]>
Commit | Line | Data |
---|---|---|
45b966db ZW |
1 | /* CPP Library - lexical analysis. |
2 | Copyright (C) 2000 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 | |
c5a04734 | 7 | Single-pass line tokenization by Neil Booth, April 2000 |
45b966db ZW |
8 | |
9 | This program is free software; you can redistribute it and/or modify it | |
10 | under the terms of the GNU General Public License as published by the | |
11 | Free Software Foundation; either version 2, or (at your option) any | |
12 | later version. | |
13 | ||
14 | This program is distributed in the hope that it will be useful, | |
15 | but WITHOUT ANY WARRANTY; without even the implied warranty of | |
16 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
17 | GNU General Public License for more details. | |
18 | ||
19 | You should have received a copy of the GNU General Public License | |
20 | along with this program; if not, write to the Free Software | |
21 | Foundation, 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. */ | |
22 | ||
93c80368 NB |
23 | /* This lexer works with a single pass of the file. Recently I |
24 | re-wrote it to minimize the places where we step backwards in the | |
25 | input stream, to make future changes to support multi-byte | |
26 | character sets fairly straight-forward. | |
27 | ||
28 | There is now only one routine where we do step backwards: | |
29 | skip_escaped_newlines. This routine could probably also be changed | |
30 | so that it doesn't need to step back. One possibility is to use a | |
31 | trick similar to that used in lex_period and lex_percent. Two | |
32 | extra characters might be needed, but skip_escaped_newlines itself | |
33 | would probably be the only place that needs to be aware of that, | |
34 | and changes to the remaining routines would probably only be needed | |
35 | if they process a backslash. */ | |
041c3194 | 36 | |
45b966db ZW |
37 | #include "config.h" |
38 | #include "system.h" | |
45b966db ZW |
39 | #include "cpplib.h" |
40 | #include "cpphash.h" | |
41 | ||
c8a96070 NB |
42 | /* MULTIBYTE_CHARS support only works for native compilers. |
43 | ??? Ideally what we want is to model widechar support after | |
44 | the current floating point support. */ | |
45 | #ifdef CROSS_COMPILE | |
46 | #undef MULTIBYTE_CHARS | |
47 | #endif | |
48 | ||
49 | #ifdef MULTIBYTE_CHARS | |
50 | #include "mbchar.h" | |
51 | #include <locale.h> | |
52 | #endif | |
53 | ||
93c80368 NB |
54 | /* Tokens with SPELL_STRING store their spelling in the token list, |
55 | and it's length in the token->val.name.len. */ | |
56 | enum spell_type | |
f9a0e96c | 57 | { |
93c80368 NB |
58 | SPELL_OPERATOR = 0, |
59 | SPELL_CHAR, | |
60 | SPELL_IDENT, | |
47ad4138 | 61 | SPELL_NUMBER, |
93c80368 NB |
62 | SPELL_STRING, |
63 | SPELL_NONE | |
f9a0e96c ZW |
64 | }; |
65 | ||
93c80368 | 66 | struct token_spelling |
f9a0e96c | 67 | { |
93c80368 NB |
68 | enum spell_type category; |
69 | const unsigned char *name; | |
f9a0e96c ZW |
70 | }; |
71 | ||
8206c799 ZW |
72 | static const unsigned char *const digraph_spellings[] = |
73 | { U"%:", U"%:%:", U"<:", U":>", U"<%", U"%>" }; | |
93c80368 NB |
74 | |
75 | #define OP(e, s) { SPELL_OPERATOR, U s }, | |
76 | #define TK(e, s) { s, U STRINGX (e) }, | |
8206c799 | 77 | static const struct token_spelling token_spellings[N_TTYPES] = { TTYPE_TABLE }; |
93c80368 NB |
78 | #undef OP |
79 | #undef TK | |
80 | ||
81 | #define TOKEN_SPELL(token) (token_spellings[(token)->type].category) | |
82 | #define TOKEN_NAME(token) (token_spellings[(token)->type].name) | |
f2d5f0cc | 83 | |
1444f2ed | 84 | static cppchar_t handle_newline PARAMS ((cpp_reader *, cppchar_t)); |
29401c30 NB |
85 | static cppchar_t skip_escaped_newlines PARAMS ((cpp_reader *, cppchar_t)); |
86 | static cppchar_t get_effective_char PARAMS ((cpp_reader *)); | |
0d9f234d | 87 | |
041c3194 | 88 | static int skip_block_comment PARAMS ((cpp_reader *)); |
cbcff6df | 89 | static int skip_line_comment PARAMS ((cpp_reader *)); |
0d9f234d NB |
90 | static void adjust_column PARAMS ((cpp_reader *)); |
91 | static void skip_whitespace PARAMS ((cpp_reader *, cppchar_t)); | |
2c3fcba6 ZW |
92 | static cpp_hashnode *parse_identifier PARAMS ((cpp_reader *)); |
93 | static cpp_hashnode *parse_identifier_slow PARAMS ((cpp_reader *, | |
94 | const U_CHAR *)); | |
93c80368 NB |
95 | static void parse_number PARAMS ((cpp_reader *, cpp_string *, cppchar_t, int)); |
96 | static int unescaped_terminator_p PARAMS ((cpp_reader *, const U_CHAR *)); | |
0d9f234d | 97 | static void parse_string PARAMS ((cpp_reader *, cpp_token *, cppchar_t)); |
93c80368 | 98 | static void unterminated PARAMS ((cpp_reader *, int)); |
0d9f234d NB |
99 | static int trigraph_ok PARAMS ((cpp_reader *, cppchar_t)); |
100 | static void save_comment PARAMS ((cpp_reader *, cpp_token *, const U_CHAR *)); | |
29401c30 | 101 | static void lex_percent PARAMS ((cpp_reader *, cpp_token *)); |
cbcff6df | 102 | static void lex_dot PARAMS ((cpp_reader *, cpp_token *)); |
93c80368 | 103 | static int name_p PARAMS ((cpp_reader *, const cpp_string *)); |
62729350 NB |
104 | static int maybe_read_ucs PARAMS ((cpp_reader *, const unsigned char **, |
105 | const unsigned char *, unsigned int *)); | |
5fddcffc | 106 | static tokenrun *next_tokenrun PARAMS ((tokenrun *)); |
f617b8e2 | 107 | |
c8a96070 | 108 | static unsigned int hex_digit_value PARAMS ((unsigned int)); |
6142088c | 109 | static _cpp_buff *new_buff PARAMS ((size_t)); |
15dad1d9 | 110 | |
041c3194 | 111 | /* Utility routine: |
9e62c811 | 112 | |
bfb9dc7f ZW |
113 | Compares, the token TOKEN to the NUL-terminated string STRING. |
114 | TOKEN must be a CPP_NAME. Returns 1 for equal, 0 for unequal. */ | |
15dad1d9 | 115 | |
041c3194 | 116 | int |
bfb9dc7f ZW |
117 | cpp_ideq (token, string) |
118 | const cpp_token *token; | |
041c3194 ZW |
119 | const char *string; |
120 | { | |
bfb9dc7f | 121 | if (token->type != CPP_NAME) |
041c3194 | 122 | return 0; |
bfb9dc7f | 123 | |
a28c5035 | 124 | return !ustrcmp (NODE_NAME (token->val.node), (const U_CHAR *) string); |
15dad1d9 | 125 | } |
1368ee70 | 126 | |
0d9f234d NB |
127 | /* Call when meeting a newline. Returns the character after the newline |
128 | (or carriage-return newline combination), or EOF. */ | |
129 | static cppchar_t | |
1444f2ed NB |
130 | handle_newline (pfile, newline_char) |
131 | cpp_reader *pfile; | |
0d9f234d NB |
132 | cppchar_t newline_char; |
133 | { | |
1444f2ed | 134 | cpp_buffer *buffer; |
0d9f234d NB |
135 | cppchar_t next = EOF; |
136 | ||
1444f2ed | 137 | pfile->line++; |
1444f2ed | 138 | buffer = pfile->buffer; |
0d9f234d | 139 | buffer->col_adjust = 0; |
0d9f234d NB |
140 | buffer->line_base = buffer->cur; |
141 | ||
142 | /* Handle CR-LF and LF-CR combinations, get the next character. */ | |
143 | if (buffer->cur < buffer->rlimit) | |
144 | { | |
145 | next = *buffer->cur++; | |
146 | if (next + newline_char == '\r' + '\n') | |
147 | { | |
148 | buffer->line_base = buffer->cur; | |
149 | if (buffer->cur < buffer->rlimit) | |
150 | next = *buffer->cur++; | |
151 | else | |
152 | next = EOF; | |
153 | } | |
154 | } | |
155 | ||
156 | buffer->read_ahead = next; | |
157 | return next; | |
158 | } | |
159 | ||
160 | /* Subroutine of skip_escaped_newlines; called when a trigraph is | |
161 | encountered. It warns if necessary, and returns true if the | |
162 | trigraph should be honoured. FROM_CHAR is the third character of a | |
163 | trigraph, and presumed to be the previous character for position | |
164 | reporting. */ | |
45b966db | 165 | static int |
0d9f234d | 166 | trigraph_ok (pfile, from_char) |
45b966db | 167 | cpp_reader *pfile; |
0d9f234d | 168 | cppchar_t from_char; |
45b966db | 169 | { |
041c3194 ZW |
170 | int accept = CPP_OPTION (pfile, trigraphs); |
171 | ||
cbcff6df NB |
172 | /* Don't warn about trigraphs in comments. */ |
173 | if (CPP_OPTION (pfile, warn_trigraphs) && !pfile->state.lexing_comment) | |
45b966db | 174 | { |
0d9f234d | 175 | cpp_buffer *buffer = pfile->buffer; |
67821e3a | 176 | |
041c3194 | 177 | if (accept) |
67821e3a | 178 | cpp_warning_with_line (pfile, pfile->line, CPP_BUF_COL (buffer) - 2, |
041c3194 | 179 | "trigraph ??%c converted to %c", |
0d9f234d NB |
180 | (int) from_char, |
181 | (int) _cpp_trigraph_map[from_char]); | |
4a5b68a2 NB |
182 | else if (buffer->cur != buffer->last_Wtrigraphs) |
183 | { | |
184 | buffer->last_Wtrigraphs = buffer->cur; | |
67821e3a | 185 | cpp_warning_with_line (pfile, pfile->line, |
4a5b68a2 NB |
186 | CPP_BUF_COL (buffer) - 2, |
187 | "trigraph ??%c ignored", (int) from_char); | |
188 | } | |
45b966db | 189 | } |
0d9f234d | 190 | |
041c3194 | 191 | return accept; |
45b966db ZW |
192 | } |
193 | ||
0d9f234d NB |
194 | /* Assumes local variables buffer and result. */ |
195 | #define ACCEPT_CHAR(t) \ | |
196 | do { result->type = t; buffer->read_ahead = EOF; } while (0) | |
197 | ||
198 | /* When we move to multibyte character sets, add to these something | |
199 | that saves and restores the state of the multibyte conversion | |
200 | library. This probably involves saving and restoring a "cookie". | |
201 | In the case of glibc it is an 8-byte structure, so is not a high | |
202 | overhead operation. In any case, it's out of the fast path. */ | |
203 | #define SAVE_STATE() do { saved_cur = buffer->cur; } while (0) | |
204 | #define RESTORE_STATE() do { buffer->cur = saved_cur; } while (0) | |
205 | ||
206 | /* Skips any escaped newlines introduced by NEXT, which is either a | |
207 | '?' or a '\\'. Returns the next character, which will also have | |
a5c3cccd NB |
208 | been placed in buffer->read_ahead. This routine performs |
209 | preprocessing stages 1 and 2 of the ISO C standard. */ | |
0d9f234d | 210 | static cppchar_t |
29401c30 NB |
211 | skip_escaped_newlines (pfile, next) |
212 | cpp_reader *pfile; | |
0d9f234d | 213 | cppchar_t next; |
45b966db | 214 | { |
29401c30 NB |
215 | cpp_buffer *buffer = pfile->buffer; |
216 | ||
a5c3cccd NB |
217 | /* Only do this if we apply stages 1 and 2. */ |
218 | if (!buffer->from_stage3) | |
041c3194 | 219 | { |
a5c3cccd NB |
220 | cppchar_t next1; |
221 | const unsigned char *saved_cur; | |
222 | int space; | |
223 | ||
224 | do | |
0d9f234d | 225 | { |
a5c3cccd NB |
226 | if (buffer->cur == buffer->rlimit) |
227 | break; | |
228 | ||
229 | SAVE_STATE (); | |
230 | if (next == '?') | |
0d9f234d | 231 | { |
a5c3cccd NB |
232 | next1 = *buffer->cur++; |
233 | if (next1 != '?' || buffer->cur == buffer->rlimit) | |
234 | { | |
235 | RESTORE_STATE (); | |
236 | break; | |
237 | } | |
238 | ||
239 | next1 = *buffer->cur++; | |
240 | if (!_cpp_trigraph_map[next1] | |
29401c30 | 241 | || !trigraph_ok (pfile, next1)) |
a5c3cccd NB |
