]>
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, | |
61 | SPELL_STRING, | |
62 | SPELL_NONE | |
f9a0e96c ZW |
63 | }; |
64 | ||
93c80368 | 65 | struct token_spelling |
f9a0e96c | 66 | { |
93c80368 NB |
67 | enum spell_type category; |
68 | const unsigned char *name; | |
f9a0e96c ZW |
69 | }; |
70 | ||
93c80368 NB |
71 | const unsigned char *digraph_spellings [] = {U"%:", U"%:%:", U"<:", |
72 | U":>", U"<%", U"%>"}; | |
73 | ||
74 | #define OP(e, s) { SPELL_OPERATOR, U s }, | |
75 | #define TK(e, s) { s, U STRINGX (e) }, | |
76 | const struct token_spelling token_spellings [N_TTYPES] = {TTYPE_TABLE }; | |
77 | #undef OP | |
78 | #undef TK | |
79 | ||
80 | #define TOKEN_SPELL(token) (token_spellings[(token)->type].category) | |
81 | #define TOKEN_NAME(token) (token_spellings[(token)->type].name) | |
f2d5f0cc | 82 | |
1444f2ed | 83 | static cppchar_t handle_newline PARAMS ((cpp_reader *, cppchar_t)); |
29401c30 NB |
84 | static cppchar_t skip_escaped_newlines PARAMS ((cpp_reader *, cppchar_t)); |
85 | static cppchar_t get_effective_char PARAMS ((cpp_reader *)); | |
0d9f234d | 86 | |
041c3194 | 87 | static int skip_block_comment PARAMS ((cpp_reader *)); |
cbcff6df | 88 | static int skip_line_comment PARAMS ((cpp_reader *)); |
0d9f234d NB |
89 | static void adjust_column PARAMS ((cpp_reader *)); |
90 | static void skip_whitespace PARAMS ((cpp_reader *, cppchar_t)); | |
2c3fcba6 ZW |
91 | static cpp_hashnode *parse_identifier PARAMS ((cpp_reader *)); |
92 | static cpp_hashnode *parse_identifier_slow PARAMS ((cpp_reader *, | |
93 | const U_CHAR *)); | |
93c80368 NB |
94 | static void parse_number PARAMS ((cpp_reader *, cpp_string *, cppchar_t, int)); |
95 | static int unescaped_terminator_p PARAMS ((cpp_reader *, const U_CHAR *)); | |
0d9f234d | 96 | static void parse_string PARAMS ((cpp_reader *, cpp_token *, cppchar_t)); |
93c80368 | 97 | static void unterminated PARAMS ((cpp_reader *, int)); |
0d9f234d NB |
98 | static int trigraph_ok PARAMS ((cpp_reader *, cppchar_t)); |
99 | static void save_comment PARAMS ((cpp_reader *, cpp_token *, const U_CHAR *)); | |
29401c30 | 100 | static void lex_percent PARAMS ((cpp_reader *, cpp_token *)); |
cbcff6df | 101 | static void lex_dot PARAMS ((cpp_reader *, cpp_token *)); |
93c80368 | 102 | static int name_p PARAMS ((cpp_reader *, const cpp_string *)); |
62729350 NB |
103 | static int maybe_read_ucs PARAMS ((cpp_reader *, const unsigned char **, |
104 | const unsigned char *, unsigned int *)); | |
5fddcffc | 105 | static tokenrun *next_tokenrun PARAMS ((tokenrun *)); |
f617b8e2 | 106 | |
93c80368 | 107 | static cpp_chunk *new_chunk PARAMS ((unsigned int)); |
1e013d2e | 108 | static int chunk_suitable PARAMS ((cpp_chunk *, unsigned int)); |
c8a96070 | 109 | static unsigned int hex_digit_value PARAMS ((unsigned int)); |
b8af0ca5 | 110 | static _cpp_buff *new_buff PARAMS ((unsigned int)); |
15dad1d9 | 111 | |
041c3194 | 112 | /* Utility routine: |
9e62c811 | 113 | |
bfb9dc7f ZW |
114 | Compares, the token TOKEN to the NUL-terminated string STRING. |
115 | TOKEN must be a CPP_NAME. Returns 1 for equal, 0 for unequal. */ | |
15dad1d9 | 116 | |
041c3194 | 117 | int |
bfb9dc7f ZW |
118 | cpp_ideq (token, string) |
119 | const cpp_token *token; | |
041c3194 ZW |
120 | const char *string; |
121 | { | |
bfb9dc7f | 122 | if (token->type != CPP_NAME) |
041c3194 | 123 | return 0; |
bfb9dc7f | 124 | |
a28c5035 | 125 | return !ustrcmp (NODE_NAME (token->val.node), (const U_CHAR *) string); |
15dad1d9 | 126 | } |
1368ee70 | 127 | |
0d9f234d NB |
128 | /* Call when meeting a newline. Returns the character after the newline |
129 | (or carriage-return newline combination), or EOF. */ | |
130 | static cppchar_t | |
1444f2ed NB |
131 | handle_newline (pfile, newline_char) |
132 | cpp_reader *pfile; | |
0d9f234d NB |
133 | cppchar_t newline_char; |
134 | { | |
1444f2ed | 135 | cpp_buffer *buffer; |
0d9f234d NB |
136 | cppchar_t next = EOF; |
137 | ||
1444f2ed | 138 | pfile->line++; |
1444f2ed | 139 | buffer = pfile->buffer; |
0d9f234d | 140 | buffer->col_adjust = 0; |
0d9f234d NB |
141 | buffer->line_base = buffer->cur; |
142 | ||
143 | /* Handle CR-LF and LF-CR combinations, get the next character. */ | |
144 | if (buffer->cur < buffer->rlimit) | |
145 | { | |
146 | next = *buffer->cur++; | |
147 | if (next + newline_char == '\r' + '\n') | |
148 | { | |
149 | buffer->line_base = buffer->cur; | |
150 | if (buffer->cur < buffer->rlimit) | |
151 | next = *buffer->cur++; | |
152 | else | |
153 | next = EOF; | |
154 | } | |
155 | } | |
156 | ||
157 | buffer->read_ahead = next; | |
158 | return next; | |
159 | } | |
160 | ||
161 | /* Subroutine of skip_escaped_newlines; called when a trigraph is | |
162 | encountered. It warns if necessary, and returns true if the | |
163 | trigraph should be honoured. FROM_CHAR is the third character of a | |
164 | trigraph, and presumed to be the previous character for position | |
165 | reporting. */ | |
45b966db | 166 | static int |
0d9f234d | 167 | trigraph_ok (pfile, from_char) |
45b966db | 168 | cpp_reader *pfile; |
0d9f234d | 169 | cppchar_t from_char; |
45b966db | 170 | { |
041c3194 ZW |
171 | int accept = CPP_OPTION (pfile, trigraphs); |
172 | ||
cbcff6df NB |
173 | /* Don't warn about trigraphs in comments. */ |
174 | if (CPP_OPTION (pfile, warn_trigraphs) && !pfile->state.lexing_comment) | |
45b966db | 175 | { |
0d9f234d | 176 | cpp_buffer *buffer = pfile->buffer; |
67821e3a | 177 | |
041c3194 | 178 | if (accept) |
67821e3a | 179 | cpp_warning_with_line (pfile, pfile->line, CPP_BUF_COL (buffer) - 2, |
041c3194 | 180 | "trigraph ??%c converted to %c", |
0d9f234d NB |
181 | (int) from_char, |
182 | (int) _cpp_trigraph_map[from_char]); | |
4a5b68a2 NB |
183 | else if (buffer->cur != buffer->last_Wtrigraphs) |
184 | { | |
185 | buffer->last_Wtrigraphs = buffer->cur; | |
67821e3a | 186 | cpp_warning_with_line (pfile, pfile->line, |
4a5b68a2 NB |
187 | CPP_BUF_COL (buffer) - 2, |
188 | "trigraph ??%c ignored", (int) from_char); | |
189 | } | |
45b966db | 190 | } |
0d9f234d | 191 | |
041c3194 | 192 | return accept; |
45b966db ZW |
193 | } |
194 | ||
0d9f234d NB |
195 | /* Assumes local variables buffer and result. */ |
196 | #define ACCEPT_CHAR(t) \ | |
197 | do { result->type = t; buffer->read_ahead = EOF; } while (0) | |
198 | ||
199 | /* When we move to multibyte character sets, add to these something | |
200 | that saves and restores the state of the multibyte conversion | |
201 | library. This probably involves saving and restoring a "cookie". | |
202 | In the case of glibc it is an 8-byte structure, so is not a high | |
203 | overhead operation. In any case, it's out of the fast path. */ | |
204 | #define SAVE_STATE() do { saved_cur = buffer->cur; } while (0) | |
205 | #define RESTORE_STATE() do { buffer->cur = saved_cur; } while (0) | |
206 | ||
207 | /* Skips any escaped newlines introduced by NEXT, which is either a | |
208 | '?' or a '\\'. Returns the next character, which will also have | |
a5c3cccd NB |
209 | been placed in buffer->read_ahead. This routine performs |
210 | preprocessing stages 1 and 2 of the ISO C standard. */ | |
0d9f234d | 211 | static cppchar_t |
29401c30 NB |
212 | skip_escaped_newlines (pfile, next) |
213 | cpp_reader *pfile; | |
0d9f234d | 214 | cppchar_t next; |
45b966db | 215 | { |
29401c30 NB |
216 | cpp_buffer *buffer = pfile->buffer; |
217 | ||
a5c3cccd NB |
218 | /* Only do this if we apply stages 1 and 2. */ |
219 | if (!buffer->from_stage3) | |
041c3194 | 220 | { |
a5c3cccd NB |
221 | cppchar_t next1; |
222 | const unsigned char *saved_cur; | |
223 | int space; | |
224 | ||
225 | do | |
0d9f234d | 226 | { |
a5c3cccd NB |
227 | if (buffer->cur == buffer->rlimit) |
228 | break; | |
229 | ||
230 | SAVE_STATE (); | |
231 | if (next == '?') | |
0d9f234d | 232 | { |
a5c3cccd NB |
233 | next1 = *buffer->cur++; |
234 | if (next1 != '?' || buffer->cur == buffer->rlimit) | |
235 | { | |
236 | RESTORE_STATE (); | |
237 | break; | |
238 | } | |
239 | ||
240 | next1 = *buffer->cur++; | |
241 | if (!_cpp_trigraph_map[next1] | |
29401c30 | 242 | || !trigraph_ok (pfile, next1)) |
a5c3cccd NB |
243 | { |
244 | RESTORE_STATE (); | |
245 | break; | |
246 | } | |
247 | ||
248 | /* We have a full trigraph here. */ | |
249 | next = _cpp_trigraph_map[next1]; | |
250 | if (next != '\\' || buffer->cur == buffer->rlimit) | |
251 | break; | |
252 | SAVE_STATE (); | |
253 | } | |
254 | ||
255 | /* We have a backslash, and room for at least one more character. */ | |
256 | space = 0; | |
257 | do | |
258 | { | |
259 | next1 = *buffer->cur++; | |
260 | if (!is_nvspace (next1)) | |
261 | break; | |
262 | space = 1; | |
0d9f234d | 263 | } |
a5c3cccd | 264 | while (buffer->cur < buffer->rlimit); |
041c3194 | 265 | |
a5c3cccd | 266 | if (!is_vspace (next1)) |
0d9f234d NB |
267 | { |
268 | RESTORE_STATE (); | |
269 | break; | |
270 | } | |
45b966db | 271 | |
29401c30 NB |
272 | if (space && !pfile->state.lexing_comment) |
273 | cpp_warning (pfile, "backslash and newline separated by space"); | |
0d9f234d | 274 | |
29401c30 | 275 | next = handle_newline (pfile, next1); |
a5c3cccd | 276 | if (next == EOF) |
29401c30 | 277 | cpp_pedwarn (pfile, "backslash-newline at end of file"); |
0d9f234d | 278 | } |
a5c3cccd | 279 | while (next == '\\' || next == '?'); |
041c3194 | 280 | } |
45b966db | 281 | |
0d9f234d NB |
282 | buffer->read_ahead = next; |
283 | return next; | |
45b966db ZW |
284 | } |
285 | ||
0d9f234d NB |
286 | /* Obtain the next character, after trigraph conversion and skipping |
287 | an arbitrary string of escaped newlines. The common case of no | |
288 | trigraphs or escaped newlines falls through quickly. */ | |
289 | static cppchar_t | |
29401c30 NB |
290 | get_effective_char (pfile) |
291 | cpp_reader *pfile; | |
64aaf407 | 292 | { |
29401c30 | 293 | cpp_buffer *buffer = pfile->buffer; |
0d9f234d NB |
294 | cppchar_t next = EOF; |
295 | ||
296 | if (buffer->cur < buffer->rlimit) | |
297 | { | |
298 | next = *buffer->cur++; | |
299 | ||
300 | /* '?' can introduce trigraphs (and therefore backslash); '\\' | |
301 | can introduce escaped newlines, which we want to skip, or | |
302 | UCNs, which, depending upon lexer state, we will handle in | |
303 | the future. */ | |
304 | if (next == '?' || next == '\\') | |
29401c30 | 305 | next = skip_escaped_newlines (pfile, next); |
0d9f234d NB |
306 | } |
307 | ||
308 | buffer->read_ahead = next; | |
309 | return next; | |
64aaf407 NB |
310 | } |
311 | ||
