]>
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 Check line numbers assigned to all errors. |
f80e83a9 | 28 | o Distinguish integers, floats, and 'other' pp-numbers. |
29 | o Store ints and char constants as binary values. | |
30 | o New command-line assertion syntax. | |
f80e83a9 | 31 | o Work towards functions in cpperror.c taking a message level parameter. |
32 | If we do this, merge the common code of do_warning and do_error. | |
33 | o Comment all functions, and describe macro expansion algorithm. | |
34 | o Move as much out of header files as possible. | |
35 | o Remove single quote pairs `', and some '', from diagnostics. | |
36 | o Correct pastability test for CPP_NAME and CPP_NUMBER. | |
37 | ||
38 | */ | |
39 | ||
0578f103 | 40 | #include "config.h" |
41 | #include "system.h" | |
42 | #include "intl.h" | |
43 | #include "cpplib.h" | |
44 | #include "cpphash.h" | |
f80e83a9 | 45 | #include "symcat.h" |
0578f103 | 46 | |
338fa5f7 | 47 | static const cpp_token placemarker_token = {0, 0, CPP_PLACEMARKER, |
48 | 0 UNION_INIT_ZERO}; | |
241e762e | 49 | static const cpp_token eof_token = {0, 0, CPP_EOF, 0 UNION_INIT_ZERO}; |
50 | ||
51 | /* Flags for cpp_context. */ | |
52 | #define CONTEXT_PASTEL (1 << 0) /* An argument context on LHS of ##. */ | |
53 | #define CONTEXT_PASTER (1 << 1) /* An argument context on RHS of ##. */ | |
54 | #define CONTEXT_RAW (1 << 2) /* If argument tokens already expanded. */ | |
55 | #define CONTEXT_ARG (1 << 3) /* If an argument context. */ | |
56 | ||
57 | typedef struct cpp_context cpp_context; | |
58 | struct cpp_context | |
59 | { | |
60 | union | |
61 | { | |
62 | const cpp_toklist *list; /* Used for macro contexts only. */ | |
63 | const cpp_token **arg; /* Used for arg contexts only. */ | |
64 | } u; | |
65 | ||
66 | /* Pushed token to be returned by next call to get_raw_token. */ | |
67 | const cpp_token *pushed_token; | |
68 | ||
fdf3a98f | 69 | struct macro_args *args; /* The arguments for a function-like |
70 | macro. NULL otherwise. */ | |
241e762e | 71 | unsigned short posn; /* Current posn, index into u. */ |
72 | unsigned short count; /* No. of tokens in u. */ | |
73 | unsigned short level; | |
74 | unsigned char flags; | |
75 | }; | |
76 | ||
77 | typedef struct macro_args macro_args; | |
78 | struct macro_args | |
79 | { | |
80 | unsigned int *ends; | |
81 | const cpp_token **tokens; | |
82 | unsigned int capacity; | |
83 | unsigned int used; | |
84 | unsigned short level; | |
85 | }; | |
86 | ||
87 | static const cpp_token *get_raw_token PARAMS ((cpp_reader *)); | |
88 | static const cpp_token *parse_arg PARAMS ((cpp_reader *, int, unsigned int, | |
89 | macro_args *, unsigned int *)); | |
90 | static int parse_args PARAMS ((cpp_reader *, cpp_hashnode *, macro_args *)); | |
91 | static void save_token PARAMS ((macro_args *, const cpp_token *)); | |
92 | static int pop_context PARAMS ((cpp_reader *)); | |
93 | static int push_macro_context PARAMS ((cpp_reader *, const cpp_token *)); | |
94 | static void push_arg_context PARAMS ((cpp_reader *, const cpp_token *)); | |
95 | static void free_macro_args PARAMS ((macro_args *)); | |
6cae2504 | 96 | static void dump_param_spelling PARAMS ((FILE *, const cpp_toklist *, |
50fd6b48 | 97 | unsigned int)); |
f80e83a9 | 98 | static void output_line_command PARAMS ((cpp_reader *, cpp_printer *, |
2c63d6c8 | 99 | unsigned int)); |
e2f9a79f | 100 | |
338fa5f7 | 101 | static cppchar_t handle_newline PARAMS ((cpp_buffer *, cppchar_t)); |
102 | static cppchar_t skip_escaped_newlines PARAMS ((cpp_buffer *, cppchar_t)); | |
103 | static cppchar_t get_effective_char PARAMS ((cpp_buffer *)); | |
104 | ||
f80e83a9 | 105 | static int skip_block_comment PARAMS ((cpp_reader *)); |
338fa5f7 | 106 | static int skip_line_comment PARAMS ((cpp_buffer *)); |
107 | static void adjust_column PARAMS ((cpp_reader *)); | |
108 | static void skip_whitespace PARAMS ((cpp_reader *, cppchar_t)); | |
109 | static cpp_hashnode *parse_identifier PARAMS ((cpp_reader *, cppchar_t)); | |
110 | static void parse_number PARAMS ((cpp_reader *, cpp_string *, cppchar_t)); | |
111 | static void parse_string PARAMS ((cpp_reader *, cpp_token *, cppchar_t)); | |
112 | static void unterminated PARAMS ((cpp_reader *, unsigned int, int)); | |
113 | static int trigraph_ok PARAMS ((cpp_reader *, cppchar_t)); | |
114 | static void save_comment PARAMS ((cpp_reader *, cpp_token *, const U_CHAR *)); | |
f80e83a9 | 115 | static void lex_line PARAMS ((cpp_reader *, cpp_toklist *)); |
338fa5f7 | 116 | static void check_long_token PARAMS ((cpp_buffer *, |
117 | cpp_token *, | |
118 | cppchar_t, | |
119 | enum cpp_ttype)); | |
120 | static void lex_token PARAMS ((cpp_reader *, cpp_token *)); | |
f80e83a9 | 121 | static int lex_next PARAMS ((cpp_reader *, int)); |
338fa5f7 | 122 | |
123 | static void process_directive PARAMS ((cpp_reader *, const cpp_token *)); | |
f80e83a9 | 124 | static int is_macro_disabled PARAMS ((cpp_reader *, const cpp_toklist *, |
125 | const cpp_token *)); | |
6bc03ce3 | 126 | |
f80e83a9 | 127 | static cpp_token *stringify_arg PARAMS ((cpp_reader *, const cpp_token *)); |
128 | static void expand_context_stack PARAMS ((cpp_reader *)); | |
d2efb5ed | 129 | static unsigned char * spell_token PARAMS ((cpp_reader *, const cpp_token *, |
f80e83a9 | 130 | unsigned char *)); |
6cae2504 | 131 | static void output_token PARAMS ((cpp_reader *, FILE *, const cpp_token *, |
132 | const cpp_token *, int)); | |
6bc03ce3 | 133 | typedef unsigned int (* speller) PARAMS ((unsigned char *, cpp_toklist *, |
134 | cpp_token *)); | |
f80e83a9 | 135 | static cpp_token *make_string_token PARAMS ((cpp_token *, const U_CHAR *, |
136 | unsigned int)); | |
137 | static cpp_token *alloc_number_token PARAMS ((cpp_reader *, int number)); | |
138 | static const cpp_token *special_symbol PARAMS ((cpp_reader *, cpp_hashnode *, | |
139 | const cpp_token *)); | |
140 | static cpp_token *duplicate_token PARAMS ((cpp_reader *, const cpp_token *)); | |
141 | static const cpp_token *maybe_paste_with_next PARAMS ((cpp_reader *, | |
142 | const cpp_token *)); | |
143 | static enum cpp_ttype can_paste PARAMS ((cpp_reader *, const cpp_token *, | |
144 | const cpp_token *, int *)); | |
145 | static unsigned int prevent_macro_expansion PARAMS ((cpp_reader *)); | |
146 | static void restore_macro_expansion PARAMS ((cpp_reader *, unsigned int)); | |
147 | static cpp_token *get_temp_token PARAMS ((cpp_reader *)); | |
148 | static void release_temp_tokens PARAMS ((cpp_reader *)); | |
149 | static U_CHAR * quote_string PARAMS ((U_CHAR *, const U_CHAR *, unsigned int)); | |
150 | static void process_directive PARAMS ((cpp_reader *, const cpp_token *)); | |
6bc03ce3 | 151 | |
f80e83a9 | 152 | #define VALID_SIGN(c, prevc) \ |
153 | (((c) == '+' || (c) == '-') && \ | |
154 | ((prevc) == 'e' || (prevc) == 'E' \ | |
155 | || (((prevc) == 'p' || (prevc) == 'P') && !CPP_OPTION (pfile, c89)))) | |
156 | ||
e916a356 | 157 | /* An upper bound on the number of bytes needed to spell a token, |
158 | including preceding whitespace. */ | |
6cae2504 | 159 | static inline size_t TOKEN_LEN PARAMS ((const cpp_token *)); |
160 | static inline size_t | |
161 | TOKEN_LEN (token) | |
162 | const cpp_token *token; | |
163 | { | |
164 | size_t len; | |
165 | ||
166 | switch (TOKEN_SPELL (token)) | |
167 | { | |
168 | default: len = 0; break; | |
169 | case SPELL_STRING: len = token->val.str.len; break; | |
170 | case SPELL_IDENT: len = token->val.node->length; break; | |
171 | } | |
172 | return len + 5; | |
173 | } | |
e916a356 | 174 | |
241e762e | 175 | #define IS_ARG_CONTEXT(c) ((c)->flags & CONTEXT_ARG) |
176 | #define CURRENT_CONTEXT(pfile) ((pfile)->contexts + (pfile)->cur_context) | |
e0a859f1 | 177 | #define ON_REST_ARG(c) \ |
58fe658a | 178 | (((c)->u.list->flags & VAR_ARGS) \ |
179 | && (c)->u.list->tokens[(c)->posn - 1].val.aux \ | |
180 | == (unsigned int) ((c)->u.list->paramc - 1)) | |
241e762e | 181 | |
182 | #define ASSIGN_FLAGS_AND_POS(d, s) \ | |
183 | do {(d)->flags = (s)->flags & (PREV_WHITE | BOL | PASTE_LEFT); \ | |
184 | if ((d)->flags & BOL) {(d)->col = (s)->col; (d)->line = (s)->line;} \ | |
185 | } while (0) | |
186 | ||
187 | /* f is flags, just consisting of PREV_WHITE | BOL. */ | |
188 | #define MODIFY_FLAGS_AND_POS(d, s, f) \ | |
189 | do {(d)->flags &= ~(PREV_WHITE | BOL); (d)->flags |= (f); \ | |
190 | if ((f) & BOL) {(d)->col = (s)->col; (d)->line = (s)->line;} \ | |
191 | } while (0) | |
192 | ||
7e842f95 | 193 | #define OP(e, s) { SPELL_OPERATOR, U s }, |
194 | #define TK(e, s) { s, U STRINGX (e) }, | |
6bc03ce3 | 195 | |
f80e83a9 | 196 | const struct token_spelling |
7e842f95 | 197 | _cpp_token_spellings [N_TTYPES] = {TTYPE_TABLE }; |
f80e83a9 | 198 | |
7e842f95 | 199 | #undef OP |
200 | #undef TK | |
6bc03ce3 | 201 | |
e2f9a79f | 202 | /* Notify the compiler proper that the current line number has jumped, |
203 | or the current file name has changed. */ | |
204 | ||
205 | static void | |
50fd6b48 | 206 | output_line_command (pfile, print, line) |
0578f103 | 207 | cpp_reader *pfile; |
e2f9a79f | 208 | cpp_printer *print; |
50fd6b48 | 209 | unsigned int line; |
0578f103 | 210 | { |
f80e83a9 | 211 | cpp_buffer *ip = CPP_BUFFER (pfile); |
e2f9a79f | 212 | |
