1 | /* CPP Library - charsets |
2 | Copyright (C) 1998-2023 Free Software Foundation, Inc. |
3 | |
4 | Broken out of c-lex.cc 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 | { |
171 | static const uchar masks[6] = { 0x7F, 0x1F, 0x0F, 0x07, 0x03, 0x01 }; |
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, cp: &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 (c: 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, cp: &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 | |
356 | if (s <= 0xFFFF) |
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 (c: 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_conversion: 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_conversion: 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_conversion: 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_conversion: 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; |
540 | to->asize += to->asize / 4; |
541 | to->text = XRESIZEVEC (uchar, to->text, to->asize); |
542 | } |
543 | memcpy (dest: to->text + to->len, src: from, n: flen); |
544 | to->len += flen; |
545 | return true; |
546 | } |
547 | |
548 | /* And this one uses the system iconv primitive. It's a little |
549 | different, since iconv's interface is a little different. */ |
550 | #if HAVE_ICONV |
551 | |
552 | #define CONVERT_ICONV_GROW_BUFFER \ |
553 | do { \ |
554 | outbytesleft += OUTBUF_BLOCK_SIZE; \ |
555 | to->asize += OUTBUF_BLOCK_SIZE; \ |
556 | to->text = XRESIZEVEC (uchar, to->text, to->asize); \ |
557 | outbuf = (char *)to->text + to->asize - outbytesleft; \ |
558 | } while (0) |
559 | |
560 | static bool |
561 | convert_using_iconv (iconv_t cd, const uchar *from, size_t flen, |
562 | struct _cpp_strbuf *to) |
563 | { |
564 | ICONV_CONST char *inbuf; |
565 | char *outbuf; |
566 | size_t inbytesleft, outbytesleft; |
567 | |
568 | /* Reset conversion descriptor and check that it is valid. */ |
569 | if (iconv (cd, 0, 0, 0, 0) == (size_t)-1) |
570 | return false; |
571 | |
572 | inbuf = (ICONV_CONST char *)from; |
573 | inbytesleft = flen; |
574 | outbuf = (char *)to->text + to->len; |
575 | outbytesleft = to->asize - to->len; |
576 | |
577 | for (;;) |
578 | { |
579 | iconv (cd, &inbuf, &inbytesleft, &outbuf, &outbytesleft); |
580 | if (__builtin_expect (inbytesleft == 0, 1)) |
581 | { |
582 | /* Close out any shift states, returning to the initial state. */ |
583 | if (iconv (cd, 0, 0, &outbuf, &outbytesleft) == (size_t)-1) |
584 | { |
585 | if (errno != E2BIG) |
586 | return false; |
587 | |
588 | CONVERT_ICONV_GROW_BUFFER; |
589 | if (iconv (cd, 0, 0, &outbuf, &outbytesleft) == (size_t)-1) |
590 | return false; |
591 | } |
592 | |
593 | to->len = to->asize - outbytesleft; |
594 | return true; |
595 | } |
596 | if (errno != E2BIG) |
597 | return false; |
598 | |
599 | CONVERT_ICONV_GROW_BUFFER; |
600 | } |
601 | } |
602 | #else |
603 | #define convert_using_iconv 0 /* prevent undefined symbol error below */ |
604 | #endif |
605 | |
606 | /* Arrange for the above custom conversion logic to be used automatically |
607 | when conversion between a suitable pair of character sets is requested. */ |
608 | |
609 | #define APPLY_CONVERSION(CONVERTER, FROM, FLEN, TO) \ |
610 | CONVERTER.func (CONVERTER.cd, FROM, FLEN, TO) |
611 | |
612 | struct cpp_conversion |
613 | { |
614 | const char *pair; |
615 | convert_f func; |
616 | iconv_t fake_cd; |
617 | }; |
618 | static const struct cpp_conversion conversion_tab[] = { |
619 | { .pair: "UTF-8/UTF-32LE" , .func: convert_utf8_utf32, .fake_cd: (iconv_t)0 }, |
620 | { .pair: "UTF-8/UTF-32BE" , .func: convert_utf8_utf32, .fake_cd: (iconv_t)1 }, |
621 | { .pair: "UTF-8/UTF-16LE" , .func: convert_utf8_utf16, .fake_cd: (iconv_t)0 }, |
622 | { .pair: "UTF-8/UTF-16BE" , .func: convert_utf8_utf16, .fake_cd: (iconv_t)1 }, |
623 | { .pair: "UTF-32LE/UTF-8" , .func: convert_utf32_utf8, .fake_cd: (iconv_t)0 }, |
624 | { .pair: "UTF-32BE/UTF-8" , .func: convert_utf32_utf8, .fake_cd: (iconv_t)1 }, |
625 | { .pair: "UTF-16LE/UTF-8" , .func: convert_utf16_utf8, .fake_cd: (iconv_t)0 }, |
626 | { .pair: "UTF-16BE/UTF-8" , .func: convert_utf16_utf8, .fake_cd: (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 | |
635 | PFILE is only used for generating diagnostics; setting it to NULL |
636 | suppresses diagnostics. */ |
637 | |
638 | static struct cset_converter |
639 | init_iconv_desc (cpp_reader *pfile, const char *to, const char *from) |
640 | { |
641 | struct cset_converter ret; |
642 | char *pair; |
643 | size_t i; |
644 | |
645 | ret.to = to; |
646 | ret.from = from; |
647 | |
648 | if (!strcasecmp (s1: to, s2: from)) |
649 | { |
650 | ret.func = convert_no_conversion; |
651 | ret.cd = (iconv_t) -1; |
652 | ret.width = -1; |
653 | return ret; |
654 | } |
655 | |
656 | pair = (char *) alloca(strlen(to) + strlen(from) + 2); |
657 | |
658 | strcpy(dest: pair, src: from); |
659 | strcat(dest: pair, src: "/" ); |
660 | strcat(dest: pair, src: to); |
661 | for (i = 0; i < ARRAY_SIZE (conversion_tab); i++) |
662 | if (!strcasecmp (s1: pair, s2: conversion_tab[i].pair)) |
663 | { |
664 | ret.func = conversion_tab[i].func; |
665 | ret.cd = conversion_tab[i].fake_cd; |
666 | ret.width = -1; |
667 | return ret; |
668 | } |
669 | |
670 | /* No custom converter - try iconv. */ |
671 | if (HAVE_ICONV) |
672 | { |
673 | ret.func = convert_using_iconv; |
674 | ret.cd = iconv_open (to, from); |
675 | ret.width = -1; |
676 | |
677 | if (ret.cd == (iconv_t) -1) |
678 | { |
679 | if (pfile) |
680 | { |
681 | if (errno == EINVAL) |
682 | cpp_error (pfile, CPP_DL_ERROR, /* FIXME should be DL_SORRY */ |
683 | msgid: "conversion from %s to %s not supported by iconv" , |
684 | from, to); |
685 | else |
686 | cpp_errno (pfile, CPP_DL_ERROR, msgid: "iconv_open" ); |
687 | } |
688 | ret.func = convert_no_conversion; |
689 | } |
690 | } |
691 | else |
692 | { |
693 | if (pfile) |
694 | { |
695 | cpp_error (pfile, CPP_DL_ERROR, /* FIXME: should be DL_SORRY */ |
696 | msgid: "no iconv implementation, cannot convert from %s to %s" , |
697 | from, to); |
698 | } |
699 | ret.func = convert_no_conversion; |
700 | ret.cd = (iconv_t) -1; |
701 | ret.width = -1; |
702 | } |
703 | |
704 | return ret; |
705 | } |
706 | |
707 | /* If charset conversion is requested, initialize iconv(3) descriptors |
708 | for conversion from the source character set to the execution |
709 | character sets. If iconv is not present in the C library, and |
710 | conversion is requested, issue an error. */ |
711 | |
712 | void |
713 | cpp_init_iconv (cpp_reader *pfile) |
714 | { |
715 | const char *ncset = CPP_OPTION (pfile, narrow_charset); |
716 | const char *wcset = CPP_OPTION (pfile, wide_charset); |
717 | const char *default_wcset; |
718 | |
719 | bool be = CPP_OPTION (pfile, bytes_big_endian); |
720 | |
721 | if (CPP_OPTION (pfile, wchar_precision) >= 32) |
722 | default_wcset = be ? "UTF-32BE" : "UTF-32LE" ; |
723 | else if (CPP_OPTION (pfile, wchar_precision) >= 16) |
724 | default_wcset = be ? "UTF-16BE" : "UTF-16LE" ; |
725 | else |
726 | /* This effectively means that wide strings are not supported, |
727 | so don't do any conversion at all. */ |
728 | default_wcset = SOURCE_CHARSET; |
729 | |
730 | if (!ncset) |
731 | ncset = SOURCE_CHARSET; |
732 | if (!wcset) |
733 | wcset = default_wcset; |
734 | |
735 | pfile->narrow_cset_desc = init_iconv_desc (pfile, to: ncset, SOURCE_CHARSET); |
736 | pfile->narrow_cset_desc.width = CPP_OPTION (pfile, char_precision); |
737 | pfile->utf8_cset_desc = init_iconv_desc (pfile, to: "UTF-8" , SOURCE_CHARSET); |
738 | pfile->utf8_cset_desc.width = CPP_OPTION (pfile, char_precision); |
739 | pfile->char16_cset_desc = init_iconv_desc (pfile, |
740 | to: be ? "UTF-16BE" : "UTF-16LE" , |
741 | SOURCE_CHARSET); |
742 | pfile->char16_cset_desc.width = 16; |
743 | pfile->char32_cset_desc = init_iconv_desc (pfile, |
744 | to: be ? "UTF-32BE" : "UTF-32LE" , |
745 | SOURCE_CHARSET); |
746 | pfile->char32_cset_desc.width = 32; |
747 | pfile->wide_cset_desc = init_iconv_desc (pfile, to: wcset, SOURCE_CHARSET); |
748 | pfile->wide_cset_desc.width = CPP_OPTION (pfile, wchar_precision); |
749 | } |
750 | |
751 | /* Destroy iconv(3) descriptors set up by cpp_init_iconv, if necessary. */ |
752 | void |
753 | _cpp_destroy_iconv (cpp_reader *pfile) |
754 | { |
755 | if (HAVE_ICONV) |
756 | { |
757 | if (pfile->narrow_cset_desc.func == convert_using_iconv) |
758 | iconv_close (pfile->narrow_cset_desc.cd); |
759 | if (pfile->utf8_cset_desc.func == convert_using_iconv) |
760 | iconv_close (pfile->utf8_cset_desc.cd); |
761 | if (pfile->char16_cset_desc.func == convert_using_iconv) |
762 | iconv_close (pfile->char16_cset_desc.cd); |
763 | if (pfile->char32_cset_desc.func == convert_using_iconv) |
764 | iconv_close (pfile->char32_cset_desc.cd); |
765 | if (pfile->wide_cset_desc.func == convert_using_iconv) |
766 | iconv_close (pfile->wide_cset_desc.cd); |
767 | } |
768 | } |
769 | |
770 | /* Utility routine for use by a full compiler. C is a character taken |
771 | from the *basic* source character set, encoded in the host's |
772 | execution encoding. Convert it to (the target's) execution |
773 | encoding, and return that value. |
774 | |
775 | Issues an internal error if C's representation in the narrow |
776 | execution character set fails to be a single-byte value (C99 |
777 | 5.2.1p3: "The representation of each member of the source and |
778 | execution character sets shall fit in a byte.") May also issue an |
779 | internal error if C fails to be a member of the basic source |
780 | character set (testing this exactly is too hard, especially when |
781 | the host character set is EBCDIC). */ |
782 | cppchar_t |
783 | cpp_host_to_exec_charset (cpp_reader *pfile, cppchar_t c) |
784 | { |
785 | uchar sbuf[1]; |
786 | struct _cpp_strbuf tbuf; |
787 | |
788 | /* This test is merely an approximation, but it suffices to catch |
789 | the most important thing, which is that we don't get handed a |
790 | character outside the unibyte range of the host character set. */ |
791 | if (c > LAST_POSSIBLY_BASIC_SOURCE_CHAR) |
792 | { |
793 | cpp_error (pfile, CPP_DL_ICE, |
794 | msgid: "character 0x%lx is not in the basic source character set\n" , |
795 | (unsigned long)c); |
796 | return 0; |
797 | } |
798 | |
799 | /* Being a character in the unibyte range of the host character set, |
800 | we can safely splat it into a one-byte buffer and trust that that |
801 | is a well-formed string. */ |
802 | sbuf[0] = c; |
803 | |
804 | /* This should never need to reallocate, but just in case... */ |
805 | tbuf.asize = 1; |
806 | tbuf.text = XNEWVEC (uchar, tbuf.asize); |
807 | tbuf.len = 0; |
808 | |
809 | if (!APPLY_CONVERSION (pfile->narrow_cset_desc, sbuf, 1, &tbuf)) |
810 | { |
811 | cpp_errno (pfile, CPP_DL_ICE, msgid: "converting to execution character set" ); |
812 | return 0; |
813 | } |
814 | if (tbuf.len != 1) |
815 | { |
816 | cpp_error (pfile, CPP_DL_ICE, |
817 | msgid: "character 0x%lx is not unibyte in execution character set" , |
818 | (unsigned long)c); |
819 | return 0; |
820 | } |
821 | c = tbuf.text[0]; |
822 | free(ptr: tbuf.text); |
823 | return c; |
824 | } |
825 | |
826 | |
827 | |
828 | /* cpp_substring_ranges's constructor. */ |
829 | |
830 | cpp_substring_ranges::cpp_substring_ranges () : |
831 | m_ranges (NULL), |
832 | m_num_ranges (0), |
833 | m_alloc_ranges (8) |
834 | { |
835 | m_ranges = XNEWVEC (source_range, m_alloc_ranges); |
836 | } |
837 | |
838 | /* cpp_substring_ranges's destructor. */ |
839 | |
840 | cpp_substring_ranges::~cpp_substring_ranges () |
841 | { |
842 | free (ptr: m_ranges); |
843 | } |
844 | |
845 | /* Add RANGE to the vector of source_range information. */ |
846 | |
847 | void |
848 | cpp_substring_ranges::add_range (source_range range) |
849 | { |
850 | if (m_num_ranges >= m_alloc_ranges) |
851 | { |
852 | m_alloc_ranges *= 2; |
853 | m_ranges |
854 | = (source_range *)xrealloc (m_ranges, |
855 | sizeof (source_range) * m_alloc_ranges); |
856 | } |
857 | m_ranges[m_num_ranges++] = range; |
858 | } |
859 | |
860 | /* Read NUM ranges from LOC_READER, adding them to the vector of source_range |
861 | information. */ |
862 | |
863 | void |
864 | cpp_substring_ranges::add_n_ranges (int num, |
865 | cpp_string_location_reader &loc_reader) |
866 | { |
867 | for (int i = 0; i < num; i++) |
868 | add_range (range: loc_reader.get_next ()); |
