = MAX (OUTBUF_BLOCK_SIZE, from->len); | |
1628 tbuf.text = XNEWVEC (uchar, tbuf.asize); | |
1629 tbuf.len = 0; | |
1630 } | |
1631 | |
1632 cpp_string_location_reader *loc_reader = NULL; | |
0 | 1633 for (i = 0; i < count; i++) |
1634 { | |
111 | 1635 if (loc_readers) |
1636 loc_reader = &loc_readers[i]; | |
1637 | |
0 | 1638 p = from[i].text; |
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1639 if (*p == 'u') |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1640 { |
111 | 1641 p++; |
1642 if (loc_reader) | |
1643 loc_reader->get_next (); | |
1644 if (*p == '8') | |
1645 { | |
1646 p++; | |
1647 if (loc_reader) | |
1648 loc_reader->get_next (); | |
1649 } | |
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1650 } |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1651 else if (*p == 'L' || *p == 'U') p++; |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1652 if (*p == 'R') |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1653 { |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1654 const uchar *prefix; |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1655 |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1656 /* Skip over 'R"'. */ |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1657 p += 2; |
111 | 1658 if (loc_reader) |
1659 { | |
1660 loc_reader->get_next (); | |
1661 loc_reader->get_next (); | |
1662 } | |
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1663 prefix = p; |
63
b7f97abdc517
update gcc from gcc-4.5.0 to gcc-4.6
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
55
diff
changeset
|
1664 while (*p != '(') |
111 | 1665 { |
1666 p++; | |
1667 if (loc_reader) | |
1668 loc_reader->get_next (); | |
1669 } | |
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1670 p++; |
111 | 1671 if (loc_reader) |
1672 loc_reader->get_next (); | |
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1673 limit = from[i].text + from[i].len; |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1674 if (limit >= p + (p - prefix) + 1) |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1675 limit -= (p - prefix) + 1; |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1676 |
63
b7f97abdc517
update gcc from gcc-4.5.0 to gcc-4.6
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
55
diff
changeset
|
1677 /* Raw strings are all normal characters; these can be fed |
b7f97abdc517
update gcc from gcc-4.5.0 to gcc-4.6
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
55
diff
changeset
|
1678 directly to convert_cset. */ |
111 | 1679 if (to) |
1680 if (!APPLY_CONVERSION (cvt, p, limit - p, &tbuf)) | |
1681 goto fail; | |
1682 | |
1683 if (loc_reader) | |
1684 { | |
1685 /* If generating source ranges, assume we have a 1:1 | |
1686 correspondence between bytes in the source encoding and bytes | |
1687 in the execution encoding (e.g. if we have a UTF-8 to UTF-8 | |
1688 conversion), so that this run of bytes in the source file | |
1689 corresponds to a run of bytes in the execution string. | |
1690 This requirement is guaranteed by an early-reject in | |
1691 cpp_interpret_string_ranges. */ | |
1692 gcc_assert (cvt.func == convert_no_conversion); | |
1693 out->add_n_ranges (limit - p, *loc_reader); | |
1694 } | |
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1695 |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1696 continue; |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1697 } |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1698 |
111 | 1699 /* If we don't now have a leading quote, something has gone wrong. |
1700 This can occur if cpp_interpret_string_ranges is handling a | |
1701 stringified macro argument, but should not be possible otherwise. */ | |
1702 if (*p != '"' && *p != '\'') | |
1703 { | |
1704 gcc_assert (out != NULL); | |
1705 cpp_error (pfile, CPP_DL_ERROR, "missing open quote"); | |
1706 if (to) | |
1707 free (tbuf.text); | |