f80e83a9 | 213 | if (line == 0) |
214 | return; | |
215 | ||
216 | /* End the previous line of text. */ | |
217 | if (pfile->need_newline) | |
6cae2504 | 218 | { |
219 | putc ('\n', print->outf); | |
220 | print->lineno++; | |
221 | } | |
f80e83a9 | 222 | pfile->need_newline = 0; |
223 | ||
e2f9a79f | 224 | if (CPP_OPTION (pfile, no_line_commands)) |
225 | return; | |
226 | ||
e2f9a79f | 227 | /* If the current file has not changed, we can output a few newlines |
228 | instead if we want to increase the line number by a small amount. | |
229 | We cannot do this if print->lineno is zero, because that means we | |
230 | haven't output any line commands yet. (The very first line | |
6cae2504 | 231 | command output is a `same_file' command.) |
232 | ||
233 | 'nominal_fname' values are unique, so they can be compared by | |
234 | comparing pointers. */ | |
235 | if (ip->nominal_fname == print->last_fname && print->lineno > 0 | |
e2f9a79f | 236 | && line >= print->lineno && line < print->lineno + 8) |
0578f103 | 237 | { |
e2f9a79f | 238 | while (line > print->lineno) |
0578f103 | 239 | { |
e2f9a79f | 240 | putc ('\n', print->outf); |
241 | print->lineno++; | |
0578f103 | 242 | } |
e2f9a79f | 243 | return; |
0578f103 | 244 | } |
e2f9a79f | 245 | |
6cae2504 | 246 | fprintf (print->outf, "# %u \"%s\"%s\n", line, ip->nominal_fname, |
247 | cpp_syshdr_flags (pfile, ip)); | |
248 | ||
249 | print->last_fname = ip->nominal_fname; | |
e2f9a79f | 250 | print->lineno = line; |
251 | } | |
252 | ||
6cae2504 | 253 | /* Like fprintf, but writes to a printer object. You should be sure |
254 | always to generate a complete line when you use this function. */ | |
e2f9a79f | 255 | void |
6cae2504 | 256 | cpp_printf VPARAMS ((cpp_reader *pfile, cpp_printer *print, |
257 | const char *fmt, ...)) | |
258 | { | |
259 | va_list ap; | |
260 | #ifndef ANSI_PROTOTYPES | |
261 | cpp_reader *pfile; | |
262 | cpp_printer *print; | |
263 | const char *fmt; | |
264 | #endif | |
0578f103 | 265 | |
6cae2504 | 266 | VA_START (ap, fmt); |
267 | ||
268 | #ifndef ANSI_PROTOTYPES | |
269 | pfile = va_arg (ap, cpp_reader *); | |
270 | print = va_arg (ap, cpp_printer *); | |
271 | fmt = va_arg (ap, const char *); | |
272 | #endif | |
273 | ||
274 | /* End the previous line of text. */ | |
275 | if (pfile->need_newline) | |
bad6bf53 | 276 | { |
277 | putc ('\n', print->outf); | |
278 | print->lineno++; | |
279 | } | |
6cae2504 | 280 | pfile->need_newline = 0; |
281 | ||
282 | vfprintf (print->outf, fmt, ap); | |
283 | va_end (ap); | |
0578f103 | 284 | } |
285 | ||
d1678bc6 | 286 | /* Scan until CPP_BUFFER (PFILE) is exhausted, discarding output. */ |
e2f9a79f | 287 | |
288 | void | |
289 | cpp_scan_buffer_nooutput (pfile) | |
290 | cpp_reader *pfile; | |
291 | { | |
f80e83a9 | 292 | cpp_buffer *stop = CPP_PREV_BUFFER (CPP_BUFFER (pfile)); |
dfa5c7fa | 293 | const cpp_token *token; |
f80e83a9 | 294 | |
dfa5c7fa | 295 | /* In no-output mode, we can ignore everything but directives. */ |
e2f9a79f | 296 | for (;;) |
297 | { | |
dfa5c7fa | 298 | token = _cpp_get_token (pfile); |
299 | ||
f80e83a9 | 300 | if (token->type == CPP_EOF) |
301 | { | |
302 | cpp_pop_buffer (pfile); | |
303 | if (CPP_BUFFER (pfile) == stop) | |
304 | break; | |
305 | } | |
dfa5c7fa | 306 | |
307 | if (token->type == CPP_HASH && token->flags & BOL | |
308 | && pfile->token_list.directive) | |
309 | { | |
310 | process_directive (pfile, token); | |
311 | continue; | |
312 | } | |
313 | ||
f80e83a9 | 314 | _cpp_skip_rest_of_line (pfile); |
e2f9a79f | 315 | } |
e2f9a79f | 316 | } |
317 | ||
d1678bc6 | 318 | /* Scan until CPP_BUFFER (pfile) is exhausted, writing output to PRINT. */ |
e2f9a79f | 319 | void |
320 | cpp_scan_buffer (pfile, print) | |
321 | cpp_reader *pfile; | |
322 | cpp_printer *print; | |
323 | { | |
d1678bc6 | 324 | cpp_buffer *stop = CPP_PREV_BUFFER (CPP_BUFFER (pfile)); |
f80e83a9 | 325 | const cpp_token *token, *prev = 0; |
e2f9a79f | 326 | |
327 | for (;;) | |
328 | { | |
dfa5c7fa | 329 | token = _cpp_get_token (pfile); |
f80e83a9 | 330 | if (token->type == CPP_EOF) |
e2f9a79f | 331 | { |
f80e83a9 | 332 | cpp_pop_buffer (pfile); |
e9d42dd4 | 333 | |
f80e83a9 | 334 | if (CPP_BUFFER (pfile) == stop) |
e2f9a79f | 335 | return; |
dfa5c7fa | 336 | |
f80e83a9 | 337 | prev = 0; |
338 | continue; | |
339 | } | |
340 | ||
341 | if (token->flags & BOL) | |
342 | { | |
338fa5f7 | 343 | output_line_command (pfile, print, token->line); |
bad6bf53 | 344 | prev = 0; |
345 | ||
dfa5c7fa | 346 | if (token->type == CPP_HASH && pfile->token_list.directive) |
347 | { | |
348 | process_directive (pfile, token); | |
349 | continue; | |
350 | } | |
e2f9a79f | 351 | } |
f80e83a9 | 352 | |
dfa5c7fa | 353 | if (token->type != CPP_PLACEMARKER) |
241e762e | 354 | { |
6cae2504 | 355 | output_token (pfile, print->outf, token, prev, 1); |
356 | pfile->need_newline = 1; | |
241e762e | 357 | } |
358 | ||
241e762e | 359 | prev = token; |
360 | } | |
241e762e | 361 | } |
362 | ||
f80e83a9 | 363 | /* Helper routine used by parse_include, which can't see spell_token. |
364 | Reinterpret the current line as an h-char-sequence (< ... >); we are | |
365 | looking at the first token after the <. */ | |
366 | const cpp_token * | |
367 | _cpp_glue_header_name (pfile) | |
0578f103 | 368 | cpp_reader *pfile; |
369 | { | |
f80e83a9 | 370 | const cpp_token *t; |
371 | cpp_token *hdr; | |
6cae2504 | 372 | U_CHAR *buf, *p; |
373 | size_t len, avail; | |
374 | ||
375 | avail = 40; | |
376 | len = 0; | |
377 | buf = xmalloc (avail); | |
f80e83a9 | 378 | |
379 | for (;;) | |
380 | { | |
deb356cf | 381 | t = _cpp_get_token (pfile); |
f80e83a9 | 382 | if (t->type == CPP_GREATER || t->type == CPP_EOF) |
383 | break; | |
384 | ||
6cae2504 | 385 | if (len + TOKEN_LEN (t) > avail) |
386 | { | |
387 | avail = len + TOKEN_LEN (t) + 40; | |
388 | buf = xrealloc (buf, avail); | |
389 | } | |
390 | ||
f80e83a9 | 391 | if (t->flags & PREV_WHITE) |
6cae2504 | 392 | buf[len++] = ' '; |
393 | ||
394 | p = spell_token (pfile, t, buf + len); | |
395 | len = (size_t) (p - buf); /* p known >= buf */ | |
f80e83a9 | 396 | } |
397 | ||
398 | if (t->type == CPP_EOF) | |
399 | cpp_error (pfile, "missing terminating > character"); | |
0578f103 | 400 | |
6cae2504 | 401 | buf = xrealloc (buf, len); |
f80e83a9 | 402 | |
403 | hdr = get_temp_token (pfile); | |
404 | hdr->type = CPP_HEADER_NAME; | |
405 | hdr->flags = 0; | |
76faa4c0 | 406 | hdr->val.str.text = buf; |
407 | hdr->val.str.len = len; | |
f80e83a9 | 408 | return hdr; |
0578f103 | 409 | } |
410 | ||
50fd6b48 | 411 | /* Token-buffer helper functions. */ |
412 | ||
d2efb5ed | 413 | /* Expand a token list's string space. It is *vital* that |
414 | list->tokens_used is correct, to get pointer fix-up right. */ | |
f80e83a9 | 415 | void |
416 | _cpp_expand_name_space (list, len) | |
50fd6b48 | 417 | cpp_toklist *list; |
6060326b | 418 | unsigned int len; |
419 | { | |
e916a356 | 420 | const U_CHAR *old_namebuf; |
e916a356 | 421 | |
422 | old_namebuf = list->namebuf; | |
6060326b | 423 | list->name_cap += len; |
424 | list->namebuf = (unsigned char *) xrealloc (list->namebuf, list->name_cap); | |
e916a356 | 425 | |
426 | /* Fix up token text pointers. */ | |
6dafd80a | 427 | if (list->namebuf != old_namebuf) |
e916a356 | 428 | { |
429 | unsigned int i; | |
430 | ||
431 | for (i = 0; i < list->tokens_used; i++) | |
7e842f95 | 432 | if (TOKEN_SPELL (&list->tokens[i]) == SPELL_STRING) |
76faa4c0 | 433 | list->tokens[i].val.str.text += (list->namebuf - old_namebuf); |
e916a356 | 434 | } |
50fd6b48 | 435 | } |
436 | ||
f80e83a9 | 437 | /* If there is not enough room for LEN more characters, expand the |
438 | list by just enough to have room for LEN characters. */ | |
439 | void | |
440 | _cpp_reserve_name_space (list, len) | |
441 | cpp_toklist *list; | |
442 | unsigned int len; | |
443 | { | |
444 | unsigned int room = list->name_cap - list->name_used; | |
445 | ||
446 | if (room < len) | |
447 | _cpp_expand_name_space (list, len - room); | |
448 | } | |
449 | ||
50fd6b48 | 450 | /* Expand the number of tokens in a list. */ |
d2efb5ed | 451 | void |
452 | _cpp_expand_token_space (list, count) | |
50fd6b48 | 453 | cpp_toklist *list; |
d2efb5ed | 454 | unsigned int count; |
50fd6b48 | 455 | { |
d2efb5ed | 456 | unsigned int n; |
457 | ||
458 | list->tokens_cap += count; | |
459 | n = list->tokens_cap; | |
bce8e0c0 | 460 | if (list->flags & LIST_OFFSET) |
d2efb5ed | 461 | list->tokens--, n++; |
50fd6b48 | 462 | list->tokens = (cpp_token *) |
d2efb5ed | 463 | xrealloc (list->tokens, n * sizeof (cpp_token)); |
bce8e0c0 | 464 | if (list->flags & LIST_OFFSET) |
465 | list->tokens++; /* Skip the dummy. */ | |
50fd6b48 | 466 | } |
467 | ||
d2efb5ed | 468 | /* Initialize a token list. If flags is DUMMY_TOKEN, we allocate |
469 | an extra token in front of the token list, as this allows the lexer | |
470 | to always peek at the previous token without worrying about | |
471 | underflowing the list, and some initial space. Otherwise, no | |
472 | token- or name-space is allocated, and there is no dummy token. */ | |
bce8e0c0 | 473 | void |
d2efb5ed | 474 | _cpp_init_toklist (list, flags) |
50fd6b48 | 475 | cpp_toklist *list; |
d2efb5ed | 476 | int flags; |
50fd6b48 | 477 | { |