869 | } |
870 | |
871 | |
872 | |
873 | /* Utility routine that computes a mask of the form 0000...111... with |
874 | WIDTH 1-bits. */ |
875 | static inline size_t |
876 | width_to_mask (size_t width) |
877 | { |
878 | width = MIN (width, BITS_PER_CPPCHAR_T); |
879 | if (width >= CHAR_BIT * sizeof (size_t)) |
880 | return ~(size_t) 0; |
881 | else |
882 | return ((size_t) 1 << width) - 1; |
883 | } |
884 | |
885 | /* A large table of unicode character information. */ |
886 | enum { |
887 | /* Valid in a C99 identifier? */ |
888 | C99 = 1, |
889 | /* Valid in a C99 identifier, but not as the first character? */ |
890 | N99 = 2, |
891 | /* Valid in a C++ identifier? */ |
892 | CXX = 4, |
893 | /* Valid in a C11/C++11 identifier? */ |
894 | C11 = 8, |
895 | /* Valid in a C11/C++11 identifier, but not as the first character? */ |
896 | N11 = 16, |
897 | /* Valid in a C++23 identifier? */ |
898 | CXX23 = 32, |
899 | /* Valid in a C++23 identifier, but not as the first character? */ |
900 | NXX23 = 64, |
901 | /* NFC representation is not valid in an identifier? */ |
902 | CID = 128, |
903 | /* Might be valid NFC form? */ |
904 | NFC = 256, |
905 | /* Might be valid NFKC form? */ |
906 | NKC = 512, |
907 | /* Certain preceding characters might make it not valid NFC/NKFC form? */ |
908 | CTX = 1024 |
909 | }; |
910 | |
911 | struct ucnrange { |
912 | /* Bitmap of flags above. */ |
913 | unsigned short flags; |
914 | /* Combining class of the character. */ |
915 | unsigned char combine; |
916 | /* Last character in the range described by this entry. */ |
917 | unsigned int end; |
918 | }; |
919 | #include "ucnid.h" |
920 | |
921 | /* ISO 10646 defines the UCS codespace as the range 0-0x10FFFF inclusive. */ |
922 | #define UCS_LIMIT 0x10FFFF |
923 | |
924 | #include "uname2c.h" |
925 | |
926 | static const char hangul_syllables[][4] = { |
927 | /* L */ |
928 | "G" , "GG" , "N" , "D" , "DD" , "R" , "M" , "B" , "BB" , "S" , "SS" , "" , |
929 | "J" , "JJ" , "C" , "K" , "T" , "P" , "H" , |
930 | /* V */ |
931 | "A" , "AE" , "YA" , "YAE" , "EO" , "E" , "YEO" , "YE" , "O" , "WA" , "WAE" , |
932 | "OE" , "YO" , "U" , "WEO" , "WE" , "WI" , "YU" , "EU" , "YI" , "I" , |
933 | /* T */ |
934 | "" , "G" , "GG" , "GS" , "N" , "NJ" , "NH" , "D" , "L" , "LG" , "LM" , "LB" , |
935 | "LS" , "LT" , "LP" , "LH" , "M" , "B" , "BS" , "S" , "SS" , "NG" , "J" , "C" , |
936 | "K" , "T" , "P" , "H" |
937 | }; |
938 | |
939 | static const short hangul_count[6] = { 19, 21, 28 }; |
940 | |
941 | /* Used for Unicode loose matching rule UAX44-LM2 matching. */ |
942 | |
943 | struct uname2c_data |
944 | { |
945 | char *canon_name; |
946 | char prev_char; |
947 | }; |
948 | |
949 | /* Map NAME, a Unicode character name or correction/control/alternate |
950 | alias, to a Unicode codepoint, or return (cppchar_t) -1 if |
951 | not found. This uses a space optimized radix tree precomputed |
952 | by the makeuname2c utility, with binary format documented in its |
953 | source makeuname2c.cc. */ |
954 | |
955 | static cppchar_t |
956 | _cpp_uname2c (const char *name, size_t len, const unsigned char *n, |
957 | struct uname2c_data *data) |
958 | { |
959 | do |
960 | { |
961 | char k; |
962 | const char *key; |
963 | size_t key_len, len_adj; |
964 | bool has_value = *n & 0x40; |
965 | bool has_children, no_sibling = false; |
966 | cppchar_t codepoint = -1; |
967 | const unsigned char *child = NULL; |
968 | int ret; |
969 | |
970 | if (*n & 0x80) |
971 | { |
972 | k = ' ' + (*n++ & 0x3f); |
973 | key = &k; |
974 | key_len = 1; |
975 | } |
976 | else |
977 | { |
978 | key_len = *n++ & 0x3f; |
979 | key = &uname2c_dict[*n++]; |
980 | key += (*n++ << 8); |
981 | } |
982 | if (has_value) |
983 | { |
984 | codepoint = *n + (n[1] << 8) + ((n[2] & 0x1f) << 16); |
985 | has_children = n[2] & 0x80; |
986 | no_sibling = n[2] & 0x40; |
987 | n += 3; |
988 | } |
989 | else |
990 | has_children = true; |
991 | if (has_children) |
992 | { |
993 | unsigned int shift = 0; |
994 | size_t child_off = 0; |
995 | |
996 | do |
997 | { |
998 | child_off |= (*n & 0x7f) << shift; |
999 | shift += 7; |
1000 | } |
1001 | while ((*n++ & 0x80) != 0); |
1002 | child = n + child_off; |
1003 | } |
1004 | if (__builtin_expect (data == NULL, 1)) |
1005 | { |
1006 | ret = memcmp (s1: name, s2: key, n: len > key_len ? key_len : len); |
1007 | len_adj = key_len; |
1008 | } |
1009 | else |
1010 | { |
1011 | const char *p = name, *q = key; |
1012 | |
1013 | while (1) |
1014 | { |
1015 | if ((size_t) (p - name) == len || (size_t) (q - key) == key_len) |
1016 | break; |
1017 | if (*q == ' ') |
1018 | { |
1019 | ++q; |
1020 | continue; |
1021 | } |
1022 | if (*q == '-') |
1023 | { |
1024 | /* This is the hard case. Only medial hyphens |
1025 | should be removed, where medial means preceded |
1026 | and followed by alnum. */ |
1027 | if (ISALNUM (q == key ? data->prev_char : q[-1])) |
1028 | { |
1029 | if (q + 1 == key + key_len) |
1030 | { |
1031 | /* We don't know what the next letter will be. |
1032 | It could be ISALNUM, then we are supposed |
1033 | to omit it, or it could be a space and then |
1034 | we should not omit it and need to compare it. |
1035 | Fortunately the only 3 names with hyphen |
1036 | followed by non-letter are |
1037 | U+0F0A TIBETAN MARK BKA- SHOG YIG MGO |
1038 | U+0FD0 TIBETAN MARK BKA- SHOG GI MGO RGYAN |
1039 | U+0FD0 TIBETAN MARK BSKA- SHOG GI MGO RGYAN |
1040 | Furthermore, prefixes of NR2 generated |
1041 | ranges all end with a hyphen, but the generated |
1042 | part is then followed by alpha-numeric. |
1043 | So, let's just assume that - at the end of |
1044 | key is always followed by alphanumeric and |
1045 | so should be omitted. |
1046 | makeuname2c.cc verifies that this is true. */ |
1047 | ++q; |
1048 | continue; |
1049 | } |
1050 | else if (ISALNUM (q[1])) |
1051 | { |
1052 | ++q; |
1053 | continue; |
1054 | } |
1055 | } |
1056 | } |
1057 | if (*p != *q) |
1058 | break; |
1059 | ++p; |
1060 | ++q; |
1061 | } |
1062 | len_adj = p - name; |
1063 | /* If we don't consume the whole key, signal a mismatch, |
1064 | but always with ret = 1, so that we keep looking through |
1065 | siblings. */ |
1066 | ret = q < key + key_len; |
1067 | } |
1068 | if (ret < 0) |
1069 | return -1; |
1070 | else if (ret == 0) |
1071 | { |
1072 | if (len < len_adj) |
1073 | return -1; |
1074 | else if (codepoint >= 0xd800 |
1075 | && codepoint < 0xd800 + ARRAY_SIZE (uname2c_generated)) |
1076 | { |
1077 | name += len_adj; |
1078 | len -= len_adj; |
1079 | if (codepoint == 0xd800) |
1080 | { |
1081 | /* NR1 - Hangul syllables. */ |
1082 | size_t start = 0, end, i, j; |
1083 | int this_len, max_len; |
1084 | char winner[3]; |
1085 | |
1086 | for (i = 0; i < 3; ++i) |
1087 | { |
1088 | end = start + hangul_count[i]; |
1089 | max_len = -1; |
1090 | winner[i] = -1; |
1091 | for (j = start; j < end; j++) |
1092 | { |
1093 | this_len = strlen (s: hangul_syllables[j]); |
1094 | if (len >= (size_t) this_len |
1095 | && this_len > max_len |
1096 | && memcmp (s1: name, s2: hangul_syllables[j], |
1097 | n: this_len) == 0) |
1098 | { |
1099 | max_len = this_len; |
1100 | winner[i] = j - start; |
1101 | } |
1102 | } |
1103 | if (max_len == -1) |
1104 | return -1; |
1105 | name += max_len; |
1106 | len -= max_len; |
1107 | start = end; |
1108 | } |
1109 | if (__builtin_expect (data != NULL, 0)) |
1110 | { |
1111 | memcpy (dest: data->canon_name, src: key, n: key_len); |
1112 | data->canon_name[key_len] = '\0'; |
1113 | for (i = 0, start = 0; i < 3; ++i) |
1114 | { |
1115 | strcat (dest: data->canon_name, |
1116 | src: hangul_syllables[start + winner[i]]); |
1117 | start += hangul_count[i]; |
1118 | } |
1119 | } |
1120 | return (0xac00 + 21 * 28 * winner[0] |
1121 | + 28 * winner[1] + winner[2]); |
1122 | } |
1123 | else |
1124 | { |
1125 | /* NR2 - prefix followed by hexadecimal codepoint. */ |
1126 | const cppchar_t *p; |
1127 | size_t i; |
1128 | |
1129 | if (len < 4 || len > 5) |
1130 | return -1; |
1131 | p = uname2c_pairs + uname2c_generated[codepoint - 0xd800]; |
1132 | codepoint = 0; |
1133 | for (i = 0; i < len; ++i) |
1134 | { |
1135 | codepoint <<= 4; |
1136 | if (!ISXDIGIT (name[i])) |
1137 | return -1; |
1138 | codepoint += hex_value (name[i]); |
1139 | } |
1140 | for (; *p; p += 2) |
1141 | if (codepoint < *p) |
1142 | return -1; |
1143 | else if (codepoint <= p[1]) |
1144 | { |
1145 | if (__builtin_expect (data != NULL, 0)) |
1146 | { |
1147 | memcpy (dest: data->canon_name, src: key, n: key_len); |
1148 | memcpy (dest: data->canon_name + key_len, src: name, n: len); |
1149 | data->canon_name[key_len + len] = '\0'; |
1150 | } |
1151 | return codepoint; |
1152 | } |
1153 | return -1; |
1154 | } |
1155 | } |
1156 | else if (__builtin_expect (data != NULL, 0)) |
1157 | { |
1158 | if (len == len_adj) |
1159 | { |
1160 | memcpy (dest: data->canon_name, src: key, n: key_len); |
1161 | data->canon_name[key_len] = '\0'; |
1162 | return codepoint; |
1163 | } |
1164 | if (has_children) |
1165 | { |
1166 | struct uname2c_data save = *data; |
1167 | memcpy (dest: data->canon_name, src: key, n: key_len); |
1168 | data->canon_name += key_len; |
1169 | data->prev_char = key[key_len - 1]; |
1170 | codepoint = _cpp_uname2c (name: name + len_adj, len: len - len_adj, |
1171 | n: child, data); |
1172 | if (codepoint != (cppchar_t) -1) |
1173 | return codepoint; |
1174 | *data = save; |
1175 | } |
1176 | } |
1177 | else if (len == len_adj) |
1178 | return codepoint; |
1179 | else if (!has_children) |
1180 | return -1; |
1181 | else |
1182 | { |
1183 | name += len_adj; |
1184 | len -= len_adj; |
1185 | n = child; |
1186 | continue; |
1187 | } |
1188 | } |
1189 | if (no_sibling || (!has_value && *n == 0xff)) |
1190 | break; |
1191 | } |
1192 | while (1); |
1193 | return -1; |
1194 | } |
1195 | |
1196 | /* Try to do a loose name lookup according to Unicode loose matching rule |
1197 | UAX44-LM2. First ignore medial hyphens, whitespace, underscore |
1198 | characters and convert to upper case. */ |
1199 | |
1200 | static cppchar_t |
1201 | _cpp_uname2c_uax44_lm2 (const char *name, size_t len, char *canon_name) |
1202 | { |
1203 | char name_after_uax44_lm2[uname2c_max_name_len]; |
1204 | char *q = name_after_uax44_lm2; |
1205 | const char *p; |
1206 | |
1207 | for (p = name; p < name + len; p++) |
1208 | if (*p == '_' || *p == ' ') |
1209 | continue; |
1210 | else if (*p == '-' && p != name && ISALNUM (p[-1]) && ISALNUM (p[1])) |
1211 | continue; |
1212 | else if (q == name_after_uax44_lm2 + uname2c_max_name_len) |
1213 | return -1; |
1214 | else if (ISLOWER (*p)) |
1215 | *q++ = TOUPPER (*p); |
1216 | else |
1217 | *q++ = *p; |
1218 | |
1219 | struct uname2c_data data; |
1220 | data.canon_name = canon_name; |
1221 | data.prev_char = ' '; |
1222 | /* Hangul Jungseong O- E after UAX44-LM2 should be HANGULJUNGSEONGO-E |
1223 | and so should match U+1180. */ |
1224 | if (q - name_after_uax44_lm2 == sizeof ("HANGULJUNGSEONGO-E" ) - 1 |
1225 | && memcmp (s1: name_after_uax44_lm2, s2: "HANGULJUNGSEONGO-E" , |
1226 | n: sizeof ("HANGULJUNGSEONGO-E" ) - 1) == 0) |
1227 | { |
1228 | name_after_uax44_lm2[sizeof ("HANGULJUNGSEONGO" ) - 1] = 'E'; |
1229 | --q; |
1230 | } |
1231 | cppchar_t result |
1232 | = _cpp_uname2c (name: name_after_uax44_lm2, len: q - name_after_uax44_lm2, |
1233 | n: uname2c_tree, data: &data); |
1234 | |
1235 | /* Unicode UAX44-LM2 exception: |
1236 | U+116C HANGUL JUNGSEONG OE |
1237 | U+1180 HANGUL JUNGSEONG O-E |
1238 | We remove all medial hyphens when we shouldn't remote the U+1180 one. |
1239 | The U+1180 entry sorts before U+116C lexicographilly, so we get U+1180 |
1240 | in both cases. Thus, if result is U+1180, check if user's name doesn't |
1241 | have a hyphen there and adjust. */ |
1242 | if (result == 0x1180) |
1243 | { |
1244 | while (p[-1] == ' ' || p[-1] == '_') |
1245 | --p; |
1246 | gcc_assert (TOUPPER (p[-1]) == 'E'); |
1247 | --p; |
1248 | while (p[-1] == ' ' || p[-1] == '_') |
1249 | --p; |
1250 | if (p[-1] != '-') |
1251 | { |
1252 | result = 0x116c; |
1253 | memcpy (dest: canon_name + sizeof ("HANGUL JUNGSEONG O" ) - 1, src: "E" , n: 2); |
1254 | } |
1255 | } |
1256 | return result; |
1257 | } |
1258 | |
1259 | |
1260 | /* Returns 1 if C is valid in an identifier, 2 if C is valid except at |
1261 | the start of an identifier, and 0 if C is not valid in an |
1262 | identifier. We assume C has already gone through the checks of |
