]>
Commit | Line | Data |
---|---|---|
0578f103 | 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 | |
6060326b | 7 | Single-pass line tokenization by Neil Booth, April 2000 |
0578f103 | 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 | ||
79bd622b | 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. */ | |
f80e83a9 | 36 | |
0578f103 | 37 | #include "config.h" |
38 | #include "system.h" | |
0578f103 | 39 | #include "cpplib.h" |
40 | #include "cpphash.h" | |
41 | ||
8330799c | 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 | ||
79bd622b | 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 | |
241e762e | 57 | { |
79bd622b | 58 | SPELL_OPERATOR = 0, |
59 | SPELL_CHAR, | |
60 | SPELL_IDENT, | |
61 | SPELL_STRING, | |
62 | SPELL_NONE | |
241e762e | 63 | }; |
64 | ||
79bd622b | 65 | struct token_spelling |
241e762e | 66 | { |
79bd622b | 67 | enum spell_type category; |
68 | const unsigned char *name; | |
241e762e | 69 | }; |
70 | ||
79bd622b | 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) | |
e2f9a79f | 82 | |
36a0aa7c | 83 | static cppchar_t handle_newline PARAMS ((cpp_reader *, cppchar_t)); |
c808d026 | 84 | static cppchar_t skip_escaped_newlines PARAMS ((cpp_reader *, cppchar_t)); |
85 | static cppchar_t get_effective_char PARAMS ((cpp_reader *)); | |
338fa5f7 | 86 | |
f80e83a9 | 87 | static int skip_block_comment PARAMS ((cpp_reader *)); |
f669338a | 88 | static int skip_line_comment PARAMS ((cpp_reader *)); |
338fa5f7 | 89 | static void adjust_column PARAMS ((cpp_reader *)); |
90 | static void skip_whitespace PARAMS ((cpp_reader *, cppchar_t)); | |
66a5287e | 91 | static cpp_hashnode *parse_identifier PARAMS ((cpp_reader *)); |
92 | static cpp_hashnode *parse_identifier_slow PARAMS ((cpp_reader *, | |
93 | const U_CHAR *)); | |
79bd622b | 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 *)); | |
338fa5f7 | 96 | static void parse_string PARAMS ((cpp_reader *, cpp_token *, cppchar_t)); |
79bd622b | 97 | static void unterminated PARAMS ((cpp_reader *, int)); |
338fa5f7 | 98 | static int trigraph_ok PARAMS ((cpp_reader *, cppchar_t)); |
99 | static void save_comment PARAMS ((cpp_reader *, cpp_token *, const U_CHAR *)); | |
c808d026 | 100 | static void lex_percent PARAMS ((cpp_reader *, cpp_token *)); |
f669338a | 101 | static void lex_dot PARAMS ((cpp_reader *, cpp_token *)); |
79bd622b | 102 | static int name_p PARAMS ((cpp_reader *, const cpp_string *)); |
c8342759 | 103 | static int maybe_read_ucs PARAMS ((cpp_reader *, const unsigned char **, |
104 | const unsigned char *, unsigned int *)); | |
83dcbb5c | 105 | static tokenrun *next_tokenrun PARAMS ((tokenrun *)); |
e916a356 | 106 | |
79bd622b | 107 | static cpp_chunk *new_chunk PARAMS ((unsigned int)); |
084163dc | 108 | static int chunk_suitable PARAMS ((cpp_chunk *, unsigned int)); |
8330799c | 109 | static unsigned int hex_digit_value PARAMS ((unsigned int)); |
06c92cbc | 110 | static _cpp_buff *new_buff PARAMS ((unsigned int)); |
bce8e0c0 | 111 | |
f80e83a9 | 112 | /* Utility routine: |
2c63d6c8 | 113 | |
76faa4c0 | 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. */ | |
bce8e0c0 | 116 | |
f80e83a9 | 117 | int |
76faa4c0 | 118 | cpp_ideq (token, string) |
119 | const cpp_token *token; | |
f80e83a9 | 120 | const char *string; |
121 | { | |
76faa4c0 | 122 | if (token->type != CPP_NAME) |
f80e83a9 | 123 | return 0; |
76faa4c0 | 124 | |
c86dbc5b | 125 | return !ustrcmp (NODE_NAME (token->val.node), (const U_CHAR *) string); |
bce8e0c0 | 126 | } |
50fd6b48 | 127 | |
338fa5f7 | 128 | /* Call when meeting a newline. Returns the character after the newline |
129 | (or carriage-return newline combination), or EOF. */ | |
130 | static cppchar_t | |
36a0aa7c | 131 | handle_newline (pfile, newline_char) |
132 | cpp_reader *pfile; | |
338fa5f7 | 133 | cppchar_t newline_char; |
134 | { | |
36a0aa7c | 135 | cpp_buffer *buffer; |
338fa5f7 | 136 | cppchar_t next = EOF; |
137 | ||
36a0aa7c | 138 | pfile->line++; |
36a0aa7c | 139 | buffer = pfile->buffer; |
338fa5f7 | 140 | buffer->col_adjust = 0; |
338fa5f7 | 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. */ | |
0578f103 | 166 | static int |
338fa5f7 | 167 | trigraph_ok (pfile, from_char) |
0578f103 | 168 | cpp_reader *pfile; |
338fa5f7 | 169 | cppchar_t from_char; |
0578f103 | 170 | { |
f80e83a9 | 171 | int accept = CPP_OPTION (pfile, trigraphs); |
172 | ||
f669338a | 173 | /* Don't warn about trigraphs in comments. */ |
174 | if (CPP_OPTION (pfile, warn_trigraphs) && !pfile->state.lexing_comment) | |
0578f103 | 175 | { |
338fa5f7 | 176 | cpp_buffer *buffer = pfile->buffer; |
1ea7ed21 | 177 | |
f80e83a9 | 178 | if (accept) |
1ea7ed21 | 179 | cpp_warning_with_line (pfile, pfile->line, CPP_BUF_COL (buffer) - 2, |
f80e83a9 | 180 | "trigraph ??%c converted to %c", |
338fa5f7 | 181 | (int) from_char, |
182 | (int) _cpp_trigraph_map[from_char]); | |
4b912310 | 183 | else if (buffer->cur != buffer->last_Wtrigraphs) |
184 | { | |
185 | buffer->last_Wtrigraphs = buffer->cur; | |
1ea7ed21 | 186 | cpp_warning_with_line (pfile, pfile->line, |
4b912310 | 187 | CPP_BUF_COL (buffer) - 2, |
188 | "trigraph ??%c ignored", (int) from_char); | |
189 | } | |
0578f103 | 190 | } |
338fa5f7 | 191 | |
f80e83a9 | 192 | return accept; |
0578f103 | 193 | } |
194 | ||
338fa5f7 | 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 | |
396ffa86 | 209 | been placed in buffer->read_ahead. This routine performs |
210 | preprocessing stages 1 and 2 of the ISO C standard. */ | |
338fa5f7 | 211 | static cppchar_t |
c808d026 | 212 | skip_escaped_newlines (pfile, next) |
213 | cpp_reader *pfile; | |
338fa5f7 | 214 | cppchar_t next; |
0578f103 | 215 | { |
c808d026 | 216 | cpp_buffer *buffer = pfile->buffer; |
217 | ||
396ffa86 | 218 | /* Only do this if we apply stages 1 and 2. */ |
219 | if (!buffer->from_stage3) | |
f80e83a9 | 220 | { |
396ffa86 | 221 | cppchar_t next1; |
222 | const unsigned char *saved_cur; | |
223 | int space; | |
224 | ||
225 | do | |
338fa5f7 | 226 | { |
396ffa86 | 227 | if (buffer->cur == buffer->rlimit) |
228 | break; | |
229 | ||
230 | SAVE_STATE (); | |
231 | if (next == '?') | |
338fa5f7 | 232 | { |
396ffa86 | 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] | |
c808d026 | 242 | || !trigraph_ok (pfile, next1)) |
396ffa86 | 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; | |
338fa5f7 | 263 | } |
396ffa86 | 264 | while (buffer->cur < buffer->rlimit); |
f80e83a9 | 265 | |
396ffa86 | 266 | if (!is_vspace (next1)) |
338fa5f7 | 267 | { |
268 | RESTORE_STATE (); | |
269 | break; | |
270 | } | |
0578f103 | 271 | |
c808d026 | 272 | if (space && !pfile->state.lexing_comment) |
273 | cpp_warning (pfile, "backslash and newline separated by space"); | |
338fa5f7 | 274 | |
c808d026 | 275 | next = handle_newline (pfile, next1); |
396ffa86 | 276 | if (next == EOF) |
c808d026 | 277 | cpp_pedwarn (pfile, "backslash-newline at end of file"); |
338fa5f7 | 278 | } |
396ffa86 | 279 | while (next == '\\' || next == '?'); |
f80e83a9 | 280 | } |
0578f103 | 281 | |
338fa5f7 | 282 | buffer->read_ahead = next; |
283 | return next; | |
0578f103 | 284 | } |
285 | ||
338fa5f7 | 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 | |
c808d026 | 290 | get_effective_char (pfile) |
291 | cpp_reader *pfile; | |
852d1b04 | 292 | { |
c808d026 | 293 | cpp_buffer *buffer = pfile->buffer; |
338fa5f7 | 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 == '\\') | |
c808d026 | 305 | next = skip_escaped_newlines (pfile, next); |
338fa5f7 | 306 | } |
307 | ||
308 | buffer->read_ahead = next; | |
309 | return next; | |
852d1b04 | 310 | } |
311 | ||
338fa5f7 | 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. */ | |
f80e83a9 | 315 | static int |
316 | skip_block_comment (pfile) | |
0578f103 | 317 | cpp_reader *pfile; |
318 | { | |
f80e83a9 | 319 | cpp_buffer *buffer = pfile->buffer; |
63e1abce | 320 | cppchar_t c = EOF, prevc = EOF; |
338fa5f7 | 321 | |
f669338a | 322 | pfile->state.lexing_comment = 1; |
338fa5f7 | 323 | while (buffer->cur != buffer->rlimit) |
0578f103 | 324 | { |
338fa5f7 | 325 | prevc = c, c = *buffer->cur++; |
326 | ||
327 | next_char: | |
328 | /* FIXME: For speed, create a new character class of characters | |
79bd622b | 329 | of interest inside block comments. */ |
338fa5f7 | 330 | if (c == '?' || c == '\\') |
c808d026 | 331 | c = skip_escaped_newlines (pfile, c); |
f80e83a9 | 332 | |
338fa5f7 | 333 | /* People like decorating comments with '*', so check for '/' |
334 | instead for efficiency. */ | |
f80e83a9 | 335 | if (c == '/') |
0578f103 | 336 | { |
338fa5f7 | 337 | if (prevc == '*') |
338 | break; | |
f80e83a9 | 339 | |
338fa5f7 | 340 | /* Warn about potential nested comments, but not if the '/' |
341 | comes immediately before the true comment delimeter. | |
f80e83a9 | 342 | Don't bother to get it right across escaped newlines. */ |
338fa5f7 | 343 | if (CPP_OPTION (pfile, warn_comments) |
344 | && buffer->cur != buffer->rlimit) | |
0578f103 | 345 | { |
338fa5f7 | 346 | prevc = c, c = *buffer->cur++; |
347 | if (c == '*' && buffer->cur != buffer->rlimit) | |
348 | { | |
349 | prevc = c, c = *buffer->cur++; | |
350 | if (c != '/') | |
1ea7ed21 | 351 | cpp_warning_with_line (pfile, pfile->line, |
352 | CPP_BUF_COL (buffer) - 2, | |
338fa5f7 | 353 | "\"/*\" within comment"); |
354 | } | |
355 | goto next_char; | |
0578f103 | 356 | } |
0578f103 | 357 | } |
78719282 | 358 | else if (is_vspace (c)) |
0578f103 | 359 | { |
36a0aa7c | 360 | prevc = c, c = handle_newline (pfile, c); |
338fa5f7 | 361 | goto next_char; |
0578f103 | 362 | } |
b86584f6 | 363 | else if (c == '\t') |