d2efb5ed | 478 | if (flags == NO_DUMMY_TOKEN) |
479 | { | |
480 | list->tokens_cap = 0; | |
f80e83a9 | 481 | list->tokens = 0; |
d2efb5ed | 482 | list->name_cap = 0; |
f80e83a9 | 483 | list->namebuf = 0; |
d2efb5ed | 484 | list->flags = 0; |
485 | } | |
486 | else | |
487 | { | |
488 | /* Initialize token space. Put a dummy token before the start | |
489 | that will fail matches. */ | |
490 | list->tokens_cap = 256; /* 4K's worth. */ | |
491 | list->tokens = (cpp_token *) | |
492 | xmalloc ((list->tokens_cap + 1) * sizeof (cpp_token)); | |
493 | list->tokens[0].type = CPP_EOF; | |
494 | list->tokens++; | |
495 | ||
496 | /* Initialize name space. */ | |
497 | list->name_cap = 1024; | |
f80e83a9 | 498 | list->namebuf = (unsigned char *) xmalloc (list->name_cap); |
d2efb5ed | 499 | list->flags = LIST_OFFSET; |
500 | } | |
bce8e0c0 | 501 | |
bce8e0c0 | 502 | _cpp_clear_toklist (list); |
503 | } | |
50fd6b48 | 504 | |
bce8e0c0 | 505 | /* Clear a token list. */ |
506 | void | |
507 | _cpp_clear_toklist (list) | |
508 | cpp_toklist *list; | |
509 | { | |
6060326b | 510 | list->tokens_used = 0; |
511 | list->name_used = 0; | |
f80e83a9 | 512 | list->directive = 0; |
513 | list->paramc = 0; | |
514 | list->params_len = 0; | |
bce8e0c0 | 515 | list->flags &= LIST_OFFSET; /* clear all but that one */ |
516 | } | |
517 | ||
518 | /* Free a token list. Does not free the list itself, which may be | |
519 | embedded in a larger structure. */ | |
520 | void | |
521 | _cpp_free_toklist (list) | |
f80e83a9 | 522 | const cpp_toklist *list; |
bce8e0c0 | 523 | { |
bce8e0c0 | 524 | if (list->flags & LIST_OFFSET) |
525 | free (list->tokens - 1); /* Backup over dummy token. */ | |
526 | else | |
527 | free (list->tokens); | |
528 | free (list->namebuf); | |
50fd6b48 | 529 | } |
530 | ||
bce8e0c0 | 531 | /* Compare two tokens. */ |
532 | int | |
533 | _cpp_equiv_tokens (a, b) | |
534 | const cpp_token *a, *b; | |
535 | { | |
f80e83a9 | 536 | if (a->type == b->type && a->flags == b->flags) |
7e842f95 | 537 | switch (TOKEN_SPELL (a)) |
f80e83a9 | 538 | { |
539 | default: /* Keep compiler happy. */ | |
540 | case SPELL_OPERATOR: | |
541 | return 1; | |
542 | case SPELL_CHAR: | |
543 | case SPELL_NONE: | |
544 | return a->val.aux == b->val.aux; /* arg_no or character. */ | |
545 | case SPELL_IDENT: | |
76faa4c0 | 546 | return a->val.node == b->val.node; |
f80e83a9 | 547 | case SPELL_STRING: |
76faa4c0 | 548 | return (a->val.str.len == b->val.str.len |
549 | && !memcmp (a->val.str.text, b->val.str.text, | |
550 | a->val.str.len)); | |
f80e83a9 | 551 | } |
bce8e0c0 | 552 | |
f80e83a9 | 553 | return 0; |
bce8e0c0 | 554 | } |
555 | ||
556 | /* Compare two token lists. */ | |
557 | int | |
558 | _cpp_equiv_toklists (a, b) | |
559 | const cpp_toklist *a, *b; | |
560 | { | |
561 | unsigned int i; | |
562 | ||
f80e83a9 | 563 | if (a->tokens_used != b->tokens_used |
564 | || a->flags != b->flags | |
565 | || a->paramc != b->paramc) | |
bce8e0c0 | 566 | return 0; |
567 | ||
568 | for (i = 0; i < a->tokens_used; i++) | |
569 | if (! _cpp_equiv_tokens (&a->tokens[i], &b->tokens[i])) | |
570 | return 0; | |
571 | return 1; | |
572 | } | |
573 | ||
f80e83a9 | 574 | /* Utility routine: |
2c63d6c8 | 575 | |
76faa4c0 | 576 | Compares, the token TOKEN to the NUL-terminated string STRING. |
577 | TOKEN must be a CPP_NAME. Returns 1 for equal, 0 for unequal. */ | |
bce8e0c0 | 578 | |
f80e83a9 | 579 | int |
76faa4c0 | 580 | cpp_ideq (token, string) |
581 | const cpp_token *token; | |
f80e83a9 | 582 | const char *string; |
583 | { | |
76faa4c0 | 584 | if (token->type != CPP_NAME) |
f80e83a9 | 585 | return 0; |
76faa4c0 | 586 | |
587 | return !ustrcmp (token->val.node->name, (const U_CHAR *)string); | |
bce8e0c0 | 588 | } |
50fd6b48 | 589 | |
f80e83a9 | 590 | static const unsigned char *digraph_spellings [] = {U"%:", U"%:%:", U"<:", |
591 | U":>", U"<%", U"%>"}; | |
0578f103 | 592 | |
338fa5f7 | 593 | /* Call when meeting a newline. Returns the character after the newline |
594 | (or carriage-return newline combination), or EOF. */ | |
595 | static cppchar_t | |
596 | handle_newline (buffer, newline_char) | |
597 | cpp_buffer *buffer; | |
598 | cppchar_t newline_char; | |
599 | { | |
600 | cppchar_t next = EOF; | |
601 | ||
602 | buffer->col_adjust = 0; | |
603 | buffer->lineno++; | |
604 | buffer->line_base = buffer->cur; | |
605 | ||
606 | /* Handle CR-LF and LF-CR combinations, get the next character. */ | |
607 | if (buffer->cur < buffer->rlimit) | |
608 | { | |
609 | next = *buffer->cur++; | |
610 | if (next + newline_char == '\r' + '\n') | |
611 | { | |
612 | buffer->line_base = buffer->cur; | |
613 | if (buffer->cur < buffer->rlimit) | |
614 | next = *buffer->cur++; | |
615 | else | |
616 | next = EOF; | |
617 | } | |
618 | } | |
619 | ||
620 | buffer->read_ahead = next; | |
621 | return next; | |
622 | } | |
623 | ||
624 | /* Subroutine of skip_escaped_newlines; called when a trigraph is | |
625 | encountered. It warns if necessary, and returns true if the | |
626 | trigraph should be honoured. FROM_CHAR is the third character of a | |
627 | trigraph, and presumed to be the previous character for position | |
628 | reporting. */ | |
0578f103 | 629 | static int |
338fa5f7 | 630 | trigraph_ok (pfile, from_char) |
0578f103 | 631 | cpp_reader *pfile; |
338fa5f7 | 632 | cppchar_t from_char; |
0578f103 | 633 | { |
f80e83a9 | 634 | int accept = CPP_OPTION (pfile, trigraphs); |
635 | ||
636 | if (CPP_OPTION (pfile, warn_trigraphs)) | |
0578f103 | 637 | { |
338fa5f7 | 638 | cpp_buffer *buffer = pfile->buffer; |
f80e83a9 | 639 | if (accept) |
338fa5f7 | 640 | cpp_warning_with_line (pfile, buffer->lineno, CPP_BUF_COL (buffer) - 2, |
f80e83a9 | 641 | "trigraph ??%c converted to %c", |
338fa5f7 | 642 | (int) from_char, |
643 | (int) _cpp_trigraph_map[from_char]); | |
0578f103 | 644 | else |
338fa5f7 | 645 | cpp_warning_with_line (pfile, buffer->lineno, CPP_BUF_COL (buffer) - 2, |
646 | "trigraph ??%c ignored", (int) from_char); | |
0578f103 | 647 | } |
338fa5f7 | 648 | |
f80e83a9 | 649 | return accept; |
0578f103 | 650 | } |
651 | ||
338fa5f7 | 652 | /* Assumes local variables buffer and result. */ |
653 | #define ACCEPT_CHAR(t) \ | |
654 | do { result->type = t; buffer->read_ahead = EOF; } while (0) | |
655 | ||
656 | /* When we move to multibyte character sets, add to these something | |
657 | that saves and restores the state of the multibyte conversion | |
658 | library. This probably involves saving and restoring a "cookie". | |
659 | In the case of glibc it is an 8-byte structure, so is not a high | |
660 | overhead operation. In any case, it's out of the fast path. */ | |
661 | #define SAVE_STATE() do { saved_cur = buffer->cur; } while (0) | |
662 | #define RESTORE_STATE() do { buffer->cur = saved_cur; } while (0) | |
663 | ||
664 | /* Skips any escaped newlines introduced by NEXT, which is either a | |
665 | '?' or a '\\'. Returns the next character, which will also have | |
666 | been placed in buffer->read_ahead. */ | |
667 | static cppchar_t | |
668 | skip_escaped_newlines (buffer, next) | |
669 | cpp_buffer *buffer; | |
670 | cppchar_t next; | |
0578f103 | 671 | { |
338fa5f7 | 672 | cppchar_t next1; |
673 | const unsigned char *saved_cur; | |
674 | int space; | |
f80e83a9 | 675 | |
338fa5f7 | 676 | do |
f80e83a9 | 677 | { |
338fa5f7 | 678 | if (buffer->cur == buffer->rlimit) |
679 | break; | |
680 | ||
681 | SAVE_STATE (); | |
682 | if (next == '?') | |
683 | { | |
684 | next1 = *buffer->cur++; | |
685 | if (next1 != '?' || buffer->cur == buffer->rlimit) | |
686 | { | |
687 | RESTORE_STATE (); | |
688 | break; | |
689 | } | |
f80e83a9 | 690 | |
338fa5f7 | 691 | next1 = *buffer->cur++; |
692 | if (!_cpp_trigraph_map[next1] || !trigraph_ok (buffer->pfile, next1)) | |
693 | { | |
694 | RESTORE_STATE (); | |
695 | break; | |
696 | } | |
0578f103 | 697 | |
338fa5f7 | 698 | /* We have a full trigraph here. */ |
699 | next = _cpp_trigraph_map[next1]; | |
700 | if (next != '\\' || buffer->cur == buffer->rlimit) | |
701 | break; | |
702 | SAVE_STATE (); | |
703 | } | |
704 | ||
705 | /* We have a backslash, and room for at least one more character. */ | |
706 | space = 0; | |
707 | do | |
708 | { | |
709 | next1 = *buffer->cur++; | |
710 | if (!is_nvspace (next1)) | |
711 | break; | |
712 | space = 1; | |
713 | } | |
714 | while (buffer->cur < buffer->rlimit); | |
715 | ||
716 | if (!is_vspace (next1)) | |
717 | { | |
718 | RESTORE_STATE (); | |
719 | break; | |
720 | } | |
0578f103 | 721 | |
338fa5f7 | 722 | if (space) |
723 | cpp_warning (buffer->pfile, | |
724 | "backslash and newline separated by space"); | |
725 | ||
726 | next = handle_newline (buffer, next1); | |
727 | if (next == EOF) | |
728 | cpp_pedwarn (buffer->pfile, "backslash-newline at end of file"); | |
f80e83a9 | 729 | } |
338fa5f7 | 730 | while (next == '\\' || next == '?'); |
0578f103 | 731 | |
338fa5f7 | 732 | buffer->read_ahead = next; |
733 | return next; | |
0578f103 | 734 | } |
735 | ||
338fa5f7 | 736 | /* Obtain the next character, after trigraph conversion and skipping |
737 | an arbitrary string of escaped newlines. The common case of no | |
738 | trigraphs or escaped newlines falls through quickly. */ | |
739 | static cppchar_t | |
740 | get_effective_char (buffer) | |
741 | cpp_buffer *buffer; | |