242 | { |
243 | RESTORE_STATE (); | |
244 | break; | |
245 | } | |
246 | ||
247 | /* We have a full trigraph here. */ | |
248 | next = _cpp_trigraph_map[next1]; | |
249 | if (next != '\\' || buffer->cur == buffer->rlimit) | |
250 | break; | |
251 | SAVE_STATE (); | |
252 | } | |
253 | ||
254 | /* We have a backslash, and room for at least one more character. */ | |
255 | space = 0; | |
256 | do | |
257 | { | |
258 | next1 = *buffer->cur++; | |
259 | if (!is_nvspace (next1)) | |
260 | break; | |
261 | space = 1; | |
0d9f234d | 262 | } |
a5c3cccd | 263 | while (buffer->cur < buffer->rlimit); |
041c3194 | 264 | |
a5c3cccd | 265 | if (!is_vspace (next1)) |
0d9f234d NB |
266 | { |
267 | RESTORE_STATE (); | |
268 | break; | |
269 | } | |
45b966db | 270 | |
29401c30 NB |
271 | if (space && !pfile->state.lexing_comment) |
272 | cpp_warning (pfile, "backslash and newline separated by space"); | |
0d9f234d | 273 | |
29401c30 | 274 | next = handle_newline (pfile, next1); |
a5c3cccd | 275 | if (next == EOF) |
29401c30 | 276 | cpp_pedwarn (pfile, "backslash-newline at end of file"); |
0d9f234d | 277 | } |
a5c3cccd | 278 | while (next == '\\' || next == '?'); |
041c3194 | 279 | } |
45b966db | 280 | |
0d9f234d NB |
281 | buffer->read_ahead = next; |
282 | return next; | |
45b966db ZW |
283 | } |
284 | ||
0d9f234d NB |
285 | /* Obtain the next character, after trigraph conversion and skipping |
286 | an arbitrary string of escaped newlines. The common case of no | |
287 | trigraphs or escaped newlines falls through quickly. */ | |
288 | static cppchar_t | |
29401c30 NB |
289 | get_effective_char (pfile) |
290 | cpp_reader *pfile; | |
64aaf407 | 291 | { |
29401c30 | 292 | cpp_buffer *buffer = pfile->buffer; |
0d9f234d NB |
293 | cppchar_t next = EOF; |
294 | ||
295 | if (buffer->cur < buffer->rlimit) | |
296 | { | |
297 | next = *buffer->cur++; | |
298 | ||
299 | /* '?' can introduce trigraphs (and therefore backslash); '\\' | |
300 | can introduce escaped newlines, which we want to skip, or | |
301 | UCNs, which, depending upon lexer state, we will handle in | |
302 | the future. */ | |
303 | if (next == '?' || next == '\\') | |
29401c30 | 304 | next = skip_escaped_newlines (pfile, next); |
0d9f234d NB |
305 | } |
306 | ||
307 | buffer->read_ahead = next; | |
308 | return next; | |
64aaf407 NB |
309 | } |
310 | ||
0d9f234d NB |
311 | /* Skip a C-style block comment. We find the end of the comment by |
312 | seeing if an asterisk is before every '/' we encounter. Returns | |
313 | non-zero if comment terminated by EOF, zero otherwise. */ | |
041c3194 ZW |
314 | static int |
315 | skip_block_comment (pfile) | |
45b966db ZW |
316 | cpp_reader *pfile; |
317 | { | |
041c3194 | 318 | cpp_buffer *buffer = pfile->buffer; |
d8090680 | 319 | cppchar_t c = EOF, prevc = EOF; |
0d9f234d | 320 | |
cbcff6df | 321 | pfile->state.lexing_comment = 1; |
0d9f234d | 322 | while (buffer->cur != buffer->rlimit) |
45b966db | 323 | { |
0d9f234d NB |
324 | prevc = c, c = *buffer->cur++; |
325 | ||
326 | next_char: | |
327 | /* FIXME: For speed, create a new character class of characters | |
93c80368 | 328 | of interest inside block comments. */ |
0d9f234d | 329 | if (c == '?' || c == '\\') |
29401c30 | 330 | c = skip_escaped_newlines (pfile, c); |
041c3194 | 331 | |
0d9f234d NB |
332 | /* People like decorating comments with '*', so check for '/' |
333 | instead for efficiency. */ | |
041c3194 | 334 | if (c == '/') |
45b966db | 335 | { |
0d9f234d NB |
336 | if (prevc == '*') |
337 | break; | |
041c3194 | 338 | |
0d9f234d NB |
339 | /* Warn about potential nested comments, but not if the '/' |
340 | comes immediately before the true comment delimeter. | |
041c3194 | 341 | Don't bother to get it right across escaped newlines. */ |
0d9f234d NB |
342 | if (CPP_OPTION (pfile, warn_comments) |
343 | && buffer->cur != buffer->rlimit) | |
45b966db | 344 | { |
0d9f234d NB |
345 | prevc = c, c = *buffer->cur++; |
346 | if (c == '*' && buffer->cur != buffer->rlimit) | |
347 | { | |
348 | prevc = c, c = *buffer->cur++; | |
349 | if (c != '/') | |
67821e3a NB |
350 | cpp_warning_with_line (pfile, pfile->line, |
351 | CPP_BUF_COL (buffer) - 2, | |
0d9f234d NB |
352 | "\"/*\" within comment"); |
353 | } | |
354 | goto next_char; | |
45b966db | 355 | } |
45b966db | 356 | } |
91fcd158 | 357 | else if (is_vspace (c)) |
45b966db | 358 | { |
1444f2ed | 359 | prevc = c, c = handle_newline (pfile, c); |
0d9f234d | 360 | goto next_char; |
45b966db | 361 | } |
52fadca8 | 362 | else if (c == '\t') |
0d9f234d | 363 | adjust_column (pfile); |
45b966db | 364 | } |
041c3194 | 365 | |
cbcff6df | 366 | pfile->state.lexing_comment = 0; |
0d9f234d NB |
367 | buffer->read_ahead = EOF; |
368 | return c != '/' || prevc != '*'; | |
45b966db ZW |
369 | } |
370 | ||
f9a0e96c | 371 | /* Skip a C++ line comment. Handles escaped newlines. Returns |
0d9f234d NB |
372 | non-zero if a multiline comment. The following new line, if any, |
373 | is left in buffer->read_ahead. */ | |
041c3194 | 374 | static int |
cbcff6df NB |
375 | skip_line_comment (pfile) |
376 | cpp_reader *pfile; | |
45b966db | 377 | { |
cbcff6df | 378 | cpp_buffer *buffer = pfile->buffer; |
67821e3a | 379 | unsigned int orig_line = pfile->line; |
0d9f234d | 380 | cppchar_t c; |
041c3194 | 381 | |
cbcff6df | 382 | pfile->state.lexing_comment = 1; |
0d9f234d | 383 | do |
041c3194 | 384 | { |
0d9f234d NB |
385 | c = EOF; |
386 | if (buffer->cur == buffer->rlimit) | |
387 | break; | |
041c3194 | 388 | |
0d9f234d NB |
389 | c = *buffer->cur++; |
390 | if (c == '?' || c == '\\') | |
29401c30 | 391 | c = skip_escaped_newlines (pfile, c); |
041c3194 | 392 | } |
0d9f234d | 393 | while (!is_vspace (c)); |
45b966db | 394 | |
cbcff6df | 395 | pfile->state.lexing_comment = 0; |
0d9f234d | 396 | buffer->read_ahead = c; /* Leave any newline for caller. */ |
67821e3a | 397 | return orig_line != pfile->line; |
041c3194 | 398 | } |
45b966db | 399 | |
0d9f234d NB |
400 | /* pfile->buffer->cur is one beyond the \t character. Update |
401 | col_adjust so we track the column correctly. */ | |
52fadca8 | 402 | static void |
0d9f234d | 403 | adjust_column (pfile) |
52fadca8 | 404 | cpp_reader *pfile; |
52fadca8 | 405 | { |
0d9f234d NB |
406 | cpp_buffer *buffer = pfile->buffer; |
407 | unsigned int col = CPP_BUF_COL (buffer) - 1; /* Zero-based column. */ | |
52fadca8 NB |
408 | |
409 | /* Round it up to multiple of the tabstop, but subtract 1 since the | |
410 | tab itself occupies a character position. */ | |
0d9f234d NB |
411 | buffer->col_adjust += (CPP_OPTION (pfile, tabstop) |
412 | - col % CPP_OPTION (pfile, tabstop)) - 1; | |
52fadca8 NB |
413 | } |
414 | ||
0d9f234d NB |
415 | /* Skips whitespace, saving the next non-whitespace character. |
416 | Adjusts pfile->col_adjust to account for tabs. Without this, | |
417 | tokens might be assigned an incorrect column. */ | |
041c3194 | 418 | static void |
0d9f234d | 419 | skip_whitespace (pfile, c) |
041c3194 | 420 | cpp_reader *pfile; |
0d9f234d | 421 | cppchar_t c; |
041c3194 ZW |
422 | { |
423 | cpp_buffer *buffer = pfile->buffer; | |
0d9f234d | 424 | unsigned int warned = 0; |
45b966db | 425 | |
0d9f234d | 426 | do |
041c3194 | 427 | { |
91fcd158 NB |
428 | /* Horizontal space always OK. */ |
429 | if (c == ' ') | |
0d9f234d | 430 | ; |
91fcd158 | 431 | else if (c == '\t') |
0d9f234d NB |
432 | adjust_column (pfile); |
433 | /* Just \f \v or \0 left. */ | |
91fcd158 | 434 | else if (c == '\0') |
041c3194 | 435 | { |
91fcd158 | 436 | if (!warned) |
0d9f234d NB |
437 | { |
438 | cpp_warning (pfile, "null character(s) ignored"); | |
439 | warned = 1; | |
440 | } | |
45b966db | 441 | } |
93c80368 | 442 | else if (pfile->state.in_directive && CPP_PEDANTIC (pfile)) |
67821e3a | 443 | cpp_pedwarn_with_line (pfile, pfile->line, |
91fcd158 NB |
444 | CPP_BUF_COL (buffer), |
445 | "%s in preprocessing directive", | |
446 | c == '\f' ? "form feed" : "vertical tab"); | |
0d9f234d NB |
447 | |
448 | c = EOF; | |
449 | if (buffer->cur == buffer->rlimit) | |
450 | break; | |
451 | c = *buffer->cur++; | |
45b966db | 452 | } |
ec5c56db | 453 | /* We only want non-vertical space, i.e. ' ' \t \f \v \0. */ |
0d9f234d NB |
454 | while (is_nvspace (c)); |
455 | ||
456 | /* Remember the next character. */ | |
457 | buffer->read_ahead = c; | |
041c3194 | 458 | } |
45b966db | 459 | |
93c80368 NB |
460 | /* See if the characters of a number token are valid in a name (no |
461 | '.', '+' or '-'). */ | |
462 | static int | |
463 | name_p (pfile, string) | |
464 | cpp_reader *pfile; | |
465 | const cpp_string *string; | |
466 | { | |
467 | unsigned int i; | |
468 | ||
469 | for (i = 0; i < string->len; i++) | |
470 | if (!is_idchar (string->text[i])) | |
471 | return 0; | |
472 | ||
473 | return 1; | |
474 | } | |
475 | ||
2c3fcba6 ZW |
476 | /* Parse an identifier, skipping embedded backslash-newlines. This is |
477 | a critical inner loop. The common case is an identifier which has | |
478 | not been split by backslash-newline, does not contain a dollar | |
479 | sign, and has already been scanned (roughly 10:1 ratio of | |
480 | seen:unseen identifiers in normal code; the distribution is | |
481 | Poisson-like). Second most common case is a new identifier, not | |
482 | split and no dollar sign. The other possibilities are rare and | |
483 | have been relegated to parse_identifier_slow. */ | |
0d9f234d NB |
484 | |
485 | static cpp_hashnode * | |
2c3fcba6 | 486 | parse_identifier (pfile) |
45b966db | 487 | cpp_reader *pfile; |
45b966db | 488 | { |
93c80368 | 489 | cpp_hashnode *result; |
2c3fcba6 ZW |
490 | const U_CHAR *cur, *rlimit; |
491 | ||
492 | /* Fast-path loop. Skim over a normal identifier. | |
493 | N.B. ISIDNUM does not include $. */ | |
494 | cur = pfile->buffer->cur - 1; | |
495 | rlimit = pfile->buffer->rlimit; | |
496 | do | |
497 | cur++; | |
498 | while (cur < rlimit && ISIDNUM (*cur)); | |
499 | ||
500 | /* Check for slow-path cases. */ | |
501 | if (cur < rlimit && (*cur == '?' || *cur == '\\' || *cur == '$')) | |
502 | result = parse_identifier_slow (pfile, cur); | |
503 | else | |
504 | { | |
505 | const U_CHAR *base = pfile->buffer->cur - 1; | |
506 | result = (cpp_hashnode *) | |
507 | ht_lookup (pfile->hash_table, base, cur - base, HT_ALLOC); | |
508 | pfile->buffer->cur = cur; | |
509 | } | |
510 | ||
511 | /* Rarely, identifiers require diagnostics when lexed. | |
512 | XXX Has to be forced out of the fast path. */ | |
513 | if (__builtin_expect ((result->flags & NODE_DIAGNOSTIC) | |
514 | && !pfile->state.skipping, 0)) | |
515 | { | |
516 | /* It is allowed to poison the same identifier twice. */ | |
517 | if ((result->flags & NODE_POISONED) && !pfile->state.poisoned_ok) | |