338fa5f7 | 364 | adjust_column (pfile); |
0578f103 | 365 | } |
f80e83a9 | 366 | |
f669338a | 367 | pfile->state.lexing_comment = 0; |
338fa5f7 | 368 | buffer->read_ahead = EOF; |
369 | return c != '/' || prevc != '*'; | |
0578f103 | 370 | } |
371 | ||
241e762e | 372 | /* Skip a C++ line comment. Handles escaped newlines. Returns |
338fa5f7 | 373 | non-zero if a multiline comment. The following new line, if any, |
374 | is left in buffer->read_ahead. */ | |
f80e83a9 | 375 | static int |
f669338a | 376 | skip_line_comment (pfile) |
377 | cpp_reader *pfile; | |
0578f103 | 378 | { |
f669338a | 379 | cpp_buffer *buffer = pfile->buffer; |
1ea7ed21 | 380 | unsigned int orig_line = pfile->line; |
338fa5f7 | 381 | cppchar_t c; |
f80e83a9 | 382 | |
f669338a | 383 | pfile->state.lexing_comment = 1; |
338fa5f7 | 384 | do |
f80e83a9 | 385 | { |
338fa5f7 | 386 | c = EOF; |
387 | if (buffer->cur == buffer->rlimit) | |
388 | break; | |
f80e83a9 | 389 | |
338fa5f7 | 390 | c = *buffer->cur++; |
391 | if (c == '?' || c == '\\') | |
c808d026 | 392 | c = skip_escaped_newlines (pfile, c); |
f80e83a9 | 393 | } |
338fa5f7 | 394 | while (!is_vspace (c)); |
0578f103 | 395 | |
f669338a | 396 | pfile->state.lexing_comment = 0; |
338fa5f7 | 397 | buffer->read_ahead = c; /* Leave any newline for caller. */ |
1ea7ed21 | 398 | return orig_line != pfile->line; |
f80e83a9 | 399 | } |
0578f103 | 400 | |
338fa5f7 | 401 | /* pfile->buffer->cur is one beyond the \t character. Update |
402 | col_adjust so we track the column correctly. */ | |
b86584f6 | 403 | static void |
338fa5f7 | 404 | adjust_column (pfile) |
b86584f6 | 405 | cpp_reader *pfile; |
b86584f6 | 406 | { |
338fa5f7 | 407 | cpp_buffer *buffer = pfile->buffer; |
408 | unsigned int col = CPP_BUF_COL (buffer) - 1; /* Zero-based column. */ | |
b86584f6 | 409 | |
410 | /* Round it up to multiple of the tabstop, but subtract 1 since the | |
411 | tab itself occupies a character position. */ | |
338fa5f7 | 412 | buffer->col_adjust += (CPP_OPTION (pfile, tabstop) |
413 | - col % CPP_OPTION (pfile, tabstop)) - 1; | |
b86584f6 | 414 | } |
415 | ||
338fa5f7 | 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. */ | |
f80e83a9 | 419 | static void |
338fa5f7 | 420 | skip_whitespace (pfile, c) |
f80e83a9 | 421 | cpp_reader *pfile; |
338fa5f7 | 422 | cppchar_t c; |
f80e83a9 | 423 | { |
424 | cpp_buffer *buffer = pfile->buffer; | |
338fa5f7 | 425 | unsigned int warned = 0; |
0578f103 | 426 | |
338fa5f7 | 427 | do |
f80e83a9 | 428 | { |
78719282 | 429 | /* Horizontal space always OK. */ |
430 | if (c == ' ') | |
338fa5f7 | 431 | ; |
78719282 | 432 | else if (c == '\t') |
338fa5f7 | 433 | adjust_column (pfile); |
434 | /* Just \f \v or \0 left. */ | |
78719282 | 435 | else if (c == '\0') |
f80e83a9 | 436 | { |
78719282 | 437 | if (!warned) |
338fa5f7 | 438 | { |
439 | cpp_warning (pfile, "null character(s) ignored"); | |
440 | warned = 1; | |
441 | } | |
0578f103 | 442 | } |
79bd622b | 443 | else if (pfile->state.in_directive && CPP_PEDANTIC (pfile)) |
1ea7ed21 | 444 | cpp_pedwarn_with_line (pfile, pfile->line, |
78719282 | 445 | CPP_BUF_COL (buffer), |
446 | "%s in preprocessing directive", | |
447 | c == '\f' ? "form feed" : "vertical tab"); | |
338fa5f7 | 448 | |
449 | c = EOF; | |
450 | if (buffer->cur == buffer->rlimit) | |
451 | break; | |
452 | c = *buffer->cur++; | |
0578f103 | 453 | } |
2c0e001b | 454 | /* We only want non-vertical space, i.e. ' ' \t \f \v \0. */ |
338fa5f7 | 455 | while (is_nvspace (c)); |
456 | ||
457 | /* Remember the next character. */ | |
458 | buffer->read_ahead = c; | |
f80e83a9 | 459 | } |
0578f103 | 460 | |
79bd622b | 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 | ||
66a5287e | 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. */ | |
338fa5f7 | 485 | |
486 | static cpp_hashnode * | |
66a5287e | 487 | parse_identifier (pfile) |
0578f103 | 488 | cpp_reader *pfile; |
0578f103 | 489 | { |
79bd622b | 490 | cpp_hashnode *result; |
66a5287e | 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 | { | |
338fa5f7 | 541 | cpp_buffer *buffer = pfile->buffer; |
66a5287e | 542 | const U_CHAR *base = buffer->cur - 1; |
0d086e18 | 543 | struct obstack *stack = &pfile->hash_table->stack; |
66a5287e | 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); | |
f80e83a9 | 548 | |
66a5287e | 549 | /* Now process the part which isn't. We are looking at one of |
550 | '$', '\\', or '?' on entry to this loop. */ | |
551 | c = *cur++; | |
552 | buffer->cur = cur; | |
338fa5f7 | 553 | do |
f80e83a9 | 554 | { |
66a5287e | 555 | while (is_idchar (c)) |
556 | { | |
557 | obstack_1grow (stack, c); | |
0578f103 | 558 | |
66a5287e | 559 | if (c == '$') |
560 | saw_dollar++; | |
71aa9da4 | 561 | |
66a5287e | 562 | c = EOF; |
563 | if (buffer->cur == buffer->rlimit) | |
564 | break; | |
71aa9da4 | 565 | |
66a5287e | 566 | c = *buffer->cur++; |
567 | } | |
71aa9da4 | 568 | |
338fa5f7 | 569 | /* Potential escaped newline? */ |
570 | if (c != '?' && c != '\\') | |
66a5287e | 571 | break; |
c808d026 | 572 | c = skip_escaped_newlines (pfile, c); |
f80e83a9 | 573 | } |
338fa5f7 | 574 | while (is_idchar (c)); |
575 | ||
79bd622b | 576 | /* Remember the next character. */ |
577 | buffer->read_ahead = c; | |
578 | ||
338fa5f7 | 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. */ | |
5e878517 | 582 | if (saw_dollar && CPP_PEDANTIC (pfile) && ! pfile->state.skipping) |
338fa5f7 | 583 | cpp_pedwarn (pfile, "'$' character(s) in identifier"); |
584 | ||
79bd622b | 585 | /* Identifiers are null-terminated. */ |
0d086e18 | 586 | len = obstack_object_size (stack); |
587 | obstack_1grow (stack, '\0'); | |
79bd622b | 588 | |
66a5287e | 589 | return (cpp_hashnode *) |
0d086e18 | 590 | ht_lookup (pfile->hash_table, obstack_finish (stack), len, HT_ALLOCED); |
0578f103 | 591 | } |
592 | ||
338fa5f7 | 593 | /* Parse a number, skipping embedded backslash-newlines. */ |
0578f103 | 594 | static void |
79bd622b | 595 | parse_number (pfile, number, c, leading_period) |
0578f103 | 596 | cpp_reader *pfile; |
338fa5f7 | 597 | cpp_string *number; |
598 | cppchar_t c; | |
79bd622b | 599 | int leading_period; |
0578f103 | 600 | { |
f80e83a9 | 601 | cpp_buffer *buffer = pfile->buffer; |
5f3f0010 | 602 | cpp_pool *pool = &pfile->ident_pool; |
79bd622b | 603 | unsigned char *dest, *limit; |
0578f103 | 604 | |
79bd622b | 605 | dest = POOL_FRONT (pool); |
606 | limit = POOL_LIMIT (pool); | |
f669338a | 607 | |
79bd622b | 608 | /* Place a leading period. */ |
609 | if (leading_period) | |
610 | { | |
611 | if (dest >= limit) | |
612 | limit = _cpp_next_chunk (pool, 0, &dest); | |
613 | *dest++ = '.'; | |
614 | } | |
615 | ||
338fa5f7 | 616 | do |
f80e83a9 | 617 | { |
338fa5f7 | 618 | do |
619 | { | |
79bd622b | 620 | /* Need room for terminating null. */ |
621 | if (dest + 1 >= limit) | |
622 | limit = _cpp_next_chunk (pool, 0, &dest); | |
623 | *dest++ = c; | |
338fa5f7 | 624 | |
338fa5f7 | 625 | c = EOF; |
626 | if (buffer->cur == buffer->rlimit) | |
627 | break; | |
0578f103 | 628 | |
338fa5f7 | 629 | c = *buffer->cur++; |
630 | } | |
79bd622b | 631 | while (is_numchar (c) || c == '.' || VALID_SIGN (c, dest[-1])); |
0578f103 | 632 | |
338fa5f7 | 633 | /* Potential escaped newline? */ |
634 | if (c != '?' && c != '\\') | |
635 | break; | |
c808d026 | 636 | c = skip_escaped_newlines (pfile, c); |
0578f103 | 637 | } |
79bd622b | 638 | while (is_numchar (c) || c == '.' || VALID_SIGN (c, dest[-1])); |
f669338a | 639 | |
338fa5f7 | 640 | /* Remember the next character. */ |
641 | buffer->read_ahead = c; | |
852d1b04 | 642 | |
79bd622b | 643 | /* Null-terminate the number. */ |
644 | *dest = '\0'; | |
645 | ||
646 | number->text = POOL_FRONT (pool); | |
647 | number->len = dest - number->text; | |
648 | POOL_COMMIT (pool, number->len + 1); | |
338fa5f7 | 649 | } |
650 | ||
651 | /* Subroutine of parse_string. Emits error for unterminated strings. */ | |
652 | static void | |
79bd622b | 653 | unterminated (pfile, term) |
338fa5f7 | 654 | cpp_reader *pfile; |
338fa5f7 | 655 | int term; |
656 | { | |
657 | cpp_error (pfile, "missing terminating %c character", term); | |
658 | ||
729d2022 | 659 | if (term == '\"' && pfile->mls_line && pfile->mls_line != pfile->line) |
f80e83a9 | 660 | { |
729d2022 | 661 | cpp_error_with_line (pfile, pfile->mls_line, pfile->mls_col, |
338fa5f7 | 662 | "possible start of unterminated string literal"); |
729d2022 | 663 | pfile->mls_line = 0; |
f80e83a9 | 664 | } |
0578f103 | 665 | } |
666 | ||
79bd622b | 667 | /* Subroutine of parse_string. */ |
668 | static int | |
669 | unescaped_terminator_p (pfile, dest) | |
670 | cpp_reader *pfile; | |
671 | const unsigned char *dest; | |
672 | { | |
673 | const unsigned char *start, *temp; | |
674 | ||
675 | /* In #include-style directives, terminators are not escapeable. */ | |
676 | if (pfile->state.angled_headers) | |
677 | return 1; | |
678 | ||
5f3f0010 | 679 | start = POOL_FRONT (&pfile->ident_pool); |
79bd622b | 680 | |
681 | /* An odd number of consecutive backslashes represents an escaped | |
682 | terminator. */ | |
683 | for (temp = dest; temp > start && temp[-1] == '\\'; temp--) | |
684 | ; | |
685 | ||
686 | return ((dest - temp) & 1) == 0; | |
687 | } | |
688 | ||
338fa5f7 | 689 | /* Parses a string, character constant, or angle-bracketed header file |
4b0c16ee | 690 | name. Handles embedded trigraphs and escaped newlines. The stored |
691 | string is guaranteed NUL-terminated, but it is not guaranteed that | |
692 | this is the first NUL since embedded NULs are preserved. | |
0578f103 | 693 | |
