]>
Commit | Line | Data |
---|---|---|
1 | /* CPP Library - lexical analysis. | |
2 | Copyright (C) 2000, 2001, 2002, 2003, 2004, 2005 Free Software Foundation, Inc. | |
3 | Contributed by Per Bothner, 1994-95. | |
4 | Based on CCCP program by Paul Rubin, June 1986 | |
5 | Adapted to ANSI C, Richard Stallman, Jan 1987 | |
6 | Broken out to separate file, Zack Weinberg, Mar 2000 | |
7 | ||
8 | This program is free software; you can redistribute it and/or modify it | |
9 | under the terms of the GNU General Public License as published by the | |
10 | Free Software Foundation; either version 2, or (at your option) any | |
11 | later version. | |
12 | ||
13 | This program is distributed in the hope that it will be useful, | |
14 | but WITHOUT ANY WARRANTY; without even the implied warranty of | |
15 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
16 | GNU General Public License for more details. | |
17 | ||
18 | You should have received a copy of the GNU General Public License | |
19 | along with this program; if not, write to the Free Software | |
20 | Foundation, 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA. */ | |
21 | ||
22 | #include "config.h" | |
23 | #include "system.h" | |
24 | #include "cpplib.h" | |
25 | #include "internal.h" | |
26 | ||
27 | enum spell_type | |
28 | { | |
29 | SPELL_OPERATOR = 0, | |
30 | SPELL_IDENT, | |
31 | SPELL_LITERAL, | |
32 | SPELL_NONE | |
33 | }; | |
34 | ||
35 | struct token_spelling | |
36 | { | |
37 | enum spell_type category; | |
38 | const unsigned char *name; | |
39 | }; | |
40 | ||
41 | static const unsigned char *const digraph_spellings[] = | |
42 | { U"%:", U"%:%:", U"<:", U":>", U"<%", U"%>" }; | |
43 | ||
44 | #define OP(e, s) { SPELL_OPERATOR, U s }, | |
45 | #define TK(e, s) { SPELL_ ## s, U #e }, | |
46 | static const struct token_spelling token_spellings[N_TTYPES] = { TTYPE_TABLE }; | |
47 | #undef OP | |
48 | #undef TK | |
49 | ||
50 | #define TOKEN_SPELL(token) (token_spellings[(token)->type].category) | |
51 | #define TOKEN_NAME(token) (token_spellings[(token)->type].name) | |
52 | ||
53 | static void add_line_note (cpp_buffer *, const uchar *, unsigned int); | |
54 | static int skip_line_comment (cpp_reader *); | |
55 | static void skip_whitespace (cpp_reader *, cppchar_t); | |
56 | static void lex_string (cpp_reader *, cpp_token *, const uchar *); | |
57 | static void save_comment (cpp_reader *, cpp_token *, const uchar *, cppchar_t); | |
58 | static void create_literal (cpp_reader *, cpp_token *, const uchar *, | |
59 | unsigned int, enum cpp_ttype); | |
60 | static bool warn_in_comment (cpp_reader *, _cpp_line_note *); | |
61 | static int name_p (cpp_reader *, const cpp_string *); | |
62 | static tokenrun *next_tokenrun (tokenrun *); | |
63 | ||
64 | static _cpp_buff *new_buff (size_t); | |
65 | ||
66 | ||
67 | /* Utility routine: | |
68 | ||
69 | Compares, the token TOKEN to the NUL-terminated string STRING. | |
70 | TOKEN must be a CPP_NAME. Returns 1 for equal, 0 for unequal. */ | |
71 | int | |
72 | cpp_ideq (const cpp_token *token, const char *string) | |
73 | { | |
74 | if (token->type != CPP_NAME) | |
75 | return 0; | |
76 | ||
77 | return !ustrcmp (NODE_NAME (token->val.node), (const uchar *) string); | |
78 | } | |
79 | ||
80 | /* Record a note TYPE at byte POS into the current cleaned logical | |
81 | line. */ | |
82 | static void | |
83 | add_line_note (cpp_buffer *buffer, const uchar *pos, unsigned int type) | |
84 | { | |
85 | if (buffer->notes_used == buffer->notes_cap) | |
86 | { | |
87 | buffer->notes_cap = buffer->notes_cap * 2 + 200; | |
88 | buffer->notes = XRESIZEVEC (_cpp_line_note, buffer->notes, | |
89 | buffer->notes_cap); | |
90 | } | |
91 | ||
92 | buffer->notes[buffer->notes_used].pos = pos; | |
93 | buffer->notes[buffer->notes_used].type = type; | |
94 | buffer->notes_used++; | |
95 | } | |
96 | ||
97 | /* Returns with a logical line that contains no escaped newlines or | |
98 | trigraphs. This is a time-critical inner loop. */ | |
99 | void | |
100 | _cpp_clean_line (cpp_reader *pfile) | |
101 | { | |
102 | cpp_buffer *buffer; | |
103 | const uchar *s; | |
104 | uchar c, *d, *p; | |
105 | ||
106 | buffer = pfile->buffer; | |
107 | buffer->cur_note = buffer->notes_used = 0; | |
108 | buffer->cur = buffer->line_base = buffer->next_line; | |
109 | buffer->need_line = false; | |
110 | s = buffer->next_line - 1; | |
111 | ||
112 | if (!buffer->from_stage3) | |
113 | { | |
114 | /* Short circuit for the common case of an un-escaped line with | |
115 | no trigraphs. The primary win here is by not writing any | |
116 | data back to memory until we have to. */ | |
117 | for (;;) | |
118 | { | |
119 | c = *++s; | |
120 | if (c == '\n' || c == '\r') | |
121 | { | |
122 | d = (uchar *) s; | |
123 | ||
124 | if (s == buffer->rlimit) | |
125 | goto done; | |
126 | ||
127 | /* DOS line ending? */ | |
128 | if (c == '\r' && s[1] == '\n') | |
129 | s++; | |
130 | ||
131 | if (s == buffer->rlimit) | |
132 | goto done; | |
133 | ||
134 | /* check for escaped newline */ | |
135 | p = d; | |
136 | while (p != buffer->next_line && is_nvspace (p[-1])) | |
137 | p--; | |
138 | if (p == buffer->next_line || p[-1] != '\\') | |
139 | goto done; | |
140 | ||
141 | /* Have an escaped newline; process it and proceed to | |
142 | the slow path. */ | |
143 | add_line_note (buffer, p - 1, p != d ? ' ' : '\\'); | |
144 | d = p - 2; | |
145 | buffer->next_line = p - 1; | |
146 | break; | |
147 | } | |
148 | if (c == '?' && s[1] == '?' && _cpp_trigraph_map[s[2]]) | |
149 | { | |
150 | /* Have a trigraph. We may or may not have to convert | |
151 | it. Add a line note regardless, for -Wtrigraphs. */ | |
152 | add_line_note (buffer, s, s[2]); | |
153 | if (CPP_OPTION (pfile, trigraphs)) | |
154 | { | |
155 | /* We do, and that means we have to switch to the | |
156 | slow path. */ | |
157 | d = (uchar *) s; | |
158 | *d = _cpp_trigraph_map[s[2]]; | |
159 | s += 2; | |
160 | break; | |
161 | } | |
162 | } | |
163 | } | |
164 | ||
165 | ||
166 | for (;;) | |
167 | { | |
168 | c = *++s; | |
169 | *++d = c; | |
170 | ||
171 | if (c == '\n' || c == '\r') | |
172 | { | |
173 | /* Handle DOS line endings. */ | |
174 | if (c == '\r' && s != buffer->rlimit && s[1] == '\n') | |
175 | s++; | |
176 | if (s == buffer->rlimit) | |
177 | break; | |
178 | ||
179 | /* Escaped? */ | |
180 | p = d; | |
181 | while (p != buffer->next_line && is_nvspace (p[-1])) | |
182 | p--; | |
183 | if (p == buffer->next_line || p[-1] != '\\') | |
184 | break; | |
185 | ||
186 | add_line_note (buffer, p - 1, p != d ? ' ': '\\'); | |
187 | d = p - 2; | |
188 | buffer->next_line = p - 1; | |
189 | } | |
190 | else if (c == '?' && s[1] == '?' && _cpp_trigraph_map[s[2]]) | |
191 | { | |
192 | /* Add a note regardless, for the benefit of -Wtrigraphs. */ | |
193 | add_line_note (buffer, d, s[2]); | |
194 | if (CPP_OPTION (pfile, trigraphs)) | |
195 | { | |
196 | *d = _cpp_trigraph_map[s[2]]; | |
197 | s += 2; | |
198 | } | |
199 | } | |
200 | } | |
201 | } | |
202 | else | |
203 | { | |
204 | do | |
205 | s++; | |
206 | while (*s != '\n' && *s != '\r'); | |
207 | d = (uchar *) s; | |
208 | ||
209 | /* Handle DOS line endings. */ | |
210 | if (*s == '\r' && s != buffer->rlimit && s[1] == '\n') | |
211 | s++; | |
212 | } | |
213 | ||
214 | done: | |
215 | *d = '\n'; | |
216 | /* A sentinel note that should never be processed. */ | |
217 | add_line_note (buffer, d + 1, '\n'); | |