1263 | _cpp_valid_ucn. Also update NST for C if returning nonzero. The |
1264 | algorithm is a simple binary search on the table defined in |
1265 | ucnid.h. */ |
1266 | |
1267 | static int |
1268 | ucn_valid_in_identifier (cpp_reader *pfile, cppchar_t c, |
1269 | struct normalize_state *nst) |
1270 | { |
1271 | int mn, mx, md; |
1272 | unsigned short valid_flags, invalid_start_flags; |
1273 | |
1274 | if (c > UCS_LIMIT) |
1275 | return 0; |
1276 | |
1277 | mn = 0; |
1278 | mx = ARRAY_SIZE (ucnranges) - 1; |
1279 | while (mx != mn) |
1280 | { |
1281 | md = (mn + mx) / 2; |
1282 | if (c <= ucnranges[md].end) |
1283 | mx = md; |
1284 | else |
1285 | mn = md + 1; |
1286 | } |
1287 | |
1288 | /* When -pedantic, we require the character to have been listed by |
1289 | the standard for the current language. Otherwise, we accept the |
1290 | union of the acceptable sets for all supported language versions. */ |
1291 | valid_flags = C99 | CXX | C11 | CXX23; |
1292 | if (CPP_PEDANTIC (pfile)) |
1293 | { |
1294 | if (CPP_OPTION (pfile, xid_identifiers)) |
1295 | valid_flags = CXX23; |
1296 | else if (CPP_OPTION (pfile, c11_identifiers)) |
1297 | valid_flags = C11; |
1298 | else if (CPP_OPTION (pfile, c99)) |
1299 | valid_flags = C99; |
1300 | } |
1301 | if (! (ucnranges[mn].flags & valid_flags)) |
1302 | return 0; |
1303 | |
1304 | /* Update NST. */ |
1305 | if (ucnranges[mn].combine != 0 && ucnranges[mn].combine < nst->prev_class) |
1306 | nst->level = normalized_none; |
1307 | else if (ucnranges[mn].flags & CTX) |
1308 | { |
1309 | bool safe; |
1310 | cppchar_t p = nst->previous; |
1311 | |
1312 | /* For Hangul, characters in the range AC00-D7A3 are NFC/NFKC, |
1313 | and are combined algorithmically from a sequence of the form |
1314 | 1100-1112 1161-1175 11A8-11C2 |
1315 | (if the third is not present, it is treated as 11A7, which is not |
1316 | really a valid character). |
1317 | Unfortunately, C99 allows (only) the NFC form, but C++ allows |
1318 | only the combining characters. */ |
1319 | if (c >= 0x1161 && c <= 0x1175) |
1320 | safe = p < 0x1100 || p > 0x1112; |
1321 | else if (c >= 0x11A8 && c <= 0x11C2) |
1322 | safe = (p < 0xAC00 || p > 0xD7A3 || (p - 0xAC00) % 28 != 0); |
1323 | else |
1324 | safe = check_nfc (pfile, c, p); |
1325 | if (!safe) |
1326 | { |
1327 | if ((c >= 0x1161 && c <= 0x1175) || (c >= 0x11A8 && c <= 0x11C2)) |
1328 | nst->level = MAX (nst->level, normalized_identifier_C); |
1329 | else |
1330 | nst->level = normalized_none; |
1331 | } |
1332 | } |
1333 | else if (ucnranges[mn].flags & NKC) |
1334 | ; |
1335 | else if (ucnranges[mn].flags & NFC) |
1336 | nst->level = MAX (nst->level, normalized_C); |
1337 | else if (ucnranges[mn].flags & CID) |
1338 | nst->level = MAX (nst->level, normalized_identifier_C); |
1339 | else |
1340 | nst->level = normalized_none; |
1341 | if (ucnranges[mn].combine == 0) |
1342 | nst->previous = c; |
1343 | nst->prev_class = ucnranges[mn].combine; |
1344 | |
1345 | if (!CPP_PEDANTIC (pfile)) |
1346 | { |
1347 | /* If not -pedantic, accept as character that may |
1348 | begin an identifier a union of characters allowed |
1349 | at that position in each of the character sets. */ |
1350 | if ((ucnranges[mn].flags & (C99 | N99)) == C99 |
1351 | || (ucnranges[mn].flags & CXX) != 0 |
1352 | || (ucnranges[mn].flags & (C11 | N11)) == C11 |
1353 | || (ucnranges[mn].flags & (CXX23 | NXX23)) == CXX23) |
1354 | return 1; |
1355 | return 2; |
1356 | } |
1357 | |
1358 | if (CPP_OPTION (pfile, xid_identifiers)) |
1359 | invalid_start_flags = NXX23; |
1360 | else if (CPP_OPTION (pfile, c11_identifiers)) |
1361 | invalid_start_flags = N11; |
1362 | else if (CPP_OPTION (pfile, c99)) |
1363 | invalid_start_flags = N99; |
1364 | else |
1365 | invalid_start_flags = 0; |
1366 | |
1367 | /* In C99, UCN digits may not begin identifiers. In C11 and C++11, |
1368 | UCN combining characters may not begin identifiers. */ |
1369 | if (ucnranges[mn].flags & invalid_start_flags) |
1370 | return 2; |
1371 | |
1372 | return 1; |
1373 | } |
1374 | |
1375 | /* Increment char_range->m_finish by a single character. */ |
1376 | |
1377 | static void |
1378 | extend_char_range (source_range *char_range, |
1379 | cpp_string_location_reader *loc_reader) |
1380 | { |
1381 | if (loc_reader) |
1382 | { |
1383 | gcc_assert (char_range); |
1384 | char_range->m_finish = loc_reader->get_next ().m_finish; |
1385 | } |
1386 | } |
1387 | |
1388 | /* [lex.charset]: The character designated by the universal character |
1389 | name \UNNNNNNNN is that character whose character short name in |
1390 | ISO/IEC 10646 is NNNNNNNN; the character designated by the |
1391 | universal character name \uNNNN is that character whose character |
1392 | short name in ISO/IEC 10646 is 0000NNNN. If the hexadecimal value |
1393 | for a universal character name corresponds to a surrogate code point |
1394 | (in the range 0xD800-0xDFFF, inclusive), the program is ill-formed. |
1395 | Additionally, if the hexadecimal value for a universal-character-name |
1396 | outside a character or string literal corresponds to a control character |
1397 | (in either of the ranges 0x00-0x1F or 0x7F-0x9F, both inclusive) or to a |
1398 | character in the basic source character set, the program is ill-formed. |
1399 | |
1400 | C99 6.4.3: A universal character name shall not specify a character |
1401 | whose short identifier is less than 00A0 other than 0024 ($), 0040 (@), |
1402 | or 0060 (`), nor one in the range D800 through DFFF inclusive. |
1403 | |
1404 | If the hexadecimal value is larger than the upper bound of the UCS |
1405 | codespace specified in ISO/IEC 10646, a pedantic warning is issued |
1406 | in all versions of C and in the C++20 or later versions of C++. |
1407 | |
1408 | *PSTR must be preceded by "\u" or "\U"; it is assumed that the |
1409 | buffer end is delimited by a non-hex digit. Returns false if the |
1410 | UCN has not been consumed, true otherwise. |
1411 | |
1412 | The value of the UCN, whether valid or invalid, is returned in *CP. |
1413 | Diagnostics are emitted for invalid values. PSTR is updated to point |
1414 | one beyond the UCN, or to the syntactically invalid character. |
1415 | |
1416 | IDENTIFIER_POS is 0 when not in an identifier, 1 for the start of |
1417 | an identifier, or 2 otherwise. |
1418 | |
1419 | If LOC_READER is non-NULL, then position information is |
1420 | read from *LOC_READER and CHAR_RANGE->m_finish is updated accordingly. */ |
1421 | |
1422 | bool |
1423 | _cpp_valid_ucn (cpp_reader *pfile, const uchar **pstr, |
1424 | const uchar *limit, int identifier_pos, |
1425 | struct normalize_state *nst, cppchar_t *cp, |
1426 | source_range *char_range, |
1427 | cpp_string_location_reader *loc_reader) |
1428 | { |
1429 | cppchar_t result, c; |
1430 | unsigned int length; |
1431 | const uchar *str = *pstr; |
1432 | const uchar *base = str - 2; |
1433 | bool delimited = false, named = false; |
1434 | |
1435 | if (!CPP_OPTION (pfile, cplusplus) && !CPP_OPTION (pfile, c99)) |
1436 | cpp_error (pfile, CPP_DL_WARNING, |
1437 | msgid: "universal character names are only valid in C++ and C99" ); |
1438 | else if (CPP_OPTION (pfile, cpp_warn_c90_c99_compat) > 0 |
1439 | && !CPP_OPTION (pfile, cplusplus)) |
1440 | cpp_error (pfile, CPP_DL_WARNING, |
1441 | msgid: "C99's universal character names are incompatible with C90" ); |
1442 | else if (CPP_WTRADITIONAL (pfile) && identifier_pos == 0) |
1443 | cpp_warning (pfile, CPP_W_TRADITIONAL, |
1444 | msgid: "the meaning of '\\%c' is different in traditional C" , |
1445 | (int) str[-1]); |
1446 | |
1447 | result = 0; |
1448 | if (str[-1] == 'u') |
1449 | { |
1450 | length = 4; |
1451 | if (str < limit |
1452 | && *str == '{' |
1453 | && (!identifier_pos |
1454 | || CPP_OPTION (pfile, delimited_escape_seqs) |
1455 | || !CPP_OPTION (pfile, std))) |
1456 | { |
1457 | str++; |
1458 | /* Magic value to indicate no digits seen. */ |
1459 | length = 32; |
1460 | delimited = true; |
1461 | extend_char_range (char_range, loc_reader); |
1462 | } |
1463 | } |
1464 | else if (str[-1] == 'U') |
1465 | length = 8; |
1466 | else if (str[-1] == 'N') |
1467 | { |
1468 | length = 4; |
1469 | if (identifier_pos |
1470 | && !CPP_OPTION (pfile, delimited_escape_seqs) |
1471 | && CPP_OPTION (pfile, std)) |
1472 | { |
1473 | *cp = 0; |
1474 | return false; |
1475 | } |
1476 | if (str == limit || *str != '{') |
1477 | { |
1478 | if (identifier_pos) |
1479 | { |
1480 | *cp = 0; |
1481 | return false; |
1482 | } |
1483 | cpp_error (pfile, CPP_DL_ERROR, msgid: "'\\N' not followed by '{'" ); |
1484 | } |
1485 | else |
1486 | { |
1487 | str++; |
1488 | named = true; |
1489 | extend_char_range (char_range, loc_reader); |
1490 | length = 0; |
1491 | const uchar *name = str; |
1492 | bool strict = true; |
1493 | |
1494 | do |
1495 | { |
1496 | if (str == limit) |
1497 | break; |
1498 | c = *str; |
1499 | if (!ISIDNUM (c) && c != ' ' && c != '-') |
1500 | break; |
1501 | if (ISLOWER (c) || c == '_') |
1502 | strict = false; |
1503 | str++; |
1504 | extend_char_range (char_range, loc_reader); |
1505 | } |
1506 | while (1); |
1507 | |
1508 | if (str < limit && *str == '}') |
1509 | { |
1510 | if (identifier_pos && name == str) |
1511 | { |
1512 | cpp_warning (pfile, CPP_W_UNICODE, |
1513 | msgid: "empty named universal character escape " |
1514 | "sequence; treating it as separate tokens" ); |
1515 | *cp = 0; |
1516 | return false; |
1517 | } |
1518 | if (name == str) |
1519 | cpp_error (pfile, CPP_DL_ERROR, |
1520 | msgid: "empty named universal character escape sequence" ); |
1521 | else if ((!identifier_pos || strict) |
1522 | && !CPP_OPTION (pfile, delimited_escape_seqs) |
1523 | && CPP_OPTION (pfile, cpp_pedantic)) |
1524 | cpp_error (pfile, CPP_DL_PEDWARN, |
1525 | msgid: "named universal character escapes are only valid " |
1526 | "in C++23" ); |
1527 | if (name == str) |
1528 | result = 0x40; |
1529 | else |
1530 | { |
1531 | /* If the name is longer than maximum length of a Unicode |
1532 | name, it can't be strictly valid. */ |
1533 | if ((size_t) (str - name) > uname2c_max_name_len || !strict) |
1534 | result = -1; |
1535 | else |
1536 | result = _cpp_uname2c (name: (const char *) name, len: str - name, |
1537 | n: uname2c_tree, NULL); |
1538 | if (result == (cppchar_t) -1) |
1539 | { |
1540 | bool ret = true; |
1541 | if (identifier_pos |
1542 | && (!CPP_OPTION (pfile, delimited_escape_seqs) |
1543 | || !strict)) |
1544 | ret = cpp_warning (pfile, CPP_W_UNICODE, |
1545 | msgid: "\\N{%.*s} is not a valid " |
1546 | "universal character; treating it " |
1547 | "as separate tokens" , |
1548 | (int) (str - name), name); |
1549 | else |
1550 | cpp_error (pfile, CPP_DL_ERROR, |
1551 | msgid: "\\N{%.*s} is not a valid universal " |
1552 | "character" , (int) (str - name), name); |
1553 | |
1554 | /* Try to do a loose name lookup according to |
1555 | Unicode loose matching rule UAX44-LM2. */ |
1556 | char canon_name[uname2c_max_name_len + 1]; |
1557 | result = _cpp_uname2c_uax44_lm2 (name: (const char *) name, |
1558 | len: str - name, canon_name); |
1559 | if (result != (cppchar_t) -1 && ret) |
1560 | cpp_error (pfile, CPP_DL_NOTE, |
1561 | msgid: "did you mean \\N{%s}?" , canon_name); |
1562 | else |
1563 | result = 0xC0; |
1564 | if (identifier_pos |
1565 | && (!CPP_OPTION (pfile, delimited_escape_seqs) |
1566 | || !strict)) |
1567 | { |
1568 | *cp = 0; |
1569 | return false; |
1570 | } |
1571 | } |
1572 | } |
1573 | str++; |
1574 | extend_char_range (char_range, loc_reader); |
1575 | } |
1576 | else if (identifier_pos) |
1577 | { |
1578 | cpp_warning (pfile, CPP_W_UNICODE, |
1579 | msgid: "'\\N{' not terminated with '}' after %.*s; " |
1580 | "treating it as separate tokens" , |
1581 | (int) (str - base), base); |
1582 | *cp = 0; |
1583 | return false; |
1584 | } |
1585 | else |
1586 | { |
1587 | cpp_error (pfile, CPP_DL_ERROR, |
1588 | msgid: "'\\N{' not terminated with '}' after %.*s" , |
1589 | (int) (str - base), base); |
1590 | result = 1; |
1591 | } |
1592 | } |
1593 | } |
1594 | else |
1595 | { |
1596 | cpp_error (pfile, CPP_DL_ICE, msgid: "In _cpp_valid_ucn but not a UCN" ); |
1597 | length = 4; |
1598 | } |
1599 | |
1600 | if (!named) |
1601 | do |
1602 | { |
1603 | if (str == limit) |
1604 | break; |
1605 | c = *str; |
1606 | if (!ISXDIGIT (c)) |
1607 | break; |
1608 | str++; |