852d1b04 | 742 | { |
338fa5f7 | 743 | cppchar_t next = EOF; |
744 | ||
745 | if (buffer->cur < buffer->rlimit) | |
746 | { | |
747 | next = *buffer->cur++; | |
748 | ||
749 | /* '?' can introduce trigraphs (and therefore backslash); '\\' | |
750 | can introduce escaped newlines, which we want to skip, or | |
751 | UCNs, which, depending upon lexer state, we will handle in | |
752 | the future. */ | |
753 | if (next == '?' || next == '\\') | |
754 | next = skip_escaped_newlines (buffer, next); | |
755 | } | |
756 | ||
757 | buffer->read_ahead = next; | |
758 | return next; | |
852d1b04 | 759 | } |
760 | ||
338fa5f7 | 761 | /* Skip a C-style block comment. We find the end of the comment by |
762 | seeing if an asterisk is before every '/' we encounter. Returns | |
763 | non-zero if comment terminated by EOF, zero otherwise. */ | |
f80e83a9 | 764 | static int |
765 | skip_block_comment (pfile) | |
0578f103 | 766 | cpp_reader *pfile; |
767 | { | |
f80e83a9 | 768 | cpp_buffer *buffer = pfile->buffer; |
338fa5f7 | 769 | cppchar_t c = EOF, prevc; |
770 | ||
771 | while (buffer->cur != buffer->rlimit) | |
0578f103 | 772 | { |
338fa5f7 | 773 | prevc = c, c = *buffer->cur++; |
774 | ||
775 | next_char: | |
776 | /* FIXME: For speed, create a new character class of characters | |
777 | of no interest inside block comments. */ | |
778 | if (c == '?' || c == '\\') | |
779 | c = skip_escaped_newlines (buffer, c); | |
f80e83a9 | 780 | |
338fa5f7 | 781 | /* People like decorating comments with '*', so check for '/' |
782 | instead for efficiency. */ | |
f80e83a9 | 783 | if (c == '/') |
0578f103 | 784 | { |
338fa5f7 | 785 | if (prevc == '*') |
786 | break; | |
f80e83a9 | 787 | |
338fa5f7 | 788 | /* Warn about potential nested comments, but not if the '/' |
789 | comes immediately before the true comment delimeter. | |
f80e83a9 | 790 | Don't bother to get it right across escaped newlines. */ |
338fa5f7 | 791 | if (CPP_OPTION (pfile, warn_comments) |
792 | && buffer->cur != buffer->rlimit) | |
0578f103 | 793 | { |
338fa5f7 | 794 | prevc = c, c = *buffer->cur++; |
795 | if (c == '*' && buffer->cur != buffer->rlimit) | |
796 | { | |
797 | prevc = c, c = *buffer->cur++; | |
798 | if (c != '/') | |
799 | cpp_warning_with_line (pfile, CPP_BUF_LINE (buffer), | |
800 | CPP_BUF_COL (buffer), | |
801 | "\"/*\" within comment"); | |
802 | } | |
803 | goto next_char; | |
0578f103 | 804 | } |
0578f103 | 805 | } |
78719282 | 806 | else if (is_vspace (c)) |
0578f103 | 807 | { |
338fa5f7 | 808 | prevc = c, c = handle_newline (buffer, c); |
809 | goto next_char; | |
0578f103 | 810 | } |
b86584f6 | 811 | else if (c == '\t') |
338fa5f7 | 812 | adjust_column (pfile); |
0578f103 | 813 | } |
f80e83a9 | 814 | |
338fa5f7 | 815 | buffer->read_ahead = EOF; |
816 | return c != '/' || prevc != '*'; | |
0578f103 | 817 | } |
818 | ||
241e762e | 819 | /* Skip a C++ line comment. Handles escaped newlines. Returns |
338fa5f7 | 820 | non-zero if a multiline comment. The following new line, if any, |
821 | is left in buffer->read_ahead. */ | |
f80e83a9 | 822 | static int |
338fa5f7 | 823 | skip_line_comment (buffer) |
824 | cpp_buffer *buffer; | |
0578f103 | 825 | { |
338fa5f7 | 826 | unsigned int orig_lineno = buffer->lineno; |
827 | cppchar_t c; | |
f80e83a9 | 828 | |
338fa5f7 | 829 | do |
f80e83a9 | 830 | { |
338fa5f7 | 831 | c = EOF; |
832 | if (buffer->cur == buffer->rlimit) | |
833 | break; | |
f80e83a9 | 834 | |
338fa5f7 | 835 | c = *buffer->cur++; |
836 | if (c == '?' || c == '\\') | |
837 | c = skip_escaped_newlines (buffer, c); | |
f80e83a9 | 838 | } |
338fa5f7 | 839 | while (!is_vspace (c)); |
0578f103 | 840 | |
338fa5f7 | 841 | buffer->read_ahead = c; /* Leave any newline for caller. */ |
842 | return orig_lineno != buffer->lineno; | |
f80e83a9 | 843 | } |
0578f103 | 844 | |
338fa5f7 | 845 | /* pfile->buffer->cur is one beyond the \t character. Update |
846 | col_adjust so we track the column correctly. */ | |
b86584f6 | 847 | static void |
338fa5f7 | 848 | adjust_column (pfile) |
b86584f6 | 849 | cpp_reader *pfile; |
b86584f6 | 850 | { |
338fa5f7 | 851 | cpp_buffer *buffer = pfile->buffer; |
852 | unsigned int col = CPP_BUF_COL (buffer) - 1; /* Zero-based column. */ | |
b86584f6 | 853 | |
854 | /* Round it up to multiple of the tabstop, but subtract 1 since the | |
855 | tab itself occupies a character position. */ | |
338fa5f7 | 856 | buffer->col_adjust += (CPP_OPTION (pfile, tabstop) |
857 | - col % CPP_OPTION (pfile, tabstop)) - 1; | |
b86584f6 | 858 | } |
859 | ||
338fa5f7 | 860 | /* Skips whitespace, saving the next non-whitespace character. |
861 | Adjusts pfile->col_adjust to account for tabs. Without this, | |
862 | tokens might be assigned an incorrect column. */ | |
f80e83a9 | 863 | static void |
338fa5f7 | 864 | skip_whitespace (pfile, c) |
f80e83a9 | 865 | cpp_reader *pfile; |
338fa5f7 | 866 | cppchar_t c; |
f80e83a9 | 867 | { |
868 | cpp_buffer *buffer = pfile->buffer; | |
338fa5f7 | 869 | unsigned int warned = 0; |
0578f103 | 870 | |
338fa5f7 | 871 | do |
f80e83a9 | 872 | { |
78719282 | 873 | /* Horizontal space always OK. */ |
874 | if (c == ' ') | |
338fa5f7 | 875 | ; |
78719282 | 876 | else if (c == '\t') |
338fa5f7 | 877 | adjust_column (pfile); |
878 | /* Just \f \v or \0 left. */ | |
78719282 | 879 | else if (c == '\0') |
f80e83a9 | 880 | { |
78719282 | 881 | if (!warned) |
338fa5f7 | 882 | { |
883 | cpp_warning (pfile, "null character(s) ignored"); | |
884 | warned = 1; | |
885 | } | |
0578f103 | 886 | } |
338fa5f7 | 887 | else if (IN_DIRECTIVE (pfile) && CPP_PEDANTIC (pfile)) |
78719282 | 888 | cpp_pedwarn_with_line (pfile, CPP_BUF_LINE (buffer), |
889 | CPP_BUF_COL (buffer), | |
890 | "%s in preprocessing directive", | |
891 | c == '\f' ? "form feed" : "vertical tab"); | |
338fa5f7 | 892 | |
893 | c = EOF; | |
894 | if (buffer->cur == buffer->rlimit) | |
895 | break; | |
896 | c = *buffer->cur++; | |
0578f103 | 897 | } |
338fa5f7 | 898 | /* We only want non-vertical space, i.e. ' ' \t \f \v \0. */ |
899 | while (is_nvspace (c)); | |
900 | ||
901 | /* Remember the next character. */ | |
902 | buffer->read_ahead = c; | |
f80e83a9 | 903 | } |
0578f103 | 904 | |
338fa5f7 | 905 | /* Parse an identifier, skipping embedded backslash-newlines. |
906 | Calculate the hash value of the token while parsing, for improved | |
907 | performance. The hashing algorithm *must* match cpp_lookup(). */ | |
908 | ||
909 | static cpp_hashnode * | |
910 | parse_identifier (pfile, c) | |
0578f103 | 911 | cpp_reader *pfile; |
338fa5f7 | 912 | cppchar_t c; |
0578f103 | 913 | { |
338fa5f7 | 914 | cpp_buffer *buffer = pfile->buffer; |
915 | unsigned int r = 0, saw_dollar = 0; | |
916 | unsigned int orig_used = pfile->token_list.name_used; | |
f80e83a9 | 917 | |
338fa5f7 | 918 | do |
f80e83a9 | 919 | { |
338fa5f7 | 920 | do |
f80e83a9 | 921 | { |
338fa5f7 | 922 | if (pfile->token_list.name_used == pfile->token_list.name_cap) |
923 | _cpp_expand_name_space (&pfile->token_list, | |
924 | pfile->token_list.name_used + 256); | |
925 | pfile->token_list.namebuf[pfile->token_list.name_used++] = c; | |
926 | r = HASHSTEP (r, c); | |
0578f103 | 927 | |
338fa5f7 | 928 | if (c == '$') |
929 | saw_dollar++; | |
71aa9da4 | 930 | |
338fa5f7 | 931 | c = EOF; |
932 | if (buffer->cur == buffer->rlimit) | |
933 | break; | |
71aa9da4 | 934 | |
338fa5f7 | 935 | c = *buffer->cur++; |
936 | } | |
937 | while (is_idchar (c)); | |
71aa9da4 | 938 | |
338fa5f7 | 939 | /* Potential escaped newline? */ |
940 | if (c != '?' && c != '\\') | |
941 | break; | |
942 | c = skip_escaped_newlines (buffer, c); | |
f80e83a9 | 943 | } |
338fa5f7 | 944 | while (is_idchar (c)); |
945 | ||
946 | /* $ is not a identifier character in the standard, but is commonly | |
947 | accepted as an extension. Don't warn about it in skipped | |
948 | conditional blocks. */ | |
949 | if (saw_dollar && CPP_PEDANTIC (pfile) && ! pfile->skipping) | |
950 | cpp_pedwarn (pfile, "'$' character(s) in identifier"); | |
951 | ||
952 | /* Remember the next character. */ | |
953 | buffer->read_ahead = c; | |
954 | return _cpp_lookup_with_hash (pfile, &pfile->token_list.namebuf[orig_used], | |
955 | pfile->token_list.name_used - orig_used, r); | |
0578f103 | 956 | } |
957 | ||
338fa5f7 | 958 | /* Parse a number, skipping embedded backslash-newlines. */ |
0578f103 | 959 | static void |
338fa5f7 | 960 | parse_number (pfile, number, c) |
0578f103 | 961 | cpp_reader *pfile; |
338fa5f7 | 962 | cpp_string *number; |
963 | cppchar_t c; | |
0578f103 | 964 | { |
338fa5f7 | 965 | cppchar_t prevc; |
f80e83a9 | 966 | cpp_buffer *buffer = pfile->buffer; |
338fa5f7 | 967 | unsigned int orig_used = pfile->token_list.name_used; |
0578f103 | 968 | |
338fa5f7 | 969 | do |
f80e83a9 | 970 | { |
338fa5f7 | 971 | do |
972 | { | |
973 | if (pfile->token_list.name_used == pfile->token_list.name_cap) | |
974 | _cpp_expand_name_space (&pfile->token_list, | |
975 | pfile->token_list.name_used + 256); | |
976 | pfile->token_list.namebuf[pfile->token_list.name_used++] = c; | |
977 | ||
978 | prevc = c; | |
979 | c = EOF; | |
980 | if (buffer->cur == buffer->rlimit) | |
981 | break; | |
0578f103 | 982 | |
338fa5f7 | 983 | c = *buffer->cur++; |
984 | } | |
985 | while (is_numchar (c) || c == '.' || VALID_SIGN (c, prevc)); | |