218 | buffer->next_line = s + 1; | |
219 | } | |
220 | ||
221 | /* Return true if the trigraph indicated by NOTE should be warned | |
222 | about in a comment. */ | |
223 | static bool | |
224 | warn_in_comment (cpp_reader *pfile, _cpp_line_note *note) | |
225 | { | |
226 | const uchar *p; | |
227 | ||
228 | /* Within comments we don't warn about trigraphs, unless the | |
229 | trigraph forms an escaped newline, as that may change | |
230 | behavior. */ | |
231 | if (note->type != '/') | |
232 | return false; | |
233 | ||
234 | /* If -trigraphs, then this was an escaped newline iff the next note | |
235 | is coincident. */ | |
236 | if (CPP_OPTION (pfile, trigraphs)) | |
237 | return note[1].pos == note->pos; | |
238 | ||
239 | /* Otherwise, see if this forms an escaped newline. */ | |
240 | p = note->pos + 3; | |
241 | while (is_nvspace (*p)) | |
242 | p++; | |
243 | ||
244 | /* There might have been escaped newlines between the trigraph and the | |
245 | newline we found. Hence the position test. */ | |
246 | return (*p == '\n' && p < note[1].pos); | |
247 | } | |
248 | ||
249 | /* Process the notes created by add_line_note as far as the current | |
250 | location. */ | |
251 | void | |
252 | _cpp_process_line_notes (cpp_reader *pfile, int in_comment) | |
253 | { | |
254 | cpp_buffer *buffer = pfile->buffer; | |
255 | ||
256 | for (;;) | |
257 | { | |
258 | _cpp_line_note *note = &buffer->notes[buffer->cur_note]; | |
259 | unsigned int col; | |
260 | ||
261 | if (note->pos > buffer->cur) | |
262 | break; | |
263 | ||
264 | buffer->cur_note++; | |
265 | col = CPP_BUF_COLUMN (buffer, note->pos + 1); | |
266 | ||
267 | if (note->type == '\\' || note->type == ' ') | |
268 | { | |
269 | if (note->type == ' ' && !in_comment) | |
270 | cpp_error_with_line (pfile, CPP_DL_WARNING, pfile->line_table->highest_line, col, | |
271 | "backslash and newline separated by space"); | |
272 | ||
273 | if (buffer->next_line > buffer->rlimit) | |
274 | { | |
275 | cpp_error_with_line (pfile, CPP_DL_PEDWARN, pfile->line_table->highest_line, col, | |
276 | "backslash-newline at end of file"); | |
277 | /* Prevent "no newline at end of file" warning. */ | |
278 | buffer->next_line = buffer->rlimit; | |
279 | } | |
280 | ||
281 | buffer->line_base = note->pos; | |
282 | CPP_INCREMENT_LINE (pfile, 0); | |
283 | } | |
284 | else if (_cpp_trigraph_map[note->type]) | |
285 | { | |
286 | if (CPP_OPTION (pfile, warn_trigraphs) | |
287 | && (!in_comment || warn_in_comment (pfile, note))) | |
288 | { | |
289 | if (CPP_OPTION (pfile, trigraphs)) | |
290 | cpp_error_with_line (pfile, CPP_DL_WARNING, pfile->line_table->highest_line, col, | |
291 | "trigraph ??%c converted to %c", | |
292 | note->type, | |
293 | (int) _cpp_trigraph_map[note->type]); | |
294 | else | |
295 | { | |
296 | cpp_error_with_line | |
297 | (pfile, CPP_DL_WARNING, pfile->line_table->highest_line, col, | |
298 | "trigraph ??%c ignored, use -trigraphs to enable", | |
299 | note->type); | |
300 | } | |
301 | } | |
302 | } | |
303 | else | |
304 | abort (); | |
305 | } | |
306 | } | |
307 | ||
308 | /* Skip a C-style block comment. We find the end of the comment by | |
309 | seeing if an asterisk is before every '/' we encounter. Returns | |
310 | nonzero if comment terminated by EOF, zero otherwise. | |
311 | ||
312 | Buffer->cur points to the initial asterisk of the comment. */ | |
313 | bool | |
314 | _cpp_skip_block_comment (cpp_reader *pfile) | |
315 | { | |
316 | cpp_buffer *buffer = pfile->buffer; | |
317 | const uchar *cur = buffer->cur; | |
318 | uchar c; | |
319 | ||
320 | cur++; | |
321 | if (*cur == '/') | |
322 | cur++; | |
323 | ||
324 | for (;;) | |
325 | { | |
326 | /* People like decorating comments with '*', so check for '/' | |
327 | instead for efficiency. */ | |
328 | c = *cur++; | |
329 | ||
330 | if (c == '/') | |
331 | { | |
332 | if (cur[-2] == '*') | |
333 | break; | |
334 | ||
335 | /* Warn about potential nested comments, but not if the '/' | |
336 | comes immediately before the true comment delimiter. | |
337 | Don't bother to get it right across escaped newlines. */ | |
338 | if (CPP_OPTION (pfile, warn_comments) | |
339 | && cur[0] == '*' && cur[1] != '/') | |
340 | { | |
341 | buffer->cur = cur; | |
342 | cpp_error_with_line (pfile, CPP_DL_WARNING, | |
343 | pfile->line_table->highest_line, CPP_BUF_COL (buffer), | |
344 | "\"/*\" within comment"); | |
345 | } | |
346 | } | |
347 | else if (c == '\n') | |
348 | { | |
349 | unsigned int cols; | |
350 | buffer->cur = cur - 1; | |
351 | _cpp_process_line_notes (pfile, true); | |
352 | if (buffer->next_line >= buffer->rlimit) | |
353 | return true; | |
354 | _cpp_clean_line (pfile); | |
355 | ||
356 | cols = buffer->next_line - buffer->line_base; | |
357 | CPP_INCREMENT_LINE (pfile, cols); | |
358 | ||
359 | cur = buffer->cur; | |
360 | } | |
361 | } | |
362 | ||
363 | buffer->cur = cur; | |
364 | _cpp_process_line_notes (pfile, true); | |
365 | return false; | |
366 | } | |
367 | ||
368 | /* Skip a C++ line comment, leaving buffer->cur pointing to the | |
369 | terminating newline. Handles escaped newlines. Returns nonzero | |
370 | if a multiline comment. */ | |
371 | static int | |
372 | skip_line_comment (cpp_reader *pfile) | |
373 | { | |
374 | cpp_buffer *buffer = pfile->buffer; | |
375 | unsigned int orig_line = pfile->line_table->highest_line; | |
376 | ||
377 | while (*buffer->cur != '\n') | |
378 | buffer->cur++; | |
379 | ||
380 | _cpp_process_line_notes (pfile, true); | |
381 | return orig_line != pfile->line_table->highest_line; | |
382 | } | |
383 | ||
384 | /* Skips whitespace, saving the next non-whitespace character. */ | |
385 | static void | |
386 | skip_whitespace (cpp_reader *pfile, cppchar_t c) | |
387 | { | |
388 | cpp_buffer *buffer = pfile->buffer; | |
389 | bool saw_NUL = false; | |
390 | ||
391 | do | |
392 | { | |
393 | /* Horizontal space always OK. */ | |
394 | if (c == ' ' || c == '\t') | |
395 | ; | |
396 | /* Just \f \v or \0 left. */ | |
397 | else if (c == '\0') | |
398 | saw_NUL = true; | |
399 | else if (pfile->state.in_directive && CPP_PEDANTIC (pfile)) | |
400 | cpp_error_with_line (pfile, CPP_DL_PEDWARN, pfile->line_table->highest_line, | |
401 | CPP_BUF_COL (buffer), | |
402 | "%s in preprocessing directive", | |
403 | c == '\f' ? "form feed" : "vertical tab"); | |
404 | ||
405 | c = *buffer->cur++; | |
406 | } | |
407 | /* We only want non-vertical space, i.e. ' ' \t \f \v \0. */ | |
408 | while (is_nvspace (c)); | |
409 | ||
410 | if (saw_NUL) | |
411 | cpp_error (pfile, CPP_DL_WARNING, "null character(s) ignored"); | |
412 | ||
413 | buffer->cur--; | |
414 | } | |
415 | ||
416 | /* See if the characters of a number token are valid in a name (no | |
417 | '.', '+' or '-'). */ | |
418 | static int | |
419 | name_p (cpp_reader *pfile, const cpp_string *string) | |
420 | { | |
421 | unsigned int i; | |
422 | ||
423 | for (i = 0; i < string->len; i++) | |
424 | if (!is_idchar (string->text[i])) | |
425 | return 0; | |
426 | ||
427 | return 1; | |
428 | } | |
429 | ||
430 | /* After parsing an identifier or other sequence, produce a warning about | |
431 | sequences not in NFC/NFKC. */ | |
432 | static void | |
433 | warn_about_normalization (cpp_reader *pfile, | |
434 | const cpp_token *token, | |
435 | const struct normalize_state *s) | |
436 | { | |
437 | if (CPP_OPTION (pfile, warn_normalize) < NORMALIZE_STATE_RESULT (s) | |
