]>
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 | ||
f80e83a9 | 23 | /* |
24 | ||
25 | Cleanups to do:- | |
26 | ||
f80e83a9 | 27 | o Distinguish integers, floats, and 'other' pp-numbers. |
28 | o Store ints and char constants as binary values. | |
29 | o New command-line assertion syntax. | |
f80e83a9 | 30 | o Comment all functions, and describe macro expansion algorithm. |
31 | o Move as much out of header files as possible. | |
32 | o Remove single quote pairs `', and some '', from diagnostics. | |
33 | o Correct pastability test for CPP_NAME and CPP_NUMBER. | |
34 | ||
35 | */ | |
36 | ||
0578f103 | 37 | #include "config.h" |
38 | #include "system.h" | |
39 | #include "intl.h" | |
40 | #include "cpplib.h" | |
41 | #include "cpphash.h" | |
f80e83a9 | 42 | #include "symcat.h" |
0578f103 | 43 | |
524f0c40 | 44 | const unsigned char *_cpp_digraph_spellings [] = {U"%:", U"%:%:", U"<:", |
45 | U":>", U"<%", U"%>"}; | |
338fa5f7 | 46 | static const cpp_token placemarker_token = {0, 0, CPP_PLACEMARKER, |
47 | 0 UNION_INIT_ZERO}; | |
241e762e | 48 | static const cpp_token eof_token = {0, 0, CPP_EOF, 0 UNION_INIT_ZERO}; |
49 | ||
50 | /* Flags for cpp_context. */ | |
51 | #define CONTEXT_PASTEL (1 << 0) /* An argument context on LHS of ##. */ | |
52 | #define CONTEXT_PASTER (1 << 1) /* An argument context on RHS of ##. */ | |
53 | #define CONTEXT_RAW (1 << 2) /* If argument tokens already expanded. */ | |
54 | #define CONTEXT_ARG (1 << 3) /* If an argument context. */ | |
55 | ||
56 | typedef struct cpp_context cpp_context; | |
57 | struct cpp_context | |
58 | { | |
59 | union | |
60 | { | |
61 | const cpp_toklist *list; /* Used for macro contexts only. */ | |
62 | const cpp_token **arg; /* Used for arg contexts only. */ | |
63 | } u; | |
64 | ||
65 | /* Pushed token to be returned by next call to get_raw_token. */ | |
66 | const cpp_token *pushed_token; | |
67 | ||
fdf3a98f | 68 | struct macro_args *args; /* The arguments for a function-like |
69 | macro. NULL otherwise. */ | |
241e762e | 70 | unsigned short posn; /* Current posn, index into u. */ |
71 | unsigned short count; /* No. of tokens in u. */ | |
72 | unsigned short level; | |
73 | unsigned char flags; | |
74 | }; | |
75 | ||
76 | typedef struct macro_args macro_args; | |
77 | struct macro_args | |
78 | { | |
79 | unsigned int *ends; | |
80 | const cpp_token **tokens; | |
81 | unsigned int capacity; | |
82 | unsigned int used; | |
83 | unsigned short level; | |
84 | }; | |
85 | ||
86 | static const cpp_token *get_raw_token PARAMS ((cpp_reader *)); | |
87 | static const cpp_token *parse_arg PARAMS ((cpp_reader *, int, unsigned int, | |
88 | macro_args *, unsigned int *)); | |
89 | static int parse_args PARAMS ((cpp_reader *, cpp_hashnode *, macro_args *)); | |
90 | static void save_token PARAMS ((macro_args *, const cpp_token *)); | |
91 | static int pop_context PARAMS ((cpp_reader *)); | |
92 | static int push_macro_context PARAMS ((cpp_reader *, const cpp_token *)); | |
93 | static void push_arg_context PARAMS ((cpp_reader *, const cpp_token *)); | |
94 | static void free_macro_args PARAMS ((macro_args *)); | |
e2f9a79f | 95 | |
338fa5f7 | 96 | static cppchar_t handle_newline PARAMS ((cpp_buffer *, cppchar_t)); |
97 | static cppchar_t skip_escaped_newlines PARAMS ((cpp_buffer *, cppchar_t)); | |
98 | static cppchar_t get_effective_char PARAMS ((cpp_buffer *)); | |
99 | ||
f80e83a9 | 100 | static int skip_block_comment PARAMS ((cpp_reader *)); |
f669338a | 101 | static int skip_line_comment PARAMS ((cpp_reader *)); |
338fa5f7 | 102 | static void adjust_column PARAMS ((cpp_reader *)); |
103 | static void skip_whitespace PARAMS ((cpp_reader *, cppchar_t)); | |
104 | static cpp_hashnode *parse_identifier PARAMS ((cpp_reader *, cppchar_t)); | |
105 | static void parse_number PARAMS ((cpp_reader *, cpp_string *, cppchar_t)); | |
106 | static void parse_string PARAMS ((cpp_reader *, cpp_token *, cppchar_t)); | |
107 | static void unterminated PARAMS ((cpp_reader *, unsigned int, int)); | |
108 | static int trigraph_ok PARAMS ((cpp_reader *, cppchar_t)); | |
109 | static void save_comment PARAMS ((cpp_reader *, cpp_token *, const U_CHAR *)); | |
f669338a | 110 | static void lex_percent PARAMS ((cpp_buffer *, cpp_token *)); |
111 | static void lex_dot PARAMS ((cpp_reader *, cpp_token *)); | |
f80e83a9 | 112 | static void lex_line PARAMS ((cpp_reader *, cpp_toklist *)); |
338fa5f7 | 113 | static void lex_token PARAMS ((cpp_reader *, cpp_token *)); |
f80e83a9 | 114 | static int lex_next PARAMS ((cpp_reader *, int)); |
338fa5f7 | 115 | |
f80e83a9 | 116 | static int is_macro_disabled PARAMS ((cpp_reader *, const cpp_toklist *, |
117 | const cpp_token *)); | |
6bc03ce3 | 118 | |
f80e83a9 | 119 | static cpp_token *stringify_arg PARAMS ((cpp_reader *, const cpp_token *)); |
120 | static void expand_context_stack PARAMS ((cpp_reader *)); | |
d2efb5ed | 121 | static unsigned char * spell_token PARAMS ((cpp_reader *, const cpp_token *, |
f80e83a9 | 122 | unsigned char *)); |
6bc03ce3 | 123 | typedef unsigned int (* speller) PARAMS ((unsigned char *, cpp_toklist *, |
124 | cpp_token *)); | |
f80e83a9 | 125 | static cpp_token *make_string_token PARAMS ((cpp_token *, const U_CHAR *, |
126 | unsigned int)); | |
127 | static cpp_token *alloc_number_token PARAMS ((cpp_reader *, int number)); | |
128 | static const cpp_token *special_symbol PARAMS ((cpp_reader *, cpp_hashnode *, | |
129 | const cpp_token *)); | |
130 | static cpp_token *duplicate_token PARAMS ((cpp_reader *, const cpp_token *)); | |
131 | static const cpp_token *maybe_paste_with_next PARAMS ((cpp_reader *, | |
132 | const cpp_token *)); | |
f80e83a9 | 133 | static unsigned int prevent_macro_expansion PARAMS ((cpp_reader *)); |
134 | static void restore_macro_expansion PARAMS ((cpp_reader *, unsigned int)); | |
135 | static cpp_token *get_temp_token PARAMS ((cpp_reader *)); | |
136 | static void release_temp_tokens PARAMS ((cpp_reader *)); | |
137 | static U_CHAR * quote_string PARAMS ((U_CHAR *, const U_CHAR *, unsigned int)); | |
6bc03ce3 | 138 | |
f80e83a9 | 139 | #define VALID_SIGN(c, prevc) \ |
140 | (((c) == '+' || (c) == '-') && \ | |
141 | ((prevc) == 'e' || (prevc) == 'E' \ | |
142 | || (((prevc) == 'p' || (prevc) == 'P') && !CPP_OPTION (pfile, c89)))) | |
143 | ||
e916a356 | 144 | /* An upper bound on the number of bytes needed to spell a token, |
145 | including preceding whitespace. */ | |
6cae2504 | 146 | static inline size_t TOKEN_LEN PARAMS ((const cpp_token *)); |
147 | static inline size_t | |
148 | TOKEN_LEN (token) | |
149 | const cpp_token *token; | |
150 | { | |
151 | size_t len; | |
152 | ||
153 | switch (TOKEN_SPELL (token)) | |
154 | { | |
155 | default: len = 0; break; | |
156 | case SPELL_STRING: len = token->val.str.len; break; | |
157 | case SPELL_IDENT: len = token->val.node->length; break; | |
158 | } | |
159 | return len + 5; | |
160 | } | |
e916a356 | 161 | |
241e762e | 162 | #define IS_ARG_CONTEXT(c) ((c)->flags & CONTEXT_ARG) |
163 | #define CURRENT_CONTEXT(pfile) ((pfile)->contexts + (pfile)->cur_context) | |
e0a859f1 | 164 | #define ON_REST_ARG(c) \ |
58fe658a | 165 | (((c)->u.list->flags & VAR_ARGS) \ |
166 | && (c)->u.list->tokens[(c)->posn - 1].val.aux \ | |
167 | == (unsigned int) ((c)->u.list->paramc - 1)) | |
241e762e | 168 | |
169 | #define ASSIGN_FLAGS_AND_POS(d, s) \ | |
170 | do {(d)->flags = (s)->flags & (PREV_WHITE | BOL | PASTE_LEFT); \ | |
171 | if ((d)->flags & BOL) {(d)->col = (s)->col; (d)->line = (s)->line;} \ | |
172 | } while (0) | |
173 | ||
174 | /* f is flags, just consisting of PREV_WHITE | BOL. */ | |
175 | #define MODIFY_FLAGS_AND_POS(d, s, f) \ | |
176 | do {(d)->flags &= ~(PREV_WHITE | BOL); (d)->flags |= (f); \ | |
177 | if ((f) & BOL) {(d)->col = (s)->col; (d)->line = (s)->line;} \ | |
178 | } while (0) | |
179 | ||
7e842f95 | 180 | #define OP(e, s) { SPELL_OPERATOR, U s }, |
181 | #define TK(e, s) { s, U STRINGX (e) }, | |
6bc03ce3 | 182 | |
f80e83a9 | 183 | const struct token_spelling |
7e842f95 | 184 | _cpp_token_spellings [N_TTYPES] = {TTYPE_TABLE }; |
f80e83a9 | 185 | |
7e842f95 | 186 | #undef OP |
187 | #undef TK | |
6bc03ce3 | 188 | |
f80e83a9 | 189 | /* Helper routine used by parse_include, which can't see spell_token. |
190 | Reinterpret the current line as an h-char-sequence (< ... >); we are | |
191 | looking at the first token after the <. */ | |
192 | const cpp_token * | |
193 | _cpp_glue_header_name (pfile) | |
0578f103 | 194 | cpp_reader *pfile; |
195 | { | |
f80e83a9 | 196 | const cpp_token *t; |
197 | cpp_token *hdr; | |
6cae2504 | 198 | U_CHAR *buf, *p; |
199 | size_t len, avail; | |
200 | ||
201 | avail = 40; | |
202 | len = 0; | |
203 | buf = xmalloc (avail); | |
f80e83a9 | 204 | |
205 | for (;;) | |
206 | { | |
deb356cf | 207 | t = _cpp_get_token (pfile); |
f80e83a9 | 208 | if (t->type == CPP_GREATER || t->type == CPP_EOF) |
209 | break; | |
210 | ||
6cae2504 | 211 | if (len + TOKEN_LEN (t) > avail) |
212 | { | |
213 | avail = len + TOKEN_LEN (t) + 40; | |
214 | buf = xrealloc (buf, avail); | |
215 | } | |
216 | ||
f80e83a9 | 217 | if (t->flags & PREV_WHITE) |
6cae2504 | 218 | buf[len++] = ' '; |
219 | ||
220 | p = spell_token (pfile, t, buf + len); | |
221 | len = (size_t) (p - buf); /* p known >= buf */ | |
f80e83a9 | 222 | } |
223 | ||
224 | if (t->type == CPP_EOF) | |
225 | cpp_error (pfile, "missing terminating > character"); | |
0578f103 | 226 | |
6cae2504 | 227 | buf = xrealloc (buf, len); |
f80e83a9 | 228 | |
229 | hdr = get_temp_token (pfile); | |
230 | hdr->type = CPP_HEADER_NAME; | |
231 | hdr->flags = 0; | |
76faa4c0 | 232 | hdr->val.str.text = buf; |
233 | hdr->val.str.len = len; | |
f80e83a9 | 234 | return hdr; |
0578f103 | 235 | } |
236 | ||
50fd6b48 | 237 | /* Token-buffer helper functions. */ |
238 | ||
d2efb5ed | 239 | /* Expand a token list's string space. It is *vital* that |
240 | list->tokens_used is correct, to get pointer fix-up right. */ | |
f80e83a9 | 241 | void |
242 | _cpp_expand_name_space (list, len) | |
50fd6b48 | 243 | cpp_toklist *list; |
6060326b | 244 | unsigned int len; |
245 | { | |
e916a356 | 246 | const U_CHAR *old_namebuf; |
e916a356 | 247 | |
248 | old_namebuf = list->namebuf; | |
6060326b | 249 | list->name_cap += len; |
250 | list->namebuf = (unsigned char *) xrealloc (list->namebuf, list->name_cap); | |
e916a356 | 251 | |
252 | /* Fix up token text pointers. */ | |
6dafd80a | 253 | if (list->namebuf != old_namebuf) |
e916a356 | 254 | { |
255 | unsigned int i; | |
256 | ||
257 | for (i = 0; i < list->tokens_used; i++) | |
7e842f95 | 258 | if (TOKEN_SPELL (&list->tokens[i]) == SPELL_STRING) |
76faa4c0 | 259 | list->tokens[i].val.str.text += (list->namebuf - old_namebuf); |
e916a356 | 260 | } |
50fd6b48 | 261 | } |
262 | ||
f80e83a9 | 263 | /* If there is not enough room for LEN more characters, expand the |
264 | list by just enough to have room for LEN characters. */ | |
265 | void | |
266 | _cpp_reserve_name_space (list, len) | |
267 | cpp_toklist *list; | |
268 | unsigned int len; | |
269 | { | |
270 | unsigned int room = list->name_cap - list->name_used; | |
271 | ||
272 | if (room < len) | |
273 | _cpp_expand_name_space (list, len - room); | |
274 | } | |
275 | ||
50fd6b48 | 276 | /* Expand the number of tokens in a list. */ |
d2efb5ed | 277 | void |
278 | _cpp_expand_token_space (list, count) | |
50fd6b48 | 279 | cpp_toklist *list; |
d2efb5ed | 280 | unsigned int count; |
50fd6b48 | 281 | { |
d2efb5ed | 282 | list->tokens_cap += count; |
50fd6b48 | 283 | list->tokens = (cpp_token *) |
f669338a | 284 | xrealloc (list->tokens, list->tokens_cap * sizeof (cpp_token)); |
50fd6b48 | 285 | } |
286 | ||
f669338a | 287 | /* Initialize a token list. If EMPTY is false, some token and name |
288 | space is provided. */ | |
bce8e0c0 | 289 | void |
f669338a | 290 | _cpp_init_toklist (list, empty) |
50fd6b48 | 291 | cpp_toklist *list; |
f669338a | 292 | int empty; |
50fd6b48 | 293 | { |
f669338a | 294 | if (empty) |
d2efb5ed | 295 | { |
296 | list->tokens_cap = 0; | |
f80e83a9 | 297 | list->tokens = 0; |
d2efb5ed | 298 | list->name_cap = 0; |
f80e83a9 | 299 | list->namebuf = 0; |
d2efb5ed | 300 | } |
301 | else | |
302 | { | |
f669338a | 303 | /* Initialize token space. */ |
d2efb5ed | 304 | list->tokens_cap = 256; /* 4K's worth. */ |
305 | list->tokens = (cpp_token *) | |
306 | xmalloc ((list->tokens_cap + 1) * sizeof (cpp_token)); | |
d2efb5ed | 307 | |
308 | /* Initialize name space. */ | |
309 | list->name_cap = 1024; | |
f80e83a9 | 310 | list->namebuf = (unsigned char *) xmalloc (list->name_cap); |
d2efb5ed | 311 | } |
bce8e0c0 | 312 | |
bce8e0c0 | 313 | _cpp_clear_toklist (list); |
314 | } | |
50fd6b48 | 315 | |
bce8e0c0 | 316 | /* Clear a token list. */ |
317 | void | |
318 | _cpp_clear_toklist (list) | |
319 | cpp_toklist *list; | |
320 | { | |
6060326b | 321 | list->tokens_used = 0; |
322 | list->name_used = 0; | |
f80e83a9 | 323 | list->directive = 0; |
324 | list->paramc = 0; | |
325 | list->params_len = 0; | |
f669338a | 326 | list->flags = 0; |
bce8e0c0 | 327 | } |
328 | ||
329 | /* Free a token list. Does not free the list itself, which may be | |
330 | embedded in a larger structure. */ | |
331 | void | |
332 | _cpp_free_toklist (list) | |
f80e83a9 | 333 | const cpp_toklist *list; |
bce8e0c0 | 334 | { |
f669338a | 335 | free (list->tokens); |
bce8e0c0 | 336 | free (list->namebuf); |
50fd6b48 | 337 | } |
338 | ||
bce8e0c0 | 339 | /* Compare two tokens. */ |
340 | int | |
341 | _cpp_equiv_tokens (a, b) | |
342 | const cpp_token *a, *b; | |
343 | { | |
f80e83a9 | 344 | if (a->type == b->type && a->flags == b->flags) |
7e842f95 | 345 | switch (TOKEN_SPELL (a)) |
f80e83a9 | 346 | { |
347 | default: /* Keep compiler happy. */ | |
348 | case SPELL_OPERATOR: | |
349 | return 1; | |
350 | case SPELL_CHAR: | |
351 | case SPELL_NONE: | |
352 | return a->val.aux == b->val.aux; /* arg_no or character. */ | |
353 | case SPELL_IDENT: | |
76faa4c0 | 354 | return a->val.node == b->val.node; |
f80e83a9 | 355 | case SPELL_STRING: |
76faa4c0 | 356 | return (a->val.str.len == b->val.str.len |
357 | && !memcmp (a->val.str.text, b->val.str.text, | |
358 | a->val.str.len)); | |
f80e83a9 | 359 | } |
bce8e0c0 | 360 | |
f80e83a9 | 361 | return 0; |
bce8e0c0 | 362 | } |
363 | ||
364 | /* Compare two token lists. */ | |
365 | int | |
366 | _cpp_equiv_toklists (a, b) | |
367 | const cpp_toklist *a, *b; | |
368 | { | |
369 | unsigned int i; | |
370 | ||
f80e83a9 | 371 | if (a->tokens_used != b->tokens_used |
372 | || a->flags != b->flags | |
373 | || a->paramc != b->paramc) | |
bce8e0c0 | 374 | return 0; |
375 | ||
376 | for (i = 0; i < a->tokens_used; i++) | |
377 | if (! _cpp_equiv_tokens (&a->tokens[i], &b->tokens[i])) | |
378 | return 0; | |
379 | return 1; | |
380 | } | |
381 | ||
f80e83a9 | 382 | /* Utility routine: |
2c63d6c8 | 383 | |
76faa4c0 | 384 | Compares, the token TOKEN to the NUL-terminated string STRING. |
385 | TOKEN must be a CPP_NAME. Returns 1 for equal, 0 for unequal. */ | |
bce8e0c0 | 386 | |
f80e83a9 | 387 | int |
76faa4c0 | 388 | cpp_ideq (token, string) |
389 | const cpp_token *token; | |
f80e83a9 | 390 | const char *string; |
391 | { | |
76faa4c0 | 392 | if (token->type != CPP_NAME) |
f80e83a9 | 393 | return 0; |
76faa4c0 | 394 | |
395 | return !ustrcmp (token->val.node->name, (const U_CHAR *)string); | |
bce8e0c0 | 396 | } |
50fd6b48 | 397 | |