1609 | extend_char_range (char_range, loc_reader); |
1610 | if (delimited) |
1611 | { |
1612 | if (!result) |
1613 | /* Accept arbitrary number of leading zeros. |
1614 | 16 is another magic value, smaller than 32 above |
1615 | and bigger than 8, so that upon encountering first |
1616 | non-zero digit we can count 8 digits and after that |
1617 | or in overflow bit and ensure length doesn't decrease |
1618 | to 0, as delimited escape sequence doesn't have upper |
1619 | bound on the number of hex digits. */ |
1620 | length = 16; |
1621 | else if (length == 16 - 8) |
1622 | { |
1623 | /* Make sure we detect overflows. */ |
1624 | result |= 0x8000000; |
1625 | ++length; |
1626 | } |
1627 | } |
1628 | |
1629 | result = (result << 4) + hex_value (c); |
1630 | } |
1631 | while (--length); |
1632 | |
1633 | if (delimited && str < limit && *str == '}') |
1634 | { |
1635 | if (length == 32 && identifier_pos) |
1636 | { |
1637 | cpp_warning (pfile, CPP_W_UNICODE, |
1638 | msgid: "empty delimited escape sequence; " |
1639 | "treating it as separate tokens" ); |
1640 | *cp = 0; |
1641 | return false; |
1642 | } |
1643 | else if (length == 32) |
1644 | cpp_error (pfile, CPP_DL_ERROR, |
1645 | msgid: "empty delimited escape sequence" ); |
1646 | else if (!CPP_OPTION (pfile, delimited_escape_seqs) |
1647 | && CPP_OPTION (pfile, cpp_pedantic)) |
1648 | cpp_error (pfile, CPP_DL_PEDWARN, |
1649 | msgid: "delimited escape sequences are only valid in C++23" ); |
1650 | str++; |
1651 | length = 0; |
1652 | delimited = false; |
1653 | extend_char_range (char_range, loc_reader); |
1654 | } |
1655 | |
1656 | /* Partial UCNs are not valid in strings, but decompose into |
1657 | multiple tokens in identifiers, so we can't give a helpful |
1658 | error message in that case. */ |
1659 | if (length && identifier_pos) |
1660 | { |
1661 | if (delimited) |
1662 | cpp_warning (pfile, CPP_W_UNICODE, |
1663 | msgid: "'\\u{' not terminated with '}' after %.*s; " |
1664 | "treating it as separate tokens" , |
1665 | (int) (str - base), base); |
1666 | *cp = 0; |
1667 | return false; |
1668 | } |
1669 | |
1670 | *pstr = str; |
1671 | if (length) |
1672 | { |
1673 | if (!delimited) |
1674 | cpp_error (pfile, CPP_DL_ERROR, |
1675 | msgid: "incomplete universal character name %.*s" , |
1676 | (int) (str - base), base); |
1677 | else |
1678 | cpp_error (pfile, CPP_DL_ERROR, |
1679 | msgid: "'\\u{' not terminated with '}' after %.*s" , |
1680 | (int) (str - base), base); |
1681 | result = 1; |
1682 | } |
1683 | /* The C99 standard permits $, @ and ` to be specified as UCNs. We use |
1684 | hex escapes so that this also works with EBCDIC hosts. |
1685 | C++0x permits everything below 0xa0 within literals; |
1686 | ucn_valid_in_identifier will complain about identifiers. */ |
1687 | else if ((result < 0xa0 |
1688 | && !CPP_OPTION (pfile, cplusplus) |
1689 | && (result != 0x24 && result != 0x40 && result != 0x60)) |
1690 | || (result & 0x80000000) |
1691 | || (result >= 0xD800 && result <= 0xDFFF)) |
1692 | { |
1693 | cpp_error (pfile, CPP_DL_ERROR, |
1694 | msgid: "%.*s is not a valid universal character" , |
1695 | (int) (str - base), base); |
1696 | result = 1; |
1697 | } |
1698 | else if (identifier_pos && result == 0x24 |
1699 | && CPP_OPTION (pfile, dollars_in_ident)) |
1700 | { |
1701 | if (CPP_OPTION (pfile, warn_dollars) && !pfile->state.skipping) |
1702 | { |
1703 | CPP_OPTION (pfile, warn_dollars) = 0; |
1704 | cpp_error (pfile, CPP_DL_PEDWARN, msgid: "'$' in identifier or number" ); |
1705 | } |
1706 | NORMALIZE_STATE_UPDATE_IDNUM (nst, result); |
1707 | } |
1708 | else if (identifier_pos) |
1709 | { |
1710 | int validity = ucn_valid_in_identifier (pfile, c: result, nst); |
1711 | |
1712 | if (validity == 0) |
1713 | cpp_error (pfile, CPP_DL_ERROR, |
1714 | msgid: "universal character %.*s is not valid in an identifier" , |
1715 | (int) (str - base), base); |
1716 | else if (validity == 2 && identifier_pos == 1) |
1717 | cpp_error (pfile, CPP_DL_ERROR, |
1718 | msgid: "universal character %.*s is not valid at the start of an identifier" , |
1719 | (int) (str - base), base); |
1720 | } |
1721 | else if (result > UCS_LIMIT |
1722 | && (!CPP_OPTION (pfile, cplusplus) |
1723 | || CPP_OPTION (pfile, lang) > CLK_CXX17)) |
1724 | cpp_error (pfile, CPP_DL_PEDWARN, |
1725 | msgid: "%.*s is outside the UCS codespace" , |
1726 | (int) (str - base), base); |
1727 | |
1728 | *cp = result; |
1729 | return true; |
1730 | } |
1731 | |
1732 | /* Convert an UCN, pointed to by FROM, to UTF-8 encoding, then translate |
1733 | it to the execution character set and write the result into TBUF, |
1734 | if TBUF is non-NULL. |
1735 | An advanced pointer is returned. Issues all relevant diagnostics. |
1736 | If LOC_READER is non-NULL, then RANGES must be non-NULL and CHAR_RANGE |
1737 | contains the location of the character so far: location information |
1738 | is read from *LOC_READER, and *RANGES is updated accordingly. */ |
1739 | static const uchar * |
1740 | convert_ucn (cpp_reader *pfile, const uchar *from, const uchar *limit, |
1741 | struct _cpp_strbuf *tbuf, struct cset_converter cvt, |
1742 | source_range char_range, |
1743 | cpp_string_location_reader *loc_reader, |
1744 | cpp_substring_ranges *ranges) |
1745 | { |
1746 | cppchar_t ucn; |
1747 | uchar buf[6]; |
1748 | uchar *bufp = buf; |
1749 | size_t bytesleft = 6; |
1750 | int rval; |
1751 | struct normalize_state nst = INITIAL_NORMALIZE_STATE; |
1752 | |
1753 | /* loc_reader and ranges must either be both NULL, or both be non-NULL. */ |
1754 | gcc_assert ((loc_reader != NULL) == (ranges != NULL)); |
1755 | |
1756 | from++; /* Skip u/U/N. */ |
1757 | |
1758 | /* The u/U is part of the spelling of this character. */ |
1759 | extend_char_range (char_range: &char_range, loc_reader); |
1760 | |
1761 | _cpp_valid_ucn (pfile, pstr: &from, limit, identifier_pos: 0, nst: &nst, |
1762 | cp: &ucn, char_range: &char_range, loc_reader); |
1763 | |
1764 | rval = one_cppchar_to_utf8 (c: ucn, outbufp: &bufp, outbytesleftp: &bytesleft); |
1765 | if (rval) |
1766 | { |
1767 | errno = rval; |
1768 | cpp_errno (pfile, CPP_DL_ERROR, |
1769 | msgid: "converting UCN to source character set" ); |
1770 | } |
1771 | else |
1772 | { |
1773 | if (tbuf) |
1774 | if (!APPLY_CONVERSION (cvt, buf, 6 - bytesleft, tbuf)) |
1775 | cpp_errno (pfile, CPP_DL_ERROR, |
1776 | msgid: "converting UCN to execution character set" ); |
1777 | |
1778 | if (loc_reader) |
1779 | { |
1780 | int num_encoded_bytes = 6 - bytesleft; |
1781 | for (int i = 0; i < num_encoded_bytes; i++) |
1782 | ranges->add_range (range: char_range); |
1783 | } |
1784 | } |
1785 | |
1786 | return from; |
1787 | } |
1788 | |
1789 | /* Performs a similar task as _cpp_valid_ucn, but parses UTF-8-encoded |
1790 | extended characters rather than UCNs. If the return value is TRUE, then a |
1791 | character was successfully decoded and stored in *CP; *PSTR has been |
1792 | updated to point one past the valid UTF-8 sequence. Diagnostics may have |
1793 | been emitted if the character parsed is not allowed in the current context. |
1794 | If the return value is FALSE, then *PSTR has not been modified and *CP may |
1795 | equal 0, to indicate that *PSTR does not form a valid UTF-8 sequence, or it |
1796 | may, when processing an identifier in C mode, equal a codepoint that was |
1797 | validly encoded but is not allowed to appear in an identifier. In either |
1798 | case, no diagnostic is emitted, and the return value of FALSE should cause |
1799 | a new token to be formed. |
1800 | |
1801 | _cpp_valid_utf8 can be called when lexing a potential identifier, or a |
1802 | CPP_OTHER token or for the purposes of -Winvalid-utf8 warning in string or |
1803 | character literals. NST is unused when not in a potential identifier. |
1804 | |
1805 | As in _cpp_valid_ucn, IDENTIFIER_POS is 0 when not in an identifier, 1 for |
1806 | the start of an identifier, or 2 otherwise. */ |
1807 | |
1808 | extern bool |
1809 | _cpp_valid_utf8 (cpp_reader *pfile, |
1810 | const uchar **pstr, |
1811 | const uchar *limit, |
1812 | int identifier_pos, |
1813 | struct normalize_state *nst, |
1814 | cppchar_t *cp) |
1815 | { |
1816 | const uchar *base = *pstr; |
1817 | size_t inbytesleft = limit - base; |
1818 | if (one_utf8_to_cppchar (inbufp: pstr, inbytesleftp: &inbytesleft, cp)) |
1819 | { |
1820 | /* No diagnostic here as this byte will rather become a |
1821 | new token. */ |
1822 | *cp = 0; |
1823 | return false; |
1824 | } |
1825 | |
1826 | if (identifier_pos) |
1827 | { |
1828 | switch (ucn_valid_in_identifier (pfile, c: *cp, nst)) |
1829 | { |
1830 | |
1831 | case 0: |
1832 | /* In C++, this is an error for invalid character in an identifier |
1833 | because logically, the UTF-8 was converted to a UCN during |
1834 | translation phase 1 (even though we don't physically do it that |
1835 | way). In C, this byte rather becomes grammatically a separate |
1836 | token. */ |
1837 | |
1838 | if (CPP_OPTION (pfile, cplusplus)) |
1839 | cpp_error (pfile, CPP_DL_ERROR, |
1840 | msgid: "extended character %.*s is not valid in an identifier" , |
1841 | (int) (*pstr - base), base); |
1842 | else |
1843 | { |
1844 | *pstr = base; |
1845 | return false; |
1846 | } |
1847 | |
1848 | break; |
1849 | |
1850 | case 2: |
1851 | if (identifier_pos == 1) |
1852 | { |
1853 | /* This is treated the same way in C++ or C99 -- lexed as an |
1854 | identifier which is then invalid because an identifier is |
1855 | not allowed to start with this character. */ |
1856 | cpp_error (pfile, CPP_DL_ERROR, |
1857 | msgid: "extended character %.*s is not valid at the start of an identifier" , |
1858 | (int) (*pstr - base), base); |
1859 | } |
1860 | break; |
1861 | } |
1862 | } |
1863 | |
1864 | return true; |
1865 | } |
1866 | |
1867 | /* Return true iff BUFFER of size NUM_BYTES is validly-encoded UTF-8. */ |
1868 | |
1869 | extern bool |
1870 | cpp_valid_utf8_p (const char *buffer, size_t num_bytes) |
1871 | { |
1872 | const uchar *iter = (const uchar *)buffer; |
1873 | size_t bytesleft = num_bytes; |
1874 | while (bytesleft > 0) |
1875 | { |
1876 | /* one_utf8_to_cppchar implements 5-byte and 6 byte sequences as per |
1877 | RFC 2279, but this has been superceded by RFC 3629, which |
1878 | restricts UTF-8 to 1-byte through 4-byte sequences, and |
1879 | states "the octet values C0, C1, F5 to FF never appear". |
1880 | |
1881 | Reject such values. */ |
1882 | if (*iter >= 0xf4) |
1883 | return false; |
1884 | |
1885 | cppchar_t cp; |
1886 | int err = one_utf8_to_cppchar (inbufp: &iter, inbytesleftp: &bytesleft, cp: &cp); |
1887 | if (err) |
1888 | return false; |
1889 | |
1890 | /* Additionally, Unicode declares that all codepoints above 0010FFFF are |
1891 | invalid because they cannot be represented in UTF-16. |
1892 | |
1893 | Reject such values.*/ |
1894 | if (cp > UCS_LIMIT) |
1895 | return false; |
1896 | } |
1897 | /* No problems encountered. */ |
1898 | return true; |
1899 | } |
1900 | |
1901 | /* Subroutine of convert_hex and convert_oct. N is the representation |
1902 | in the execution character set of a numeric escape; write it into the |
1903 | string buffer TBUF and update the end-of-string pointer therein. WIDE |
1904 | is true if it's a wide string that's being assembled in TBUF. This |
1905 | function issues no diagnostics and never fails. */ |
1906 | static void |
1907 | emit_numeric_escape (cpp_reader *pfile, cppchar_t n, |
1908 | struct _cpp_strbuf *tbuf, struct cset_converter cvt) |
1909 | { |
1910 | size_t width = cvt.width; |
1911 | |
1912 | if (width != CPP_OPTION (pfile, char_precision)) |
1913 | { |
1914 | /* We have to render this into the target byte order, which may not |
1915 | be our byte order. */ |
1916 | bool bigend = CPP_OPTION (pfile, bytes_big_endian); |
1917 | size_t cwidth = CPP_OPTION (pfile, char_precision); |
1918 | size_t cmask = width_to_mask (width: cwidth); |
1919 | size_t nbwc = width / cwidth; |
1920 | size_t i; |
1921 | size_t off = tbuf->len; |
1922 | cppchar_t c; |
1923 | |
1924 | if (tbuf->len + nbwc > tbuf->asize) |
1925 | { |
1926 | tbuf->asize += OUTBUF_BLOCK_SIZE; |
1927 | tbuf->text = XRESIZEVEC (uchar, tbuf->text, tbuf->asize); |
1928 | } |
1929 | |
1930 | for (i = 0; i < nbwc; i++) |
1931 | { |
1932 | c = n & cmask; |
1933 | n >>= cwidth; |
1934 | tbuf->text[off + (bigend ? nbwc - i - 1 : i)] = c; |
1935 | } |