438 | && !pfile->state.skipping) | |
439 | { | |
440 | /* Make sure that the token is printed using UCNs, even | |
441 | if we'd otherwise happily print UTF-8. */ | |
442 | unsigned char *buf = XNEWVEC (unsigned char, cpp_token_len (token)); | |
443 | size_t sz; | |
444 | ||
445 | sz = cpp_spell_token (pfile, token, buf, false) - buf; | |
446 | if (NORMALIZE_STATE_RESULT (s) == normalized_C) | |
447 | cpp_error_with_line (pfile, CPP_DL_WARNING, token->src_loc, 0, | |
448 | "`%.*s' is not in NFKC", (int) sz, buf); | |
449 | else | |
450 | cpp_error_with_line (pfile, CPP_DL_WARNING, token->src_loc, 0, | |
451 | "`%.*s' is not in NFC", (int) sz, buf); | |
452 | } | |
453 | } | |
454 | ||
455 | /* Returns TRUE if the sequence starting at buffer->cur is invalid in | |
456 | an identifier. FIRST is TRUE if this starts an identifier. */ | |
457 | static bool | |
458 | forms_identifier_p (cpp_reader *pfile, int first, | |
459 | struct normalize_state *state) | |
460 | { | |
461 | cpp_buffer *buffer = pfile->buffer; | |
462 | ||
463 | if (*buffer->cur == '$') | |
464 | { | |
465 | if (!CPP_OPTION (pfile, dollars_in_ident)) | |
466 | return false; | |
467 | ||
468 | buffer->cur++; | |
469 | if (CPP_OPTION (pfile, warn_dollars) && !pfile->state.skipping) | |
470 | { | |
471 | CPP_OPTION (pfile, warn_dollars) = 0; | |
472 | cpp_error (pfile, CPP_DL_PEDWARN, "'$' in identifier or number"); | |
473 | } | |
474 | ||
475 | return true; | |
476 | } | |
477 | ||
478 | /* Is this a syntactically valid UCN? */ | |
479 | if (CPP_OPTION (pfile, extended_identifiers) | |
480 | && *buffer->cur == '\\' | |
481 | && (buffer->cur[1] == 'u' || buffer->cur[1] == 'U')) | |
482 | { | |
483 | buffer->cur += 2; | |
484 | if (_cpp_valid_ucn (pfile, &buffer->cur, buffer->rlimit, 1 + !first, | |
485 | state)) | |
486 | return true; | |
487 | buffer->cur -= 2; | |
488 | } | |
489 | ||
490 | return false; | |
491 | } | |
492 | ||
493 | /* Lex an identifier starting at BUFFER->CUR - 1. */ | |
494 | static cpp_hashnode * | |
495 | lex_identifier (cpp_reader *pfile, const uchar *base, bool starts_ucn, | |
496 | struct normalize_state *nst) | |
497 | { | |
498 | cpp_hashnode *result; | |
499 | const uchar *cur; | |
500 | unsigned int len; | |
501 | unsigned int hash = HT_HASHSTEP (0, *base); | |
502 | ||
503 | cur = pfile->buffer->cur; | |
504 | if (! starts_ucn) | |
505 | while (ISIDNUM (*cur)) | |
506 | { | |
507 | hash = HT_HASHSTEP (hash, *cur); | |
508 | cur++; | |
509 | } | |
510 | pfile->buffer->cur = cur; | |
511 | if (starts_ucn || forms_identifier_p (pfile, false, nst)) | |
512 | { | |
513 | /* Slower version for identifiers containing UCNs (or $). */ | |
514 | do { | |
515 | while (ISIDNUM (*pfile->buffer->cur)) | |
516 | { | |
517 | pfile->buffer->cur++; | |
518 | NORMALIZE_STATE_UPDATE_IDNUM (nst); | |
519 | } | |
520 | } while (forms_identifier_p (pfile, false, nst)); | |
521 | result = _cpp_interpret_identifier (pfile, base, | |
522 | pfile->buffer->cur - base); | |
523 | } | |
524 | else | |
525 | { | |
526 | len = cur - base; | |
527 | hash = HT_HASHFINISH (hash, len); | |
528 | ||
529 | result = (cpp_hashnode *) | |
530 | ht_lookup_with_hash (pfile->hash_table, base, len, hash, HT_ALLOC); | |
531 | } | |
532 | ||
533 | /* Rarely, identifiers require diagnostics when lexed. */ | |
534 | if (__builtin_expect ((result->flags & NODE_DIAGNOSTIC) | |
535 | && !pfile->state.skipping, 0)) | |
536 | { | |
537 | /* It is allowed to poison the same identifier twice. */ | |
538 | if ((result->flags & NODE_POISONED) && !pfile->state.poisoned_ok) | |
539 | cpp_error (pfile, CPP_DL_ERROR, "attempt to use poisoned \"%s\"", | |
540 | NODE_NAME (result)); | |
541 | ||
542 | /* Constraint 6.10.3.5: __VA_ARGS__ should only appear in the | |
543 | replacement list of a variadic macro. */ | |
544 | if (result == pfile->spec_nodes.n__VA_ARGS__ | |
545 | && !pfile->state.va_args_ok) | |
546 | cpp_error (pfile, CPP_DL_PEDWARN, | |
547 | "__VA_ARGS__ can only appear in the expansion" | |
548 | " of a C99 variadic macro"); | |
549 | } | |
550 | ||
551 | return result; | |
552 | } | |
553 | ||
554 | /* Lex a number to NUMBER starting at BUFFER->CUR - 1. */ | |
555 | static void | |
556 | lex_number (cpp_reader *pfile, cpp_string *number, | |
557 | struct normalize_state *nst) | |
558 | { | |
559 | const uchar *cur; | |
560 | const uchar *base; | |
561 | uchar *dest; | |
562 | ||
563 | base = pfile->buffer->cur - 1; | |
564 | do | |
565 | { | |
566 | cur = pfile->buffer->cur; | |
567 | ||
568 | /* N.B. ISIDNUM does not include $. */ | |
569 | while (ISIDNUM (*cur) || *cur == '.' || VALID_SIGN (*cur, cur[-1])) | |
570 | { | |
571 | cur++; | |
572 | NORMALIZE_STATE_UPDATE_IDNUM (nst); | |
573 | } | |
574 | ||
575 | pfile->buffer->cur = cur; | |
576 | } | |
577 | while (forms_identifier_p (pfile, false, nst)); | |
578 | ||
579 | number->len = cur - base; | |
580 | dest = _cpp_unaligned_alloc (pfile, number->len + 1); | |
581 | memcpy (dest, base, number->len); | |
582 | dest[number->len] = '\0'; | |
583 | number->text = dest; | |
584 | } | |
585 | ||
586 | /* Create a token of type TYPE with a literal spelling. */ | |
587 | static void | |
588 | create_literal (cpp_reader *pfile, cpp_token *token, const uchar *base, | |
589 | unsigned int len, enum cpp_ttype type) | |
590 | { | |
591 | uchar *dest = _cpp_unaligned_alloc (pfile, len + 1); | |
592 | ||
593 | memcpy (dest, base, len); | |
594 | dest[len] = '\0'; | |
595 | token->type = type; | |
596 | token->val.str.len = len; | |
597 | token->val.str.text = dest; | |
598 | } | |
599 | ||
600 | /* Lexes a string, character constant, or angle-bracketed header file | |
601 | name. The stored string contains the spelling, including opening | |
602 | quote and leading any leading 'L'. It returns the type of the | |
603 | literal, or CPP_OTHER if it was not properly terminated. | |
604 | ||
605 | The spelling is NUL-terminated, but it is not guaranteed that this | |
606 | is the first NUL since embedded NULs are preserved. */ | |
607 | static void | |
608 | lex_string (cpp_reader *pfile, cpp_token *token, const uchar *base) | |
609 | { | |
610 | bool saw_NUL = false; | |
611 | const uchar *cur; | |
612 | cppchar_t terminator; | |
613 | enum cpp_ttype type; | |
614 | ||
615 | cur = base; | |
616 | terminator = *cur++; | |
617 | if (terminator == 'L') | |
618 | terminator = *cur++; | |
619 | if (terminator == '\"') | |
620 | type = *base == 'L' ? CPP_WSTRING: CPP_STRING; | |
621 | else if (terminator == '\'') | |
622 | type = *base == 'L' ? CPP_WCHAR: CPP_CHAR; | |
623 | else | |
624 | terminator = '>', type = CPP_HEADER_NAME; | |
625 | ||
626 | for (;;) | |
627 | { | |
628 | cppchar_t c = *cur++; | |
629 | ||
630 | /* In #include-style directives, terminators are not escapable. */ | |
631 | if (c == '\\' && !pfile->state.angled_headers && *cur != '\n') | |
632 | cur++; | |
633 | else if (c == terminator) | |
634 | break; | |
635 | else if (c == '\n') | |
636 | { | |
637 | cur--; | |
638 | type = CPP_OTHER; | |
639 | break; | |
640 | } | |
641 | else if (c == '\0') | |
642 | saw_NUL = true; | |
643 | } | |
644 | ||
645 | if (saw_NUL && !pfile->state.skipping) | |
646 | cpp_error (pfile, CPP_DL_WARNING, | |
647 | "null character(s) preserved in literal"); | |
648 | ||
649 | pfile->buffer->cur = cur; | |
650 | create_literal (pfile, token, base, cur - base, type); | |
651 | } | |
652 | ||