338fa5f7 | 398 | /* Call when meeting a newline. Returns the character after the newline |
399 | (or carriage-return newline combination), or EOF. */ | |
400 | static cppchar_t | |
401 | handle_newline (buffer, newline_char) | |
402 | cpp_buffer *buffer; | |
403 | cppchar_t newline_char; | |
404 | { | |
405 | cppchar_t next = EOF; | |
406 | ||
407 | buffer->col_adjust = 0; | |
408 | buffer->lineno++; | |
409 | buffer->line_base = buffer->cur; | |
410 | ||
411 | /* Handle CR-LF and LF-CR combinations, get the next character. */ | |
412 | if (buffer->cur < buffer->rlimit) | |
413 | { | |
414 | next = *buffer->cur++; | |
415 | if (next + newline_char == '\r' + '\n') | |
416 | { | |
417 | buffer->line_base = buffer->cur; | |
418 | if (buffer->cur < buffer->rlimit) | |
419 | next = *buffer->cur++; | |
420 | else | |
421 | next = EOF; | |
422 | } | |
423 | } | |
424 | ||
425 | buffer->read_ahead = next; | |
426 | return next; | |
427 | } | |
428 | ||
429 | /* Subroutine of skip_escaped_newlines; called when a trigraph is | |
430 | encountered. It warns if necessary, and returns true if the | |
431 | trigraph should be honoured. FROM_CHAR is the third character of a | |
432 | trigraph, and presumed to be the previous character for position | |
433 | reporting. */ | |
0578f103 | 434 | static int |
338fa5f7 | 435 | trigraph_ok (pfile, from_char) |
0578f103 | 436 | cpp_reader *pfile; |
338fa5f7 | 437 | cppchar_t from_char; |
0578f103 | 438 | { |
f80e83a9 | 439 | int accept = CPP_OPTION (pfile, trigraphs); |
440 | ||
f669338a | 441 | /* Don't warn about trigraphs in comments. */ |
442 | if (CPP_OPTION (pfile, warn_trigraphs) && !pfile->state.lexing_comment) | |
0578f103 | 443 | { |
338fa5f7 | 444 | cpp_buffer *buffer = pfile->buffer; |
f80e83a9 | 445 | if (accept) |
338fa5f7 | 446 | cpp_warning_with_line (pfile, buffer->lineno, CPP_BUF_COL (buffer) - 2, |
f80e83a9 | 447 | "trigraph ??%c converted to %c", |
338fa5f7 | 448 | (int) from_char, |
449 | (int) _cpp_trigraph_map[from_char]); | |
0578f103 | 450 | else |
338fa5f7 | 451 | cpp_warning_with_line (pfile, buffer->lineno, CPP_BUF_COL (buffer) - 2, |
452 | "trigraph ??%c ignored", (int) from_char); | |
0578f103 | 453 | } |
338fa5f7 | 454 | |
f80e83a9 | 455 | return accept; |
0578f103 | 456 | } |
457 | ||
338fa5f7 | 458 | /* Assumes local variables buffer and result. */ |
459 | #define ACCEPT_CHAR(t) \ | |
460 | do { result->type = t; buffer->read_ahead = EOF; } while (0) | |
461 | ||
462 | /* When we move to multibyte character sets, add to these something | |
463 | that saves and restores the state of the multibyte conversion | |
464 | library. This probably involves saving and restoring a "cookie". | |
465 | In the case of glibc it is an 8-byte structure, so is not a high | |
466 | overhead operation. In any case, it's out of the fast path. */ | |
467 | #define SAVE_STATE() do { saved_cur = buffer->cur; } while (0) | |
468 | #define RESTORE_STATE() do { buffer->cur = saved_cur; } while (0) | |
469 | ||
470 | /* Skips any escaped newlines introduced by NEXT, which is either a | |
471 | '?' or a '\\'. Returns the next character, which will also have | |
472 | been placed in buffer->read_ahead. */ | |
473 | static cppchar_t | |
474 | skip_escaped_newlines (buffer, next) | |
475 | cpp_buffer *buffer; | |
476 | cppchar_t next; | |
0578f103 | 477 | { |
338fa5f7 | 478 | cppchar_t next1; |
479 | const unsigned char *saved_cur; | |
480 | int space; | |
f80e83a9 | 481 | |
338fa5f7 | 482 | do |
f80e83a9 | 483 | { |
338fa5f7 | 484 | if (buffer->cur == buffer->rlimit) |
485 | break; | |
486 | ||
487 | SAVE_STATE (); | |
488 | if (next == '?') | |
489 | { | |
490 | next1 = *buffer->cur++; | |
491 | if (next1 != '?' || buffer->cur == buffer->rlimit) | |
492 | { | |
493 | RESTORE_STATE (); | |
494 | break; | |
495 | } | |
f80e83a9 | 496 | |
338fa5f7 | 497 | next1 = *buffer->cur++; |
498 | if (!_cpp_trigraph_map[next1] || !trigraph_ok (buffer->pfile, next1)) | |
499 | { | |
500 | RESTORE_STATE (); | |
501 | break; | |
502 | } | |
0578f103 | 503 | |
338fa5f7 | 504 | /* We have a full trigraph here. */ |
505 | next = _cpp_trigraph_map[next1]; | |
506 | if (next != '\\' || buffer->cur == buffer->rlimit) | |
507 | break; | |
508 | SAVE_STATE (); | |
509 | } | |
510 | ||
511 | /* We have a backslash, and room for at least one more character. */ | |
512 | space = 0; | |
513 | do | |
514 | { | |
515 | next1 = *buffer->cur++; | |
516 | if (!is_nvspace (next1)) | |
517 | break; | |
518 | space = 1; | |
519 | } | |
520 | while (buffer->cur < buffer->rlimit); | |
521 | ||
522 | if (!is_vspace (next1)) | |
523 | { | |
524 | RESTORE_STATE (); | |
525 | break; | |
526 | } | |
0578f103 | 527 | |
338fa5f7 | 528 | if (space) |
529 | cpp_warning (buffer->pfile, | |
530 | "backslash and newline separated by space"); | |
531 | ||
532 | next = handle_newline (buffer, next1); | |
533 | if (next == EOF) | |
534 | cpp_pedwarn (buffer->pfile, "backslash-newline at end of file"); | |
f80e83a9 | 535 | } |
338fa5f7 | 536 | while (next == '\\' || next == '?'); |
0578f103 | 537 | |
338fa5f7 | 538 | buffer->read_ahead = next; |
539 | return next; | |
0578f103 | 540 | } |
541 | ||
338fa5f7 | 542 | /* Obtain the next character, after trigraph conversion and skipping |
543 | an arbitrary string of escaped newlines. The common case of no | |
544 | trigraphs or escaped newlines falls through quickly. */ | |
545 | static cppchar_t | |
546 | get_effective_char (buffer) | |
547 | cpp_buffer *buffer; | |
852d1b04 | 548 | { |
338fa5f7 | 549 | cppchar_t next = EOF; |
550 | ||
551 | if (buffer->cur < buffer->rlimit) | |
552 | { | |
553 | next = *buffer->cur++; | |
554 | ||
555 | /* '?' can introduce trigraphs (and therefore backslash); '\\' | |
556 | can introduce escaped newlines, which we want to skip, or | |
557 | UCNs, which, depending upon lexer state, we will handle in | |
558 | the future. */ | |
559 | if (next == '?' || next == '\\') | |
560 | next = skip_escaped_newlines (buffer, next); | |
561 | } | |
562 | ||
563 | buffer->read_ahead = next; | |
564 | return next; | |
852d1b04 | 565 | } |
566 | ||
338fa5f7 | 567 | /* Skip a C-style block comment. We find the end of the comment by |
568 | seeing if an asterisk is before every '/' we encounter. Returns | |
569 | non-zero if comment terminated by EOF, zero otherwise. */ | |
f80e83a9 | 570 | static int |
571 | skip_block_comment (pfile) | |
0578f103 | 572 | cpp_reader *pfile; |
573 | { | |
f80e83a9 | 574 | cpp_buffer *buffer = pfile->buffer; |
338fa5f7 | 575 | cppchar_t c = EOF, prevc; |
576 | ||
f669338a | 577 | pfile->state.lexing_comment = 1; |
338fa5f7 | 578 | while (buffer->cur != buffer->rlimit) |
0578f103 | 579 | { |
338fa5f7 | 580 | prevc = c, c = *buffer->cur++; |
581 | ||
582 | next_char: | |
583 | /* FIXME: For speed, create a new character class of characters | |
584 | of no interest inside block comments. */ | |
585 | if (c == '?' || c == '\\') | |
586 | c = skip_escaped_newlines (buffer, c); | |
f80e83a9 | 587 | |
338fa5f7 | 588 | /* People like decorating comments with '*', so check for '/' |
589 | instead for efficiency. */ | |
f80e83a9 | 590 | if (c == '/') |
0578f103 | 591 | { |
338fa5f7 | 592 | if (prevc == '*') |
593 | break; | |
f80e83a9 | 594 | |
338fa5f7 | 595 | /* Warn about potential nested comments, but not if the '/' |
596 | comes immediately before the true comment delimeter. | |
f80e83a9 | 597 | Don't bother to get it right across escaped newlines. */ |
338fa5f7 | 598 | if (CPP_OPTION (pfile, warn_comments) |
599 | && buffer->cur != buffer->rlimit) | |
0578f103 | 600 | { |
338fa5f7 | 601 | prevc = c, c = *buffer->cur++; |
602 | if (c == '*' && buffer->cur != buffer->rlimit) | |
603 | { | |
604 | prevc = c, c = *buffer->cur++; | |
605 | if (c != '/') | |
606 | cpp_warning_with_line (pfile, CPP_BUF_LINE (buffer), | |
607 | CPP_BUF_COL (buffer), | |
608 | "\"/*\" within comment"); | |
609 | } | |
610 | goto next_char; | |
0578f103 | 611 | } |
0578f103 | 612 | } |
78719282 | 613 | else if (is_vspace (c)) |
0578f103 | 614 | { |
338fa5f7 | 615 | prevc = c, c = handle_newline (buffer, c); |
616 | goto next_char; | |
0578f103 | 617 | } |
b86584f6 | 618 | else if (c == '\t') |
338fa5f7 | 619 | adjust_column (pfile); |
0578f103 | 620 | } |
f80e83a9 | 621 | |
f669338a | 622 | pfile->state.lexing_comment = 0; |
338fa5f7 | 623 | buffer->read_ahead = EOF; |
624 | return c != '/' || prevc != '*'; | |
0578f103 | 625 | } |
626 | ||
241e762e | 627 | /* Skip a C++ line comment. Handles escaped newlines. Returns |
338fa5f7 | 628 | non-zero if a multiline comment. The following new line, if any, |
629 | is left in buffer->read_ahead. */ | |
f80e83a9 | 630 | static int |
f669338a | 631 | skip_line_comment (pfile) |
632 | cpp_reader *pfile; | |
0578f103 | 633 | { |
f669338a | 634 | cpp_buffer *buffer = pfile->buffer; |
338fa5f7 | 635 | unsigned int orig_lineno = buffer->lineno; |
636 | cppchar_t c; | |
f80e83a9 | 637 | |
f669338a | 638 | pfile->state.lexing_comment = 1; |
338fa5f7 | 639 | do |
f80e83a9 | 640 | { |
338fa5f7 | 641 | c = EOF; |
642 | if (buffer->cur == buffer->rlimit) | |
643 | break; | |
f80e83a9 | 644 | |
338fa5f7 | 645 | c = *buffer->cur++; |
646 | if (c == '?' || c == '\\') | |
647 | c = skip_escaped_newlines (buffer, c); | |
f80e83a9 | 648 | } |
338fa5f7 | 649 | while (!is_vspace (c)); |
0578f103 | 650 | |
f669338a | 651 | pfile->state.lexing_comment = 0; |
338fa5f7 | 652 | buffer->read_ahead = c; /* Leave any newline for caller. */ |
653 | return orig_lineno != buffer->lineno; | |
f80e83a9 | 654 | } |
0578f103 | 655 | |
338fa5f7 | 656 | /* pfile->buffer->cur is one beyond the \t character. Update |
657 | col_adjust so we track the column correctly. */ | |
b86584f6 | 658 | static void |
338fa5f7 | 659 | adjust_column (pfile) |
b86584f6 | 660 | cpp_reader *pfile; |
b86584f6 | 661 | { |
338fa5f7 | 662 | cpp_buffer *buffer = pfile->buffer; |
663 | unsigned int col = CPP_BUF_COL (buffer) - 1; /* Zero-based column. */ | |
b86584f6 | 664 | |
665 | /* Round it up to multiple of the tabstop, but subtract 1 since the | |
666 | tab itself occupies a character position. */ | |
338fa5f7 | 667 | buffer->col_adjust += (CPP_OPTION (pfile, tabstop) |
668 | - col % CPP_OPTION (pfile, tabstop)) - 1; | |
b86584f6 | 669 | } |
670 | ||
338fa5f7 | 671 | /* Skips whitespace, saving the next non-whitespace character. |
672 | Adjusts pfile->col_adjust to account for tabs. Without this, | |
673 | tokens might be assigned an incorrect column. */ | |
f80e83a9 | 674 | static void |
338fa5f7 | 675 | skip_whitespace (pfile, c) |
f80e83a9 | 676 | cpp_reader *pfile; |
338fa5f7 | 677 | cppchar_t c; |
f80e83a9 | 678 | { |
679 | cpp_buffer *buffer = pfile->buffer; | |
338fa5f7 | 680 | unsigned int warned = 0; |
0578f103 | 681 | |
338fa5f7 | 682 | do |
f80e83a9 | 683 | { |
78719282 | 684 | /* Horizontal space always OK. */ |
685 | if (c == ' ') | |
338fa5f7 | 686 | ; |
78719282 | 687 | else if (c == '\t') |
338fa5f7 | 688 | adjust_column (pfile); |
689 | /* Just \f \v or \0 left. */ | |
78719282 | 690 | else if (c == '\0') |
f80e83a9 | 691 | { |
78719282 | 692 | if (!warned) |
338fa5f7 | 693 | { |
694 | cpp_warning (pfile, "null character(s) ignored"); | |
695 | warned = 1; | |
696 | } | |
0578f103 | 697 | } |
338fa5f7 | 698 | else if (IN_DIRECTIVE (pfile) && CPP_PEDANTIC (pfile)) |
78719282 | 699 | cpp_pedwarn_with_line (pfile, CPP_BUF_LINE (buffer), |
700 | CPP_BUF_COL (buffer), | |
701 | "%s in preprocessing directive", | |
702 | c == '\f' ? "form feed" : "vertical tab"); | |
338fa5f7 | 703 | |
704 | c = EOF; | |
705 | if (buffer->cur == buffer->rlimit) | |
706 | break; | |
707 | c = *buffer->cur++; | |
0578f103 | 708 | } |
338fa5f7 | 709 | /* We only want non-vertical space, i.e. ' ' \t \f \v \0. */ |
710 | while (is_nvspace (c)); | |
711 | ||
712 | /* Remember the next character. */ | |
713 | buffer->read_ahead = c; | |
f80e83a9 | 714 | } |
0578f103 | 715 | |
338fa5f7 | 716 | /* Parse an identifier, skipping embedded backslash-newlines. |
717 | Calculate the hash value of the token while parsing, for improved | |
718 | performance. The hashing algorithm *must* match cpp_lookup(). */ | |
719 | ||
720 | static cpp_hashnode * | |
721 | parse_identifier (pfile, c) | |
0578f103 | 722 | cpp_reader *pfile; |
338fa5f7 | 723 | cppchar_t c; |
0578f103 | 724 | { |
338fa5f7 | 725 | cpp_buffer *buffer = pfile->buffer; |
726 | unsigned int r = 0, saw_dollar = 0; | |
727 | unsigned int orig_used = pfile->token_list.name_used; | |
f80e83a9 | 728 | |
338fa5f7 | 729 | do |
f80e83a9 | 730 | { |
338fa5f7 | 731 | do |
f80e83a9 | 732 | { |
338fa5f7 | 733 | if (pfile->token_list.name_used == pfile->token_list.name_cap) |
734 | _cpp_expand_name_space (&pfile->token_list, | |
735 | pfile->token_list.name_used + 256); | |
736 | pfile->token_list.namebuf[pfile->token_list.name_used++] = c; | |
737 | r = HASHSTEP (r, c); | |
0578f103 | 738 | |
338fa5f7 | 739 | if (c == '$') |
740 | saw_dollar++; | |
71aa9da4 | 741 | |
338fa5f7 | 742 | c = EOF; |
743 | if (buffer->cur == buffer->rlimit) | |
744 | break; | |
71aa9da4 | 745 | |
338fa5f7 | 746 | c = *buffer->cur++; |
747 | } | |
748 | while (is_idchar (c)); | |
71aa9da4 | 749 | |
338fa5f7 | 750 | /* Potential escaped newline? */ |
751 | if (c != '?' && c != '\\') | |
752 | break; | |
753 | c = skip_escaped_newlines (buffer, c); | |
f80e83a9 | 754 | } |
338fa5f7 | 755 | while (is_idchar (c)); |
756 | ||
757 | /* $ is not a identifier character in the standard, but is commonly | |
758 | accepted as an extension. Don't warn about it in skipped | |
759 | conditional blocks. */ | |
760 | if (saw_dollar && CPP_PEDANTIC (pfile) && ! pfile->skipping) | |
761 | cpp_pedwarn (pfile, "'$' character(s) in identifier"); | |
762 | ||
763 | /* Remember the next character. */ | |
764 | buffer->read_ahead = c; | |
765 | return _cpp_lookup_with_hash (pfile, &pfile->token_list.namebuf[orig_used], | |
766 | pfile->token_list.name_used - orig_used, r); | |
0578f103 | 767 | } |
768 | ||
338fa5f7 | 769 | /* Parse a number, skipping embedded backslash-newlines. */ |
0578f103 | 770 | static void |
338fa5f7 | 771 | parse_number (pfile, number, c) |
0578f103 | 772 | cpp_reader *pfile; |
338fa5f7 | 773 | cpp_string *number; |
774 | cppchar_t c; | |
0578f103 | 775 | { |
338fa5f7 | 776 | cppchar_t prevc; |
f80e83a9 | 777 | cpp_buffer *buffer = pfile->buffer; |
338fa5f7 | 778 | unsigned int orig_used = pfile->token_list.name_used; |
0578f103 | 779 | |
f669338a | 780 | /* Reserve space for a leading period. */ |
781 | if (pfile->state.seen_dot) | |
782 | pfile->token_list.name_used++; | |
783 | ||
338fa5f7 | 784 | do |
f80e83a9 | 785 | { |
338fa5f7 | 786 | do |
787 | { | |
f669338a | 788 | if (pfile->token_list.name_used >= pfile->token_list.name_cap) |
338fa5f7 | 789 | _cpp_expand_name_space (&pfile->token_list, |
790 | pfile->token_list.name_used + 256); | |
791 | pfile->token_list.namebuf[pfile->token_list.name_used++] = c; | |
792 | ||
793 | prevc = c; | |
794 | c = EOF; | |
795 | if (buffer->cur == buffer->rlimit) | |
796 | break; | |
0578f103 | 797 | |
338fa5f7 | 798 | c = *buffer->cur++; |
799 | } | |
800 | while (is_numchar (c) || c == '.' || VALID_SIGN (c, prevc)); | |
0578f103 | 801 | |
338fa5f7 | 802 | /* Potential escaped newline? */ |
803 | if (c != '?' && c != '\\') | |
804 | break; | |
805 | c = skip_escaped_newlines (buffer, c); | |
0578f103 | 806 | } |