518 | cpp_error (pfile, "attempt to use poisoned \"%s\"", | |
519 | NODE_NAME (result)); | |
520 | ||
521 | /* Constraint 6.10.3.5: __VA_ARGS__ should only appear in the | |
522 | replacement list of a variadic macro. */ | |
523 | if (result == pfile->spec_nodes.n__VA_ARGS__ | |
524 | && !pfile->state.va_args_ok) | |
525 | cpp_pedwarn (pfile, | |
526 | "__VA_ARGS__ can only appear in the expansion of a C99 variadic macro"); | |
527 | } | |
528 | ||
529 | return result; | |
530 | } | |
531 | ||
532 | /* Slow path. This handles identifiers which have been split, and | |
533 | identifiers which contain dollar signs. The part of the identifier | |
534 | from PFILE->buffer->cur-1 to CUR has already been scanned. */ | |
535 | static cpp_hashnode * | |
536 | parse_identifier_slow (pfile, cur) | |
537 | cpp_reader *pfile; | |
538 | const U_CHAR *cur; | |
539 | { | |
0d9f234d | 540 | cpp_buffer *buffer = pfile->buffer; |
2c3fcba6 | 541 | const U_CHAR *base = buffer->cur - 1; |
2a967f3d | 542 | struct obstack *stack = &pfile->hash_table->stack; |
2c3fcba6 ZW |
543 | unsigned int c, saw_dollar = 0, len; |
544 | ||
545 | /* Copy the part of the token which is known to be okay. */ | |
546 | obstack_grow (stack, base, cur - base); | |
041c3194 | 547 | |
2c3fcba6 ZW |
548 | /* Now process the part which isn't. We are looking at one of |
549 | '$', '\\', or '?' on entry to this loop. */ | |
550 | c = *cur++; | |
551 | buffer->cur = cur; | |
0d9f234d | 552 | do |
041c3194 | 553 | { |
2c3fcba6 ZW |
554 | while (is_idchar (c)) |
555 | { | |
556 | obstack_1grow (stack, c); | |
45b966db | 557 | |
2c3fcba6 ZW |
558 | if (c == '$') |
559 | saw_dollar++; | |
ba89d661 | 560 | |
2c3fcba6 ZW |
561 | c = EOF; |
562 | if (buffer->cur == buffer->rlimit) | |
563 | break; | |
ba89d661 | 564 | |
2c3fcba6 ZW |
565 | c = *buffer->cur++; |
566 | } | |
ba89d661 | 567 | |
0d9f234d NB |
568 | /* Potential escaped newline? */ |
569 | if (c != '?' && c != '\\') | |
2c3fcba6 | 570 | break; |
29401c30 | 571 | c = skip_escaped_newlines (pfile, c); |
041c3194 | 572 | } |
0d9f234d NB |
573 | while (is_idchar (c)); |
574 | ||
93c80368 NB |
575 | /* Remember the next character. */ |
576 | buffer->read_ahead = c; | |
577 | ||
4fe9b91c | 578 | /* $ is not an identifier character in the standard, but is commonly |
0d9f234d NB |
579 | accepted as an extension. Don't warn about it in skipped |
580 | conditional blocks. */ | |
cef0d199 | 581 | if (saw_dollar && CPP_PEDANTIC (pfile) && ! pfile->state.skipping) |
0d9f234d NB |
582 | cpp_pedwarn (pfile, "'$' character(s) in identifier"); |
583 | ||
93c80368 | 584 | /* Identifiers are null-terminated. */ |
2a967f3d NB |
585 | len = obstack_object_size (stack); |
586 | obstack_1grow (stack, '\0'); | |
93c80368 | 587 | |
2c3fcba6 | 588 | return (cpp_hashnode *) |
2a967f3d | 589 | ht_lookup (pfile->hash_table, obstack_finish (stack), len, HT_ALLOCED); |
45b966db ZW |
590 | } |
591 | ||
0d9f234d | 592 | /* Parse a number, skipping embedded backslash-newlines. */ |
45b966db | 593 | static void |
93c80368 | 594 | parse_number (pfile, number, c, leading_period) |
45b966db | 595 | cpp_reader *pfile; |
0d9f234d NB |
596 | cpp_string *number; |
597 | cppchar_t c; | |
93c80368 | 598 | int leading_period; |
45b966db | 599 | { |
041c3194 | 600 | cpp_buffer *buffer = pfile->buffer; |
93c80368 | 601 | unsigned char *dest, *limit; |
45b966db | 602 | |
ece54d54 NB |
603 | dest = BUFF_FRONT (pfile->u_buff); |
604 | limit = BUFF_LIMIT (pfile->u_buff); | |
cbcff6df | 605 | |
93c80368 NB |
606 | /* Place a leading period. */ |
607 | if (leading_period) | |
608 | { | |
ece54d54 NB |
609 | if (dest == limit) |
610 | { | |
8c3b2693 | 611 | _cpp_extend_buff (pfile, &pfile->u_buff, 1); |
ece54d54 NB |
612 | dest = BUFF_FRONT (pfile->u_buff); |
613 | limit = BUFF_LIMIT (pfile->u_buff); | |
614 | } | |
93c80368 NB |
615 | *dest++ = '.'; |
616 | } | |
617 | ||
0d9f234d | 618 | do |
041c3194 | 619 | { |
0d9f234d NB |
620 | do |
621 | { | |
93c80368 | 622 | /* Need room for terminating null. */ |
ece54d54 NB |
623 | if ((size_t) (limit - dest) < 2) |
624 | { | |
625 | size_t len_so_far = dest - BUFF_FRONT (pfile->u_buff); | |
8c3b2693 | 626 | _cpp_extend_buff (pfile, &pfile->u_buff, 2); |
ece54d54 NB |
627 | dest = BUFF_FRONT (pfile->u_buff) + len_so_far; |
628 | limit = BUFF_LIMIT (pfile->u_buff); | |
629 | } | |
93c80368 | 630 | *dest++ = c; |
0d9f234d | 631 | |
0d9f234d NB |
632 | c = EOF; |
633 | if (buffer->cur == buffer->rlimit) | |
634 | break; | |
45b966db | 635 | |
0d9f234d NB |
636 | c = *buffer->cur++; |
637 | } | |
93c80368 | 638 | while (is_numchar (c) || c == '.' || VALID_SIGN (c, dest[-1])); |
45b966db | 639 | |
0d9f234d NB |
640 | /* Potential escaped newline? */ |
641 | if (c != '?' && c != '\\') | |
642 | break; | |
29401c30 | 643 | c = skip_escaped_newlines (pfile, c); |
45b966db | 644 | } |
93c80368 | 645 | while (is_numchar (c) || c == '.' || VALID_SIGN (c, dest[-1])); |
cbcff6df | 646 | |
0d9f234d NB |
647 | /* Remember the next character. */ |
648 | buffer->read_ahead = c; | |
64aaf407 | 649 | |
93c80368 NB |
650 | /* Null-terminate the number. */ |
651 | *dest = '\0'; | |
652 | ||
ece54d54 | 653 | number->text = BUFF_FRONT (pfile->u_buff); |
93c80368 | 654 | number->len = dest - number->text; |
ece54d54 | 655 | BUFF_FRONT (pfile->u_buff) = dest + 1; |
0d9f234d NB |
656 | } |
657 | ||
658 | /* Subroutine of parse_string. Emits error for unterminated strings. */ | |
659 | static void | |
93c80368 | 660 | unterminated (pfile, term) |
0d9f234d | 661 | cpp_reader *pfile; |
0d9f234d NB |
662 | int term; |
663 | { | |
664 | cpp_error (pfile, "missing terminating %c character", term); | |
665 | ||
50410426 | 666 | if (term == '\"' && pfile->mls_line && pfile->mls_line != pfile->line) |
041c3194 | 667 | { |
50410426 | 668 | cpp_error_with_line (pfile, pfile->mls_line, pfile->mls_col, |
0d9f234d | 669 | "possible start of unterminated string literal"); |
50410426 | 670 | pfile->mls_line = 0; |
041c3194 | 671 | } |
45b966db ZW |
672 | } |
673 | ||
93c80368 NB |
674 | /* Subroutine of parse_string. */ |
675 | static int | |
676 | unescaped_terminator_p (pfile, dest) | |
677 | cpp_reader *pfile; | |
678 | const unsigned char *dest; | |
679 | { | |
680 | const unsigned char *start, *temp; | |
681 | ||
682 | /* In #include-style directives, terminators are not escapeable. */ | |
683 | if (pfile->state.angled_headers) | |
684 | return 1; | |
685 | ||
ece54d54 | 686 | start = BUFF_FRONT (pfile->u_buff); |
93c80368 NB |
687 | |
688 | /* An odd number of consecutive backslashes represents an escaped | |
689 | terminator. */ | |
690 | for (temp = dest; temp > start && temp[-1] == '\\'; temp--) | |
691 | ; | |
692 | ||
693 | return ((dest - temp) & 1) == 0; | |
694 | } | |
695 | ||
0d9f234d | 696 | /* Parses a string, character constant, or angle-bracketed header file |
7868b4a2 NB |
697 | name. Handles embedded trigraphs and escaped newlines. The stored |
698 | string is guaranteed NUL-terminated, but it is not guaranteed that | |
699 | this is the first NUL since embedded NULs are preserved. | |
45b966db | 700 | |
7868b4a2 | 701 | Multi-line strings are allowed, but they are deprecated. */ |
041c3194 | 702 | static void |
0d9f234d | 703 | parse_string (pfile, token, terminator) |
45b966db | 704 | cpp_reader *pfile; |
041c3194 | 705 | cpp_token *token; |
0d9f234d | 706 | cppchar_t terminator; |
45b966db | 707 | { |
041c3194 | 708 | cpp_buffer *buffer = pfile->buffer; |
93c80368 | 709 | unsigned char *dest, *limit; |
0d9f234d | 710 | cppchar_t c; |
d82fc108 | 711 | bool warned_nulls = false, warned_multi = false; |
0d9f234d | 712 | |
ece54d54 NB |
713 | dest = BUFF_FRONT (pfile->u_buff); |
714 | limit = BUFF_LIMIT (pfile->u_buff); | |
93c80368 | 715 | |
0d9f234d | 716 | for (;;) |
45b966db | 717 | { |
0d9f234d | 718 | if (buffer->cur == buffer->rlimit) |
7868b4a2 NB |
719 | c = EOF; |
720 | else | |
721 | c = *buffer->cur++; | |
722 | ||
723 | have_char: | |
724 | /* We need space for the terminating NUL. */ | |
ece54d54 NB |
725 | if ((size_t) (limit - dest) < 1) |
726 | { | |
727 | size_t len_so_far = dest - BUFF_FRONT (pfile->u_buff); | |
8c3b2693 | 728 | _cpp_extend_buff (pfile, &pfile->u_buff, 2); |
ece54d54 NB |
729 | dest = BUFF_FRONT (pfile->u_buff) + len_so_far; |
730 | limit = BUFF_LIMIT (pfile->u_buff); | |
731 | } | |
7868b4a2 NB |
732 | |
733 | if (c == EOF) | |
0d9f234d | 734 | { |
93c80368 | 735 | unterminated (pfile, terminator); |
0d9f234d NB |
736 | break; |
737 | } | |
0d9f234d | 738 | |
0d9f234d NB |
739 | /* Handle trigraphs, escaped newlines etc. */ |
740 | if (c == '?' || c == '\\') | |
29401c30 | 741 | c = skip_escaped_newlines (pfile, c); |
45b966db | 742 | |
93c80368 | 743 | if (c == terminator && unescaped_terminator_p (pfile, dest)) |
45b966db | 744 | { |
93c80368 NB |
745 | c = EOF; |
746 | break; | |
0d9f234d NB |
747 | } |
748 | else if (is_vspace (c)) | |
749 | { | |
750 | /* In assembly language, silently terminate string and | |
751 | character literals at end of line. This is a kludge | |
752 | around not knowing where comments are. */ | |
bdb05a7b | 753 | if (CPP_OPTION (pfile, lang) == CLK_ASM && terminator != '>') |
0d9f234d | 754 | break; |
45b966db | 755 | |
0d9f234d NB |
756 | /* Character constants and header names may not extend over |
757 | multiple lines. In Standard C, neither may strings. | |
758 | Unfortunately, we accept multiline strings as an | |
16eb2788 NB |
759 | extension, except in #include family directives. */ |
760 | if (terminator != '"' || pfile->state.angled_headers) | |
45b966db | 761 | { |
93c80368 | 762 | unterminated (pfile, terminator); |
0d9f234d | 763 | break; |
45b966db | 764 | } |
45b966db | 765 | |
d82fc108 NB |
766 | if (!warned_multi) |
767 | { | |
768 | warned_multi = true; | |
769 | cpp_pedwarn (pfile, "multi-line string literals are deprecated"); | |
770 | } | |
771 | ||
50410426 NB |
772 | if (pfile->mls_line == 0) |
773 | { | |
774 | pfile->mls_line = token->line; | |
775 | pfile->mls_col = token->col; | |
776 | } | |
0d9f234d | 777 | |
1444f2ed | 778 | c = handle_newline (pfile, c); |
7868b4a2 NB |
779 | *dest++ = '\n'; |
780 | goto have_char; | |
0d9f234d | 781 | } |
d82fc108 | 782 | else if (c == '\0' && !warned_nulls) |
0d9f234d | 783 | { |
d82fc108 NB |
784 | warned_nulls = true; |
785 | cpp_warning (pfile, "null character(s) preserved in literal"); | |
45b966db | 786 | } |
45b966db | 787 | |
93c80368 | 788 | *dest++ = c; |
45b966db ZW |
789 | } |
790 | ||
93c80368 | 791 | /* Remember the next character. */ |
0d9f234d | 792 | buffer->read_ahead = c; |
7868b4a2 | 793 | *dest = '\0'; |
45b966db | 794 | |
ece54d54 NB |
795 | token->val.str.text = BUFF_FRONT (pfile->u_buff); |