0d9f234d NB |
312 | /* Skip a C-style block comment. We find the end of the comment by |
313 | seeing if an asterisk is before every '/' we encounter. Returns | |
314 | non-zero if comment terminated by EOF, zero otherwise. */ | |
041c3194 ZW |
315 | static int |
316 | skip_block_comment (pfile) | |
45b966db ZW |
317 | cpp_reader *pfile; |
318 | { | |
041c3194 | 319 | cpp_buffer *buffer = pfile->buffer; |
d8090680 | 320 | cppchar_t c = EOF, prevc = EOF; |
0d9f234d | 321 | |
cbcff6df | 322 | pfile->state.lexing_comment = 1; |
0d9f234d | 323 | while (buffer->cur != buffer->rlimit) |
45b966db | 324 | { |
0d9f234d NB |
325 | prevc = c, c = *buffer->cur++; |
326 | ||
327 | next_char: | |
328 | /* FIXME: For speed, create a new character class of characters | |
93c80368 | 329 | of interest inside block comments. */ |
0d9f234d | 330 | if (c == '?' || c == '\\') |
29401c30 | 331 | c = skip_escaped_newlines (pfile, c); |
041c3194 | 332 | |
0d9f234d NB |
333 | /* People like decorating comments with '*', so check for '/' |
334 | instead for efficiency. */ | |
041c3194 | 335 | if (c == '/') |
45b966db | 336 | { |
0d9f234d NB |
337 | if (prevc == '*') |
338 | break; | |
041c3194 | 339 | |
0d9f234d NB |
340 | /* Warn about potential nested comments, but not if the '/' |
341 | comes immediately before the true comment delimeter. | |
041c3194 | 342 | Don't bother to get it right across escaped newlines. */ |
0d9f234d NB |
343 | if (CPP_OPTION (pfile, warn_comments) |
344 | && buffer->cur != buffer->rlimit) | |
45b966db | 345 | { |
0d9f234d NB |
346 | prevc = c, c = *buffer->cur++; |
347 | if (c == '*' && buffer->cur != buffer->rlimit) | |
348 | { | |
349 | prevc = c, c = *buffer->cur++; | |
350 | if (c != '/') | |
67821e3a NB |
351 | cpp_warning_with_line (pfile, pfile->line, |
352 | CPP_BUF_COL (buffer) - 2, | |
0d9f234d NB |
353 | "\"/*\" within comment"); |
354 | } | |
355 | goto next_char; | |
45b966db | 356 | } |
45b966db | 357 | } |
91fcd158 | 358 | else if (is_vspace (c)) |
45b966db | 359 | { |
1444f2ed | 360 | prevc = c, c = handle_newline (pfile, c); |
0d9f234d | 361 | goto next_char; |
45b966db | 362 | } |
52fadca8 | 363 | else if (c == '\t') |
0d9f234d | 364 | adjust_column (pfile); |
45b966db | 365 | } |
041c3194 | 366 | |
cbcff6df | 367 | pfile->state.lexing_comment = 0; |
0d9f234d NB |
368 | buffer->read_ahead = EOF; |
369 | return c != '/' || prevc != '*'; | |
45b966db ZW |
370 | } |
371 | ||
f9a0e96c | 372 | /* Skip a C++ line comment. Handles escaped newlines. Returns |
0d9f234d NB |
373 | non-zero if a multiline comment. The following new line, if any, |
374 | is left in buffer->read_ahead. */ | |
041c3194 | 375 | static int |
cbcff6df NB |
376 | skip_line_comment (pfile) |
377 | cpp_reader *pfile; | |
45b966db | 378 | { |
cbcff6df | 379 | cpp_buffer *buffer = pfile->buffer; |
67821e3a | 380 | unsigned int orig_line = pfile->line; |
0d9f234d | 381 | cppchar_t c; |
041c3194 | 382 | |
cbcff6df | 383 | pfile->state.lexing_comment = 1; |
0d9f234d | 384 | do |
041c3194 | 385 | { |
0d9f234d NB |
386 | c = EOF; |
387 | if (buffer->cur == buffer->rlimit) | |
388 | break; | |
041c3194 | 389 | |
0d9f234d NB |
390 | c = *buffer->cur++; |
391 | if (c == '?' || c == '\\') | |
29401c30 | 392 | c = skip_escaped_newlines (pfile, c); |
041c3194 | 393 | } |
0d9f234d | 394 | while (!is_vspace (c)); |
45b966db | 395 | |
cbcff6df | 396 | pfile->state.lexing_comment = 0; |
0d9f234d | 397 | buffer->read_ahead = c; /* Leave any newline for caller. */ |
67821e3a | 398 | return orig_line != pfile->line; |
041c3194 | 399 | } |
45b966db | 400 | |
0d9f234d NB |
401 | /* pfile->buffer->cur is one beyond the \t character. Update |
402 | col_adjust so we track the column correctly. */ | |
52fadca8 | 403 | static void |
0d9f234d | 404 | adjust_column (pfile) |
52fadca8 | 405 | cpp_reader *pfile; |
52fadca8 | 406 | { |
0d9f234d NB |
407 | cpp_buffer *buffer = pfile->buffer; |
408 | unsigned int col = CPP_BUF_COL (buffer) - 1; /* Zero-based column. */ | |
52fadca8 NB |
409 | |
410 | /* Round it up to multiple of the tabstop, but subtract 1 since the | |
411 | tab itself occupies a character position. */ | |
0d9f234d NB |
412 | buffer->col_adjust += (CPP_OPTION (pfile, tabstop) |
413 | - col % CPP_OPTION (pfile, tabstop)) - 1; | |
52fadca8 NB |
414 | } |
415 | ||
0d9f234d NB |
416 | /* Skips whitespace, saving the next non-whitespace character. |
417 | Adjusts pfile->col_adjust to account for tabs. Without this, | |
418 | tokens might be assigned an incorrect column. */ | |
041c3194 | 419 | static void |
0d9f234d | 420 | skip_whitespace (pfile, c) |
041c3194 | 421 | cpp_reader *pfile; |
0d9f234d | 422 | cppchar_t c; |
041c3194 ZW |
423 | { |
424 | cpp_buffer *buffer = pfile->buffer; | |
0d9f234d | 425 | unsigned int warned = 0; |
45b966db | 426 | |
0d9f234d | 427 | do |
041c3194 | 428 | { |
91fcd158 NB |
429 | /* Horizontal space always OK. */ |
430 | if (c == ' ') | |
0d9f234d | 431 | ; |
91fcd158 | 432 | else if (c == '\t') |
0d9f234d NB |
433 | adjust_column (pfile); |
434 | /* Just \f \v or \0 left. */ | |
91fcd158 | 435 | else if (c == '\0') |
041c3194 | 436 | { |
91fcd158 | 437 | if (!warned) |
0d9f234d NB |
438 | { |
439 | cpp_warning (pfile, "null character(s) ignored"); | |
440 | warned = 1; | |
441 | } | |
45b966db | 442 | } |
93c80368 | 443 | else if (pfile->state.in_directive && CPP_PEDANTIC (pfile)) |
67821e3a | 444 | cpp_pedwarn_with_line (pfile, pfile->line, |
91fcd158 NB |
445 | CPP_BUF_COL (buffer), |
446 | "%s in preprocessing directive", | |
447 | c == '\f' ? "form feed" : "vertical tab"); | |
0d9f234d NB |
448 | |
449 | c = EOF; | |
450 | if (buffer->cur == buffer->rlimit) | |
451 | break; | |
452 | c = *buffer->cur++; | |
45b966db | 453 | } |
ec5c56db | 454 | /* We only want non-vertical space, i.e. ' ' \t \f \v \0. */ |
0d9f234d NB |
455 | while (is_nvspace (c)); |
456 | ||
457 | /* Remember the next character. */ | |
458 | buffer->read_ahead = c; | |
041c3194 | 459 | } |
45b966db | 460 | |
93c80368 NB |
461 | /* See if the characters of a number token are valid in a name (no |
462 | '.', '+' or '-'). */ | |
463 | static int | |
464 | name_p (pfile, string) | |
465 | cpp_reader *pfile; | |
466 | const cpp_string *string; | |
467 | { | |
468 | unsigned int i; | |
469 | ||
470 | for (i = 0; i < string->len; i++) | |
471 | if (!is_idchar (string->text[i])) | |
472 | return 0; | |
473 | ||
474 | return 1; | |
475 | } | |
476 | ||
2c3fcba6 ZW |
477 | /* Parse an identifier, skipping embedded backslash-newlines. This is |
478 | a critical inner loop. The common case is an identifier which has | |
479 | not been split by backslash-newline, does not contain a dollar | |
480 | sign, and has already been scanned (roughly 10:1 ratio of | |
481 | seen:unseen identifiers in normal code; the distribution is | |
482 | Poisson-like). Second most common case is a new identifier, not | |
483 | split and no dollar sign. The other possibilities are rare and | |
484 | have been relegated to parse_identifier_slow. */ | |
0d9f234d NB |
485 | |
486 | static cpp_hashnode * | |
2c3fcba6 | 487 | parse_identifier (pfile) |
45b966db | 488 | cpp_reader *pfile; |
45b966db | 489 | { |
93c80368 | 490 | cpp_hashnode *result; |
2c3fcba6 ZW |
491 | const U_CHAR *cur, *rlimit; |
492 | ||
493 | /* Fast-path loop. Skim over a normal identifier. | |
494 | N.B. ISIDNUM does not include $. */ | |
495 | cur = pfile->buffer->cur - 1; | |
496 | rlimit = pfile->buffer->rlimit; | |
497 | do | |
498 | cur++; | |
499 | while (cur < rlimit && ISIDNUM (*cur)); | |
500 | ||
501 | /* Check for slow-path cases. */ | |
502 | if (cur < rlimit && (*cur == '?' || *cur == '\\' || *cur == '$')) | |
503 | result = parse_identifier_slow (pfile, cur); | |
504 | else | |
505 | { | |
506 | const U_CHAR *base = pfile->buffer->cur - 1; | |
507 | result = (cpp_hashnode *) | |
508 | ht_lookup (pfile->hash_table, base, cur - base, HT_ALLOC); | |
509 | pfile->buffer->cur = cur; | |
510 | } | |
511 | ||
512 | /* Rarely, identifiers require diagnostics when lexed. | |
513 | XXX Has to be forced out of the fast path. */ | |
514 | if (__builtin_expect ((result->flags & NODE_DIAGNOSTIC) | |
515 | && !pfile->state.skipping, 0)) | |
516 | { | |
517 | /* It is allowed to poison the same identifier twice. */ | |
518 | if ((result->flags & NODE_POISONED) && !pfile->state.poisoned_ok) | |
519 | cpp_error (pfile, "attempt to use poisoned \"%s\"", | |
520 | NODE_NAME (result)); | |
521 | ||
522 | /* Constraint 6.10.3.5: __VA_ARGS__ should only appear in the | |
523 | replacement list of a variadic macro. */ | |
524 | if (result == pfile->spec_nodes.n__VA_ARGS__ | |
525 | && !pfile->state.va_args_ok) | |
526 | cpp_pedwarn (pfile, | |
527 | "__VA_ARGS__ can only appear in the expansion of a C99 variadic macro"); | |
528 | } | |
529 | ||
530 | return result; | |
531 | } | |
532 | ||
533 | /* Slow path. This handles identifiers which have been split, and | |
534 | identifiers which contain dollar signs. The part of the identifier | |
535 | from PFILE->buffer->cur-1 to CUR has already been scanned. */ | |
536 | static cpp_hashnode * | |
537 | parse_identifier_slow (pfile, cur) | |
538 | cpp_reader *pfile; | |
539 | const U_CHAR *cur; | |
540 | { | |
0d9f234d | 541 | cpp_buffer *buffer = pfile->buffer; |
2c3fcba6 | 542 | const U_CHAR *base = buffer->cur - 1; |
2a967f3d | 543 | struct obstack *stack = &pfile->hash_table->stack; |
2c3fcba6 ZW |
544 | unsigned int c, saw_dollar = 0, len; |
545 | ||
546 | /* Copy the part of the token which is known to be okay. */ | |
547 | obstack_grow (stack, base, cur - base); | |
041c3194 | 548 | |
2c3fcba6 ZW |
549 | /* Now process the part which isn't. We are looking at one of |
550 | '$', '\\', or '?' on entry to this loop. */ | |
551 | c = *cur++; | |
552 | buffer->cur = cur; | |
0d9f234d | 553 | do |
041c3194 | 554 | { |
2c3fcba6 ZW |
555 | while (is_idchar (c)) |
556 | { | |
557 | obstack_1grow (stack, c); | |
45b966db | 558 | |
2c3fcba6 ZW |
559 | if (c == '$') |
560 | saw_dollar++; | |
ba89d661 | 561 | |
2c3fcba6 ZW |
562 | c = EOF; |
563 | if (buffer->cur == buffer->rlimit) | |
564 | break; | |
ba89d661 | 565 | |
2c3fcba6 ZW |
566 | c = *buffer->cur++; |
567 | } | |
ba89d661 | 568 | |
0d9f234d NB |
569 | /* Potential escaped newline? */ |
570 | if (c != '?' && c != '\\') | |
2c3fcba6 | 571 | break; |
29401c30 | 572 | c = skip_escaped_newlines (pfile, c); |
041c3194 | 573 | } |
0d9f234d NB |
574 | while (is_idchar (c)); |
575 | ||