1708 return false; | |
1709 } | |
1710 | |
1711 /* Skip leading quote. */ | |
1712 p++; | |
1713 if (loc_reader) | |
1714 loc_reader->get_next (); | |
1715 | |
0 | 1716 limit = from[i].text + from[i].len - 1; /* Skip trailing quote. */ |
1717 | |
1718 for (;;) | |
1719 { | |
1720 base = p; | |
1721 while (p < limit && *p != '\\') | |
1722 p++; | |
1723 if (p > base) | |
1724 { | |
1725 /* We have a run of normal characters; these can be fed | |
1726 directly to convert_cset. */ | |
111 | 1727 if (to) |
1728 if (!APPLY_CONVERSION (cvt, base, p - base, &tbuf)) | |
1729 goto fail; | |
1730 /* Similar to above: assumes we have a 1:1 correspondence | |
1731 between bytes in the source encoding and bytes in the | |
1732 execution encoding. */ | |
1733 if (loc_reader) | |
1734 { | |
1735 gcc_assert (cvt.func == convert_no_conversion); | |
1736 out->add_n_ranges (p - base, *loc_reader); | |
1737 } | |
0 | 1738 } |
111 | 1739 if (p >= limit) |
0 | 1740 break; |
1741 | |
111 | 1742 struct _cpp_strbuf *tbuf_ptr = to ? &tbuf : NULL; |
1743 p = convert_escape (pfile, p + 1, limit, tbuf_ptr, cvt, | |
1744 loc_reader, out); | |
0 | 1745 } |
1746 } | |
111 | 1747 |
1748 if (to) | |
1749 { | |
1750 /* NUL-terminate the 'to' buffer and translate it to a cpp_string | |
1751 structure. */ | |
1752 emit_numeric_escape (pfile, 0, &tbuf, cvt); | |
1753 tbuf.text = XRESIZEVEC (uchar, tbuf.text, tbuf.len); | |
1754 to->text = tbuf.text; | |
1755 to->len = tbuf.len; | |
1756 } | |
1757 /* Use the location of the trailing quote as the location of the | |
1758 NUL-terminator. */ | |
1759 if (loc_reader) | |
1760 { | |
1761 source_range range = loc_reader->get_next (); | |
1762 out->add_range (range); | |
1763 } | |
1764 | |
0 | 1765 return true; |
1766 | |
1767 fail: | |
1768 cpp_errno (pfile, CPP_DL_ERROR, "converting to execution character set"); | |
111 | 1769 if (to) |
1770 free (tbuf.text); | |
0 | 1771 return false; |
1772 } | |
1773 | |
111 | 1774 /* FROM is an array of cpp_string structures of length COUNT. These |
1775 are to be converted from the source to the execution character set, | |
1776 escape sequences translated, and finally all are to be | |
1777 concatenated. WIDE indicates whether or not to produce a wide | |
1778 string. The result is written into TO. Returns true for success, | |
1779 false for failure. */ | |
1780 bool | |
1781 cpp_interpret_string (cpp_reader *pfile, const cpp_string *from, size_t count, | |
1782 cpp_string *to, enum cpp_ttype type) | |
1783 { | |
1784 return cpp_interpret_string_1 (pfile, from, count, to, type, NULL, NULL); | |
1785 } | |
1786 | |
131 | 1787 /* A "do nothing" diagnostic-handling callback for use by |
111 | 1788 cpp_interpret_string_ranges, so that it can temporarily suppress |
131 | 1789 diagnostic-handling. */ |
111 | 1790 |
1791 static bool | |
131 | 1792 noop_diagnostic_cb (cpp_reader *, enum cpp_diagnostic_level, |
1793 enum cpp_warning_reason, rich_location *, | |
1794 const char *, va_list *) | |
111 | 1795 { |
1796 /* no-op. */ | |
1797 return true; | |
1798 } | |
1799 | |
1800 /* This function mimics the behavior of cpp_interpret_string, but | |
1801 rather than generating a string in the execution character set, | |
1802 *OUT is written to with the source code ranges of the characters | |
1803 in such a string. | |
1804 FROM and LOC_READERS should both be arrays of length COUNT. | |
1805 Returns NULL for success, or an error message for failure. */ | |
1806 | |
1807 const char * | |
1808 cpp_interpret_string_ranges (cpp_reader *pfile, const cpp_string *from, | |
1809 cpp_string_location_reader *loc_readers, | |
1810 size_t count, | |
1811 cpp_substring_ranges *out, | |
1812 enum cpp_ttype type) | |
1813 { | |
1814 /* There are a couple of cases in the range-handling in | |
1815 cpp_interpret_string_1 that rely on there being a 1:1 correspondence | |