653 | /* The stored comment includes the comment start and any terminator. */ | |
654 | static void | |
655 | save_comment (cpp_reader *pfile, cpp_token *token, const unsigned char *from, | |
656 | cppchar_t type) | |
657 | { | |
658 | unsigned char *buffer; | |
659 | unsigned int len, clen; | |
660 | ||
661 | len = pfile->buffer->cur - from + 1; /* + 1 for the initial '/'. */ | |
662 | ||
663 | /* C++ comments probably (not definitely) have moved past a new | |
664 | line, which we don't want to save in the comment. */ | |
665 | if (is_vspace (pfile->buffer->cur[-1])) | |
666 | len--; | |
667 | ||
668 | /* If we are currently in a directive, then we need to store all | |
669 | C++ comments as C comments internally, and so we need to | |
670 | allocate a little extra space in that case. | |
671 | ||
672 | Note that the only time we encounter a directive here is | |
673 | when we are saving comments in a "#define". */ | |
674 | clen = (pfile->state.in_directive && type == '/') ? len + 2 : len; | |
675 | ||
676 | buffer = _cpp_unaligned_alloc (pfile, clen); | |
677 | ||
678 | token->type = CPP_COMMENT; | |
679 | token->val.str.len = clen; | |
680 | token->val.str.text = buffer; | |
681 | ||
682 | buffer[0] = '/'; | |
683 | memcpy (buffer + 1, from, len - 1); | |
684 | ||
685 | /* Finish conversion to a C comment, if necessary. */ | |
686 | if (pfile->state.in_directive && type == '/') | |
687 | { | |
688 | buffer[1] = '*'; | |
689 | buffer[clen - 2] = '*'; | |
690 | buffer[clen - 1] = '/'; | |
691 | } | |
692 | } | |
693 | ||
694 | /* Allocate COUNT tokens for RUN. */ | |
695 | void | |
696 | _cpp_init_tokenrun (tokenrun *run, unsigned int count) | |
697 | { | |
698 | run->base = XNEWVEC (cpp_token, count); | |
699 | run->limit = run->base + count; | |
700 | run->next = NULL; | |
701 | } | |
702 | ||
703 | /* Returns the next tokenrun, or creates one if there is none. */ | |
704 | static tokenrun * | |
705 | next_tokenrun (tokenrun *run) | |
706 | { | |
707 | if (run->next == NULL) | |
708 | { | |
709 | run->next = XNEW (tokenrun); | |
710 | run->next->prev = run; | |
711 | _cpp_init_tokenrun (run->next, 250); | |
712 | } | |
713 | ||
714 | return run->next; | |
715 | } | |
716 | ||
717 | /* Allocate a single token that is invalidated at the same time as the | |
718 | rest of the tokens on the line. Has its line and col set to the | |
719 | same as the last lexed token, so that diagnostics appear in the | |
720 | right place. */ | |
721 | cpp_token * | |
722 | _cpp_temp_token (cpp_reader *pfile) | |
723 | { | |
724 | cpp_token *old, *result; | |
725 | ||
726 | old = pfile->cur_token - 1; | |
727 | if (pfile->cur_token == pfile->cur_run->limit) | |
728 | { | |
729 | pfile->cur_run = next_tokenrun (pfile->cur_run); | |
730 | pfile->cur_token = pfile->cur_run->base; | |
731 | } | |
732 | ||
733 | result = pfile->cur_token++; | |
734 | result->src_loc = old->src_loc; | |
735 | return result; | |
736 | } | |
737 | ||
738 | /* Lex a token into RESULT (external interface). Takes care of issues | |
739 | like directive handling, token lookahead, multiple include | |
740 | optimization and skipping. */ | |
741 | const cpp_token * | |
742 | _cpp_lex_token (cpp_reader *pfile) | |
743 | { | |
744 | cpp_token *result; | |
745 | ||
746 | for (;;) | |
747 | { | |
748 | if (pfile->cur_token == pfile->cur_run->limit) | |
749 | { | |
750 | pfile->cur_run = next_tokenrun (pfile->cur_run); | |
751 | pfile->cur_token = pfile->cur_run->base; | |
752 | } | |
753 | ||
754 | if (pfile->lookaheads) | |
755 | { | |
756 | pfile->lookaheads--; | |
757 | result = pfile->cur_token++; | |
758 | } | |
759 | else | |
760 | result = _cpp_lex_direct (pfile); | |
761 | ||
762 | if (result->flags & BOL) | |
763 | { | |
764 | /* Is this a directive. If _cpp_handle_directive returns | |
765 | false, it is an assembler #. */ | |
766 | if (result->type == CPP_HASH | |
767 | /* 6.10.3 p 11: Directives in a list of macro arguments | |
768 | gives undefined behavior. This implementation | |
769 | handles the directive as normal. */ | |
770 | && pfile->state.parsing_args != 1) | |
771 | { | |
772 | if (_cpp_handle_directive (pfile, result->flags & PREV_WHITE)) | |
773 | { | |
774 | if (pfile->directive_result.type == CPP_PADDING) | |
775 | continue; | |
776 | result = &pfile->directive_result; | |
777 | } | |
778 | } | |
779 | else if (pfile->state.in_deferred_pragma) | |
780 | result = &pfile->directive_result; | |
781 | ||
782 | if (pfile->cb.line_change && !pfile->state.skipping) | |
783 | pfile->cb.line_change (pfile, result, pfile->state.parsing_args); | |
784 | } | |
785 | ||
786 | /* We don't skip tokens in directives. */ | |
787 | if (pfile->state.in_directive || pfile->state.in_deferred_pragma) | |
788 | break; | |
789 | ||
790 | /* Outside a directive, invalidate controlling macros. At file | |
791 | EOF, _cpp_lex_direct takes care of popping the buffer, so we never | |
792 | get here and MI optimization works. */ | |
793 | pfile->mi_valid = false; | |
794 | ||
795 | if (!pfile->state.skipping || result->type == CPP_EOF) | |
796 | break; | |
797 | } | |
798 | ||
799 | return result; | |
800 | } | |
801 | ||
802 | /* Returns true if a fresh line has been loaded. */ | |
803 | bool | |
804 | _cpp_get_fresh_line (cpp_reader *pfile) | |
805 | { | |
806 | int return_at_eof; | |
807 | ||
808 | /* We can't get a new line until we leave the current directive. */ | |
809 | if (pfile->state.in_directive) | |
810 | return false; | |
811 | ||
812 | for (;;) | |
813 | { | |
814 | cpp_buffer *buffer = pfile->buffer; | |
815 | ||
816 | if (!buffer->need_line) | |
817 | return true; | |
818 | ||
819 | if (buffer->next_line < buffer->rlimit) | |
820 | { | |
821 | _cpp_clean_line (pfile); | |
822 | return true; | |
823 | } | |
824 | ||
825 | /* First, get out of parsing arguments state. */ | |
826 | if (pfile->state.parsing_args) | |
827 | return false; | |
828 | ||
829 | /* End of buffer. Non-empty files should end in a newline. */ | |
830 | if (buffer->buf != buffer->rlimit | |
831 | && buffer->next_line > buffer->rlimit | |
832 | && !buffer->from_stage3) | |
833 | { | |
834 | /* Only warn once. */ | |
835 | buffer->next_line = buffer->rlimit; | |
836 | cpp_error_with_line (pfile, CPP_DL_PEDWARN, pfile->line_table->highest_line, | |
837 | CPP_BUF_COLUMN (buffer, buffer->cur), | |
838 | "no newline at end of file"); | |
839 | } | |
840 | ||
841 | return_at_eof = buffer->return_at_eof; | |
842 | _cpp_pop_buffer (pfile); | |
843 | if (pfile->buffer == NULL || return_at_eof) | |
844 | return false; | |
845 | } | |
846 | } | |
847 | ||
848 | #define IF_NEXT_IS(CHAR, THEN_TYPE, ELSE_TYPE) \ | |
849 | do \ | |
850 | { \ | |
851 | result->type = ELSE_TYPE; \ | |
852 | if (*buffer->cur == CHAR) \ | |
853 | buffer->cur++, result->type = THEN_TYPE; \ | |
854 | } \ | |
855 | while (0) | |
856 | ||
857 | /* Lex a token into pfile->cur_token, which is also incremented, to | |
858 | get diagnostics pointing to the correct location. | |
859 | ||
860 | Does not handle issues such as token lookahead, multiple-include | |
861 | optimization, directives, skipping etc. This function is only | |
862 | suitable for use by _cpp_lex_token, and in special cases like | |
863 | lex_expansion_token which doesn't care for any of these issues. | |
864 | ||
865 | When meeting a newline, returns CPP_EOF if parsing a directive, | |