796 | token->val.str.len = dest - BUFF_FRONT (pfile->u_buff); | |
797 | BUFF_FRONT (pfile->u_buff) = dest + 1; | |
0d9f234d | 798 | } |
041c3194 | 799 | |
93c80368 | 800 | /* The stored comment includes the comment start and any terminator. */ |
9e62c811 | 801 | static void |
0d9f234d NB |
802 | save_comment (pfile, token, from) |
803 | cpp_reader *pfile; | |
041c3194 ZW |
804 | cpp_token *token; |
805 | const unsigned char *from; | |
9e62c811 | 806 | { |
041c3194 | 807 | unsigned char *buffer; |
0d9f234d | 808 | unsigned int len; |
0d9f234d | 809 | |
1c6d33ef | 810 | len = pfile->buffer->cur - from + 1; /* + 1 for the initial '/'. */ |
3542203b NB |
811 | /* C++ comments probably (not definitely) have moved past a new |
812 | line, which we don't want to save in the comment. */ | |
813 | if (pfile->buffer->read_ahead != EOF) | |
814 | len--; | |
ece54d54 | 815 | buffer = _cpp_unaligned_alloc (pfile, len); |
041c3194 | 816 | |
041c3194 | 817 | token->type = CPP_COMMENT; |
bfb9dc7f | 818 | token->val.str.len = len; |
0d9f234d | 819 | token->val.str.text = buffer; |
45b966db | 820 | |
1c6d33ef NB |
821 | buffer[0] = '/'; |
822 | memcpy (buffer + 1, from, len - 1); | |
0d9f234d | 823 | } |
45b966db | 824 | |
14baae01 | 825 | /* Subroutine of _cpp_lex_direct to handle '%'. A little tricky, since we |
cbcff6df | 826 | want to avoid stepping back when lexing %:%X. */ |
0d9f234d | 827 | static void |
29401c30 NB |
828 | lex_percent (pfile, result) |
829 | cpp_reader *pfile; | |
0d9f234d | 830 | cpp_token *result; |
0d9f234d | 831 | { |
29401c30 | 832 | cpp_buffer *buffer= pfile->buffer; |
cbcff6df NB |
833 | cppchar_t c; |
834 | ||
835 | result->type = CPP_MOD; | |
836 | /* Parsing %:%X could leave an extra character. */ | |
837 | if (buffer->extra_char == EOF) | |
29401c30 | 838 | c = get_effective_char (pfile); |
cbcff6df NB |
839 | else |
840 | { | |
841 | c = buffer->read_ahead = buffer->extra_char; | |
842 | buffer->extra_char = EOF; | |
843 | } | |
844 | ||
845 | if (c == '=') | |
846 | ACCEPT_CHAR (CPP_MOD_EQ); | |
29401c30 | 847 | else if (CPP_OPTION (pfile, digraphs)) |
cbcff6df NB |
848 | { |
849 | if (c == ':') | |
850 | { | |
851 | result->flags |= DIGRAPH; | |
852 | ACCEPT_CHAR (CPP_HASH); | |
29401c30 | 853 | if (get_effective_char (pfile) == '%') |
cbcff6df | 854 | { |
29401c30 | 855 | buffer->extra_char = get_effective_char (pfile); |
cbcff6df NB |
856 | if (buffer->extra_char == ':') |
857 | { | |
858 | buffer->extra_char = EOF; | |
859 | ACCEPT_CHAR (CPP_PASTE); | |
860 | } | |
861 | else | |
862 | /* We'll catch the extra_char when we're called back. */ | |
863 | buffer->read_ahead = '%'; | |
864 | } | |
865 | } | |
866 | else if (c == '>') | |
867 | { | |
868 | result->flags |= DIGRAPH; | |
869 | ACCEPT_CHAR (CPP_CLOSE_BRACE); | |
870 | } | |
871 | } | |
872 | } | |
873 | ||
14baae01 | 874 | /* Subroutine of _cpp_lex_direct to handle '.'. This is tricky, since we |
cbcff6df NB |
875 | want to avoid stepping back when lexing '...' or '.123'. In the |
876 | latter case we should also set a flag for parse_number. */ | |
877 | static void | |
878 | lex_dot (pfile, result) | |
879 | cpp_reader *pfile; | |
880 | cpp_token *result; | |
881 | { | |
882 | cpp_buffer *buffer = pfile->buffer; | |
883 | cppchar_t c; | |
884 | ||
885 | /* Parsing ..X could leave an extra character. */ | |
886 | if (buffer->extra_char == EOF) | |
29401c30 | 887 | c = get_effective_char (pfile); |
cbcff6df NB |
888 | else |
889 | { | |
890 | c = buffer->read_ahead = buffer->extra_char; | |
891 | buffer->extra_char = EOF; | |
892 | } | |
0d9f234d | 893 | |
cbcff6df NB |
894 | /* All known character sets have 0...9 contiguous. */ |
895 | if (c >= '0' && c <= '9') | |
896 | { | |
897 | result->type = CPP_NUMBER; | |
93c80368 | 898 | parse_number (pfile, &result->val.str, c, 1); |
cbcff6df | 899 | } |
041c3194 | 900 | else |
ea4a453b | 901 | { |
cbcff6df NB |
902 | result->type = CPP_DOT; |
903 | if (c == '.') | |
904 | { | |
29401c30 | 905 | buffer->extra_char = get_effective_char (pfile); |
cbcff6df NB |
906 | if (buffer->extra_char == '.') |
907 | { | |
908 | buffer->extra_char = EOF; | |
909 | ACCEPT_CHAR (CPP_ELLIPSIS); | |
910 | } | |
911 | else | |
912 | /* We'll catch the extra_char when we're called back. */ | |
913 | buffer->read_ahead = '.'; | |
914 | } | |
915 | else if (c == '*' && CPP_OPTION (pfile, cplusplus)) | |
916 | ACCEPT_CHAR (CPP_DOT_STAR); | |
ea4a453b | 917 | } |
45b966db ZW |
918 | } |
919 | ||
5fddcffc NB |
920 | /* Allocate COUNT tokens for RUN. */ |
921 | void | |
922 | _cpp_init_tokenrun (run, count) | |
923 | tokenrun *run; | |
924 | unsigned int count; | |
925 | { | |
926 | run->base = xnewvec (cpp_token, count); | |
927 | run->limit = run->base + count; | |
928 | run->next = NULL; | |
929 | } | |
930 | ||
931 | /* Returns the next tokenrun, or creates one if there is none. */ | |
932 | static tokenrun * | |
933 | next_tokenrun (run) | |
934 | tokenrun *run; | |
935 | { | |
936 | if (run->next == NULL) | |
937 | { | |
938 | run->next = xnew (tokenrun); | |
bdcbe496 | 939 | run->next->prev = run; |
5fddcffc NB |
940 | _cpp_init_tokenrun (run->next, 250); |
941 | } | |
942 | ||
943 | return run->next; | |
944 | } | |
945 | ||
4ed5bcfb NB |
946 | /* Allocate a single token that is invalidated at the same time as the |
947 | rest of the tokens on the line. Has its line and col set to the | |
948 | same as the last lexed token, so that diagnostics appear in the | |
949 | right place. */ | |
950 | cpp_token * | |
951 | _cpp_temp_token (pfile) | |
952 | cpp_reader *pfile; | |
953 | { | |
954 | cpp_token *old, *result; | |
955 | ||
956 | old = pfile->cur_token - 1; | |
957 | if (pfile->cur_token == pfile->cur_run->limit) | |
958 | { | |
959 | pfile->cur_run = next_tokenrun (pfile->cur_run); | |
960 | pfile->cur_token = pfile->cur_run->base; | |
961 | } | |
962 | ||
963 | result = pfile->cur_token++; | |
964 | result->line = old->line; | |
965 | result->col = old->col; | |
966 | return result; | |
967 | } | |
968 | ||
14baae01 NB |
969 | /* Lex a token into RESULT (external interface). Takes care of issues |
970 | like directive handling, token lookahead, multiple include | |
971 | opimisation and skipping. */ | |
345894b4 NB |
972 | const cpp_token * |
973 | _cpp_lex_token (pfile) | |
45b966db | 974 | cpp_reader *pfile; |
5fddcffc | 975 | { |
bdcbe496 | 976 | cpp_token *result; |
5fddcffc | 977 | |
bdcbe496 | 978 | for (;;) |
5fddcffc | 979 | { |
bdcbe496 | 980 | if (pfile->cur_token == pfile->cur_run->limit) |
5fddcffc | 981 | { |
bdcbe496 NB |
982 | pfile->cur_run = next_tokenrun (pfile->cur_run); |
983 | pfile->cur_token = pfile->cur_run->base; | |
5fddcffc NB |
984 | } |
985 | ||
bdcbe496 | 986 | if (pfile->lookaheads) |
14baae01 NB |
987 | { |
988 | pfile->lookaheads--; | |
989 | result = pfile->cur_token++; | |
990 | } | |
bdcbe496 | 991 | else |
14baae01 | 992 | result = _cpp_lex_direct (pfile); |
bdcbe496 NB |
993 | |
994 | if (result->flags & BOL) | |
5fddcffc | 995 | { |
bdcbe496 NB |
996 | /* Is this a directive. If _cpp_handle_directive returns |
997 | false, it is an assembler #. */ | |
998 | if (result->type == CPP_HASH | |
999 | && !pfile->state.parsing_args | |
1000 | && _cpp_handle_directive (pfile, result->flags & PREV_WHITE)) | |
1001 | continue; | |
97293897 NB |
1002 | if (pfile->cb.line_change && !pfile->state.skipping) |
1003 | (*pfile->cb.line_change)(pfile, result, pfile->state.parsing_args); | |
5fddcffc | 1004 | } |
5fddcffc | 1005 | |
bdcbe496 NB |
1006 | /* We don't skip tokens in directives. */ |
1007 | if (pfile->state.in_directive) | |
1008 | break; | |
5fddcffc | 1009 | |
bdcbe496 | 1010 | /* Outside a directive, invalidate controlling macros. At file |
14baae01 | 1011 | EOF, _cpp_lex_direct takes care of popping the buffer, so we never |
bdcbe496 | 1012 | get here and MI optimisation works. */ |
5fddcffc | 1013 | pfile->mi_valid = false; |
bdcbe496 NB |
1014 | |
1015 | if (!pfile->state.skipping || result->type == CPP_EOF) | |
1016 | break; | |
5fddcffc NB |
1017 | } |
1018 | ||
345894b4 | 1019 | return result; |
5fddcffc NB |
1020 | } |
1021 | ||
14baae01 NB |
1022 | /* Lex a token into pfile->cur_token, which is also incremented, to |
1023 | get diagnostics pointing to the correct location. | |
1024 | ||
1025 | Does not handle issues such as token lookahead, multiple-include | |
1026 | optimisation, directives, skipping etc. This function is only | |
1027 | suitable for use by _cpp_lex_token, and in special cases like | |
1028 | lex_expansion_token which doesn't care for any of these issues. | |
1029 | ||
1030 | When meeting a newline, returns CPP_EOF if parsing a directive, | |
1031 | otherwise returns to the start of the token buffer if permissible. | |
1032 | Returns the location of the lexed token. */ | |
1033 | cpp_token * | |
1034 | _cpp_lex_direct (pfile) | |
5fddcffc | 1035 | cpp_reader *pfile; |
45b966db | 1036 | { |
0d9f234d | 1037 | cppchar_t c; |
adb84b42 | 1038 | cpp_buffer *buffer; |
0d9f234d | 1039 | const unsigned char *comment_start; |
14baae01 | 1040 | cpp_token *result = pfile->cur_token++; |
9ec7291f | 1041 | |
5fddcffc | 1042 | fresh_line: |
adb84b42 | 1043 | buffer = pfile->buffer; |
bd969772 NB |
1044 | result->flags = buffer->saved_flags; |
1045 | buffer->saved_flags = 0; | |
5fddcffc | 1046 | update_tokens_line: |
1444f2ed | 1047 | result->line = pfile->line; |
041c3194 | 1048 | |
5fddcffc | 1049 | skipped_white: |
0d9f234d NB |
1050 | c = buffer->read_ahead; |
1051 | if (c == EOF && buffer->cur < buffer->rlimit) | |
5fddcffc NB |
1052 | c = *buffer->cur++; |
1053 | result->col = CPP_BUF_COLUMN (buffer, buffer->cur); | |
0d9f234d | 1054 | buffer->read_ahead = EOF; |
5fddcffc NB |
1055 | |
1056 | trigraph: | |
0d9f234d | 1057 | switch (c) |
45b966db | 1058 | { |
0d9f234d | 1059 | case EOF: |
bdcbe496 | 1060 | buffer->saved_flags = BOL; |
5fddcffc | 1061 | if (!pfile->state.parsing_args && !pfile->state.in_directive) |
ef6e958a | 1062 | { |
bdcbe496 | 1063 | if (buffer->cur != buffer->line_base) |
5fddcffc NB |
1064 | { |
1065 | /* Non-empty files should end in a newline. Don't warn | |
1066 | for command line and _Pragma buffers. */ | |
1067 | if (!buffer->from_stage3) | |
1068 | cpp_pedwarn (pfile, "no newline at end of file"); | |
1069 | handle_newline (pfile, '\n'); | |
7364fdd8 | 1070 | } |
bdcbe496 NB |
1071 | |
1072 | /* Don't pop the last buffer. */ | |
1073 | if (buffer->prev) | |
1074 | { | |
1075 | unsigned char stop = buffer->return_at_eof; | |
1076 | ||
1077 | _cpp_pop_buffer (pfile); | |
1078 | if (!stop) | |
1079 | goto fresh_line; | |
1080 | } | |
ef6e958a | 1081 | } |
0d9f234d | 1082 | result->type = CPP_EOF; |
5fddcffc | 1083 | break; |