338fa5f7 | 807 | while (is_numchar (c) || c == '.' || VALID_SIGN (c, prevc)); |
808 | ||
f669338a | 809 | /* Put any leading period in place, now we have the room. */ |
810 | if (pfile->state.seen_dot) | |
811 | pfile->token_list.namebuf[orig_used] = '.'; | |
812 | ||
338fa5f7 | 813 | /* Remember the next character. */ |
814 | buffer->read_ahead = c; | |
852d1b04 | 815 | |
338fa5f7 | 816 | number->text = &pfile->token_list.namebuf[orig_used]; |
817 | number->len = pfile->token_list.name_used - orig_used; | |
818 | } | |
819 | ||
820 | /* Subroutine of parse_string. Emits error for unterminated strings. */ | |
821 | static void | |
822 | unterminated (pfile, line, term) | |
823 | cpp_reader *pfile; | |
824 | unsigned int line; | |
825 | int term; | |
826 | { | |
827 | cpp_error (pfile, "missing terminating %c character", term); | |
828 | ||
829 | if (term == '\"' && pfile->mls_line && pfile->mls_line != line) | |
f80e83a9 | 830 | { |
338fa5f7 | 831 | cpp_error_with_line (pfile, pfile->mls_line, pfile->mls_column, |
832 | "possible start of unterminated string literal"); | |
833 | pfile->mls_line = 0; | |
f80e83a9 | 834 | } |
0578f103 | 835 | } |
836 | ||
338fa5f7 | 837 | /* Parses a string, character constant, or angle-bracketed header file |
838 | name. Handles embedded trigraphs and escaped newlines. | |
0578f103 | 839 | |
338fa5f7 | 840 | Multi-line strings are allowed, but they are deprecated within |
841 | directives. */ | |
f80e83a9 | 842 | static void |
338fa5f7 | 843 | parse_string (pfile, token, terminator) |
0578f103 | 844 | cpp_reader *pfile; |
f80e83a9 | 845 | cpp_token *token; |
338fa5f7 | 846 | cppchar_t terminator; |
0578f103 | 847 | { |
f80e83a9 | 848 | cpp_buffer *buffer = pfile->buffer; |
338fa5f7 | 849 | unsigned int orig_used = pfile->token_list.name_used; |
850 | cppchar_t c; | |
851 | unsigned int nulls = 0; | |
852 | ||
853 | for (;;) | |
0578f103 | 854 | { |
338fa5f7 | 855 | if (buffer->cur == buffer->rlimit) |
856 | { | |
857 | c = EOF; | |
858 | unterminated (pfile, token->line, terminator); | |
859 | break; | |
860 | } | |
861 | c = *buffer->cur++; | |
862 | ||
863 | have_char: | |
864 | /* Handle trigraphs, escaped newlines etc. */ | |
865 | if (c == '?' || c == '\\') | |
866 | c = skip_escaped_newlines (buffer, c); | |
0578f103 | 867 | |
338fa5f7 | 868 | if (c == terminator) |
0578f103 | 869 | { |
338fa5f7 | 870 | unsigned int u = pfile->token_list.name_used; |
01f93cf7 | 871 | |
338fa5f7 | 872 | /* An odd number of consecutive backslashes represents an |
873 | escaped terminator. */ | |
874 | while (u > orig_used && pfile->token_list.namebuf[u - 1] == '\\') | |
875 | u--; | |
876 | ||
877 | if ((pfile->token_list.name_used - u) % 2 == 0) | |
0578f103 | 878 | { |
338fa5f7 | 879 | c = EOF; |
880 | break; | |
0578f103 | 881 | } |
338fa5f7 | 882 | } |
883 | else if (is_vspace (c)) | |
884 | { | |
885 | /* In assembly language, silently terminate string and | |
886 | character literals at end of line. This is a kludge | |
887 | around not knowing where comments are. */ | |
888 | if (CPP_OPTION (pfile, lang_asm) && terminator != '>') | |
889 | break; | |
0578f103 | 890 | |
338fa5f7 | 891 | /* Character constants and header names may not extend over |
892 | multiple lines. In Standard C, neither may strings. | |
893 | Unfortunately, we accept multiline strings as an | |
894 | extension. (Deprecatedly even in directives - otherwise, | |
895 | glibc's longlong.h breaks.) */ | |
896 | if (terminator != '"') | |
0578f103 | 897 | { |
338fa5f7 | 898 | unterminated (pfile, token->line, terminator); |
899 | break; | |
0578f103 | 900 | } |
0578f103 | 901 | |
338fa5f7 | 902 | if (pfile->mls_line == 0) |
903 | { | |
904 | pfile->mls_line = token->line; | |
905 | pfile->mls_column = token->col; | |
906 | if (CPP_PEDANTIC (pfile)) | |
907 | cpp_pedwarn (pfile, "multi-line string constant"); | |
f80e83a9 | 908 | } |
338fa5f7 | 909 | |
910 | handle_newline (buffer, c); /* Stores to read_ahead. */ | |
911 | c = '\n'; | |
912 | } | |
913 | else if (c == '\0') | |
914 | { | |
915 | if (nulls++ == 0) | |
916 | cpp_warning (pfile, "null character(s) preserved in literal"); | |
0578f103 | 917 | } |
0578f103 | 918 | |
338fa5f7 | 919 | if (pfile->token_list.name_used == pfile->token_list.name_cap) |
920 | _cpp_expand_name_space (&pfile->token_list, | |
921 | pfile->token_list.name_used + 256); | |
9fb5b53d | 922 | |
338fa5f7 | 923 | pfile->token_list.namebuf[pfile->token_list.name_used++] = c; |
924 | /* If we had a new line, the next character is in read_ahead. */ | |
925 | if (c != '\n') | |
926 | continue; | |
927 | c = buffer->read_ahead; | |
928 | if (c != EOF) | |
929 | goto have_char; | |
0578f103 | 930 | } |
931 | ||
338fa5f7 | 932 | buffer->read_ahead = c; |
0578f103 | 933 | |
338fa5f7 | 934 | token->val.str.text = &pfile->token_list.namebuf[orig_used]; |
935 | token->val.str.len = pfile->token_list.name_used - orig_used; | |
936 | } | |
f80e83a9 | 937 | |
338fa5f7 | 938 | /* For output routine simplicity, the stored comment includes the |
939 | comment start and any terminator. */ | |
2c63d6c8 | 940 | static void |
338fa5f7 | 941 | save_comment (pfile, token, from) |
942 | cpp_reader *pfile; | |
f80e83a9 | 943 | cpp_token *token; |
944 | const unsigned char *from; | |
2c63d6c8 | 945 | { |
f80e83a9 | 946 | unsigned char *buffer; |
338fa5f7 | 947 | unsigned int len; |
948 | cpp_toklist *list = &pfile->token_list; | |
949 | ||
950 | #define COMMENT_START_LEN 2 | |
951 | len = pfile->buffer->cur - from + COMMENT_START_LEN; | |
952 | _cpp_reserve_name_space (list, len); | |
953 | buffer = list->namebuf + list->name_used; | |
954 | list->name_used += len; | |
f80e83a9 | 955 | |
f80e83a9 | 956 | token->type = CPP_COMMENT; |
76faa4c0 | 957 | token->val.str.len = len; |
338fa5f7 | 958 | token->val.str.text = buffer; |
0578f103 | 959 | |
338fa5f7 | 960 | /* from[-1] is '/' or '*' depending on the comment type. */ |
961 | *buffer++ = '/'; | |
962 | *buffer++ = from[-1]; | |
963 | memcpy (buffer, from, len - COMMENT_START_LEN); | |
964 | } | |
0578f103 | 965 | |
f669338a | 966 | /* Subroutine of lex_token to handle '%'. A little tricky, since we |
967 | want to avoid stepping back when lexing %:%X. */ | |
338fa5f7 | 968 | static void |
f669338a | 969 | lex_percent (buffer, result) |
338fa5f7 | 970 | cpp_buffer *buffer; |
971 | cpp_token *result; | |
338fa5f7 | 972 | { |
f669338a | 973 | cppchar_t c; |
974 | ||
975 | result->type = CPP_MOD; | |
976 | /* Parsing %:%X could leave an extra character. */ | |
977 | if (buffer->extra_char == EOF) | |
978 | c = get_effective_char (buffer); | |
979 | else | |
980 | { | |
981 | c = buffer->read_ahead = buffer->extra_char; | |
982 | buffer->extra_char = EOF; | |
983 | } | |
984 | ||
985 | if (c == '=') | |
986 | ACCEPT_CHAR (CPP_MOD_EQ); | |
987 | else if (CPP_OPTION (buffer->pfile, digraphs)) | |
988 | { | |
989 | if (c == ':') | |
990 | { | |
991 | result->flags |= DIGRAPH; | |
992 | ACCEPT_CHAR (CPP_HASH); | |
993 | if (get_effective_char (buffer) == '%') | |
994 | { | |
995 | buffer->extra_char = get_effective_char (buffer); | |
996 | if (buffer->extra_char == ':') | |
997 | { | |
998 | buffer->extra_char = EOF; | |
999 | ACCEPT_CHAR (CPP_PASTE); | |
1000 | } | |
1001 | else | |
1002 | /* We'll catch the extra_char when we're called back. */ | |
1003 | buffer->read_ahead = '%'; | |
1004 | } | |
1005 | } | |
1006 | else if (c == '>') | |
1007 | { | |
1008 | result->flags |= DIGRAPH; | |
1009 | ACCEPT_CHAR (CPP_CLOSE_BRACE); | |
1010 | } | |
1011 | } | |
1012 | } | |
1013 | ||
1014 | /* Subroutine of lex_token to handle '.'. This is tricky, since we | |
1015 | want to avoid stepping back when lexing '...' or '.123'. In the | |
1016 | latter case we should also set a flag for parse_number. */ | |
1017 | static void | |
1018 | lex_dot (pfile, result) | |
1019 | cpp_reader *pfile; | |
1020 | cpp_token *result; | |
1021 | { | |
1022 | cpp_buffer *buffer = pfile->buffer; | |
1023 | cppchar_t c; | |
1024 | ||
1025 | /* Parsing ..X could leave an extra character. */ | |
1026 | if (buffer->extra_char == EOF) | |
1027 | c = get_effective_char (buffer); | |
1028 | else | |
1029 | { | |
1030 | c = buffer->read_ahead = buffer->extra_char; | |
1031 | buffer->extra_char = EOF; | |
1032 | } | |
338fa5f7 | 1033 | |
f669338a | 1034 | /* All known character sets have 0...9 contiguous. */ |
1035 | if (c >= '0' && c <= '9') | |
1036 | { | |
1037 | result->type = CPP_NUMBER; | |
1038 | buffer->pfile->state.seen_dot = 1; | |
1039 | parse_number (pfile, &result->val.str, c); | |
1040 | buffer->pfile->state.seen_dot = 0; | |
1041 | } | |
f80e83a9 | 1042 | else |
c4357c92 | 1043 | { |
f669338a | 1044 | result->type = CPP_DOT; |
1045 | if (c == '.') | |
1046 | { | |
1047 | buffer->extra_char = get_effective_char (buffer); | |
1048 | if (buffer->extra_char == '.') | |
1049 | { | |
1050 | buffer->extra_char = EOF; | |
1051 | ACCEPT_CHAR (CPP_ELLIPSIS); | |
1052 | } | |
1053 | else | |
1054 | /* We'll catch the extra_char when we're called back. */ | |
1055 | buffer->read_ahead = '.'; | |
1056 | } | |
1057 | else if (c == '*' && CPP_OPTION (pfile, cplusplus)) | |
1058 | ACCEPT_CHAR (CPP_DOT_STAR); | |
c4357c92 | 1059 | } |
0578f103 | 1060 | } |
1061 | ||
f80e83a9 | 1062 | static void |
338fa5f7 | 1063 | lex_token (pfile, result) |
0578f103 | 1064 | cpp_reader *pfile; |
338fa5f7 | 1065 | cpp_token *result; |
0578f103 | 1066 | { |
338fa5f7 | 1067 | cppchar_t c; |
f80e83a9 | 1068 | cpp_buffer *buffer = pfile->buffer; |
338fa5f7 | 1069 | const unsigned char *comment_start; |
0653b94e | 1070 | |
338fa5f7 | 1071 | result->flags = 0; |
1072 | next_char: | |
1073 | result->line = CPP_BUF_LINE (buffer); | |
1074 | next_char2: | |
1075 | result->col = CPP_BUF_COLUMN (buffer, buffer->cur); | |
f80e83a9 | 1076 | |
338fa5f7 | 1077 | c = buffer->read_ahead; |
1078 | if (c == EOF && buffer->cur < buffer->rlimit) | |
1079 | { | |
1080 | c = *buffer->cur++; | |
1081 | result->col++; | |
1082 | } | |
0578f103 | 1083 | |
338fa5f7 | 1084 | do_switch: |
1085 | buffer->read_ahead = EOF; | |
1086 | switch (c) | |
0578f103 | 1087 | { |
338fa5f7 | 1088 | case EOF: |
1089 | /* Non-empty files should end in a newline. Testing | |
1090 | skip_newlines ensures we only emit the warning once. */ | |
1091 | if (buffer->cur != buffer->line_base && buffer->cur != buffer->buf | |
1092 | && pfile->state.skip_newlines) | |
1093 | cpp_pedwarn_with_line (pfile, buffer->lineno, CPP_BUF_COL (buffer), | |
1094 | "no newline at end of file"); | |
1095 | result->type = CPP_EOF; | |
1096 | break; | |
0578f103 | 1097 | |
338fa5f7 | 1098 | case ' ': case '\t': case '\f': case '\v': case '\0': |
1099 | skip_whitespace (pfile, c); | |
1100 | result->flags |= PREV_WHITE; | |
1101 | goto next_char2; | |
1102 | ||
1103 | case '\n': case '\r': | |
1104 | result->type = CPP_EOF; | |
1105 | handle_newline (buffer, c); | |
1106 | /* Handling here will change significantly when moving to | |
1107 | token-at-a-time. */ | |
1108 | if (pfile->state.skip_newlines) | |
0578f103 | 1109 | { |
338fa5f7 | 1110 | result->flags &= ~PREV_WHITE; /* Clear any whitespace flag. */ |
1111 | goto next_char; | |
0578f103 | 1112 | } |
338fa5f7 | 1113 | break; |
732cb4c9 | 1114 | |
338fa5f7 | 1115 | case '?': |
1116 | case '\\': | |
1117 | /* These could start an escaped newline, or '?' a trigraph. Let | |
1118 | skip_escaped_newlines do all the work. */ | |
1119 | { | |
1120 | unsigned int lineno = buffer->lineno; | |
1121 | ||
1122 | c = skip_escaped_newlines (buffer, c); | |
1123 | if (lineno != buffer->lineno) | |
1124 | /* We had at least one escaped newline of some sort, and the | |
1125 | next character is in buffer->read_ahead. Update the | |
1126 | token's line and column. */ | |
1127 | goto next_char; | |
1128 | ||
1129 | /* We are either the original '?' or '\\', or a trigraph. */ | |
1130 | result->type = CPP_QUERY; | |
1131 | buffer->read_ahead = EOF; | |
1132 | if (c == '\\') | |
1133 | result->type = CPP_BACKSLASH; | |
1134 | else if (c != '?') | |
1135 | goto do_switch; | |
1136 | } | |
1137 | break; | |
732cb4c9 | 1138 | |
338fa5f7 | 1139 | case '0': case '1': case '2': case '3': case '4': |
1140 | case '5': case '6': case '7': case '8': case '9': | |
1141 | result->type = CPP_NUMBER; | |
1142 | parse_number (pfile, &result->val.str, c); | |
1143 | break; | |
732cb4c9 | 1144 | |
338fa5f7 | 1145 | case '$': |
1146 | if (!CPP_OPTION (pfile, dollars_in_ident)) | |
1147 | goto random_char; | |
1148 | /* Fall through... */ | |
1149 | ||
1150 | case '_': | |
1151 | case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': | |
1152 | case 'g': case 'h': case 'i': case 'j': case 'k': case 'l': | |
1153 | case 'm': case 'n': case 'o': case 'p': case 'q': case 'r': | |
1154 | case 's': case 't': case 'u': case 'v': case 'w': case 'x': | |
1155 | case 'y': case 'z': | |
1156 | case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': | |
1157 | case 'G': case 'H': case 'I': case 'J': case 'K': case 'L': | |
1158 | case 'M': case 'N': case 'O': case 'P': case 'Q': case 'R': | |
1159 | case 'S': case 'T': case 'U': case 'V': case 'W': case 'X': | |
1160 | case 'Y': case 'Z': | |
1161 | result->type = CPP_NAME; | |
1162 | result->val.node = parse_identifier (pfile, c); | |
1163 | ||
1164 | /* 'L' may introduce wide characters or strings. */ | |
1165 | if (result->val.node == pfile->spec_nodes->n_L) | |
1166 | { | |
1167 | c = buffer->read_ahead; /* For make_string. */ | |
1168 | if (c == '\'' || c == '"') | |
71aa9da4 | 1169 | { |
338fa5f7 | 1170 | ACCEPT_CHAR (c == '"' ? CPP_WSTRING: CPP_WCHAR); |
1171 | goto make_string; | |
71aa9da4 | 1172 | } |
338fa5f7 | 1173 | } |
1174 | /* Convert named operators to their proper types. */ | |
1175 | else if (result->val.node->type == T_OPERATOR) | |
1176 | { | |
1177 | result->flags |= NAMED_OP; | |
1178 | result->type = result->val.node->value.code; | |
1179 | } | |
1180 | break; | |
1181 | ||
1182 | case '\'': | |
1183 | case '"': | |
1184 | result->type = c == '"' ? CPP_STRING: CPP_CHAR; | |
1185 | make_string: | |
1186 | parse_string (pfile, result, c); | |
1187 | break; | |
f80e83a9 | 1188 | |
338fa5f7 | 1189 | case '/': |
1190 | result->type = CPP_DIV; | |
1191 | c = get_effective_char (buffer); | |
1192 | if (c == '=') | |
1193 | ACCEPT_CHAR (CPP_DIV_EQ); | |
1194 | else if (c == '*') | |
1195 | { | |
1196 | comment_start = buffer->cur; | |
0578f103 | 1197 | |
338fa5f7 | 1198 | /* Skip_block_comment updates buffer->read_ahead. */ |
1199 | if (skip_block_comment (pfile)) | |
1200 | cpp_error_with_line (pfile, result->line, result->col, | |
1201 | "unterminated comment"); | |
1202 | if (!pfile->state.save_comments) | |
31674461 | 1203 | { |
338fa5f7 | 1204 | result->flags |= PREV_WHITE; |
1205 | goto next_char; | |
31674461 | 1206 | } |
1207 | ||
338fa5f7 | 1208 | /* Save the comment as a token in its own right. */ |
1209 | save_comment (pfile, result, comment_start); | |
1210 | } | |
1211 | else if (c == '/') | |
1212 | { | |
1213 | /* We silently allow C++ comments in system headers, | |
1214 | irrespective of conformance mode, because lots of | |
1215 | broken systems do that and trying to clean it up in | |
1216 | fixincludes is a nightmare. */ | |
1217 | if (CPP_IN_SYSTEM_HEADER (pfile)) | |
1218 | goto do_line_comment; | |
1219 | if (CPP_OPTION (pfile, cplusplus_comments)) | |
f80e83a9 | 1220 | { |
338fa5f7 | 1221 | if (CPP_OPTION (pfile, c89) && CPP_PEDANTIC (pfile) |