866 | otherwise returns to the start of the token buffer if permissible. | |
867 | Returns the location of the lexed token. */ | |
868 | cpp_token * | |
869 | _cpp_lex_direct (cpp_reader *pfile) | |
870 | { | |
871 | cppchar_t c; | |
872 | cpp_buffer *buffer; | |
873 | const unsigned char *comment_start; | |
874 | cpp_token *result = pfile->cur_token++; | |
875 | ||
876 | fresh_line: | |
877 | result->flags = 0; | |
878 | buffer = pfile->buffer; | |
879 | if (buffer->need_line) | |
880 | { | |
881 | if (pfile->state.in_deferred_pragma) | |
882 | { | |
883 | result->type = CPP_PRAGMA_EOL; | |
884 | pfile->state.in_deferred_pragma = false; | |
885 | if (!pfile->state.pragma_allow_expansion) | |
886 | pfile->state.prevent_expansion--; | |
887 | return result; | |
888 | } | |
889 | if (!_cpp_get_fresh_line (pfile)) | |
890 | { | |
891 | result->type = CPP_EOF; | |
892 | if (!pfile->state.in_directive) | |
893 | { | |
894 | /* Tell the compiler the line number of the EOF token. */ | |
895 | result->src_loc = pfile->line_table->highest_line; | |
896 | result->flags = BOL; | |
897 | } | |
898 | return result; | |
899 | } | |
900 | if (!pfile->keep_tokens) | |
901 | { | |
902 | pfile->cur_run = &pfile->base_run; | |
903 | result = pfile->base_run.base; | |
904 | pfile->cur_token = result + 1; | |
905 | } | |
906 | result->flags = BOL; | |
907 | if (pfile->state.parsing_args == 2) | |
908 | result->flags |= PREV_WHITE; | |
909 | } | |
910 | buffer = pfile->buffer; | |
911 | update_tokens_line: | |
912 | result->src_loc = pfile->line_table->highest_line; | |
913 | ||
914 | skipped_white: | |
915 | if (buffer->cur >= buffer->notes[buffer->cur_note].pos | |
916 | && !pfile->overlaid_buffer) | |
917 | { | |
918 | _cpp_process_line_notes (pfile, false); | |
919 | result->src_loc = pfile->line_table->highest_line; | |
920 | } | |
921 | c = *buffer->cur++; | |
922 | ||
923 | LINEMAP_POSITION_FOR_COLUMN (result->src_loc, pfile->line_table, | |
924 | CPP_BUF_COLUMN (buffer, buffer->cur)); | |
925 | ||
926 | switch (c) | |
927 | { | |
928 | case ' ': case '\t': case '\f': case '\v': case '\0': | |
929 | result->flags |= PREV_WHITE; | |
930 | skip_whitespace (pfile, c); | |
931 | goto skipped_white; | |
932 | ||
933 | case '\n': | |
934 | if (buffer->cur < buffer->rlimit) | |
935 | CPP_INCREMENT_LINE (pfile, 0); | |
936 | buffer->need_line = true; | |
937 | goto fresh_line; | |
938 | ||
939 | case '0': case '1': case '2': case '3': case '4': | |
940 | case '5': case '6': case '7': case '8': case '9': | |
941 | { | |
942 | struct normalize_state nst = INITIAL_NORMALIZE_STATE; | |
943 | result->type = CPP_NUMBER; | |
944 | lex_number (pfile, &result->val.str, &nst); | |
945 | warn_about_normalization (pfile, result, &nst); | |
946 | break; | |
947 | } | |
948 | ||
949 | case 'L': | |
950 | /* 'L' may introduce wide characters or strings. */ | |
951 | if (*buffer->cur == '\'' || *buffer->cur == '"') | |
952 | { | |
953 | lex_string (pfile, result, buffer->cur - 1); | |
954 | break; | |
955 | } | |
956 | /* Fall through. */ | |
957 | ||
958 | case '_': | |
959 | case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': | |
960 | case 'g': case 'h': case 'i': case 'j': case 'k': case 'l': | |
961 | case 'm': case 'n': case 'o': case 'p': case 'q': case 'r': | |
962 | case 's': case 't': case 'u': case 'v': case 'w': case 'x': | |
963 | case 'y': case 'z': | |
964 | case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': | |
965 | case 'G': case 'H': case 'I': case 'J': case 'K': | |
966 | case 'M': case 'N': case 'O': case 'P': case 'Q': case 'R': | |
967 | case 'S': case 'T': case 'U': case 'V': case 'W': case 'X': | |
968 | case 'Y': case 'Z': | |
969 | result->type = CPP_NAME; | |
970 | { | |
971 | struct normalize_state nst = INITIAL_NORMALIZE_STATE; | |
972 | result->val.node = lex_identifier (pfile, buffer->cur - 1, false, | |
973 | &nst); | |
974 | warn_about_normalization (pfile, result, &nst); | |
975 | } | |
976 | ||
977 | /* Convert named operators to their proper types. */ | |
978 | if (result->val.node->flags & NODE_OPERATOR) | |
979 | { | |
980 | result->flags |= NAMED_OP; | |
981 | result->type = (enum cpp_ttype) result->val.node->directive_index; | |
982 | } | |
983 | break; | |
984 | ||
985 | case '\'': | |
986 | case '"': | |
987 | lex_string (pfile, result, buffer->cur - 1); | |
988 | break; | |
989 | ||
990 | case '/': | |
991 | /* A potential block or line comment. */ | |
992 | comment_start = buffer->cur; | |
993 | c = *buffer->cur; | |
994 | ||
995 | if (c == '*') | |
996 | { | |
997 | if (_cpp_skip_block_comment (pfile)) | |
998 | cpp_error (pfile, CPP_DL_ERROR, "unterminated comment"); | |
999 | } | |
1000 | else if (c == '/' && (CPP_OPTION (pfile, cplusplus_comments) | |
1001 | || cpp_in_system_header (pfile))) | |
1002 | { | |
1003 | /* Warn about comments only if pedantically GNUC89, and not | |
1004 | in system headers. */ | |
1005 | if (CPP_OPTION (pfile, lang) == CLK_GNUC89 && CPP_PEDANTIC (pfile) | |
1006 | && ! buffer->warned_cplusplus_comments) | |
1007 | { | |
1008 | cpp_error (pfile, CPP_DL_PEDWARN, | |
1009 | "C++ style comments are not allowed in ISO C90"); | |
1010 | cpp_error (pfile, CPP_DL_PEDWARN, | |
1011 | "(this will be reported only once per input file)"); | |
1012 | buffer->warned_cplusplus_comments = 1; | |
1013 | } | |
1014 | ||
1015 | if (skip_line_comment (pfile) && CPP_OPTION (pfile, warn_comments)) | |
1016 | cpp_error (pfile, CPP_DL_WARNING, "multi-line comment"); | |
1017 | } | |
1018 | else if (c == '=') | |
1019 | { | |
1020 | buffer->cur++; | |
1021 | result->type = CPP_DIV_EQ; | |
1022 | break; | |
1023 | } | |
1024 | else | |
1025 | { | |
1026 | result->type = CPP_DIV; | |
1027 | break; | |
1028 | } | |
1029 | ||
1030 | if (!pfile->state.save_comments) | |
1031 | { | |
1032 | result->flags |= PREV_WHITE; | |
1033 | goto update_tokens_line; | |
1034 | } | |
1035 | ||
1036 | /* Save the comment as a token in its own right. */ | |
1037 | save_comment (pfile, result, comment_start, c); | |
1038 | break; | |
1039 | ||
1040 | case '<': | |
1041 | if (pfile->state.angled_headers) | |
1042 | { | |
1043 | lex_string (pfile, result, buffer->cur - 1); | |
1044 | break; | |
1045 | } | |
1046 | ||
1047 | result->type = CPP_LESS; | |
1048 | if (*buffer->cur == '=') | |
1049 | buffer->cur++, result->type = CPP_LESS_EQ; | |
1050 | else if (*buffer->cur == '<') | |
1051 | { | |
1052 | buffer->cur++; | |
1053 | IF_NEXT_IS ('=', CPP_LSHIFT_EQ, CPP_LSHIFT); | |
1054 | } | |
1055 | else if (*buffer->cur == '?' && CPP_OPTION (pfile, cplusplus)) | |
1056 | { | |
1057 | buffer->cur++; | |
1058 | IF_NEXT_IS ('=', CPP_MIN_EQ, CPP_MIN); | |
1059 | } | |
1060 | else if (CPP_OPTION (pfile, digraphs)) | |
1061 | { | |
1062 | if (*buffer->cur == ':') | |
1063 | { | |
1064 | buffer->cur++; | |
1065 | result->flags |= DIGRAPH; | |
1066 | result->type = CPP_OPEN_SQUARE; | |
1067 | } | |
1068 | else if (*buffer->cur == '%') | |
1069 | { | |
1070 | buffer->cur++; | |
1071 | result->flags |= DIGRAPH; | |
1072 | result->type = CPP_OPEN_BRACE; | |
1073 | } | |
1074 | } | |
1075 | break; | |
1076 | ||
1077 | case '>': | |
1078 | result->type = CPP_GREATER; | |
1079 | if (*buffer->cur == '=') | |
1080 | buffer->cur++, result->type = CPP_GREATER_EQ; | |
1081 | else if (*buffer->cur == '>') | |
1082 | { | |
1083 | buffer->cur++; | |
1084 | IF_NEXT_IS ('=', CPP_RSHIFT_EQ, CPP_RSHIFT); | |
1085 | } | |