45b966db | 1084 | |
0d9f234d NB |
1085 | case ' ': case '\t': case '\f': case '\v': case '\0': |
1086 | skip_whitespace (pfile, c); | |
1087 | result->flags |= PREV_WHITE; | |
5fddcffc | 1088 | goto skipped_white; |
0d9f234d NB |
1089 | |
1090 | case '\n': case '\r': | |
bdcbe496 NB |
1091 | handle_newline (pfile, c); |
1092 | buffer->saved_flags = BOL; | |
1093 | if (! pfile->state.in_directive) | |
45b966db | 1094 | { |
4ed5bcfb NB |
1095 | if (pfile->state.parsing_args == 2) |
1096 | buffer->saved_flags |= PREV_WHITE; | |
bdcbe496 NB |
1097 | if (!pfile->keep_tokens) |
1098 | { | |
1099 | pfile->cur_run = &pfile->base_run; | |
1100 | result = pfile->base_run.base; | |
1101 | pfile->cur_token = result + 1; | |
1102 | } | |
1103 | goto fresh_line; | |
45b966db | 1104 | } |
5fddcffc NB |
1105 | result->type = CPP_EOF; |
1106 | break; | |
46d07497 | 1107 | |
0d9f234d NB |
1108 | case '?': |
1109 | case '\\': | |
1110 | /* These could start an escaped newline, or '?' a trigraph. Let | |
1111 | skip_escaped_newlines do all the work. */ | |
1112 | { | |
67821e3a | 1113 | unsigned int line = pfile->line; |
0d9f234d | 1114 | |
29401c30 | 1115 | c = skip_escaped_newlines (pfile, c); |
67821e3a | 1116 | if (line != pfile->line) |
0d9f234d NB |
1117 | /* We had at least one escaped newline of some sort, and the |
1118 | next character is in buffer->read_ahead. Update the | |
1119 | token's line and column. */ | |
5fddcffc | 1120 | goto update_tokens_line; |
0d9f234d NB |
1121 | |
1122 | /* We are either the original '?' or '\\', or a trigraph. */ | |
1123 | result->type = CPP_QUERY; | |
1124 | buffer->read_ahead = EOF; | |
1125 | if (c == '\\') | |
12c4f523 | 1126 | goto random_char; |
0d9f234d | 1127 | else if (c != '?') |
5fddcffc | 1128 | goto trigraph; |
0d9f234d NB |
1129 | } |
1130 | break; | |
46d07497 | 1131 | |
0d9f234d NB |
1132 | case '0': case '1': case '2': case '3': case '4': |
1133 | case '5': case '6': case '7': case '8': case '9': | |
1134 | result->type = CPP_NUMBER; | |
93c80368 | 1135 | parse_number (pfile, &result->val.str, c, 0); |
0d9f234d | 1136 | break; |
46d07497 | 1137 | |
0d9f234d NB |
1138 | case '$': |
1139 | if (!CPP_OPTION (pfile, dollars_in_ident)) | |
1140 | goto random_char; | |
ec5c56db | 1141 | /* Fall through... */ |
0d9f234d NB |
1142 | |
1143 | case '_': | |
1144 | case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': | |
1145 | case 'g': case 'h': case 'i': case 'j': case 'k': case 'l': | |
1146 | case 'm': case 'n': case 'o': case 'p': case 'q': case 'r': | |
1147 | case 's': case 't': case 'u': case 'v': case 'w': case 'x': | |
1148 | case 'y': case 'z': | |
1149 | case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': | |
1150 | case 'G': case 'H': case 'I': case 'J': case 'K': case 'L': | |
1151 | case 'M': case 'N': case 'O': case 'P': case 'Q': case 'R': | |
1152 | case 'S': case 'T': case 'U': case 'V': case 'W': case 'X': | |
1153 | case 'Y': case 'Z': | |
1154 | result->type = CPP_NAME; | |
2c3fcba6 | 1155 | result->val.node = parse_identifier (pfile); |
0d9f234d NB |
1156 | |
1157 | /* 'L' may introduce wide characters or strings. */ | |
93c80368 | 1158 | if (result->val.node == pfile->spec_nodes.n_L) |
0d9f234d | 1159 | { |
2c3fcba6 ZW |
1160 | c = buffer->read_ahead; |
1161 | if (c == EOF && buffer->cur < buffer->rlimit) | |
1162 | c = *buffer->cur; | |
0d9f234d | 1163 | if (c == '\'' || c == '"') |
ba89d661 | 1164 | { |
2c3fcba6 | 1165 | buffer->cur++; |
0d9f234d NB |
1166 | ACCEPT_CHAR (c == '"' ? CPP_WSTRING: CPP_WCHAR); |
1167 | goto make_string; | |
ba89d661 | 1168 | } |
0d9f234d NB |
1169 | } |
1170 | /* Convert named operators to their proper types. */ | |
93c80368 | 1171 | else if (result->val.node->flags & NODE_OPERATOR) |
0d9f234d NB |
1172 | { |
1173 | result->flags |= NAMED_OP; | |
93c80368 | 1174 | result->type = result->val.node->value.operator; |
0d9f234d NB |
1175 | } |
1176 | break; | |
1177 | ||
1178 | case '\'': | |
1179 | case '"': | |
1180 | result->type = c == '"' ? CPP_STRING: CPP_CHAR; | |
1181 | make_string: | |
1182 | parse_string (pfile, result, c); | |
1183 | break; | |
041c3194 | 1184 | |
0d9f234d | 1185 | case '/': |
1c6d33ef NB |
1186 | /* A potential block or line comment. */ |
1187 | comment_start = buffer->cur; | |
0d9f234d | 1188 | result->type = CPP_DIV; |
29401c30 | 1189 | c = get_effective_char (pfile); |
0d9f234d NB |
1190 | if (c == '=') |
1191 | ACCEPT_CHAR (CPP_DIV_EQ); | |
1c6d33ef NB |
1192 | if (c != '/' && c != '*') |
1193 | break; | |
e61fc951 | 1194 | |
1c6d33ef NB |
1195 | if (c == '*') |
1196 | { | |
0d9f234d | 1197 | if (skip_block_comment (pfile)) |
67821e3a | 1198 | cpp_error (pfile, "unterminated comment"); |
0d9f234d | 1199 | } |
1c6d33ef | 1200 | else |
0d9f234d | 1201 | { |
1c6d33ef NB |
1202 | if (!CPP_OPTION (pfile, cplusplus_comments) |
1203 | && !CPP_IN_SYSTEM_HEADER (pfile)) | |
1204 | break; | |
1205 | ||
bdb05a7b NB |
1206 | /* Warn about comments only if pedantically GNUC89, and not |
1207 | in system headers. */ | |
1208 | if (CPP_OPTION (pfile, lang) == CLK_GNUC89 && CPP_PEDANTIC (pfile) | |
a94c1199 | 1209 | && ! buffer->warned_cplusplus_comments) |
041c3194 | 1210 | { |
1c6d33ef NB |
1211 | cpp_pedwarn (pfile, |
1212 | "C++ style comments are not allowed in ISO C89"); | |
1213 | cpp_pedwarn (pfile, | |
1214 | "(this will be reported only once per input file)"); | |
1215 | buffer->warned_cplusplus_comments = 1; | |
1216 | } | |
0d9f234d | 1217 | |
a94c1199 | 1218 | /* Skip_line_comment updates buffer->read_ahead. */ |
01ef6563 | 1219 | if (skip_line_comment (pfile) && CPP_OPTION (pfile, warn_comments)) |
50410426 | 1220 | cpp_warning (pfile, "multi-line comment"); |
1c6d33ef | 1221 | } |
0d9f234d | 1222 | |
1c6d33ef NB |
1223 | /* Skipping the comment has updated buffer->read_ahead. */ |
1224 | if (!pfile->state.save_comments) | |
1225 | { | |
1226 | result->flags |= PREV_WHITE; | |
5fddcffc | 1227 | goto update_tokens_line; |
0d9f234d | 1228 | } |
1c6d33ef NB |
1229 | |
1230 | /* Save the comment as a token in its own right. */ | |
1231 | save_comment (pfile, result, comment_start); | |
bdcbe496 | 1232 | break; |
0d9f234d NB |
1233 | |
1234 | case '<': | |
1235 | if (pfile->state.angled_headers) | |
1236 | { | |
1237 | result->type = CPP_HEADER_NAME; | |
1238 | c = '>'; /* terminator. */ | |
1239 | goto make_string; | |
1240 | } | |
45b966db | 1241 | |
0d9f234d | 1242 | result->type = CPP_LESS; |
29401c30 | 1243 | c = get_effective_char (pfile); |
0d9f234d NB |
1244 | if (c == '=') |
1245 | ACCEPT_CHAR (CPP_LESS_EQ); | |
1246 | else if (c == '<') | |
1247 | { | |
1248 | ACCEPT_CHAR (CPP_LSHIFT); | |
29401c30 | 1249 | if (get_effective_char (pfile) == '=') |
0d9f234d NB |
1250 | ACCEPT_CHAR (CPP_LSHIFT_EQ); |
1251 | } | |
1252 | else if (c == '?' && CPP_OPTION (pfile, cplusplus)) | |
1253 | { | |
1254 | ACCEPT_CHAR (CPP_MIN); | |
29401c30 | 1255 | if (get_effective_char (pfile) == '=') |
0d9f234d NB |
1256 | ACCEPT_CHAR (CPP_MIN_EQ); |
1257 | } | |
1258 | else if (c == ':' && CPP_OPTION (pfile, digraphs)) | |
1259 | { | |
1260 | ACCEPT_CHAR (CPP_OPEN_SQUARE); | |
1261 | result->flags |= DIGRAPH; | |
1262 | } | |
1263 | else if (c == '%' && CPP_OPTION (pfile, digraphs)) | |
1264 | { | |
1265 | ACCEPT_CHAR (CPP_OPEN_BRACE); | |
1266 | result->flags |= DIGRAPH; | |
1267 | } | |
1268 | break; | |
1269 | ||
1270 | case '>': | |
1271 | result->type = CPP_GREATER; | |
29401c30 | 1272 | c = get_effective_char (pfile); |
0d9f234d NB |
1273 | if (c == '=') |
1274 | ACCEPT_CHAR (CPP_GREATER_EQ); | |
1275 | else if (c == '>') | |
1276 | { | |
1277 | ACCEPT_CHAR (CPP_RSHIFT); | |
29401c30 | 1278 | if (get_effective_char (pfile) == '=') |
0d9f234d NB |
1279 | ACCEPT_CHAR (CPP_RSHIFT_EQ); |
1280 | } | |
1281 | else if (c == '?' && CPP_OPTION (pfile, cplusplus)) | |
1282 | { | |
1283 | ACCEPT_CHAR (CPP_MAX); | |
29401c30 | 1284 | if (get_effective_char (pfile) == '=') |
0d9f234d NB |
1285 | ACCEPT_CHAR (CPP_MAX_EQ); |
1286 | } | |
1287 | break; | |
1288 | ||
cbcff6df | 1289 | case '%': |
29401c30 | 1290 | lex_percent (pfile, result); |
0d9f234d NB |
1291 | break; |
1292 | ||
cbcff6df NB |
1293 | case '.': |
1294 | lex_dot (pfile, result); | |
0d9f234d | 1295 | break; |
45b966db | 1296 | |
0d9f234d NB |
1297 | case '+': |
1298 | result->type = CPP_PLUS; | |
29401c30 | 1299 | c = get_effective_char (pfile); |
0d9f234d NB |
1300 | if (c == '=') |
1301 | ACCEPT_CHAR (CPP_PLUS_EQ); | |
1302 | else if (c == '+') | |
1303 | ACCEPT_CHAR (CPP_PLUS_PLUS); | |
1304 | break; | |
04e3ec78 | 1305 | |
0d9f234d NB |
1306 | case '-': |
1307 | result->type = CPP_MINUS; | |
29401c30 | 1308 | c = get_effective_char (pfile); |
0d9f234d NB |
1309 | if (c == '>') |
1310 | { | |
1311 | ACCEPT_CHAR (CPP_DEREF); | |
1312 | if (CPP_OPTION (pfile, cplusplus) | |
29401c30 | 1313 | && get_effective_char (pfile) == '*') |
0d9f234d NB |
1314 | ACCEPT_CHAR (CPP_DEREF_STAR); |
1315 | } | |
1316 | else if (c == '=') | |
1317 | ACCEPT_CHAR (CPP_MINUS_EQ); | |
1318 | else if (c == '-') | |
1319 | ACCEPT_CHAR (CPP_MINUS_MINUS); | |
1320 | break; | |
45b966db | 1321 | |
0d9f234d NB |
1322 | case '*': |
1323 | result->type = CPP_MULT; | |
29401c30 | 1324 | if (get_effective_char (pfile) == '=') |
0d9f234d NB |
1325 | ACCEPT_CHAR (CPP_MULT_EQ); |
1326 | break; | |
04e3ec78 | 1327 | |
0d9f234d NB |
1328 | case '=': |
1329 | result->type = CPP_EQ; | |
29401c30 | 1330 | if (get_effective_char (pfile) == '=') |
0d9f234d NB |
1331 | ACCEPT_CHAR (CPP_EQ_EQ); |
1332 | break; | |
f8f769ea | 1333 | |
0d9f234d NB |
1334 | case '!': |
1335 | result->type = CPP_NOT; | |
29401c30 | 1336 | if (get_effective_char (pfile) == '=') |
0d9f234d NB |
1337 | ACCEPT_CHAR (CPP_NOT_EQ); |
1338 | break; | |
45b966db | 1339 | |
0d9f234d NB |
1340 | case '&': |
1341 | result->type = CPP_AND; | |
29401c30 | 1342 | c = get_effective_char (pfile); |
0d9f234d NB |
1343 | if (c == '=') |
1344 | ACCEPT_CHAR (CPP_AND_EQ); | |
1345 | else if (c == '&') | |
1346 | ACCEPT_CHAR (CPP_AND_AND); | |
1347 | break; | |
1348 | ||
1349 | case '#': | |
a949941c | 1350 | result->type = CPP_HASH; |
5fddcffc NB |
1351 | if (get_effective_char (pfile) == '#') |
1352 | ACCEPT_CHAR (CPP_PASTE); | |
0d9f234d | 1353 | break; |
45b966db | 1354 | |
0d9f234d NB |
1355 | case '|': |
1356 | result->type = CPP_OR; | |
29401c30 | 1357 | c = get_effective_char (pfile); |
0d9f234d NB |
1358 | if (c == '=') |
1359 | ACCEPT_CHAR (CPP_OR_EQ); | |
1360 | else if (c == '|') | |
1361 | ACCEPT_CHAR (CPP_OR_OR); | |
1362 | break; | |
45b966db | 1363 | |
0d9f234d NB |
1364 | case '^': |
1365 | result->type = CPP_XOR; | |
29401c30 | 1366 | if (get_effective_char (pfile) == '=') |
0d9f234d NB |
1367 | ACCEPT_CHAR (CPP_XOR_EQ); |