93c80368 NB |
576 | /* Remember the next character. */ |
577 | buffer->read_ahead = c; | |
578 | ||
0d9f234d NB |
579 | /* $ is not a identifier character in the standard, but is commonly |
580 | accepted as an extension. Don't warn about it in skipped | |
581 | conditional blocks. */ | |
cef0d199 | 582 | if (saw_dollar && CPP_PEDANTIC (pfile) && ! pfile->state.skipping) |
0d9f234d NB |
583 | cpp_pedwarn (pfile, "'$' character(s) in identifier"); |
584 | ||
93c80368 | 585 | /* Identifiers are null-terminated. */ |
2a967f3d NB |
586 | len = obstack_object_size (stack); |
587 | obstack_1grow (stack, '\0'); | |
93c80368 | 588 | |
2c3fcba6 | 589 | return (cpp_hashnode *) |
2a967f3d | 590 | ht_lookup (pfile->hash_table, obstack_finish (stack), len, HT_ALLOCED); |
45b966db ZW |
591 | } |
592 | ||
0d9f234d | 593 | /* Parse a number, skipping embedded backslash-newlines. */ |
45b966db | 594 | static void |
93c80368 | 595 | parse_number (pfile, number, c, leading_period) |
45b966db | 596 | cpp_reader *pfile; |
0d9f234d NB |
597 | cpp_string *number; |
598 | cppchar_t c; | |
93c80368 | 599 | int leading_period; |
45b966db | 600 | { |
041c3194 | 601 | cpp_buffer *buffer = pfile->buffer; |
93c80368 | 602 | unsigned char *dest, *limit; |
45b966db | 603 | |
ece54d54 NB |
604 | dest = BUFF_FRONT (pfile->u_buff); |
605 | limit = BUFF_LIMIT (pfile->u_buff); | |
cbcff6df | 606 | |
93c80368 NB |
607 | /* Place a leading period. */ |
608 | if (leading_period) | |
609 | { | |
ece54d54 NB |
610 | if (dest == limit) |
611 | { | |
612 | pfile->u_buff = _cpp_extend_buff (pfile, pfile->u_buff, 1); | |
613 | dest = BUFF_FRONT (pfile->u_buff); | |
614 | limit = BUFF_LIMIT (pfile->u_buff); | |
615 | } | |
93c80368 NB |
616 | *dest++ = '.'; |
617 | } | |
618 | ||
0d9f234d | 619 | do |
041c3194 | 620 | { |
0d9f234d NB |
621 | do |
622 | { | |
93c80368 | 623 | /* Need room for terminating null. */ |
ece54d54 NB |
624 | if ((size_t) (limit - dest) < 2) |
625 | { | |
626 | size_t len_so_far = dest - BUFF_FRONT (pfile->u_buff); | |
627 | pfile->u_buff = _cpp_extend_buff (pfile, pfile->u_buff, 2); | |
628 | dest = BUFF_FRONT (pfile->u_buff) + len_so_far; | |
629 | limit = BUFF_LIMIT (pfile->u_buff); | |
630 | } | |
93c80368 | 631 | *dest++ = c; |
0d9f234d | 632 | |
0d9f234d NB |
633 | c = EOF; |
634 | if (buffer->cur == buffer->rlimit) | |
635 | break; | |
45b966db | 636 | |
0d9f234d NB |
637 | c = *buffer->cur++; |
638 | } | |
93c80368 | 639 | while (is_numchar (c) || c == '.' || VALID_SIGN (c, dest[-1])); |
45b966db | 640 | |
0d9f234d NB |
641 | /* Potential escaped newline? */ |
642 | if (c != '?' && c != '\\') | |
643 | break; | |
29401c30 | 644 | c = skip_escaped_newlines (pfile, c); |
45b966db | 645 | } |
93c80368 | 646 | while (is_numchar (c) || c == '.' || VALID_SIGN (c, dest[-1])); |
cbcff6df | 647 | |
0d9f234d NB |
648 | /* Remember the next character. */ |
649 | buffer->read_ahead = c; | |
64aaf407 | 650 | |
93c80368 NB |
651 | /* Null-terminate the number. */ |
652 | *dest = '\0'; | |
653 | ||
ece54d54 | 654 | number->text = BUFF_FRONT (pfile->u_buff); |
93c80368 | 655 | number->len = dest - number->text; |
ece54d54 | 656 | BUFF_FRONT (pfile->u_buff) = dest + 1; |
0d9f234d NB |
657 | } |
658 | ||
659 | /* Subroutine of parse_string. Emits error for unterminated strings. */ | |
660 | static void | |
93c80368 | 661 | unterminated (pfile, term) |
0d9f234d | 662 | cpp_reader *pfile; |
0d9f234d NB |
663 | int term; |
664 | { | |
665 | cpp_error (pfile, "missing terminating %c character", term); | |
666 | ||
50410426 | 667 | if (term == '\"' && pfile->mls_line && pfile->mls_line != pfile->line) |
041c3194 | 668 | { |
50410426 | 669 | cpp_error_with_line (pfile, pfile->mls_line, pfile->mls_col, |
0d9f234d | 670 | "possible start of unterminated string literal"); |
50410426 | 671 | pfile->mls_line = 0; |
041c3194 | 672 | } |
45b966db ZW |
673 | } |
674 | ||
93c80368 NB |
675 | /* Subroutine of parse_string. */ |
676 | static int | |
677 | unescaped_terminator_p (pfile, dest) | |
678 | cpp_reader *pfile; | |
679 | const unsigned char *dest; | |
680 | { | |
681 | const unsigned char *start, *temp; | |
682 | ||
683 | /* In #include-style directives, terminators are not escapeable. */ | |
684 | if (pfile->state.angled_headers) | |
685 | return 1; | |
686 | ||
ece54d54 | 687 | start = BUFF_FRONT (pfile->u_buff); |
93c80368 NB |
688 | |
689 | /* An odd number of consecutive backslashes represents an escaped | |
690 | terminator. */ | |
691 | for (temp = dest; temp > start && temp[-1] == '\\'; temp--) | |
692 | ; | |
693 | ||
694 | return ((dest - temp) & 1) == 0; | |
695 | } | |
696 | ||
0d9f234d | 697 | /* Parses a string, character constant, or angle-bracketed header file |
7868b4a2 NB |
698 | name. Handles embedded trigraphs and escaped newlines. The stored |
699 | string is guaranteed NUL-terminated, but it is not guaranteed that | |
700 | this is the first NUL since embedded NULs are preserved. | |
45b966db | 701 | |
7868b4a2 | 702 | Multi-line strings are allowed, but they are deprecated. */ |
041c3194 | 703 | static void |
0d9f234d | 704 | parse_string (pfile, token, terminator) |
45b966db | 705 | cpp_reader *pfile; |
041c3194 | 706 | cpp_token *token; |
0d9f234d | 707 | cppchar_t terminator; |
45b966db | 708 | { |
041c3194 | 709 | cpp_buffer *buffer = pfile->buffer; |
93c80368 | 710 | unsigned char *dest, *limit; |
0d9f234d | 711 | cppchar_t c; |
d82fc108 | 712 | bool warned_nulls = false, warned_multi = false; |
0d9f234d | 713 | |
ece54d54 NB |
714 | dest = BUFF_FRONT (pfile->u_buff); |
715 | limit = BUFF_LIMIT (pfile->u_buff); | |
93c80368 | 716 | |
0d9f234d | 717 | for (;;) |
45b966db | 718 | { |
0d9f234d | 719 | if (buffer->cur == buffer->rlimit) |
7868b4a2 NB |
720 | c = EOF; |
721 | else | |
722 | c = *buffer->cur++; | |
723 | ||
724 | have_char: | |
725 | /* We need space for the terminating NUL. */ | |
ece54d54 NB |
726 | if ((size_t) (limit - dest) < 1) |
727 | { | |
728 | size_t len_so_far = dest - BUFF_FRONT (pfile->u_buff); | |
729 | pfile->u_buff = _cpp_extend_buff (pfile, pfile->u_buff, 2); | |
730 | dest = BUFF_FRONT (pfile->u_buff) + len_so_far; | |
731 | limit = BUFF_LIMIT (pfile->u_buff); | |
732 | } | |
7868b4a2 NB |
733 | |
734 | if (c == EOF) | |
0d9f234d | 735 | { |
93c80368 | 736 | unterminated (pfile, terminator); |
0d9f234d NB |
737 | break; |
738 | } | |
0d9f234d | 739 | |
0d9f234d NB |
740 | /* Handle trigraphs, escaped newlines etc. */ |
741 | if (c == '?' || c == '\\') | |
29401c30 | 742 | c = skip_escaped_newlines (pfile, c); |
45b966db | 743 | |
93c80368 | 744 | if (c == terminator && unescaped_terminator_p (pfile, dest)) |
45b966db | 745 | { |
93c80368 NB |
746 | c = EOF; |
747 | break; | |
0d9f234d NB |
748 | } |
749 | else if (is_vspace (c)) | |
750 | { | |
751 | /* In assembly language, silently terminate string and | |
752 | character literals at end of line. This is a kludge | |
753 | around not knowing where comments are. */ | |
bdb05a7b | 754 | if (CPP_OPTION (pfile, lang) == CLK_ASM && terminator != '>') |
0d9f234d | 755 | break; |
45b966db | 756 | |
0d9f234d NB |
757 | /* Character constants and header names may not extend over |
758 | multiple lines. In Standard C, neither may strings. | |
759 | Unfortunately, we accept multiline strings as an | |
16eb2788 NB |
760 | extension, except in #include family directives. */ |
761 | if (terminator != '"' || pfile->state.angled_headers) | |
45b966db | 762 | { |
93c80368 | 763 | unterminated (pfile, terminator); |
0d9f234d | 764 | break; |
45b966db | 765 | } |
45b966db | 766 | |
d82fc108 NB |
767 | if (!warned_multi) |
768 | { | |
769 | warned_multi = true; | |
770 | cpp_pedwarn (pfile, "multi-line string literals are deprecated"); | |
771 | } | |
772 | ||
50410426 NB |
773 | if (pfile->mls_line == 0) |
774 | { | |
775 | pfile->mls_line = token->line; | |
776 | pfile->mls_col = token->col; | |
777 | } | |
0d9f234d | 778 | |
1444f2ed | 779 | c = handle_newline (pfile, c); |
7868b4a2 NB |
780 | *dest++ = '\n'; |
781 | goto have_char; | |
0d9f234d | 782 | } |
d82fc108 | 783 | else if (c == '\0' && !warned_nulls) |
0d9f234d | 784 | { |
d82fc108 NB |
785 | warned_nulls = true; |
786 | cpp_warning (pfile, "null character(s) preserved in literal"); | |
45b966db | 787 | } |
45b966db | 788 | |
93c80368 | 789 | *dest++ = c; |
45b966db ZW |
790 | } |
791 | ||
93c80368 | 792 | /* Remember the next character. */ |
0d9f234d | 793 | buffer->read_ahead = c; |
7868b4a2 | 794 | *dest = '\0'; |
45b966db | 795 | |
ece54d54 NB |
796 | token->val.str.text = BUFF_FRONT (pfile->u_buff); |
797 | token->val.str.len = dest - BUFF_FRONT (pfile->u_buff); | |
798 | BUFF_FRONT (pfile->u_buff) = dest + 1; | |
0d9f234d | 799 | } |
041c3194 | 800 | |
93c80368 | 801 | /* The stored comment includes the comment start and any terminator. */ |
9e62c811 | 802 | static void |
0d9f234d NB |
803 | save_comment (pfile, token, from) |
804 | cpp_reader *pfile; | |
041c3194 ZW |
805 | cpp_token *token; |
806 | const unsigned char *from; | |
9e62c811 | 807 | { |
041c3194 | 808 | unsigned char *buffer; |
0d9f234d | 809 | unsigned int len; |
0d9f234d | 810 | |
1c6d33ef | 811 | len = pfile->buffer->cur - from + 1; /* + 1 for the initial '/'. */ |
3542203b NB |
812 | /* C++ comments probably (not definitely) have moved past a new |
813 | line, which we don't want to save in the comment. */ | |
814 | if (pfile->buffer->read_ahead != EOF) | |
815 | len--; | |
ece54d54 | 816 | buffer = _cpp_unaligned_alloc (pfile, len); |
041c3194 | 817 | |
041c3194 | 818 | token->type = CPP_COMMENT; |
bfb9dc7f | 819 | token->val.str.len = len; |
0d9f234d | 820 | token->val.str.text = buffer; |
45b966db | 821 | |
1c6d33ef NB |
822 | buffer[0] = '/'; |
823 | memcpy (buffer + 1, from, len - 1); | |
0d9f234d | 824 | } |
45b966db | 825 | |
14baae01 | 826 | /* Subroutine of _cpp_lex_direct to handle '%'. A little tricky, since we |
cbcff6df | 827 | want to avoid stepping back when lexing %:%X. */ |
0d9f234d | 828 | static void |
29401c30 NB |
829 | lex_percent (pfile, result) |
830 | cpp_reader *pfile; | |
0d9f234d | 831 | cpp_token *result; |
0d9f234d | 832 | { |
29401c30 | 833 | cpp_buffer *buffer= pfile->buffer; |
cbcff6df NB |
834 | cppchar_t c; |
835 | ||
836 | result->type = CPP_MOD; | |
837 | /* Parsing %:%X could leave an extra character. */ | |
838 | if (buffer->extra_char == EOF) | |
29401c30 | 839 | c = get_effective_char (pfile); |
cbcff6df NB |
840 | else |
841 | { | |