4b0c16ee | 694 | Multi-line strings are allowed, but they are deprecated. */ |
f80e83a9 | 695 | static void |
338fa5f7 | 696 | parse_string (pfile, token, terminator) |
0578f103 | 697 | cpp_reader *pfile; |
f80e83a9 | 698 | cpp_token *token; |
338fa5f7 | 699 | cppchar_t terminator; |
0578f103 | 700 | { |
f80e83a9 | 701 | cpp_buffer *buffer = pfile->buffer; |
5f3f0010 | 702 | cpp_pool *pool = &pfile->ident_pool; |
79bd622b | 703 | unsigned char *dest, *limit; |
338fa5f7 | 704 | cppchar_t c; |
38692459 | 705 | bool warned_nulls = false, warned_multi = false; |
338fa5f7 | 706 | |
79bd622b | 707 | dest = POOL_FRONT (pool); |
708 | limit = POOL_LIMIT (pool); | |
709 | ||
338fa5f7 | 710 | for (;;) |
0578f103 | 711 | { |
338fa5f7 | 712 | if (buffer->cur == buffer->rlimit) |
4b0c16ee | 713 | c = EOF; |
714 | else | |
715 | c = *buffer->cur++; | |
716 | ||
717 | have_char: | |
718 | /* We need space for the terminating NUL. */ | |
719 | if (dest >= limit) | |
720 | limit = _cpp_next_chunk (pool, 0, &dest); | |
721 | ||
722 | if (c == EOF) | |
338fa5f7 | 723 | { |
79bd622b | 724 | unterminated (pfile, terminator); |
338fa5f7 | 725 | break; |
726 | } | |
338fa5f7 | 727 | |
338fa5f7 | 728 | /* Handle trigraphs, escaped newlines etc. */ |
729 | if (c == '?' || c == '\\') | |
c808d026 | 730 | c = skip_escaped_newlines (pfile, c); |
0578f103 | 731 | |
79bd622b | 732 | if (c == terminator && unescaped_terminator_p (pfile, dest)) |
0578f103 | 733 | { |
79bd622b | 734 | c = EOF; |
735 | break; | |
338fa5f7 | 736 | } |
737 | else if (is_vspace (c)) | |
738 | { | |
739 | /* In assembly language, silently terminate string and | |
740 | character literals at end of line. This is a kludge | |
741 | around not knowing where comments are. */ | |
5db5d057 | 742 | if (CPP_OPTION (pfile, lang) == CLK_ASM && terminator != '>') |
338fa5f7 | 743 | break; |
0578f103 | 744 | |
338fa5f7 | 745 | /* Character constants and header names may not extend over |
746 | multiple lines. In Standard C, neither may strings. | |
747 | Unfortunately, we accept multiline strings as an | |
cc8770bf | 748 | extension, except in #include family directives. */ |
749 | if (terminator != '"' || pfile->state.angled_headers) | |
0578f103 | 750 | { |
79bd622b | 751 | unterminated (pfile, terminator); |
338fa5f7 | 752 | break; |
0578f103 | 753 | } |
0578f103 | 754 | |
38692459 | 755 | if (!warned_multi) |
756 | { | |
757 | warned_multi = true; | |
758 | cpp_pedwarn (pfile, "multi-line string literals are deprecated"); | |
759 | } | |
760 | ||
729d2022 | 761 | if (pfile->mls_line == 0) |
762 | { | |
763 | pfile->mls_line = token->line; | |
764 | pfile->mls_col = token->col; | |
765 | } | |
338fa5f7 | 766 | |
36a0aa7c | 767 | c = handle_newline (pfile, c); |
4b0c16ee | 768 | *dest++ = '\n'; |
769 | goto have_char; | |
338fa5f7 | 770 | } |
38692459 | 771 | else if (c == '\0' && !warned_nulls) |
338fa5f7 | 772 | { |
38692459 | 773 | warned_nulls = true; |
774 | cpp_warning (pfile, "null character(s) preserved in literal"); | |
0578f103 | 775 | } |
0578f103 | 776 | |
79bd622b | 777 | *dest++ = c; |
0578f103 | 778 | } |
779 | ||
79bd622b | 780 | /* Remember the next character. */ |
338fa5f7 | 781 | buffer->read_ahead = c; |
4b0c16ee | 782 | *dest = '\0'; |
0578f103 | 783 | |
79bd622b | 784 | token->val.str.text = POOL_FRONT (pool); |
785 | token->val.str.len = dest - token->val.str.text; | |
4b0c16ee | 786 | POOL_COMMIT (pool, token->val.str.len + 1); |
338fa5f7 | 787 | } |
f80e83a9 | 788 | |
79bd622b | 789 | /* The stored comment includes the comment start and any terminator. */ |
2c63d6c8 | 790 | static void |
338fa5f7 | 791 | save_comment (pfile, token, from) |
792 | cpp_reader *pfile; | |
f80e83a9 | 793 | cpp_token *token; |
794 | const unsigned char *from; | |
2c63d6c8 | 795 | { |
f80e83a9 | 796 | unsigned char *buffer; |
338fa5f7 | 797 | unsigned int len; |
338fa5f7 | 798 | |
f0495c2c | 799 | len = pfile->buffer->cur - from + 1; /* + 1 for the initial '/'. */ |
a543b315 | 800 | /* C++ comments probably (not definitely) have moved past a new |
801 | line, which we don't want to save in the comment. */ | |
802 | if (pfile->buffer->read_ahead != EOF) | |
803 | len--; | |
5f3f0010 | 804 | buffer = _cpp_pool_alloc (&pfile->ident_pool, len); |
f80e83a9 | 805 | |
f80e83a9 | 806 | token->type = CPP_COMMENT; |
76faa4c0 | 807 | token->val.str.len = len; |
338fa5f7 | 808 | token->val.str.text = buffer; |
0578f103 | 809 | |
f0495c2c | 810 | buffer[0] = '/'; |
811 | memcpy (buffer + 1, from, len - 1); | |
338fa5f7 | 812 | } |
0578f103 | 813 | |
10b4496a | 814 | /* Subroutine of _cpp_lex_direct to handle '%'. A little tricky, since we |
f669338a | 815 | want to avoid stepping back when lexing %:%X. */ |
338fa5f7 | 816 | static void |
c808d026 | 817 | lex_percent (pfile, result) |
818 | cpp_reader *pfile; | |
338fa5f7 | 819 | cpp_token *result; |
338fa5f7 | 820 | { |
c808d026 | 821 | cpp_buffer *buffer= pfile->buffer; |
f669338a | 822 | cppchar_t c; |
823 | ||
824 | result->type = CPP_MOD; | |
825 | /* Parsing %:%X could leave an extra character. */ | |
826 | if (buffer->extra_char == EOF) | |
c808d026 | 827 | c = get_effective_char (pfile); |
f669338a | 828 | else |
829 | { | |
830 | c = buffer->read_ahead = buffer->extra_char; | |
831 | buffer->extra_char = EOF; | |
832 | } | |
833 | ||
834 | if (c == '=') | |
835 | ACCEPT_CHAR (CPP_MOD_EQ); | |
c808d026 | 836 | else if (CPP_OPTION (pfile, digraphs)) |
f669338a | 837 | { |
838 | if (c == ':') | |
839 | { | |
840 | result->flags |= DIGRAPH; | |
841 | ACCEPT_CHAR (CPP_HASH); | |
c808d026 | 842 | if (get_effective_char (pfile) == '%') |
f669338a | 843 | { |
c808d026 | 844 | buffer->extra_char = get_effective_char (pfile); |
f669338a | 845 | if (buffer->extra_char == ':') |
846 | { | |
847 | buffer->extra_char = EOF; | |
848 | ACCEPT_CHAR (CPP_PASTE); | |
849 | } | |
850 | else | |
851 | /* We'll catch the extra_char when we're called back. */ | |
852 | buffer->read_ahead = '%'; | |
853 | } | |
854 | } | |
855 | else if (c == '>') | |
856 | { | |
857 | result->flags |= DIGRAPH; | |
858 | ACCEPT_CHAR (CPP_CLOSE_BRACE); | |
859 | } | |
860 | } | |
861 | } | |
862 | ||
10b4496a | 863 | /* Subroutine of _cpp_lex_direct to handle '.'. This is tricky, since we |
f669338a | 864 | want to avoid stepping back when lexing '...' or '.123'. In the |
865 | latter case we should also set a flag for parse_number. */ | |
866 | static void | |
867 | lex_dot (pfile, result) | |
868 | cpp_reader *pfile; | |
869 | cpp_token *result; | |
870 | { | |
871 | cpp_buffer *buffer = pfile->buffer; | |
872 | cppchar_t c; | |
873 | ||
874 | /* Parsing ..X could leave an extra character. */ | |
875 | if (buffer->extra_char == EOF) | |
c808d026 | 876 | c = get_effective_char (pfile); |
f669338a | 877 | else |
878 | { | |
879 | c = buffer->read_ahead = buffer->extra_char; | |
880 | buffer->extra_char = EOF; | |
881 | } | |
338fa5f7 | 882 | |
f669338a | 883 | /* All known character sets have 0...9 contiguous. */ |
884 | if (c >= '0' && c <= '9') | |
885 | { | |
886 | result->type = CPP_NUMBER; | |
79bd622b | 887 | parse_number (pfile, &result->val.str, c, 1); |
f669338a | 888 | } |
f80e83a9 | 889 | else |
c4357c92 | 890 | { |
f669338a | 891 | result->type = CPP_DOT; |
892 | if (c == '.') | |
893 | { | |
c808d026 | 894 | buffer->extra_char = get_effective_char (pfile); |
f669338a | 895 | if (buffer->extra_char == '.') |
896 | { | |
897 | buffer->extra_char = EOF; | |
898 | ACCEPT_CHAR (CPP_ELLIPSIS); | |
899 | } | |
900 | else | |
901 | /* We'll catch the extra_char when we're called back. */ | |
902 | buffer->read_ahead = '.'; | |
903 | } | |
904 | else if (c == '*' && CPP_OPTION (pfile, cplusplus)) | |
905 | ACCEPT_CHAR (CPP_DOT_STAR); | |
c4357c92 | 906 | } |
0578f103 | 907 | } |
908 | ||
83dcbb5c | 909 | /* Allocate COUNT tokens for RUN. */ |
910 | void | |
911 | _cpp_init_tokenrun (run, count) | |
912 | tokenrun *run; | |
913 | unsigned int count; | |
914 | { | |
915 | run->base = xnewvec (cpp_token, count); | |
916 | run->limit = run->base + count; | |
917 | run->next = NULL; | |
918 | } | |
919 | ||
920 | /* Returns the next tokenrun, or creates one if there is none. */ | |
921 | static tokenrun * | |
922 | next_tokenrun (run) | |
923 | tokenrun *run; | |
924 | { | |
925 | if (run->next == NULL) | |
926 | { | |
927 | run->next = xnew (tokenrun); | |
fb5ab82c | 928 | run->next->prev = run; |
83dcbb5c | 929 | _cpp_init_tokenrun (run->next, 250); |
930 | } | |
931 | ||
932 | return run->next; | |
933 | } | |
934 | ||
f9b5f742 | 935 | /* Allocate a single token that is invalidated at the same time as the |
936 | rest of the tokens on the line. Has its line and col set to the | |
937 | same as the last lexed token, so that diagnostics appear in the | |
938 | right place. */ | |
939 | cpp_token * | |
940 | _cpp_temp_token (pfile) | |
941 | cpp_reader *pfile; | |
942 | { | |
943 | cpp_token *old, *result; | |
944 | ||
945 | old = pfile->cur_token - 1; | |
946 | if (pfile->cur_token == pfile->cur_run->limit) | |
947 | { | |
948 | pfile->cur_run = next_tokenrun (pfile->cur_run); | |
949 | pfile->cur_token = pfile->cur_run->base; | |
950 | } | |
951 | ||
952 | result = pfile->cur_token++; | |
953 | result->line = old->line; | |
954 | result->col = old->col; | |
955 | return result; | |
956 | } | |
957 | ||
10b4496a | 958 | /* Lex a token into RESULT (external interface). Takes care of issues |
959 | like directive handling, token lookahead, multiple include | |
960 | opimisation and skipping. */ | |
c00e481c | 961 | const cpp_token * |
962 | _cpp_lex_token (pfile) | |