1086 | else if (*buffer->cur == '?' && CPP_OPTION (pfile, cplusplus)) | |
1087 | { | |
1088 | buffer->cur++; | |
1089 | IF_NEXT_IS ('=', CPP_MAX_EQ, CPP_MAX); | |
1090 | } | |
1091 | break; | |
1092 | ||
1093 | case '%': | |
1094 | result->type = CPP_MOD; | |
1095 | if (*buffer->cur == '=') | |
1096 | buffer->cur++, result->type = CPP_MOD_EQ; | |
1097 | else if (CPP_OPTION (pfile, digraphs)) | |
1098 | { | |
1099 | if (*buffer->cur == ':') | |
1100 | { | |
1101 | buffer->cur++; | |
1102 | result->flags |= DIGRAPH; | |
1103 | result->type = CPP_HASH; | |
1104 | if (*buffer->cur == '%' && buffer->cur[1] == ':') | |
1105 | buffer->cur += 2, result->type = CPP_PASTE; | |
1106 | } | |
1107 | else if (*buffer->cur == '>') | |
1108 | { | |
1109 | buffer->cur++; | |
1110 | result->flags |= DIGRAPH; | |
1111 | result->type = CPP_CLOSE_BRACE; | |
1112 | } | |
1113 | } | |
1114 | break; | |
1115 | ||
1116 | case '.': | |
1117 | result->type = CPP_DOT; | |
1118 | if (ISDIGIT (*buffer->cur)) | |
1119 | { | |
1120 | struct normalize_state nst = INITIAL_NORMALIZE_STATE; | |
1121 | result->type = CPP_NUMBER; | |
1122 | lex_number (pfile, &result->val.str, &nst); | |
1123 | warn_about_normalization (pfile, result, &nst); | |
1124 | } | |
1125 | else if (*buffer->cur == '.' && buffer->cur[1] == '.') | |
1126 | buffer->cur += 2, result->type = CPP_ELLIPSIS; | |
1127 | else if (*buffer->cur == '*' && CPP_OPTION (pfile, cplusplus)) | |
1128 | buffer->cur++, result->type = CPP_DOT_STAR; | |
1129 | break; | |
1130 | ||
1131 | case '+': | |
1132 | result->type = CPP_PLUS; | |
1133 | if (*buffer->cur == '+') | |
1134 | buffer->cur++, result->type = CPP_PLUS_PLUS; | |
1135 | else if (*buffer->cur == '=') | |
1136 | buffer->cur++, result->type = CPP_PLUS_EQ; | |
1137 | break; | |
1138 | ||
1139 | case '-': | |
1140 | result->type = CPP_MINUS; | |
1141 | if (*buffer->cur == '>') | |
1142 | { | |
1143 | buffer->cur++; | |
1144 | result->type = CPP_DEREF; | |
1145 | if (*buffer->cur == '*' && CPP_OPTION (pfile, cplusplus)) | |
1146 | buffer->cur++, result->type = CPP_DEREF_STAR; | |
1147 | } | |
1148 | else if (*buffer->cur == '-') | |
1149 | buffer->cur++, result->type = CPP_MINUS_MINUS; | |
1150 | else if (*buffer->cur == '=') | |
1151 | buffer->cur++, result->type = CPP_MINUS_EQ; | |
1152 | break; | |
1153 | ||
1154 | case '&': | |
1155 | result->type = CPP_AND; | |
1156 | if (*buffer->cur == '&') | |
1157 | buffer->cur++, result->type = CPP_AND_AND; | |
1158 | else if (*buffer->cur == '=') | |
1159 | buffer->cur++, result->type = CPP_AND_EQ; | |
1160 | break; | |
1161 | ||
1162 | case '|': | |
1163 | result->type = CPP_OR; | |
1164 | if (*buffer->cur == '|') | |
1165 | buffer->cur++, result->type = CPP_OR_OR; | |
1166 | else if (*buffer->cur == '=') | |
1167 | buffer->cur++, result->type = CPP_OR_EQ; | |
1168 | break; | |
1169 | ||
1170 | case ':': | |
1171 | result->type = CPP_COLON; | |
1172 | if (*buffer->cur == ':' && CPP_OPTION (pfile, cplusplus)) | |
1173 | buffer->cur++, result->type = CPP_SCOPE; | |
1174 | else if (*buffer->cur == '>' && CPP_OPTION (pfile, digraphs)) | |
1175 | { | |
1176 | buffer->cur++; | |
1177 | result->flags |= DIGRAPH; | |
1178 | result->type = CPP_CLOSE_SQUARE; | |
1179 | } | |
1180 | break; | |
1181 | ||
1182 | case '*': IF_NEXT_IS ('=', CPP_MULT_EQ, CPP_MULT); break; | |
1183 | case '=': IF_NEXT_IS ('=', CPP_EQ_EQ, CPP_EQ); break; | |
1184 | case '!': IF_NEXT_IS ('=', CPP_NOT_EQ, CPP_NOT); break; | |
1185 | case '^': IF_NEXT_IS ('=', CPP_XOR_EQ, CPP_XOR); break; | |
1186 | case '#': IF_NEXT_IS ('#', CPP_PASTE, CPP_HASH); break; | |
1187 | ||
1188 | case '?': result->type = CPP_QUERY; break; | |
1189 | case '~': result->type = CPP_COMPL; break; | |
1190 | case ',': result->type = CPP_COMMA; break; | |
1191 | case '(': result->type = CPP_OPEN_PAREN; break; | |
1192 | case ')': result->type = CPP_CLOSE_PAREN; break; | |
1193 | case '[': result->type = CPP_OPEN_SQUARE; break; | |
1194 | case ']': result->type = CPP_CLOSE_SQUARE; break; | |
1195 | case '{': result->type = CPP_OPEN_BRACE; break; | |
1196 | case '}': result->type = CPP_CLOSE_BRACE; break; | |
1197 | case ';': result->type = CPP_SEMICOLON; break; | |
1198 | ||
1199 | /* @ is a punctuator in Objective-C. */ | |
1200 | case '@': result->type = CPP_ATSIGN; break; | |
1201 | ||
1202 | case '$': | |
1203 | case '\\': | |
1204 | { | |
1205 | const uchar *base = --buffer->cur; | |
1206 | struct normalize_state nst = INITIAL_NORMALIZE_STATE; | |
1207 | ||
1208 | if (forms_identifier_p (pfile, true, &nst)) | |
1209 | { | |
1210 | result->type = CPP_NAME; | |
1211 | result->val.node = lex_identifier (pfile, base, true, &nst); | |
1212 | warn_about_normalization (pfile, result, &nst); | |
1213 | break; | |
1214 | } | |
1215 | buffer->cur++; | |
1216 | } | |
1217 | ||
1218 | default: | |
1219 | create_literal (pfile, result, buffer->cur - 1, 1, CPP_OTHER); | |
1220 | break; | |
1221 | } | |
1222 | ||
1223 | return result; | |
1224 | } | |
1225 | ||
1226 | /* An upper bound on the number of bytes needed to spell TOKEN. | |
1227 | Does not include preceding whitespace. */ | |
1228 | unsigned int | |
1229 | cpp_token_len (const cpp_token *token) | |
1230 | { | |
1231 | unsigned int len; | |
1232 | ||
1233 | switch (TOKEN_SPELL (token)) | |
1234 | { | |
1235 | default: len = 4; break; | |
1236 | case SPELL_LITERAL: len = token->val.str.len; break; | |
1237 | case SPELL_IDENT: len = NODE_LEN (token->val.node) * 10; break; | |
1238 | } | |
1239 | ||
1240 | return len; | |
1241 | } | |
1242 | ||
1243 | /* Parse UTF-8 out of NAMEP and place a \U escape in BUFFER. | |
1244 | Return the number of bytes read out of NAME. (There are always | |
1245 | 10 bytes written to BUFFER.) */ | |
1246 | ||
1247 | static size_t | |
1248 | utf8_to_ucn (unsigned char *buffer, const unsigned char *name) | |
1249 | { | |
1250 | int j; | |
1251 | int ucn_len = 0; | |
1252 | int ucn_len_c; | |
1253 | unsigned t; | |
1254 | unsigned long utf32; | |
1255 | ||
1256 | /* Compute the length of the UTF-8 sequence. */ | |
1257 | for (t = *name; t & 0x80; t <<= 1) | |
1258 | ucn_len++; | |
1259 | ||
1260 | utf32 = *name & (0x7F >> ucn_len); | |
1261 | for (ucn_len_c = 1; ucn_len_c < ucn_len; ucn_len_c++) | |
1262 | { | |
1263 | utf32 = (utf32 << 6) | (*++name & 0x3F); | |
1264 | ||
1265 | /* Ill-formed UTF-8. */ | |
1266 | if ((*name & ~0x3F) != 0x80) | |
1267 | abort (); | |
1268 | } | |
1269 | ||
1270 | *buffer++ = '\\'; | |
1271 | *buffer++ = 'U'; | |
1272 | for (j = 7; j >= 0; j--) | |
1273 | *buffer++ = "0123456789abcdef"[(utf32 >> (4 * j)) & 0xF]; | |
1274 | return ucn_len; | |
1275 | } | |
1276 | ||
1277 | ||
1278 | /* Write the spelling of a token TOKEN to BUFFER. The buffer must | |
1279 | already contain the enough space to hold the token's spelling. | |
1280 | Returns a pointer to the character after the last character written. | |
1281 | FORSTRING is true if this is to be the spelling after translation | |
1282 | phase 1 (this is different for UCNs). | |
1283 | FIXME: Would be nice if we didn't need the PFILE argument. */ | |
1284 | unsigned char * | |
1285 | cpp_spell_token (cpp_reader *pfile, const cpp_token *token, | |
1286 | unsigned char *buffer, bool forstring) | |
1287 | { | |
1288 | switch (TOKEN_SPELL (token)) | |
1289 | { | |
1290 | case SPELL_OPERATOR: | |
1291 | { | |
1292 | const unsigned char *spelling; | |
1293 | unsigned char c; | |
1294 | ||