842 | c = buffer->read_ahead = buffer->extra_char; | |
843 | buffer->extra_char = EOF; | |
844 | } | |
845 | ||
846 | if (c == '=') | |
847 | ACCEPT_CHAR (CPP_MOD_EQ); | |
29401c30 | 848 | else if (CPP_OPTION (pfile, digraphs)) |
cbcff6df NB |
849 | { |
850 | if (c == ':') | |
851 | { | |
852 | result->flags |= DIGRAPH; | |
853 | ACCEPT_CHAR (CPP_HASH); | |
29401c30 | 854 | if (get_effective_char (pfile) == '%') |
cbcff6df | 855 | { |
29401c30 | 856 | buffer->extra_char = get_effective_char (pfile); |
cbcff6df NB |
857 | if (buffer->extra_char == ':') |
858 | { | |
859 | buffer->extra_char = EOF; | |
860 | ACCEPT_CHAR (CPP_PASTE); | |
861 | } | |
862 | else | |
863 | /* We'll catch the extra_char when we're called back. */ | |
864 | buffer->read_ahead = '%'; | |
865 | } | |
866 | } | |
867 | else if (c == '>') | |
868 | { | |
869 | result->flags |= DIGRAPH; | |
870 | ACCEPT_CHAR (CPP_CLOSE_BRACE); | |
871 | } | |
872 | } | |
873 | } | |
874 | ||
14baae01 | 875 | /* Subroutine of _cpp_lex_direct to handle '.'. This is tricky, since we |
cbcff6df NB |
876 | want to avoid stepping back when lexing '...' or '.123'. In the |
877 | latter case we should also set a flag for parse_number. */ | |
878 | static void | |
879 | lex_dot (pfile, result) | |
880 | cpp_reader *pfile; | |
881 | cpp_token *result; | |
882 | { | |
883 | cpp_buffer *buffer = pfile->buffer; | |
884 | cppchar_t c; | |
885 | ||
886 | /* Parsing ..X could leave an extra character. */ | |
887 | if (buffer->extra_char == EOF) | |
29401c30 | 888 | c = get_effective_char (pfile); |
cbcff6df NB |
889 | else |
890 | { | |
891 | c = buffer->read_ahead = buffer->extra_char; | |
892 | buffer->extra_char = EOF; | |
893 | } | |
0d9f234d | 894 | |
cbcff6df NB |
895 | /* All known character sets have 0...9 contiguous. */ |
896 | if (c >= '0' && c <= '9') | |
897 | { | |
898 | result->type = CPP_NUMBER; | |
93c80368 | 899 | parse_number (pfile, &result->val.str, c, 1); |
cbcff6df | 900 | } |
041c3194 | 901 | else |
ea4a453b | 902 | { |
cbcff6df NB |
903 | result->type = CPP_DOT; |
904 | if (c == '.') | |
905 | { | |
29401c30 | 906 | buffer->extra_char = get_effective_char (pfile); |
cbcff6df NB |
907 | if (buffer->extra_char == '.') |
908 | { | |
909 | buffer->extra_char = EOF; | |
910 | ACCEPT_CHAR (CPP_ELLIPSIS); | |
911 | } | |
912 | else | |
913 | /* We'll catch the extra_char when we're called back. */ | |
914 | buffer->read_ahead = '.'; | |
915 | } | |
916 | else if (c == '*' && CPP_OPTION (pfile, cplusplus)) | |
917 | ACCEPT_CHAR (CPP_DOT_STAR); | |
ea4a453b | 918 | } |
45b966db ZW |
919 | } |
920 | ||
5fddcffc NB |
921 | /* Allocate COUNT tokens for RUN. */ |
922 | void | |
923 | _cpp_init_tokenrun (run, count) | |
924 | tokenrun *run; | |
925 | unsigned int count; | |
926 | { | |
927 | run->base = xnewvec (cpp_token, count); | |
928 | run->limit = run->base + count; | |
929 | run->next = NULL; | |
930 | } | |
931 | ||
932 | /* Returns the next tokenrun, or creates one if there is none. */ | |
933 | static tokenrun * | |
934 | next_tokenrun (run) | |
935 | tokenrun *run; | |
936 | { | |
937 | if (run->next == NULL) | |
938 | { | |
939 | run->next = xnew (tokenrun); | |
bdcbe496 | 940 | run->next->prev = run; |
5fddcffc NB |
941 | _cpp_init_tokenrun (run->next, 250); |
942 | } | |
943 | ||
944 | return run->next; | |
945 | } | |
946 | ||
4ed5bcfb NB |
947 | /* Allocate a single token that is invalidated at the same time as the |
948 | rest of the tokens on the line. Has its line and col set to the | |
949 | same as the last lexed token, so that diagnostics appear in the | |
950 | right place. */ | |
951 | cpp_token * | |
952 | _cpp_temp_token (pfile) | |
953 | cpp_reader *pfile; | |
954 | { | |
955 | cpp_token *old, *result; | |
956 | ||
957 | old = pfile->cur_token - 1; | |
958 | if (pfile->cur_token == pfile->cur_run->limit) | |
959 | { | |
960 | pfile->cur_run = next_tokenrun (pfile->cur_run); | |
961 | pfile->cur_token = pfile->cur_run->base; | |
962 | } | |
963 | ||
964 | result = pfile->cur_token++; | |
965 | result->line = old->line; | |
966 | result->col = old->col; | |
967 | return result; | |
968 | } | |
969 | ||
14baae01 NB |
970 | /* Lex a token into RESULT (external interface). Takes care of issues |
971 | like directive handling, token lookahead, multiple include | |
972 | opimisation and skipping. */ | |
345894b4 NB |
973 | const cpp_token * |
974 | _cpp_lex_token (pfile) | |
45b966db | 975 | cpp_reader *pfile; |
5fddcffc | 976 | { |
bdcbe496 | 977 | cpp_token *result; |
5fddcffc | 978 | |
bdcbe496 | 979 | for (;;) |
5fddcffc | 980 | { |
bdcbe496 | 981 | if (pfile->cur_token == pfile->cur_run->limit) |
5fddcffc | 982 | { |
bdcbe496 NB |
983 | pfile->cur_run = next_tokenrun (pfile->cur_run); |
984 | pfile->cur_token = pfile->cur_run->base; | |
5fddcffc NB |
985 | } |
986 | ||
bdcbe496 | 987 | if (pfile->lookaheads) |
14baae01 NB |
988 | { |
989 | pfile->lookaheads--; | |
990 | result = pfile->cur_token++; | |
991 | } | |
bdcbe496 | 992 | else |
14baae01 | 993 | result = _cpp_lex_direct (pfile); |
bdcbe496 NB |
994 | |
995 | if (result->flags & BOL) | |
5fddcffc | 996 | { |
bdcbe496 NB |
997 | /* Is this a directive. If _cpp_handle_directive returns |
998 | false, it is an assembler #. */ | |
999 | if (result->type == CPP_HASH | |
1000 | && !pfile->state.parsing_args | |
1001 | && _cpp_handle_directive (pfile, result->flags & PREV_WHITE)) | |
1002 | continue; | |
97293897 NB |
1003 | if (pfile->cb.line_change && !pfile->state.skipping) |
1004 | (*pfile->cb.line_change)(pfile, result, pfile->state.parsing_args); | |
5fddcffc | 1005 | } |
5fddcffc | 1006 | |
bdcbe496 NB |
1007 | /* We don't skip tokens in directives. */ |
1008 | if (pfile->state.in_directive) | |
1009 | break; | |
5fddcffc | 1010 | |
bdcbe496 | 1011 | /* Outside a directive, invalidate controlling macros. At file |
14baae01 | 1012 | EOF, _cpp_lex_direct takes care of popping the buffer, so we never |
bdcbe496 | 1013 | get here and MI optimisation works. */ |
5fddcffc | 1014 | pfile->mi_valid = false; |
bdcbe496 NB |
1015 | |
1016 | if (!pfile->state.skipping || result->type == CPP_EOF) | |
1017 | break; | |
5fddcffc NB |
1018 | } |
1019 | ||
345894b4 | 1020 | return result; |
5fddcffc NB |
1021 | } |
1022 | ||
14baae01 NB |
1023 | /* Lex a token into pfile->cur_token, which is also incremented, to |
1024 | get diagnostics pointing to the correct location. | |
1025 | ||
1026 | Does not handle issues such as token lookahead, multiple-include | |
1027 | optimisation, directives, skipping etc. This function is only | |
1028 | suitable for use by _cpp_lex_token, and in special cases like | |
1029 | lex_expansion_token which doesn't care for any of these issues. | |
1030 | ||
1031 | When meeting a newline, returns CPP_EOF if parsing a directive, | |
1032 | otherwise returns to the start of the token buffer if permissible. | |
1033 | Returns the location of the lexed token. */ | |
1034 | cpp_token * | |
1035 | _cpp_lex_direct (pfile) | |
5fddcffc | 1036 | cpp_reader *pfile; |
45b966db | 1037 | { |
0d9f234d | 1038 | cppchar_t c; |
adb84b42 | 1039 | cpp_buffer *buffer; |
0d9f234d | 1040 | const unsigned char *comment_start; |
14baae01 | 1041 | cpp_token *result = pfile->cur_token++; |
9ec7291f | 1042 | |
5fddcffc | 1043 | fresh_line: |
adb84b42 | 1044 | buffer = pfile->buffer; |
bd969772 NB |
1045 | result->flags = buffer->saved_flags; |
1046 | buffer->saved_flags = 0; | |
5fddcffc | 1047 | update_tokens_line: |
1444f2ed | 1048 | result->line = pfile->line; |
041c3194 | 1049 | |
5fddcffc | 1050 | skipped_white: |
0d9f234d NB |
1051 | c = buffer->read_ahead; |
1052 | if (c == EOF && buffer->cur < buffer->rlimit) | |
5fddcffc NB |
1053 | c = *buffer->cur++; |
1054 | result->col = CPP_BUF_COLUMN (buffer, buffer->cur); | |
0d9f234d | 1055 | buffer->read_ahead = EOF; |
5fddcffc NB |
1056 | |
1057 | trigraph: | |
0d9f234d | 1058 | switch (c) |
45b966db | 1059 | { |
0d9f234d | 1060 | case EOF: |
bdcbe496 | 1061 | buffer->saved_flags = BOL; |
5fddcffc | 1062 | if (!pfile->state.parsing_args && !pfile->state.in_directive) |
ef6e958a | 1063 | { |
bdcbe496 | 1064 | if (buffer->cur != buffer->line_base) |
5fddcffc NB |
1065 | { |
1066 | /* Non-empty files should end in a newline. Don't warn | |
1067 | for command line and _Pragma buffers. */ | |
1068 | if (!buffer->from_stage3) | |
1069 | cpp_pedwarn (pfile, "no newline at end of file"); | |
1070 | handle_newline (pfile, '\n'); | |
7364fdd8 | 1071 | } |
bdcbe496 NB |
1072 | |
1073 | /* Don't pop the last buffer. */ | |
1074 | if (buffer->prev) | |
1075 | { | |
1076 | unsigned char stop = buffer->return_at_eof; | |
1077 | ||
1078 | _cpp_pop_buffer (pfile); | |
1079 | if (!stop) | |
1080 | goto fresh_line; | |
1081 | } | |
ef6e958a | 1082 | } |
0d9f234d | 1083 | result->type = CPP_EOF; |
5fddcffc | 1084 | break; |
45b966db | 1085 | |
0d9f234d NB |
1086 | case ' ': case '\t': case '\f': case '\v': case '\0': |
1087 | skip_whitespace (pfile, c); | |
1088 | result->flags |= PREV_WHITE; | |
5fddcffc | 1089 | goto skipped_white; |
0d9f234d NB |
1090 | |
1091 | case '\n': case '\r': | |
bdcbe496 NB |
1092 | handle_newline (pfile, c); |
1093 | buffer->saved_flags = BOL; | |
1094 | if (! pfile->state.in_directive) | |
45b966db | 1095 | { |
4ed5bcfb NB |
1096 | if (pfile->state.parsing_args == 2) |
1097 | buffer->saved_flags |= PREV_WHITE; | |
bdcbe496 NB |
1098 | if (!pfile->keep_tokens) |
1099 | { | |
1100 | pfile->cur_run = &pfile->base_run; | |
1101 | result = pfile->base_run.base; | |
1102 | pfile->cur_token = result + 1; | |
1103 | } | |
1104 | goto fresh_line; | |
45b966db | 1105 | } |
5fddcffc NB |
1106 | result->type = CPP_EOF; |
1107 | break; | |
46d07497 | 1108 | |
0d9f234d NB |
1109 | case '?': |
1110 | case '\\': | |
1111 | /* These could start an escaped newline, or '?' a trigraph. Let | |
1112 | skip_escaped_newlines do all the work. */ | |
1113 | { | |
67821e3a | 1114 | unsigned int line = pfile->line; |
0d9f234d | 1115 | |
29401c30 | 1116 | c = skip_escaped_newlines (pfile, c); |
67821e3a | 1117 | if (line != pfile->line) |
0d9f234d NB |
1118 | /* We had at least one escaped newline of some sort, and the |
1119 | next character is in buffer->read_ahead. Update the | |
1120 | token's line and column. */ | |
5fddcffc | 1121 | goto update_tokens_line; |
0d9f234d NB |
1122 | |
1123 | /* We are either the original '?' or '\\', or a trigraph. */ | |
1124 | result->type = CPP_QUERY; | |
1125 | buffer->read_ahead = EOF; | |