0578f103 | 986 | |
338fa5f7 | 987 | /* Potential escaped newline? */ |
988 | if (c != '?' && c != '\\') | |
989 | break; | |
990 | c = skip_escaped_newlines (buffer, c); | |
0578f103 | 991 | } |
338fa5f7 | 992 | while (is_numchar (c) || c == '.' || VALID_SIGN (c, prevc)); |
993 | ||
994 | /* Remember the next character. */ | |
995 | buffer->read_ahead = c; | |
852d1b04 | 996 | |
338fa5f7 | 997 | number->text = &pfile->token_list.namebuf[orig_used]; |
998 | number->len = pfile->token_list.name_used - orig_used; | |
999 | } | |
1000 | ||
1001 | /* Subroutine of parse_string. Emits error for unterminated strings. */ | |
1002 | static void | |
1003 | unterminated (pfile, line, term) | |
1004 | cpp_reader *pfile; | |
1005 | unsigned int line; | |
1006 | int term; | |
1007 | { | |
1008 | cpp_error (pfile, "missing terminating %c character", term); | |
1009 | ||
1010 | if (term == '\"' && pfile->mls_line && pfile->mls_line != line) | |
f80e83a9 | 1011 | { |
338fa5f7 | 1012 | cpp_error_with_line (pfile, pfile->mls_line, pfile->mls_column, |
1013 | "possible start of unterminated string literal"); | |
1014 | pfile->mls_line = 0; | |
f80e83a9 | 1015 | } |
0578f103 | 1016 | } |
1017 | ||
338fa5f7 | 1018 | /* Parses a string, character constant, or angle-bracketed header file |
1019 | name. Handles embedded trigraphs and escaped newlines. | |
0578f103 | 1020 | |
338fa5f7 | 1021 | Multi-line strings are allowed, but they are deprecated within |
1022 | directives. */ | |
f80e83a9 | 1023 | static void |
338fa5f7 | 1024 | parse_string (pfile, token, terminator) |
0578f103 | 1025 | cpp_reader *pfile; |
f80e83a9 | 1026 | cpp_token *token; |
338fa5f7 | 1027 | cppchar_t terminator; |
0578f103 | 1028 | { |
f80e83a9 | 1029 | cpp_buffer *buffer = pfile->buffer; |
338fa5f7 | 1030 | unsigned int orig_used = pfile->token_list.name_used; |
1031 | cppchar_t c; | |
1032 | unsigned int nulls = 0; | |
1033 | ||
1034 | for (;;) | |
0578f103 | 1035 | { |
338fa5f7 | 1036 | if (buffer->cur == buffer->rlimit) |
1037 | { | |
1038 | c = EOF; | |
1039 | unterminated (pfile, token->line, terminator); | |
1040 | break; | |
1041 | } | |
1042 | c = *buffer->cur++; | |
1043 | ||
1044 | have_char: | |
1045 | /* Handle trigraphs, escaped newlines etc. */ | |
1046 | if (c == '?' || c == '\\') | |
1047 | c = skip_escaped_newlines (buffer, c); | |
0578f103 | 1048 | |
338fa5f7 | 1049 | if (c == terminator) |
0578f103 | 1050 | { |
338fa5f7 | 1051 | unsigned int u = pfile->token_list.name_used; |
01f93cf7 | 1052 | |
338fa5f7 | 1053 | /* An odd number of consecutive backslashes represents an |
1054 | escaped terminator. */ | |
1055 | while (u > orig_used && pfile->token_list.namebuf[u - 1] == '\\') | |
1056 | u--; | |
1057 | ||
1058 | if ((pfile->token_list.name_used - u) % 2 == 0) | |
0578f103 | 1059 | { |
338fa5f7 | 1060 | c = EOF; |
1061 | break; | |
0578f103 | 1062 | } |
338fa5f7 | 1063 | } |
1064 | else if (is_vspace (c)) | |
1065 | { | |
1066 | /* In assembly language, silently terminate string and | |
1067 | character literals at end of line. This is a kludge | |
1068 | around not knowing where comments are. */ | |
1069 | if (CPP_OPTION (pfile, lang_asm) && terminator != '>') | |
1070 | break; | |
0578f103 | 1071 | |
338fa5f7 | 1072 | /* Character constants and header names may not extend over |
1073 | multiple lines. In Standard C, neither may strings. | |
1074 | Unfortunately, we accept multiline strings as an | |
1075 | extension. (Deprecatedly even in directives - otherwise, | |
1076 | glibc's longlong.h breaks.) */ | |
1077 | if (terminator != '"') | |
0578f103 | 1078 | { |
338fa5f7 | 1079 | unterminated (pfile, token->line, terminator); |
1080 | break; | |
0578f103 | 1081 | } |
0578f103 | 1082 | |
338fa5f7 | 1083 | if (pfile->mls_line == 0) |
1084 | { | |
1085 | pfile->mls_line = token->line; | |
1086 | pfile->mls_column = token->col; | |
1087 | if (CPP_PEDANTIC (pfile)) | |
1088 | cpp_pedwarn (pfile, "multi-line string constant"); | |
f80e83a9 | 1089 | } |
338fa5f7 | 1090 | |
1091 | handle_newline (buffer, c); /* Stores to read_ahead. */ | |
1092 | c = '\n'; | |
1093 | } | |
1094 | else if (c == '\0') | |
1095 | { | |
1096 | if (nulls++ == 0) | |
1097 | cpp_warning (pfile, "null character(s) preserved in literal"); | |
0578f103 | 1098 | } |
0578f103 | 1099 | |
338fa5f7 | 1100 | if (pfile->token_list.name_used == pfile->token_list.name_cap) |
1101 | _cpp_expand_name_space (&pfile->token_list, | |
1102 | pfile->token_list.name_used + 256); | |
9fb5b53d | 1103 | |
338fa5f7 | 1104 | pfile->token_list.namebuf[pfile->token_list.name_used++] = c; |
1105 | /* If we had a new line, the next character is in read_ahead. */ | |
1106 | if (c != '\n') | |
1107 | continue; | |
1108 | c = buffer->read_ahead; | |
1109 | if (c != EOF) | |
1110 | goto have_char; | |
0578f103 | 1111 | } |
1112 | ||
338fa5f7 | 1113 | buffer->read_ahead = c; |
0578f103 | 1114 | |
338fa5f7 | 1115 | token->val.str.text = &pfile->token_list.namebuf[orig_used]; |
1116 | token->val.str.len = pfile->token_list.name_used - orig_used; | |
1117 | } | |
f80e83a9 | 1118 | |
338fa5f7 | 1119 | /* For output routine simplicity, the stored comment includes the |
1120 | comment start and any terminator. */ | |
2c63d6c8 | 1121 | static void |
338fa5f7 | 1122 | save_comment (pfile, token, from) |
1123 | cpp_reader *pfile; | |
f80e83a9 | 1124 | cpp_token *token; |
1125 | const unsigned char *from; | |
2c63d6c8 | 1126 | { |
f80e83a9 | 1127 | unsigned char *buffer; |
338fa5f7 | 1128 | unsigned int len; |
1129 | cpp_toklist *list = &pfile->token_list; | |
1130 | ||
1131 | #define COMMENT_START_LEN 2 | |
1132 | len = pfile->buffer->cur - from + COMMENT_START_LEN; | |
1133 | _cpp_reserve_name_space (list, len); | |
1134 | buffer = list->namebuf + list->name_used; | |
1135 | list->name_used += len; | |
f80e83a9 | 1136 | |
f80e83a9 | 1137 | token->type = CPP_COMMENT; |
76faa4c0 | 1138 | token->val.str.len = len; |
338fa5f7 | 1139 | token->val.str.text = buffer; |
0578f103 | 1140 | |
338fa5f7 | 1141 | /* from[-1] is '/' or '*' depending on the comment type. */ |
1142 | *buffer++ = '/'; | |
1143 | *buffer++ = from[-1]; | |
1144 | memcpy (buffer, from, len - COMMENT_START_LEN); | |
1145 | } | |
0578f103 | 1146 | |
338fa5f7 | 1147 | /* A helper routine for lex_token. With some long tokens, we need |
1148 | to read ahead to see if that is the token we have, but back-track | |
1149 | if not. */ | |
1150 | static void | |
1151 | check_long_token (buffer, result, wanted, type) | |
1152 | cpp_buffer *buffer; | |
1153 | cpp_token *result; | |
1154 | cppchar_t wanted; | |
1155 | enum cpp_ttype type; | |
1156 | { | |
1157 | const unsigned char *saved_cur; | |
1158 | cppchar_t c = buffer->read_ahead; | |
1159 | ||
1160 | SAVE_STATE (); | |
1161 | if (get_effective_char (buffer) == wanted) | |
1162 | ACCEPT_CHAR (type); | |
f80e83a9 | 1163 | else |
c4357c92 | 1164 | { |
338fa5f7 | 1165 | /* Restore state. */ |
1166 | RESTORE_STATE (); | |
1167 | buffer->read_ahead = c; | |
c4357c92 | 1168 | } |
0578f103 | 1169 | } |
1170 | ||
f80e83a9 | 1171 | static void |
338fa5f7 | 1172 | lex_token (pfile, result) |
0578f103 | 1173 | cpp_reader *pfile; |
338fa5f7 | 1174 | cpp_token *result; |
0578f103 | 1175 | { |
338fa5f7 | 1176 | cppchar_t c; |
f80e83a9 | 1177 | cpp_buffer *buffer = pfile->buffer; |
338fa5f7 | 1178 | const unsigned char *comment_start; |
0653b94e | 1179 | |
338fa5f7 | 1180 | result->flags = 0; |
1181 | next_char: | |
1182 | result->line = CPP_BUF_LINE (buffer); | |
1183 | next_char2: | |
1184 | result->col = CPP_BUF_COLUMN (buffer, buffer->cur); | |
f80e83a9 | 1185 | |
338fa5f7 | 1186 | c = buffer->read_ahead; |
1187 | if (c == EOF && buffer->cur < buffer->rlimit) | |
1188 | { | |
1189 | c = *buffer->cur++; | |
1190 | result->col++; | |
1191 | } | |
0578f103 | 1192 | |
338fa5f7 | 1193 | do_switch: |
1194 | buffer->read_ahead = EOF; | |
1195 | switch (c) | |
0578f103 | 1196 | { |
338fa5f7 | 1197 | case EOF: |
1198 | /* Non-empty files should end in a newline. Testing | |
1199 | skip_newlines ensures we only emit the warning once. */ | |
1200 | if (buffer->cur != buffer->line_base && buffer->cur != buffer->buf | |
1201 | && pfile->state.skip_newlines) | |
1202 | cpp_pedwarn_with_line (pfile, buffer->lineno, CPP_BUF_COL (buffer), | |
1203 | "no newline at end of file"); | |
1204 | result->type = CPP_EOF; | |
1205 | break; | |
0578f103 | 1206 | |
338fa5f7 | 1207 | case ' ': case '\t': case '\f': case '\v': case '\0': |
1208 | skip_whitespace (pfile, c); | |
1209 | result->flags |= PREV_WHITE; | |
1210 | goto next_char2; | |
1211 | ||
1212 | case '\n': case '\r': | |
1213 | result->type = CPP_EOF; | |
1214 | handle_newline (buffer, c); | |
1215 | /* Handling here will change significantly when moving to | |
1216 | token-at-a-time. */ | |
1217 | if (pfile->state.skip_newlines) | |
0578f103 | 1218 | { |
338fa5f7 | 1219 | result->flags &= ~PREV_WHITE; /* Clear any whitespace flag. */ |
1220 | goto next_char; | |
0578f103 | 1221 | } |
338fa5f7 | 1222 | break; |
732cb4c9 | 1223 | |
338fa5f7 | 1224 | case '?': |
1225 | case '\\': | |
1226 | /* These could start an escaped newline, or '?' a trigraph. Let | |
1227 | skip_escaped_newlines do all the work. */ | |
1228 | { | |
1229 | unsigned int lineno = buffer->lineno; | |