0578f103 | 963 | cpp_reader *pfile; |
83dcbb5c | 964 | { |
fb5ab82c | 965 | cpp_token *result; |
83dcbb5c | 966 | |
fb5ab82c | 967 | for (;;) |
83dcbb5c | 968 | { |
fb5ab82c | 969 | if (pfile->cur_token == pfile->cur_run->limit) |
83dcbb5c | 970 | { |
fb5ab82c | 971 | pfile->cur_run = next_tokenrun (pfile->cur_run); |
972 | pfile->cur_token = pfile->cur_run->base; | |
83dcbb5c | 973 | } |
974 | ||
fb5ab82c | 975 | if (pfile->lookaheads) |
10b4496a | 976 | { |
977 | pfile->lookaheads--; | |
978 | result = pfile->cur_token++; | |
979 | } | |
fb5ab82c | 980 | else |
10b4496a | 981 | result = _cpp_lex_direct (pfile); |
fb5ab82c | 982 | |
983 | if (result->flags & BOL) | |
83dcbb5c | 984 | { |
fb5ab82c | 985 | /* Is this a directive. If _cpp_handle_directive returns |
986 | false, it is an assembler #. */ | |
987 | if (result->type == CPP_HASH | |
988 | && !pfile->state.parsing_args | |
989 | && _cpp_handle_directive (pfile, result->flags & PREV_WHITE)) | |
990 | continue; | |
5621a364 | 991 | if (pfile->cb.line_change && !pfile->state.skipping) |
992 | (*pfile->cb.line_change)(pfile, result, pfile->state.parsing_args); | |
83dcbb5c | 993 | } |
83dcbb5c | 994 | |
fb5ab82c | 995 | /* We don't skip tokens in directives. */ |
996 | if (pfile->state.in_directive) | |
997 | break; | |
83dcbb5c | 998 | |
fb5ab82c | 999 | /* Outside a directive, invalidate controlling macros. At file |
10b4496a | 1000 | EOF, _cpp_lex_direct takes care of popping the buffer, so we never |
fb5ab82c | 1001 | get here and MI optimisation works. */ |
83dcbb5c | 1002 | pfile->mi_valid = false; |
fb5ab82c | 1003 | |
1004 | if (!pfile->state.skipping || result->type == CPP_EOF) | |
1005 | break; | |
83dcbb5c | 1006 | } |
1007 | ||
c00e481c | 1008 | return result; |
83dcbb5c | 1009 | } |
1010 | ||
10b4496a | 1011 | /* Lex a token into pfile->cur_token, which is also incremented, to |
1012 | get diagnostics pointing to the correct location. | |
1013 | ||
1014 | Does not handle issues such as token lookahead, multiple-include | |
1015 | optimisation, directives, skipping etc. This function is only | |
1016 | suitable for use by _cpp_lex_token, and in special cases like | |
1017 | lex_expansion_token which doesn't care for any of these issues. | |
1018 | ||
1019 | When meeting a newline, returns CPP_EOF if parsing a directive, | |
1020 | otherwise returns to the start of the token buffer if permissible. | |
1021 | Returns the location of the lexed token. */ | |
1022 | cpp_token * | |
1023 | _cpp_lex_direct (pfile) | |
83dcbb5c | 1024 | cpp_reader *pfile; |
0578f103 | 1025 | { |
338fa5f7 | 1026 | cppchar_t c; |
230f0943 | 1027 | cpp_buffer *buffer; |
338fa5f7 | 1028 | const unsigned char *comment_start; |
10b4496a | 1029 | cpp_token *result = pfile->cur_token++; |
0653b94e | 1030 | |
83dcbb5c | 1031 | fresh_line: |
230f0943 | 1032 | buffer = pfile->buffer; |
8c2e2fc5 | 1033 | result->flags = buffer->saved_flags; |
1034 | buffer->saved_flags = 0; | |
83dcbb5c | 1035 | update_tokens_line: |
36a0aa7c | 1036 | result->line = pfile->line; |
f80e83a9 | 1037 | |
83dcbb5c | 1038 | skipped_white: |
338fa5f7 | 1039 | c = buffer->read_ahead; |
1040 | if (c == EOF && buffer->cur < buffer->rlimit) | |
83dcbb5c | 1041 | c = *buffer->cur++; |
1042 | result->col = CPP_BUF_COLUMN (buffer, buffer->cur); | |
338fa5f7 | 1043 | buffer->read_ahead = EOF; |
83dcbb5c | 1044 | |
1045 | trigraph: | |
338fa5f7 | 1046 | switch (c) |
0578f103 | 1047 | { |
338fa5f7 | 1048 | case EOF: |
fb5ab82c | 1049 | buffer->saved_flags = BOL; |
83dcbb5c | 1050 | if (!pfile->state.parsing_args && !pfile->state.in_directive) |
4dfe8b74 | 1051 | { |
fb5ab82c | 1052 | if (buffer->cur != buffer->line_base) |
83dcbb5c | 1053 | { |
1054 | /* Non-empty files should end in a newline. Don't warn | |
1055 | for command line and _Pragma buffers. */ | |
1056 | if (!buffer->from_stage3) | |
1057 | cpp_pedwarn (pfile, "no newline at end of file"); | |
1058 | handle_newline (pfile, '\n'); | |
5475a165 | 1059 | } |
fb5ab82c | 1060 | |
1061 | /* Don't pop the last buffer. */ | |
1062 | if (buffer->prev) | |
1063 | { | |
1064 | unsigned char stop = buffer->return_at_eof; | |
1065 | ||
1066 | _cpp_pop_buffer (pfile); | |
1067 | if (!stop) | |
1068 | goto fresh_line; | |
1069 | } | |
4dfe8b74 | 1070 | } |
338fa5f7 | 1071 | result->type = CPP_EOF; |
83dcbb5c | 1072 | break; |
0578f103 | 1073 | |
338fa5f7 | 1074 | case ' ': case '\t': case '\f': case '\v': case '\0': |
1075 | skip_whitespace (pfile, c); | |
1076 | result->flags |= PREV_WHITE; | |
83dcbb5c | 1077 | goto skipped_white; |
338fa5f7 | 1078 | |
1079 | case '\n': case '\r': | |
fb5ab82c | 1080 | handle_newline (pfile, c); |
1081 | buffer->saved_flags = BOL; | |
1082 | if (! pfile->state.in_directive) | |
0578f103 | 1083 | { |
f9b5f742 | 1084 | if (pfile->state.parsing_args == 2) |
1085 | buffer->saved_flags |= PREV_WHITE; | |
fb5ab82c | 1086 | if (!pfile->keep_tokens) |
1087 | { | |
1088 | pfile->cur_run = &pfile->base_run; | |
1089 | result = pfile->base_run.base; | |
1090 | pfile->cur_token = result + 1; | |
1091 | } | |
1092 | goto fresh_line; | |
0578f103 | 1093 | } |
83dcbb5c | 1094 | result->type = CPP_EOF; |
1095 | break; | |
732cb4c9 | 1096 | |
338fa5f7 | 1097 | case '?': |
1098 | case '\\': | |
1099 | /* These could start an escaped newline, or '?' a trigraph. Let | |
1100 | skip_escaped_newlines do all the work. */ | |
1101 | { | |
1ea7ed21 | 1102 | unsigned int line = pfile->line; |
338fa5f7 | 1103 | |
c808d026 | 1104 | c = skip_escaped_newlines (pfile, c); |
1ea7ed21 | 1105 | if (line != pfile->line) |
338fa5f7 | 1106 | /* We had at least one escaped newline of some sort, and the |
1107 | next character is in buffer->read_ahead. Update the | |
1108 | token's line and column. */ | |
83dcbb5c | 1109 | goto update_tokens_line; |
338fa5f7 | 1110 | |
1111 | /* We are either the original '?' or '\\', or a trigraph. */ | |
1112 | result->type = CPP_QUERY; | |
1113 | buffer->read_ahead = EOF; | |
1114 | if (c == '\\') | |
3f90a920 | 1115 | goto random_char; |
338fa5f7 | 1116 | else if (c != '?') |
83dcbb5c | 1117 | goto trigraph; |
338fa5f7 | 1118 | } |
1119 | break; | |
732cb4c9 | 1120 | |
338fa5f7 | 1121 | case '0': case '1': case '2': case '3': case '4': |
1122 | case '5': case '6': case '7': case '8': case '9': | |
1123 | result->type = CPP_NUMBER; | |
79bd622b | 1124 | parse_number (pfile, &result->val.str, c, 0); |
338fa5f7 | 1125 | break; |
732cb4c9 | 1126 | |
338fa5f7 | 1127 | case '$': |
1128 | if (!CPP_OPTION (pfile, dollars_in_ident)) | |
1129 | goto random_char; | |
2c0e001b | 1130 | /* Fall through... */ |
338fa5f7 | 1131 | |
1132 | case '_': | |
1133 | case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': | |
1134 | case 'g': case 'h': case 'i': case 'j': case 'k': case 'l': | |
1135 | case 'm': case 'n': case 'o': case 'p': case 'q': case 'r': | |
1136 | case 's': case 't': case 'u': case 'v': case 'w': case 'x': | |
1137 | case 'y': case 'z': | |
1138 | case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': | |
1139 | case 'G': case 'H': case 'I': case 'J': case 'K': case 'L': | |
1140 | case 'M': case 'N': case 'O': case 'P': case 'Q': case 'R': | |
1141 | case 'S': case 'T': case 'U': case 'V': case 'W': case 'X': | |
1142 | case 'Y': case 'Z': | |
1143 | result->type = CPP_NAME; | |
66a5287e | 1144 | result->val.node = parse_identifier (pfile); |
338fa5f7 | 1145 | |
1146 | /* 'L' may introduce wide characters or strings. */ | |
79bd622b | 1147 | if (result->val.node == pfile->spec_nodes.n_L) |
338fa5f7 | 1148 | { |
66a5287e | 1149 | c = buffer->read_ahead; |
1150 | if (c == EOF && buffer->cur < buffer->rlimit) | |
1151 | c = *buffer->cur; | |
338fa5f7 | 1152 | if (c == '\'' || c == '"') |
71aa9da4 | 1153 | { |
66a5287e | 1154 | buffer->cur++; |
338fa5f7 | 1155 | ACCEPT_CHAR (c == '"' ? CPP_WSTRING: CPP_WCHAR); |
1156 | goto make_string; | |
71aa9da4 | 1157 | } |
338fa5f7 | 1158 | } |
1159 | /* Convert named operators to their proper types. */ | |
79bd622b | 1160 | else if (result->val.node->flags & NODE_OPERATOR) |
338fa5f7 | 1161 | { |
1162 | result->flags |= NAMED_OP; | |
79bd622b | 1163 | result->type = result->val.node->value.operator; |
338fa5f7 | 1164 | } |
1165 | break; | |
1166 | ||
1167 | case '\'': | |
1168 | case '"': | |
1169 | result->type = c == '"' ? CPP_STRING: CPP_CHAR; | |
1170 | make_string: | |
1171 | parse_string (pfile, result, c); | |
1172 | break; | |
f80e83a9 | 1173 | |
338fa5f7 | 1174 | case '/': |
f0495c2c | 1175 | /* A potential block or line comment. */ |
1176 | comment_start = buffer->cur; | |
338fa5f7 | 1177 | result->type = CPP_DIV; |
c808d026 | 1178 | c = get_effective_char (pfile); |
338fa5f7 | 1179 | if (c == '=') |
1180 | ACCEPT_CHAR (CPP_DIV_EQ); | |
f0495c2c | 1181 | if (c != '/' && c != '*') |
1182 | break; | |
20b8f8ff | 1183 | |
f0495c2c | 1184 | if (c == '*') |
1185 | { | |
338fa5f7 | 1186 | if (skip_block_comment (pfile)) |
1ea7ed21 | 1187 | cpp_error (pfile, "unterminated comment"); |
338fa5f7 | 1188 | } |
f0495c2c | 1189 | else |
338fa5f7 | 1190 | { |
f0495c2c | 1191 | if (!CPP_OPTION (pfile, cplusplus_comments) |
1192 | && !CPP_IN_SYSTEM_HEADER (pfile)) | |
1193 | break; | |
1194 | ||
5db5d057 | 1195 | /* Warn about comments only if pedantically GNUC89, and not |
1196 | in system headers. */ | |
1197 | if (CPP_OPTION (pfile, lang) == CLK_GNUC89 && CPP_PEDANTIC (pfile) | |
66914e49 | 1198 | && ! buffer->warned_cplusplus_comments) |
f80e83a9 | 1199 | { |
f0495c2c | 1200 | cpp_pedwarn (pfile, |
1201 | "C++ style comments are not allowed in ISO C89"); | |