1126 | if (c == '\\') | |
12c4f523 | 1127 | goto random_char; |
0d9f234d | 1128 | else if (c != '?') |
5fddcffc | 1129 | goto trigraph; |
0d9f234d NB |
1130 | } |
1131 | break; | |
46d07497 | 1132 | |
0d9f234d NB |
1133 | case '0': case '1': case '2': case '3': case '4': |
1134 | case '5': case '6': case '7': case '8': case '9': | |
1135 | result->type = CPP_NUMBER; | |
93c80368 | 1136 | parse_number (pfile, &result->val.str, c, 0); |
0d9f234d | 1137 | break; |
46d07497 | 1138 | |
0d9f234d NB |
1139 | case '$': |
1140 | if (!CPP_OPTION (pfile, dollars_in_ident)) | |
1141 | goto random_char; | |
ec5c56db | 1142 | /* Fall through... */ |
0d9f234d NB |
1143 | |
1144 | case '_': | |
1145 | case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': | |
1146 | case 'g': case 'h': case 'i': case 'j': case 'k': case 'l': | |
1147 | case 'm': case 'n': case 'o': case 'p': case 'q': case 'r': | |
1148 | case 's': case 't': case 'u': case 'v': case 'w': case 'x': | |
1149 | case 'y': case 'z': | |
1150 | case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': | |
1151 | case 'G': case 'H': case 'I': case 'J': case 'K': case 'L': | |
1152 | case 'M': case 'N': case 'O': case 'P': case 'Q': case 'R': | |
1153 | case 'S': case 'T': case 'U': case 'V': case 'W': case 'X': | |
1154 | case 'Y': case 'Z': | |
1155 | result->type = CPP_NAME; | |
2c3fcba6 | 1156 | result->val.node = parse_identifier (pfile); |
0d9f234d NB |
1157 | |
1158 | /* 'L' may introduce wide characters or strings. */ | |
93c80368 | 1159 | if (result->val.node == pfile->spec_nodes.n_L) |
0d9f234d | 1160 | { |
2c3fcba6 ZW |
1161 | c = buffer->read_ahead; |
1162 | if (c == EOF && buffer->cur < buffer->rlimit) | |
1163 | c = *buffer->cur; | |
0d9f234d | 1164 | if (c == '\'' || c == '"') |
ba89d661 | 1165 | { |
2c3fcba6 | 1166 | buffer->cur++; |
0d9f234d NB |
1167 | ACCEPT_CHAR (c == '"' ? CPP_WSTRING: CPP_WCHAR); |
1168 | goto make_string; | |
ba89d661 | 1169 | } |
0d9f234d NB |
1170 | } |
1171 | /* Convert named operators to their proper types. */ | |
93c80368 | 1172 | else if (result->val.node->flags & NODE_OPERATOR) |
0d9f234d NB |
1173 | { |
1174 | result->flags |= NAMED_OP; | |
93c80368 | 1175 | result->type = result->val.node->value.operator; |
0d9f234d NB |
1176 | } |
1177 | break; | |
1178 | ||
1179 | case '\'': | |
1180 | case '"': | |
1181 | result->type = c == '"' ? CPP_STRING: CPP_CHAR; | |
1182 | make_string: | |
1183 | parse_string (pfile, result, c); | |
1184 | break; | |
041c3194 | 1185 | |
0d9f234d | 1186 | case '/': |
1c6d33ef NB |
1187 | /* A potential block or line comment. */ |
1188 | comment_start = buffer->cur; | |
0d9f234d | 1189 | result->type = CPP_DIV; |
29401c30 | 1190 | c = get_effective_char (pfile); |
0d9f234d NB |
1191 | if (c == '=') |
1192 | ACCEPT_CHAR (CPP_DIV_EQ); | |
1c6d33ef NB |
1193 | if (c != '/' && c != '*') |
1194 | break; | |
e61fc951 | 1195 | |
1c6d33ef NB |
1196 | if (c == '*') |
1197 | { | |
0d9f234d | 1198 | if (skip_block_comment (pfile)) |
67821e3a | 1199 | cpp_error (pfile, "unterminated comment"); |
0d9f234d | 1200 | } |
1c6d33ef | 1201 | else |
0d9f234d | 1202 | { |
1c6d33ef NB |
1203 | if (!CPP_OPTION (pfile, cplusplus_comments) |
1204 | && !CPP_IN_SYSTEM_HEADER (pfile)) | |
1205 | break; | |
1206 | ||
bdb05a7b NB |
1207 | /* Warn about comments only if pedantically GNUC89, and not |
1208 | in system headers. */ | |
1209 | if (CPP_OPTION (pfile, lang) == CLK_GNUC89 && CPP_PEDANTIC (pfile) | |
a94c1199 | 1210 | && ! buffer->warned_cplusplus_comments) |
041c3194 | 1211 | { |
1c6d33ef NB |
1212 | cpp_pedwarn (pfile, |
1213 | "C++ style comments are not allowed in ISO C89"); | |
1214 | cpp_pedwarn (pfile, | |
1215 | "(this will be reported only once per input file)"); | |
1216 | buffer->warned_cplusplus_comments = 1; | |
1217 | } | |
0d9f234d | 1218 | |
a94c1199 | 1219 | /* Skip_line_comment updates buffer->read_ahead. */ |
01ef6563 | 1220 | if (skip_line_comment (pfile) && CPP_OPTION (pfile, warn_comments)) |
50410426 | 1221 | cpp_warning (pfile, "multi-line comment"); |
1c6d33ef | 1222 | } |
0d9f234d | 1223 | |
1c6d33ef NB |
1224 | /* Skipping the comment has updated buffer->read_ahead. */ |
1225 | if (!pfile->state.save_comments) | |
1226 | { | |
1227 | result->flags |= PREV_WHITE; | |
5fddcffc | 1228 | goto update_tokens_line; |
0d9f234d | 1229 | } |
1c6d33ef NB |
1230 | |
1231 | /* Save the comment as a token in its own right. */ | |
1232 | save_comment (pfile, result, comment_start); | |
bdcbe496 | 1233 | break; |
0d9f234d NB |
1234 | |
1235 | case '<': | |
1236 | if (pfile->state.angled_headers) | |
1237 | { | |
1238 | result->type = CPP_HEADER_NAME; | |
1239 | c = '>'; /* terminator. */ | |
1240 | goto make_string; | |
1241 | } | |
45b966db | 1242 | |
0d9f234d | 1243 | result->type = CPP_LESS; |
29401c30 | 1244 | c = get_effective_char (pfile); |
0d9f234d NB |
1245 | if (c == '=') |
1246 | ACCEPT_CHAR (CPP_LESS_EQ); | |
1247 | else if (c == '<') | |
1248 | { | |
1249 | ACCEPT_CHAR (CPP_LSHIFT); | |
29401c30 | 1250 | if (get_effective_char (pfile) == '=') |
0d9f234d NB |
1251 | ACCEPT_CHAR (CPP_LSHIFT_EQ); |
1252 | } | |
1253 | else if (c == '?' && CPP_OPTION (pfile, cplusplus)) | |
1254 | { | |
1255 | ACCEPT_CHAR (CPP_MIN); | |
29401c30 | 1256 | if (get_effective_char (pfile) == '=') |
0d9f234d NB |
1257 | ACCEPT_CHAR (CPP_MIN_EQ); |
1258 | } | |
1259 | else if (c == ':' && CPP_OPTION (pfile, digraphs)) | |
1260 | { | |
1261 | ACCEPT_CHAR (CPP_OPEN_SQUARE); | |
1262 | result->flags |= DIGRAPH; | |
1263 | } | |
1264 | else if (c == '%' && CPP_OPTION (pfile, digraphs)) | |
1265 | { | |
1266 | ACCEPT_CHAR (CPP_OPEN_BRACE); | |
1267 | result->flags |= DIGRAPH; | |
1268 | } | |
1269 | break; | |
1270 | ||
1271 | case '>': | |
1272 | result->type = CPP_GREATER; | |
29401c30 | 1273 | c = get_effective_char (pfile); |
0d9f234d NB |
1274 | if (c == '=') |
1275 | ACCEPT_CHAR (CPP_GREATER_EQ); | |
1276 | else if (c == '>') | |
1277 | { | |
1278 | ACCEPT_CHAR (CPP_RSHIFT); | |
29401c30 | 1279 | if (get_effective_char (pfile) == '=') |
0d9f234d NB |
1280 | ACCEPT_CHAR (CPP_RSHIFT_EQ); |
1281 | } | |
1282 | else if (c == '?' && CPP_OPTION (pfile, cplusplus)) | |
1283 | { | |
1284 | ACCEPT_CHAR (CPP_MAX); | |
29401c30 | 1285 | if (get_effective_char (pfile) == '=') |
0d9f234d NB |
1286 | ACCEPT_CHAR (CPP_MAX_EQ); |
1287 | } | |
1288 | break; | |
1289 | ||
cbcff6df | 1290 | case '%': |
29401c30 | 1291 | lex_percent (pfile, result); |
0d9f234d NB |
1292 | break; |
1293 | ||
cbcff6df NB |
1294 | case '.': |
1295 | lex_dot (pfile, result); | |
0d9f234d | 1296 | break; |
45b966db | 1297 | |
0d9f234d NB |
1298 | case '+': |
1299 | result->type = CPP_PLUS; | |
29401c30 | 1300 | c = get_effective_char (pfile); |
0d9f234d NB |
1301 | if (c == '=') |
1302 | ACCEPT_CHAR (CPP_PLUS_EQ); | |
1303 | else if (c == '+') | |
1304 | ACCEPT_CHAR (CPP_PLUS_PLUS); | |
1305 | break; | |
04e3ec78 | 1306 | |
0d9f234d NB |
1307 | case '-': |
1308 | result->type = CPP_MINUS; | |
29401c30 | 1309 | c = get_effective_char (pfile); |
0d9f234d NB |
1310 | if (c == '>') |
1311 | { | |
1312 | ACCEPT_CHAR (CPP_DEREF); | |
1313 | if (CPP_OPTION (pfile, cplusplus) | |
29401c30 | 1314 | && get_effective_char (pfile) == '*') |
0d9f234d NB |
1315 | ACCEPT_CHAR (CPP_DEREF_STAR); |
1316 | } | |
1317 | else if (c == '=') | |
1318 | ACCEPT_CHAR (CPP_MINUS_EQ); | |
1319 | else if (c == '-') | |
1320 | ACCEPT_CHAR (CPP_MINUS_MINUS); | |
1321 | break; | |
45b966db | 1322 | |
0d9f234d NB |
1323 | case '*': |
1324 | result->type = CPP_MULT; | |
29401c30 | 1325 | if (get_effective_char (pfile) == '=') |
0d9f234d NB |
1326 | ACCEPT_CHAR (CPP_MULT_EQ); |
1327 | break; | |
04e3ec78 | 1328 | |
0d9f234d NB |
1329 | case '=': |
1330 | result->type = CPP_EQ; | |
29401c30 | 1331 | if (get_effective_char (pfile) == '=') |
0d9f234d NB |
1332 | ACCEPT_CHAR (CPP_EQ_EQ); |
1333 | break; | |
f8f769ea | 1334 | |
0d9f234d NB |
1335 | case '!': |
1336 | result->type = CPP_NOT; | |
29401c30 | 1337 | if (get_effective_char (pfile) == '=') |
0d9f234d NB |
1338 | ACCEPT_CHAR (CPP_NOT_EQ); |
1339 | break; | |
45b966db | 1340 | |
0d9f234d NB |
1341 | case '&': |
1342 | result->type = CPP_AND; | |
29401c30 | 1343 | c = get_effective_char (pfile); |
0d9f234d NB |
1344 | if (c == '=') |
1345 | ACCEPT_CHAR (CPP_AND_EQ); | |
1346 | else if (c == '&') | |
1347 | ACCEPT_CHAR (CPP_AND_AND); | |
1348 | break; | |
1349 | ||
1350 | case '#': | |
a949941c | 1351 | result->type = CPP_HASH; |
5fddcffc NB |
1352 | if (get_effective_char (pfile) == '#') |
1353 | ACCEPT_CHAR (CPP_PASTE); | |
0d9f234d | 1354 | break; |
45b966db | 1355 | |
0d9f234d NB |
1356 | case '|': |
1357 | result->type = CPP_OR; | |
29401c30 | 1358 | c = get_effective_char (pfile); |
0d9f234d NB |
1359 | if (c == '=') |
1360 | ACCEPT_CHAR (CPP_OR_EQ); | |
1361 | else if (c == '|') | |
1362 | ACCEPT_CHAR (CPP_OR_OR); | |
1363 | break; | |
45b966db | 1364 | |
0d9f234d NB |
1365 | case '^': |
1366 | result->type = CPP_XOR; | |
29401c30 | 1367 | if (get_effective_char (pfile) == '=') |
0d9f234d NB |
1368 | ACCEPT_CHAR (CPP_XOR_EQ); |
1369 | break; | |
45b966db | 1370 | |
0d9f234d NB |
1371 | case ':': |
1372 | result->type = CPP_COLON; | |
29401c30 | 1373 | c = get_effective_char (pfile); |
0d9f234d NB |
1374 | if (c == ':' && CPP_OPTION (pfile, cplusplus)) |
1375 | ACCEPT_CHAR (CPP_SCOPE); | |
1376 | else if (c == '>' && CPP_OPTION (pfile, digraphs)) | |
1377 | { | |
1378 | result->flags |= DIGRAPH; | |
1379 | ACCEPT_CHAR (CPP_CLOSE_SQUARE); | |
1380 | } | |
1381 | break; | |
45b966db | 1382 | |
0d9f234d NB |
1383 | case '~': result->type = CPP_COMPL; break; |
1384 | case ',': result->type = CPP_COMMA; break; | |
1385 | case '(': result->type = CPP_OPEN_PAREN; break; | |
1386 | case ')': result->type = CPP_CLOSE_PAREN; break; | |
1387 | case '[': result->type = CPP_OPEN_SQUARE; break; | |
1388 | case ']': result->type = CPP_CLOSE_SQUARE; break; | |
1389 | case '{': result->type = CPP_OPEN_BRACE; break; | |
1390 | case '}': result->type = CPP_CLOSE_BRACE; break; | |
1391 | case ';': result->type = CPP_SEMICOLON; break; | |
1392 | ||
cc937581 ZW |
1393 | /* @ is a punctuator in Objective C. */ |
1394 | case '@': result->type = CPP_ATSIGN; break; | |
0d9f234d NB |
1395 | |
1396 | random_char: | |
1397 | default: | |
1398 | result->type = CPP_OTHER; | |
6c53ebff | 1399 | result->val.c = c; |
0d9f234d NB |
1400 | break; |
1401 | } | |
bdcbe496 NB |
1402 | |
1403 | return result; | |
0d9f234d NB |
1404 | } |