1222 | && ! buffer->warned_cplusplus_comments) | |
f80e83a9 | 1223 | { |
338fa5f7 | 1224 | cpp_pedwarn (pfile, |
1225 | "C++ style comments are not allowed in ISO C89"); | |
1226 | cpp_pedwarn (pfile, | |
1227 | "(this will be reported only once per input file)"); | |
1228 | buffer->warned_cplusplus_comments = 1; | |
f80e83a9 | 1229 | } |
f80e83a9 | 1230 | |
338fa5f7 | 1231 | do_line_comment: |
1232 | comment_start = buffer->cur; | |
1233 | ||
1234 | /* Skip_line_comment updates buffer->read_ahead. */ | |
f669338a | 1235 | if (skip_line_comment (pfile)) |
338fa5f7 | 1236 | cpp_warning_with_line (pfile, result->line, result->col, |
1237 | "multi-line comment"); | |
1238 | ||
1239 | if (!pfile->state.save_comments) | |
f80e83a9 | 1240 | { |
338fa5f7 | 1241 | result->flags |= PREV_WHITE; |
1242 | goto next_char; | |
f80e83a9 | 1243 | } |
338fa5f7 | 1244 | |
1245 | /* Save the comment as a token in its own right. */ | |
1246 | save_comment (pfile, result, comment_start); | |
0578f103 | 1247 | } |
338fa5f7 | 1248 | } |
1249 | break; | |
1250 | ||
1251 | case '<': | |
1252 | if (pfile->state.angled_headers) | |
1253 | { | |
1254 | result->type = CPP_HEADER_NAME; | |
1255 | c = '>'; /* terminator. */ | |
1256 | goto make_string; | |
1257 | } | |
0578f103 | 1258 | |
338fa5f7 | 1259 | result->type = CPP_LESS; |
1260 | c = get_effective_char (buffer); | |
1261 | if (c == '=') | |
1262 | ACCEPT_CHAR (CPP_LESS_EQ); | |
1263 | else if (c == '<') | |
1264 | { | |
1265 | ACCEPT_CHAR (CPP_LSHIFT); | |
1266 | if (get_effective_char (buffer) == '=') | |
1267 | ACCEPT_CHAR (CPP_LSHIFT_EQ); | |
1268 | } | |
1269 | else if (c == '?' && CPP_OPTION (pfile, cplusplus)) | |
1270 | { | |
1271 | ACCEPT_CHAR (CPP_MIN); | |
1272 | if (get_effective_char (buffer) == '=') | |
1273 | ACCEPT_CHAR (CPP_MIN_EQ); | |
1274 | } | |
1275 | else if (c == ':' && CPP_OPTION (pfile, digraphs)) | |
1276 | { | |
1277 | ACCEPT_CHAR (CPP_OPEN_SQUARE); | |
1278 | result->flags |= DIGRAPH; | |
1279 | } | |
1280 | else if (c == '%' && CPP_OPTION (pfile, digraphs)) | |
1281 | { | |
1282 | ACCEPT_CHAR (CPP_OPEN_BRACE); | |
1283 | result->flags |= DIGRAPH; | |
1284 | } | |
1285 | break; | |
1286 | ||
1287 | case '>': | |
1288 | result->type = CPP_GREATER; | |
1289 | c = get_effective_char (buffer); | |
1290 | if (c == '=') | |
1291 | ACCEPT_CHAR (CPP_GREATER_EQ); | |
1292 | else if (c == '>') | |
1293 | { | |
1294 | ACCEPT_CHAR (CPP_RSHIFT); | |
1295 | if (get_effective_char (buffer) == '=') | |
1296 | ACCEPT_CHAR (CPP_RSHIFT_EQ); | |
1297 | } | |
1298 | else if (c == '?' && CPP_OPTION (pfile, cplusplus)) | |
1299 | { | |
1300 | ACCEPT_CHAR (CPP_MAX); | |
1301 | if (get_effective_char (buffer) == '=') | |
1302 | ACCEPT_CHAR (CPP_MAX_EQ); | |
1303 | } | |
1304 | break; | |
1305 | ||
f669338a | 1306 | case '%': |
1307 | lex_percent (buffer, result); | |
338fa5f7 | 1308 | break; |
1309 | ||
f669338a | 1310 | case '.': |
1311 | lex_dot (pfile, result); | |
338fa5f7 | 1312 | break; |
0578f103 | 1313 | |
338fa5f7 | 1314 | case '+': |
1315 | result->type = CPP_PLUS; | |
1316 | c = get_effective_char (buffer); | |
1317 | if (c == '=') | |
1318 | ACCEPT_CHAR (CPP_PLUS_EQ); | |
1319 | else if (c == '+') | |
1320 | ACCEPT_CHAR (CPP_PLUS_PLUS); | |
1321 | break; | |
ac0749c7 | 1322 | |
338fa5f7 | 1323 | case '-': |
1324 | result->type = CPP_MINUS; | |
1325 | c = get_effective_char (buffer); | |
1326 | if (c == '>') | |
1327 | { | |
1328 | ACCEPT_CHAR (CPP_DEREF); | |
1329 | if (CPP_OPTION (pfile, cplusplus) | |
1330 | && get_effective_char (buffer) == '*') | |
1331 | ACCEPT_CHAR (CPP_DEREF_STAR); | |
1332 | } | |
1333 | else if (c == '=') | |
1334 | ACCEPT_CHAR (CPP_MINUS_EQ); | |
1335 | else if (c == '-') | |
1336 | ACCEPT_CHAR (CPP_MINUS_MINUS); | |
1337 | break; | |
0578f103 | 1338 | |
338fa5f7 | 1339 | case '*': |
1340 | result->type = CPP_MULT; | |
1341 | if (get_effective_char (buffer) == '=') | |
1342 | ACCEPT_CHAR (CPP_MULT_EQ); | |
1343 | break; | |
ac0749c7 | 1344 | |
338fa5f7 | 1345 | case '=': |
1346 | result->type = CPP_EQ; | |
1347 | if (get_effective_char (buffer) == '=') | |
1348 | ACCEPT_CHAR (CPP_EQ_EQ); | |
1349 | break; | |
c4abf88d | 1350 | |
338fa5f7 | 1351 | case '!': |
1352 | result->type = CPP_NOT; | |
1353 | if (get_effective_char (buffer) == '=') | |
1354 | ACCEPT_CHAR (CPP_NOT_EQ); | |
1355 | break; | |
0578f103 | 1356 | |
338fa5f7 | 1357 | case '&': |
1358 | result->type = CPP_AND; | |
1359 | c = get_effective_char (buffer); | |
1360 | if (c == '=') | |
1361 | ACCEPT_CHAR (CPP_AND_EQ); | |
1362 | else if (c == '&') | |
1363 | ACCEPT_CHAR (CPP_AND_AND); | |
1364 | break; | |
1365 | ||
1366 | case '#': | |
1367 | result->type = CPP_HASH; | |
1368 | if (get_effective_char (buffer) == '#') | |
1369 | ACCEPT_CHAR (CPP_PASTE); | |
1370 | break; | |
0578f103 | 1371 | |
338fa5f7 | 1372 | case '|': |
1373 | result->type = CPP_OR; | |
1374 | c = get_effective_char (buffer); | |
1375 | if (c == '=') | |
1376 | ACCEPT_CHAR (CPP_OR_EQ); | |
1377 | else if (c == '|') | |
1378 | ACCEPT_CHAR (CPP_OR_OR); | |
1379 | break; | |
0578f103 | 1380 | |
338fa5f7 | 1381 | case '^': |
1382 | result->type = CPP_XOR; | |
1383 | if (get_effective_char (buffer) == '=') | |
1384 | ACCEPT_CHAR (CPP_XOR_EQ); | |
1385 | break; | |
0578f103 | 1386 | |
338fa5f7 | 1387 | case ':': |
1388 | result->type = CPP_COLON; | |
1389 | c = get_effective_char (buffer); | |
1390 | if (c == ':' && CPP_OPTION (pfile, cplusplus)) | |
1391 | ACCEPT_CHAR (CPP_SCOPE); | |
1392 | else if (c == '>' && CPP_OPTION (pfile, digraphs)) | |
1393 | { | |
1394 | result->flags |= DIGRAPH; | |
1395 | ACCEPT_CHAR (CPP_CLOSE_SQUARE); | |
1396 | } | |
1397 | break; | |
0578f103 | 1398 | |
338fa5f7 | 1399 | case '~': result->type = CPP_COMPL; break; |
1400 | case ',': result->type = CPP_COMMA; break; | |
1401 | case '(': result->type = CPP_OPEN_PAREN; break; | |
1402 | case ')': result->type = CPP_CLOSE_PAREN; break; | |
1403 | case '[': result->type = CPP_OPEN_SQUARE; break; | |
1404 | case ']': result->type = CPP_CLOSE_SQUARE; break; | |
1405 | case '{': result->type = CPP_OPEN_BRACE; break; | |
1406 | case '}': result->type = CPP_CLOSE_BRACE; break; | |
1407 | case ';': result->type = CPP_SEMICOLON; break; | |
1408 | ||
1409 | case '@': | |
1410 | if (CPP_OPTION (pfile, objc)) | |
1411 | { | |
1412 | /* In Objective C, '@' may begin keywords or strings, like | |
1413 | @keyword or @"string". It would be nice to call | |
1414 | get_effective_char here and test the result. However, we | |
1415 | would then need to pass 2 characters to parse_identifier, | |
1416 | making it ugly and slowing down its main loop. Instead, | |
1417 | we assume we have an identifier, and recover if not. */ | |
1418 | result->type = CPP_NAME; | |
1419 | result->val.node = parse_identifier (pfile, c); | |
1420 | if (result->val.node->length != 1) | |
1421 | break; | |
ac0749c7 | 1422 | |
338fa5f7 | 1423 | /* OK, so it wasn't an identifier. Maybe a string? */ |
1424 | if (buffer->read_ahead == '"') | |
f80e83a9 | 1425 | { |
338fa5f7 | 1426 | c = '"'; |
1427 | ACCEPT_CHAR (CPP_OSTRING); | |
1428 | goto make_string; | |
f80e83a9 | 1429 | } |
338fa5f7 | 1430 | } |
1431 | goto random_char; | |
1432 | ||
1433 | random_char: | |
1434 | default: | |
1435 | result->type = CPP_OTHER; | |
1436 | result->val.aux = c; | |
1437 | break; | |
1438 | } | |
1439 | } | |
1440 | ||
1441 | /* | |
1442 | * The tokenizer's main loop. Returns a token list, representing a | |
1443 | * logical line in the input file. On EOF after some tokens have | |
1444 | * been processed, we return immediately. Then in next call, or if | |
1445 | * EOF occurred at the beginning of a logical line, a single CPP_EOF | |
1446 | * token is placed in the list. | |
1447 | */ | |
1448 | ||
1449 | static void | |
1450 | lex_line (pfile, list) | |
1451 | cpp_reader *pfile; | |
1452 | cpp_toklist *list; | |
1453 | { | |
1454 | unsigned int first_token; | |
1455 | cpp_token *cur_token, *first; | |
1456 | cpp_buffer *buffer = pfile->buffer; | |
0578f103 | 1457 | |
338fa5f7 | 1458 | pfile->state.in_lex_line = 1; |
1459 | if (pfile->buffer->cur == pfile->buffer->buf) | |
1460 | list->flags |= BEG_OF_FILE; | |
1461 | ||
1462 | retry: | |
1463 | pfile->state.in_directive = 0; | |
1464 | pfile->state.angled_headers = 0; | |
1465 | pfile->state.skip_newlines = 1; | |
1466 | pfile->state.save_comments = ! CPP_OPTION (pfile, discard_comments); | |
1467 | first_token = list->tokens_used; | |
1468 | list->file = buffer->nominal_fname; | |
1469 | ||
1470 | do | |
1471 | { | |
1472 | if (list->tokens_used >= list->tokens_cap) | |
1473 | _cpp_expand_token_space (list, 256); | |
1474 | ||
1475 | cur_token = list->tokens + list->tokens_used; | |
1476 | lex_token (pfile, cur_token); | |
1477 | ||
1478 | if (pfile->state.skip_newlines) | |
1479 | { | |
1480 | pfile->state.skip_newlines = 0; | |
1481 | list->line = buffer->lineno; | |
1482 | if (cur_token->type == CPP_HASH) | |
f80e83a9 | 1483 | { |
338fa5f7 | 1484 | pfile->state.in_directive = 1; |
1485 | pfile->state.save_comments = 0; | |
1486 | pfile->state.indented = cur_token->flags & PREV_WHITE; | |
f80e83a9 | 1487 | } |
338fa5f7 | 1488 | /* 6.10.3.10: Within the sequence of preprocessing tokens |
1489 | making up the invocation of a function-like macro, new | |
1490 | line is considered a normal white-space character. */ | |
1491 | else if (first_token != 0) | |
1492 | cur_token->flags |= PREV_WHITE; | |
1493 | } | |
1494 | else if (IN_DIRECTIVE (pfile) && list->tokens_used == first_token + 1) | |
1495 | { | |
1496 | if (cur_token->type == CPP_NUMBER) | |
1497 | list->directive = _cpp_check_linemarker (pfile, cur_token); | |
f80e83a9 | 1498 | else |
338fa5f7 | 1499 | list->directive = _cpp_check_directive (pfile, cur_token); |
f80e83a9 | 1500 | } |
cfad5579 | 1501 | |
338fa5f7 | 1502 | /* _cpp_get_line assumes list->tokens_used refers to the current |
1503 | token being lexed. So do this after _cpp_check_directive to | |
1504 | get the warnings therein correct. */ | |
1505 | list->tokens_used++; | |
f80e83a9 | 1506 | } |
338fa5f7 | 1507 | while (cur_token->type != CPP_EOF); |
cfad5579 | 1508 | |
f80e83a9 | 1509 | /* All tokens are allocated, so the memory location is fixed. */ |
1510 | first = &list->tokens[first_token]; | |
338fa5f7 | 1511 | first->flags |= BOL; |
1512 | pfile->first_directive_token = first; | |
1513 | ||
f80e83a9 | 1514 | /* Don't complain about the null directive, nor directives in |
1515 | assembly source: we don't know where the comments are, and # may | |
1516 | introduce assembler pseudo-ops. Don't complain about invalid | |
1517 | directives in skipped conditional groups (6.10 p4). */ | |
338fa5f7 | 1518 | if (IN_DIRECTIVE (pfile) && !KNOWN_DIRECTIVE (list) && !pfile->skipping |
1519 | && !CPP_OPTION (pfile, lang_asm)) | |
f80e83a9 | 1520 | { |
338fa5f7 | 1521 | if (cur_token > first + 1) |
1522 | { | |
1523 | if (first[1].type == CPP_NAME) | |
1524 | cpp_error_with_line (pfile, first->line, first->col, | |
1525 | "invalid preprocessing directive #%s", | |
1526 | first[1].val.node->name); | |
1527 | else | |
1528 | cpp_error_with_line (pfile, first->line, first->col, | |
1529 | "invalid preprocessing directive"); | |
1530 | } | |
0653b94e | 1531 | |
1532 | /* Discard this line to prevent further errors from cc1. */ | |
1533 | _cpp_clear_toklist (list); | |
1534 | goto retry; | |
f80e83a9 | 1535 | } |
1536 | ||
be6e77fe | 1537 | /* Drop the EOF unless really at EOF or in a directive. */ |
1538 | if (cur_token != first && !KNOWN_DIRECTIVE (list) | |
1539 | && pfile->done_initializing) | |
1540 | list->tokens_used--; | |
1541 | ||
338fa5f7 | 1542 | pfile->state.in_lex_line = 0; |
cfad5579 | 1543 | } |
1544 | ||
f80e83a9 | 1545 | /* Write the spelling of a token TOKEN to BUFFER. The buffer must |
c5ea33a8 | 1546 | already contain the enough space to hold the token's spelling. |
1547 | Returns a pointer to the character after the last character | |
1548 | written. */ | |
ab12a39c | 1549 | |
f80e83a9 | 1550 | static unsigned char * |
1551 | spell_token (pfile, token, buffer) | |
1552 | cpp_reader *pfile; /* Would be nice to be rid of this... */ | |
1553 | const cpp_token *token; | |
1554 | unsigned char *buffer; | |
1555 | { | |
7e842f95 | 1556 | switch (TOKEN_SPELL (token)) |
f80e83a9 | 1557 | { |
1558 | case SPELL_OPERATOR: | |
1559 | { | |
1560 | const unsigned char *spelling; | |
1561 | unsigned char c; | |
ab12a39c | 1562 | |
f80e83a9 | 1563 | if (token->flags & DIGRAPH) |
524f0c40 | 1564 | spelling = _cpp_digraph_spellings[token->type - CPP_FIRST_DIGRAPH]; |
31674461 | 1565 | else if (token->flags & NAMED_OP) |
1566 | goto spell_ident; | |
f80e83a9 | 1567 | else |
7e842f95 | 1568 | spelling = TOKEN_NAME (token); |
f80e83a9 | 1569 | |
1570 | while ((c = *spelling++) != '\0') | |
1571 | *buffer++ = c; | |
1572 | } | |
1573 | break; | |
ab12a39c | 1574 | |
f80e83a9 | 1575 | case SPELL_IDENT: |
31674461 | 1576 | spell_ident: |
76faa4c0 | 1577 | memcpy (buffer, token->val.node->name, token->val.node->length); |
1578 | buffer += token->val.node->length; | |
f80e83a9 | 1579 | break; |
ab12a39c | 1580 | |
f80e83a9 | 1581 | case SPELL_STRING: |
1582 | { | |
71aa9da4 | 1583 | int left, right, tag; |
1584 | switch (token->type) | |
1585 | { | |
1586 | case CPP_STRING: left = '"'; right = '"'; tag = '\0'; break; | |
1587 | case CPP_WSTRING: left = '"'; right = '"'; tag = 'L'; break; | |
1588 | case CPP_OSTRING: left = '"'; right = '"'; tag = '@'; break; | |
1589 | case CPP_CHAR: left = '\''; right = '\''; tag = '\0'; break; | |
1590 | case CPP_WCHAR: left = '\''; right = '\''; tag = 'L'; break; | |
1591 | case CPP_HEADER_NAME: left = '<'; right = '>'; tag = '\0'; break; | |
1592 | default: left = '\0'; right = '\0'; tag = '\0'; break; | |
1593 | } | |
1594 | if (tag) *buffer++ = tag; | |
1595 | if (left) *buffer++ = left; | |
76faa4c0 | 1596 | memcpy (buffer, token->val.str.text, token->val.str.len); |
1597 | buffer += token->val.str.len; | |
71aa9da4 | 1598 | if (right) *buffer++ = right; |
f80e83a9 | 1599 | } |
1600 | break; | |
ab12a39c | 1601 | |
f80e83a9 | 1602 | case SPELL_CHAR: |
1603 | *buffer++ = token->val.aux; | |
1604 | break; | |
ab12a39c | 1605 | |
f80e83a9 | 1606 | case SPELL_NONE: |
7e842f95 | 1607 | cpp_ice (pfile, "Unspellable token %s", TOKEN_NAME (token)); |
f80e83a9 | 1608 | break; |
1609 | } | |
ab12a39c | 1610 | |
f80e83a9 | 1611 | return buffer; |
1612 | } | |
ab12a39c | 1613 | |
fdf3a98f | 1614 | /* Macro expansion algorithm. |
1615 | ||
1616 | Macro expansion is implemented by a single-pass algorithm; there are | |
1617 | no rescan passes involved. cpp_get_token expands just enough to be | |
1618 | able to return a token to the caller, a consequence is that when it | |
1619 | returns the preprocessor can be in a state of mid-expansion. The | |
1620 | algorithm does not work by fully expanding a macro invocation into | |
1621 | some kind of token list, and then returning them one by one. | |
1622 | ||
1623 | Our expansion state is recorded in a context stack. We start out with | |
1624 | a single context on the stack, let's call it base context. This | |
1625 | consists of the token list returned by lex_line that forms the next | |
1626 | logical line in the source file. | |
1627 | ||
1628 | The current level in the context stack is stored in the cur_context | |