1936 | tbuf->len += nbwc; |
1937 | } |
1938 | else |
1939 | { |
1940 | /* Note: this code does not handle the case where the target |
1941 | and host have a different number of bits in a byte. */ |
1942 | if (tbuf->len + 1 > tbuf->asize) |
1943 | { |
1944 | tbuf->asize += OUTBUF_BLOCK_SIZE; |
1945 | tbuf->text = XRESIZEVEC (uchar, tbuf->text, tbuf->asize); |
1946 | } |
1947 | tbuf->text[tbuf->len++] = n; |
1948 | } |
1949 | } |
1950 | |
1951 | /* Convert a hexadecimal escape, pointed to by FROM, to the execution |
1952 | character set and write it into the string buffer TBUF (if non-NULL). |
1953 | Returns an advanced pointer, and issues diagnostics as necessary. |
1954 | No character set translation occurs; this routine always produces the |
1955 | execution-set character with numeric value equal to the given hex |
1956 | number. You can, e.g. generate surrogate pairs this way. |
1957 | If LOC_READER is non-NULL, then RANGES must be non-NULL and CHAR_RANGE |
1958 | contains the location of the character so far: location information |
1959 | is read from *LOC_READER, and *RANGES is updated accordingly. */ |
1960 | static const uchar * |
1961 | convert_hex (cpp_reader *pfile, const uchar *from, const uchar *limit, |
1962 | struct _cpp_strbuf *tbuf, struct cset_converter cvt, |
1963 | source_range char_range, |
1964 | cpp_string_location_reader *loc_reader, |
1965 | cpp_substring_ranges *ranges) |
1966 | { |
1967 | cppchar_t c, n = 0, overflow = 0; |
1968 | int digits_found = 0; |
1969 | size_t width = cvt.width; |
1970 | size_t mask = width_to_mask (width); |
1971 | bool delimited = false; |
1972 | const uchar *base = from - 1; |
1973 | |
1974 | /* loc_reader and ranges must either be both NULL, or both be non-NULL. */ |
1975 | gcc_assert ((loc_reader != NULL) == (ranges != NULL)); |
1976 | |
1977 | if (CPP_WTRADITIONAL (pfile)) |
1978 | cpp_warning (pfile, CPP_W_TRADITIONAL, |
1979 | msgid: "the meaning of '\\x' is different in traditional C" ); |
1980 | |
1981 | /* Skip 'x'. */ |
1982 | from++; |
1983 | |
1984 | /* The 'x' is part of the spelling of this character. */ |
1985 | extend_char_range (char_range: &char_range, loc_reader); |
1986 | |
1987 | if (from < limit && *from == '{') |
1988 | { |
1989 | delimited = true; |
1990 | from++; |
1991 | extend_char_range (char_range: &char_range, loc_reader); |
1992 | } |
1993 | |
1994 | while (from < limit) |
1995 | { |
1996 | c = *from; |
1997 | if (! hex_p (c)) |
1998 | break; |
1999 | from++; |
2000 | extend_char_range (char_range: &char_range, loc_reader); |
2001 | overflow |= n ^ (n << 4 >> 4); |
2002 | n = (n << 4) + hex_value (c); |
2003 | digits_found = 1; |
2004 | } |
2005 | |
2006 | if (delimited && from < limit && *from == '}') |
2007 | { |
2008 | from++; |
2009 | if (!digits_found) |
2010 | { |
2011 | cpp_error (pfile, CPP_DL_ERROR, |
2012 | msgid: "empty delimited escape sequence" ); |
2013 | return from; |
2014 | } |
2015 | else if (!CPP_OPTION (pfile, delimited_escape_seqs) |
2016 | && CPP_OPTION (pfile, cpp_pedantic)) |
2017 | cpp_error (pfile, CPP_DL_PEDWARN, |
2018 | msgid: "delimited escape sequences are only valid in C++23" ); |
2019 | delimited = false; |
2020 | extend_char_range (char_range: &char_range, loc_reader); |
2021 | } |
2022 | |
2023 | if (!digits_found) |
2024 | { |
2025 | cpp_error (pfile, CPP_DL_ERROR, |
2026 | msgid: "\\x used with no following hex digits" ); |
2027 | return from; |
2028 | } |
2029 | else if (delimited) |
2030 | { |
2031 | cpp_error (pfile, CPP_DL_ERROR, |
2032 | msgid: "'\\x{' not terminated with '}' after %.*s" , |
2033 | (int) (from - base), base); |
2034 | return from; |
2035 | } |
2036 | |
2037 | if (overflow | (n != (n & mask))) |
2038 | { |
2039 | cpp_error (pfile, CPP_DL_PEDWARN, |
2040 | msgid: "hex escape sequence out of range" ); |
2041 | n &= mask; |
2042 | } |
2043 | |
2044 | if (tbuf) |
2045 | emit_numeric_escape (pfile, n, tbuf, cvt); |
2046 | if (ranges) |
2047 | ranges->add_range (range: char_range); |
2048 | |
2049 | return from; |
2050 | } |
2051 | |
2052 | /* Convert an octal escape, pointed to by FROM, to the execution |
2053 | character set and write it into the string buffer TBUF. Returns an |
2054 | advanced pointer, and issues diagnostics as necessary. |
2055 | No character set translation occurs; this routine always produces the |
2056 | execution-set character with numeric value equal to the given octal |
2057 | number. |
2058 | If LOC_READER is non-NULL, then RANGES must be non-NULL and CHAR_RANGE |
2059 | contains the location of the character so far: location information |
2060 | is read from *LOC_READER, and *RANGES is updated accordingly. */ |
2061 | static const uchar * |
2062 | convert_oct (cpp_reader *pfile, const uchar *from, const uchar *limit, |
2063 | struct _cpp_strbuf *tbuf, struct cset_converter cvt, |
2064 | source_range char_range, |
2065 | cpp_string_location_reader *loc_reader, |
2066 | cpp_substring_ranges *ranges) |
2067 | { |
2068 | size_t count = 0; |
2069 | cppchar_t c, n = 0, overflow = 0; |
2070 | size_t width = cvt.width; |
2071 | size_t mask = width_to_mask (width); |
2072 | bool delimited = false; |
2073 | const uchar *base = from - 1; |
2074 | |
2075 | /* loc_reader and ranges must either be both NULL, or both be non-NULL. */ |
2076 | gcc_assert ((loc_reader != NULL) == (ranges != NULL)); |
2077 | |
2078 | if (from < limit && *from == 'o') |
2079 | { |
2080 | from++; |
2081 | extend_char_range (char_range: &char_range, loc_reader); |
2082 | if (from == limit || *from != '{') |
2083 | cpp_error (pfile, CPP_DL_ERROR, msgid: "'\\o' not followed by '{'" ); |
2084 | else |
2085 | { |
2086 | from++; |
2087 | extend_char_range (char_range: &char_range, loc_reader); |
2088 | delimited = true; |
2089 | } |
2090 | } |
2091 | |
2092 | while (from < limit && count++ < 3) |
2093 | { |
2094 | c = *from; |
2095 | if (c < '0' || c > '7') |
2096 | break; |
2097 | from++; |
2098 | extend_char_range (char_range: &char_range, loc_reader); |
2099 | if (delimited) |
2100 | { |
2101 | count = 2; |
2102 | overflow |= n ^ (n << 3 >> 3); |
2103 | } |
2104 | n = (n << 3) + c - '0'; |
2105 | } |
2106 | |
2107 | if (delimited) |
2108 | { |
2109 | if (from < limit && *from == '}') |
2110 | { |
2111 | from++; |
2112 | if (count == 1) |
2113 | { |
2114 | cpp_error (pfile, CPP_DL_ERROR, |
2115 | msgid: "empty delimited escape sequence" ); |
2116 | return from; |
2117 | } |
2118 | else if (!CPP_OPTION (pfile, delimited_escape_seqs) |
2119 | && CPP_OPTION (pfile, cpp_pedantic)) |
2120 | cpp_error (pfile, CPP_DL_PEDWARN, |
2121 | msgid: "delimited escape sequences are only valid in C++23" ); |
2122 | extend_char_range (char_range: &char_range, loc_reader); |
2123 | } |
2124 | else |
2125 | { |
2126 | cpp_error (pfile, CPP_DL_ERROR, |
2127 | msgid: "'\\o{' not terminated with '}' after %.*s" , |
2128 | (int) (from - base), base); |
2129 | return from; |
2130 | } |
2131 | } |
2132 | |
2133 | if (overflow | (n != (n & mask))) |
2134 | { |
2135 | cpp_error (pfile, CPP_DL_PEDWARN, |
2136 | msgid: "octal escape sequence out of range" ); |
2137 | n &= mask; |
2138 | } |
2139 | |
2140 | if (tbuf) |
2141 | emit_numeric_escape (pfile, n, tbuf, cvt); |
2142 | if (ranges) |
2143 | ranges->add_range (range: char_range); |
2144 | |
2145 | return from; |
2146 | } |
2147 | |
2148 | /* Convert an escape sequence (pointed to by FROM) to its value on |
2149 | the target, and to the execution character set. Do not scan past |
2150 | LIMIT. Write the converted value into TBUF, if TBUF is non-NULL. |
2151 | Returns an advanced pointer. Handles all relevant diagnostics. |
2152 | If LOC_READER is non-NULL, then RANGES must be non-NULL: location |
2153 | information is read from *LOC_READER, and *RANGES is updated |
2154 | accordingly. */ |
2155 | static const uchar * |
2156 | convert_escape (cpp_reader *pfile, const uchar *from, const uchar *limit, |
2157 | struct _cpp_strbuf *tbuf, struct cset_converter cvt, |
2158 | cpp_string_location_reader *loc_reader, |
2159 | cpp_substring_ranges *ranges, bool uneval) |
2160 | { |
2161 | /* Values of \a \b \e \f \n \r \t \v respectively. */ |
2162 | #if HOST_CHARSET == HOST_CHARSET_ASCII |
2163 | static const uchar charconsts[] = { 7, 8, 27, 12, 10, 13, 9, 11 }; |
2164 | #elif HOST_CHARSET == HOST_CHARSET_EBCDIC |
2165 | static const uchar charconsts[] = { 47, 22, 39, 12, 21, 13, 5, 11 }; |
2166 | #else |
2167 | #error "unknown host character set" |
2168 | #endif |
2169 | |
2170 | uchar c; |
2171 | |
2172 | /* Record the location of the backslash. */ |
2173 | source_range char_range; |
2174 | if (loc_reader) |
2175 | char_range = loc_reader->get_next (); |
2176 | |
2177 | c = *from; |
2178 | switch (c) |
2179 | { |
2180 | /* UCNs, hex escapes, and octal escapes are processed separately. */ |
2181 | case 'u': case 'U': case 'N': |
2182 | return convert_ucn (pfile, from, limit, tbuf, cvt, |
2183 | char_range, loc_reader, ranges); |
2184 | |
2185 | case 'x': |
2186 | if (uneval && CPP_PEDANTIC (pfile)) |
2187 | cpp_error (pfile, CPP_DL_PEDWARN, |
2188 | msgid: "numeric escape sequence in unevaluated string: " |
2189 | "'\\%c'" , (int) c); |
2190 | return convert_hex (pfile, from, limit, tbuf, cvt, |
2191 | char_range, loc_reader, ranges); |
2192 | |
2193 | case '0': case '1': case '2': case '3': |
2194 | case '4': case '5': case '6': case '7': |
2195 | case 'o': |
2196 | if (uneval && CPP_PEDANTIC (pfile)) |
2197 | cpp_error (pfile, CPP_DL_PEDWARN, |
2198 | msgid: "numeric escape sequence in unevaluated string: " |
2199 | "'\\%c'" , (int) c); |
2200 | return convert_oct (pfile, from, limit, tbuf, cvt, |
2201 | char_range, loc_reader, ranges); |
2202 | |
2203 | /* Various letter escapes. Get the appropriate host-charset |
2204 | value into C. */ |
2205 | case '\\': case '\'': case '"': case '?': break; |
2206 | |
2207 | case '(': case '{': case '[': case '%': |
2208 | /* '\(', etc, can be used at the beginning of a line in a long |
2209 | string split onto multiple lines with \-newline, to prevent |
2210 | Emacs or other text editors from getting confused. '\%' can |
2211 | be used to prevent SCCS from mangling printf format strings. */ |
2212 | if (CPP_PEDANTIC (pfile)) |
2213 | goto unknown; |
2214 | break; |
2215 | |
2216 | case 'b': c = charconsts[1]; break; |
2217 | case 'f': c = charconsts[3]; break; |
2218 | case 'n': c = charconsts[4]; break; |
2219 | case 'r': c = charconsts[5]; break; |
2220 | case 't': c = charconsts[6]; break; |
2221 | case 'v': c = charconsts[7]; break; |
2222 | |
2223 | case 'a': |
2224 | if (CPP_WTRADITIONAL (pfile)) |
2225 | cpp_warning (pfile, CPP_W_TRADITIONAL, |
2226 | msgid: "the meaning of '\\a' is different in traditional C" ); |
2227 | c = charconsts[0]; |
2228 | break; |
2229 | |
2230 | case 'e': case 'E': |
2231 | if (CPP_PEDANTIC (pfile)) |
2232 | cpp_error (pfile, CPP_DL_PEDWARN, |
2233 | msgid: "non-ISO-standard escape sequence, '\\%c'" , (int) c); |
2234 | c = charconsts[2]; |
2235 | break; |
2236 | |
2237 | default: |
2238 | unknown: |
2239 | if (ISGRAPH (c)) |
2240 | cpp_error (pfile, CPP_DL_PEDWARN, |
2241 | msgid: "unknown escape sequence: '\\%c'" , (int) c); |
2242 | else |
2243 | { |
2244 | encoding_rich_location rich_loc (pfile); |
2245 | |
2246 | /* diagnostic.cc does not support "%03o". When it does, this |
2247 | code can use %03o directly in the diagnostic again. */ |
2248 | char buf[32]; |
2249 | sprintf(s: buf, format: "%03o" , (int) c); |
2250 | cpp_error_at (pfile, CPP_DL_PEDWARN, richloc: &rich_loc, |
2251 | msgid: "unknown escape sequence: '\\%s'" , buf); |
2252 | } |
2253 | } |
2254 | |
2255 | if (tbuf) |
2256 | /* Now convert what we have to the execution character set. */ |
2257 | if (!APPLY_CONVERSION (cvt, &c, 1, tbuf)) |
2258 | cpp_errno (pfile, CPP_DL_ERROR, |
2259 | msgid: "converting escape sequence to execution character set" ); |
2260 | |
2261 | if (loc_reader) |
2262 | { |
2263 | char_range.m_finish = loc_reader->get_next ().m_finish; |
2264 | ranges->add_range (range: char_range); |
2265 | } |
2266 | |
2267 | return from + 1; |
2268 | } |
2269 | |
2270 | /* TYPE is a token type. The return value is the conversion needed to |
2271 | convert from source to execution character set for the given type. */ |
2272 | static struct cset_converter |
2273 | converter_for_type (cpp_reader *pfile, enum cpp_ttype type) |
2274 | { |
2275 | switch (type) |
2276 | { |