1295 | if (token->flags & DIGRAPH) | |
1296 | spelling | |
1297 | = digraph_spellings[(int) token->type - (int) CPP_FIRST_DIGRAPH]; | |
1298 | else if (token->flags & NAMED_OP) | |
1299 | goto spell_ident; | |
1300 | else | |
1301 | spelling = TOKEN_NAME (token); | |
1302 | ||
1303 | while ((c = *spelling++) != '\0') | |
1304 | *buffer++ = c; | |
1305 | } | |
1306 | break; | |
1307 | ||
1308 | spell_ident: | |
1309 | case SPELL_IDENT: | |
1310 | if (forstring) | |
1311 | { | |
1312 | memcpy (buffer, NODE_NAME (token->val.node), | |
1313 | NODE_LEN (token->val.node)); | |
1314 | buffer += NODE_LEN (token->val.node); | |
1315 | } | |
1316 | else | |
1317 | { | |
1318 | size_t i; | |
1319 | const unsigned char * name = NODE_NAME (token->val.node); | |
1320 | ||
1321 | for (i = 0; i < NODE_LEN (token->val.node); i++) | |
1322 | if (name[i] & ~0x7F) | |
1323 | { | |
1324 | i += utf8_to_ucn (buffer, name + i) - 1; | |
1325 | buffer += 10; | |
1326 | } | |
1327 | else | |
1328 | *buffer++ = NODE_NAME (token->val.node)[i]; | |
1329 | } | |
1330 | break; | |
1331 | ||
1332 | case SPELL_LITERAL: | |
1333 | memcpy (buffer, token->val.str.text, token->val.str.len); | |
1334 | buffer += token->val.str.len; | |
1335 | break; | |
1336 | ||
1337 | case SPELL_NONE: | |
1338 | cpp_error (pfile, CPP_DL_ICE, | |
1339 | "unspellable token %s", TOKEN_NAME (token)); | |
1340 | break; | |
1341 | } | |
1342 | ||
1343 | return buffer; | |
1344 | } | |
1345 | ||
1346 | /* Returns TOKEN spelt as a null-terminated string. The string is | |
1347 | freed when the reader is destroyed. Useful for diagnostics. */ | |
1348 | unsigned char * | |
1349 | cpp_token_as_text (cpp_reader *pfile, const cpp_token *token) | |
1350 | { | |
1351 | unsigned int len = cpp_token_len (token) + 1; | |
1352 | unsigned char *start = _cpp_unaligned_alloc (pfile, len), *end; | |
1353 | ||
1354 | end = cpp_spell_token (pfile, token, start, false); | |
1355 | end[0] = '\0'; | |
1356 | ||
1357 | return start; | |
1358 | } | |
1359 | ||
1360 | /* Used by C front ends, which really should move to using | |
1361 | cpp_token_as_text. */ | |
1362 | const char * | |
1363 | cpp_type2name (enum cpp_ttype type) | |
1364 | { | |
1365 | return (const char *) token_spellings[type].name; | |
1366 | } | |
1367 | ||
1368 | /* Writes the spelling of token to FP, without any preceding space. | |
1369 | Separated from cpp_spell_token for efficiency - to avoid stdio | |
1370 | double-buffering. */ | |
1371 | void | |
1372 | cpp_output_token (const cpp_token *token, FILE *fp) | |
1373 | { | |
1374 | switch (TOKEN_SPELL (token)) | |
1375 | { | |
1376 | case SPELL_OPERATOR: | |
1377 | { | |
1378 | const unsigned char *spelling; | |
1379 | int c; | |
1380 | ||
1381 | if (token->flags & DIGRAPH) | |
1382 | spelling | |
1383 | = digraph_spellings[(int) token->type - (int) CPP_FIRST_DIGRAPH]; | |
1384 | else if (token->flags & NAMED_OP) | |
1385 | goto spell_ident; | |
1386 | else | |
1387 | spelling = TOKEN_NAME (token); | |
1388 | ||
1389 | c = *spelling; | |
1390 | do | |
1391 | putc (c, fp); | |
1392 | while ((c = *++spelling) != '\0'); | |
1393 | } | |
1394 | break; | |
1395 | ||
1396 | spell_ident: | |
1397 | case SPELL_IDENT: | |
1398 | { | |
1399 | size_t i; | |
1400 | const unsigned char * name = NODE_NAME (token->val.node); | |
1401 | ||
1402 | for (i = 0; i < NODE_LEN (token->val.node); i++) | |
1403 | if (name[i] & ~0x7F) | |
1404 | { | |
1405 | unsigned char buffer[10]; | |
1406 | i += utf8_to_ucn (buffer, name + i) - 1; | |
1407 | fwrite (buffer, 1, 10, fp); | |
1408 | } | |
1409 | else | |
1410 | fputc (NODE_NAME (token->val.node)[i], fp); | |
1411 | } | |
1412 | break; | |
1413 | ||
1414 | case SPELL_LITERAL: | |
1415 | fwrite (token->val.str.text, 1, token->val.str.len, fp); | |
1416 | break; | |
1417 | ||
1418 | case SPELL_NONE: | |
1419 | /* An error, most probably. */ | |
1420 | break; | |
1421 | } | |
1422 | } | |
1423 | ||
1424 | /* Compare two tokens. */ | |
1425 | int | |
1426 | _cpp_equiv_tokens (const cpp_token *a, const cpp_token *b) | |
1427 | { | |
1428 | if (a->type == b->type && a->flags == b->flags) | |
1429 | switch (TOKEN_SPELL (a)) | |
1430 | { | |
1431 | default: /* Keep compiler happy. */ | |
1432 | case SPELL_OPERATOR: | |
1433 | return 1; | |
1434 | case SPELL_NONE: | |
1435 | return (a->type != CPP_MACRO_ARG || a->val.arg_no == b->val.arg_no); | |
1436 | case SPELL_IDENT: | |
1437 | return a->val.node == b->val.node; | |
1438 | case SPELL_LITERAL: | |
1439 | return (a->val.str.len == b->val.str.len | |
1440 | && !memcmp (a->val.str.text, b->val.str.text, | |
1441 | a->val.str.len)); | |
1442 | } | |
1443 | ||
1444 | return 0; | |
1445 | } | |
1446 | ||
1447 | /* Returns nonzero if a space should be inserted to avoid an | |
1448 | accidental token paste for output. For simplicity, it is | |
1449 | conservative, and occasionally advises a space where one is not | |
1450 | needed, e.g. "." and ".2". */ | |
1451 | int | |
1452 | cpp_avoid_paste (cpp_reader *pfile, const cpp_token *token1, | |
1453 | const cpp_token *token2) | |
1454 | { | |
1455 | enum cpp_ttype a = token1->type, b = token2->type; | |
1456 | cppchar_t c; | |
1457 | ||
1458 | if (token1->flags & NAMED_OP) | |
1459 | a = CPP_NAME; | |
1460 | if (token2->flags & NAMED_OP) | |
1461 | b = CPP_NAME; | |
1462 | ||
1463 | c = EOF; | |
1464 | if (token2->flags & DIGRAPH) | |
1465 | c = digraph_spellings[(int) b - (int) CPP_FIRST_DIGRAPH][0]; | |
1466 | else if (token_spellings[b].category == SPELL_OPERATOR) | |
1467 | c = token_spellings[b].name[0]; | |
1468 | ||
1469 | /* Quickly get everything that can paste with an '='. */ | |
1470 | if ((int) a <= (int) CPP_LAST_EQ && c == '=') | |
1471 | return 1; | |
1472 | ||
1473 | switch (a) | |
1474 | { | |
1475 | case CPP_GREATER: return c == '>' || c == '?'; | |
1476 | case CPP_LESS: return c == '<' || c == '?' || c == '%' || c == ':'; | |
1477 | case CPP_PLUS: return c == '+'; | |
1478 | case CPP_MINUS: return c == '-' || c == '>'; | |
1479 | case CPP_DIV: return c == '/' || c == '*'; /* Comments. */ | |
1480 | case CPP_MOD: return c == ':' || c == '>'; | |
1481 | case CPP_AND: return c == '&'; | |
1482 | case CPP_OR: return c == '|'; | |
1483 | case CPP_COLON: return c == ':' || c == '>'; | |
1484 | case CPP_DEREF: return c == '*'; | |
1485 | case CPP_DOT: return c == '.' || c == '%' || b == CPP_NUMBER; | |
1486 | case CPP_HASH: return c == '#' || c == '%'; /* Digraph form. */ | |
1487 | case CPP_NAME: return ((b == CPP_NUMBER | |
1488 | && name_p (pfile, &token2->val.str)) | |
1489 | || b == CPP_NAME | |
1490 | || b == CPP_CHAR || b == CPP_STRING); /* L */ | |
1491 | case CPP_NUMBER: return (b == CPP_NUMBER || b == CPP_NAME | |
1492 | || c == '.' || c == '+' || c == '-'); | |
1493 | /* UCNs */ | |
1494 | case CPP_OTHER: return ((token1->val.str.text[0] == '\\' | |
1495 | && b == CPP_NAME) | |
1496 | || (CPP_OPTION (pfile, objc) | |
1497 | && token1->val.str.text[0] == '@' | |
1498 | && (b == CPP_NAME || b == CPP_STRING))); | |
1499 | default: break; | |
1500 | } | |
1501 | ||
1502 | return 0; | |
1503 | } | |
1504 | ||
1505 | /* Output all the remaining tokens on the current line, and a newline | |
1506 | character, to FP. Leading whitespace is removed. If there are | |
1507 | macros, special token padding is not performed. */ | |
1508 | void | |
1509 | cpp_output_line (cpp_reader *pfile, FILE *fp) | |
1510 | { | |
1511 | const cpp_token *token; | |