1629 | member of the cpp_reader structure. The context it references keeps, | |
1630 | amongst other things, a count of how many tokens form that context and | |
1631 | our position within those tokens. | |
1632 | ||
1633 | Fundamentally, calling cpp_get_token will return the next token from | |
1634 | the current context. If we're at the end of the current context, that | |
1635 | context is popped from the stack first, unless it is the base context, | |
1636 | in which case the next logical line is lexed from the source file. | |
1637 | ||
1638 | However, before returning the token, if it is a CPP_NAME token | |
1639 | _cpp_get_token checks to see if it is a macro and if it is enabled. | |
1640 | Each time it encounters a macro name, it calls push_macro_context. | |
1641 | This function checks that the macro should be expanded (with | |
1642 | is_macro_enabled), and if so pushes a new macro context on the stack | |
1643 | which becomes the current context. It then loops back to read the | |
1644 | first token of the macro context. | |
1645 | ||
1646 | A macro context basically consists of the token list representing the | |
1647 | macro's replacement list, which was saved in the hash table by | |
1648 | save_macro_expansion when its #define statement was parsed. If the | |
1649 | macro is function-like, it also contains the tokens that form the | |
1650 | arguments to the macro. I say more about macro arguments below, but | |
1651 | for now just saying that each argument is a set of pointers to tokens | |
1652 | is enough. | |
1653 | ||
1654 | When taking tokens from a macro context, we may get a CPP_MACRO_ARG | |
1655 | token. This represents an argument passed to the macro, with the | |
1656 | argument number stored in the token's AUX field. The argument should | |
1657 | be substituted, this is achieved by pushing an "argument context". An | |
1658 | argument context is just refers to the tokens forming the argument, | |
1659 | which are obtained directly from the macro context. The STRINGIFY | |
1660 | flag on a CPP_MACRO_ARG token indicates that the argument should be | |
1661 | stringified. | |
1662 | ||
1663 | Here's a few simple rules the context stack obeys:- | |
1664 | ||
1665 | 1) The lex_line token list is always context zero. | |
1666 | ||
1667 | 2) Context 1, if it exists, must be a macro context. | |
1668 | ||
1669 | 3) An argument context can only appear above a macro context. | |
1670 | ||
1671 | 4) A macro context can appear above the base context, another macro | |
1672 | context, or an argument context. | |
1673 | ||
1674 | 5) These imply that the minimal level of an argument context is 2. | |
1675 | ||
1676 | The only tricky thing left is ensuring that macros are enabled and | |
1677 | disabled correctly. The algorithm controls macro expansion by the | |
1678 | level of the context a token is taken from in the context stack. If a | |
1679 | token is taken from a level equal to no_expand_level (a member of | |
1680 | struct cpp_reader), no expansion is performed. | |
1681 | ||
1682 | When popping a context off the stack, if no_expand_level equals the | |
1683 | level of the popped context, it is reduced by one to match the new | |
1684 | context level, so that expansion is still disabled. It does not | |
1685 | increase if a context is pushed, though. It starts out life as | |
1686 | UINT_MAX, which has the effect that initially macro expansion is | |
1687 | enabled. I explain how this mechanism works below. | |
1688 | ||
1689 | The standard requires:- | |
1690 | ||
1691 | 1) Arguments to be fully expanded before substitution. | |
1692 | ||
1693 | 2) Stringified arguments to not be expanded, nor the tokens | |
1694 | immediately surrounding a ## operator. | |
1695 | ||
1696 | 3) Continual rescanning until there are no more macros left to | |
1697 | replace. | |
1698 | ||
1699 | 4) Once a macro has been expanded in stage 1) or 3), it cannot be | |
1700 | expanded again during later rescans. This prevents infinite | |
1701 | recursion. | |
1702 | ||
1703 | The first thing to observe is that stage 3) is mostly redundant. | |
1704 | Since a macro is disabled once it has been expanded, how can a rescan | |
1705 | find an unexpanded macro name? There are only two cases where this is | |
1706 | possible:- | |
1707 | ||
1708 | a) If the macro name results from a token paste operation. | |
1709 | ||
1710 | b) If the macro in question is a function-like macro that hasn't | |
1711 | already been expanded because previously there was not the required | |
1712 | '(' token immediately following it. This is only possible when an | |
1713 | argument is substituted, and after substitution the last token of | |
1714 | the argument can bind with a parenthesis appearing in the tokens | |
1715 | following the substitution. Note that if the '(' appears within the | |
1716 | argument, the ')' must too, as expanding macro arguments cannot | |
1717 | "suck in" tokens outside the argument. | |
1718 | ||
1719 | So we tackle this as follows. When parsing the macro invocation for | |
1720 | arguments, we record the tokens forming each argument as a list of | |
1721 | pointers to those tokens. We do not expand any tokens that are "raw", | |
1722 | i.e. directly from the macro invocation, but other tokens that come | |
1723 | from (nested) argument substitution are fully expanded. | |
1724 | ||
1725 | This is achieved by setting the no_expand_level to that of the macro | |
1726 | invocation. A CPP_MACRO_ARG token never appears in the list of tokens | |
1727 | forming an argument, because parse_args (indirectly) calls | |
1728 | get_raw_token which automatically pushes argument contexts and traces | |
1729 | into them. Since these contexts are at a higher level than the | |
1730 | no_expand_level, they get fully macro expanded. | |
1731 | ||
1732 | "Raw" and non-raw tokens are separated in arguments by null pointers, | |
1733 | with the policy that the initial state of an argument is raw. If the | |
1734 | first token is not raw, it should be preceded by a null pointer. When | |
1735 | tracing through the tokens of an argument context, each time | |
1736 | get_raw_token encounters a null pointer, it toggles the flag | |
1737 | CONTEXT_RAW. | |
1738 | ||
1739 | This flag, when set, indicates to is_macro_disabled that we are | |
1740 | reading raw tokens which should be macro-expanded. Similarly, if | |
1741 | clear, is_macro_disabled suppresses re-expansion. | |
1742 | ||
1743 | It's probably time for an example. | |
1744 | ||
1745 | #define hash # | |
1746 | #define str(x) #x | |
1747 | #define xstr(y) str(y hash) | |
1748 | str(hash) // "hash" | |
1749 | xstr(hash) // "# hash" | |
1750 | ||
1751 | In the invocation of str, parse_args turns off macro expansion and so | |
1752 | parses the argument as <hash>. This is the only token (pointer) | |
1753 | passed as the argument to str. Since <hash> is raw there is no need | |
1754 | for an initial null pointer. stringify_arg is called from | |
1755 | get_raw_token when tracing through the expansion of str, since the | |
1756 | argument has the STRINGIFY flag set. stringify_arg turns off | |
1757 | macro_expansion by setting the no_expand_level to that of the argument | |
1758 | context. Thus it gets the token <hash> and stringifies it to "hash" | |
1759 | correctly. | |
1760 | ||
1761 | Similary xstr is passed <hash>. However, when parse_args is parsing | |
1762 | the invocation of str() in xstr's expansion, get_raw_token encounters | |
1763 | a CPP_MACRO_ARG token for y. Transparently to parse_args, it pushes | |
1764 | an argument context, and enters the tokens of the argument, | |
1765 | i.e. <hash>. This is at a higher context level than parse_args | |
1766 | disabled, and so is_macro_disabled permits expansion of it and a macro | |
1767 | context is pushed on top of the argument context. This contains the | |
1768 | <#> token, and the end result is that <hash> is macro expanded. | |
1769 | However, after popping off the argument context, the <hash> of xstr's | |
1770 | expansion does not get macro expanded because we're back at the | |
1771 | no_expand_level. The end result is that the argument passed to str is | |
1772 | <NULL> <#> <NULL> <hash>. Note the nulls - policy is we start off | |
1773 | raw, <#> is not raw, but then <hash> is. | |
1774 | ||
1775 | */ | |
ab12a39c | 1776 | |
f80e83a9 | 1777 | |
1778 | /* Free the storage allocated for macro arguments. */ | |
1779 | static void | |
1780 | free_macro_args (args) | |
1781 | macro_args *args; | |
6060326b | 1782 | { |
f80e83a9 | 1783 | if (args->tokens) |
deb356cf | 1784 | free ((PTR) args->tokens); |
f80e83a9 | 1785 | free (args->ends); |
1786 | free (args); | |
6060326b | 1787 | } |
1788 | ||
f80e83a9 | 1789 | /* Determines if a macro has been already used (and is therefore |
1790 | disabled). */ | |
6060326b | 1791 | static int |
f80e83a9 | 1792 | is_macro_disabled (pfile, expansion, token) |
6060326b | 1793 | cpp_reader *pfile; |
f80e83a9 | 1794 | const cpp_toklist *expansion; |
1795 | const cpp_token *token; | |
6060326b | 1796 | { |
f80e83a9 | 1797 | cpp_context *context = CURRENT_CONTEXT (pfile); |
1798 | ||
1799 | /* Arguments on either side of ## are inserted in place without | |
1800 | macro expansion (6.10.3.3.2). Conceptually, any macro expansion | |
1801 | occurs during a later rescan pass. The effect is that we expand | |
1802 | iff we would as part of the macro's expansion list, so we should | |
1803 | drop to the macro's context. */ | |
1804 | if (IS_ARG_CONTEXT (context)) | |
6060326b | 1805 | { |
f80e83a9 | 1806 | if (token->flags & PASTED) |
1807 | context--; | |
1808 | else if (!(context->flags & CONTEXT_RAW)) | |
1809 | return 1; | |
1810 | else if (context->flags & (CONTEXT_PASTEL | CONTEXT_PASTER)) | |
1811 | context--; | |
6060326b | 1812 | } |
6060326b | 1813 | |
f80e83a9 | 1814 | /* Have we already used this macro? */ |
1815 | while (context->level > 0) | |
6060326b | 1816 | { |
f80e83a9 | 1817 | if (!IS_ARG_CONTEXT (context) && context->u.list == expansion) |
1818 | return 1; | |
1819 | /* Raw argument tokens are judged based on the token list they | |
1820 | came from. */ | |
1821 | if (context->flags & CONTEXT_RAW) | |
1822 | context = pfile->contexts + context->level; | |
1823 | else | |
1824 | context--; | |
1825 | } | |
6060326b | 1826 | |
f80e83a9 | 1827 | /* Function-like macros may be disabled if the '(' is not in the |
1828 | current context. We check this without disrupting the context | |
1829 | stack. */ | |
1830 | if (expansion->paramc >= 0) | |
1831 | { | |
1832 | const cpp_token *next; | |
1833 | unsigned int prev_nme; | |
6060326b | 1834 | |
f80e83a9 | 1835 | context = CURRENT_CONTEXT (pfile); |
1836 | /* Drop down any contexts we're at the end of: the '(' may | |
1837 | appear in lower macro expansions, or in the rest of the file. */ | |
1838 | while (context->posn == context->count && context > pfile->contexts) | |
1839 | { | |
1840 | context--; | |
1841 | /* If we matched, we are disabled, as we appear in the | |
1842 | expansion of each macro we meet. */ | |
1843 | if (!IS_ARG_CONTEXT (context) && context->u.list == expansion) | |
1844 | return 1; | |
1845 | } | |
6060326b | 1846 | |
f80e83a9 | 1847 | prev_nme = pfile->no_expand_level; |
1848 | pfile->no_expand_level = context - pfile->contexts; | |
deb356cf | 1849 | next = _cpp_get_token (pfile); |
f80e83a9 | 1850 | restore_macro_expansion (pfile, prev_nme); |
1851 | if (next->type != CPP_OPEN_PAREN) | |
1852 | { | |
1853 | _cpp_push_token (pfile, next); | |
1449850f | 1854 | if (CPP_WTRADITIONAL (pfile)) |
f80e83a9 | 1855 | cpp_warning (pfile, |
31674461 | 1856 | "function macro %s must be used with arguments in traditional C", |
1857 | token->val.node->name); | |
f80e83a9 | 1858 | return 1; |
1859 | } | |
6060326b | 1860 | } |
6060326b | 1861 | |
f80e83a9 | 1862 | return 0; |
6060326b | 1863 | } |
1864 | ||
f80e83a9 | 1865 | /* Add a token to the set of tokens forming the arguments to the macro |
1866 | being parsed in parse_args. */ | |
1867 | static void | |
1868 | save_token (args, token) | |
1869 | macro_args *args; | |
1870 | const cpp_token *token; | |
6060326b | 1871 | { |
f80e83a9 | 1872 | if (args->used == args->capacity) |
1873 | { | |
1874 | args->capacity += args->capacity + 100; | |
1875 | args->tokens = (const cpp_token **) | |
deb356cf | 1876 | xrealloc ((PTR) args->tokens, |
1877 | args->capacity * sizeof (const cpp_token *)); | |
f80e83a9 | 1878 | } |
1879 | args->tokens[args->used++] = token; | |
6060326b | 1880 | } |
1881 | ||
f80e83a9 | 1882 | /* Take and save raw tokens until we finish one argument. Empty |
1883 | arguments are saved as a single CPP_PLACEMARKER token. */ | |
1884 | static const cpp_token * | |
1885 | parse_arg (pfile, var_args, paren_context, args, pcount) | |
6060326b | 1886 | cpp_reader *pfile; |
f80e83a9 | 1887 | int var_args; |
1888 | unsigned int paren_context; | |
1889 | macro_args *args; | |
1890 | unsigned int *pcount; | |
6060326b | 1891 | { |
f80e83a9 | 1892 | const cpp_token *token; |
1893 | unsigned int paren = 0, count = 0; | |
1894 | int raw, was_raw = 1; | |
6060326b | 1895 | |
f80e83a9 | 1896 | for (count = 0;; count++) |
6060326b | 1897 | { |
deb356cf | 1898 | token = _cpp_get_token (pfile); |
6060326b | 1899 | |
f80e83a9 | 1900 | switch (token->type) |
6060326b | 1901 | { |
f80e83a9 | 1902 | default: |
1903 | break; | |
6060326b | 1904 | |
f80e83a9 | 1905 | case CPP_OPEN_PAREN: |
1906 | paren++; | |
1907 | break; | |
1908 | ||
1909 | case CPP_CLOSE_PAREN: | |
1910 | if (paren-- != 0) | |
1911 | break; | |
1912 | goto out; | |
1913 | ||
1914 | case CPP_COMMA: | |
1915 | /* Commas are not terminators within parantheses or var_args. */ | |
1916 | if (paren || var_args) | |
1917 | break; | |
1918 | goto out; | |
1919 | ||
1920 | case CPP_EOF: /* Error reported by caller. */ | |
1921 | goto out; | |
6060326b | 1922 | } |
6060326b | 1923 | |
f80e83a9 | 1924 | raw = pfile->cur_context <= paren_context; |
1925 | if (raw != was_raw) | |
1926 | { | |
1927 | was_raw = raw; | |
1928 | save_token (args, 0); | |
1929 | count++; | |
6060326b | 1930 | } |
f80e83a9 | 1931 | save_token (args, token); |
6060326b | 1932 | } |
6060326b | 1933 | |
1934 | out: | |
f80e83a9 | 1935 | if (count == 0) |
1936 | { | |
1937 | /* Duplicate the placemarker. Then we can set its flags and | |
1938 | position and safely be using more than one. */ | |
1939 | save_token (args, duplicate_token (pfile, &placemarker_token)); | |
1940 | count++; | |
1941 | } | |
1942 | ||
1943 | *pcount = count; | |
1944 | return token; | |
6060326b | 1945 | } |
1946 | ||
f80e83a9 | 1947 | /* This macro returns true if the argument starting at offset O of arglist |
1948 | A is empty - that is, it's either a single PLACEMARKER token, or a null | |
1949 | pointer followed by a PLACEMARKER. */ | |
1950 | ||
1951 | #define empty_argument(A, O) \ | |
1952 | ((A)->tokens[O] ? (A)->tokens[O]->type == CPP_PLACEMARKER \ | |
1953 | : (A)->tokens[(O)+1]->type == CPP_PLACEMARKER) | |
1954 | ||
1955 | /* Parse the arguments making up a macro invocation. Nested arguments | |
1956 | are automatically macro expanded, but immediate macros are not | |
1957 | expanded; this enables e.g. operator # to work correctly. Returns | |
1958 | non-zero on error. */ | |
6060326b | 1959 | static int |
f80e83a9 | 1960 | parse_args (pfile, hp, args) |