1202 | cpp_pedwarn (pfile, | |
1203 | "(this will be reported only once per input file)"); | |
1204 | buffer->warned_cplusplus_comments = 1; | |
1205 | } | |
338fa5f7 | 1206 | |
66914e49 | 1207 | /* Skip_line_comment updates buffer->read_ahead. */ |
e1caf668 | 1208 | if (skip_line_comment (pfile) && CPP_OPTION (pfile, warn_comments)) |
729d2022 | 1209 | cpp_warning (pfile, "multi-line comment"); |
f0495c2c | 1210 | } |
338fa5f7 | 1211 | |
f0495c2c | 1212 | /* Skipping the comment has updated buffer->read_ahead. */ |
1213 | if (!pfile->state.save_comments) | |
1214 | { | |
1215 | result->flags |= PREV_WHITE; | |
83dcbb5c | 1216 | goto update_tokens_line; |
338fa5f7 | 1217 | } |
f0495c2c | 1218 | |
1219 | /* Save the comment as a token in its own right. */ | |
1220 | save_comment (pfile, result, comment_start); | |
fb5ab82c | 1221 | break; |
338fa5f7 | 1222 | |
1223 | case '<': | |
1224 | if (pfile->state.angled_headers) | |
1225 | { | |
1226 | result->type = CPP_HEADER_NAME; | |
1227 | c = '>'; /* terminator. */ | |
1228 | goto make_string; | |
1229 | } | |
0578f103 | 1230 | |
338fa5f7 | 1231 | result->type = CPP_LESS; |
c808d026 | 1232 | c = get_effective_char (pfile); |
338fa5f7 | 1233 | if (c == '=') |
1234 | ACCEPT_CHAR (CPP_LESS_EQ); | |
1235 | else if (c == '<') | |
1236 | { | |
1237 | ACCEPT_CHAR (CPP_LSHIFT); | |
c808d026 | 1238 | if (get_effective_char (pfile) == '=') |
338fa5f7 | 1239 | ACCEPT_CHAR (CPP_LSHIFT_EQ); |
1240 | } | |
1241 | else if (c == '?' && CPP_OPTION (pfile, cplusplus)) | |
1242 | { | |
1243 | ACCEPT_CHAR (CPP_MIN); | |
c808d026 | 1244 | if (get_effective_char (pfile) == '=') |
338fa5f7 | 1245 | ACCEPT_CHAR (CPP_MIN_EQ); |
1246 | } | |
1247 | else if (c == ':' && CPP_OPTION (pfile, digraphs)) | |
1248 | { | |
1249 | ACCEPT_CHAR (CPP_OPEN_SQUARE); | |
1250 | result->flags |= DIGRAPH; | |
1251 | } | |
1252 | else if (c == '%' && CPP_OPTION (pfile, digraphs)) | |
1253 | { | |
1254 | ACCEPT_CHAR (CPP_OPEN_BRACE); | |
1255 | result->flags |= DIGRAPH; | |
1256 | } | |
1257 | break; | |
1258 | ||
1259 | case '>': | |
1260 | result->type = CPP_GREATER; | |
c808d026 | 1261 | c = get_effective_char (pfile); |
338fa5f7 | 1262 | if (c == '=') |
1263 | ACCEPT_CHAR (CPP_GREATER_EQ); | |
1264 | else if (c == '>') | |
1265 | { | |
1266 | ACCEPT_CHAR (CPP_RSHIFT); | |
c808d026 | 1267 | if (get_effective_char (pfile) == '=') |
338fa5f7 | 1268 | ACCEPT_CHAR (CPP_RSHIFT_EQ); |
1269 | } | |
1270 | else if (c == '?' && CPP_OPTION (pfile, cplusplus)) | |
1271 | { | |
1272 | ACCEPT_CHAR (CPP_MAX); | |
c808d026 | 1273 | if (get_effective_char (pfile) == '=') |
338fa5f7 | 1274 | ACCEPT_CHAR (CPP_MAX_EQ); |
1275 | } | |
1276 | break; | |
1277 | ||
f669338a | 1278 | case '%': |
c808d026 | 1279 | lex_percent (pfile, result); |
338fa5f7 | 1280 | break; |
1281 | ||
f669338a | 1282 | case '.': |
1283 | lex_dot (pfile, result); | |
338fa5f7 | 1284 | break; |
0578f103 | 1285 | |
338fa5f7 | 1286 | case '+': |
1287 | result->type = CPP_PLUS; | |
c808d026 | 1288 | c = get_effective_char (pfile); |
338fa5f7 | 1289 | if (c == '=') |
1290 | ACCEPT_CHAR (CPP_PLUS_EQ); | |
1291 | else if (c == '+') | |
1292 | ACCEPT_CHAR (CPP_PLUS_PLUS); | |
1293 | break; | |
ac0749c7 | 1294 | |
338fa5f7 | 1295 | case '-': |
1296 | result->type = CPP_MINUS; | |
c808d026 | 1297 | c = get_effective_char (pfile); |
338fa5f7 | 1298 | if (c == '>') |
1299 | { | |
1300 | ACCEPT_CHAR (CPP_DEREF); | |
1301 | if (CPP_OPTION (pfile, cplusplus) | |
c808d026 | 1302 | && get_effective_char (pfile) == '*') |
338fa5f7 | 1303 | ACCEPT_CHAR (CPP_DEREF_STAR); |
1304 | } | |
1305 | else if (c == '=') | |
1306 | ACCEPT_CHAR (CPP_MINUS_EQ); | |
1307 | else if (c == '-') | |
1308 | ACCEPT_CHAR (CPP_MINUS_MINUS); | |
1309 | break; | |
0578f103 | 1310 | |
338fa5f7 | 1311 | case '*': |
1312 | result->type = CPP_MULT; | |
c808d026 | 1313 | if (get_effective_char (pfile) == '=') |
338fa5f7 | 1314 | ACCEPT_CHAR (CPP_MULT_EQ); |
1315 | break; | |
ac0749c7 | 1316 | |
338fa5f7 | 1317 | case '=': |
1318 | result->type = CPP_EQ; | |
c808d026 | 1319 | if (get_effective_char (pfile) == '=') |
338fa5f7 | 1320 | ACCEPT_CHAR (CPP_EQ_EQ); |
1321 | break; | |
c4abf88d | 1322 | |
338fa5f7 | 1323 | case '!': |
1324 | result->type = CPP_NOT; | |
c808d026 | 1325 | if (get_effective_char (pfile) == '=') |
338fa5f7 | 1326 | ACCEPT_CHAR (CPP_NOT_EQ); |
1327 | break; | |
0578f103 | 1328 | |
338fa5f7 | 1329 | case '&': |
1330 | result->type = CPP_AND; | |
c808d026 | 1331 | c = get_effective_char (pfile); |
338fa5f7 | 1332 | if (c == '=') |
1333 | ACCEPT_CHAR (CPP_AND_EQ); | |
1334 | else if (c == '&') | |
1335 | ACCEPT_CHAR (CPP_AND_AND); | |
1336 | break; | |
1337 | ||
1338 | case '#': | |
e14c5993 | 1339 | result->type = CPP_HASH; |
83dcbb5c | 1340 | if (get_effective_char (pfile) == '#') |
1341 | ACCEPT_CHAR (CPP_PASTE); | |
338fa5f7 | 1342 | break; |
0578f103 | 1343 | |
338fa5f7 | 1344 | case '|': |
1345 | result->type = CPP_OR; | |
c808d026 | 1346 | c = get_effective_char (pfile); |
338fa5f7 | 1347 | if (c == '=') |
1348 | ACCEPT_CHAR (CPP_OR_EQ); | |
1349 | else if (c == '|') | |
1350 | ACCEPT_CHAR (CPP_OR_OR); | |
1351 | break; | |
0578f103 | 1352 | |
338fa5f7 | 1353 | case '^': |
1354 | result->type = CPP_XOR; | |
c808d026 | 1355 | if (get_effective_char (pfile) == '=') |
338fa5f7 | 1356 | ACCEPT_CHAR (CPP_XOR_EQ); |
1357 | break; | |
0578f103 | 1358 | |
338fa5f7 | 1359 | case ':': |
1360 | result->type = CPP_COLON; | |
c808d026 | 1361 | c = get_effective_char (pfile); |
338fa5f7 | 1362 | if (c == ':' && CPP_OPTION (pfile, cplusplus)) |
1363 | ACCEPT_CHAR (CPP_SCOPE); | |
1364 | else if (c == '>' && CPP_OPTION (pfile, digraphs)) | |
1365 | { | |
1366 | result->flags |= DIGRAPH; | |
1367 | ACCEPT_CHAR (CPP_CLOSE_SQUARE); | |
1368 | } | |
1369 | break; | |
0578f103 | 1370 | |
338fa5f7 | 1371 | case '~': result->type = CPP_COMPL; break; |
1372 | case ',': result->type = CPP_COMMA; break; | |
1373 | case '(': result->type = CPP_OPEN_PAREN; break; | |
1374 | case ')': result->type = CPP_CLOSE_PAREN; break; | |
1375 | case '[': result->type = CPP_OPEN_SQUARE; break; | |
1376 | case ']': result->type = CPP_CLOSE_SQUARE; break; | |
1377 | case '{': result->type = CPP_OPEN_BRACE; break; | |
1378 | case '}': result->type = CPP_CLOSE_BRACE; break; | |
1379 | case ';': result->type = CPP_SEMICOLON; break; | |
1380 | ||
9ee99ac6 | 1381 | /* @ is a punctuator in Objective C. */ |
1382 | case '@': result->type = CPP_ATSIGN; break; | |
338fa5f7 | 1383 | |
1384 | random_char: | |
1385 | default: | |
1386 | result->type = CPP_OTHER; | |
33344a1c | 1387 | result->val.c = c; |
338fa5f7 | 1388 | break; |
1389 | } | |
fb5ab82c | 1390 | |
1391 | return result; | |
338fa5f7 | 1392 | } |
1393 | ||
79bd622b | 1394 | /* An upper bound on the number of bytes needed to spell a token, |
1395 | including preceding whitespace. */ | |
1396 | unsigned int | |
1397 | cpp_token_len (token) | |
1398 | const cpp_token *token; | |
338fa5f7 | 1399 | { |
79bd622b | 1400 | unsigned int len; |
cfad5579 | 1401 | |
79bd622b | 1402 | switch (TOKEN_SPELL (token)) |
f80e83a9 | 1403 | { |
c86dbc5b | 1404 | default: len = 0; break; |
1405 | case SPELL_STRING: len = token->val.str.len; break; | |
1406 | case SPELL_IDENT: len = NODE_LEN (token->val.node); break; | |
f80e83a9 | 1407 | } |
79bd622b | 1408 | /* 1 for whitespace, 4 for comment delimeters. */ |
1409 | return len + 5; | |
cfad5579 | 1410 | } |
1411 | ||
f80e83a9 | 1412 | /* Write the spelling of a token TOKEN to BUFFER. The buffer must |
c5ea33a8 | 1413 | already contain the enough space to hold the token's spelling. |
1414 | Returns a pointer to the character after the last character | |
1415 | written. */ | |
79bd622b | 1416 | unsigned char * |
1417 | cpp_spell_token (pfile, token, buffer) | |
f80e83a9 | 1418 | cpp_reader *pfile; /* Would be nice to be rid of this... */ |
1419 | const cpp_token *token; | |
1420 | unsigned char *buffer; | |
1421 | { | |
7e842f95 | 1422 | switch (TOKEN_SPELL (token)) |
f80e83a9 | 1423 | { |
1424 | case SPELL_OPERATOR: | |
1425 | { | |
1426 | const unsigned char *spelling; | |
1427 | unsigned char c; | |
ab12a39c | 1428 | |
f80e83a9 | 1429 | if (token->flags & DIGRAPH) |
ee6c4e4b | 1430 | spelling |
1431 | = digraph_spellings[(int) token->type - (int) CPP_FIRST_DIGRAPH]; | |
31674461 | 1432 | else if (token->flags & NAMED_OP) |
1433 | goto spell_ident; | |
f80e83a9 | 1434 | else |
7e842f95 | 1435 | spelling = TOKEN_NAME (token); |
f80e83a9 | 1436 | |
1437 | while ((c = *spelling++) != '\0') | |
1438 | *buffer++ = c; | |
1439 | } | |
1440 | break; | |
ab12a39c | 1441 | |
f80e83a9 | 1442 | case SPELL_IDENT: |
31674461 | 1443 | spell_ident: |
c86dbc5b | 1444 | memcpy (buffer, NODE_NAME (token->val.node), NODE_LEN (token->val.node)); |
1445 | buffer += NODE_LEN (token->val.node); | |
f80e83a9 | 1446 | break; |
ab12a39c | 1447 | |
f80e83a9 | 1448 | case SPELL_STRING: |
1449 | { | |
71aa9da4 | 1450 | int left, right, tag; |
1451 | switch (token->type) | |
1452 | { | |
1453 | case CPP_STRING: left = '"'; right = '"'; tag = '\0'; break; | |
1454 | case CPP_WSTRING: left = '"'; right = '"'; tag = 'L'; break; | |
71aa9da4 | 1455 | case CPP_CHAR: left = '\''; right = '\''; tag = '\0'; break; |
1456 | case CPP_WCHAR: left = '\''; right = '\''; tag = 'L'; break; | |
1457 | case CPP_HEADER_NAME: left = '<'; right = '>'; tag = '\0'; break; | |
1458 | default: left = '\0'; right = '\0'; tag = '\0'; break; | |
1459 | } | |
1460 | if (tag) *buffer++ = tag; | |
1461 | if (left) *buffer++ = left; | |
76faa4c0 | 1462 | memcpy (buffer, token->val.str.text, token->val.str.len); |