1816 between bytes in the source encoding and bytes in the execution | |
1817 encoding, so that each byte in the execution string can correspond | |
1818 to the location of a byte in the source string. | |
1819 | |
1820 This holds for the typical case of a UTF-8 to UTF-8 conversion. | |
1821 Enforce this requirement by only attempting to track substring | |
1822 locations if we have source encoding == execution encoding. | |
1823 | |
1824 This is a stronger condition than we need, since we could e.g. | |
1825 have ASCII to EBCDIC (with 1 byte per character before and after), | |
1826 but it seems to be a reasonable restriction. */ | |
1827 struct cset_converter cvt = converter_for_type (pfile, type); | |
1828 if (cvt.func != convert_no_conversion) | |
1829 return "execution character set != source character set"; | |
1830 | |
1831 /* For on-demand strings we have already lexed the strings, so there | |
131 | 1832 should be no diagnostics. However, if we have bogus source location |
111 | 1833 data (or stringified macro arguments), the attempt to lex the |
131 | 1834 strings could fail with an diagnostic. Temporarily install an |
1835 diagnostic-handler to catch the diagnostic, so that it can lead to this call | |
111 | 1836 failing, rather than being emitted as a user-visible diagnostic. |
131 | 1837 If an diagnostic does occur, we should see it via the return value of |
111 | 1838 cpp_interpret_string_1. */ |
131 | 1839 bool (*saved_diagnostic_handler) (cpp_reader *, enum cpp_diagnostic_level, |
1840 enum cpp_warning_reason, rich_location *, | |
1841 const char *, va_list *) | |
111 | 1842 ATTRIBUTE_FPTR_PRINTF(5,0); |
1843 | |
131 | 1844 saved_diagnostic_handler = pfile->cb.diagnostic; |
1845 pfile->cb.diagnostic = noop_diagnostic_cb; | |
111 | 1846 |
1847 bool result = cpp_interpret_string_1 (pfile, from, count, NULL, type, | |
1848 loc_readers, out); | |
1849 | |
131 | 1850 /* Restore the saved diagnostic-handler. */ |
1851 pfile->cb.diagnostic = saved_diagnostic_handler; | |
111 | 1852 |
1853 if (!result) | |
1854 return "cpp_interpret_string_1 failed"; | |
1855 | |
1856 /* Success. */ | |
1857 return NULL; | |
1858 } | |
1859 | |
0 | 1860 /* Subroutine of do_line and do_linemarker. Convert escape sequences |
1861 in a string, but do not perform character set conversion. */ | |
1862 bool | |
1863 cpp_interpret_string_notranslate (cpp_reader *pfile, const cpp_string *from, | |
1864 size_t count, cpp_string *to, | |
1865 enum cpp_ttype type ATTRIBUTE_UNUSED) | |
1866 { | |
1867 struct cset_converter save_narrow_cset_desc = pfile->narrow_cset_desc; | |
1868 bool retval; | |
1869 | |
1870 pfile->narrow_cset_desc.func = convert_no_conversion; | |
1871 pfile->narrow_cset_desc.cd = (iconv_t) -1; | |
1872 pfile->narrow_cset_desc.width = CPP_OPTION (pfile, char_precision); | |
1873 | |
1874 retval = cpp_interpret_string (pfile, from, count, to, CPP_STRING); | |
1875 | |
1876 pfile->narrow_cset_desc = save_narrow_cset_desc; | |
1877 return retval; | |
1878 } | |
1879 | |
1880 | |
1881 /* Subroutine of cpp_interpret_charconst which performs the conversion | |
1882 to a number, for narrow strings. STR is the string structure returned | |
1883 by cpp_interpret_string. PCHARS_SEEN and UNSIGNEDP are as for | |
145 | 1884 cpp_interpret_charconst. TYPE is the token type. */ |
0 | 1885 static cppchar_t |
1886 narrow_str_to_charconst (cpp_reader *pfile, cpp_string str, | |
145 | 1887 unsigned int *pchars_seen, int *unsignedp, |
1888 enum cpp_ttype type) | |
0 | 1889 { |
1890 size_t width = CPP_OPTION (pfile, char_precision); | |
1891 size_t max_chars = CPP_OPTION (pfile, int_precision) / width; | |
1892 size_t mask = width_to_mask (width); | |
1893 size_t i; | |
1894 cppchar_t result, c; | |
1895 bool unsigned_p; | |
1896 | |
1897 /* The value of a multi-character character constant, or a | |
1898 single-character character constant whose representation in the | |