6060326b | 1961 | cpp_reader *pfile; |
f80e83a9 | 1962 | cpp_hashnode *hp; |
1963 | macro_args *args; | |
6060326b | 1964 | { |
f80e83a9 | 1965 | const cpp_token *token; |
1966 | const cpp_toklist *macro; | |
1967 | unsigned int total = 0; | |
1968 | unsigned int paren_context = pfile->cur_context; | |
1969 | int argc = 0; | |
1970 | ||
1971 | macro = hp->value.expansion; | |
1972 | do | |
6060326b | 1973 | { |
f80e83a9 | 1974 | unsigned int count; |
6060326b | 1975 | |
f80e83a9 | 1976 | token = parse_arg (pfile, (argc + 1 == macro->paramc |
1977 | && (macro->flags & VAR_ARGS)), | |
1978 | paren_context, args, &count); | |
1979 | if (argc < macro->paramc) | |
6060326b | 1980 | { |
f80e83a9 | 1981 | total += count; |
1982 | args->ends[argc] = total; | |
6060326b | 1983 | } |
f80e83a9 | 1984 | argc++; |
6060326b | 1985 | } |
f80e83a9 | 1986 | while (token->type != CPP_CLOSE_PAREN && token->type != CPP_EOF); |
6060326b | 1987 | |
f80e83a9 | 1988 | if (token->type == CPP_EOF) |
1989 | { | |
31674461 | 1990 | cpp_error(pfile, "unterminated argument list for macro \"%s\"", hp->name); |
f80e83a9 | 1991 | return 1; |
1992 | } | |
1993 | else if (argc < macro->paramc) | |
1994 | { | |
1995 | /* A rest argument is allowed to not appear in the invocation at all. | |
1996 | e.g. #define debug(format, args...) ... | |
1997 | debug("string"); | |
1998 | This is exactly the same as if the rest argument had received no | |
57ba19ba | 1999 | tokens - debug("string",); This extension is deprecated. */ |
35677230 | 2000 | |
2001 | if (argc + 1 == macro->paramc && (macro->flags & VAR_ARGS)) | |
f80e83a9 | 2002 | { |
2003 | /* Duplicate the placemarker. Then we can set its flags and | |
2004 | position and safely be using more than one. */ | |
35677230 | 2005 | cpp_token *pm = duplicate_token (pfile, &placemarker_token); |
2006 | pm->flags = VOID_REST; | |
2007 | save_token (args, pm); | |
f80e83a9 | 2008 | args->ends[argc] = total + 1; |
35677230 | 2009 | |
2010 | if (CPP_OPTION (pfile, c99) && CPP_PEDANTIC (pfile)) | |
2011 | cpp_pedwarn (pfile, "ISO C99 requires rest arguments to be used"); | |
2012 | ||
f80e83a9 | 2013 | return 0; |
2014 | } | |
2015 | else | |
2016 | { | |
31674461 | 2017 | cpp_error (pfile, "not enough arguments for macro \"%s\"", hp->name); |
f80e83a9 | 2018 | return 1; |
2019 | } | |
2020 | } | |
2021 | /* An empty argument to an empty function-like macro is fine. */ | |
2022 | else if (argc > macro->paramc | |
2023 | && !(macro->paramc == 0 && argc == 1 && empty_argument (args, 0))) | |
2024 | { | |
31674461 | 2025 | cpp_error (pfile, "too many arguments for macro \"%s\"", hp->name); |
f80e83a9 | 2026 | return 1; |
2027 | } | |
6060326b | 2028 | |
f80e83a9 | 2029 | return 0; |
2030 | } | |
2031 | ||
2032 | /* Adds backslashes before all backslashes and double quotes appearing | |
2033 | in strings. Non-printable characters are converted to octal. */ | |
2034 | static U_CHAR * | |
2035 | quote_string (dest, src, len) | |
2036 | U_CHAR *dest; | |
2037 | const U_CHAR *src; | |
2038 | unsigned int len; | |
2039 | { | |
2040 | while (len--) | |
6060326b | 2041 | { |
f80e83a9 | 2042 | U_CHAR c = *src++; |
6060326b | 2043 | |
f80e83a9 | 2044 | if (c == '\\' || c == '"') |
3b304865 | 2045 | { |
f80e83a9 | 2046 | *dest++ = '\\'; |
2047 | *dest++ = c; | |
2048 | } | |
2049 | else | |
2050 | { | |
2051 | if (ISPRINT (c)) | |
2052 | *dest++ = c; | |
2053 | else | |
2054 | { | |
2055 | sprintf ((char *) dest, "\\%03o", c); | |
2056 | dest += 4; | |
2057 | } | |
3b304865 | 2058 | } |
6060326b | 2059 | } |
6060326b | 2060 | |
f80e83a9 | 2061 | return dest; |
6060326b | 2062 | } |
2063 | ||
f80e83a9 | 2064 | /* Allocates a buffer to hold a token's TEXT, and converts TOKEN to a |
2065 | CPP_STRING token containing TEXT in quoted form. */ | |
2066 | static cpp_token * | |
2067 | make_string_token (token, text, len) | |
2068 | cpp_token *token; | |
2069 | const U_CHAR *text; | |
2070 | unsigned int len; | |
2071 | { | |
2072 | U_CHAR *buf; | |
2073 | ||
2074 | buf = (U_CHAR *) xmalloc (len * 4); | |
2075 | token->type = CPP_STRING; | |
2076 | token->flags = 0; | |
76faa4c0 | 2077 | token->val.str.text = buf; |
2078 | token->val.str.len = quote_string (buf, text, len) - buf; | |
f80e83a9 | 2079 | return token; |
2080 | } | |
2081 | ||
2082 | /* Allocates and converts a temporary token to a CPP_NUMBER token, | |
2083 | evaluating to NUMBER. */ | |
2084 | static cpp_token * | |
2085 | alloc_number_token (pfile, number) | |
6060326b | 2086 | cpp_reader *pfile; |
f80e83a9 | 2087 | int number; |
6060326b | 2088 | { |
f80e83a9 | 2089 | cpp_token *result; |
2090 | char *buf; | |
2091 | ||
2092 | result = get_temp_token (pfile); | |
2093 | buf = xmalloc (20); | |
2094 | sprintf (buf, "%d", number); | |
2095 | ||
2096 | result->type = CPP_NUMBER; | |
2097 | result->flags = 0; | |
76faa4c0 | 2098 | result->val.str.text = (U_CHAR *) buf; |
2099 | result->val.str.len = strlen (buf); | |
f80e83a9 | 2100 | return result; |
2101 | } | |
6060326b | 2102 | |
f80e83a9 | 2103 | /* Returns a temporary token from the temporary token store of PFILE. */ |
2104 | static cpp_token * | |
2105 | get_temp_token (pfile) | |
2106 | cpp_reader *pfile; | |
2107 | { | |
2108 | if (pfile->temp_used == pfile->temp_alloced) | |
2109 | { | |
2110 | if (pfile->temp_used == pfile->temp_cap) | |
2111 | { | |
2112 | pfile->temp_cap += pfile->temp_cap + 20; | |
2113 | pfile->temp_tokens = (cpp_token **) xrealloc | |
2114 | (pfile->temp_tokens, pfile->temp_cap * sizeof (cpp_token *)); | |
2115 | } | |
2116 | pfile->temp_tokens[pfile->temp_alloced++] = (cpp_token *) xmalloc | |
2117 | (sizeof (cpp_token)); | |
2118 | } | |
6060326b | 2119 | |
f80e83a9 | 2120 | return pfile->temp_tokens[pfile->temp_used++]; |
2121 | } | |
2122 | ||
2123 | /* Release (not free) for re-use the temporary tokens of PFILE. */ | |
2124 | static void | |
2125 | release_temp_tokens (pfile) | |
2126 | cpp_reader *pfile; | |
2127 | { | |
2128 | while (pfile->temp_used) | |
6060326b | 2129 | { |
f80e83a9 | 2130 | cpp_token *token = pfile->temp_tokens[--pfile->temp_used]; |
6060326b | 2131 | |
7e842f95 | 2132 | if (TOKEN_SPELL (token) == SPELL_STRING) |
6060326b | 2133 | { |
76faa4c0 | 2134 | free ((char *) token->val.str.text); |
2135 | token->val.str.text = 0; | |
6060326b | 2136 | } |
2137 | } | |
f80e83a9 | 2138 | } |
6060326b | 2139 | |
f80e83a9 | 2140 | /* Free all of PFILE's dynamically-allocated temporary tokens. */ |
2141 | void | |
2142 | _cpp_free_temp_tokens (pfile) | |
2143 | cpp_reader *pfile; | |
2144 | { | |
2145 | if (pfile->temp_tokens) | |
6060326b | 2146 | { |
f80e83a9 | 2147 | /* It is possible, though unlikely (looking for '(' of a funlike |
2148 | macro into EOF), that we haven't released the tokens yet. */ | |
2149 | release_temp_tokens (pfile); | |
2150 | while (pfile->temp_alloced) | |
2151 | free (pfile->temp_tokens[--pfile->temp_alloced]); | |
2152 | free (pfile->temp_tokens); | |
6060326b | 2153 | } |
2154 | ||
f80e83a9 | 2155 | if (pfile->date) |
2156 | { | |
76faa4c0 | 2157 | free ((char *) pfile->date->val.str.text); |
f80e83a9 | 2158 | free (pfile->date); |
76faa4c0 | 2159 | free ((char *) pfile->time->val.str.text); |
f80e83a9 | 2160 | free (pfile->time); |
2161 | } | |
6060326b | 2162 | } |
2163 | ||
f80e83a9 | 2164 | /* Copy TOKEN into a temporary token from PFILE's store. */ |
2165 | static cpp_token * | |
2166 | duplicate_token (pfile, token) | |
2167 | cpp_reader *pfile; | |
2168 | const cpp_token *token; | |
2169 | { | |
2170 | cpp_token *result = get_temp_token (pfile); | |
6060326b | 2171 | |
f80e83a9 | 2172 | *result = *token; |
7e842f95 | 2173 | if (TOKEN_SPELL (token) == SPELL_STRING) |
f80e83a9 | 2174 | { |
76faa4c0 | 2175 | U_CHAR *buff = (U_CHAR *) xmalloc (token->val.str.len); |
2176 | memcpy (buff, token->val.str.text, token->val.str.len); | |
2177 | result->val.str.text = buff; | |
f80e83a9 | 2178 | } |
2179 | return result; | |
2180 | } | |
6060326b | 2181 | |
f80e83a9 | 2182 | /* Determine whether two tokens can be pasted together, and if so, |
2183 | what the resulting token is. Returns CPP_EOF if the tokens cannot | |
2184 | be pasted, or the appropriate type for the merged token if they | |
2185 | can. */ | |
524f0c40 | 2186 | enum cpp_ttype |
2187 | _cpp_can_paste (pfile, token1, token2, digraph) | |
f80e83a9 | 2188 | cpp_reader * pfile; |
2189 | const cpp_token *token1, *token2; | |
2190 | int* digraph; | |
6060326b | 2191 | { |
f80e83a9 | 2192 | enum cpp_ttype a = token1->type, b = token2->type; |
2193 | int cxx = CPP_OPTION (pfile, cplusplus); | |
6060326b | 2194 | |
31674461 | 2195 | /* Treat named operators as if they were ordinary NAMEs. */ |
2196 | if (token1->flags & NAMED_OP) | |
2197 | a = CPP_NAME; | |
2198 | if (token2->flags & NAMED_OP) | |
2199 | b = CPP_NAME; | |
2200 | ||
f80e83a9 | 2201 | if (a <= CPP_LAST_EQ && b == CPP_EQ) |
2202 | return a + (CPP_EQ_EQ - CPP_EQ); | |
6060326b | 2203 | |
f80e83a9 | 2204 | switch (a) |
6060326b | 2205 | { |
f80e83a9 | 2206 | case CPP_GREATER: |
2207 | if (b == a) return CPP_RSHIFT; | |
2208 | if (b == CPP_QUERY && cxx) return CPP_MAX; | |
2209 | if (b == CPP_GREATER_EQ) return CPP_RSHIFT_EQ; | |
2210 | break; | |
2211 | case CPP_LESS: | |
2212 | if (b == a) return CPP_LSHIFT; | |
2213 | if (b == CPP_QUERY && cxx) return CPP_MIN; | |
2214 | if (b == CPP_LESS_EQ) return CPP_LSHIFT_EQ; | |
27fdc0b6 | 2215 | if (CPP_OPTION (pfile, digraphs)) |
2216 | { | |
2217 | if (b == CPP_COLON) | |
2218 | {*digraph = 1; return CPP_OPEN_SQUARE;} /* <: digraph */ | |
2219 | if (b == CPP_MOD) | |
2220 | {*digraph = 1; return CPP_OPEN_BRACE;} /* <% digraph */ | |
2221 | } | |
f80e83a9 | 2222 | break; |
6060326b | 2223 | |
f80e83a9 | 2224 | case CPP_PLUS: if (b == a) return CPP_PLUS_PLUS; break; |
2225 | case CPP_AND: if (b == a) return CPP_AND_AND; break; | |
2226 | case CPP_OR: if (b == a) return CPP_OR_OR; break; | |
6060326b | 2227 | |
f80e83a9 | 2228 | case CPP_MINUS: |
2229 | if (b == a) return CPP_MINUS_MINUS; | |
2230 | if (b == CPP_GREATER) return CPP_DEREF; | |
2231 | break; | |
2232 | case CPP_COLON: | |
2233 | if (b == a && cxx) return CPP_SCOPE; | |
27fdc0b6 | 2234 | if (b == CPP_GREATER && CPP_OPTION (pfile, digraphs)) |
f80e83a9 | 2235 | {*digraph = 1; return CPP_CLOSE_SQUARE;} /* :> digraph */ |
2236 | break; | |
2237 | ||
2238 | case CPP_MOD: | |
27fdc0b6 | 2239 | if (CPP_OPTION (pfile, digraphs)) |
2240 | { | |
2241 | if (b == CPP_GREATER) | |
2242 | {*digraph = 1; return CPP_CLOSE_BRACE;} /* %> digraph */ | |
2243 | if (b == CPP_COLON) | |
2244 | {*digraph = 1; return CPP_HASH;} /* %: digraph */ | |
2245 | } | |
f80e83a9 | 2246 | break; |
2247 | case CPP_DEREF: | |
2248 | if (b == CPP_MULT && cxx) return CPP_DEREF_STAR; | |
2249 | break; | |
2250 | case CPP_DOT: | |
2251 | if (b == CPP_MULT && cxx) return CPP_DOT_STAR; | |
2252 | if (b == CPP_NUMBER) return CPP_NUMBER; | |
2253 | break; | |
2254 | ||
2255 | case CPP_HASH: | |
2256 | if (b == a && (token1->flags & DIGRAPH) == (token2->flags & DIGRAPH)) | |
2257 | /* %:%: digraph */ | |
2258 | {*digraph = (token1->flags & DIGRAPH); return CPP_PASTE;} | |
2259 | break; | |
2260 | ||
2261 | case CPP_NAME: | |
2262 | if (b == CPP_NAME) return CPP_NAME; | |
2263 | if (b == CPP_NUMBER | |
76faa4c0 | 2264 | && is_numstart(token2->val.str.text[0])) return CPP_NAME; |
f80e83a9 | 2265 | if (b == CPP_CHAR |
76faa4c0 | 2266 | && token1->val.node == pfile->spec_nodes->n_L) return CPP_WCHAR; |
f80e83a9 | 2267 | if (b == CPP_STRING |
76faa4c0 | 2268 | && token1->val.node == pfile->spec_nodes->n_L) return CPP_WSTRING; |
f80e83a9 | 2269 | break; |
2270 | ||
2271 | case CPP_NUMBER: | |
2272 | if (b == CPP_NUMBER) return CPP_NUMBER; | |
2273 | if (b == CPP_NAME) return CPP_NUMBER; | |
2274 | if (b == CPP_DOT) return CPP_NUMBER; | |
2275 | /* Numbers cannot have length zero, so this is safe. */ | |
2276 | if ((b == CPP_PLUS || b == CPP_MINUS) | |
76faa4c0 | 2277 | && VALID_SIGN ('+', token1->val.str.text[token1->val.str.len - 1])) |
f80e83a9 | 2278 | return CPP_NUMBER; |
2279 | break; | |
2280 | ||
71aa9da4 | 2281 | case CPP_OTHER: |
2282 | if (CPP_OPTION (pfile, objc) && token1->val.aux == '@') | |
2283 | { | |
2284 | if (b == CPP_NAME) return CPP_NAME; | |
2285 | if (b == CPP_STRING) return CPP_OSTRING; | |
2286 | } | |
2287 | ||
f80e83a9 | 2288 | default: |
2289 | break; | |
6060326b | 2290 | } |
2291 | ||
f80e83a9 | 2292 | return CPP_EOF; |
2293 | } | |
2294 | ||
2295 | /* Check if TOKEN is to be ##-pasted with the token after it. */ | |
2296 | static const cpp_token * | |
2297 | maybe_paste_with_next (pfile, token) | |
2298 | cpp_reader *pfile; | |
2299 | const cpp_token *token; | |
2300 | { | |
2301 | cpp_token *pasted; | |
2302 | const cpp_token *second; | |
2303 | cpp_context *context = CURRENT_CONTEXT (pfile); | |
2304 | ||
2305 | /* Is this token on the LHS of ## ? */ | |
f80e83a9 | 2306 | |
deb356cf | 2307 | while ((token->flags & PASTE_LEFT) |
2308 | || ((context->flags & CONTEXT_PASTEL) | |
2309 | && context->posn == context->count)) | |
6060326b | 2310 | { |
deb356cf | 2311 | /* Suppress macro expansion for next token, but don't conflict |
2312 | with the other method of suppression. If it is an argument, | |
2313 | macro expansion within the argument will still occur. */ | |
2314 | pfile->paste_level = pfile->cur_context; | |
2315 | second = _cpp_get_token (pfile); | |
2316 | pfile->paste_level = 0; | |
2317 | ||
2318 | /* Ignore placemarker argument tokens (cannot be from an empty | |
2319 | macro since macros are not expanded). */ | |
2320 | if (token->type == CPP_PLACEMARKER) | |
2321 | pasted = duplicate_token (pfile, second); | |
2322 | else if (second->type == CPP_PLACEMARKER) | |
f80e83a9 | 2323 | { |
e0a859f1 | 2324 | /* GCC has special extended semantics for , ## b where b is |
2325 | a varargs parameter: the comma disappears if b was given | |
2326 | no actual arguments (not merely if b is an empty | |
2327 | argument). */ | |
2328 | if (token->type == CPP_COMMA && second->flags & VOID_REST) | |
35677230 | 2329 | pasted = duplicate_token (pfile, second); |
deb356cf | 2330 | else |
2331 | pasted = duplicate_token (pfile, token); | |
f80e83a9 | 2332 | } |
2333 | else | |
f80e83a9 | 2334 | { |
deb356cf | 2335 | int digraph = 0; |
524f0c40 | 2336 | enum cpp_ttype type = _cpp_can_paste (pfile, token, second, &digraph); |
6060326b | 2337 | |
deb356cf | 2338 | if (type == CPP_EOF) |
2339 | { | |
2340 | if (CPP_OPTION (pfile, warn_paste)) | |
e0a859f1 | 2341 | { |
2342 | /* Do not complain about , ## <whatever> if | |
2343 | <whatever> came from a variable argument, because | |
2344 | the author probably intended the ## to trigger | |
2345 | the special extended semantics (see above). */ | |
2346 | if (token->type == CPP_COMMA | |
2347 | && IS_ARG_CONTEXT (CURRENT_CONTEXT (pfile)) | |
58fe658a | 2348 | && ON_REST_ARG (CURRENT_CONTEXT (pfile) - 1)) |
e0a859f1 | 2349 | /* no warning */; |
2350 | else | |
2351 | cpp_warning (pfile, | |
deb356cf | 2352 | "pasting would not give a valid preprocessing token"); |
e0a859f1 | 2353 | } |
deb356cf | 2354 | _cpp_push_token (pfile, second); |
58fe658a | 2355 | /* A short term hack to safely clear the PASTE_LEFT flag. */ |
2356 | pasted = duplicate_token (pfile, token); | |
2357 | pasted->flags &= ~PASTE_LEFT; | |
2358 | return pasted; | |
deb356cf | 2359 | } |
6060326b | 2360 | |