2277 | default: |
2278 | return pfile->narrow_cset_desc; |
2279 | case CPP_UTF8CHAR: |
2280 | case CPP_UTF8STRING: |
2281 | return pfile->utf8_cset_desc; |
2282 | case CPP_CHAR16: |
2283 | case CPP_STRING16: |
2284 | return pfile->char16_cset_desc; |
2285 | case CPP_CHAR32: |
2286 | case CPP_STRING32: |
2287 | return pfile->char32_cset_desc; |
2288 | case CPP_WCHAR: |
2289 | case CPP_WSTRING: |
2290 | return pfile->wide_cset_desc; |
2291 | } |
2292 | } |
2293 | |
2294 | /* FROM is an array of cpp_string structures of length COUNT. These |
2295 | are to be converted from the source to the execution character set, |
2296 | escape sequences translated, and finally all are to be |
2297 | concatenated. WIDE indicates whether or not to produce a wide |
2298 | string. If TO is non-NULL, the result is written into TO. |
2299 | If LOC_READERS and OUT are non-NULL, then location information |
2300 | is read from LOC_READERS (which must be an array of length COUNT), |
2301 | and location information is written to *RANGES. |
2302 | |
2303 | Returns true for success, false for failure. */ |
2304 | |
2305 | static bool |
2306 | cpp_interpret_string_1 (cpp_reader *pfile, const cpp_string *from, size_t count, |
2307 | cpp_string *to, enum cpp_ttype type, |
2308 | cpp_string_location_reader *loc_readers, |
2309 | cpp_substring_ranges *out) |
2310 | { |
2311 | struct _cpp_strbuf tbuf; |
2312 | const uchar *p, *base, *limit; |
2313 | size_t i; |
2314 | struct cset_converter cvt = converter_for_type (pfile, type); |
2315 | |
2316 | /* loc_readers and out must either be both NULL, or both be non-NULL. */ |
2317 | gcc_assert ((loc_readers != NULL) == (out != NULL)); |
2318 | |
2319 | if (to) |
2320 | { |
2321 | tbuf.asize = MAX (OUTBUF_BLOCK_SIZE, from->len); |
2322 | tbuf.text = XNEWVEC (uchar, tbuf.asize); |
2323 | tbuf.len = 0; |
2324 | } |
2325 | |
2326 | cpp_string_location_reader *loc_reader = NULL; |
2327 | for (i = 0; i < count; i++) |
2328 | { |
2329 | if (loc_readers) |
2330 | loc_reader = &loc_readers[i]; |
2331 | |
2332 | p = from[i].text; |
2333 | if (*p == 'u') |
2334 | { |
2335 | p++; |
2336 | if (loc_reader) |
2337 | loc_reader->get_next (); |
2338 | if (*p == '8') |
2339 | { |
2340 | p++; |
2341 | if (loc_reader) |
2342 | loc_reader->get_next (); |
2343 | } |
2344 | } |
2345 | else if (*p == 'L' || *p == 'U') p++; |
2346 | if (*p == 'R') |
2347 | { |
2348 | const uchar *prefix; |
2349 | |
2350 | /* Skip over 'R"'. */ |
2351 | p += 2; |
2352 | if (loc_reader) |
2353 | { |
2354 | loc_reader->get_next (); |
2355 | loc_reader->get_next (); |
2356 | } |
2357 | prefix = p; |
2358 | while (*p != '(') |
2359 | { |
2360 | p++; |
2361 | if (loc_reader) |
2362 | loc_reader->get_next (); |
2363 | } |
2364 | p++; |
2365 | if (loc_reader) |
2366 | loc_reader->get_next (); |
2367 | limit = from[i].text + from[i].len; |
2368 | if (limit >= p + (p - prefix) + 1) |
2369 | limit -= (p - prefix) + 1; |
2370 | |
2371 | /* Raw strings are all normal characters; these can be fed |
2372 | directly to convert_cset. */ |
2373 | if (to) |
2374 | if (!APPLY_CONVERSION (cvt, p, limit - p, &tbuf)) |
2375 | goto fail; |
2376 | |
2377 | if (loc_reader) |
2378 | { |
2379 | /* If generating source ranges, assume we have a 1:1 |
2380 | correspondence between bytes in the source encoding and bytes |
2381 | in the execution encoding (e.g. if we have a UTF-8 to UTF-8 |
2382 | conversion), so that this run of bytes in the source file |
2383 | corresponds to a run of bytes in the execution string. |
2384 | This requirement is guaranteed by an early-reject in |
2385 | cpp_interpret_string_ranges. */ |
2386 | gcc_assert (cvt.func == convert_no_conversion); |
2387 | out->add_n_ranges (num: limit - p, loc_reader&: *loc_reader); |
2388 | } |
2389 | |
2390 | continue; |
2391 | } |
2392 | |
2393 | /* If we don't now have a leading quote, something has gone wrong. |
2394 | This can occur if cpp_interpret_string_ranges is handling a |
2395 | stringified macro argument, but should not be possible otherwise. */ |
2396 | if (*p != '"' && *p != '\'') |
2397 | { |
2398 | gcc_assert (out != NULL); |
2399 | cpp_error (pfile, CPP_DL_ERROR, msgid: "missing open quote" ); |
2400 | if (to) |
2401 | free (ptr: tbuf.text); |
2402 | return false; |
2403 | } |
2404 | |
2405 | /* Skip leading quote. */ |
2406 | p++; |
2407 | if (loc_reader) |
2408 | loc_reader->get_next (); |
2409 | |
2410 | limit = from[i].text + from[i].len - 1; /* Skip trailing quote. */ |
2411 | |
2412 | for (;;) |
2413 | { |
2414 | base = p; |
2415 | while (p < limit && *p != '\\') |
2416 | p++; |
2417 | if (p > base) |
2418 | { |
2419 | /* We have a run of normal characters; these can be fed |
2420 | directly to convert_cset. */ |
2421 | if (to) |
2422 | if (!APPLY_CONVERSION (cvt, base, p - base, &tbuf)) |
2423 | goto fail; |
2424 | /* Similar to above: assumes we have a 1:1 correspondence |
2425 | between bytes in the source encoding and bytes in the |
2426 | execution encoding. */ |
2427 | if (loc_reader) |
2428 | { |
2429 | gcc_assert (cvt.func == convert_no_conversion); |
2430 | out->add_n_ranges (num: p - base, loc_reader&: *loc_reader); |
2431 | } |
2432 | } |
2433 | if (p >= limit) |
2434 | break; |
2435 | |
2436 | struct _cpp_strbuf *tbuf_ptr = to ? &tbuf : NULL; |
2437 | p = convert_escape (pfile, from: p + 1, limit, tbuf: tbuf_ptr, cvt, |
2438 | loc_reader, ranges: out, uneval: type == CPP_UNEVAL_STRING); |
2439 | } |
2440 | } |
2441 | |
2442 | if (to) |
2443 | { |
2444 | /* NUL-terminate the 'to' buffer and translate it to a cpp_string |
2445 | structure. */ |
2446 | emit_numeric_escape (pfile, n: 0, tbuf: &tbuf, cvt); |
2447 | tbuf.text = XRESIZEVEC (uchar, tbuf.text, tbuf.len); |
2448 | to->text = tbuf.text; |
2449 | to->len = tbuf.len; |
2450 | } |
2451 | /* Use the location of the trailing quote as the location of the |
2452 | NUL-terminator. */ |
2453 | if (loc_reader) |
2454 | { |
2455 | source_range range = loc_reader->get_next (); |
2456 | out->add_range (range); |
2457 | } |
2458 | |
2459 | return true; |
2460 | |
2461 | fail: |
2462 | cpp_errno (pfile, CPP_DL_ERROR, msgid: "converting to execution character set" ); |
2463 | if (to) |
2464 | free (ptr: tbuf.text); |
2465 | return false; |
2466 | } |
2467 | |
2468 | /* FROM is an array of cpp_string structures of length COUNT. These |
2469 | are to be converted from the source to the execution character set, |
2470 | escape sequences translated, and finally all are to be |
2471 | concatenated. WIDE indicates whether or not to produce a wide |
2472 | string. The result is written into TO. Returns true for success, |
2473 | false for failure. */ |
2474 | bool |
2475 | cpp_interpret_string (cpp_reader *pfile, const cpp_string *from, size_t count, |
2476 | cpp_string *to, enum cpp_ttype type) |
2477 | { |
2478 | return cpp_interpret_string_1 (pfile, from, count, to, type, NULL, NULL); |
2479 | } |
2480 | |
2481 | /* A "do nothing" diagnostic-handling callback for use by |
2482 | cpp_interpret_string_ranges, so that it can temporarily suppress |
2483 | diagnostic-handling. */ |
2484 | |
2485 | static bool |
2486 | noop_diagnostic_cb (cpp_reader *, enum cpp_diagnostic_level, |
2487 | enum cpp_warning_reason, rich_location *, |
2488 | const char *, va_list *) |
2489 | { |
2490 | /* no-op. */ |
2491 | return true; |
2492 | } |
2493 | |
2494 | /* This function mimics the behavior of cpp_interpret_string, but |
2495 | rather than generating a string in the execution character set, |
2496 | *OUT is written to with the source code ranges of the characters |
2497 | in such a string. |
2498 | FROM and LOC_READERS should both be arrays of length COUNT. |
2499 | Returns NULL for success, or an error message for failure. */ |
2500 | |
2501 | const char * |
2502 | cpp_interpret_string_ranges (cpp_reader *pfile, const cpp_string *from, |
2503 | cpp_string_location_reader *loc_readers, |
2504 | size_t count, |
2505 | cpp_substring_ranges *out, |
2506 | enum cpp_ttype type) |
2507 | { |
2508 | /* There are a couple of cases in the range-handling in |
2509 | cpp_interpret_string_1 that rely on there being a 1:1 correspondence |
2510 | between bytes in the source encoding and bytes in the execution |
2511 | encoding, so that each byte in the execution string can correspond |
2512 | to the location of a byte in the source string. |
2513 | |
2514 | This holds for the typical case of a UTF-8 to UTF-8 conversion. |
2515 | Enforce this requirement by only attempting to track substring |
2516 | locations if we have source encoding == execution encoding. |
2517 | |
2518 | This is a stronger condition than we need, since we could e.g. |
2519 | have ASCII to EBCDIC (with 1 byte per character before and after), |
2520 | but it seems to be a reasonable restriction. */ |
2521 | struct cset_converter cvt = converter_for_type (pfile, type); |
2522 | if (cvt.func != convert_no_conversion) |
2523 | return "execution character set != source character set" ; |
2524 | |
2525 | /* For on-demand strings we have already lexed the strings, so there |
2526 | should be no diagnostics. However, if we have bogus source location |
2527 | data (or stringified macro arguments), the attempt to lex the |
2528 | strings could fail with an diagnostic. Temporarily install an |
2529 | diagnostic-handler to catch the diagnostic, so that it can lead to this call |
2530 | failing, rather than being emitted as a user-visible diagnostic. |
2531 | If an diagnostic does occur, we should see it via the return value of |
2532 | cpp_interpret_string_1. */ |
2533 | bool (*saved_diagnostic_handler) (cpp_reader *, enum cpp_diagnostic_level, |
2534 | enum cpp_warning_reason, rich_location *, |
2535 | const char *, va_list *) |
2536 | ATTRIBUTE_FPTR_PRINTF(5,0); |
2537 | |
2538 | saved_diagnostic_handler = pfile->cb.diagnostic; |
2539 | pfile->cb.diagnostic = noop_diagnostic_cb; |
2540 | |
2541 | bool result = cpp_interpret_string_1 (pfile, from, count, NULL, type, |
2542 | loc_readers, out); |
2543 | |
2544 | /* Restore the saved diagnostic-handler. */ |
2545 | pfile->cb.diagnostic = saved_diagnostic_handler; |
2546 | |
2547 | if (!result) |
2548 | return "cpp_interpret_string_1 failed" ; |
2549 | |
2550 | /* Success. */ |
2551 | return NULL; |
2552 | } |
2553 | |
2554 | /* Subroutine of do_line and do_linemarker. Convert escape sequences |
2555 | in a string, but do not perform character set conversion. */ |
2556 | bool |
2557 | cpp_interpret_string_notranslate (cpp_reader *pfile, const cpp_string *from, |
2558 | size_t count, cpp_string *to, |
2559 | enum cpp_ttype type) |
2560 | { |
2561 | struct cset_converter save_narrow_cset_desc = pfile->narrow_cset_desc; |
2562 | bool retval; |
2563 | |
2564 | pfile->narrow_cset_desc.func = convert_no_conversion; |
2565 | pfile->narrow_cset_desc.cd = (iconv_t) -1; |
2566 | pfile->narrow_cset_desc.width = CPP_OPTION (pfile, char_precision); |
2567 | |
2568 | retval = cpp_interpret_string (pfile, from, count, to, |
2569 | type: type == CPP_UNEVAL_STRING |
2570 | ? CPP_UNEVAL_STRING : CPP_STRING); |
2571 | |
2572 | pfile->narrow_cset_desc = save_narrow_cset_desc; |
2573 | return retval; |
2574 | } |
2575 | |
2576 | |
2577 | /* Subroutine of cpp_interpret_charconst which performs the conversion |
2578 | to a number, for narrow strings. STR is the string structure returned |
2579 | by cpp_interpret_string. PCHARS_SEEN and UNSIGNEDP are as for |
2580 | cpp_interpret_charconst. TYPE is the token type. */ |
2581 | static cppchar_t |
2582 | narrow_str_to_charconst (cpp_reader *pfile, cpp_string str, |
2583 | unsigned int *pchars_seen, int *unsignedp, |
2584 | enum cpp_ttype type) |
2585 | { |
2586 | size_t width = CPP_OPTION (pfile, char_precision); |
2587 | size_t max_chars = CPP_OPTION (pfile, int_precision) / width; |
2588 | size_t mask = width_to_mask (width); |
2589 | size_t i; |
2590 | cppchar_t result, c; |
2591 | bool unsigned_p; |
2592 | |
2593 | /* The value of a multi-character character constant, or a |
2594 | single-character character constant whose representation in the |
2595 | execution character set is more than one byte long, is |
2596 | implementation defined. This implementation defines it to be the |
2597 | number formed by interpreting the byte sequence in memory as a |
2598 | big-endian binary number. If overflow occurs, the high bytes are |
2599 | lost, and a warning is issued. |
2600 | |