1899 execution character set is more than one byte long, is | |
1900 implementation defined. This implementation defines it to be the | |
1901 number formed by interpreting the byte sequence in memory as a | |
1902 big-endian binary number. If overflow occurs, the high bytes are | |
1903 lost, and a warning is issued. | |
1904 | |
1905 We don't want to process the NUL terminator handed back by | |
1906 cpp_interpret_string. */ | |
1907 result = 0; | |
1908 for (i = 0; i < str.len - 1; i++) | |
1909 { | |
1910 c = str.text[i] & mask; | |
1911 if (width < BITS_PER_CPPCHAR_T) | |
1912 result = (result << width) | c; | |
1913 else | |
1914 result = c; | |
1915 } | |
1916 | |
145 | 1917 if (type == CPP_UTF8CHAR) |
1918 max_chars = 1; | |
0 | 1919 if (i > max_chars) |
1920 { | |
1921 i = max_chars; | |
145 | 1922 cpp_error (pfile, type == CPP_UTF8CHAR ? CPP_DL_ERROR : CPP_DL_WARNING, |
0 | 1923 "character constant too long for its type"); |
1924 } | |
1925 else if (i > 1 && CPP_OPTION (pfile, warn_multichar)) | |
63
b7f97abdc517
update gcc from gcc-4.5.0 to gcc-4.6
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
55
diff
changeset
|
1926 cpp_warning (pfile, CPP_W_MULTICHAR, "multi-character character constant"); |
0 | 1927 |
1928 /* Multichar constants are of type int and therefore signed. */ | |
1929 if (i > 1) | |
1930 unsigned_p = 0; | |
145 | 1931 else if (type == CPP_UTF8CHAR && !CPP_OPTION (pfile, cplusplus)) |
1932 unsigned_p = 1; | |
0 | 1933 else |
1934 unsigned_p = CPP_OPTION (pfile, unsigned_char); | |
1935 | |
1936 /* Truncate the constant to its natural width, and simultaneously | |
1937 sign- or zero-extend to the full width of cppchar_t. | |
1938 For single-character constants, the value is WIDTH bits wide. | |
1939 For multi-character constants, the value is INT_PRECISION bits wide. */ | |
1940 if (i > 1) | |
1941 width = CPP_OPTION (pfile, int_precision); | |
1942 if (width < BITS_PER_CPPCHAR_T) | |
1943 { | |
1944 mask = ((cppchar_t) 1 << width) - 1; | |
1945 if (unsigned_p || !(result & (1 << (width - 1)))) | |
1946 result &= mask; | |
1947 else | |
1948 result |= ~mask; | |
1949 } | |
1950 *pchars_seen = i; | |
1951 *unsignedp = unsigned_p; | |
1952 return result; | |
1953 } | |
1954 | |
1955 /* Subroutine of cpp_interpret_charconst which performs the conversion | |
1956 to a number, for wide strings. STR is the string structure returned | |
1957 by cpp_interpret_string. PCHARS_SEEN and UNSIGNEDP are as for | |
1958 cpp_interpret_charconst. TYPE is the token type. */ | |
1959 static cppchar_t | |
1960 wide_str_to_charconst (cpp_reader *pfile, cpp_string str, | |
1961 unsigned int *pchars_seen, int *unsignedp, | |
1962 enum cpp_ttype type) | |
1963 { | |
1964 bool bigend = CPP_OPTION (pfile, bytes_big_endian); | |
1965 size_t width = converter_for_type (pfile, type).width; | |
1966 size_t cwidth = CPP_OPTION (pfile, char_precision); | |
1967 size_t mask = width_to_mask (width); | |
1968 size_t cmask = width_to_mask (cwidth); | |
1969 size_t nbwc = width / cwidth; | |
1970 size_t off, i; | |
1971 cppchar_t result = 0, c; | |
1972 | |
145 | 1973 if (str.len <= nbwc) |
1974 { | |
1975 /* Error recovery, if no errors have been diagnosed previously, | |
1976 there should be at least two wide characters. Empty literals | |
1977 are diagnosed earlier and we can get just the zero terminator | |
1978 only if there were errors diagnosed during conversion. */ | |
1979 *pchars_seen = 0; | |
1980 *unsignedp = 0; | |
1981 return 0; | |
1982 } | |
1983 | |
0 | 1984 /* This is finicky because the string is in the target's byte order, |
1985 which may not be our byte order. Only the last character, ignoring | |
1986 the NUL terminator, is relevant. */ | |
1987 off = str.len - (nbwc * 2); | |
1988 result = 0; | |
1989 for (i = 0; i < nbwc; i++) | |
1990 { | |
1991 c = bigend ? str.text[off + i] : str.text[off + nbwc - i - 1]; | |