1463 | buffer += token->val.str.len; | |
71aa9da4 | 1464 | if (right) *buffer++ = right; |
f80e83a9 | 1465 | } |
1466 | break; | |
ab12a39c | 1467 | |
f80e83a9 | 1468 | case SPELL_CHAR: |
33344a1c | 1469 | *buffer++ = token->val.c; |
f80e83a9 | 1470 | break; |
ab12a39c | 1471 | |
f80e83a9 | 1472 | case SPELL_NONE: |
7e842f95 | 1473 | cpp_ice (pfile, "Unspellable token %s", TOKEN_NAME (token)); |
f80e83a9 | 1474 | break; |
1475 | } | |
ab12a39c | 1476 | |
f80e83a9 | 1477 | return buffer; |
1478 | } | |
ab12a39c | 1479 | |
79bd622b | 1480 | /* Returns a token as a null-terminated string. The string is |
1481 | temporary, and automatically freed later. Useful for diagnostics. */ | |
1482 | unsigned char * | |
1483 | cpp_token_as_text (pfile, token) | |
6060326b | 1484 | cpp_reader *pfile; |
f80e83a9 | 1485 | const cpp_token *token; |
6060326b | 1486 | { |
79bd622b | 1487 | unsigned int len = cpp_token_len (token); |
5f3f0010 | 1488 | unsigned char *start = _cpp_pool_alloc (&pfile->ident_pool, len), *end; |
6060326b | 1489 | |
79bd622b | 1490 | end = cpp_spell_token (pfile, token, start); |
1491 | end[0] = '\0'; | |
6060326b | 1492 | |
79bd622b | 1493 | return start; |
1494 | } | |
6060326b | 1495 | |
79bd622b | 1496 | /* Used by C front ends. Should really move to using cpp_token_as_text. */ |
1497 | const char * | |
1498 | cpp_type2name (type) | |
1499 | enum cpp_ttype type; | |
1500 | { | |
1501 | return (const char *) token_spellings[type].name; | |
1502 | } | |
6060326b | 1503 | |
f9b5f742 | 1504 | /* Writes the spelling of token to FP, without any preceding space. |
1505 | Separated from cpp_spell_token for efficiency - to avoid stdio | |
1506 | double-buffering. */ | |
79bd622b | 1507 | void |
1508 | cpp_output_token (token, fp) | |
1509 | const cpp_token *token; | |
1510 | FILE *fp; | |
1511 | { | |
79bd622b | 1512 | switch (TOKEN_SPELL (token)) |
6060326b | 1513 | { |
79bd622b | 1514 | case SPELL_OPERATOR: |
1515 | { | |
1516 | const unsigned char *spelling; | |
28874558 | 1517 | int c; |
6060326b | 1518 | |
79bd622b | 1519 | if (token->flags & DIGRAPH) |
ee6c4e4b | 1520 | spelling |
1521 | = digraph_spellings[(int) token->type - (int) CPP_FIRST_DIGRAPH]; | |
79bd622b | 1522 | else if (token->flags & NAMED_OP) |
1523 | goto spell_ident; | |
1524 | else | |
1525 | spelling = TOKEN_NAME (token); | |
f80e83a9 | 1526 | |
28874558 | 1527 | c = *spelling; |
1528 | do | |
1529 | putc (c, fp); | |
1530 | while ((c = *++spelling) != '\0'); | |
79bd622b | 1531 | } |
1532 | break; | |
f80e83a9 | 1533 | |
79bd622b | 1534 | spell_ident: |
1535 | case SPELL_IDENT: | |
28874558 | 1536 | fwrite (NODE_NAME (token->val.node), 1, NODE_LEN (token->val.node), fp); |
79bd622b | 1537 | break; |
f80e83a9 | 1538 | |
79bd622b | 1539 | case SPELL_STRING: |
1540 | { | |
1541 | int left, right, tag; | |
1542 | switch (token->type) | |
1543 | { | |
1544 | case CPP_STRING: left = '"'; right = '"'; tag = '\0'; break; | |
1545 | case CPP_WSTRING: left = '"'; right = '"'; tag = 'L'; break; | |
79bd622b | 1546 | case CPP_CHAR: left = '\''; right = '\''; tag = '\0'; break; |
1547 | case CPP_WCHAR: left = '\''; right = '\''; tag = 'L'; break; | |
1548 | case CPP_HEADER_NAME: left = '<'; right = '>'; tag = '\0'; break; | |
1549 | default: left = '\0'; right = '\0'; tag = '\0'; break; | |
1550 | } | |
1551 | if (tag) putc (tag, fp); | |
1552 | if (left) putc (left, fp); | |
1553 | fwrite (token->val.str.text, 1, token->val.str.len, fp); | |
1554 | if (right) putc (right, fp); | |
1555 | } | |
1556 | break; | |
6060326b | 1557 | |
79bd622b | 1558 | case SPELL_CHAR: |
33344a1c | 1559 | putc (token->val.c, fp); |
79bd622b | 1560 | break; |
6060326b | 1561 | |
79bd622b | 1562 | case SPELL_NONE: |
1563 | /* An error, most probably. */ | |
1564 | break; | |
f80e83a9 | 1565 | } |
6060326b | 1566 | } |
1567 | ||
79bd622b | 1568 | /* Compare two tokens. */ |
1569 | int | |
1570 | _cpp_equiv_tokens (a, b) | |
1571 | const cpp_token *a, *b; | |
6060326b | 1572 | { |
79bd622b | 1573 | if (a->type == b->type && a->flags == b->flags) |
1574 | switch (TOKEN_SPELL (a)) | |
1575 | { | |
1576 | default: /* Keep compiler happy. */ | |
1577 | case SPELL_OPERATOR: | |
1578 | return 1; | |
1579 | case SPELL_CHAR: | |
33344a1c | 1580 | return a->val.c == b->val.c; /* Character. */ |
79bd622b | 1581 | case SPELL_NONE: |
588d632b | 1582 | return (a->type != CPP_MACRO_ARG || a->val.arg_no == b->val.arg_no); |
79bd622b | 1583 | case SPELL_IDENT: |
1584 | return a->val.node == b->val.node; | |
1585 | case SPELL_STRING: | |
1586 | return (a->val.str.len == b->val.str.len | |
1587 | && !memcmp (a->val.str.text, b->val.str.text, | |
1588 | a->val.str.len)); | |
1589 | } | |
6060326b | 1590 | |
f80e83a9 | 1591 | return 0; |
1592 | } | |
1593 | ||
79bd622b | 1594 | /* Returns nonzero if a space should be inserted to avoid an |
1595 | accidental token paste for output. For simplicity, it is | |
1596 | conservative, and occasionally advises a space where one is not | |
1597 | needed, e.g. "." and ".2". */ | |
f80e83a9 | 1598 | |
79bd622b | 1599 | int |
1600 | cpp_avoid_paste (pfile, token1, token2) | |
6060326b | 1601 | cpp_reader *pfile; |
79bd622b | 1602 | const cpp_token *token1, *token2; |
6060326b | 1603 | { |
79bd622b | 1604 | enum cpp_ttype a = token1->type, b = token2->type; |
1605 | cppchar_t c; | |
6060326b | 1606 | |
79bd622b | 1607 | if (token1->flags & NAMED_OP) |
1608 | a = CPP_NAME; | |
1609 | if (token2->flags & NAMED_OP) | |
1610 | b = CPP_NAME; | |
6060326b | 1611 | |
79bd622b | 1612 | c = EOF; |
1613 | if (token2->flags & DIGRAPH) | |
ee6c4e4b | 1614 | c = digraph_spellings[(int) b - (int) CPP_FIRST_DIGRAPH][0]; |
79bd622b | 1615 | else if (token_spellings[b].category == SPELL_OPERATOR) |
1616 | c = token_spellings[b].name[0]; | |
6060326b | 1617 | |
79bd622b | 1618 | /* Quickly get everything that can paste with an '='. */ |
ee6c4e4b | 1619 | if ((int) a <= (int) CPP_LAST_EQ && c == '=') |
79bd622b | 1620 | return 1; |
6060326b | 1621 | |
79bd622b | 1622 | switch (a) |
6060326b | 1623 | { |
79bd622b | 1624 | case CPP_GREATER: return c == '>' || c == '?'; |
1625 | case CPP_LESS: return c == '<' || c == '?' || c == '%' || c == ':'; | |
1626 | case CPP_PLUS: return c == '+'; | |
1627 | case CPP_MINUS: return c == '-' || c == '>'; | |
1628 | case CPP_DIV: return c == '/' || c == '*'; /* Comments. */ | |
1629 | case CPP_MOD: return c == ':' || c == '>'; | |
1630 | case CPP_AND: return c == '&'; | |
1631 | case CPP_OR: return c == '|'; | |
1632 | case CPP_COLON: return c == ':' || c == '>'; | |
1633 | case CPP_DEREF: return c == '*'; | |
efdcc728 | 1634 | case CPP_DOT: return c == '.' || c == '%' || b == CPP_NUMBER; |
79bd622b | 1635 | case CPP_HASH: return c == '#' || c == '%'; /* Digraph form. */ |
1636 | case CPP_NAME: return ((b == CPP_NUMBER | |
1637 | && name_p (pfile, &token2->val.str)) | |
1638 | || b == CPP_NAME | |
1639 | || b == CPP_CHAR || b == CPP_STRING); /* L */ | |
1640 | case CPP_NUMBER: return (b == CPP_NUMBER || b == CPP_NAME | |
1641 | || c == '.' || c == '+' || c == '-'); | |
1642 | case CPP_OTHER: return (CPP_OPTION (pfile, objc) | |
33344a1c | 1643 | && token1->val.c == '@' |
79bd622b | 1644 | && (b == CPP_NAME || b == CPP_STRING)); |
1645 | default: break; | |
6060326b | 1646 | } |
6060326b | 1647 | |
deb356cf | 1648 | return 0; |
6060326b | 1649 | } |
1650 | ||
79bd622b | 1651 | /* Output all the remaining tokens on the current line, and a newline |
f9b5f742 | 1652 | character, to FP. Leading whitespace is removed. If there are |
1653 | macros, special token padding is not performed. */ | |
6060326b | 1654 | void |
79bd622b | 1655 | cpp_output_line (pfile, fp) |
6060326b | 1656 | cpp_reader *pfile; |
79bd622b | 1657 | FILE *fp; |
6060326b | 1658 | { |
f9b5f742 | 1659 | const cpp_token *token; |
7e842f95 | 1660 | |
f9b5f742 | 1661 | token = cpp_get_token (pfile); |
1662 | while (token->type != CPP_EOF) | |
7e842f95 | 1663 | { |
f9b5f742 | 1664 | cpp_output_token (token, fp); |
1665 | token = cpp_get_token (pfile); | |
1666 | if (token->flags & PREV_WHITE) | |
1667 | putc (' ', fp); | |
7e842f95 | 1668 | } |
1669 | ||
79bd622b | 1670 | putc ('\n', fp); |
f80e83a9 | 1671 | } |
6060326b | 1672 | |
8330799c | 1673 | /* Returns the value of a hexadecimal digit. */ |
1674 | static unsigned int | |
1675 | hex_digit_value (c) | |
1676 | unsigned int c; | |
1677 | { | |
1678 | if (c >= 'a' && c <= 'f') | |
1679 | return c - 'a' + 10; | |
1680 | if (c >= 'A' && c <= 'F') | |
1681 | return c - 'A' + 10; | |
1682 | if (c >= '0' && c <= '9') | |
1683 | return c - '0'; | |
1684 | abort (); | |
1685 | } | |
1686 | ||
c8342759 | 1687 | /* Parse a '\uNNNN' or '\UNNNNNNNN' sequence. Returns 1 to indicate |
1688 | failure if cpplib is not parsing C++ or C99. Such failure is | |
1689 | silent, and no variables are updated. Otherwise returns 0, and | |
1690 | warns if -Wtraditional. | |
8330799c | 1691 | |
1692 | [lex.charset]: The character designated by the universal character | |
1693 | name \UNNNNNNNN is that character whose character short name in | |
1694 | ISO/IEC 10646 is NNNNNNNN; the character designated by the | |
1695 | universal character name \uNNNN is that character whose character | |
1696 | short name in ISO/IEC 10646 is 0000NNNN. If the hexadecimal value | |
1697 | for a universal character name is less than 0x20 or in the range | |
1698 | 0x7F-0x9F (inclusive), or if the universal character name | |
1699 | designates a character in the basic source character set, then the | |
1700 | program is ill-formed. | |