1368 | break; | |
45b966db | 1369 | |
0d9f234d NB |
1370 | case ':': |
1371 | result->type = CPP_COLON; | |
29401c30 | 1372 | c = get_effective_char (pfile); |
0d9f234d NB |
1373 | if (c == ':' && CPP_OPTION (pfile, cplusplus)) |
1374 | ACCEPT_CHAR (CPP_SCOPE); | |
1375 | else if (c == '>' && CPP_OPTION (pfile, digraphs)) | |
1376 | { | |
1377 | result->flags |= DIGRAPH; | |
1378 | ACCEPT_CHAR (CPP_CLOSE_SQUARE); | |
1379 | } | |
1380 | break; | |
45b966db | 1381 | |
0d9f234d NB |
1382 | case '~': result->type = CPP_COMPL; break; |
1383 | case ',': result->type = CPP_COMMA; break; | |
1384 | case '(': result->type = CPP_OPEN_PAREN; break; | |
1385 | case ')': result->type = CPP_CLOSE_PAREN; break; | |
1386 | case '[': result->type = CPP_OPEN_SQUARE; break; | |
1387 | case ']': result->type = CPP_CLOSE_SQUARE; break; | |
1388 | case '{': result->type = CPP_OPEN_BRACE; break; | |
1389 | case '}': result->type = CPP_CLOSE_BRACE; break; | |
1390 | case ';': result->type = CPP_SEMICOLON; break; | |
1391 | ||
cc937581 ZW |
1392 | /* @ is a punctuator in Objective C. */ |
1393 | case '@': result->type = CPP_ATSIGN; break; | |
0d9f234d NB |
1394 | |
1395 | random_char: | |
1396 | default: | |
1397 | result->type = CPP_OTHER; | |
6c53ebff | 1398 | result->val.c = c; |
0d9f234d NB |
1399 | break; |
1400 | } | |
bdcbe496 NB |
1401 | |
1402 | return result; | |
0d9f234d NB |
1403 | } |
1404 | ||
93c80368 NB |
1405 | /* An upper bound on the number of bytes needed to spell a token, |
1406 | including preceding whitespace. */ | |
1407 | unsigned int | |
1408 | cpp_token_len (token) | |
1409 | const cpp_token *token; | |
0d9f234d | 1410 | { |
93c80368 | 1411 | unsigned int len; |
6d2c2047 | 1412 | |
93c80368 | 1413 | switch (TOKEN_SPELL (token)) |
041c3194 | 1414 | { |
a28c5035 | 1415 | default: len = 0; break; |
47ad4138 | 1416 | case SPELL_NUMBER: |
a28c5035 NB |
1417 | case SPELL_STRING: len = token->val.str.len; break; |
1418 | case SPELL_IDENT: len = NODE_LEN (token->val.node); break; | |
041c3194 | 1419 | } |
47ad4138 | 1420 | /* 1 for whitespace, 4 for comment delimiters. */ |
93c80368 | 1421 | return len + 5; |
6d2c2047 ZW |
1422 | } |
1423 | ||
041c3194 | 1424 | /* Write the spelling of a token TOKEN to BUFFER. The buffer must |
cf00a885 ZW |
1425 | already contain the enough space to hold the token's spelling. |
1426 | Returns a pointer to the character after the last character | |
1427 | written. */ | |
93c80368 NB |
1428 | unsigned char * |
1429 | cpp_spell_token (pfile, token, buffer) | |
041c3194 ZW |
1430 | cpp_reader *pfile; /* Would be nice to be rid of this... */ |
1431 | const cpp_token *token; | |
1432 | unsigned char *buffer; | |
1433 | { | |
96be6998 | 1434 | switch (TOKEN_SPELL (token)) |
041c3194 ZW |
1435 | { |
1436 | case SPELL_OPERATOR: | |
1437 | { | |
1438 | const unsigned char *spelling; | |
1439 | unsigned char c; | |
d6d5f795 | 1440 | |
041c3194 | 1441 | if (token->flags & DIGRAPH) |
37b8524c JDA |
1442 | spelling |
1443 | = digraph_spellings[(int) token->type - (int) CPP_FIRST_DIGRAPH]; | |
92936ecf ZW |
1444 | else if (token->flags & NAMED_OP) |
1445 | goto spell_ident; | |
041c3194 | 1446 | else |
96be6998 | 1447 | spelling = TOKEN_NAME (token); |
041c3194 ZW |
1448 | |
1449 | while ((c = *spelling++) != '\0') | |
1450 | *buffer++ = c; | |
1451 | } | |
1452 | break; | |
d6d5f795 | 1453 | |
47ad4138 ZW |
1454 | case SPELL_CHAR: |
1455 | *buffer++ = token->val.c; | |
1456 | break; | |
1457 | ||
1458 | spell_ident: | |
041c3194 | 1459 | case SPELL_IDENT: |
a28c5035 NB |
1460 | memcpy (buffer, NODE_NAME (token->val.node), NODE_LEN (token->val.node)); |
1461 | buffer += NODE_LEN (token->val.node); | |
041c3194 | 1462 | break; |
d6d5f795 | 1463 | |
47ad4138 ZW |
1464 | case SPELL_NUMBER: |
1465 | memcpy (buffer, token->val.str.text, token->val.str.len); | |
1466 | buffer += token->val.str.len; | |
1467 | break; | |
1468 | ||
041c3194 ZW |
1469 | case SPELL_STRING: |
1470 | { | |
ba89d661 ZW |
1471 | int left, right, tag; |
1472 | switch (token->type) | |
1473 | { | |
1474 | case CPP_STRING: left = '"'; right = '"'; tag = '\0'; break; | |
1475 | case CPP_WSTRING: left = '"'; right = '"'; tag = 'L'; break; | |
ba89d661 ZW |
1476 | case CPP_CHAR: left = '\''; right = '\''; tag = '\0'; break; |
1477 | case CPP_WCHAR: left = '\''; right = '\''; tag = 'L'; break; | |
1478 | case CPP_HEADER_NAME: left = '<'; right = '>'; tag = '\0'; break; | |
47ad4138 ZW |
1479 | default: |
1480 | cpp_ice (pfile, "unknown string token %s\n", TOKEN_NAME (token)); | |
1481 | return buffer; | |
ba89d661 ZW |
1482 | } |
1483 | if (tag) *buffer++ = tag; | |
47ad4138 | 1484 | *buffer++ = left; |
bfb9dc7f ZW |
1485 | memcpy (buffer, token->val.str.text, token->val.str.len); |
1486 | buffer += token->val.str.len; | |
47ad4138 | 1487 | *buffer++ = right; |
041c3194 ZW |
1488 | } |
1489 | break; | |
d6d5f795 | 1490 | |
041c3194 | 1491 | case SPELL_NONE: |
96be6998 | 1492 | cpp_ice (pfile, "Unspellable token %s", TOKEN_NAME (token)); |
041c3194 ZW |
1493 | break; |
1494 | } | |
d6d5f795 | 1495 | |
041c3194 ZW |
1496 | return buffer; |
1497 | } | |
d6d5f795 | 1498 | |
93c80368 NB |
1499 | /* Returns a token as a null-terminated string. The string is |
1500 | temporary, and automatically freed later. Useful for diagnostics. */ | |
1501 | unsigned char * | |
1502 | cpp_token_as_text (pfile, token) | |
c5a04734 | 1503 | cpp_reader *pfile; |
041c3194 | 1504 | const cpp_token *token; |
c5a04734 | 1505 | { |
93c80368 | 1506 | unsigned int len = cpp_token_len (token); |
ece54d54 | 1507 | unsigned char *start = _cpp_unaligned_alloc (pfile, len), *end; |
c5a04734 | 1508 | |
93c80368 NB |
1509 | end = cpp_spell_token (pfile, token, start); |
1510 | end[0] = '\0'; | |
c5a04734 | 1511 | |
93c80368 NB |
1512 | return start; |
1513 | } | |
c5a04734 | 1514 | |
93c80368 NB |
1515 | /* Used by C front ends. Should really move to using cpp_token_as_text. */ |
1516 | const char * | |
1517 | cpp_type2name (type) | |
1518 | enum cpp_ttype type; | |
1519 | { | |
1520 | return (const char *) token_spellings[type].name; | |
1521 | } | |
c5a04734 | 1522 | |
4ed5bcfb NB |
1523 | /* Writes the spelling of token to FP, without any preceding space. |
1524 | Separated from cpp_spell_token for efficiency - to avoid stdio | |
1525 | double-buffering. */ | |
93c80368 NB |
1526 | void |
1527 | cpp_output_token (token, fp) | |
1528 | const cpp_token *token; | |
1529 | FILE *fp; | |
1530 | { | |
93c80368 | 1531 | switch (TOKEN_SPELL (token)) |
c5a04734 | 1532 | { |
93c80368 NB |
1533 | case SPELL_OPERATOR: |
1534 | { | |
1535 | const unsigned char *spelling; | |
3b681e9d | 1536 | int c; |
c5a04734 | 1537 | |
93c80368 | 1538 | if (token->flags & DIGRAPH) |
37b8524c JDA |
1539 | spelling |
1540 | = digraph_spellings[(int) token->type - (int) CPP_FIRST_DIGRAPH]; | |
93c80368 NB |
1541 | else if (token->flags & NAMED_OP) |
1542 | goto spell_ident; | |
1543 | else | |
1544 | spelling = TOKEN_NAME (token); | |
041c3194 | 1545 | |
3b681e9d ZW |
1546 | c = *spelling; |
1547 | do | |
1548 | putc (c, fp); | |
1549 | while ((c = *++spelling) != '\0'); | |
93c80368 NB |
1550 | } |
1551 | break; | |
041c3194 | 1552 | |
47ad4138 ZW |
1553 | case SPELL_CHAR: |
1554 | putc (token->val.c, fp); | |
1555 | break; | |
1556 | ||
93c80368 NB |
1557 | spell_ident: |
1558 | case SPELL_IDENT: | |
3b681e9d | 1559 | fwrite (NODE_NAME (token->val.node), 1, NODE_LEN (token->val.node), fp); |
93c80368 | 1560 | break; |
041c3194 | 1561 | |
47ad4138 ZW |
1562 | case SPELL_NUMBER: |
1563 | fwrite (token->val.str.text, 1, token->val.str.len, fp); | |
1564 | break; | |
1565 | ||
93c80368 NB |
1566 | case SPELL_STRING: |
1567 | { | |
1568 | int left, right, tag; | |
1569 | switch (token->type) | |
1570 | { | |
1571 | case CPP_STRING: left = '"'; right = '"'; tag = '\0'; break; | |
1572 | case CPP_WSTRING: left = '"'; right = '"'; tag = 'L'; break; | |
93c80368 NB |
1573 | case CPP_CHAR: left = '\''; right = '\''; tag = '\0'; break; |
1574 | case CPP_WCHAR: left = '\''; right = '\''; tag = 'L'; break; | |
1575 | case CPP_HEADER_NAME: left = '<'; right = '>'; tag = '\0'; break; | |
47ad4138 ZW |
1576 | default: |
1577 | fprintf (stderr, "impossible STRING token %s\n", TOKEN_NAME (token)); | |
1578 | return; | |
93c80368 NB |
1579 | } |
1580 | if (tag) putc (tag, fp); | |
47ad4138 | 1581 | putc (left, fp); |
93c80368 | 1582 | fwrite (token->val.str.text, 1, token->val.str.len, fp); |
47ad4138 | 1583 | putc (right, fp); |
93c80368 NB |
1584 | } |
1585 | break; | |
c5a04734 | 1586 | |
93c80368 NB |
1587 | case SPELL_NONE: |
1588 | /* An error, most probably. */ | |
1589 | break; | |
041c3194 | 1590 | } |
c5a04734 ZW |
1591 | } |
1592 | ||
93c80368 NB |
1593 | /* Compare two tokens. */ |
1594 | int | |
1595 | _cpp_equiv_tokens (a, b) | |
1596 | const cpp_token *a, *b; | |
c5a04734 | 1597 | { |
93c80368 NB |
1598 | if (a->type == b->type && a->flags == b->flags) |
1599 | switch (TOKEN_SPELL (a)) | |
1600 | { | |
1601 | default: /* Keep compiler happy. */ | |
1602 | case SPELL_OPERATOR: | |
1603 | return 1; | |
1604 | case SPELL_CHAR: | |
6c53ebff | 1605 | return a->val.c == b->val.c; /* Character. */ |
93c80368 | 1606 | case SPELL_NONE: |
56051c0a | 1607 | return (a->type != CPP_MACRO_ARG || a->val.arg_no == b->val.arg_no); |
93c80368 NB |
1608 | case SPELL_IDENT: |
1609 | return a->val.node == b->val.node; | |
47ad4138 | 1610 | case SPELL_NUMBER: |
93c80368 NB |
1611 | case SPELL_STRING: |
1612 | return (a->val.str.len == b->val.str.len | |
1613 | && !memcmp (a->val.str.text, b->val.str.text, | |
1614 | a->val.str.len)); | |
1615 | } | |
c5a04734 | 1616 | |
041c3194 ZW |
1617 | return 0; |
1618 | } | |
1619 | ||
93c80368 NB |
1620 | /* Returns nonzero if a space should be inserted to avoid an |
1621 | accidental token paste for output. For simplicity, it is | |
1622 | conservative, and occasionally advises a space where one is not | |
1623 | needed, e.g. "." and ".2". */ | |
041c3194 | 1624 | |
93c80368 NB |
1625 | int |
1626 | cpp_avoid_paste (pfile, token1, token2) | |
c5a04734 | 1627 | cpp_reader *pfile; |
93c80368 | 1628 | const cpp_token *token1, *token2; |
c5a04734 | 1629 | { |
93c80368 NB |
1630 | enum cpp_ttype a = token1->type, b = token2->type; |
1631 | cppchar_t c; | |
c5a04734 | 1632 | |
93c80368 NB |
1633 | if (token1->flags & NAMED_OP) |
1634 | a = CPP_NAME; | |
1635 | if (token2->flags & NAMED_OP) | |
1636 | b = CPP_NAME; | |
c5a04734 | 1637 | |
93c80368 NB |
1638 | c = EOF; |
1639 | if (token2->flags & DIGRAPH) | |
37b8524c | 1640 | c = digraph_spellings[(int) b - (int) CPP_FIRST_DIGRAPH][0]; |
93c80368 NB |
1641 | else if (token_spellings[b].category == SPELL_OPERATOR) |