2601 | We don't want to process the NUL terminator handed back by |
2602 | cpp_interpret_string. */ |
2603 | result = 0; |
2604 | for (i = 0; i < str.len - 1; i++) |
2605 | { |
2606 | c = str.text[i] & mask; |
2607 | if (width < BITS_PER_CPPCHAR_T) |
2608 | result = (result << width) | c; |
2609 | else |
2610 | result = c; |
2611 | } |
2612 | |
2613 | if (type == CPP_UTF8CHAR) |
2614 | max_chars = 1; |
2615 | if (i > max_chars) |
2616 | { |
2617 | i = max_chars; |
2618 | cpp_error (pfile, type == CPP_UTF8CHAR ? CPP_DL_ERROR : CPP_DL_WARNING, |
2619 | msgid: "character constant too long for its type" ); |
2620 | } |
2621 | else if (i > 1 && CPP_OPTION (pfile, warn_multichar)) |
2622 | cpp_warning (pfile, CPP_W_MULTICHAR, msgid: "multi-character character constant" ); |
2623 | |
2624 | /* Multichar constants are of type int and therefore signed. */ |
2625 | if (i > 1) |
2626 | unsigned_p = 0; |
2627 | else if (type == CPP_UTF8CHAR) |
2628 | unsigned_p = CPP_OPTION (pfile, unsigned_utf8char); |
2629 | else |
2630 | unsigned_p = CPP_OPTION (pfile, unsigned_char); |
2631 | |
2632 | /* Truncate the constant to its natural width, and simultaneously |
2633 | sign- or zero-extend to the full width of cppchar_t. |
2634 | For single-character constants, the value is WIDTH bits wide. |
2635 | For multi-character constants, the value is INT_PRECISION bits wide. */ |
2636 | if (i > 1) |
2637 | width = CPP_OPTION (pfile, int_precision); |
2638 | if (width < BITS_PER_CPPCHAR_T) |
2639 | { |
2640 | mask = ((cppchar_t) 1 << width) - 1; |
2641 | if (unsigned_p || !(result & (1 << (width - 1)))) |
2642 | result &= mask; |
2643 | else |
2644 | result |= ~mask; |
2645 | } |
2646 | *pchars_seen = i; |
2647 | *unsignedp = unsigned_p; |
2648 | return result; |
2649 | } |
2650 | |
2651 | /* Subroutine of cpp_interpret_charconst which performs the conversion |
2652 | to a number, for wide strings. STR is the string structure returned |
2653 | by cpp_interpret_string. PCHARS_SEEN and UNSIGNEDP are as for |
2654 | cpp_interpret_charconst. TYPE is the token type. */ |
2655 | static cppchar_t |
2656 | wide_str_to_charconst (cpp_reader *pfile, cpp_string str, |
2657 | unsigned int *pchars_seen, int *unsignedp, |
2658 | enum cpp_ttype type) |
2659 | { |
2660 | bool bigend = CPP_OPTION (pfile, bytes_big_endian); |
2661 | size_t width = converter_for_type (pfile, type).width; |
2662 | size_t cwidth = CPP_OPTION (pfile, char_precision); |
2663 | size_t mask = width_to_mask (width); |
2664 | size_t cmask = width_to_mask (width: cwidth); |
2665 | size_t nbwc = width / cwidth; |
2666 | size_t off, i; |
2667 | cppchar_t result = 0, c; |
2668 | |
2669 | if (str.len <= nbwc) |
2670 | { |
2671 | /* Error recovery, if no errors have been diagnosed previously, |
2672 | there should be at least two wide characters. Empty literals |
2673 | are diagnosed earlier and we can get just the zero terminator |
2674 | only if there were errors diagnosed during conversion. */ |
2675 | *pchars_seen = 0; |
2676 | *unsignedp = 0; |
2677 | return 0; |
2678 | } |
2679 | |
2680 | /* This is finicky because the string is in the target's byte order, |
2681 | which may not be our byte order. Only the last character, ignoring |
2682 | the NUL terminator, is relevant. */ |
2683 | off = str.len - (nbwc * 2); |
2684 | result = 0; |
2685 | for (i = 0; i < nbwc; i++) |
2686 | { |
2687 | c = bigend ? str.text[off + i] : str.text[off + nbwc - i - 1]; |
2688 | result = (result << cwidth) | (c & cmask); |
2689 | } |
2690 | |
2691 | /* Wide character constants have type wchar_t, and a single |
2692 | character exactly fills a wchar_t, so a multi-character wide |
2693 | character constant is guaranteed to overflow. */ |
2694 | if (str.len > nbwc * 2) |
2695 | cpp_error (pfile, (CPP_OPTION (pfile, cplusplus) |
2696 | && (type == CPP_CHAR16 |
2697 | || type == CPP_CHAR32 |
2698 | /* In C++23 this is error even for L'ab'. */ |
2699 | || (type == CPP_WCHAR |
2700 | && CPP_OPTION (pfile, size_t_literals)))) |
2701 | ? CPP_DL_ERROR : CPP_DL_WARNING, |
2702 | msgid: "character constant too long for its type" ); |
2703 | |
2704 | /* Truncate the constant to its natural width, and simultaneously |
2705 | sign- or zero-extend to the full width of cppchar_t. */ |
2706 | if (width < BITS_PER_CPPCHAR_T) |
2707 | { |
2708 | if (type == CPP_CHAR16 || type == CPP_CHAR32 |
2709 | || CPP_OPTION (pfile, unsigned_wchar) |
2710 | || !(result & (1 << (width - 1)))) |
2711 | result &= mask; |
2712 | else |
2713 | result |= ~mask; |
2714 | } |
2715 | |
2716 | if (type == CPP_CHAR16 || type == CPP_CHAR32 |
2717 | || CPP_OPTION (pfile, unsigned_wchar)) |
2718 | *unsignedp = 1; |
2719 | else |
2720 | *unsignedp = 0; |
2721 | |
2722 | *pchars_seen = 1; |
2723 | return result; |
2724 | } |
2725 | |
2726 | /* Interpret a (possibly wide) character constant in TOKEN. |
2727 | PCHARS_SEEN points to a variable that is filled in with the number |
2728 | of characters seen, and UNSIGNEDP to a variable that indicates |
2729 | whether the result has signed type. */ |
2730 | cppchar_t |
2731 | cpp_interpret_charconst (cpp_reader *pfile, const cpp_token *token, |
2732 | unsigned int *pchars_seen, int *unsignedp) |
2733 | { |
2734 | cpp_string str = { .len: 0, .text: 0 }; |
2735 | bool wide = (token->type != CPP_CHAR && token->type != CPP_UTF8CHAR); |
2736 | int u8 = 2 * int(token->type == CPP_UTF8CHAR); |
2737 | cppchar_t result; |
2738 | |
2739 | /* An empty constant will appear as L'', u'', U'', u8'', or '' */ |
2740 | if (token->val.str.len == (size_t) (2 + wide + u8)) |
2741 | { |
2742 | cpp_error (pfile, CPP_DL_ERROR, msgid: "empty character constant" ); |
2743 | *pchars_seen = 0; |
2744 | *unsignedp = 0; |
2745 | return 0; |
2746 | } |
2747 | else if (!cpp_interpret_string (pfile, from: &token->val.str, count: 1, to: &str, |
2748 | type: token->type)) |
2749 | { |
2750 | *pchars_seen = 0; |
2751 | *unsignedp = 0; |
2752 | return 0; |
2753 | } |
2754 | |
2755 | if (wide) |
2756 | result = wide_str_to_charconst (pfile, str, pchars_seen, unsignedp, |
2757 | type: token->type); |
2758 | else |
2759 | result = narrow_str_to_charconst (pfile, str, pchars_seen, unsignedp, |
2760 | type: token->type); |
2761 | |
2762 | if (str.text != token->val.str.text) |
2763 | free (ptr: (void *)str.text); |
2764 | |
2765 | return result; |
2766 | } |
2767 | |
2768 | /* Convert an identifier denoted by ID and LEN, which might contain |
2769 | UCN escapes or UTF-8 multibyte chars, to the source character set, |
2770 | either UTF-8 or UTF-EBCDIC. Assumes that the identifier is actually |
2771 | a valid identifier. */ |
2772 | cpp_hashnode * |
2773 | _cpp_interpret_identifier (cpp_reader *pfile, const uchar *id, size_t len) |
2774 | { |
2775 | /* It turns out that a UCN escape always turns into fewer characters |
2776 | than the escape itself, so we can allocate a temporary in advance. */ |
2777 | uchar * buf = (uchar *) alloca (len + 1); |
2778 | uchar * bufp = buf; |
2779 | size_t idp; |
2780 | |
2781 | for (idp = 0; idp < len; idp++) |
2782 | if (id[idp] != '\\') |
2783 | *bufp++ = id[idp]; |
2784 | else |
2785 | { |
2786 | unsigned length = id[idp + 1] == 'u' ? 4 : 8; |
2787 | cppchar_t value = 0; |
2788 | size_t bufleft = len - (bufp - buf); |
2789 | int rval; |
2790 | bool delimited = false; |
2791 | |
2792 | idp += 2; |
2793 | if (id[idp - 1] == 'N' && id[idp] == '{') |
2794 | { |
2795 | idp++; |
2796 | const uchar *name = &id[idp]; |
2797 | while (idp < len |
2798 | && (ISIDNUM (id[idp]) || id[idp] == ' ' || id[idp] == '-')) |
2799 | idp++; |
2800 | if (id[idp] == '}') |
2801 | { |
2802 | value = _cpp_uname2c (name: (const char *) name, len: &id[idp] - name, |
2803 | n: uname2c_tree, NULL); |
2804 | if (value == (cppchar_t) -1) |
2805 | value = 1; |
2806 | } |
2807 | else |
2808 | idp--; |
2809 | } |
2810 | else |
2811 | { |
2812 | if (length == 4 && id[idp] == '{') |
2813 | { |
2814 | delimited = true; |
2815 | idp++; |
2816 | } |
2817 | while (length && idp < len && ISXDIGIT (id[idp])) |
2818 | { |
2819 | value = (value << 4) + hex_value (id[idp]); |
2820 | idp++; |
2821 | if (!delimited) |
2822 | length--; |
2823 | } |
2824 | if (!delimited || id[idp] != '}') |
2825 | idp--; |
2826 | } |
2827 | |
2828 | /* Special case for EBCDIC: if the identifier contains |
2829 | a '$' specified using a UCN, translate it to EBCDIC. */ |
2830 | if (value == 0x24) |
2831 | { |
2832 | *bufp++ = '$'; |
2833 | continue; |
2834 | } |
2835 | |
2836 | rval = one_cppchar_to_utf8 (c: value, outbufp: &bufp, outbytesleftp: &bufleft); |
2837 | if (rval) |
2838 | { |
2839 | errno = rval; |
2840 | cpp_errno (pfile, CPP_DL_ERROR, |
2841 | msgid: "converting UCN to source character set" ); |
2842 | break; |
2843 | } |
2844 | } |
2845 | |
2846 | return CPP_HASHNODE (ht_lookup (pfile->hash_table, |
2847 | buf, bufp - buf, HT_ALLOC)); |
2848 | } |
2849 | |
2850 | |
2851 | /* Utility to strip a UTF-8 byte order marking from the beginning |
2852 | of a buffer. Returns the number of bytes to skip, which currently |
2853 | will be either 0 or 3. */ |
2854 | int |
2855 | cpp_check_utf8_bom (const char *data, size_t data_length) |
2856 | { |
2857 | |
2858 | #if HOST_CHARSET == HOST_CHARSET_ASCII |
2859 | const unsigned char *udata = (const unsigned char *) data; |
2860 | if (data_length >= 3 && udata[0] == 0xef && udata[1] == 0xbb |
2861 | && udata[2] == 0xbf) |
2862 | return 3; |
2863 | #endif |
2864 | |
2865 | return 0; |
2866 | } |
2867 | |
2868 | |
2869 | /* Convert an input buffer (containing the complete contents of one |
2870 | source file) from INPUT_CHARSET to the source character set. INPUT |
2871 | points to the input buffer, SIZE is its allocated size, and LEN is |
2872 | the length of the meaningful data within the buffer. The |
2873 | translated buffer is returned, *ST_SIZE is set to the length of |
2874 | the meaningful data within the translated buffer, and *BUFFER_START |
2875 | is set to the start of the returned buffer. *BUFFER_START may |
2876 | differ from the return value in the case of a BOM or other ignored |
2877 | marker information. |
2878 | |
2879 | INPUT is expected to have been allocated with xmalloc. This |
2880 | function will either set *BUFFER_START to INPUT, or free it and set |
2881 | *BUFFER_START to a pointer to another xmalloc-allocated block of |
2882 | memory. |
2883 | |
2884 | PFILE is only used to generate diagnostics; setting it to NULL suppresses |
2885 | diagnostics, and causes a return of NULL if there was any error instead. */ |
2886 | |
2887 | uchar * |
2888 | _cpp_convert_input (cpp_reader *pfile, const char *input_charset, |
2889 | uchar *input, size_t size, size_t len, |
2890 | const unsigned char **buffer_start, off_t *st_size) |
2891 | { |
2892 | struct cset_converter input_cset; |
2893 | struct _cpp_strbuf to; |
2894 | unsigned char *buffer; |
2895 | |
2896 | input_cset = init_iconv_desc (pfile, SOURCE_CHARSET, from: input_charset); |
2897 | if (input_cset.func == convert_no_conversion) |
2898 | { |
2899 | to.text = input; |
2900 | to.asize = size; |
2901 | to.len = len; |
2902 | } |
2903 | else |
2904 | { |
2905 | to.asize = MAX (65536, len); |
2906 | to.text = XNEWVEC (uchar, to.asize); |
2907 | to.len = 0; |
2908 | |
2909 | const bool ok = APPLY_CONVERSION (input_cset, input, len, &to); |
2910 | free (ptr: input); |
2911 | |
2912 | /* Clean up the mess. */ |
2913 | if (input_cset.func == convert_using_iconv) |
2914 | iconv_close (input_cset.cd); |
2915 | |
2916 | /* Handle conversion failure. */ |
2917 | if (!ok) |
2918 | { |
2919 | if (!pfile) |
2920 | { |
2921 | XDELETEVEC (to.text); |
2922 | *buffer_start = NULL; |
2923 | *st_size = 0; |
2924 | return NULL; |
2925 | } |
2926 | cpp_error (pfile, CPP_DL_ERROR, msgid: "failure to convert %s to %s" , |
2927 | input_charset, SOURCE_CHARSET); |
2928 | } |
2929 | } |
2930 | |
2931 | /* Resize buffer if we allocated substantially too much, or if we |
2932 | haven't enough space for the \n-terminator or following |
2933 | 15 bytes of padding (used to quiet warnings from valgrind or |
2934 | Address Sanitizer, when the optimized lexer accesses aligned |
2935 | 16-byte memory chunks, including the bytes after the malloced, |
2936 | area, and stops lexing on '\n'). */ |
2937 | if (to.len + 4096 < to.asize || to.len + 16 > to.asize) |
2938 | to.text = XRESIZEVEC (uchar, to.text, to.len + 16); |
2939 | |