1701 | ||
1702 | We assume that wchar_t is Unicode, so we don't need to do any | |
c8342759 | 1703 | mapping. Is this ever wrong? |
8330799c | 1704 | |
c8342759 | 1705 | PC points to the 'u' or 'U', PSTR is points to the byte after PC, |
1706 | LIMIT is the end of the string or charconst. PSTR is updated to | |
1707 | point after the UCS on return, and the UCS is written into PC. */ | |
1708 | ||
1709 | static int | |
1710 | maybe_read_ucs (pfile, pstr, limit, pc) | |
8330799c | 1711 | cpp_reader *pfile; |
1712 | const unsigned char **pstr; | |
1713 | const unsigned char *limit; | |
c8342759 | 1714 | unsigned int *pc; |
8330799c | 1715 | { |
1716 | const unsigned char *p = *pstr; | |
c8342759 | 1717 | unsigned int code = 0; |
1718 | unsigned int c = *pc, length; | |
1719 | ||
1720 | /* Only attempt to interpret a UCS for C++ and C99. */ | |
1721 | if (! (CPP_OPTION (pfile, cplusplus) || CPP_OPTION (pfile, c99))) | |
1722 | return 1; | |
8330799c | 1723 | |
c8342759 | 1724 | if (CPP_WTRADITIONAL (pfile)) |
1725 | cpp_warning (pfile, "the meaning of '\\%c' varies with -traditional", c); | |
8330799c | 1726 | |
f73bab03 | 1727 | length = (c == 'u' ? 4: 8); |
1728 | ||
1729 | if ((size_t) (limit - p) < length) | |
1730 | { | |
1731 | cpp_error (pfile, "incomplete universal-character-name"); | |
1732 | /* Skip to the end to avoid more diagnostics. */ | |
1733 | p = limit; | |
1734 | } | |
1735 | else | |
1736 | { | |
1737 | for (; length; length--, p++) | |
8330799c | 1738 | { |
f73bab03 | 1739 | c = *p; |
1740 | if (ISXDIGIT (c)) | |
1741 | code = (code << 4) + hex_digit_value (c); | |
1742 | else | |
1743 | { | |
1744 | cpp_error (pfile, | |
1745 | "non-hex digit '%c' in universal-character-name", c); | |
1746 | /* We shouldn't skip in case there are multibyte chars. */ | |
1747 | break; | |
1748 | } | |
8330799c | 1749 | } |
8330799c | 1750 | } |
1751 | ||
1752 | #ifdef TARGET_EBCDIC | |
1753 | cpp_error (pfile, "universal-character-name on EBCDIC target"); | |
1754 | code = 0x3f; /* EBCDIC invalid character */ | |
1755 | #else | |
f73bab03 | 1756 | /* True extended characters are OK. */ |
1757 | if (code >= 0xa0 | |
1758 | && !(code & 0x80000000) | |
1759 | && !(code >= 0xD800 && code <= 0xDFFF)) | |
1760 | ; | |
1761 | /* The standard permits $, @ and ` to be specified as UCNs. We use | |
1762 | hex escapes so that this also works with EBCDIC hosts. */ | |
1763 | else if (code == 0x24 || code == 0x40 || code == 0x60) | |
1764 | ; | |
1765 | /* Don't give another error if one occurred above. */ | |
1766 | else if (length == 0) | |
1767 | cpp_error (pfile, "universal-character-name out of range"); | |
8330799c | 1768 | #endif |
1769 | ||
1770 | *pstr = p; | |
c8342759 | 1771 | *pc = code; |
1772 | return 0; | |
8330799c | 1773 | } |
1774 | ||
1775 | /* Interpret an escape sequence, and return its value. PSTR points to | |
1776 | the input pointer, which is just after the backslash. LIMIT is how | |
c8342759 | 1777 | much text we have. MASK is a bitmask for the precision for the |
1778 | destination type (char or wchar_t). TRADITIONAL, if true, does not | |
1779 | interpret escapes that did not exist in traditional C. | |
8330799c | 1780 | |
c8342759 | 1781 | Handles all relevant diagnostics. */ |
1782 | ||
1783 | unsigned int | |
1784 | cpp_parse_escape (pfile, pstr, limit, mask, traditional) | |
8330799c | 1785 | cpp_reader *pfile; |
1786 | const unsigned char **pstr; | |
1787 | const unsigned char *limit; | |
c8342759 | 1788 | unsigned HOST_WIDE_INT mask; |
8330799c | 1789 | int traditional; |
1790 | { | |
1791 | int unknown = 0; | |
1792 | const unsigned char *str = *pstr; | |
1793 | unsigned int c = *str++; | |
1794 | ||
1795 | switch (c) | |
1796 | { | |
1797 | case '\\': case '\'': case '"': case '?': break; | |
1798 | case 'b': c = TARGET_BS; break; | |
1799 | case 'f': c = TARGET_FF; break; | |
1800 | case 'n': c = TARGET_NEWLINE; break; | |
1801 | case 'r': c = TARGET_CR; break; | |
1802 | case 't': c = TARGET_TAB; break; | |
1803 | case 'v': c = TARGET_VT; break; | |
1804 | ||
1805 | case '(': case '{': case '[': case '%': | |
1806 | /* '\(', etc, are used at beginning of line to avoid confusing Emacs. | |
1807 | '\%' is used to prevent SCCS from getting confused. */ | |
1808 | unknown = CPP_PEDANTIC (pfile); | |
1809 | break; | |
1810 | ||
1811 | case 'a': | |
1812 | if (CPP_WTRADITIONAL (pfile)) | |
1813 | cpp_warning (pfile, "the meaning of '\\a' varies with -traditional"); | |
1814 | if (!traditional) | |
1815 | c = TARGET_BELL; | |
1816 | break; | |
1817 | ||
1818 | case 'e': case 'E': | |
1819 | if (CPP_PEDANTIC (pfile)) | |
1820 | cpp_pedwarn (pfile, "non-ISO-standard escape sequence, '\\%c'", c); | |
1821 | c = TARGET_ESC; | |
1822 | break; | |
1823 | ||
8330799c | 1824 | case 'u': case 'U': |
c8342759 | 1825 | unknown = maybe_read_ucs (pfile, &str, limit, &c); |
8330799c | 1826 | break; |
1827 | ||
1828 | case 'x': | |
1829 | if (CPP_WTRADITIONAL (pfile)) | |
1830 | cpp_warning (pfile, "the meaning of '\\x' varies with -traditional"); | |
1831 | ||
1832 | if (!traditional) | |
1833 | { | |
1834 | unsigned int i = 0, overflow = 0; | |
1835 | int digits_found = 0; | |
1836 | ||
1837 | while (str < limit) | |
1838 | { | |
1839 | c = *str; | |
1840 | if (! ISXDIGIT (c)) | |
1841 | break; | |
1842 | str++; | |
1843 | overflow |= i ^ (i << 4 >> 4); | |
1844 | i = (i << 4) + hex_digit_value (c); | |
1845 | digits_found = 1; | |
1846 | } | |
1847 | ||
1848 | if (!digits_found) | |
1849 | cpp_error (pfile, "\\x used with no following hex digits"); | |
1850 | ||
1851 | if (overflow | (i != (i & mask))) | |
1852 | { | |
1853 | cpp_pedwarn (pfile, "hex escape sequence out of range"); | |
1854 | i &= mask; | |
1855 | } | |
1856 | c = i; | |
1857 | } | |
1858 | break; | |
1859 | ||
1860 | case '0': case '1': case '2': case '3': | |
1861 | case '4': case '5': case '6': case '7': | |
1862 | { | |
1863 | unsigned int i = c - '0'; | |
1864 | int count = 0; | |
1865 | ||
1866 | while (str < limit && ++count < 3) | |
1867 | { | |
1868 | c = *str; | |
1869 | if (c < '0' || c > '7') | |
1870 | break; | |
1871 | str++; | |
1872 | i = (i << 3) + c - '0'; | |
1873 | } | |
1874 | ||
1875 | if (i != (i & mask)) | |
1876 | { | |
1877 | cpp_pedwarn (pfile, "octal escape sequence out of range"); | |
1878 | i &= mask; | |
1879 | } | |
1880 | c = i; | |
1881 | } | |
1882 | break; | |
1883 | ||
1884 | default: | |
1885 | unknown = 1; | |
1886 | break; | |
1887 | } | |
1888 | ||
1889 | if (unknown) | |
1890 | { | |
1891 | if (ISGRAPH (c)) | |
1892 | cpp_pedwarn (pfile, "unknown escape sequence '\\%c'", c); | |
1893 | else | |
1894 | cpp_pedwarn (pfile, "unknown escape sequence: '\\%03o'", c); | |
1895 | } | |
1896 | ||
c8342759 | 1897 | if (c > mask) |
1898 | cpp_pedwarn (pfile, "escape sequence out of range for character"); | |
1899 | ||
8330799c | 1900 | *pstr = str; |
1901 | return c; | |
1902 | } | |
1903 | ||
1904 | #ifndef MAX_CHAR_TYPE_SIZE | |
1905 | #define MAX_CHAR_TYPE_SIZE CHAR_TYPE_SIZE | |
1906 | #endif | |
1907 | ||
1908 | #ifndef MAX_WCHAR_TYPE_SIZE | |
1909 | #define MAX_WCHAR_TYPE_SIZE WCHAR_TYPE_SIZE | |
1910 | #endif | |
1911 | ||
1912 | /* Interpret a (possibly wide) character constant in TOKEN. | |
1913 | WARN_MULTI warns about multi-character charconsts, if not | |
1914 | TRADITIONAL. TRADITIONAL also indicates not to interpret escapes | |
1915 | that did not exist in traditional C. PCHARS_SEEN points to a | |
1916 | variable that is filled in with the number of characters seen. */ | |
1917 | HOST_WIDE_INT | |
1918 | cpp_interpret_charconst (pfile, token, warn_multi, traditional, pchars_seen) | |
1919 | cpp_reader *pfile; | |
1920 | const cpp_token *token; | |
1921 | int warn_multi; | |
1922 | int traditional; | |
1923 | unsigned int *pchars_seen; | |
1924 | { | |
1925 | const unsigned char *str = token->val.str.text; | |
1926 | const unsigned char *limit = str + token->val.str.len; | |
1927 | unsigned int chars_seen = 0; | |
1928 | unsigned int width, max_chars, c; | |
0d086e18 | 1929 | unsigned HOST_WIDE_INT mask; |
1930 | HOST_WIDE_INT result = 0; | |
8330799c | 1931 | |
1932 | #ifdef MULTIBYTE_CHARS | |
1933 | (void) local_mbtowc (NULL, NULL, 0); | |
1934 | #endif | |
1935 | ||
1936 | /* Width in bits. */ | |
1937 | if (token->type == CPP_CHAR) | |
1938 | width = MAX_CHAR_TYPE_SIZE; | |
1939 | else | |
1940 | width = MAX_WCHAR_TYPE_SIZE; | |
1941 | ||
1942 | if (width < HOST_BITS_PER_WIDE_INT) | |
1943 | mask = ((unsigned HOST_WIDE_INT) 1 << width) - 1; | |
1944 | else | |
1945 | mask = ~0; | |
1946 | max_chars = HOST_BITS_PER_WIDE_INT / width; | |
1947 | ||
1948 | while (str < limit) | |
1949 | { | |
1950 | #ifdef MULTIBYTE_CHARS | |
1951 | wchar_t wc; | |
1952 | int char_len; | |
1953 | ||
1954 | char_len = local_mbtowc (&wc, str, limit - str); | |
1955 | if (char_len == -1) | |
1956 | { | |
1957 | cpp_warning (pfile, "ignoring invalid multibyte character"); | |
1958 | c = *str++; | |
1959 | } | |
1960 | else | |
1961 | { | |
1962 | str += char_len; | |
1963 | c = wc; | |
1964 | } | |
1965 | #else | |
1966 | c = *str++; | |
1967 | #endif | |
1968 | ||
1969 | if (c == '\\') | |
c8342759 | 1970 | c = cpp_parse_escape (pfile, &str, limit, mask, traditional); |
8330799c | 1971 | |
1972 | #ifdef MAP_CHARACTER | |
1973 | if (ISPRINT (c)) | |
1974 | c = MAP_CHARACTER (c); | |
1975 | #endif | |
1976 | ||
1977 | /* Merge character into result; ignore excess chars. */ | |
1978 | if (++chars_seen <= max_chars) | |
1979 | { | |
1980 | if (width < HOST_BITS_PER_WIDE_INT) | |