1642 | c = token_spellings[b].name[0]; | |
c5a04734 | 1643 | |
93c80368 | 1644 | /* Quickly get everything that can paste with an '='. */ |
37b8524c | 1645 | if ((int) a <= (int) CPP_LAST_EQ && c == '=') |
93c80368 | 1646 | return 1; |
c5a04734 | 1647 | |
93c80368 | 1648 | switch (a) |
c5a04734 | 1649 | { |
93c80368 NB |
1650 | case CPP_GREATER: return c == '>' || c == '?'; |
1651 | case CPP_LESS: return c == '<' || c == '?' || c == '%' || c == ':'; | |
1652 | case CPP_PLUS: return c == '+'; | |
1653 | case CPP_MINUS: return c == '-' || c == '>'; | |
1654 | case CPP_DIV: return c == '/' || c == '*'; /* Comments. */ | |
1655 | case CPP_MOD: return c == ':' || c == '>'; | |
1656 | case CPP_AND: return c == '&'; | |
1657 | case CPP_OR: return c == '|'; | |
1658 | case CPP_COLON: return c == ':' || c == '>'; | |
1659 | case CPP_DEREF: return c == '*'; | |
26ec42ee | 1660 | case CPP_DOT: return c == '.' || c == '%' || b == CPP_NUMBER; |
93c80368 NB |
1661 | case CPP_HASH: return c == '#' || c == '%'; /* Digraph form. */ |
1662 | case CPP_NAME: return ((b == CPP_NUMBER | |
1663 | && name_p (pfile, &token2->val.str)) | |
1664 | || b == CPP_NAME | |
1665 | || b == CPP_CHAR || b == CPP_STRING); /* L */ | |
1666 | case CPP_NUMBER: return (b == CPP_NUMBER || b == CPP_NAME | |
1667 | || c == '.' || c == '+' || c == '-'); | |
1668 | case CPP_OTHER: return (CPP_OPTION (pfile, objc) | |
6c53ebff | 1669 | && token1->val.c == '@' |
93c80368 NB |
1670 | && (b == CPP_NAME || b == CPP_STRING)); |
1671 | default: break; | |
c5a04734 | 1672 | } |
c5a04734 | 1673 | |
417f3e3a | 1674 | return 0; |
c5a04734 ZW |
1675 | } |
1676 | ||
93c80368 | 1677 | /* Output all the remaining tokens on the current line, and a newline |
4ed5bcfb NB |
1678 | character, to FP. Leading whitespace is removed. If there are |
1679 | macros, special token padding is not performed. */ | |
c5a04734 | 1680 | void |
93c80368 | 1681 | cpp_output_line (pfile, fp) |
c5a04734 | 1682 | cpp_reader *pfile; |
93c80368 | 1683 | FILE *fp; |
c5a04734 | 1684 | { |
4ed5bcfb | 1685 | const cpp_token *token; |
96be6998 | 1686 | |
4ed5bcfb NB |
1687 | token = cpp_get_token (pfile); |
1688 | while (token->type != CPP_EOF) | |
96be6998 | 1689 | { |
4ed5bcfb NB |
1690 | cpp_output_token (token, fp); |
1691 | token = cpp_get_token (pfile); | |
1692 | if (token->flags & PREV_WHITE) | |
1693 | putc (' ', fp); | |
96be6998 ZW |
1694 | } |
1695 | ||
93c80368 | 1696 | putc ('\n', fp); |
041c3194 | 1697 | } |
c5a04734 | 1698 | |
c8a96070 NB |
1699 | /* Returns the value of a hexadecimal digit. */ |
1700 | static unsigned int | |
1701 | hex_digit_value (c) | |
1702 | unsigned int c; | |
1703 | { | |
1704 | if (c >= 'a' && c <= 'f') | |
1705 | return c - 'a' + 10; | |
1706 | if (c >= 'A' && c <= 'F') | |
1707 | return c - 'A' + 10; | |
1708 | if (c >= '0' && c <= '9') | |
1709 | return c - '0'; | |
1710 | abort (); | |
1711 | } | |
1712 | ||
62729350 NB |
1713 | /* Parse a '\uNNNN' or '\UNNNNNNNN' sequence. Returns 1 to indicate |
1714 | failure if cpplib is not parsing C++ or C99. Such failure is | |
1715 | silent, and no variables are updated. Otherwise returns 0, and | |
1716 | warns if -Wtraditional. | |
c8a96070 NB |
1717 | |
1718 | [lex.charset]: The character designated by the universal character | |
1719 | name \UNNNNNNNN is that character whose character short name in | |
1720 | ISO/IEC 10646 is NNNNNNNN; the character designated by the | |
1721 | universal character name \uNNNN is that character whose character | |
1722 | short name in ISO/IEC 10646 is 0000NNNN. If the hexadecimal value | |
1723 | for a universal character name is less than 0x20 or in the range | |
1724 | 0x7F-0x9F (inclusive), or if the universal character name | |
1725 | designates a character in the basic source character set, then the | |
1726 | program is ill-formed. | |
1727 | ||
1728 | We assume that wchar_t is Unicode, so we don't need to do any | |
62729350 | 1729 | mapping. Is this ever wrong? |
c8a96070 | 1730 | |
62729350 NB |
1731 | PC points to the 'u' or 'U', PSTR is points to the byte after PC, |
1732 | LIMIT is the end of the string or charconst. PSTR is updated to | |
1733 | point after the UCS on return, and the UCS is written into PC. */ | |
1734 | ||
1735 | static int | |
1736 | maybe_read_ucs (pfile, pstr, limit, pc) | |
c8a96070 NB |
1737 | cpp_reader *pfile; |
1738 | const unsigned char **pstr; | |
1739 | const unsigned char *limit; | |
62729350 | 1740 | unsigned int *pc; |
c8a96070 NB |
1741 | { |
1742 | const unsigned char *p = *pstr; | |
62729350 NB |
1743 | unsigned int code = 0; |
1744 | unsigned int c = *pc, length; | |
1745 | ||
1746 | /* Only attempt to interpret a UCS for C++ and C99. */ | |
1747 | if (! (CPP_OPTION (pfile, cplusplus) || CPP_OPTION (pfile, c99))) | |
1748 | return 1; | |
c8a96070 | 1749 | |
62729350 NB |
1750 | if (CPP_WTRADITIONAL (pfile)) |
1751 | cpp_warning (pfile, "the meaning of '\\%c' varies with -traditional", c); | |
c8a96070 | 1752 | |
f8710242 NB |
1753 | length = (c == 'u' ? 4: 8); |
1754 | ||
1755 | if ((size_t) (limit - p) < length) | |
1756 | { | |
1757 | cpp_error (pfile, "incomplete universal-character-name"); | |
1758 | /* Skip to the end to avoid more diagnostics. */ | |
1759 | p = limit; | |
1760 | } | |
1761 | else | |
1762 | { | |
1763 | for (; length; length--, p++) | |
c8a96070 | 1764 | { |
f8710242 NB |
1765 | c = *p; |
1766 | if (ISXDIGIT (c)) | |
1767 | code = (code << 4) + hex_digit_value (c); | |
1768 | else | |
1769 | { | |
1770 | cpp_error (pfile, | |
1771 | "non-hex digit '%c' in universal-character-name", c); | |
1772 | /* We shouldn't skip in case there are multibyte chars. */ | |
1773 | break; | |
1774 | } | |
c8a96070 | 1775 | } |
c8a96070 NB |
1776 | } |
1777 | ||
1778 | #ifdef TARGET_EBCDIC | |
1779 | cpp_error (pfile, "universal-character-name on EBCDIC target"); | |
1780 | code = 0x3f; /* EBCDIC invalid character */ | |
1781 | #else | |
f8710242 NB |
1782 | /* True extended characters are OK. */ |
1783 | if (code >= 0xa0 | |
1784 | && !(code & 0x80000000) | |
1785 | && !(code >= 0xD800 && code <= 0xDFFF)) | |
1786 | ; | |
1787 | /* The standard permits $, @ and ` to be specified as UCNs. We use | |
1788 | hex escapes so that this also works with EBCDIC hosts. */ | |
1789 | else if (code == 0x24 || code == 0x40 || code == 0x60) | |
1790 | ; | |
1791 | /* Don't give another error if one occurred above. */ | |
1792 | else if (length == 0) | |
1793 | cpp_error (pfile, "universal-character-name out of range"); | |
c8a96070 NB |
1794 | #endif |
1795 | ||
1796 | *pstr = p; | |
62729350 NB |
1797 | *pc = code; |
1798 | return 0; | |
c8a96070 NB |
1799 | } |
1800 | ||
1801 | /* Interpret an escape sequence, and return its value. PSTR points to | |
1802 | the input pointer, which is just after the backslash. LIMIT is how | |
62729350 NB |
1803 | much text we have. MASK is a bitmask for the precision for the |
1804 | destination type (char or wchar_t). TRADITIONAL, if true, does not | |
1805 | interpret escapes that did not exist in traditional C. | |
c8a96070 | 1806 | |
62729350 NB |
1807 | Handles all relevant diagnostics. */ |
1808 | ||
1809 | unsigned int | |
1810 | cpp_parse_escape (pfile, pstr, limit, mask, traditional) | |
c8a96070 NB |
1811 | cpp_reader *pfile; |
1812 | const unsigned char **pstr; | |
1813 | const unsigned char *limit; | |
62729350 | 1814 | unsigned HOST_WIDE_INT mask; |
c8a96070 NB |
1815 | int traditional; |
1816 | { | |
1817 | int unknown = 0; | |
1818 | const unsigned char *str = *pstr; | |
1819 | unsigned int c = *str++; | |
1820 | ||
1821 | switch (c) | |
1822 | { | |
1823 | case '\\': case '\'': case '"': case '?': break; | |
1824 | case 'b': c = TARGET_BS; break; | |
1825 | case 'f': c = TARGET_FF; break; | |
1826 | case 'n': c = TARGET_NEWLINE; break; | |
1827 | case 'r': c = TARGET_CR; break; | |
1828 | case 't': c = TARGET_TAB; break; | |
1829 | case 'v': c = TARGET_VT; break; | |
1830 | ||
1831 | case '(': case '{': case '[': case '%': | |
1832 | /* '\(', etc, are used at beginning of line to avoid confusing Emacs. | |
1833 | '\%' is used to prevent SCCS from getting confused. */ | |
1834 | unknown = CPP_PEDANTIC (pfile); | |
1835 | break; | |
1836 | ||
1837 | case 'a': | |
1838 | if (CPP_WTRADITIONAL (pfile)) | |
1839 | cpp_warning (pfile, "the meaning of '\\a' varies with -traditional"); | |
1840 | if (!traditional) | |
1841 | c = TARGET_BELL; | |
1842 | break; | |
1843 | ||
1844 | case 'e': case 'E': | |
1845 | if (CPP_PEDANTIC (pfile)) | |
1846 | cpp_pedwarn (pfile, "non-ISO-standard escape sequence, '\\%c'", c); | |
1847 | c = TARGET_ESC; | |
1848 | break; | |
1849 | ||
c8a96070 | 1850 | case 'u': case 'U': |
62729350 | 1851 | unknown = maybe_read_ucs (pfile, &str, limit, &c); |
c8a96070 NB |
1852 | break; |
1853 | ||
1854 | case 'x': | |
1855 | if (CPP_WTRADITIONAL (pfile)) | |
1856 | cpp_warning (pfile, "the meaning of '\\x' varies with -traditional"); | |
1857 | ||
1858 | if (!traditional) | |
1859 | { | |
1860 | unsigned int i = 0, overflow = 0; | |
1861 | int digits_found = 0; | |
1862 | ||
1863 | while (str < limit) | |
1864 | { | |
1865 | c = *str; | |
1866 | if (! ISXDIGIT (c)) | |
1867 | break; | |
1868 | str++; | |
1869 | overflow |= i ^ (i << 4 >> 4); | |
1870 | i = (i << 4) + hex_digit_value (c); | |
1871 | digits_found = 1; | |
1872 | } | |
1873 | ||
1874 | if (!digits_found) | |
1875 | cpp_error (pfile, "\\x used with no following hex digits"); | |
1876 | ||
1877 | if (overflow | (i != (i & mask))) | |
1878 | { | |
1879 | cpp_pedwarn (pfile, "hex escape sequence out of range"); | |
1880 | i &= mask; | |
1881 | } | |
1882 | c = i; | |
1883 | } | |
1884 | break; | |
1885 | ||
1886 | case '0': case '1': case '2': case '3': | |
1887 | case '4': case '5': case '6': case '7': | |
1888 | { | |
1889 | unsigned int i = c - '0'; | |
1890 | int count = 0; | |
1891 | ||
1892 | while (str < limit && ++count < 3) | |
1893 | { | |
1894 | c = *str; | |
1895 | if (c < '0' || c > '7') | |
1896 | break; | |
1897 | str++; | |
1898 | i = (i << 3) + c - '0'; | |
1899 | } | |
1900 | ||
1901 | if (i != (i & mask)) | |
1902 | { | |
1903 | cpp_pedwarn (pfile, "octal escape sequence out of range"); | |
1904 | i &= mask; | |
1905 | } | |
1906 | c = i; | |
1907 | } | |
1908 | break; | |
1909 | ||
1910 | default: | |
1911 | unknown = 1; | |
1912 | break; | |
1913 | } | |
1914 | ||
1915 | if (unknown) | |
1916 | { | |
1917 | if (ISGRAPH (c)) | |
1918 | cpp_pedwarn (pfile, "unknown escape sequence '\\%c'", c); | |
1919 | else | |
1920 | cpp_pedwarn (pfile, "unknown escape sequence: '\\%03o'", c); | |
1921 | } | |
1922 | ||
62729350 NB |
1923 | if (c > mask) |
1924 | cpp_pedwarn (pfile, "escape sequence out of range for character"); | |