1992 result = (result << cwidth) | (c & cmask); | |
1993 } | |
1994 | |
1995 /* Wide character constants have type wchar_t, and a single | |
1996 character exactly fills a wchar_t, so a multi-character wide | |
1997 character constant is guaranteed to overflow. */ | |
1998 if (str.len > nbwc * 2) | |
145 | 1999 cpp_error (pfile, (CPP_OPTION (pfile, cplusplus) |
2000 && (type == CPP_CHAR16 || type == CPP_CHAR32)) | |
2001 ? CPP_DL_ERROR : CPP_DL_WARNING, | |
0 | 2002 "character constant too long for its type"); |
2003 | |
2004 /* Truncate the constant to its natural width, and simultaneously | |
2005 sign- or zero-extend to the full width of cppchar_t. */ | |
2006 if (width < BITS_PER_CPPCHAR_T) | |
2007 { | |
2008 if (type == CPP_CHAR16 || type == CPP_CHAR32 | |
2009 || CPP_OPTION (pfile, unsigned_wchar) | |
2010 || !(result & (1 << (width - 1)))) | |
2011 result &= mask; | |
2012 else | |
2013 result |= ~mask; | |
2014 } | |
2015 | |
2016 if (type == CPP_CHAR16 || type == CPP_CHAR32 | |
2017 || CPP_OPTION (pfile, unsigned_wchar)) | |
2018 *unsignedp = 1; | |
2019 else | |
2020 *unsignedp = 0; | |
2021 | |
2022 *pchars_seen = 1; | |
2023 return result; | |
2024 } | |
2025 | |
2026 /* Interpret a (possibly wide) character constant in TOKEN. | |
2027 PCHARS_SEEN points to a variable that is filled in with the number | |
2028 of characters seen, and UNSIGNEDP to a variable that indicates | |
2029 whether the result has signed type. */ | |
2030 cppchar_t | |
2031 cpp_interpret_charconst (cpp_reader *pfile, const cpp_token *token, | |
2032 unsigned int *pchars_seen, int *unsignedp) | |
2033 { | |
2034 cpp_string str = { 0, 0 }; | |
111 | 2035 bool wide = (token->type != CPP_CHAR && token->type != CPP_UTF8CHAR); |
2036 int u8 = 2 * int(token->type == CPP_UTF8CHAR); | |
0 | 2037 cppchar_t result; |
2038 | |
111 | 2039 /* An empty constant will appear as L'', u'', U'', u8'', or '' */ |
2040 if (token->val.str.len == (size_t) (2 + wide + u8)) | |
0 | 2041 { |
2042 cpp_error (pfile, CPP_DL_ERROR, "empty character constant"); | |
111 | 2043 *pchars_seen = 0; |
2044 *unsignedp = 0; | |
0 | 2045 return 0; |
2046 } | |
111 | 2047 else if (!cpp_interpret_string (pfile, &token->val.str, 1, &str, |
2048 token->type)) | |
2049 { | |
2050 *pchars_seen = 0; | |
2051 *unsignedp = 0; | |
2052 return 0; | |
2053 } | |
0 | 2054 |
2055 if (wide) | |
2056 result = wide_str_to_charconst (pfile, str, pchars_seen, unsignedp, | |
2057 token->type); | |
2058 else | |
145 | 2059 result = narrow_str_to_charconst (pfile, str, pchars_seen, unsignedp, |
2060 token->type); | |
0 | 2061 |
2062 if (str.text != token->val.str.text) | |
2063 free ((void *)str.text); | |
2064 | |
2065 return result; | |
2066 } | |
2067 | |
2068 /* Convert an identifier denoted by ID and LEN, which might contain | |
145 | 2069 UCN escapes or UTF-8 multibyte chars, to the source character set, |
2070 either UTF-8 or UTF-EBCDIC. Assumes that the identifier is actually | |
2071 a valid identifier. */ | |
0 | 2072 cpp_hashnode * |
2073 _cpp_interpret_identifier (cpp_reader *pfile, const uchar *id, size_t len) | |
2074 { | |
2075 /* It turns out that a UCN escape always turns into fewer characters | |
2076 than the escape itself, so we can allocate a temporary in advance. */ | |
2077 uchar * buf = (uchar *) alloca (len + 1); | |
2078 uchar * bufp = buf; | |
2079 size_t idp; | |
2080 | |
2081 for (idp = 0; idp < len; idp++) | |
2082 if (id[idp] != '\\') | |
2083 *bufp++ = id[idp]; | |
2084 else | |
2085 { | |
2086 unsigned length = id[idp+1] == 'u' ? 4 : 8; | |
2087 cppchar_t value = 0; | |
2088 size_t bufleft = len - (bufp - buf); | |
2089 int rval; | |
2090 | |
2091 idp += 2; | |
2092 while (length && idp < len && ISXDIGIT (id[idp])) | |
2093 { | |
2094 value = (value << 4) + hex_value (id[idp]); | |
2095 idp++; | |
2096 length--; | |
2097 } | |