1512 | ||
1513 | token = cpp_get_token (pfile); | |
1514 | while (token->type != CPP_EOF) | |
1515 | { | |
1516 | cpp_output_token (token, fp); | |
1517 | token = cpp_get_token (pfile); | |
1518 | if (token->flags & PREV_WHITE) | |
1519 | putc (' ', fp); | |
1520 | } | |
1521 | ||
1522 | putc ('\n', fp); | |
1523 | } | |
1524 | ||
1525 | /* Memory buffers. Changing these three constants can have a dramatic | |
1526 | effect on performance. The values here are reasonable defaults, | |
1527 | but might be tuned. If you adjust them, be sure to test across a | |
1528 | range of uses of cpplib, including heavy nested function-like macro | |
1529 | expansion. Also check the change in peak memory usage (NJAMD is a | |
1530 | good tool for this). */ | |
1531 | #define MIN_BUFF_SIZE 8000 | |
1532 | #define BUFF_SIZE_UPPER_BOUND(MIN_SIZE) (MIN_BUFF_SIZE + (MIN_SIZE) * 3 / 2) | |
1533 | #define EXTENDED_BUFF_SIZE(BUFF, MIN_EXTRA) \ | |
1534 | (MIN_EXTRA + ((BUFF)->limit - (BUFF)->cur) * 2) | |
1535 | ||
1536 | #if MIN_BUFF_SIZE > BUFF_SIZE_UPPER_BOUND (0) | |
1537 | #error BUFF_SIZE_UPPER_BOUND must be at least as large as MIN_BUFF_SIZE! | |
1538 | #endif | |
1539 | ||
1540 | /* Create a new allocation buffer. Place the control block at the end | |
1541 | of the buffer, so that buffer overflows will cause immediate chaos. */ | |
1542 | static _cpp_buff * | |
1543 | new_buff (size_t len) | |
1544 | { | |
1545 | _cpp_buff *result; | |
1546 | unsigned char *base; | |
1547 | ||
1548 | if (len < MIN_BUFF_SIZE) | |
1549 | len = MIN_BUFF_SIZE; | |
1550 | len = CPP_ALIGN (len); | |
1551 | ||
1552 | base = XNEWVEC (unsigned char, len + sizeof (_cpp_buff)); | |
1553 | result = (_cpp_buff *) (base + len); | |
1554 | result->base = base; | |
1555 | result->cur = base; | |
1556 | result->limit = base + len; | |
1557 | result->next = NULL; | |
1558 | return result; | |
1559 | } | |
1560 | ||
1561 | /* Place a chain of unwanted allocation buffers on the free list. */ | |
1562 | void | |
1563 | _cpp_release_buff (cpp_reader *pfile, _cpp_buff *buff) | |
1564 | { | |
1565 | _cpp_buff *end = buff; | |
1566 | ||
1567 | while (end->next) | |
1568 | end = end->next; | |
1569 | end->next = pfile->free_buffs; | |
1570 | pfile->free_buffs = buff; | |
1571 | } | |
1572 | ||
1573 | /* Return a free buffer of size at least MIN_SIZE. */ | |
1574 | _cpp_buff * | |
1575 | _cpp_get_buff (cpp_reader *pfile, size_t min_size) | |
1576 | { | |
1577 | _cpp_buff *result, **p; | |
1578 | ||
1579 | for (p = &pfile->free_buffs;; p = &(*p)->next) | |
1580 | { | |
1581 | size_t size; | |
1582 | ||
1583 | if (*p == NULL) | |
1584 | return new_buff (min_size); | |
1585 | result = *p; | |
1586 | size = result->limit - result->base; | |
1587 | /* Return a buffer that's big enough, but don't waste one that's | |
1588 | way too big. */ | |
1589 | if (size >= min_size && size <= BUFF_SIZE_UPPER_BOUND (min_size)) | |
1590 | break; | |
1591 | } | |
1592 | ||
1593 | *p = result->next; | |
1594 | result->next = NULL; | |
1595 | result->cur = result->base; | |
1596 | return result; | |
1597 | } | |
1598 | ||
1599 | /* Creates a new buffer with enough space to hold the uncommitted | |
1600 | remaining bytes of BUFF, and at least MIN_EXTRA more bytes. Copies | |
1601 | the excess bytes to the new buffer. Chains the new buffer after | |
1602 | BUFF, and returns the new buffer. */ | |
1603 | _cpp_buff * | |
1604 | _cpp_append_extend_buff (cpp_reader *pfile, _cpp_buff *buff, size_t min_extra) | |
1605 | { | |
1606 | size_t size = EXTENDED_BUFF_SIZE (buff, min_extra); | |
1607 | _cpp_buff *new_buff = _cpp_get_buff (pfile, size); | |
1608 | ||
1609 | buff->next = new_buff; | |
1610 | memcpy (new_buff->base, buff->cur, BUFF_ROOM (buff)); | |
1611 | return new_buff; | |
1612 | } | |
1613 | ||
1614 | /* Creates a new buffer with enough space to hold the uncommitted | |
1615 | remaining bytes of the buffer pointed to by BUFF, and at least | |
1616 | MIN_EXTRA more bytes. Copies the excess bytes to the new buffer. | |
1617 | Chains the new buffer before the buffer pointed to by BUFF, and | |
1618 | updates the pointer to point to the new buffer. */ | |
1619 | void | |
1620 | _cpp_extend_buff (cpp_reader *pfile, _cpp_buff **pbuff, size_t min_extra) | |
1621 | { | |
1622 | _cpp_buff *new_buff, *old_buff = *pbuff; | |
1623 | size_t size = EXTENDED_BUFF_SIZE (old_buff, min_extra); | |
1624 | ||
1625 | new_buff = _cpp_get_buff (pfile, size); | |
1626 | memcpy (new_buff->base, old_buff->cur, BUFF_ROOM (old_buff)); | |
1627 | new_buff->next = old_buff; | |
1628 | *pbuff = new_buff; | |
1629 | } | |
1630 | ||
1631 | /* Free a chain of buffers starting at BUFF. */ | |
1632 | void | |
1633 | _cpp_free_buff (_cpp_buff *buff) | |
1634 | { | |
1635 | _cpp_buff *next; | |
1636 | ||
1637 | for (; buff; buff = next) | |
1638 | { | |
1639 | next = buff->next; | |
1640 | free (buff->base); | |
1641 | } | |
1642 | } | |
1643 | ||
1644 | /* Allocate permanent, unaligned storage of length LEN. */ | |
1645 | unsigned char * | |
1646 | _cpp_unaligned_alloc (cpp_reader *pfile, size_t len) | |
1647 | { | |
1648 | _cpp_buff *buff = pfile->u_buff; | |
1649 | unsigned char *result = buff->cur; | |
1650 | ||
1651 | if (len > (size_t) (buff->limit - result)) | |
1652 | { | |
1653 | buff = _cpp_get_buff (pfile, len); | |
1654 | buff->next = pfile->u_buff; | |
1655 | pfile->u_buff = buff; | |
1656 | result = buff->cur; | |
1657 | } | |
1658 | ||
1659 | buff->cur = result + len; | |
1660 | return result; | |
1661 | } | |
1662 | ||
1663 | /* Allocate permanent, unaligned storage of length LEN from a_buff. | |
1664 | That buffer is used for growing allocations when saving macro | |
1665 | replacement lists in a #define, and when parsing an answer to an | |
1666 | assertion in #assert, #unassert or #if (and therefore possibly | |
1667 | whilst expanding macros). It therefore must not be used by any | |
1668 | code that they might call: specifically the lexer and the guts of | |
1669 | the macro expander. | |
1670 | ||
1671 | All existing other uses clearly fit this restriction: storing | |
1672 | registered pragmas during initialization. */ | |
1673 | unsigned char * | |
1674 | _cpp_aligned_alloc (cpp_reader *pfile, size_t len) | |
1675 | { | |
1676 | _cpp_buff *buff = pfile->a_buff; | |
1677 | unsigned char *result = buff->cur; | |
1678 | ||
1679 | if (len > (size_t) (buff->limit - result)) | |
1680 | { | |
1681 | buff = _cpp_get_buff (pfile, len); | |
1682 | buff->next = pfile->a_buff; | |
1683 | pfile->a_buff = buff; | |
1684 | result = buff->cur; | |
1685 | } | |
1686 | ||
1687 | buff->cur = result + len; | |
1688 | return result; | |
1689 | } | |
1690 | ||
1691 | /* Say which field of TOK is in use. */ | |
1692 | ||
1693 | enum cpp_token_fld_kind | |
1694 | cpp_token_val_index (cpp_token *tok) | |
1695 | { | |
1696 | switch (TOKEN_SPELL (tok)) | |
1697 | { | |
1698 | case SPELL_IDENT: | |
1699 | return CPP_TOKEN_FLD_NODE; | |
1700 | case SPELL_LITERAL: | |
1701 | return CPP_TOKEN_FLD_STR; | |
1702 | case SPELL_NONE: | |
1703 | if (tok->type == CPP_MACRO_ARG) | |
1704 | return CPP_TOKEN_FLD_ARG_NO; | |
1705 | else if (tok->type == CPP_PADDING) | |
1706 | return CPP_TOKEN_FLD_SOURCE; | |
1707 | else if (tok->type == CPP_PRAGMA) | |
1708 | return CPP_TOKEN_FLD_PRAGMA; | |
1709 | /* else fall through */ | |
1710 | default: | |
1711 | return CPP_TOKEN_FLD_NONE; | |
1712 | } | |
1713 | } |