2940 | memset (s: to.text + to.len, c: '\0', n: 16); |
2941 | |
2942 | /* If the file is using old-school Mac line endings (\r only), |
2943 | terminate with another \r, not an \n, so that we do not mistake |
2944 | the \r\n sequence for a single DOS line ending and erroneously |
2945 | issue the "No newline at end of file" diagnostic. */ |
2946 | if (to.len && to.text[to.len - 1] == '\r') |
2947 | to.text[to.len] = '\r'; |
2948 | else |
2949 | to.text[to.len] = '\n'; |
2950 | |
2951 | buffer = to.text; |
2952 | *st_size = to.len; |
2953 | |
2954 | /* Ignore a UTF-8 BOM if we see one and the source charset is UTF-8. Note |
2955 | that glib'c UTF-8 iconv() provider (as of glibc 2.7) does not ignore a |
2956 | BOM -- however, even if it did, we would still need this code due |
2957 | to the 'convert_no_conversion' case. */ |
2958 | const int bom_len = cpp_check_utf8_bom (data: (const char *) to.text, data_length: to.len); |
2959 | *st_size -= bom_len; |
2960 | buffer += bom_len; |
2961 | |
2962 | *buffer_start = to.text; |
2963 | return buffer; |
2964 | } |
2965 | |
2966 | /* Decide on the default encoding to assume for input files. */ |
2967 | const char * |
2968 | _cpp_default_encoding (void) |
2969 | { |
2970 | const char *current_encoding = NULL; |
2971 | |
2972 | /* We disable this because the default codeset is 7-bit ASCII on |
2973 | most platforms, and this causes conversion failures on every |
2974 | file in GCC that happens to have one of the upper 128 characters |
2975 | in it -- most likely, as part of the name of a contributor. |
2976 | We should definitely recognize in-band markers of file encoding, |
2977 | like: |
2978 | - the appropriate Unicode byte-order mark (FE FF) to recognize |
2979 | UTF16 and UCS4 (in both big-endian and little-endian flavors) |
2980 | and UTF8 |
2981 | - a "#i", "#d", "/ *", "//", " #p" or "#p" (for #pragma) to |
2982 | distinguish ASCII and EBCDIC. |
2983 | - now we can parse something like "#pragma GCC encoding <xyz> |
2984 | on the first line, or even Emacs/VIM's mode line tags (there's |
2985 | a problem here in that VIM uses the last line, and Emacs has |
2986 | its more elaborate "local variables" convention). |
2987 | - investigate whether Java has another common convention, which |
2988 | would be friendly to support. |
2989 | (Zack Weinberg and Paolo Bonzini, May 20th 2004) */ |
2990 | #if defined (HAVE_LOCALE_H) && defined (HAVE_LANGINFO_CODESET) && 0 |
2991 | setlocale (LC_CTYPE, "" ); |
2992 | current_encoding = nl_langinfo (CODESET); |
2993 | #endif |
2994 | if (current_encoding == NULL || *current_encoding == '\0') |
2995 | current_encoding = SOURCE_CHARSET; |
2996 | |
2997 | return current_encoding; |
2998 | } |
2999 | |
3000 | /* Check if the configured input charset requires no conversion, other than |
3001 | possibly stripping a UTF-8 BOM. */ |
3002 | bool cpp_input_conversion_is_trivial (const char *input_charset) |
3003 | { |
3004 | return !strcasecmp (s1: input_charset, SOURCE_CHARSET); |
3005 | } |
3006 | |
3007 | /* Implementation of class cpp_string_location_reader. */ |
3008 | |
3009 | /* Constructor for cpp_string_location_reader. */ |
3010 | |
3011 | cpp_string_location_reader:: |
3012 | cpp_string_location_reader (location_t src_loc, |
3013 | line_maps *line_table) |
3014 | { |
3015 | src_loc = get_range_from_loc (set: line_table, loc: src_loc).m_start; |
3016 | |
3017 | /* SRC_LOC might be a macro location. It only makes sense to do |
3018 | column-by-column calculations on ordinary maps, so get the |
3019 | corresponding location in an ordinary map. */ |
3020 | m_loc |
3021 | = linemap_resolve_location (line_table, loc: src_loc, |
3022 | lrk: LRK_SPELLING_LOCATION, NULL); |
3023 | |
3024 | const line_map_ordinary *map |
3025 | = linemap_check_ordinary (map: linemap_lookup (line_table, m_loc)); |
3026 | m_offset_per_column = (1 << map->m_range_bits); |
3027 | } |
3028 | |
3029 | /* Get the range of the next source byte. */ |
3030 | |
3031 | source_range |
3032 | cpp_string_location_reader::get_next () |
3033 | { |
3034 | source_range result; |
3035 | result.m_start = m_loc; |
3036 | result.m_finish = m_loc; |
3037 | if (m_loc <= LINE_MAP_MAX_LOCATION_WITH_COLS) |
3038 | m_loc += m_offset_per_column; |
3039 | return result; |
3040 | } |
3041 | |
3042 | cpp_display_width_computation:: |
3043 | cpp_display_width_computation (const char *data, int data_length, |
3044 | const cpp_char_column_policy &policy) : |
3045 | m_begin (data), |
3046 | m_next (m_begin), |
3047 | m_bytes_left (data_length), |
3048 | m_policy (policy), |
3049 | m_display_cols (0) |
3050 | { |
3051 | gcc_assert (policy.m_tabstop > 0); |
3052 | gcc_assert (policy.m_width_cb); |
3053 | } |
3054 | |
3055 | |
3056 | /* The main implementation function for class cpp_display_width_computation. |
3057 | m_next points on entry to the start of the UTF-8 encoding of the next |
3058 | character, and is updated to point just after the last byte of the encoding. |
3059 | m_bytes_left contains on entry the remaining size of the buffer into which |
3060 | m_next points, and this is also updated accordingly. If m_next does not |
3061 | point to a valid UTF-8-encoded sequence, then it will be treated as a single |
3062 | byte with display width 1. m_cur_display_col is the current display column, |
3063 | relative to which tab stops should be expanded. Returns the display width of |
3064 | the codepoint just processed. |
3065 | If OUT is non-NULL, it is populated. */ |
3066 | |
3067 | int |
3068 | cpp_display_width_computation::process_next_codepoint (cpp_decoded_char *out) |
3069 | { |
3070 | cppchar_t c; |
3071 | int next_width; |
3072 | |
3073 | if (out) |
3074 | out->m_start_byte = m_next; |
3075 | |
3076 | if (*m_next == '\t') |
3077 | { |
3078 | ++m_next; |
3079 | --m_bytes_left; |
3080 | next_width = m_policy.m_tabstop - (m_display_cols % m_policy.m_tabstop); |
3081 | if (out) |
3082 | { |
3083 | out->m_ch = '\t'; |
3084 | out->m_valid_ch = true; |
3085 | } |
3086 | } |
3087 | else if (one_utf8_to_cppchar (inbufp: (const uchar **) &m_next, inbytesleftp: &m_bytes_left, cp: &c) |
3088 | != 0) |
3089 | { |
3090 | /* Input is not convertible to UTF-8. This could be fine, e.g. in a |
3091 | string literal, so don't complain. Just treat it as if it has a width |
3092 | of one. */ |
3093 | ++m_next; |
3094 | --m_bytes_left; |
3095 | next_width = m_policy.m_undecoded_byte_width; |
3096 | if (out) |
3097 | out->m_valid_ch = false; |
3098 | } |
3099 | else |
3100 | { |
3101 | /* one_utf8_to_cppchar() has updated m_next and m_bytes_left for us. */ |
3102 | next_width = m_policy.m_width_cb (c); |
3103 | if (out) |
3104 | { |
3105 | out->m_ch = c; |
3106 | out->m_valid_ch = true; |
3107 | } |
3108 | } |
3109 | |
3110 | if (out) |
3111 | out->m_next_byte = m_next; |
3112 | |
3113 | m_display_cols += next_width; |
3114 | return next_width; |
3115 | } |
3116 | |
3117 | /* Utility to advance the byte stream by the minimum amount needed to consume |
3118 | N display columns. Returns the number of display columns that were |
3119 | actually skipped. This could be less than N, if there was not enough data, |
3120 | or more than N, if the last character to be skipped had a sufficiently large |
3121 | display width. */ |
3122 | int |
3123 | cpp_display_width_computation::advance_display_cols (int n) |
3124 | { |
3125 | const int start = m_display_cols; |
3126 | const int target = start + n; |
3127 | while (m_display_cols < target && !done ()) |
3128 | process_next_codepoint (NULL); |
3129 | return m_display_cols - start; |
3130 | } |
3131 | |
3132 | /* For the string of length DATA_LENGTH bytes that begins at DATA, compute |
3133 | how many display columns are occupied by the first COLUMN bytes. COLUMN |
3134 | may exceed DATA_LENGTH, in which case the phantom bytes at the end are |
3135 | treated as if they have display width 1. Tabs are expanded to the next tab |
3136 | stop, relative to the start of DATA, and non-printable-ASCII characters |
3137 | will be escaped as per POLICY. */ |
3138 | |
3139 | int |
3140 | cpp_byte_column_to_display_column (const char *data, int data_length, |
3141 | int column, |
3142 | const cpp_char_column_policy &policy) |
3143 | { |
3144 | const int offset = MAX (0, column - data_length); |
3145 | cpp_display_width_computation dw (data, column - offset, policy); |
3146 | while (!dw.done ()) |
3147 | dw.process_next_codepoint (NULL); |
3148 | return dw.display_cols_processed () + offset; |
3149 | } |
3150 | |
3151 | /* For the string of length DATA_LENGTH bytes that begins at DATA, compute |
3152 | the least number of bytes that will result in at least DISPLAY_COL display |
3153 | columns. The return value may exceed DATA_LENGTH if the entire string does |
3154 | not occupy enough display columns. Non-printable-ASCII characters |
3155 | will be escaped as per POLICY. */ |
3156 | |
3157 | int |
3158 | cpp_display_column_to_byte_column (const char *data, int data_length, |
3159 | int display_col, |
3160 | const cpp_char_column_policy &policy) |
3161 | { |
3162 | cpp_display_width_computation dw (data, data_length, policy); |
3163 | const int avail_display = dw.advance_display_cols (n: display_col); |
3164 | return dw.bytes_processed () + MAX (0, display_col - avail_display); |
3165 | } |
3166 | |
3167 | template <typename PropertyType> |
3168 | PropertyType |
3169 | get_cppchar_property (cppchar_t c, |
3170 | const cppchar_t *range_ends, |
3171 | const PropertyType *range_values, |
3172 | size_t num_ranges, |
3173 | PropertyType default_value) |
3174 | { |
3175 | if (__builtin_expect (c <= range_ends[0], true)) |
3176 | return range_values[0]; |
3177 | |
3178 | /* Binary search the tables. */ |
3179 | int begin = 1; |
3180 | static const int end = num_ranges; |
3181 | int len = end - begin; |
3182 | do |
3183 | { |
3184 | int half = len/2; |
3185 | int middle = begin + half; |
3186 | if (c > range_ends[middle]) |
3187 | { |
3188 | begin = middle + 1; |
3189 | len -= half + 1; |
3190 | } |
3191 | else |
3192 | len = half; |
3193 | } while (len); |
3194 | |
3195 | if (__builtin_expect (begin != end, true)) |
3196 | return range_values[begin]; |
3197 | |
3198 | return default_value; |
3199 | } |
3200 | |
3201 | /* Our own version of wcwidth(). We don't use the actual wcwidth() in glibc, |
3202 | because that will inspect the user's locale, and in particular in an ASCII |
3203 | locale, it will not return anything useful for extended characters. But GCC |
3204 | in other respects (see e.g. _cpp_default_encoding()) behaves as if |
3205 | everything is UTF-8. We also make some tweaks that are useful for the way |
3206 | GCC needs to use this data, e.g. tabs and other control characters should be |
3207 | treated as having width 1. The lookup tables are generated from |
3208 | contrib/unicode/gen_wcwidth.py and were made by simply calling glibc |
3209 | wcwidth() on all codepoints, then applying the small tweaks. These tables |
3210 | are not highly optimized, but for the present purpose of outputting |
3211 | diagnostics, they are sufficient. */ |
3212 | |
3213 | #include "generated_cpp_wcwidth.h" |
3214 | |
3215 | int |
3216 | cpp_wcwidth (cppchar_t c) |
3217 | { |
3218 | const size_t num_ranges |
3219 | = sizeof wcwidth_range_ends / sizeof (*wcwidth_range_ends); |
3220 | return get_cppchar_property<unsigned char > (c, |
3221 | range_ends: &wcwidth_range_ends[0], |
3222 | range_values: &wcwidth_widths[0], |
3223 | num_ranges, |
3224 | default_value: 1); |
3225 | } |
3226 | |
3227 | #include "combining-chars.inc" |
3228 | |
3229 | bool |
3230 | cpp_is_combining_char (cppchar_t c) |
3231 | { |
3232 | const size_t num_ranges |
3233 | = sizeof combining_range_ends / sizeof (*combining_range_ends); |
3234 | return get_cppchar_property<bool> (c, |
3235 | range_ends: &combining_range_ends[0], |
3236 | range_values: &is_combining[0], |
3237 | num_ranges, |
3238 | default_value: false); |
3239 | } |
3240 | |
3241 | #include "printable-chars.inc" |
3242 | |
3243 | bool |
3244 | cpp_is_printable_char (cppchar_t c) |
3245 | { |
3246 | const size_t num_ranges |
3247 | = sizeof printable_range_ends / sizeof (*printable_range_ends); |
3248 | return get_cppchar_property<bool> (c, |
3249 | range_ends: &printable_range_ends[0], |
3250 | range_values: &is_printable[0], |
3251 | num_ranges, |
3252 | default_value: false); |
3253 | } |
3254 | |