2098 idp--; | |
2099 | |
2100 /* Special case for EBCDIC: if the identifier contains | |
2101 a '$' specified using a UCN, translate it to EBCDIC. */ | |
2102 if (value == 0x24) | |
2103 { | |
2104 *bufp++ = '$'; | |
2105 continue; | |
2106 } | |
2107 | |
2108 rval = one_cppchar_to_utf8 (value, &bufp, &bufleft); | |
2109 if (rval) | |
2110 { | |
2111 errno = rval; | |
2112 cpp_errno (pfile, CPP_DL_ERROR, | |
2113 "converting UCN to source character set"); | |
2114 break; | |
2115 } | |
2116 } | |
2117 | |
2118 return CPP_HASHNODE (ht_lookup (pfile->hash_table, | |
2119 buf, bufp - buf, HT_ALLOC)); | |
2120 } | |
2121 | |
2122 /* Convert an input buffer (containing the complete contents of one | |
2123 source file) from INPUT_CHARSET to the source character set. INPUT | |
2124 points to the input buffer, SIZE is its allocated size, and LEN is | |
2125 the length of the meaningful data within the buffer. The | |
2126 translated buffer is returned, *ST_SIZE is set to the length of | |
2127 the meaningful data within the translated buffer, and *BUFFER_START | |
2128 is set to the start of the returned buffer. *BUFFER_START may | |
2129 differ from the return value in the case of a BOM or other ignored | |
2130 marker information. | |
2131 | |
2132 INPUT is expected to have been allocated with xmalloc. This | |
2133 function will either set *BUFFER_START to INPUT, or free it and set | |
2134 *BUFFER_START to a pointer to another xmalloc-allocated block of | |
2135 memory. */ | |
2136 uchar * | |
2137 _cpp_convert_input (cpp_reader *pfile, const char *input_charset, | |
2138 uchar *input, size_t size, size_t len, | |
2139 const unsigned char **buffer_start, off_t *st_size) | |
2140 { | |
2141 struct cset_converter input_cset; | |
2142 struct _cpp_strbuf to; | |
2143 unsigned char *buffer; | |
2144 | |
2145 input_cset = init_iconv_desc (pfile, SOURCE_CHARSET, input_charset); | |
2146 if (input_cset.func == convert_no_conversion) | |
2147 { | |
2148 to.text = input; | |
2149 to.asize = size; | |
2150 to.len = len; | |
2151 } | |
2152 else | |
2153 { | |
2154 to.asize = MAX (65536, len); | |
2155 to.text = XNEWVEC (uchar, to.asize); | |
2156 to.len = 0; | |
2157 | |
2158 if (!APPLY_CONVERSION (input_cset, input, len, &to)) | |
2159 cpp_error (pfile, CPP_DL_ERROR, | |
2160 "failure to convert %s to %s", | |
2161 CPP_OPTION (pfile, input_charset), SOURCE_CHARSET); | |
2162 | |
2163 free (input); | |
2164 } | |
2165 | |
2166 /* Clean up the mess. */ | |
2167 if (input_cset.func == convert_using_iconv) | |
2168 iconv_close (input_cset.cd); | |
2169 | |
2170 /* Resize buffer if we allocated substantially too much, or if we | |
111 | 2171 haven't enough space for the \n-terminator or following |
2172 15 bytes of padding (used to quiet warnings from valgrind or | |
2173 Address Sanitizer, when the optimized lexer accesses aligned | |
2174 16-byte memory chunks, including the bytes after the malloced, | |
2175 area, and stops lexing on '\n'). */ | |
2176 if (to.len + 4096 < to.asize || to.len + 16 > to.asize) | |
2177 to.text = XRESIZEVEC (uchar, to.text, to.len + 16); | |
2178 | |
2179 memset (to.text + to.len, '\0', 16); | |
0 | 2180 |
2181 /* If the file is using old-school Mac line endings (\r only), | |
2182 terminate with another \r, not an \n, so that we do not mistake | |
2183 the \r\n sequence for a single DOS line ending and erroneously | |
2184 issue the "No newline at end of file" diagnostic. */ | |
2185 if (to.len && to.text[to.len - 1] == '\r') | |
2186 to.text[to.len] = '\r'; | |
2187 else | |
2188 to.text[to.len] = '\n'; | |
2189 | |
2190 buffer = to.text; | |
2191 *st_size = to.len; | |
2192 #if HOST_CHARSET == HOST_CHARSET_ASCII | |
2193 /* The HOST_CHARSET test just above ensures that the source charset | |
2194 is UTF-8. So, ignore a UTF-8 BOM if we see one. Note that | |
2195 glib'c UTF-8 iconv() provider (as of glibc 2.7) does not ignore a | |