1230 | ||
1231 | c = skip_escaped_newlines (buffer, c); | |
1232 | if (lineno != buffer->lineno) | |
1233 | /* We had at least one escaped newline of some sort, and the | |
1234 | next character is in buffer->read_ahead. Update the | |
1235 | token's line and column. */ | |
1236 | goto next_char; | |
1237 | ||
1238 | /* We are either the original '?' or '\\', or a trigraph. */ | |
1239 | result->type = CPP_QUERY; | |
1240 | buffer->read_ahead = EOF; | |
1241 | if (c == '\\') | |
1242 | result->type = CPP_BACKSLASH; | |
1243 | else if (c != '?') | |
1244 | goto do_switch; | |
1245 | } | |
1246 | break; | |
732cb4c9 | 1247 | |
338fa5f7 | 1248 | make_number: |
1249 | case '0': case '1': case '2': case '3': case '4': | |
1250 | case '5': case '6': case '7': case '8': case '9': | |
1251 | result->type = CPP_NUMBER; | |
1252 | parse_number (pfile, &result->val.str, c); | |
1253 | break; | |
732cb4c9 | 1254 | |
338fa5f7 | 1255 | case '$': |
1256 | if (!CPP_OPTION (pfile, dollars_in_ident)) | |
1257 | goto random_char; | |
1258 | /* Fall through... */ | |
1259 | ||
1260 | case '_': | |
1261 | case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': | |
1262 | case 'g': case 'h': case 'i': case 'j': case 'k': case 'l': | |
1263 | case 'm': case 'n': case 'o': case 'p': case 'q': case 'r': | |
1264 | case 's': case 't': case 'u': case 'v': case 'w': case 'x': | |
1265 | case 'y': case 'z': | |
1266 | case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': | |
1267 | case 'G': case 'H': case 'I': case 'J': case 'K': case 'L': | |
1268 | case 'M': case 'N': case 'O': case 'P': case 'Q': case 'R': | |
1269 | case 'S': case 'T': case 'U': case 'V': case 'W': case 'X': | |
1270 | case 'Y': case 'Z': | |
1271 | result->type = CPP_NAME; | |
1272 | result->val.node = parse_identifier (pfile, c); | |
1273 | ||
1274 | /* 'L' may introduce wide characters or strings. */ | |
1275 | if (result->val.node == pfile->spec_nodes->n_L) | |
1276 | { | |
1277 | c = buffer->read_ahead; /* For make_string. */ | |
1278 | if (c == '\'' || c == '"') | |
71aa9da4 | 1279 | { |
338fa5f7 | 1280 | ACCEPT_CHAR (c == '"' ? CPP_WSTRING: CPP_WCHAR); |
1281 | goto make_string; | |
71aa9da4 | 1282 | } |
338fa5f7 | 1283 | } |
1284 | /* Convert named operators to their proper types. */ | |
1285 | else if (result->val.node->type == T_OPERATOR) | |
1286 | { | |
1287 | result->flags |= NAMED_OP; | |
1288 | result->type = result->val.node->value.code; | |
1289 | } | |
1290 | break; | |
1291 | ||
1292 | case '\'': | |
1293 | case '"': | |
1294 | result->type = c == '"' ? CPP_STRING: CPP_CHAR; | |
1295 | make_string: | |
1296 | parse_string (pfile, result, c); | |
1297 | break; | |
f80e83a9 | 1298 | |
338fa5f7 | 1299 | case '/': |
1300 | result->type = CPP_DIV; | |
1301 | c = get_effective_char (buffer); | |
1302 | if (c == '=') | |
1303 | ACCEPT_CHAR (CPP_DIV_EQ); | |
1304 | else if (c == '*') | |
1305 | { | |
1306 | comment_start = buffer->cur; | |
0578f103 | 1307 | |
338fa5f7 | 1308 | /* Skip_block_comment updates buffer->read_ahead. */ |
1309 | if (skip_block_comment (pfile)) | |
1310 | cpp_error_with_line (pfile, result->line, result->col, | |
1311 | "unterminated comment"); | |
1312 | if (!pfile->state.save_comments) | |
31674461 | 1313 | { |
338fa5f7 | 1314 | result->flags |= PREV_WHITE; |
1315 | goto next_char; | |
31674461 | 1316 | } |
1317 | ||
338fa5f7 | 1318 | /* Save the comment as a token in its own right. */ |
1319 | save_comment (pfile, result, comment_start); | |
1320 | } | |
1321 | else if (c == '/') | |
1322 | { | |
1323 | /* We silently allow C++ comments in system headers, | |
1324 | irrespective of conformance mode, because lots of | |
1325 | broken systems do that and trying to clean it up in | |
1326 | fixincludes is a nightmare. */ | |
1327 | if (CPP_IN_SYSTEM_HEADER (pfile)) | |
1328 | goto do_line_comment; | |
1329 | if (CPP_OPTION (pfile, cplusplus_comments)) | |
f80e83a9 | 1330 | { |
338fa5f7 | 1331 | if (CPP_OPTION (pfile, c89) && CPP_PEDANTIC (pfile) |
1332 | && ! buffer->warned_cplusplus_comments) | |
f80e83a9 | 1333 | { |
338fa5f7 | 1334 | cpp_pedwarn (pfile, |
1335 | "C++ style comments are not allowed in ISO C89"); | |
1336 | cpp_pedwarn (pfile, | |
1337 | "(this will be reported only once per input file)"); | |
1338 | buffer->warned_cplusplus_comments = 1; | |
f80e83a9 | 1339 | } |
f80e83a9 | 1340 | |
338fa5f7 | 1341 | do_line_comment: |
1342 | comment_start = buffer->cur; | |
1343 | ||
1344 | /* Skip_line_comment updates buffer->read_ahead. */ | |
1345 | if (skip_line_comment (buffer)) | |
1346 | cpp_warning_with_line (pfile, result->line, result->col, | |
1347 | "multi-line comment"); | |
1348 | ||
1349 | if (!pfile->state.save_comments) | |
f80e83a9 | 1350 | { |
338fa5f7 | 1351 | result->flags |= PREV_WHITE; |
1352 | goto next_char; | |
f80e83a9 | 1353 | } |
338fa5f7 | 1354 | |
1355 | /* Save the comment as a token in its own right. */ | |
1356 | save_comment (pfile, result, comment_start); | |
0578f103 | 1357 | } |
338fa5f7 | 1358 | } |
1359 | break; | |
1360 | ||
1361 | case '<': | |
1362 | if (pfile->state.angled_headers) | |
1363 | { | |
1364 | result->type = CPP_HEADER_NAME; | |
1365 | c = '>'; /* terminator. */ | |
1366 | goto make_string; | |
1367 | } | |
0578f103 | 1368 | |
338fa5f7 | 1369 | result->type = CPP_LESS; |
1370 | c = get_effective_char (buffer); | |
1371 | if (c == '=') | |
1372 | ACCEPT_CHAR (CPP_LESS_EQ); | |
1373 | else if (c == '<') | |
1374 | { | |
1375 | ACCEPT_CHAR (CPP_LSHIFT); | |
1376 | if (get_effective_char (buffer) == '=') | |
1377 | ACCEPT_CHAR (CPP_LSHIFT_EQ); | |
1378 | } | |
1379 | else if (c == '?' && CPP_OPTION (pfile, cplusplus)) | |
1380 | { | |
1381 | ACCEPT_CHAR (CPP_MIN); | |
1382 | if (get_effective_char (buffer) == '=') | |
1383 | ACCEPT_CHAR (CPP_MIN_EQ); | |
1384 | } | |
1385 | else if (c == ':' && CPP_OPTION (pfile, digraphs)) | |
1386 | { | |
1387 | ACCEPT_CHAR (CPP_OPEN_SQUARE); | |
1388 | result->flags |= DIGRAPH; | |
1389 | } | |
1390 | else if (c == '%' && CPP_OPTION (pfile, digraphs)) | |
1391 | { | |
1392 | ACCEPT_CHAR (CPP_OPEN_BRACE); | |
1393 | result->flags |= DIGRAPH; | |
1394 | } | |
1395 | break; | |
1396 | ||
1397 | case '>': | |
1398 | result->type = CPP_GREATER; | |
1399 | c = get_effective_char (buffer); | |
1400 | if (c == '=') | |
1401 | ACCEPT_CHAR (CPP_GREATER_EQ); | |
1402 | else if (c == '>') | |
1403 | { | |
1404 | ACCEPT_CHAR (CPP_RSHIFT); | |
1405 | if (get_effective_char (buffer) == '=') | |
1406 | ACCEPT_CHAR (CPP_RSHIFT_EQ); | |
1407 | } | |
1408 | else if (c == '?' && CPP_OPTION (pfile, cplusplus)) | |
1409 | { | |
1410 | ACCEPT_CHAR (CPP_MAX); | |
1411 | if (get_effective_char (buffer) == '=') | |
1412 | ACCEPT_CHAR (CPP_MAX_EQ); | |
1413 | } | |
1414 | break; | |
1415 | ||
1416 | case '.': | |
1417 | { | |
1418 | const unsigned char *saved_cur; | |
1419 | cppchar_t c1; | |
1420 | ||
1421 | /* Save state to avoid needing to pass 2 chars to parse_number. */ | |
1422 | SAVE_STATE (); | |
1423 | c1 = get_effective_char (buffer); | |
1424 | /* All known character sets have 0...9 contiguous. */ | |
1425 | if (c1 >= '0' && c1 <= '9') | |
1426 | { | |
1427 | RESTORE_STATE (); | |
1428 | goto make_number; | |
1429 | } | |
1430 | ||
1431 | result->type = CPP_DOT; | |
1432 | if (c1 == '.') | |
1433 | { | |
1434 | if (get_effective_char (buffer) == '.') | |
1435 | ACCEPT_CHAR (CPP_ELLIPSIS); | |
1436 | else | |
1437 | { | |
1438 | buffer->read_ahead = EOF; | |
1439 | RESTORE_STATE (); | |
1440 | } | |
1441 | } | |
1442 | else if (c1 == '*' && CPP_OPTION (pfile, cplusplus)) | |
1443 | ACCEPT_CHAR (CPP_DOT_STAR); | |
1444 | } | |
1445 | break; | |
1446 | ||
1447 | case '%': | |
1448 | result->type = CPP_MOD; | |
1449 | c = get_effective_char (buffer); | |
1450 | if (c == '=') | |
1451 | ACCEPT_CHAR (CPP_MOD_EQ); | |
1452 | else if (CPP_OPTION (pfile, digraphs)) | |
1453 | { | |
1454 | if (c == ':') | |
0578f103 | 1455 | { |
338fa5f7 | 1456 | result->flags |= DIGRAPH; |
1457 | ACCEPT_CHAR (CPP_HASH); | |
1458 | if (get_effective_char (buffer) == '%') | |
1459 | check_long_token (buffer, result, ':', CPP_PASTE); | |
f80e83a9 | 1460 | } |
338fa5f7 | 1461 | else if (c == '>') |
f80e83a9 | 1462 | { |
338fa5f7 | 1463 | result->flags |= DIGRAPH; |
1464 | ACCEPT_CHAR (CPP_CLOSE_BRACE); | |
c4abf88d | 1465 | } |
338fa5f7 | 1466 | } |
1467 | break; | |
0578f103 | 1468 | |
338fa5f7 | 1469 | case '+': |
1470 | result->type = CPP_PLUS; | |
1471 | c = get_effective_char (buffer); | |
1472 | if (c == '=') | |
1473 | ACCEPT_CHAR (CPP_PLUS_EQ); | |
1474 | else if (c == '+') | |
1475 | ACCEPT_CHAR (CPP_PLUS_PLUS); | |
1476 | break; | |
ac0749c7 | 1477 | |
338fa5f7 | 1478 | case '-': |
1479 | result->type = CPP_MINUS; | |
1480 | c = get_effective_char (buffer); | |
1481 | if (c == '>') | |
1482 | { | |
1483 | ACCEPT_CHAR (CPP_DEREF); | |
1484 | if (CPP_OPTION (pfile, cplusplus) | |
1485 | && get_effective_char (buffer) == '*') | |
1486 | ACCEPT_CHAR (CPP_DEREF_STAR); | |
1487 | } | |
1488 | else if (c == '=') | |
1489 | ACCEPT_CHAR (CPP_MINUS_EQ); | |
1490 | else if (c == '-') | |
1491 | ACCEPT_CHAR (CPP_MINUS_MINUS); | |
1492 | break; | |
0578f103 | 1493 | |
338fa5f7 | 1494 | case '*': |
1495 | result->type = CPP_MULT; | |