deb356cf | 2361 | if (type == CPP_NAME || type == CPP_NUMBER) |
2362 | { | |
2363 | /* Join spellings. */ | |
2364 | U_CHAR *buf, *end; | |
2365 | ||
2366 | pasted = get_temp_token (pfile); | |
2367 | buf = (U_CHAR *) alloca (TOKEN_LEN (token) + TOKEN_LEN (second)); | |
2368 | end = spell_token (pfile, token, buf); | |
2369 | end = spell_token (pfile, second, end); | |
2370 | *end = '\0'; | |
f80e83a9 | 2371 | |
deb356cf | 2372 | if (type == CPP_NAME) |
2373 | pasted->val.node = cpp_lookup (pfile, buf, end - buf); | |
2374 | else | |
2375 | { | |
2376 | pasted->val.str.text = uxstrdup (buf); | |
2377 | pasted->val.str.len = end - buf; | |
2378 | } | |
2379 | } | |
71aa9da4 | 2380 | else if (type == CPP_WCHAR || type == CPP_WSTRING |
2381 | || type == CPP_OSTRING) | |
deb356cf | 2382 | pasted = duplicate_token (pfile, second); |
76faa4c0 | 2383 | else |
2384 | { | |
deb356cf | 2385 | pasted = get_temp_token (pfile); |
2386 | pasted->val.integer = 0; | |
76faa4c0 | 2387 | } |
deb356cf | 2388 | |
2389 | pasted->type = type; | |
2390 | pasted->flags = digraph ? DIGRAPH : 0; | |
31674461 | 2391 | |
2392 | if (type == CPP_NAME && pasted->val.node->type == T_OPERATOR) | |
2393 | { | |
2394 | pasted->type = pasted->val.node->value.code; | |
2395 | pasted->flags |= NAMED_OP; | |
2396 | } | |
f80e83a9 | 2397 | } |
2398 | ||
deb356cf | 2399 | /* The pasted token gets the whitespace flags and position of the |
2400 | first token, the PASTE_LEFT flag of the second token, plus the | |
2401 | PASTED flag to indicate it is the result of a paste. However, we | |
2402 | want to preserve the DIGRAPH flag. */ | |
2403 | pasted->flags &= ~(PREV_WHITE | BOL | PASTE_LEFT); | |
2404 | pasted->flags |= ((token->flags & (PREV_WHITE | BOL)) | |
2405 | | (second->flags & PASTE_LEFT) | PASTED); | |
2406 | pasted->col = token->col; | |
2407 | pasted->line = token->line; | |
2408 | ||
2409 | /* See if there is another token to be pasted onto the one we just | |
2410 | constructed. */ | |
2411 | token = pasted; | |
2412 | context = CURRENT_CONTEXT (pfile); | |
2413 | /* and loop */ | |
f80e83a9 | 2414 | } |
deb356cf | 2415 | return token; |
f80e83a9 | 2416 | } |
2417 | ||
2418 | /* Convert a token sequence to a single string token according to the | |
2419 | rules of the ISO C #-operator. */ | |
2420 | #define INIT_SIZE 200 | |
2421 | static cpp_token * | |
2422 | stringify_arg (pfile, token) | |
6060326b | 2423 | cpp_reader *pfile; |
f80e83a9 | 2424 | const cpp_token *token; |
6060326b | 2425 | { |
f80e83a9 | 2426 | cpp_token *result; |
2427 | unsigned char *main_buf; | |
2428 | unsigned int prev_value, backslash_count = 0; | |
2429 | unsigned int buf_used = 0, whitespace = 0, buf_cap = INIT_SIZE; | |
6060326b | 2430 | |
deb356cf | 2431 | push_arg_context (pfile, token); |
f80e83a9 | 2432 | prev_value = prevent_macro_expansion (pfile); |
2433 | main_buf = (unsigned char *) xmalloc (buf_cap); | |
6060326b | 2434 | |
f80e83a9 | 2435 | result = get_temp_token (pfile); |
2436 | ASSIGN_FLAGS_AND_POS (result, token); | |
2437 | ||
deb356cf | 2438 | for (; (token = _cpp_get_token (pfile))->type != CPP_EOF; ) |
6060326b | 2439 | { |
f80e83a9 | 2440 | int escape; |
2441 | unsigned char *buf; | |
2442 | unsigned int len = TOKEN_LEN (token); | |
6060326b | 2443 | |
7e842f95 | 2444 | if (token->type == CPP_PLACEMARKER) |
2445 | continue; | |
2446 | ||
f80e83a9 | 2447 | escape = (token->type == CPP_STRING || token->type == CPP_WSTRING |
2448 | || token->type == CPP_CHAR || token->type == CPP_WCHAR); | |
2449 | if (escape) | |
2450 | len *= 4 + 1; | |
2451 | ||
2452 | if (buf_used + len > buf_cap) | |
6060326b | 2453 | { |
f80e83a9 | 2454 | buf_cap = buf_used + len + INIT_SIZE; |
2455 | main_buf = xrealloc (main_buf, buf_cap); | |
2456 | } | |
6060326b | 2457 | |
f80e83a9 | 2458 | if (whitespace && (token->flags & PREV_WHITE)) |
2459 | main_buf[buf_used++] = ' '; | |
6060326b | 2460 | |
f80e83a9 | 2461 | if (escape) |
2462 | buf = (unsigned char *) xmalloc (len); | |
2463 | else | |
2464 | buf = main_buf + buf_used; | |
2465 | ||
2466 | len = spell_token (pfile, token, buf) - buf; | |
2467 | if (escape) | |
2468 | { | |
2469 | buf_used = quote_string (&main_buf[buf_used], buf, len) - main_buf; | |
2470 | free (buf); | |
2471 | } | |
2472 | else | |
2473 | buf_used += len; | |
6060326b | 2474 | |
f80e83a9 | 2475 | whitespace = 1; |
2476 | if (token->type == CPP_BACKSLASH) | |
2477 | backslash_count++; | |
2478 | else | |
2479 | backslash_count = 0; | |
2480 | } | |
6060326b | 2481 | |
f80e83a9 | 2482 | /* Ignore the final \ of invalid string literals. */ |
2483 | if (backslash_count & 1) | |
2484 | { | |
2485 | cpp_warning (pfile, "invalid string literal, ignoring final '\\'"); | |
2486 | buf_used--; | |
2487 | } | |
6060326b | 2488 | |
f80e83a9 | 2489 | result->type = CPP_STRING; |
76faa4c0 | 2490 | result->val.str.text = main_buf; |
2491 | result->val.str.len = buf_used; | |
f80e83a9 | 2492 | restore_macro_expansion (pfile, prev_value); |
2493 | return result; | |
2494 | } | |
6060326b | 2495 | |
f80e83a9 | 2496 | /* Allocate more room on the context stack of PFILE. */ |
2497 | static void | |
2498 | expand_context_stack (pfile) | |
2499 | cpp_reader *pfile; | |
2500 | { | |
2501 | pfile->context_cap += pfile->context_cap + 20; | |
2502 | pfile->contexts = (cpp_context *) | |
2503 | xrealloc (pfile->contexts, pfile->context_cap * sizeof (cpp_context)); | |
2504 | } | |
6060326b | 2505 | |
f80e83a9 | 2506 | /* Push the context of macro NODE onto the context stack. TOKEN is |
2507 | the CPP_NAME token invoking the macro. */ | |
deb356cf | 2508 | static int |
2509 | push_macro_context (pfile, token) | |
f80e83a9 | 2510 | cpp_reader *pfile; |
f80e83a9 | 2511 | const cpp_token *token; |
2512 | { | |
2513 | unsigned char orig_flags; | |
2514 | macro_args *args; | |
2515 | cpp_context *context; | |
deb356cf | 2516 | cpp_hashnode *node = token->val.node; |
6060326b | 2517 | |
f80e83a9 | 2518 | /* Token's flags may change when parsing args containing a nested |
2519 | invocation of this macro. */ | |
2520 | orig_flags = token->flags & (PREV_WHITE | BOL); | |
2521 | args = 0; | |
2522 | if (node->value.expansion->paramc >= 0) | |
6060326b | 2523 | { |
f80e83a9 | 2524 | unsigned int error, prev_nme; |
2525 | ||
2526 | /* Allocate room for the argument contexts, and parse them. */ | |
2527 | args = (macro_args *) xmalloc (sizeof (macro_args)); | |
2528 | args->ends = (unsigned int *) | |
2529 | xmalloc (node->value.expansion->paramc * sizeof (unsigned int)); | |
2530 | args->tokens = 0; | |
2531 | args->capacity = 0; | |
2532 | args->used = 0; | |
f80e83a9 | 2533 | |
2534 | prev_nme = prevent_macro_expansion (pfile); | |
2535 | pfile->args = args; | |
2536 | error = parse_args (pfile, node, args); | |
2537 | pfile->args = 0; | |
2538 | restore_macro_expansion (pfile, prev_nme); | |
2539 | if (error) | |
2540 | { | |
2541 | free_macro_args (args); | |
deb356cf | 2542 | return 1; |
f80e83a9 | 2543 | } |
e2ab8477 | 2544 | /* Set the level after the call to parse_args. */ |
2545 | args->level = pfile->cur_context; | |
6060326b | 2546 | } |
6060326b | 2547 | |
f80e83a9 | 2548 | /* Now push its context. */ |
2549 | pfile->cur_context++; | |
2550 | if (pfile->cur_context == pfile->context_cap) | |
2551 | expand_context_stack (pfile); | |
2552 | ||
2553 | context = CURRENT_CONTEXT (pfile); | |
2554 | context->u.list = node->value.expansion; | |
2555 | context->args = args; | |
2556 | context->posn = 0; | |
2557 | context->count = context->u.list->tokens_used; | |
2558 | context->level = pfile->cur_context; | |
2559 | context->flags = 0; | |
2560 | context->pushed_token = 0; | |
2561 | ||
2562 | /* Set the flags of the first token. We know there must | |
2563 | be one, empty macros are a single placemarker token. */ | |
2564 | MODIFY_FLAGS_AND_POS (&context->u.list->tokens[0], token, orig_flags); | |
2565 | ||
deb356cf | 2566 | return 0; |
6060326b | 2567 | } |
2568 | ||
f80e83a9 | 2569 | /* Push an argument to the current macro onto the context stack. |
2570 | TOKEN is the MACRO_ARG token representing the argument expansion. */ | |
deb356cf | 2571 | static void |
f80e83a9 | 2572 | push_arg_context (pfile, token) |
2573 | cpp_reader *pfile; | |
2574 | const cpp_token *token; | |
6060326b | 2575 | { |
f80e83a9 | 2576 | cpp_context *context; |
2577 | macro_args *args; | |
2578 | ||
2579 | pfile->cur_context++; | |
2580 | if (pfile->cur_context == pfile->context_cap) | |
2581 | expand_context_stack (pfile); | |
2582 | ||
2583 | context = CURRENT_CONTEXT (pfile); | |
2584 | args = context[-1].args; | |
2585 | ||
2586 | context->count = token->val.aux ? args->ends[token->val.aux - 1]: 0; | |
2587 | context->u.arg = args->tokens + context->count; | |
2588 | context->count = args->ends[token->val.aux] - context->count; | |
2589 | context->args = 0; | |
2590 | context->posn = 0; | |
2591 | context->level = args->level; | |
2592 | context->flags = CONTEXT_ARG | CONTEXT_RAW; | |
2593 | context->pushed_token = 0; | |
2594 | ||
2595 | /* Set the flags of the first token. There is one. */ | |
2596 | { | |
2597 | const cpp_token *first = context->u.arg[0]; | |
2598 | if (!first) | |
2599 | first = context->u.arg[1]; | |
6060326b | 2600 | |
f80e83a9 | 2601 | MODIFY_FLAGS_AND_POS ((cpp_token *) first, token, |
2602 | token->flags & (PREV_WHITE | BOL)); | |
2603 | } | |
6060326b | 2604 | |
f80e83a9 | 2605 | if (token->flags & PASTE_LEFT) |
2606 | context->flags |= CONTEXT_PASTEL; | |
2607 | if (pfile->paste_level) | |
2608 | context->flags |= CONTEXT_PASTER; | |
6060326b | 2609 | } |
2610 | ||
f80e83a9 | 2611 | /* "Unget" a token. It is effectively inserted in the token queue and |
2612 | will be returned by the next call to get_raw_token. */ | |
6060326b | 2613 | void |
f80e83a9 | 2614 | _cpp_push_token (pfile, token) |
6060326b | 2615 | cpp_reader *pfile; |
f80e83a9 | 2616 | const cpp_token *token; |
6060326b | 2617 | { |
f80e83a9 | 2618 | cpp_context *context = CURRENT_CONTEXT (pfile); |
7e842f95 | 2619 | |
2620 | if (context->posn > 0) | |
2621 | { | |
2622 | const cpp_token *prev; | |
2623 | if (IS_ARG_CONTEXT (context)) | |
2624 | prev = context->u.arg[context->posn - 1]; | |
2625 | else | |
2626 | prev = &context->u.list->tokens[context->posn - 1]; | |
2627 | ||
2628 | if (prev == token) | |
2629 | { | |
2630 | context->posn--; | |
2631 | return; | |
2632 | } | |
2633 | } | |
2634 | ||
f80e83a9 | 2635 | if (context->pushed_token) |
2636 | cpp_ice (pfile, "two tokens pushed in a row"); | |
2637 | if (token->type != CPP_EOF) | |
2638 | context->pushed_token = token; | |
2639 | /* Don't push back a directive's CPP_EOF, step back instead. */ | |
2640 | else if (pfile->cur_context == 0) | |
2641 | pfile->contexts[0].posn--; | |
2642 | } | |
6060326b | 2643 | |
f80e83a9 | 2644 | /* Handle a preprocessing directive. TOKEN is the CPP_HASH token |
2645 | introducing the directive. */ | |
524f0c40 | 2646 | void |
2647 | _cpp_process_directive (pfile, token) | |
f80e83a9 | 2648 | cpp_reader *pfile; |
2649 | const cpp_token *token; | |
2650 | { | |
2651 | const struct directive *d = pfile->token_list.directive; | |
2652 | int prev_nme = 0; | |
2653 | ||
2654 | /* Skip over the directive name. */ | |
2655 | if (token[1].type == CPP_NAME) | |
2656 | _cpp_get_raw_token (pfile); | |
2657 | else if (token[1].type != CPP_NUMBER) | |
7e842f95 | 2658 | cpp_ice (pfile, "directive begins with %s?!", TOKEN_NAME (token)); |
f80e83a9 | 2659 | |
f80e83a9 | 2660 | if (! (d->flags & EXPAND)) |
2661 | prev_nme = prevent_macro_expansion (pfile); | |
2662 | (void) (*d->handler) (pfile); | |
2663 | if (! (d->flags & EXPAND)) | |
2664 | restore_macro_expansion (pfile, prev_nme); | |
2665 | _cpp_skip_rest_of_line (pfile); | |
2666 | } | |
6060326b | 2667 | |
f80e83a9 | 2668 | /* The external interface to return the next token. All macro |
2669 | expansion and directive processing is handled internally, the | |
2670 | caller only ever sees the output after preprocessing. */ | |
2671 | const cpp_token * | |
2672 | cpp_get_token (pfile) | |
2673 | cpp_reader *pfile; | |
2674 | { | |
2675 | const cpp_token *token; | |
deb356cf | 2676 | /* Loop till we hit a non-directive, non-placemarker token. */ |
f80e83a9 | 2677 | for (;;) |
2678 | { | |
deb356cf | 2679 | token = _cpp_get_token (pfile); |
2680 | ||
2681 | if (token->type == CPP_PLACEMARKER) | |
2682 | continue; | |
2683 | ||
2684 | if (token->type == CPP_HASH && token->flags & BOL | |
f80e83a9 | 2685 | && pfile->token_list.directive) |
6060326b | 2686 | { |
524f0c40 | 2687 | _cpp_process_directive (pfile, token); |
f80e83a9 | 2688 | continue; |
6060326b | 2689 | } |
2690 | ||
deb356cf | 2691 | return token; |
2692 | } | |
2693 | } | |
2694 | ||
2695 | /* The internal interface to return the next token. There are two | |
2696 | differences between the internal and external interfaces: the | |
2697 | internal interface may return a PLACEMARKER token, and it does not | |
2698 | process directives. */ | |
2699 | const cpp_token * | |
2700 | _cpp_get_token (pfile) | |
2701 | cpp_reader *pfile; | |
2702 | { | |
c2cce424 | 2703 | const cpp_token *token, *old_token; |
deb356cf | 2704 | cpp_hashnode *node; |
2705 | ||
2706 | /* Loop until we hit a non-macro token. */ | |
2707 | for (;;) | |
2708 | { | |
2709 | token = get_raw_token (pfile); | |
2710 | ||
f80e83a9 | 2711 | /* Short circuit EOF. */ |
2712 | if (token->type == CPP_EOF) | |
2713 | return token; | |
deb356cf | 2714 | |
2715 | /* If we are skipping... */ | |
2716 | if (pfile->skipping) | |
6060326b | 2717 | { |
deb356cf | 2718 | /* we still have to process directives, */ |
2719 | if (pfile->token_list.directive) | |
2720 | return token; | |
2721 | ||
2722 | /* but everything else is ignored. */ | |
f80e83a9 | 2723 | _cpp_skip_rest_of_line (pfile); |
2724 | continue; | |
2725 | } | |
6060326b | 2726 | |
deb356cf | 2727 | /* If there's a potential control macro and we get here, then that |
2728 | #ifndef didn't cover the entire file and its argument shouldn't | |
2729 | be taken as a control macro. */ | |
2730 | pfile->potential_control_macro = 0; | |
6060326b | 2731 | |
db54da8f | 2732 | /* If we are rescanning preprocessed input, no macro expansion or |
2733 | token pasting may occur. */ | |
2734 | if (CPP_OPTION (pfile, preprocessed)) | |
2735 | return token; | |
2736 | ||
c2cce424 | 2737 | old_token = token; |
2738 | ||
deb356cf | 2739 | /* See if there's a token to paste with this one. */ |
2740 | if (!pfile->paste_level) | |
2741 | token = maybe_paste_with_next (pfile, token); | |
6060326b | 2742 | |
deb356cf | 2743 | /* If it isn't a macro, return it now. */ |
31674461 | 2744 | if (token->type != CPP_NAME || token->val.node->type == T_VOID) |
deb356cf | 2745 | return token; |
6060326b | 2746 | |
31674461 | 2747 | /* Is macro expansion disabled in general, or are we in the |
c2cce424 | 2748 | middle of a token paste, or was this token just pasted? |
2749 | (Note we don't check token->flags & PASTED, because that | |
2750 | counts tokens that were pasted at some point in the past, | |
2751 | we're only interested in tokens that were pasted by this call | |
2752 | to maybe_paste_with_next.) */ | |
2753 | if (pfile->no_expand_level == pfile->cur_context | |
2754 | || pfile->paste_level | |
2755 | || (token != old_token | |
2756 | && pfile->no_expand_level + 1 == pfile->cur_context)) | |
deb356cf | 2757 | return token; |