2196 BOM -- however, even if it did, we would still need this code due | |
2197 to the 'convert_no_conversion' case. */ | |
2198 if (to.len >= 3 && to.text[0] == 0xef && to.text[1] == 0xbb | |
2199 && to.text[2] == 0xbf) | |
2200 { | |
2201 *st_size -= 3; | |
2202 buffer += 3; | |
2203 } | |
2204 #endif | |
2205 | |
2206 *buffer_start = to.text; | |
2207 return buffer; | |
2208 } | |
2209 | |
2210 /* Decide on the default encoding to assume for input files. */ | |
2211 const char * | |
2212 _cpp_default_encoding (void) | |
2213 { | |
2214 const char *current_encoding = NULL; | |
2215 | |
2216 /* We disable this because the default codeset is 7-bit ASCII on | |
2217 most platforms, and this causes conversion failures on every | |
2218 file in GCC that happens to have one of the upper 128 characters | |
2219 in it -- most likely, as part of the name of a contributor. | |
2220 We should definitely recognize in-band markers of file encoding, | |
2221 like: | |
2222 - the appropriate Unicode byte-order mark (FE FF) to recognize | |
2223 UTF16 and UCS4 (in both big-endian and little-endian flavors) | |
2224 and UTF8 | |
2225 - a "#i", "#d", "/ *", "//", " #p" or "#p" (for #pragma) to | |
2226 distinguish ASCII and EBCDIC. | |
2227 - now we can parse something like "#pragma GCC encoding <xyz> | |
2228 on the first line, or even Emacs/VIM's mode line tags (there's | |
2229 a problem here in that VIM uses the last line, and Emacs has | |
2230 its more elaborate "local variables" convention). | |
2231 - investigate whether Java has another common convention, which | |
2232 would be friendly to support. | |
2233 (Zack Weinberg and Paolo Bonzini, May 20th 2004) */ | |
2234 #if defined (HAVE_LOCALE_H) && defined (HAVE_LANGINFO_CODESET) && 0 | |
2235 setlocale (LC_CTYPE, ""); | |
2236 current_encoding = nl_langinfo (CODESET); | |
2237 #endif | |
2238 if (current_encoding == NULL || *current_encoding == '\0') | |
2239 current_encoding = SOURCE_CHARSET; | |
2240 | |
2241 return current_encoding; | |
2242 } | |
111 | 2243 |
2244 /* Implementation of class cpp_string_location_reader. */ | |
2245 | |
2246 /* Constructor for cpp_string_location_reader. */ | |
2247 | |
2248 cpp_string_location_reader:: | |
145 | 2249 cpp_string_location_reader (location_t src_loc, |
111 | 2250 line_maps *line_table) |
2251 { | |
2252 src_loc = get_range_from_loc (line_table, src_loc).m_start; | |
2253 | |
2254 /* SRC_LOC might be a macro location. It only makes sense to do | |
2255 column-by-column calculations on ordinary maps, so get the | |
2256 corresponding location in an ordinary map. */ | |
2257 m_loc | |
2258 = linemap_resolve_location (line_table, src_loc, | |
2259 LRK_SPELLING_LOCATION, NULL); | |
2260 | |
2261 const line_map_ordinary *map | |
2262 = linemap_check_ordinary (linemap_lookup (line_table, m_loc)); | |
2263 m_offset_per_column = (1 << map->m_range_bits); | |
2264 } | |
2265 | |
2266 /* Get the range of the next source byte. */ | |
2267 | |
2268 source_range | |
2269 cpp_string_location_reader::get_next () | |
2270 { | |
2271 source_range result; | |
2272 result.m_start = m_loc; | |
2273 result.m_finish = m_loc; | |
2274 if (m_loc <= LINE_MAP_MAX_LOCATION_WITH_COLS) | |
2275 m_loc += m_offset_per_column; | |
2276 return result; | |
2277 } | |
145 | 2278 |
2279 /* Helper for cpp_byte_column_to_display_column and its inverse. Given a | |
2280 pointer to a UTF-8-encoded character, compute its display width. *INBUFP | |
2281 points on entry to the start of the UTF-8 encoding of the character, and | |
2282 is updated to point just after the last byte of the encoding. *INBYTESLEFTP | |
2283 contains on entry the remaining size of the buffer into which *INBUFP | |
2284 points, and this is also updated accordingly. If *INBUFP does not | |
2285 point to a valid UTF-8-encoded sequence, then it will be treated as a single | |