1496 | if (get_effective_char (buffer) == '=') | |
1497 | ACCEPT_CHAR (CPP_MULT_EQ); | |
1498 | break; | |
ac0749c7 | 1499 | |
338fa5f7 | 1500 | case '=': |
1501 | result->type = CPP_EQ; | |
1502 | if (get_effective_char (buffer) == '=') | |
1503 | ACCEPT_CHAR (CPP_EQ_EQ); | |
1504 | break; | |
c4abf88d | 1505 | |
338fa5f7 | 1506 | case '!': |
1507 | result->type = CPP_NOT; | |
1508 | if (get_effective_char (buffer) == '=') | |
1509 | ACCEPT_CHAR (CPP_NOT_EQ); | |
1510 | break; | |
0578f103 | 1511 | |
338fa5f7 | 1512 | case '&': |
1513 | result->type = CPP_AND; | |
1514 | c = get_effective_char (buffer); | |
1515 | if (c == '=') | |
1516 | ACCEPT_CHAR (CPP_AND_EQ); | |
1517 | else if (c == '&') | |
1518 | ACCEPT_CHAR (CPP_AND_AND); | |
1519 | break; | |
1520 | ||
1521 | case '#': | |
1522 | result->type = CPP_HASH; | |
1523 | if (get_effective_char (buffer) == '#') | |
1524 | ACCEPT_CHAR (CPP_PASTE); | |
1525 | break; | |
0578f103 | 1526 | |
338fa5f7 | 1527 | case '|': |
1528 | result->type = CPP_OR; | |
1529 | c = get_effective_char (buffer); | |
1530 | if (c == '=') | |
1531 | ACCEPT_CHAR (CPP_OR_EQ); | |
1532 | else if (c == '|') | |
1533 | ACCEPT_CHAR (CPP_OR_OR); | |
1534 | break; | |
0578f103 | 1535 | |
338fa5f7 | 1536 | case '^': |
1537 | result->type = CPP_XOR; | |
1538 | if (get_effective_char (buffer) == '=') | |
1539 | ACCEPT_CHAR (CPP_XOR_EQ); | |
1540 | break; | |
0578f103 | 1541 | |
338fa5f7 | 1542 | case ':': |
1543 | result->type = CPP_COLON; | |
1544 | c = get_effective_char (buffer); | |
1545 | if (c == ':' && CPP_OPTION (pfile, cplusplus)) | |
1546 | ACCEPT_CHAR (CPP_SCOPE); | |
1547 | else if (c == '>' && CPP_OPTION (pfile, digraphs)) | |
1548 | { | |
1549 | result->flags |= DIGRAPH; | |
1550 | ACCEPT_CHAR (CPP_CLOSE_SQUARE); | |
1551 | } | |
1552 | break; | |
0578f103 | 1553 | |
338fa5f7 | 1554 | case '~': result->type = CPP_COMPL; break; |
1555 | case ',': result->type = CPP_COMMA; break; | |
1556 | case '(': result->type = CPP_OPEN_PAREN; break; | |
1557 | case ')': result->type = CPP_CLOSE_PAREN; break; | |
1558 | case '[': result->type = CPP_OPEN_SQUARE; break; | |
1559 | case ']': result->type = CPP_CLOSE_SQUARE; break; | |
1560 | case '{': result->type = CPP_OPEN_BRACE; break; | |
1561 | case '}': result->type = CPP_CLOSE_BRACE; break; | |
1562 | case ';': result->type = CPP_SEMICOLON; break; | |
1563 | ||
1564 | case '@': | |
1565 | if (CPP_OPTION (pfile, objc)) | |
1566 | { | |
1567 | /* In Objective C, '@' may begin keywords or strings, like | |
1568 | @keyword or @"string". It would be nice to call | |
1569 | get_effective_char here and test the result. However, we | |
1570 | would then need to pass 2 characters to parse_identifier, | |
1571 | making it ugly and slowing down its main loop. Instead, | |
1572 | we assume we have an identifier, and recover if not. */ | |
1573 | result->type = CPP_NAME; | |
1574 | result->val.node = parse_identifier (pfile, c); | |
1575 | if (result->val.node->length != 1) | |
1576 | break; | |
ac0749c7 | 1577 | |
338fa5f7 | 1578 | /* OK, so it wasn't an identifier. Maybe a string? */ |
1579 | if (buffer->read_ahead == '"') | |
f80e83a9 | 1580 | { |
338fa5f7 | 1581 | c = '"'; |
1582 | ACCEPT_CHAR (CPP_OSTRING); | |
1583 | goto make_string; | |
f80e83a9 | 1584 | } |
338fa5f7 | 1585 | } |
1586 | goto random_char; | |
1587 | ||
1588 | random_char: | |
1589 | default: | |
1590 | result->type = CPP_OTHER; | |
1591 | result->val.aux = c; | |
1592 | break; | |
1593 | } | |
1594 | } | |
1595 | ||
1596 | /* | |
1597 | * The tokenizer's main loop. Returns a token list, representing a | |
1598 | * logical line in the input file. On EOF after some tokens have | |
1599 | * been processed, we return immediately. Then in next call, or if | |
1600 | * EOF occurred at the beginning of a logical line, a single CPP_EOF | |
1601 | * token is placed in the list. | |
1602 | */ | |
1603 | ||
1604 | static void | |
1605 | lex_line (pfile, list) | |
1606 | cpp_reader *pfile; | |
1607 | cpp_toklist *list; | |
1608 | { | |
1609 | unsigned int first_token; | |
1610 | cpp_token *cur_token, *first; | |
1611 | cpp_buffer *buffer = pfile->buffer; | |
0578f103 | 1612 | |
338fa5f7 | 1613 | if (!(list->flags & LIST_OFFSET)) |
1614 | (abort) (); | |
1615 | ||
1616 | pfile->state.in_lex_line = 1; | |
1617 | if (pfile->buffer->cur == pfile->buffer->buf) | |
1618 | list->flags |= BEG_OF_FILE; | |
1619 | ||
1620 | retry: | |
1621 | pfile->state.in_directive = 0; | |
1622 | pfile->state.angled_headers = 0; | |
1623 | pfile->state.skip_newlines = 1; | |
1624 | pfile->state.save_comments = ! CPP_OPTION (pfile, discard_comments); | |
1625 | first_token = list->tokens_used; | |
1626 | list->file = buffer->nominal_fname; | |
1627 | ||
1628 | do | |
1629 | { | |
1630 | if (list->tokens_used >= list->tokens_cap) | |
1631 | _cpp_expand_token_space (list, 256); | |
1632 | ||
1633 | cur_token = list->tokens + list->tokens_used; | |
1634 | lex_token (pfile, cur_token); | |
1635 | ||
1636 | if (pfile->state.skip_newlines) | |
1637 | { | |
1638 | pfile->state.skip_newlines = 0; | |
1639 | list->line = buffer->lineno; | |
1640 | if (cur_token->type == CPP_HASH) | |
f80e83a9 | 1641 | { |
338fa5f7 | 1642 | pfile->state.in_directive = 1; |
1643 | pfile->state.save_comments = 0; | |
1644 | pfile->state.indented = cur_token->flags & PREV_WHITE; | |
f80e83a9 | 1645 | } |
338fa5f7 | 1646 | /* 6.10.3.10: Within the sequence of preprocessing tokens |
1647 | making up the invocation of a function-like macro, new | |
1648 | line is considered a normal white-space character. */ | |
1649 | else if (first_token != 0) | |
1650 | cur_token->flags |= PREV_WHITE; | |
1651 | } | |
1652 | else if (IN_DIRECTIVE (pfile) && list->tokens_used == first_token + 1) | |
1653 | { | |
1654 | if (cur_token->type == CPP_NUMBER) | |
1655 | list->directive = _cpp_check_linemarker (pfile, cur_token); | |
f80e83a9 | 1656 | else |
338fa5f7 | 1657 | list->directive = _cpp_check_directive (pfile, cur_token); |
f80e83a9 | 1658 | } |
cfad5579 | 1659 | |
338fa5f7 | 1660 | /* _cpp_get_line assumes list->tokens_used refers to the current |
1661 | token being lexed. So do this after _cpp_check_directive to | |
1662 | get the warnings therein correct. */ | |
1663 | list->tokens_used++; | |
f80e83a9 | 1664 | } |
338fa5f7 | 1665 | while (cur_token->type != CPP_EOF); |
cfad5579 | 1666 | |
f80e83a9 | 1667 | /* All tokens are allocated, so the memory location is fixed. */ |
1668 | first = &list->tokens[first_token]; | |
338fa5f7 | 1669 | first->flags |= BOL; |
1670 | pfile->first_directive_token = first; | |
1671 | ||
f80e83a9 | 1672 | /* Don't complain about the null directive, nor directives in |
1673 | assembly source: we don't know where the comments are, and # may | |
1674 | introduce assembler pseudo-ops. Don't complain about invalid | |
1675 | directives in skipped conditional groups (6.10 p4). */ | |
338fa5f7 | 1676 | if (IN_DIRECTIVE (pfile) && !KNOWN_DIRECTIVE (list) && !pfile->skipping |
1677 | && !CPP_OPTION (pfile, lang_asm)) | |
f80e83a9 | 1678 | { |
338fa5f7 | 1679 | if (cur_token > first + 1) |
1680 | { | |
1681 | if (first[1].type == CPP_NAME) | |
1682 | cpp_error_with_line (pfile, first->line, first->col, | |
1683 | "invalid preprocessing directive #%s", | |
1684 | first[1].val.node->name); | |
1685 | else | |
1686 | cpp_error_with_line (pfile, first->line, first->col, | |
1687 | "invalid preprocessing directive"); | |
1688 | } | |
0653b94e | 1689 | |
1690 | /* Discard this line to prevent further errors from cc1. */ | |
1691 | _cpp_clear_toklist (list); | |
1692 | goto retry; | |
f80e83a9 | 1693 | } |
1694 | ||
be6e77fe | 1695 | /* Drop the EOF unless really at EOF or in a directive. */ |
1696 | if (cur_token != first && !KNOWN_DIRECTIVE (list) | |
1697 | && pfile->done_initializing) | |
1698 | list->tokens_used--; | |
1699 | ||
338fa5f7 | 1700 | pfile->state.in_lex_line = 0; |
cfad5579 | 1701 | } |
1702 | ||
f80e83a9 | 1703 | /* Write the spelling of a token TOKEN, with any appropriate |
6cae2504 | 1704 | whitespace before it, to FP. PREV is the previous token, which |
1705 | is used to determine if we need to shove in an extra space in order | |
1706 | to avoid accidental token paste. If WHITE is 0, do not insert any | |
1707 | leading whitespace. */ | |
f80e83a9 | 1708 | static void |
6cae2504 | 1709 | output_token (pfile, fp, token, prev, white) |
f80e83a9 | 1710 | cpp_reader *pfile; |
6cae2504 | 1711 | FILE *fp; |
f80e83a9 | 1712 | const cpp_token *token, *prev; |
6cae2504 | 1713 | int white; |
f80e83a9 | 1714 | { |
6cae2504 | 1715 | if (white) |
f80e83a9 | 1716 | { |
6cae2504 | 1717 | int dummy; |
1718 | ||
1719 | if (token->col && (token->flags & BOL)) | |
1720 | { | |
1721 | /* Supply enough whitespace to put this token in its original | |
1722 | column. Don't bother trying to reconstruct tabs; we can't | |
1723 | get it right in general, and nothing ought to care. (Yes, | |
1724 | some things do care; the fault lies with them.) */ | |
1725 | unsigned int spaces = token->col - 1; | |
1726 | ||
1727 | while (spaces--) | |
1728 | putc (' ', fp); | |
1729 | } | |
1730 | else if (token->flags & PREV_WHITE) | |
1731 | putc (' ', fp); | |
1732 | else | |
1733 | /* Check for and prevent accidental token pasting. | |