1925 | ||
c8a96070 NB |
1926 | *pstr = str; |
1927 | return c; | |
1928 | } | |
1929 | ||
1930 | #ifndef MAX_CHAR_TYPE_SIZE | |
1931 | #define MAX_CHAR_TYPE_SIZE CHAR_TYPE_SIZE | |
1932 | #endif | |
1933 | ||
1934 | #ifndef MAX_WCHAR_TYPE_SIZE | |
1935 | #define MAX_WCHAR_TYPE_SIZE WCHAR_TYPE_SIZE | |
1936 | #endif | |
1937 | ||
1938 | /* Interpret a (possibly wide) character constant in TOKEN. | |
1939 | WARN_MULTI warns about multi-character charconsts, if not | |
1940 | TRADITIONAL. TRADITIONAL also indicates not to interpret escapes | |
1941 | that did not exist in traditional C. PCHARS_SEEN points to a | |
1942 | variable that is filled in with the number of characters seen. */ | |
1943 | HOST_WIDE_INT | |
1944 | cpp_interpret_charconst (pfile, token, warn_multi, traditional, pchars_seen) | |
1945 | cpp_reader *pfile; | |
1946 | const cpp_token *token; | |
1947 | int warn_multi; | |
1948 | int traditional; | |
1949 | unsigned int *pchars_seen; | |
1950 | { | |
1951 | const unsigned char *str = token->val.str.text; | |
1952 | const unsigned char *limit = str + token->val.str.len; | |
1953 | unsigned int chars_seen = 0; | |
1954 | unsigned int width, max_chars, c; | |
2a967f3d NB |
1955 | unsigned HOST_WIDE_INT mask; |
1956 | HOST_WIDE_INT result = 0; | |
c8a96070 NB |
1957 | |
1958 | #ifdef MULTIBYTE_CHARS | |
1959 | (void) local_mbtowc (NULL, NULL, 0); | |
1960 | #endif | |
1961 | ||
1962 | /* Width in bits. */ | |
1963 | if (token->type == CPP_CHAR) | |
1964 | width = MAX_CHAR_TYPE_SIZE; | |
1965 | else | |
1966 | width = MAX_WCHAR_TYPE_SIZE; | |
1967 | ||
1968 | if (width < HOST_BITS_PER_WIDE_INT) | |
1969 | mask = ((unsigned HOST_WIDE_INT) 1 << width) - 1; | |
1970 | else | |
1971 | mask = ~0; | |
1972 | max_chars = HOST_BITS_PER_WIDE_INT / width; | |
1973 | ||
1974 | while (str < limit) | |
1975 | { | |
1976 | #ifdef MULTIBYTE_CHARS | |
1977 | wchar_t wc; | |
1978 | int char_len; | |
1979 | ||
1980 | char_len = local_mbtowc (&wc, str, limit - str); | |
1981 | if (char_len == -1) | |
1982 | { | |
1983 | cpp_warning (pfile, "ignoring invalid multibyte character"); | |
1984 | c = *str++; | |
1985 | } | |
1986 | else | |
1987 | { | |
1988 | str += char_len; | |
1989 | c = wc; | |
1990 | } | |
1991 | #else | |
1992 | c = *str++; | |
1993 | #endif | |
1994 | ||
1995 | if (c == '\\') | |
62729350 | 1996 | c = cpp_parse_escape (pfile, &str, limit, mask, traditional); |
c8a96070 NB |
1997 | |
1998 | #ifdef MAP_CHARACTER | |
1999 | if (ISPRINT (c)) | |
2000 | c = MAP_CHARACTER (c); | |
2001 | #endif | |
2002 | ||
2003 | /* Merge character into result; ignore excess chars. */ | |
2004 | if (++chars_seen <= max_chars) | |
2005 | { | |
2006 | if (width < HOST_BITS_PER_WIDE_INT) | |
2007 | result = (result << width) | (c & mask); | |
2008 | else | |
2009 | result = c; | |
2010 | } | |
2011 | } | |
2012 | ||
2013 | if (chars_seen == 0) | |
2014 | cpp_error (pfile, "empty character constant"); | |
2015 | else if (chars_seen > max_chars) | |
2016 | { | |
2017 | chars_seen = max_chars; | |
f8710242 | 2018 | cpp_warning (pfile, "character constant too long"); |
c8a96070 NB |
2019 | } |
2020 | else if (chars_seen > 1 && !traditional && warn_multi) | |
2021 | cpp_warning (pfile, "multi-character character constant"); | |
2022 | ||
2023 | /* If char type is signed, sign-extend the constant. The | |
2024 | __CHAR_UNSIGNED__ macro is set by the driver if appropriate. */ | |
2025 | if (token->type == CPP_CHAR && chars_seen) | |
2026 | { | |
2027 | unsigned int nbits = chars_seen * width; | |
2028 | unsigned int mask = (unsigned int) ~0 >> (HOST_BITS_PER_INT - nbits); | |
2029 | ||
2030 | if (pfile->spec_nodes.n__CHAR_UNSIGNED__->type == NT_MACRO | |
2031 | || ((result >> (nbits - 1)) & 1) == 0) | |
2032 | result &= mask; | |
2033 | else | |
2034 | result |= ~mask; | |
2035 | } | |
2036 | ||
2037 | *pchars_seen = chars_seen; | |
2038 | return result; | |
2039 | } | |
2040 | ||
1e013d2e NB |
2041 | /* Memory buffers. Changing these three constants can have a dramatic |
2042 | effect on performance. The values here are reasonable defaults, | |
2043 | but might be tuned. If you adjust them, be sure to test across a | |
2044 | range of uses of cpplib, including heavy nested function-like macro | |
2045 | expansion. Also check the change in peak memory usage (NJAMD is a | |
2046 | good tool for this). */ | |
2047 | #define MIN_BUFF_SIZE 8000 | |
2048 | #define BUFF_SIZE_UPPER_BOUND(MIN_SIZE) (8000 + (MIN_SIZE) * 3 / 2) | |
2049 | #define EXTENDED_BUFF_SIZE(BUFF, MIN_EXTRA) \ | |
2050 | (MIN_EXTRA + ((BUFF)->limit - (BUFF)->cur) * 2) | |
417f3e3a | 2051 | |
93c80368 | 2052 | struct dummy |
417f3e3a | 2053 | { |
93c80368 NB |
2054 | char c; |
2055 | union | |
2056 | { | |
2057 | double d; | |
2058 | int *p; | |
2059 | } u; | |
2060 | }; | |
417f3e3a | 2061 | |
93c80368 | 2062 | #define DEFAULT_ALIGNMENT (offsetof (struct dummy, u)) |
b8af0ca5 NB |
2063 | #define CPP_ALIGN(size, align) (((size) + ((align) - 1)) & ~((align) - 1)) |
2064 | ||
c9e7a609 NB |
2065 | /* Create a new allocation buffer. Place the control block at the end |
2066 | of the buffer, so that buffer overflows will cause immediate chaos. */ | |
b8af0ca5 NB |
2067 | static _cpp_buff * |
2068 | new_buff (len) | |
6142088c | 2069 | size_t len; |
b8af0ca5 NB |
2070 | { |
2071 | _cpp_buff *result; | |
ece54d54 | 2072 | unsigned char *base; |
b8af0ca5 | 2073 | |
1e013d2e NB |
2074 | if (len < MIN_BUFF_SIZE) |
2075 | len = MIN_BUFF_SIZE; | |
b8af0ca5 NB |
2076 | len = CPP_ALIGN (len, DEFAULT_ALIGNMENT); |
2077 | ||
2078 | base = xmalloc (len + sizeof (_cpp_buff)); | |
2079 | result = (_cpp_buff *) (base + len); | |
2080 | result->base = base; | |
2081 | result->cur = base; | |
2082 | result->limit = base + len; | |
2083 | result->next = NULL; | |
2084 | return result; | |
2085 | } | |
2086 | ||
2087 | /* Place a chain of unwanted allocation buffers on the free list. */ | |
2088 | void | |
2089 | _cpp_release_buff (pfile, buff) | |
2090 | cpp_reader *pfile; | |
2091 | _cpp_buff *buff; | |
2092 | { | |
2093 | _cpp_buff *end = buff; | |
2094 | ||
2095 | while (end->next) | |
2096 | end = end->next; | |
2097 | end->next = pfile->free_buffs; | |
2098 | pfile->free_buffs = buff; | |
2099 | } | |
2100 | ||
2101 | /* Return a free buffer of size at least MIN_SIZE. */ | |
2102 | _cpp_buff * | |
2103 | _cpp_get_buff (pfile, min_size) | |
2104 | cpp_reader *pfile; | |
6142088c | 2105 | size_t min_size; |
b8af0ca5 NB |
2106 | { |
2107 | _cpp_buff *result, **p; | |
2108 | ||
2109 | for (p = &pfile->free_buffs;; p = &(*p)->next) | |
2110 | { | |
6142088c | 2111 | size_t size; |
1e013d2e NB |
2112 | |
2113 | if (*p == NULL) | |
b8af0ca5 | 2114 | return new_buff (min_size); |
1e013d2e NB |
2115 | result = *p; |
2116 | size = result->limit - result->base; | |
2117 | /* Return a buffer that's big enough, but don't waste one that's | |
2118 | way too big. */ | |
2119 | if (size >= min_size && size < BUFF_SIZE_UPPER_BOUND (min_size)) | |
b8af0ca5 NB |
2120 | break; |
2121 | } | |
2122 | ||
2123 | *p = result->next; | |
2124 | result->next = NULL; | |
2125 | result->cur = result->base; | |
2126 | return result; | |
2127 | } | |
2128 | ||
4fe9b91c | 2129 | /* Creates a new buffer with enough space to hold the uncommitted |
8c3b2693 NB |
2130 | remaining bytes of BUFF, and at least MIN_EXTRA more bytes. Copies |
2131 | the excess bytes to the new buffer. Chains the new buffer after | |
2132 | BUFF, and returns the new buffer. */ | |
b8af0ca5 | 2133 | _cpp_buff * |
8c3b2693 | 2134 | _cpp_append_extend_buff (pfile, buff, min_extra) |
b8af0ca5 NB |
2135 | cpp_reader *pfile; |
2136 | _cpp_buff *buff; | |
6142088c | 2137 | size_t min_extra; |
b8af0ca5 | 2138 | { |
6142088c | 2139 | size_t size = EXTENDED_BUFF_SIZE (buff, min_extra); |
8c3b2693 | 2140 | _cpp_buff *new_buff = _cpp_get_buff (pfile, size); |
b8af0ca5 | 2141 | |
8c3b2693 NB |
2142 | buff->next = new_buff; |
2143 | memcpy (new_buff->base, buff->cur, BUFF_ROOM (buff)); | |
2144 | return new_buff; | |
2145 | } | |
2146 | ||
4fe9b91c | 2147 | /* Creates a new buffer with enough space to hold the uncommitted |
8c3b2693 NB |
2148 | remaining bytes of the buffer pointed to by BUFF, and at least |
2149 | MIN_EXTRA more bytes. Copies the excess bytes to the new buffer. | |
2150 | Chains the new buffer before the buffer pointed to by BUFF, and | |
2151 | updates the pointer to point to the new buffer. */ | |
2152 | void | |
2153 | _cpp_extend_buff (pfile, pbuff, min_extra) | |
2154 | cpp_reader *pfile; | |
2155 | _cpp_buff **pbuff; | |
2156 | size_t min_extra; | |
2157 | { | |
2158 | _cpp_buff *new_buff, *old_buff = *pbuff; | |
2159 | size_t size = EXTENDED_BUFF_SIZE (old_buff, min_extra); | |
2160 | ||
2161 | new_buff = _cpp_get_buff (pfile, size); | |
2162 | memcpy (new_buff->base, old_buff->cur, BUFF_ROOM (old_buff)); | |
2163 | new_buff->next = old_buff; | |
2164 | *pbuff = new_buff; | |
b8af0ca5 NB |
2165 | } |
2166 | ||
2167 | /* Free a chain of buffers starting at BUFF. */ | |
2168 | void | |
2169 | _cpp_free_buff (buff) | |
2170 | _cpp_buff *buff; | |
2171 | { | |
2172 | _cpp_buff *next; | |
2173 | ||
2174 | for (; buff; buff = next) | |
2175 | { | |
2176 | next = buff->next; | |
2177 | free (buff->base); | |
2178 | } | |
2179 | } | |
417f3e3a | 2180 | |
ece54d54 NB |
2181 | /* Allocate permanent, unaligned storage of length LEN. */ |
2182 | unsigned char * | |
2183 | _cpp_unaligned_alloc (pfile, len) | |
2184 | cpp_reader *pfile; | |
2185 | size_t len; | |
2186 | { | |
2187 | _cpp_buff *buff = pfile->u_buff; | |
2188 | unsigned char *result = buff->cur; | |
2189 | ||
2190 | if (len > (size_t) (buff->limit - result)) | |
2191 | { | |
2192 | buff = _cpp_get_buff (pfile, len); | |
2193 | buff->next = pfile->u_buff; | |
2194 | pfile->u_buff = buff; | |
2195 | result = buff->cur; | |
2196 | } | |
2197 | ||
2198 | buff->cur = result + len; | |
2199 | return result; | |
2200 | } | |
2201 | ||
8c3b2693 | 2202 | /* Allocate permanent, unaligned storage of length LEN. */ |
93c80368 | 2203 | unsigned char * |
8c3b2693 NB |
2204 | _cpp_aligned_alloc (pfile, len) |
2205 | cpp_reader *pfile; | |
2206 | size_t len; | |
3fef5b2b | 2207 | { |
8c3b2693 NB |
2208 | _cpp_buff *buff = pfile->a_buff; |
2209 | unsigned char *result = buff->cur; | |
3fef5b2b | 2210 | |
8c3b2693 | 2211 | if (len > (size_t) (buff->limit - result)) |
3fef5b2b | 2212 | { |
8c3b2693 NB |
2213 | buff = _cpp_get_buff (pfile, len); |
2214 | buff->next = pfile->a_buff; | |
2215 | pfile->a_buff = buff; | |
2216 | result = buff->cur; | |
3fef5b2b | 2217 | } |
041c3194 | 2218 | |
8c3b2693 | 2219 | buff->cur = result + len; |
93c80368 | 2220 | return result; |
041c3194 | 2221 | } |