1981 | result = (result << width) | (c & mask); | |
1982 | else | |
1983 | result = c; | |
1984 | } | |
1985 | } | |
1986 | ||
1987 | if (chars_seen == 0) | |
1988 | cpp_error (pfile, "empty character constant"); | |
1989 | else if (chars_seen > max_chars) | |
1990 | { | |
1991 | chars_seen = max_chars; | |
f73bab03 | 1992 | cpp_warning (pfile, "character constant too long"); |
8330799c | 1993 | } |
1994 | else if (chars_seen > 1 && !traditional && warn_multi) | |
1995 | cpp_warning (pfile, "multi-character character constant"); | |
1996 | ||
1997 | /* If char type is signed, sign-extend the constant. The | |
1998 | __CHAR_UNSIGNED__ macro is set by the driver if appropriate. */ | |
1999 | if (token->type == CPP_CHAR && chars_seen) | |
2000 | { | |
2001 | unsigned int nbits = chars_seen * width; | |
2002 | unsigned int mask = (unsigned int) ~0 >> (HOST_BITS_PER_INT - nbits); | |
2003 | ||
2004 | if (pfile->spec_nodes.n__CHAR_UNSIGNED__->type == NT_MACRO | |
2005 | || ((result >> (nbits - 1)) & 1) == 0) | |
2006 | result &= mask; | |
2007 | else | |
2008 | result |= ~mask; | |
2009 | } | |
2010 | ||
2011 | *pchars_seen = chars_seen; | |
2012 | return result; | |
2013 | } | |
2014 | ||
084163dc | 2015 | /* Memory buffers. Changing these three constants can have a dramatic |
2016 | effect on performance. The values here are reasonable defaults, | |
2017 | but might be tuned. If you adjust them, be sure to test across a | |
2018 | range of uses of cpplib, including heavy nested function-like macro | |
2019 | expansion. Also check the change in peak memory usage (NJAMD is a | |
2020 | good tool for this). */ | |
2021 | #define MIN_BUFF_SIZE 8000 | |
2022 | #define BUFF_SIZE_UPPER_BOUND(MIN_SIZE) (8000 + (MIN_SIZE) * 3 / 2) | |
2023 | #define EXTENDED_BUFF_SIZE(BUFF, MIN_EXTRA) \ | |
2024 | (MIN_EXTRA + ((BUFF)->limit - (BUFF)->cur) * 2) | |
deb356cf | 2025 | |
79bd622b | 2026 | struct dummy |
deb356cf | 2027 | { |
79bd622b | 2028 | char c; |
2029 | union | |
2030 | { | |
2031 | double d; | |
2032 | int *p; | |
2033 | } u; | |
2034 | }; | |
deb356cf | 2035 | |
79bd622b | 2036 | #define DEFAULT_ALIGNMENT (offsetof (struct dummy, u)) |
06c92cbc | 2037 | #define CPP_ALIGN(size, align) (((size) + ((align) - 1)) & ~((align) - 1)) |
2038 | ||
1785b647 | 2039 | /* Create a new allocation buffer. Place the control block at the end |
2040 | of the buffer, so that buffer overflows will cause immediate chaos. */ | |
06c92cbc | 2041 | static _cpp_buff * |
2042 | new_buff (len) | |
2043 | unsigned int len; | |
2044 | { | |
2045 | _cpp_buff *result; | |
2046 | char *base; | |
2047 | ||
084163dc | 2048 | if (len < MIN_BUFF_SIZE) |
2049 | len = MIN_BUFF_SIZE; | |
06c92cbc | 2050 | len = CPP_ALIGN (len, DEFAULT_ALIGNMENT); |
2051 | ||
2052 | base = xmalloc (len + sizeof (_cpp_buff)); | |
2053 | result = (_cpp_buff *) (base + len); | |
2054 | result->base = base; | |
2055 | result->cur = base; | |
2056 | result->limit = base + len; | |
2057 | result->next = NULL; | |
2058 | return result; | |
2059 | } | |
2060 | ||
2061 | /* Place a chain of unwanted allocation buffers on the free list. */ | |
2062 | void | |
2063 | _cpp_release_buff (pfile, buff) | |
2064 | cpp_reader *pfile; | |
2065 | _cpp_buff *buff; | |
2066 | { | |
2067 | _cpp_buff *end = buff; | |
2068 | ||
2069 | while (end->next) | |
2070 | end = end->next; | |
2071 | end->next = pfile->free_buffs; | |
2072 | pfile->free_buffs = buff; | |
2073 | } | |
2074 | ||
2075 | /* Return a free buffer of size at least MIN_SIZE. */ | |
2076 | _cpp_buff * | |
2077 | _cpp_get_buff (pfile, min_size) | |
2078 | cpp_reader *pfile; | |
2079 | unsigned int min_size; | |
2080 | { | |
2081 | _cpp_buff *result, **p; | |
2082 | ||
2083 | for (p = &pfile->free_buffs;; p = &(*p)->next) | |
2084 | { | |
084163dc | 2085 | unsigned int size; |
2086 | ||
2087 | if (*p == NULL) | |
06c92cbc | 2088 | return new_buff (min_size); |
084163dc | 2089 | result = *p; |
2090 | size = result->limit - result->base; | |
2091 | /* Return a buffer that's big enough, but don't waste one that's | |
2092 | way too big. */ | |
2093 | if (size >= min_size && size < BUFF_SIZE_UPPER_BOUND (min_size)) | |
06c92cbc | 2094 | break; |
2095 | } | |
2096 | ||
2097 | *p = result->next; | |
2098 | result->next = NULL; | |
2099 | result->cur = result->base; | |
2100 | return result; | |
2101 | } | |
2102 | ||
2103 | /* Return a buffer chained on the end of BUFF. Copy to it the | |
2104 | uncommitted remaining bytes of BUFF, with at least MIN_EXTRA more | |
2105 | bytes. */ | |
2106 | _cpp_buff * | |
2107 | _cpp_extend_buff (pfile, buff, min_extra) | |
2108 | cpp_reader *pfile; | |
2109 | _cpp_buff *buff; | |
2110 | unsigned int min_extra; | |
2111 | { | |
084163dc | 2112 | unsigned int size = EXTENDED_BUFF_SIZE (buff, min_extra); |
06c92cbc | 2113 | |
2114 | buff->next = _cpp_get_buff (pfile, size); | |
2115 | memcpy (buff->next->base, buff->cur, buff->limit - buff->cur); | |
2116 | return buff->next; | |
2117 | } | |
2118 | ||
2119 | /* Free a chain of buffers starting at BUFF. */ | |
2120 | void | |
2121 | _cpp_free_buff (buff) | |
2122 | _cpp_buff *buff; | |
2123 | { | |
2124 | _cpp_buff *next; | |
2125 | ||
2126 | for (; buff; buff = next) | |
2127 | { | |
2128 | next = buff->next; | |
2129 | free (buff->base); | |
2130 | } | |
2131 | } | |
deb356cf | 2132 | |
79bd622b | 2133 | static int |
084163dc | 2134 | chunk_suitable (chunk, size) |
79bd622b | 2135 | cpp_chunk *chunk; |
2136 | unsigned int size; | |
2137 | { | |
2138 | /* Being at least twice SIZE means we can use memcpy in | |
2139 | _cpp_next_chunk rather than memmove. Besides, it's a good idea | |
2140 | anyway. */ | |
084163dc | 2141 | return (chunk && (unsigned int) (chunk->limit - chunk->base) >= size * 2); |
f80e83a9 | 2142 | } |
6060326b | 2143 | |
79bd622b | 2144 | /* Returns the end of the new pool. PTR points to a char in the old |
2145 | pool, and is updated to point to the same char in the new pool. */ | |
2146 | unsigned char * | |
2147 | _cpp_next_chunk (pool, len, ptr) | |
2148 | cpp_pool *pool; | |
2149 | unsigned int len; | |
2150 | unsigned char **ptr; | |
f80e83a9 | 2151 | { |
79bd622b | 2152 | cpp_chunk *chunk = pool->cur->next; |
6060326b | 2153 | |
79bd622b | 2154 | /* LEN is the minimum size we want in the new pool. */ |
2155 | len += POOL_ROOM (pool); | |
084163dc | 2156 | if (! chunk_suitable (chunk, len)) |
f80e83a9 | 2157 | { |
79bd622b | 2158 | chunk = new_chunk (POOL_SIZE (pool) * 2 + len); |
6060326b | 2159 | |
79bd622b | 2160 | chunk->next = pool->cur->next; |
2161 | pool->cur->next = chunk; | |
6060326b | 2162 | } |
2163 | ||
79bd622b | 2164 | /* Update the pointer before changing chunk's front. */ |
2165 | if (ptr) | |
2166 | *ptr += chunk->base - POOL_FRONT (pool); | |
f80e83a9 | 2167 | |
79bd622b | 2168 | memcpy (chunk->base, POOL_FRONT (pool), POOL_ROOM (pool)); |
2169 | chunk->front = chunk->base; | |
f80e83a9 | 2170 | |
79bd622b | 2171 | pool->cur = chunk; |
2172 | return POOL_LIMIT (pool); | |
6060326b | 2173 | } |
2174 | ||
79bd622b | 2175 | static cpp_chunk * |
2176 | new_chunk (size) | |
2177 | unsigned int size; | |
f80e83a9 | 2178 | { |
79bd622b | 2179 | unsigned char *base; |
2180 | cpp_chunk *result; | |
89b05ef6 | 2181 | |
a28b091b | 2182 | size = POOL_ALIGN (size, DEFAULT_ALIGNMENT); |
79bd622b | 2183 | base = (unsigned char *) xmalloc (size + sizeof (cpp_chunk)); |
2184 | /* Put the chunk descriptor at the end. Then chunk overruns will | |
2185 | cause obvious chaos. */ | |
2186 | result = (cpp_chunk *) (base + size); | |
2187 | result->base = base; | |
2188 | result->front = base; | |
2189 | result->limit = base + size; | |
2190 | result->next = 0; | |
deb356cf | 2191 | |
79bd622b | 2192 | return result; |
f80e83a9 | 2193 | } |
2194 | ||
79bd622b | 2195 | void |
2196 | _cpp_init_pool (pool, size, align, temp) | |
2197 | cpp_pool *pool; | |
2198 | unsigned int size, align, temp; | |
2199 | { | |
2200 | if (align == 0) | |
2201 | align = DEFAULT_ALIGNMENT; | |
2202 | if (align & (align - 1)) | |
2203 | abort (); | |
2204 | pool->align = align; | |
162cee98 | 2205 | pool->first = new_chunk (size); |
2206 | pool->cur = pool->first; | |
79bd622b | 2207 | if (temp) |
2208 | pool->cur->next = pool->cur; | |
f80e83a9 | 2209 | } |
2210 | ||
79bd622b | 2211 | void |
2212 | _cpp_free_pool (pool) | |
2213 | cpp_pool *pool; | |
89b05ef6 | 2214 | { |
162cee98 | 2215 | cpp_chunk *chunk = pool->first, *next; |
89b05ef6 | 2216 | |
79bd622b | 2217 | do |
89b05ef6 | 2218 | { |
79bd622b | 2219 | next = chunk->next; |
2220 | free (chunk->base); | |
2221 | chunk = next; | |
89b05ef6 | 2222 | } |
162cee98 | 2223 | while (chunk && chunk != pool->first); |
f80e83a9 | 2224 | } |
f80e83a9 | 2225 | |
79bd622b | 2226 | /* Reserve LEN bytes from a memory pool. */ |
2227 | unsigned char * | |
2228 | _cpp_pool_reserve (pool, len) | |
2229 | cpp_pool *pool; | |
2230 | unsigned int len; | |
f80e83a9 | 2231 | { |
a28b091b | 2232 | len = POOL_ALIGN (len, pool->align); |
79bd622b | 2233 | if (len > (unsigned int) POOL_ROOM (pool)) |
2234 | _cpp_next_chunk (pool, len, 0); | |
f80e83a9 | 2235 | |
79bd622b | 2236 | return POOL_FRONT (pool); |
6060326b | 2237 | } |
2238 | ||
79bd622b | 2239 | /* Allocate LEN bytes from a memory pool. */ |
2240 | unsigned char * | |
2241 | _cpp_pool_alloc (pool, len) | |
2242 | cpp_pool *pool; | |
2243 | unsigned int len; | |
f80e83a9 | 2244 | { |
79bd622b | 2245 | unsigned char *result = _cpp_pool_reserve (pool, len); |
deb356cf | 2246 | |
79bd622b | 2247 | POOL_COMMIT (pool, len); |
2248 | return result; | |
f80e83a9 | 2249 | } |