c2cce424 | 2758 | |
deb356cf | 2759 | node = token->val.node; |
2760 | if (node->type != T_MACRO) | |
2761 | return special_symbol (pfile, node, token); | |
6060326b | 2762 | |
f80e83a9 | 2763 | if (is_macro_disabled (pfile, node->value.expansion, token)) |
2764 | return token; | |
6060326b | 2765 | |
deb356cf | 2766 | if (push_macro_context (pfile, token)) |
2767 | return token; | |
2768 | /* else loop */ | |
f80e83a9 | 2769 | } |
f80e83a9 | 2770 | } |
6060326b | 2771 | |
f80e83a9 | 2772 | /* Returns the next raw token, i.e. without performing macro |
2773 | expansion. Argument contexts are automatically entered. */ | |
2774 | static const cpp_token * | |
2775 | get_raw_token (pfile) | |
2776 | cpp_reader *pfile; | |
2777 | { | |
2778 | const cpp_token *result; | |
deb356cf | 2779 | cpp_context *context; |
6060326b | 2780 | |
deb356cf | 2781 | for (;;) |
f80e83a9 | 2782 | { |
deb356cf | 2783 | context = CURRENT_CONTEXT (pfile); |
2784 | if (context->pushed_token) | |
f80e83a9 | 2785 | { |
deb356cf | 2786 | result = context->pushed_token; |
2787 | context->pushed_token = 0; | |
35677230 | 2788 | return result; /* Cannot be a CPP_MACRO_ARG */ |
deb356cf | 2789 | } |
2790 | else if (context->posn == context->count) | |
2791 | { | |
2792 | if (pop_context (pfile)) | |
2793 | return &eof_token; | |
2794 | continue; | |
2795 | } | |
35677230 | 2796 | else if (IS_ARG_CONTEXT (context)) |
deb356cf | 2797 | { |
35677230 | 2798 | result = context->u.arg[context->posn++]; |
2799 | if (result == 0) | |
6060326b | 2800 | { |
35677230 | 2801 | context->flags ^= CONTEXT_RAW; |
f80e83a9 | 2802 | result = context->u.arg[context->posn++]; |
6060326b | 2803 | } |
35677230 | 2804 | return result; /* Cannot be a CPP_MACRO_ARG */ |
f80e83a9 | 2805 | } |
6060326b | 2806 | |
35677230 | 2807 | result = &context->u.list->tokens[context->posn++]; |
2808 | ||
deb356cf | 2809 | if (result->type != CPP_MACRO_ARG) |
2810 | return result; | |
2811 | ||
2812 | if (result->flags & STRINGIFY_ARG) | |
2813 | return stringify_arg (pfile, result); | |
2814 | ||
2815 | push_arg_context (pfile, result); | |
2816 | } | |
f80e83a9 | 2817 | } |
6060326b | 2818 | |
f80e83a9 | 2819 | /* Internal interface to get the token without macro expanding. */ |
2820 | const cpp_token * | |
2821 | _cpp_get_raw_token (pfile) | |
2822 | cpp_reader *pfile; | |
2823 | { | |
2824 | int prev_nme = prevent_macro_expansion (pfile); | |
deb356cf | 2825 | const cpp_token *result = _cpp_get_token (pfile); |
f80e83a9 | 2826 | restore_macro_expansion (pfile, prev_nme); |
2827 | return result; | |
2828 | } | |
6060326b | 2829 | |
f80e83a9 | 2830 | /* A thin wrapper to lex_line. CLEAR is non-zero if the current token |
2831 | list should be overwritten, or zero if we need to append | |
2832 | (typically, if we are within the arguments to a macro, or looking | |
2833 | for the '(' to start a function-like macro invocation). */ | |
2834 | static int | |
2835 | lex_next (pfile, clear) | |
2836 | cpp_reader *pfile; | |
2837 | int clear; | |
2838 | { | |
2839 | cpp_toklist *list = &pfile->token_list; | |
2840 | const cpp_token *old_list = list->tokens; | |
2841 | unsigned int old_used = list->tokens_used; | |
6060326b | 2842 | |
f80e83a9 | 2843 | if (clear) |
6060326b | 2844 | { |
f80e83a9 | 2845 | /* Release all temporary tokens. */ |
2846 | _cpp_clear_toklist (list); | |
2847 | pfile->contexts[0].posn = 0; | |
2848 | if (pfile->temp_used) | |
2849 | release_temp_tokens (pfile); | |
6060326b | 2850 | } |
f80e83a9 | 2851 | lex_line (pfile, list); |
2852 | pfile->contexts[0].count = list->tokens_used; | |
6060326b | 2853 | |
f80e83a9 | 2854 | if (!clear && pfile->args) |
6060326b | 2855 | { |
f80e83a9 | 2856 | /* Fix up argument token pointers. */ |
2857 | if (old_list != list->tokens) | |
2858 | { | |
2859 | unsigned int i; | |
2860 | ||
2861 | for (i = 0; i < pfile->args->used; i++) | |
2862 | { | |
2863 | const cpp_token *token = pfile->args->tokens[i]; | |
2864 | if (token >= old_list && token < old_list + old_used) | |
2865 | pfile->args->tokens[i] = (const cpp_token *) | |
2866 | ((char *) token + ((char *) list->tokens - (char *) old_list)); | |
2867 | } | |
2868 | } | |
2869 | ||
2870 | /* 6.10.3 paragraph 11: If there are sequences of preprocessing | |
2871 | tokens within the list of arguments that would otherwise act as | |
2872 | preprocessing directives, the behavior is undefined. | |
2873 | ||
2874 | This implementation will report a hard error and treat the | |
2875 | 'sequence of preprocessing tokens' as part of the macro argument, | |
2876 | not a directive. | |
2877 | ||
2878 | Note if pfile->args == 0, we're OK since we're only inside a | |
2879 | macro argument after a '('. */ | |
2880 | if (list->directive) | |
2881 | { | |
2882 | cpp_error_with_line (pfile, list->tokens[old_used].line, | |
2883 | list->tokens[old_used].col, | |
2884 | "#%s may not be used inside a macro argument", | |
2885 | list->directive->name); | |
78719282 | 2886 | return 1; |
f80e83a9 | 2887 | } |
6060326b | 2888 | } |
2889 | ||
f80e83a9 | 2890 | return 0; |
6060326b | 2891 | } |
2892 | ||
deb356cf | 2893 | /* Pops a context off the context stack. If we're at the bottom, lexes |
2894 | the next logical line. Returns EOF if we're at the end of the | |
e0a859f1 | 2895 | argument list to the # operator, or we should not "overflow" |
f80e83a9 | 2896 | into the rest of the file (e.g. 6.10.3.1.1). */ |
2897 | static int | |
deb356cf | 2898 | pop_context (pfile) |
f80e83a9 | 2899 | cpp_reader *pfile; |
2900 | { | |
2901 | cpp_context *context; | |
89b05ef6 | 2902 | |
f80e83a9 | 2903 | if (pfile->cur_context == 0) |
deb356cf | 2904 | { |
2905 | /* If we are currently processing a directive, do not advance. 6.10 | |
2906 | paragraph 2: A new-line character ends the directive even if it | |
2907 | occurs within what would otherwise be an invocation of a | |
2908 | function-like macro. */ | |
2909 | if (pfile->token_list.directive) | |
2910 | return 1; | |
2911 | ||
2912 | return lex_next (pfile, pfile->no_expand_level == UINT_MAX); | |
2913 | } | |
f80e83a9 | 2914 | |
2915 | /* Argument contexts, when parsing args or handling # operator | |
2916 | return CPP_EOF at the end. */ | |
2917 | context = CURRENT_CONTEXT (pfile); | |
2918 | if (IS_ARG_CONTEXT (context) && pfile->cur_context == pfile->no_expand_level) | |
2919 | return 1; | |
2920 | ||
2921 | /* Free resources when leaving macro contexts. */ | |
2922 | if (context->args) | |
2923 | free_macro_args (context->args); | |
2924 | ||
2925 | if (pfile->cur_context == pfile->no_expand_level) | |
2926 | pfile->no_expand_level--; | |
2927 | pfile->cur_context--; | |
2928 | ||
2929 | return 0; | |
2930 | } | |
2931 | ||
f80e83a9 | 2932 | /* Turn off macro expansion at the current context level. */ |
2933 | static unsigned int | |
2934 | prevent_macro_expansion (pfile) | |
2935 | cpp_reader *pfile; | |
2936 | { | |
2937 | unsigned int prev_value = pfile->no_expand_level; | |
2938 | pfile->no_expand_level = pfile->cur_context; | |
2939 | return prev_value; | |
2940 | } | |
2941 | ||
2942 | /* Restore macro expansion to its previous state. */ | |
2943 | static void | |
2944 | restore_macro_expansion (pfile, prev_value) | |
2945 | cpp_reader *pfile; | |
2946 | unsigned int prev_value; | |
2947 | { | |
2948 | pfile->no_expand_level = prev_value; | |
2949 | } | |
2950 | ||
2951 | /* Used by cpperror.c to obtain the correct line and column to report | |
2952 | in a diagnostic. */ | |
2953 | unsigned int | |
2954 | _cpp_get_line (pfile, pcol) | |
2955 | cpp_reader *pfile; | |
2956 | unsigned int *pcol; | |
2957 | { | |
2958 | unsigned int index; | |
2959 | const cpp_token *cur_token; | |
2960 | ||
338fa5f7 | 2961 | if (pfile->state.in_lex_line) |
f80e83a9 | 2962 | index = pfile->token_list.tokens_used; |
2963 | else | |
c2cce424 | 2964 | { |
338fa5f7 | 2965 | index = pfile->contexts[0].posn; |
2966 | ||
2967 | if (index == 0) | |
2968 | { | |
2969 | if (pcol) | |
2970 | *pcol = 0; | |
2971 | return 0; | |
2972 | } | |
2973 | index--; | |
c2cce424 | 2974 | } |
2975 | ||
338fa5f7 | 2976 | cur_token = &pfile->token_list.tokens[index]; |
f80e83a9 | 2977 | if (pcol) |
2978 | *pcol = cur_token->col; | |
2979 | return cur_token->line; | |
2980 | } | |
2981 | ||
2982 | #define DSC(str) (const U_CHAR *)str, sizeof str - 1 | |
2983 | static const char * const monthnames[] = | |
2984 | { | |
2985 | "Jan", "Feb", "Mar", "Apr", "May", "Jun", | |
2986 | "Jul", "Aug", "Sep", "Oct", "Nov", "Dec", | |
2987 | }; | |
2988 | ||
2989 | /* Handle builtin macros like __FILE__. */ | |
2990 | static const cpp_token * | |
2991 | special_symbol (pfile, node, token) | |
2992 | cpp_reader *pfile; | |
2993 | cpp_hashnode *node; | |
d2efb5ed | 2994 | const cpp_token *token; |
89b05ef6 | 2995 | { |
f80e83a9 | 2996 | cpp_token *result; |
2997 | cpp_buffer *ip; | |
89b05ef6 | 2998 | |
f80e83a9 | 2999 | switch (node->type) |
89b05ef6 | 3000 | { |
f80e83a9 | 3001 | case T_FILE: |
3002 | case T_BASE_FILE: | |
89b05ef6 | 3003 | { |
f80e83a9 | 3004 | const char *file; |
89b05ef6 | 3005 | |
f80e83a9 | 3006 | ip = CPP_BUFFER (pfile); |
3007 | if (ip == 0) | |
3008 | file = ""; | |
89b05ef6 | 3009 | else |
f80e83a9 | 3010 | { |
3011 | if (node->type == T_BASE_FILE) | |
3012 | while (CPP_PREV_BUFFER (ip) != NULL) | |
3013 | ip = CPP_PREV_BUFFER (ip); | |
3014 | ||
3015 | file = ip->nominal_fname; | |
3016 | } | |
3017 | result = make_string_token (get_temp_token (pfile), (U_CHAR *) file, | |
3018 | strlen (file)); | |
3019 | } | |
3020 | break; | |
89b05ef6 | 3021 | |
f80e83a9 | 3022 | case T_INCLUDE_LEVEL: |
241e762e | 3023 | /* pfile->include_depth counts the primary source as level 1, |
3024 | but historically __INCLUDE_DEPTH__ has called the primary | |
3025 | source level 0. */ | |
3026 | result = alloc_number_token (pfile, pfile->include_depth - 1); | |
89b05ef6 | 3027 | break; |
3028 | ||
f80e83a9 | 3029 | case T_SPECLINE: |
3030 | /* If __LINE__ is embedded in a macro, it must expand to the | |
3031 | line of the macro's invocation, not its definition. | |
3032 | Otherwise things like assert() will not work properly. */ | |
3033 | result = alloc_number_token (pfile, _cpp_get_line (pfile, NULL)); | |
de3d9756 | 3034 | break; |
3035 | ||
f80e83a9 | 3036 | case T_STDC: |
89b05ef6 | 3037 | { |
f80e83a9 | 3038 | int stdc = 1; |
89b05ef6 | 3039 | |
f80e83a9 | 3040 | #ifdef STDC_0_IN_SYSTEM_HEADERS |
3041 | if (CPP_IN_SYSTEM_HEADER (pfile) | |
76faa4c0 | 3042 | && pfile->spec_nodes->n__STRICT_ANSI__->type == T_VOID) |
f80e83a9 | 3043 | stdc = 0; |
3044 | #endif | |
3045 | result = alloc_number_token (pfile, stdc); | |
89b05ef6 | 3046 | } |
3047 | break; | |
3048 | ||
f80e83a9 | 3049 | case T_DATE: |
3050 | case T_TIME: | |
3051 | if (pfile->date == 0) | |
3052 | { | |
3053 | /* Allocate __DATE__ and __TIME__ from permanent storage, | |
3054 | and save them in pfile so we don't have to do this again. | |
3055 | We don't generate these strings at init time because | |
3056 | time() and localtime() are very slow on some systems. */ | |
3057 | time_t tt = time (NULL); | |
3058 | struct tm *tb = localtime (&tt); | |
3059 | ||
3060 | pfile->date = make_string_token | |
3061 | ((cpp_token *) xmalloc (sizeof (cpp_token)), DSC("Oct 11 1347")); | |
3062 | pfile->time = make_string_token | |
3063 | ((cpp_token *) xmalloc (sizeof (cpp_token)), DSC("12:34:56")); | |
3064 | ||
76faa4c0 | 3065 | sprintf ((char *) pfile->date->val.str.text, "%s %2d %4d", |
f80e83a9 | 3066 | monthnames[tb->tm_mon], tb->tm_mday, tb->tm_year + 1900); |
76faa4c0 | 3067 | sprintf ((char *) pfile->time->val.str.text, "%02d:%02d:%02d", |
f80e83a9 | 3068 | tb->tm_hour, tb->tm_min, tb->tm_sec); |
3069 | } | |
3070 | result = node->type == T_DATE ? pfile->date: pfile->time; | |
89b05ef6 | 3071 | break; |
3072 | ||
f80e83a9 | 3073 | case T_POISON: |
76faa4c0 | 3074 | cpp_error (pfile, "attempt to use poisoned \"%s\"", node->name); |
f80e83a9 | 3075 | return token; |
3076 | ||
3077 | default: | |
3078 | cpp_ice (pfile, "invalid special hash type"); | |
3079 | return token; | |
89b05ef6 | 3080 | } |
3081 | ||
f80e83a9 | 3082 | ASSIGN_FLAGS_AND_POS (result, token); |
3083 | return result; | |
3084 | } | |
3085 | #undef DSC | |
3086 | ||
b3954366 | 3087 | /* Allocate pfile->input_buffer, and initialize _cpp_trigraph_map[] |
f80e83a9 | 3088 | if it hasn't happened already. */ |
3089 | ||
3090 | void | |
3091 | _cpp_init_input_buffer (pfile) | |
3092 | cpp_reader *pfile; | |
3093 | { | |
deb356cf | 3094 | cpp_context *base; |
3095 | ||
f669338a | 3096 | _cpp_init_toklist (&pfile->token_list, 0); |
deb356cf | 3097 | pfile->no_expand_level = UINT_MAX; |
f80e83a9 | 3098 | pfile->context_cap = 20; |
f80e83a9 | 3099 | pfile->cur_context = 0; |
f80e83a9 | 3100 | |
deb356cf | 3101 | pfile->contexts = (cpp_context *) |
3102 | xmalloc (pfile->context_cap * sizeof (cpp_context)); | |
f80e83a9 | 3103 | |
deb356cf | 3104 | /* Clear the base context. */ |
3105 | base = &pfile->contexts[0]; | |
3106 | base->u.list = &pfile->token_list; | |
3107 | base->posn = 0; | |
3108 | base->count = 0; | |
3109 | base->args = 0; | |
3110 | base->level = 0; | |
3111 | base->flags = 0; | |
3112 | base->pushed_token = 0; | |
f80e83a9 | 3113 | } |
3114 | ||
3115 | /* Moves to the end of the directive line, popping contexts as | |
3116 | necessary. */ | |
3117 | void | |
3118 | _cpp_skip_rest_of_line (pfile) | |
3119 | cpp_reader *pfile; | |
3120 | { | |
deb356cf | 3121 | /* Discard all stacked contexts. */ |
3122 | int i; | |
3123 | for (i = pfile->cur_context; i > 0; i--) | |
3124 | if (pfile->contexts[i].args) | |
3125 | free_macro_args (pfile->contexts[i].args); | |
3126 | ||
3127 | if (pfile->no_expand_level <= pfile->cur_context) | |
3128 | pfile->no_expand_level = 0; | |
3129 | pfile->cur_context = 0; | |
f80e83a9 | 3130 | |
deb356cf | 3131 | /* Clear the base context, and clear the directive pointer so that |
3132 | get_raw_token will advance to the next line. */ | |
3133 | pfile->contexts[0].count = 0; | |
3134 | pfile->contexts[0].posn = 0; | |
f80e83a9 | 3135 | pfile->token_list.directive = 0; |
6060326b | 3136 | } |
3137 | ||
f80e83a9 | 3138 | /* Directive handler wrapper used by the command line option |
3139 | processor. */ | |
3140 | void | |
a336277c | 3141 | _cpp_run_directive (pfile, dir, buf, count, name) |
f80e83a9 | 3142 | cpp_reader *pfile; |
3143 | const struct directive *dir; | |
3144 | const char *buf; | |
3145 | size_t count; | |
a336277c | 3146 | const char *name; |
f80e83a9 | 3147 | { |
3148 | if (cpp_push_buffer (pfile, (const U_CHAR *)buf, count) != NULL) | |
3149 | { | |
3150 | unsigned int prev_lvl = 0; | |
deb356cf | 3151 | |
a336277c | 3152 | if (name) |
3153 | CPP_BUFFER (pfile)->nominal_fname = name; | |
3154 | else | |
3155 | CPP_BUFFER (pfile)->nominal_fname = _("<command line>"); | |
3156 | CPP_BUFFER (pfile)->lineno = (unsigned int)-1; | |
3157 | ||
deb356cf | 3158 | /* Scan the line now, else prevent_macro_expansion won't work. */ |
3159 | lex_next (pfile, 1); | |
f80e83a9 | 3160 | if (! (dir->flags & EXPAND)) |
3161 | prev_lvl = prevent_macro_expansion (pfile); | |
3162 | ||
3163 | (void) (*dir->handler) (pfile); | |
3164 | ||
3165 | if (! (dir->flags & EXPAND)) | |
3166 | restore_macro_expansion (pfile, prev_lvl); | |
3167 | ||
3168 | _cpp_skip_rest_of_line (pfile); | |
3169 | cpp_pop_buffer (pfile); | |
3170 | } | |
3171 | } |