1405 | ||
93c80368 NB |
1406 | /* An upper bound on the number of bytes needed to spell a token, |
1407 | including preceding whitespace. */ | |
1408 | unsigned int | |
1409 | cpp_token_len (token) | |
1410 | const cpp_token *token; | |
0d9f234d | 1411 | { |
93c80368 | 1412 | unsigned int len; |
6d2c2047 | 1413 | |
93c80368 | 1414 | switch (TOKEN_SPELL (token)) |
041c3194 | 1415 | { |
a28c5035 NB |
1416 | default: len = 0; break; |
1417 | case SPELL_STRING: len = token->val.str.len; break; | |
1418 | case SPELL_IDENT: len = NODE_LEN (token->val.node); break; | |
041c3194 | 1419 | } |
93c80368 NB |
1420 | /* 1 for whitespace, 4 for comment delimeters. */ |
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 | |
041c3194 | 1454 | case SPELL_IDENT: |
92936ecf | 1455 | spell_ident: |
a28c5035 NB |
1456 | memcpy (buffer, NODE_NAME (token->val.node), NODE_LEN (token->val.node)); |
1457 | buffer += NODE_LEN (token->val.node); | |
041c3194 | 1458 | break; |
d6d5f795 | 1459 | |
041c3194 ZW |
1460 | case SPELL_STRING: |
1461 | { | |
ba89d661 ZW |
1462 | int left, right, tag; |
1463 | switch (token->type) | |
1464 | { | |
1465 | case CPP_STRING: left = '"'; right = '"'; tag = '\0'; break; | |
1466 | case CPP_WSTRING: left = '"'; right = '"'; tag = 'L'; break; | |
ba89d661 ZW |
1467 | case CPP_CHAR: left = '\''; right = '\''; tag = '\0'; break; |
1468 | case CPP_WCHAR: left = '\''; right = '\''; tag = 'L'; break; | |
1469 | case CPP_HEADER_NAME: left = '<'; right = '>'; tag = '\0'; break; | |
1470 | default: left = '\0'; right = '\0'; tag = '\0'; break; | |
1471 | } | |
1472 | if (tag) *buffer++ = tag; | |
1473 | if (left) *buffer++ = left; | |
bfb9dc7f ZW |
1474 | memcpy (buffer, token->val.str.text, token->val.str.len); |
1475 | buffer += token->val.str.len; | |
ba89d661 | 1476 | if (right) *buffer++ = right; |
041c3194 ZW |
1477 | } |
1478 | break; | |
d6d5f795 | 1479 | |
041c3194 | 1480 | case SPELL_CHAR: |
6c53ebff | 1481 | *buffer++ = token->val.c; |
041c3194 | 1482 | break; |
d6d5f795 | 1483 | |
041c3194 | 1484 | case SPELL_NONE: |
96be6998 | 1485 | cpp_ice (pfile, "Unspellable token %s", TOKEN_NAME (token)); |
041c3194 ZW |
1486 | break; |
1487 | } | |
d6d5f795 | 1488 | |
041c3194 ZW |
1489 | return buffer; |
1490 | } | |
d6d5f795 | 1491 | |
93c80368 NB |
1492 | /* Returns a token as a null-terminated string. The string is |
1493 | temporary, and automatically freed later. Useful for diagnostics. */ | |
1494 | unsigned char * | |
1495 | cpp_token_as_text (pfile, token) | |
c5a04734 | 1496 | cpp_reader *pfile; |
041c3194 | 1497 | const cpp_token *token; |
c5a04734 | 1498 | { |
93c80368 | 1499 | unsigned int len = cpp_token_len (token); |
ece54d54 | 1500 | unsigned char *start = _cpp_unaligned_alloc (pfile, len), *end; |
c5a04734 | 1501 | |
93c80368 NB |
1502 | end = cpp_spell_token (pfile, token, start); |
1503 | end[0] = '\0'; | |
c5a04734 | 1504 | |
93c80368 NB |
1505 | return start; |
1506 | } | |
c5a04734 | 1507 | |
93c80368 NB |
1508 | /* Used by C front ends. Should really move to using cpp_token_as_text. */ |
1509 | const char * | |
1510 | cpp_type2name (type) | |
1511 | enum cpp_ttype type; | |
1512 | { | |
1513 | return (const char *) token_spellings[type].name; | |
1514 | } | |
c5a04734 | 1515 | |
4ed5bcfb NB |
1516 | /* Writes the spelling of token to FP, without any preceding space. |
1517 | Separated from cpp_spell_token for efficiency - to avoid stdio | |
1518 | double-buffering. */ | |
93c80368 NB |
1519 | void |
1520 | cpp_output_token (token, fp) | |
1521 | const cpp_token *token; | |
1522 | FILE *fp; | |
1523 | { | |
93c80368 | 1524 | switch (TOKEN_SPELL (token)) |
c5a04734 | 1525 | { |
93c80368 NB |
1526 | case SPELL_OPERATOR: |
1527 | { | |
1528 | const unsigned char *spelling; | |
3b681e9d | 1529 | int c; |
c5a04734 | 1530 | |
93c80368 | 1531 | if (token->flags & DIGRAPH) |
37b8524c JDA |
1532 | spelling |
1533 | = digraph_spellings[(int) token->type - (int) CPP_FIRST_DIGRAPH]; | |
93c80368 NB |
1534 | else if (token->flags & NAMED_OP) |
1535 | goto spell_ident; | |
1536 | else | |
1537 | spelling = TOKEN_NAME (token); | |
041c3194 | 1538 | |
3b681e9d ZW |
1539 | c = *spelling; |
1540 | do | |
1541 | putc (c, fp); | |
1542 | while ((c = *++spelling) != '\0'); | |
93c80368 NB |
1543 | } |
1544 | break; | |
041c3194 | 1545 | |
93c80368 NB |
1546 | spell_ident: |
1547 | case SPELL_IDENT: | |
3b681e9d | 1548 | fwrite (NODE_NAME (token->val.node), 1, NODE_LEN (token->val.node), fp); |
93c80368 | 1549 | break; |
041c3194 | 1550 | |
93c80368 NB |
1551 | case SPELL_STRING: |
1552 | { | |
1553 | int left, right, tag; | |
1554 | switch (token->type) | |
1555 | { | |
1556 | case CPP_STRING: left = '"'; right = '"'; tag = '\0'; break; | |
1557 | case CPP_WSTRING: left = '"'; right = '"'; tag = 'L'; break; | |
93c80368 NB |
1558 | case CPP_CHAR: left = '\''; right = '\''; tag = '\0'; break; |
1559 | case CPP_WCHAR: left = '\''; right = '\''; tag = 'L'; break; | |
1560 | case CPP_HEADER_NAME: left = '<'; right = '>'; tag = '\0'; break; | |
1561 | default: left = '\0'; right = '\0'; tag = '\0'; break; | |
1562 | } | |
1563 | if (tag) putc (tag, fp); | |
1564 | if (left) putc (left, fp); | |
1565 | fwrite (token->val.str.text, 1, token->val.str.len, fp); | |
1566 | if (right) putc (right, fp); | |
1567 | } | |
1568 | break; | |
c5a04734 | 1569 | |
93c80368 | 1570 | case SPELL_CHAR: |
6c53ebff | 1571 | putc (token->val.c, fp); |
93c80368 | 1572 | break; |
c5a04734 | 1573 | |
93c80368 NB |
1574 | case SPELL_NONE: |
1575 | /* An error, most probably. */ | |
1576 | break; | |
041c3194 | 1577 | } |
c5a04734 ZW |
1578 | } |
1579 | ||
93c80368 NB |
1580 | /* Compare two tokens. */ |
1581 | int | |
1582 | _cpp_equiv_tokens (a, b) | |
1583 | const cpp_token *a, *b; | |
c5a04734 | 1584 | { |
93c80368 NB |
1585 | if (a->type == b->type && a->flags == b->flags) |
1586 | switch (TOKEN_SPELL (a)) | |
1587 | { | |
1588 | default: /* Keep compiler happy. */ | |
1589 | case SPELL_OPERATOR: | |
1590 | return 1; | |
1591 | case SPELL_CHAR: | |
6c53ebff | 1592 | return a->val.c == b->val.c; /* Character. */ |
93c80368 | 1593 | case SPELL_NONE: |
56051c0a | 1594 | return (a->type != CPP_MACRO_ARG || a->val.arg_no == b->val.arg_no); |
93c80368 NB |
1595 | case SPELL_IDENT: |
1596 | return a->val.node == b->val.node; | |
1597 | case SPELL_STRING: | |
1598 | return (a->val.str.len == b->val.str.len | |
1599 | && !memcmp (a->val.str.text, b->val.str.text, | |
1600 | a->val.str.len)); | |
1601 | } | |
c5a04734 | 1602 | |
041c3194 ZW |
1603 | return 0; |
1604 | } | |
1605 | ||
93c80368 NB |
1606 | /* Returns nonzero if a space should be inserted to avoid an |
1607 | accidental token paste for output. For simplicity, it is | |
1608 | conservative, and occasionally advises a space where one is not | |
1609 | needed, e.g. "." and ".2". */ | |
041c3194 | 1610 | |
93c80368 NB |
1611 | int |
1612 | cpp_avoid_paste (pfile, token1, token2) | |
c5a04734 | 1613 | cpp_reader *pfile; |
93c80368 | 1614 | const cpp_token *token1, *token2; |
c5a04734 | 1615 | { |
93c80368 NB |
1616 | enum cpp_ttype a = token1->type, b = token2->type; |
1617 | cppchar_t c; | |
c5a04734 | 1618 | |
93c80368 NB |
1619 | if (token1->flags & NAMED_OP) |
1620 | a = CPP_NAME; | |
1621 | if (token2->flags & NAMED_OP) | |
1622 | b = CPP_NAME; | |
c5a04734 | 1623 | |
93c80368 NB |
1624 | c = EOF; |
1625 | if (token2->flags & DIGRAPH) | |
37b8524c | 1626 | c = digraph_spellings[(int) b - (int) CPP_FIRST_DIGRAPH][0]; |
93c80368 NB |
1627 | else if (token_spellings[b].category == SPELL_OPERATOR) |
1628 | c = token_spellings[b].name[0]; | |
c5a04734 | 1629 | |
93c80368 | 1630 | /* Quickly get everything that can paste with an '='. */ |
37b8524c | 1631 | if ((int) a <= (int) CPP_LAST_EQ && c == '=') |
93c80368 | 1632 | return 1; |
c5a04734 | 1633 | |
93c80368 | 1634 | switch (a) |
c5a04734 | 1635 | { |
93c80368 NB |
1636 | case CPP_GREATER: return c == '>' || c == '?'; |
1637 | case CPP_LESS: return c == '<' || c == '?' || c == '%' || c == ':'; | |
1638 | case CPP_PLUS: return c == '+'; | |
1639 | case CPP_MINUS: return c == '-' || c == '>'; | |
1640 | case CPP_DIV: return c == '/' || c == '*'; /* Comments. */ | |
1641 | case CPP_MOD: return c == ':' || c == '>'; | |
1642 | case CPP_AND: return c == '&'; | |
1643 | case CPP_OR: return c == '|'; | |
1644 | case CPP_COLON: return c == ':' || c == '>'; | |
1645 | case CPP_DEREF: return c == '*'; | |
26ec42ee | 1646 | case CPP_DOT: return c == '.' || c == '%' || b == CPP_NUMBER; |
93c80368 NB |
1647 | case CPP_HASH: return c == '#' || c == '%'; /* Digraph form. */ |
1648 | case CPP_NAME: return ((b == CPP_NUMBER | |
1649 | && name_p (pfile, &token2->val.str)) | |
1650 | || b == CPP_NAME | |
1651 | || b == CPP_CHAR || b == CPP_STRING); /* L */ | |
1652 | case CPP_NUMBER: return (b == CPP_NUMBER || b == CPP_NAME | |
1653 | || c == '.' || c == '+' || c == '-'); | |
1654 | case CPP_OTHER: return (CPP_OPTION (pfile, objc) | |
6c53ebff | 1655 | && token1->val.c == '@' |
93c80368 NB |
1656 | && (b == CPP_NAME || b == CPP_STRING)); |
1657 | default: break; | |
c5a04734 | 1658 | } |
c5a04734 | 1659 | |
417f3e3a | 1660 | return 0; |
c5a04734 ZW |
1661 | } |
1662 | ||
93c80368 | 1663 | /* Output all the remaining tokens on the current line, and a newline |
4ed5bcfb NB |
1664 | character, to FP. Leading whitespace is removed. If there are |
1665 | macros, special token padding is not performed. */ | |
c5a04734 | 1666 | void |
93c80368 | 1667 | cpp_output_line (pfile, fp) |
c5a04734 | 1668 | cpp_reader *pfile; |
93c80368 | 1669 | FILE *fp; |
c5a04734 | 1670 | { |
4ed5bcfb | 1671 | const cpp_token *token; |
96be6998 | 1672 | |
4ed5bcfb NB |
1673 | token = cpp_get_token (pfile); |
1674 | while (token->type != CPP_EOF) | |
96be6998 | 1675 | { |
4ed5bcfb NB |
1676 | cpp_output_token (token, fp); |
1677 | token = cpp_get_token (pfile); | |
1678 | if (token->flags & PREV_WHITE) | |
1679 | putc (' ', fp); | |
96be6998 ZW |
1680 | } |
1681 | ||
93c80368 | 1682 | putc ('\n', fp); |
041c3194 | 1683 | } |
c5a04734 | 1684 | |
c8a96070 NB |
1685 | /* Returns the value of a hexadecimal digit. */ |
1686 | static unsigned int | |
1687 | hex_digit_value (c) | |
1688 | unsigned int c; | |
1689 | { | |
1690 | if (c >= 'a' && c <= 'f') | |
1691 | return c - 'a' + 10; | |
1692 | if (c >= 'A' && c <= 'F') | |
1693 | return c - 'A' + 10; | |
1694 | if (c >= '0' && c <= '9') | |
1695 | return c - '0'; | |
1696 | abort (); | |