2286 byte with display width 1. */ | |
2287 | |
2288 static inline int | |
2289 compute_next_display_width (const uchar **inbufp, size_t *inbytesleftp) | |
2290 { | |
2291 cppchar_t c; | |
2292 if (one_utf8_to_cppchar (inbufp, inbytesleftp, &c) != 0) | |
2293 { | |
2294 /* Input is not convertible to UTF-8. This could be fine, e.g. in a | |
2295 string literal, so don't complain. Just treat it as if it has a width | |
2296 of one. */ | |
2297 ++*inbufp; | |
2298 --*inbytesleftp; | |
2299 return 1; | |
2300 } | |
2301 | |
2302 /* one_utf8_to_cppchar() has updated inbufp and inbytesleftp for us. */ | |
2303 return cpp_wcwidth (c); | |
2304 } | |
2305 | |
2306 /* For the string of length DATA_LENGTH bytes that begins at DATA, compute | |
2307 how many display columns are occupied by the first COLUMN bytes. COLUMN | |
2308 may exceed DATA_LENGTH, in which case the phantom bytes at the end are | |
2309 treated as if they have display width 1. */ | |
2310 | |
2311 int | |
2312 cpp_byte_column_to_display_column (const char *data, int data_length, | |
2313 int column) | |
2314 { | |
2315 int display_col = 0; | |
2316 const uchar *udata = (const uchar *) data; | |
2317 const int offset = MAX (0, column - data_length); | |
2318 size_t inbytesleft = column - offset; | |
2319 while (inbytesleft) | |
2320 display_col += compute_next_display_width (&udata, &inbytesleft); | |
2321 return display_col + offset; | |
2322 } | |
2323 | |
2324 /* For the string of length DATA_LENGTH bytes that begins at DATA, compute | |
2325 the least number of bytes that will result in at least DISPLAY_COL display | |
2326 columns. The return value may exceed DATA_LENGTH if the entire string does | |
2327 not occupy enough display columns. */ | |
2328 | |
2329 int | |
2330 cpp_display_column_to_byte_column (const char *data, int data_length, | |
2331 int display_col) | |
2332 { | |
2333 int column = 0; | |
2334 const uchar *udata = (const uchar *) data; | |
2335 size_t inbytesleft = data_length; | |
2336 while (column < display_col && inbytesleft) | |
2337 column += compute_next_display_width (&udata, &inbytesleft); | |
2338 return data_length - inbytesleft + MAX (0, display_col - column); | |
2339 } | |
2340 | |
2341 /* Our own version of wcwidth(). We don't use the actual wcwidth() in glibc, | |
2342 because that will inspect the user's locale, and in particular in an ASCII | |
2343 locale, it will not return anything useful for extended characters. But GCC | |
2344 in other respects (see e.g. _cpp_default_encoding()) behaves as if | |
2345 everything is UTF-8. We also make some tweaks that are useful for the way | |
2346 GCC needs to use this data, e.g. tabs and other control characters should be | |
2347 treated as having width 1. The lookup tables are generated from | |
2348 contrib/unicode/gen_wcwidth.py and were made by simply calling glibc | |
2349 wcwidth() on all codepoints, then applying the small tweaks. These tables | |
2350 are not highly optimized, but for the present purpose of outputting | |
2351 diagnostics, they are sufficient. */ | |
2352 | |
2353 #include "generated_cpp_wcwidth.h" | |
2354 int cpp_wcwidth (cppchar_t c) | |
2355 { | |
2356 if (__builtin_expect (c <= wcwidth_range_ends[0], true)) | |
2357 return wcwidth_widths[0]; | |
2358 | |
2359 /* Binary search the tables. */ | |
2360 int begin = 1; | |
2361 static const int end | |
2362 = sizeof wcwidth_range_ends / sizeof (*wcwidth_range_ends); | |
2363 int len = end - begin; | |
2364 do | |
2365 { | |
2366 int half = len/2; | |
2367 int middle = begin + half; | |
2368 if (c > wcwidth_range_ends[middle]) | |
2369 { | |
2370 begin = middle + 1; | |
2371 len -= half + 1; | |
2372 } | |
2373 else | |
2374 len = half; | |
2375 } while (len); | |
2376 | |
2377 if (__builtin_expect (begin != end, true)) | |
2378 return wcwidth_widths[begin]; | |
2379 return 1; | |
2380 } |