1697 | } | |
1698 | ||
62729350 NB |
1699 | /* Parse a '\uNNNN' or '\UNNNNNNNN' sequence. Returns 1 to indicate |
1700 | failure if cpplib is not parsing C++ or C99. Such failure is | |
1701 | silent, and no variables are updated. Otherwise returns 0, and | |
1702 | warns if -Wtraditional. | |
c8a96070 NB |
1703 | |
1704 | [lex.charset]: The character designated by the universal character | |
1705 | name \UNNNNNNNN is that character whose character short name in | |
1706 | ISO/IEC 10646 is NNNNNNNN; the character designated by the | |
1707 | universal character name \uNNNN is that character whose character | |
1708 | short name in ISO/IEC 10646 is 0000NNNN. If the hexadecimal value | |
1709 | for a universal character name is less than 0x20 or in the range | |
1710 | 0x7F-0x9F (inclusive), or if the universal character name | |
1711 | designates a character in the basic source character set, then the | |
1712 | program is ill-formed. | |
1713 | ||
1714 | We assume that wchar_t is Unicode, so we don't need to do any | |
62729350 | 1715 | mapping. Is this ever wrong? |
c8a96070 | 1716 | |
62729350 NB |
1717 | PC points to the 'u' or 'U', PSTR is points to the byte after PC, |
1718 | LIMIT is the end of the string or charconst. PSTR is updated to | |
1719 | point after the UCS on return, and the UCS is written into PC. */ | |
1720 | ||
1721 | static int | |
1722 | maybe_read_ucs (pfile, pstr, limit, pc) | |
c8a96070 NB |
1723 | cpp_reader *pfile; |
1724 | const unsigned char **pstr; | |
1725 | const unsigned char *limit; | |
62729350 | 1726 | unsigned int *pc; |
c8a96070 NB |
1727 | { |
1728 | const unsigned char *p = *pstr; | |
62729350 NB |
1729 | unsigned int code = 0; |
1730 | unsigned int c = *pc, length; | |
1731 | ||
1732 | /* Only attempt to interpret a UCS for C++ and C99. */ | |
1733 | if (! (CPP_OPTION (pfile, cplusplus) || CPP_OPTION (pfile, c99))) | |
1734 | return 1; | |
c8a96070 | 1735 | |
62729350 NB |
1736 | if (CPP_WTRADITIONAL (pfile)) |
1737 | cpp_warning (pfile, "the meaning of '\\%c' varies with -traditional", c); | |
c8a96070 | 1738 | |
f8710242 NB |
1739 | length = (c == 'u' ? 4: 8); |
1740 | ||
1741 | if ((size_t) (limit - p) < length) | |
1742 | { | |
1743 | cpp_error (pfile, "incomplete universal-character-name"); | |
1744 | /* Skip to the end to avoid more diagnostics. */ | |
1745 | p = limit; | |
1746 | } | |
1747 | else | |
1748 | { | |
1749 | for (; length; length--, p++) | |
c8a96070 | 1750 | { |
f8710242 NB |
1751 | c = *p; |
1752 | if (ISXDIGIT (c)) | |
1753 | code = (code << 4) + hex_digit_value (c); | |
1754 | else | |
1755 | { | |
1756 | cpp_error (pfile, | |
1757 | "non-hex digit '%c' in universal-character-name", c); | |
1758 | /* We shouldn't skip in case there are multibyte chars. */ | |
1759 | break; | |
1760 | } | |
c8a96070 | 1761 | } |
c8a96070 NB |
1762 | } |
1763 | ||
1764 | #ifdef TARGET_EBCDIC | |
1765 | cpp_error (pfile, "universal-character-name on EBCDIC target"); | |
1766 | code = 0x3f; /* EBCDIC invalid character */ | |
1767 | #else | |
f8710242 NB |
1768 | /* True extended characters are OK. */ |
1769 | if (code >= 0xa0 | |
1770 | && !(code & 0x80000000) | |
1771 | && !(code >= 0xD800 && code <= 0xDFFF)) | |
1772 | ; | |
1773 | /* The standard permits $, @ and ` to be specified as UCNs. We use | |
1774 | hex escapes so that this also works with EBCDIC hosts. */ | |
1775 | else if (code == 0x24 || code == 0x40 || code == 0x60) | |
1776 | ; | |
1777 | /* Don't give another error if one occurred above. */ | |
1778 | else if (length == 0) | |
1779 | cpp_error (pfile, "universal-character-name out of range"); | |
c8a96070 NB |
1780 | #endif |
1781 | ||
1782 | *pstr = p; | |
62729350 NB |
1783 | *pc = code; |
1784 | return 0; | |
c8a96070 NB |
1785 | } |
1786 | ||
1787 | /* Interpret an escape sequence, and return its value. PSTR points to | |
1788 | the input pointer, which is just after the backslash. LIMIT is how | |
62729350 NB |
1789 | much text we have. MASK is a bitmask for the precision for the |
1790 | destination type (char or wchar_t). TRADITIONAL, if true, does not | |
1791 | interpret escapes that did not exist in traditional C. | |
c8a96070 | 1792 | |
62729350 NB |
1793 | Handles all relevant diagnostics. */ |
1794 | ||
1795 | unsigned int | |
1796 | cpp_parse_escape (pfile, pstr, limit, mask, traditional) | |
c8a96070 NB |
1797 | cpp_reader *pfile; |
1798 | const unsigned char **pstr; | |
1799 | const unsigned char *limit; | |
62729350 | 1800 | unsigned HOST_WIDE_INT mask; |
c8a96070 NB |
1801 | int traditional; |
1802 | { | |
1803 | int unknown = 0; | |
1804 | const unsigned char *str = *pstr; | |
1805 | unsigned int c = *str++; | |
1806 | ||
1807 | switch (c) | |
1808 | { | |
1809 | case '\\': case '\'': case '"': case '?': break; | |
1810 | case 'b': c = TARGET_BS; break; | |
1811 | case 'f': c = TARGET_FF; break; | |
1812 | case 'n': c = TARGET_NEWLINE; break; | |
1813 | case 'r': c = TARGET_CR; break; | |
1814 | case 't': c = TARGET_TAB; break; | |
1815 | case 'v': c = TARGET_VT; break; | |
1816 | ||
1817 | case '(': case '{': case '[': case '%': | |
1818 | /* '\(', etc, are used at beginning of line to avoid confusing Emacs. | |
1819 | '\%' is used to prevent SCCS from getting confused. */ | |
1820 | unknown = CPP_PEDANTIC (pfile); | |
1821 | break; | |
1822 | ||
1823 | case 'a': | |
1824 | if (CPP_WTRADITIONAL (pfile)) | |
1825 | cpp_warning (pfile, "the meaning of '\\a' varies with -traditional"); | |
1826 | if (!traditional) | |
1827 | c = TARGET_BELL; | |
1828 | break; | |
1829 | ||
1830 | case 'e': case 'E': | |
1831 | if (CPP_PEDANTIC (pfile)) | |
1832 | cpp_pedwarn (pfile, "non-ISO-standard escape sequence, '\\%c'", c); | |
1833 | c = TARGET_ESC; | |
1834 | break; | |
1835 | ||
c8a96070 | 1836 | case 'u': case 'U': |
62729350 | 1837 | unknown = maybe_read_ucs (pfile, &str, limit, &c); |
c8a96070 NB |
1838 | break; |
1839 | ||
1840 | case 'x': | |
1841 | if (CPP_WTRADITIONAL (pfile)) | |
1842 | cpp_warning (pfile, "the meaning of '\\x' varies with -traditional"); | |
1843 | ||
1844 | if (!traditional) | |
1845 | { | |
1846 | unsigned int i = 0, overflow = 0; | |
1847 | int digits_found = 0; | |
1848 | ||
1849 | while (str < limit) | |
1850 | { | |
1851 | c = *str; | |
1852 | if (! ISXDIGIT (c)) | |
1853 | break; | |
1854 | str++; | |
1855 | overflow |= i ^ (i << 4 >> 4); | |
1856 | i = (i << 4) + hex_digit_value (c); | |
1857 | digits_found = 1; | |
1858 | } | |
1859 | ||
1860 | if (!digits_found) | |
1861 | cpp_error (pfile, "\\x used with no following hex digits"); | |
1862 | ||
1863 | if (overflow | (i != (i & mask))) | |
1864 | { | |
1865 | cpp_pedwarn (pfile, "hex escape sequence out of range"); | |
1866 | i &= mask; | |
1867 | } | |
1868 | c = i; | |
1869 | } | |
1870 | break; | |
1871 | ||
1872 | case '0': case '1': case '2': case '3': | |
1873 | case '4': case '5': case '6': case '7': | |
1874 | { | |
1875 | unsigned int i = c - '0'; | |
1876 | int count = 0; | |
1877 | ||
1878 | while (str < limit && ++count < 3) | |
1879 | { | |
1880 | c = *str; | |
1881 | if (c < '0' || c > '7') | |
1882 | break; | |
1883 | str++; | |
1884 | i = (i << 3) + c - '0'; | |
1885 | } | |
1886 | ||
1887 | if (i != (i & mask)) | |
1888 | { | |
1889 | cpp_pedwarn (pfile, "octal escape sequence out of range"); | |
1890 | i &= mask; | |
1891 | } | |
1892 | c = i; | |
1893 | } | |
1894 | break; | |
1895 | ||
1896 | default: | |
1897 | unknown = 1; | |
1898 | break; | |
1899 | } | |
1900 | ||
1901 | if (unknown) | |
1902 | { | |
1903 | if (ISGRAPH (c)) | |
1904 | cpp_pedwarn (pfile, "unknown escape sequence '\\%c'", c); | |
1905 | else | |
1906 | cpp_pedwarn (pfile, "unknown escape sequence: '\\%03o'", c); | |
1907 | } | |
1908 | ||
62729350 NB |
1909 | if (c > mask) |
1910 | cpp_pedwarn (pfile, "escape sequence out of range for character"); | |
1911 | ||
c8a96070 NB |
1912 | *pstr = str; |
1913 | return c; | |
1914 | } | |
1915 | ||
1916 | #ifndef MAX_CHAR_TYPE_SIZE | |
1917 | #define MAX_CHAR_TYPE_SIZE CHAR_TYPE_SIZE | |
1918 | #endif | |
1919 | ||
1920 | #ifndef MAX_WCHAR_TYPE_SIZE | |
1921 | #define MAX_WCHAR_TYPE_SIZE WCHAR_TYPE_SIZE | |
1922 | #endif | |
1923 | ||
1924 | /* Interpret a (possibly wide) character constant in TOKEN. | |
1925 | WARN_MULTI warns about multi-character charconsts, if not | |
1926 | TRADITIONAL. TRADITIONAL also indicates not to interpret escapes | |
1927 | that did not exist in traditional C. PCHARS_SEEN points to a | |
1928 | variable that is filled in with the number of characters seen. */ | |
1929 | HOST_WIDE_INT | |
1930 | cpp_interpret_charconst (pfile, token, warn_multi, traditional, pchars_seen) | |
1931 | cpp_reader *pfile; | |
1932 | const cpp_token *token; | |
1933 | int warn_multi; | |
1934 | int traditional; | |
1935 | unsigned int *pchars_seen; | |
1936 | { | |
1937 | const unsigned char *str = token->val.str.text; | |
1938 | const unsigned char *limit = str + token->val.str.len; | |
1939 | unsigned int chars_seen = 0; | |
1940 | unsigned int width, max_chars, c; | |
2a967f3d NB |
1941 | unsigned HOST_WIDE_INT mask; |
1942 | HOST_WIDE_INT result = 0; | |
c8a96070 NB |
1943 | |
1944 | #ifdef MULTIBYTE_CHARS | |
1945 | (void) local_mbtowc (NULL, NULL, 0); | |
1946 | #endif | |
1947 | ||
1948 | /* Width in bits. */ | |
1949 | if (token->type == CPP_CHAR) | |
1950 | width = MAX_CHAR_TYPE_SIZE; | |
1951 | else | |
1952 | width = MAX_WCHAR_TYPE_SIZE; | |
1953 | ||
1954 | if (width < HOST_BITS_PER_WIDE_INT) | |
1955 | mask = ((unsigned HOST_WIDE_INT) 1 << width) - 1; | |
1956 | else | |
1957 | mask = ~0; | |
1958 | max_chars = HOST_BITS_PER_WIDE_INT / width; | |
1959 | ||
1960 | while (str < limit) | |
1961 | { | |
1962 | #ifdef MULTIBYTE_CHARS | |
1963 | wchar_t wc; | |
1964 | int char_len; | |
1965 | ||
1966 | char_len = local_mbtowc (&wc, str, limit - str); | |
1967 | if (char_len == -1) | |
1968 | { | |
1969 | cpp_warning (pfile, "ignoring invalid multibyte character"); | |
1970 | c = *str++; | |
1971 | } | |
1972 | else | |
1973 | { | |
1974 | str += char_len; | |
1975 | c = wc; | |
1976 | } | |
1977 | #else | |
1978 | c = *str++; | |
1979 | #endif | |
1980 | ||
1981 | if (c == '\\') | |
62729350 | 1982 | c = cpp_parse_escape (pfile, &str, limit, mask, traditional); |
c8a96070 NB |
1983 | |
1984 | #ifdef MAP_CHARACTER | |
1985 | if (ISPRINT (c)) | |
1986 | c = MAP_CHARACTER (c); | |
1987 | #endif | |
1988 | ||
1989 | /* Merge character into result; ignore excess chars. */ | |
1990 | if (++chars_seen <= max_chars) | |
1991 | { | |
1992 | if (width < HOST_BITS_PER_WIDE_INT) | |
1993 | result = (result << width) | (c & mask); | |
1994 | else | |
1995 | result = c; | |
1996 | } | |
1997 | } | |
1998 | ||
1999 | if (chars_seen == 0) | |
2000 | cpp_error (pfile, "empty character constant"); | |
2001 | else if (chars_seen > max_chars) | |
2002 | { | |
2003 | chars_seen = max_chars; | |
f8710242 | 2004 | cpp_warning (pfile, "character constant too long"); |
c8a96070 NB |
2005 | } |
2006 | else if (chars_seen > 1 && !traditional && warn_multi) | |
2007 | cpp_warning (pfile, "multi-character character constant"); | |
2008 | ||
2009 | /* If char type is signed, sign-extend the constant. The | |
2010 | __CHAR_UNSIGNED__ macro is set by the driver if appropriate. */ | |
2011 | if (token->type == CPP_CHAR && chars_seen) | |
2012 | { | |
2013 | unsigned int nbits = chars_seen * width; | |
2014 | unsigned int mask = (unsigned int) ~0 >> (HOST_BITS_PER_INT - nbits); | |
2015 | ||
2016 | if (pfile->spec_nodes.n__CHAR_UNSIGNED__->type == NT_MACRO | |
2017 | || ((result >> (nbits - 1)) & 1) == 0) | |
2018 | result &= mask; | |
2019 | else | |
2020 | result |= ~mask; | |
2021 | } | |
2022 | ||
2023 | *pchars_seen = chars_seen; | |
2024 | return result; | |
2025 | } | |
2026 | ||
1e013d2e NB |
2027 | /* Memory buffers. Changing these three constants can have a dramatic |
2028 | effect on performance. The values here are reasonable defaults, | |
2029 | but might be tuned. If you adjust them, be sure to test across a | |
2030 | range of uses of cpplib, including heavy nested function-like macro | |
2031 | expansion. Also check the change in peak memory usage (NJAMD is a | |
2032 | good tool for this). */ | |
2033 | #define MIN_BUFF_SIZE 8000 | |
2034 | #define BUFF_SIZE_UPPER_BOUND(MIN_SIZE) (8000 + (MIN_SIZE) * 3 / 2) | |
2035 | #define EXTENDED_BUFF_SIZE(BUFF, MIN_EXTRA) \ | |
2036 | (MIN_EXTRA + ((BUFF)->limit - (BUFF)->cur) * 2) | |
417f3e3a | 2037 | |
93c80368 | 2038 | struct dummy |
417f3e3a | 2039 | { |
93c80368 NB |
2040 | char c; |
2041 | union | |
2042 | { | |
2043 | double d; | |
2044 | int *p; | |
2045 | } u; | |
2046 | }; | |
417f3e3a | 2047 | |
93c80368 | 2048 | #define DEFAULT_ALIGNMENT (offsetof (struct dummy, u)) |
b8af0ca5 NB |
2049 | #define CPP_ALIGN(size, align) (((size) + ((align) - 1)) & ~((align) - 1)) |
2050 | ||
c9e7a609 NB |
2051 | /* Create a new allocation buffer. Place the control block at the end |
2052 | of the buffer, so that buffer overflows will cause immediate chaos. */ | |
b8af0ca5 NB |
2053 | static _cpp_buff * |
2054 | new_buff (len) | |
2055 | unsigned int len; | |
2056 | { | |
2057 | _cpp_buff *result; | |
ece54d54 | 2058 | unsigned char *base; |
b8af0ca5 | 2059 | |
1e013d2e NB |
2060 | if (len < MIN_BUFF_SIZE) |
2061 | len = MIN_BUFF_SIZE; | |
b8af0ca5 NB |
2062 | len = CPP_ALIGN (len, DEFAULT_ALIGNMENT); |
2063 | ||
2064 | base = xmalloc (len + sizeof (_cpp_buff)); | |
2065 | result = (_cpp_buff *) (base + len); | |
2066 | result->base = base; | |
2067 | result->cur = base; | |
2068 | result->limit = base + len; | |
2069 | result->next = NULL; | |
2070 | return result; | |
2071 | } | |
2072 | ||
2073 | /* Place a chain of unwanted allocation buffers on the free list. */ | |
2074 | void | |
2075 | _cpp_release_buff (pfile, buff) | |
2076 | cpp_reader *pfile; | |
2077 | _cpp_buff *buff; | |
2078 | { | |
2079 | _cpp_buff *end = buff; | |
2080 | ||
2081 | while (end->next) | |
2082 | end = end->next; | |
2083 | end->next = pfile->free_buffs; | |
2084 | pfile->free_buffs = buff; | |
2085 | } | |
2086 | ||
2087 | /* Return a free buffer of size at least MIN_SIZE. */ | |
2088 | _cpp_buff * | |
2089 | _cpp_get_buff (pfile, min_size) | |
2090 | cpp_reader *pfile; | |
2091 | unsigned int min_size; | |
2092 | { | |
2093 | _cpp_buff *result, **p; | |
2094 | ||
2095 | for (p = &pfile->free_buffs;; p = &(*p)->next) | |
2096 | { | |
1e013d2e NB |
2097 | unsigned int size; |
2098 | ||
2099 | if (*p == NULL) | |
b8af0ca5 | 2100 | return new_buff (min_size); |
1e013d2e NB |
2101 | result = *p; |
2102 | size = result->limit - result->base; | |
2103 | /* Return a buffer that's big enough, but don't waste one that's | |
2104 | way too big. */ | |
2105 | if (size >= min_size && size < BUFF_SIZE_UPPER_BOUND (min_size)) | |
b8af0ca5 NB |
2106 | break; |
2107 | } | |
2108 | ||
2109 | *p = result->next; | |
2110 | result->next = NULL; | |
2111 | result->cur = result->base; | |
2112 | return result; | |
2113 | } | |
2114 | ||
2115 | /* Return a buffer chained on the end of BUFF. Copy to it the | |
2116 | uncommitted remaining bytes of BUFF, with at least MIN_EXTRA more | |
2117 | bytes. */ | |
2118 | _cpp_buff * | |
2119 | _cpp_extend_buff (pfile, buff, min_extra) | |
2120 | cpp_reader *pfile; | |
2121 | _cpp_buff *buff; | |
2122 | unsigned int min_extra; | |
2123 | { | |
1e013d2e | 2124 | unsigned int size = EXTENDED_BUFF_SIZE (buff, min_extra); |
b8af0ca5 NB |
2125 | |
2126 | buff->next = _cpp_get_buff (pfile, size); | |
2127 | memcpy (buff->next->base, buff->cur, buff->limit - buff->cur); | |
2128 | return buff->next; | |
2129 | } | |
2130 | ||
2131 | /* Free a chain of buffers starting at BUFF. */ | |
2132 | void | |
2133 | _cpp_free_buff (buff) | |
2134 | _cpp_buff *buff; | |
2135 | { | |
2136 | _cpp_buff *next; | |
2137 | ||
2138 | for (; buff; buff = next) | |
2139 | { | |
2140 | next = buff->next; | |
2141 | free (buff->base); | |
2142 | } | |
2143 | } | |
417f3e3a | 2144 | |
ece54d54 NB |
2145 | /* Allocate permanent, unaligned storage of length LEN. */ |
2146 | unsigned char * | |
2147 | _cpp_unaligned_alloc (pfile, len) | |
2148 | cpp_reader *pfile; | |
2149 | size_t len; | |
2150 | { | |
2151 | _cpp_buff *buff = pfile->u_buff; | |
2152 | unsigned char *result = buff->cur; | |
2153 | ||
2154 | if (len > (size_t) (buff->limit - result)) | |
2155 | { | |
2156 | buff = _cpp_get_buff (pfile, len); | |
2157 | buff->next = pfile->u_buff; | |
2158 | pfile->u_buff = buff; | |
2159 | result = buff->cur; | |
2160 | } | |
2161 | ||
2162 | buff->cur = result + len; | |
2163 | return result; | |
2164 | } | |
2165 | ||
93c80368 | 2166 | static int |
1e013d2e | 2167 | chunk_suitable (chunk, size) |
93c80368 NB |
2168 | cpp_chunk *chunk; |
2169 | unsigned int size; | |
2170 | { | |
2171 | /* Being at least twice SIZE means we can use memcpy in | |
2172 | _cpp_next_chunk rather than memmove. Besides, it's a good idea | |
2173 | anyway. */ | |
1e013d2e | 2174 | return (chunk && (unsigned int) (chunk->limit - chunk->base) >= size * 2); |
041c3194 | 2175 | } |
c5a04734 | 2176 | |
93c80368 NB |
2177 | /* Returns the end of the new pool. PTR points to a char in the old |
2178 | pool, and is updated to point to the same char in the new pool. */ | |
2179 | unsigned char * | |
2180 | _cpp_next_chunk (pool, len, ptr) | |
2181 | cpp_pool *pool; | |
2182 | unsigned int len; | |
2183 | unsigned char **ptr; | |
041c3194 | 2184 | { |
93c80368 | 2185 | cpp_chunk *chunk = pool->cur->next; |
c5a04734 | 2186 | |
93c80368 NB |
2187 | /* LEN is the minimum size we want in the new pool. */ |
2188 | len += POOL_ROOM (pool); | |
1e013d2e | 2189 | if (! chunk_suitable (chunk, len)) |
041c3194 | 2190 | { |
93c80368 | 2191 | chunk = new_chunk (POOL_SIZE (pool) * 2 + len); |
c5a04734 | 2192 | |
93c80368 NB |
2193 | chunk->next = pool->cur->next; |
2194 | pool->cur->next = chunk; | |
c5a04734 ZW |
2195 | } |
2196 | ||
93c80368 NB |
2197 | /* Update the pointer before changing chunk's front. */ |
2198 | if (ptr) | |
2199 | *ptr += chunk->base - POOL_FRONT (pool); | |
041c3194 | 2200 | |
93c80368 NB |
2201 | memcpy (chunk->base, POOL_FRONT (pool), POOL_ROOM (pool)); |
2202 | chunk->front = chunk->base; | |
041c3194 | 2203 | |
93c80368 NB |
2204 | pool->cur = chunk; |
2205 | return POOL_LIMIT (pool); | |
c5a04734 ZW |
2206 | } |
2207 | ||
93c80368 NB |
2208 | static cpp_chunk * |
2209 | new_chunk (size) | |
2210 | unsigned int size; | |
041c3194 | 2211 | { |
93c80368 NB |
2212 | unsigned char *base; |
2213 | cpp_chunk *result; | |
3fef5b2b | 2214 | |
269592a8 | 2215 | size = POOL_ALIGN (size, DEFAULT_ALIGNMENT); |
93c80368 NB |
2216 | base = (unsigned char *) xmalloc (size + sizeof (cpp_chunk)); |
2217 | /* Put the chunk descriptor at the end. Then chunk overruns will | |
2218 | cause obvious chaos. */ | |
2219 | result = (cpp_chunk *) (base + size); | |
2220 | result->base = base; | |
2221 | result->front = base; | |
2222 | result->limit = base + size; | |
2223 | result->next = 0; | |
417f3e3a | 2224 | |
93c80368 | 2225 | return result; |
041c3194 ZW |
2226 | } |
2227 | ||
93c80368 NB |
2228 | void |
2229 | _cpp_init_pool (pool, size, align, temp) | |
2230 | cpp_pool *pool; | |
2231 | unsigned int size, align, temp; | |
2232 | { | |
2233 | if (align == 0) | |
2234 | align = DEFAULT_ALIGNMENT; | |
2235 | if (align & (align - 1)) | |
2236 | abort (); | |
2237 | pool->align = align; | |
bef985f3 NB |
2238 | pool->first = new_chunk (size); |
2239 | pool->cur = pool->first; | |
93c80368 NB |
2240 | if (temp) |
2241 | pool->cur->next = pool->cur; | |
041c3194 ZW |
2242 | } |
2243 | ||
93c80368 NB |
2244 | void |
2245 | _cpp_free_pool (pool) | |
2246 | cpp_pool *pool; | |
3fef5b2b | 2247 | { |
bef985f3 | 2248 | cpp_chunk *chunk = pool->first, *next; |
3fef5b2b | 2249 | |
93c80368 | 2250 | do |
3fef5b2b | 2251 | { |
93c80368 NB |
2252 | next = chunk->next; |
2253 | free (chunk->base); | |
2254 | chunk = next; | |
3fef5b2b | 2255 | } |
bef985f3 | 2256 | while (chunk && chunk != pool->first); |
041c3194 | 2257 | } |
041c3194 | 2258 | |
93c80368 NB |
2259 | /* Reserve LEN bytes from a memory pool. */ |
2260 | unsigned char * | |
2261 | _cpp_pool_reserve (pool, len) | |
2262 | cpp_pool *pool; | |
2263 | unsigned int len; | |
041c3194 | 2264 | { |
269592a8 | 2265 | len = POOL_ALIGN (len, pool->align); |
93c80368 NB |
2266 | if (len > (unsigned int) POOL_ROOM (pool)) |
2267 | _cpp_next_chunk (pool, len, 0); | |
041c3194 | 2268 | |
93c80368 | 2269 | return POOL_FRONT (pool); |
c5a04734 ZW |
2270 | } |
2271 | ||
93c80368 NB |
2272 | /* Allocate LEN bytes from a memory pool. */ |
2273 | unsigned char * | |
2274 | _cpp_pool_alloc (pool, len) | |
2275 | cpp_pool *pool; | |
2276 | unsigned int len; | |
041c3194 | 2277 | { |
93c80368 | 2278 | unsigned char *result = _cpp_pool_reserve (pool, len); |
417f3e3a | 2279 | |
93c80368 NB |
2280 | POOL_COMMIT (pool, len); |
2281 | return result; | |
041c3194 | 2282 | } |