1734 | In addition to the cases handled by can_paste, consider | |
1735 | ||
1736 | a + ++b - if there is not a space between the + and ++, it | |
1737 | will be misparsed as a++ + b. But + ## ++ doesn't produce | |
1738 | a valid token. */ | |
1739 | if (prev | |
1740 | && (can_paste (pfile, prev, token, &dummy) != CPP_EOF | |
1741 | || (prev->type == CPP_PLUS && token->type == CPP_PLUS_PLUS) | |
1742 | || (prev->type == CPP_MINUS && token->type == CPP_MINUS_MINUS))) | |
1743 | putc (' ', fp); | |
f80e83a9 | 1744 | } |
6cae2504 | 1745 | |
1746 | switch (TOKEN_SPELL (token)) | |
f80e83a9 | 1747 | { |
6cae2504 | 1748 | case SPELL_OPERATOR: |
1749 | { | |
1750 | const unsigned char *spelling; | |
1751 | ||
1752 | if (token->flags & DIGRAPH) | |
1753 | spelling = digraph_spellings[token->type - CPP_FIRST_DIGRAPH]; | |
1754 | else if (token->flags & NAMED_OP) | |
1755 | goto spell_ident; | |
1756 | else | |
1757 | spelling = TOKEN_NAME (token); | |
1758 | ||
1759 | ufputs (spelling, fp); | |
1760 | } | |
1761 | break; | |
1762 | ||
1763 | case SPELL_IDENT: | |
1764 | spell_ident: | |
1765 | ufputs (token->val.node->name, fp); | |
1766 | break; | |
1767 | ||
1768 | case SPELL_STRING: | |
1769 | { | |
71aa9da4 | 1770 | int left, right, tag; |
1771 | switch (token->type) | |
1772 | { | |
1773 | case CPP_STRING: left = '"'; right = '"'; tag = '\0'; break; | |
1774 | case CPP_WSTRING: left = '"'; right = '"'; tag = 'L'; break; | |
1775 | case CPP_OSTRING: left = '"'; right = '"'; tag = '@'; break; | |
1776 | case CPP_CHAR: left = '\''; right = '\''; tag = '\0'; break; | |
1777 | case CPP_WCHAR: left = '\''; right = '\''; tag = 'L'; break; | |
1778 | case CPP_HEADER_NAME: left = '<'; right = '>'; tag = '\0'; break; | |
1779 | default: left = '\0'; right = '\0'; tag = '\0'; break; | |
1780 | } | |
1781 | if (tag) putc (tag, fp); | |
1782 | if (left) putc (left, fp); | |
6cae2504 | 1783 | fwrite (token->val.str.text, 1, token->val.str.len, fp); |
71aa9da4 | 1784 | if (right) putc (right, fp); |
6cae2504 | 1785 | } |
1786 | break; | |
1787 | ||
1788 | case SPELL_CHAR: | |
1789 | putc (token->val.aux, fp); | |
1790 | break; | |
1791 | ||
1792 | case SPELL_NONE: | |
1793 | /* Placemarker or EOF - no output. (Macro args are handled | |
1794 | elsewhere. */ | |
1795 | break; | |
f80e83a9 | 1796 | } |
6cae2504 | 1797 | } |
1798 | ||
1799 | /* Dump the original user's spelling of argument index ARG_NO to the | |
1800 | macro whose expansion is LIST. */ | |
1801 | static void | |
1802 | dump_param_spelling (fp, list, arg_no) | |
1803 | FILE *fp; | |
1804 | const cpp_toklist *list; | |
1805 | unsigned int arg_no; | |
1806 | { | |
1807 | const U_CHAR *param = list->namebuf; | |
1808 | ||
1809 | while (arg_no--) | |
1810 | param += ustrlen (param) + 1; | |
1811 | ufputs (param, fp); | |
1812 | } | |
1813 | ||
1814 | /* Output all the tokens of LIST, starting at TOKEN, to FP. */ | |
1815 | void | |
1816 | cpp_output_list (pfile, fp, list, token) | |
1817 | cpp_reader *pfile; | |
1818 | FILE *fp; | |
1819 | const cpp_toklist *list; | |
1820 | const cpp_token *token; | |
1821 | { | |
1822 | const cpp_token *limit = list->tokens + list->tokens_used; | |
1823 | const cpp_token *prev = 0; | |
1824 | int white = 0; | |
ab12a39c | 1825 | |
6cae2504 | 1826 | while (token < limit) |
1827 | { | |
1828 | /* XXX Find some way we can write macro args from inside | |
1829 | output_token/spell_token. */ | |
1830 | if (token->type == CPP_MACRO_ARG) | |
1831 | { | |
1832 | if (white && token->flags & PREV_WHITE) | |
1833 | putc (' ', fp); | |
1834 | if (token->flags & STRINGIFY_ARG) | |
1835 | putc ('#', fp); | |
1836 | dump_param_spelling (fp, list, token->val.aux); | |
1837 | } | |
1838 | else | |
1839 | output_token (pfile, fp, token, prev, white); | |
1840 | if (token->flags & PASTE_LEFT) | |
1841 | fputs (" ##", fp); | |
1842 | prev = token; | |
1843 | token++; | |
1844 | white = 1; | |
1845 | } | |
f80e83a9 | 1846 | } |
ab12a39c | 1847 | |
6cae2504 | 1848 | |
f80e83a9 | 1849 | /* Write the spelling of a token TOKEN to BUFFER. The buffer must |
c5ea33a8 | 1850 | already contain the enough space to hold the token's spelling. |
1851 | Returns a pointer to the character after the last character | |
1852 | written. */ | |
ab12a39c | 1853 | |
f80e83a9 | 1854 | static unsigned char * |
1855 | spell_token (pfile, token, buffer) | |
1856 | cpp_reader *pfile; /* Would be nice to be rid of this... */ | |
1857 | const cpp_token *token; | |
1858 | unsigned char *buffer; | |
1859 | { | |
7e842f95 | 1860 | switch (TOKEN_SPELL (token)) |
f80e83a9 | 1861 | { |
1862 | case SPELL_OPERATOR: | |
1863 | { | |
1864 | const unsigned char *spelling; | |
1865 | unsigned char c; | |
ab12a39c | 1866 | |
f80e83a9 | 1867 | if (token->flags & DIGRAPH) |
1868 | spelling = digraph_spellings[token->type - CPP_FIRST_DIGRAPH]; | |
31674461 | 1869 | else if (token->flags & NAMED_OP) |
1870 | goto spell_ident; | |
f80e83a9 | 1871 | else |
7e842f95 | 1872 | spelling = TOKEN_NAME (token); |
f80e83a9 | 1873 | |
1874 | while ((c = *spelling++) != '\0') | |
1875 | *buffer++ = c; | |
1876 | } | |
1877 | break; | |
ab12a39c | 1878 | |
f80e83a9 | 1879 | case SPELL_IDENT: |
31674461 | 1880 | spell_ident: |
76faa4c0 | 1881 | memcpy (buffer, token->val.node->name, token->val.node->length); |
1882 | buffer += token->val.node->length; | |
f80e83a9 | 1883 | break; |
ab12a39c | 1884 | |
f80e83a9 | 1885 | case SPELL_STRING: |
1886 | { | |
71aa9da4 | 1887 | int left, right, tag; |
1888 | switch (token->type) | |
1889 | { | |
1890 | case CPP_STRING: left = '"'; right = '"'; tag = '\0'; break; | |
1891 | case CPP_WSTRING: left = '"'; right = '"'; tag = 'L'; break; | |
1892 | case CPP_OSTRING: left = '"'; right = '"'; tag = '@'; break; | |
1893 | case CPP_CHAR: left = '\''; right = '\''; tag = '\0'; break; | |
1894 | case CPP_WCHAR: left = '\''; right = '\''; tag = 'L'; break; | |
1895 | case CPP_HEADER_NAME: left = '<'; right = '>'; tag = '\0'; break; | |
1896 | default: left = '\0'; right = '\0'; tag = '\0'; break; | |
1897 | } | |
1898 | if (tag) *buffer++ = tag; | |
1899 | if (left) *buffer++ = left; | |
76faa4c0 | 1900 | memcpy (buffer, token->val.str.text, token->val.str.len); |
1901 | buffer += token->val.str.len; | |
71aa9da4 | 1902 | if (right) *buffer++ = right; |
f80e83a9 | 1903 | } |
1904 | break; | |
ab12a39c | 1905 | |
f80e83a9 | 1906 | case SPELL_CHAR: |
1907 | *buffer++ = token->val.aux; | |
1908 | break; | |
ab12a39c | 1909 | |
f80e83a9 | 1910 | case SPELL_NONE: |
7e842f95 | 1911 | cpp_ice (pfile, "Unspellable token %s", TOKEN_NAME (token)); |
f80e83a9 | 1912 | break; |
1913 | } | |
ab12a39c | 1914 | |
f80e83a9 | 1915 | return buffer; |
1916 | } | |
ab12a39c | 1917 | |
fdf3a98f | 1918 | /* Macro expansion algorithm. |
1919 | ||
1920 | Macro expansion is implemented by a single-pass algorithm; there are | |
1921 | no rescan passes involved. cpp_get_token expands just enough to be | |
1922 | able to return a token to the caller, a consequence is that when it | |
1923 | returns the preprocessor can be in a state of mid-expansion. The | |
1924 | algorithm does not work by fully expanding a macro invocation into | |
1925 | some kind of token list, and then returning them one by one. | |
1926 | ||
1927 | Our expansion state is recorded in a context stack. We start out with | |
1928 | a single context on the stack, let's call it base context. This | |
1929 | consists of the token list returned by lex_line that forms the next | |
1930 | logical line in the source file. | |
1931 | ||
1932 | The current level in the context stack is stored in the cur_context | |
1933 | member of the cpp_reader structure. The context it references keeps, | |
1934 | amongst other things, a count of how many tokens form that context and | |
1935 | our position within those tokens. | |
1936 | ||
1937 | Fundamentally, calling cpp_get_token will return the next token from | |
1938 | the current context. If we're at the end of the current context, that | |
1939 | context is popped from the stack first, unless it is the base context, | |
1940 | in which case the next logical line is lexed from the source file. | |
1941 | ||
1942 | However, before returning the token, if it is a CPP_NAME token | |
1943 | _cpp_get_token checks to see if it is a macro and if it is enabled. | |
1944 | Each time it encounters a macro name, it calls push_macro_context. | |
1945 | This function checks that the macro should be expanded (with | |
1946 | is_macro_enabled), and if so pushes a new macro context on the stack | |
1947 | which becomes the current context. It then loops back to read the | |
1948 | first token of the macro context. | |
1949 | ||
1950 | A macro context basically consists of the token list representing the | |
1951 | macro's replacement list, which was saved in the hash table by | |
1952 | save_macro_expansion when its #define statement was parsed. If the | |
1953 | macro is function-like, it also contains the tokens that form the | |
1954 | arguments to the macro. I say more about macro arguments below, but | |
1955 | for now just saying that each argument is a set of pointers to tokens | |
1956 | is enough. | |
1957 | ||
1958 | When taking tokens from a macro context, we may get a CPP_MACRO_ARG | |
1959 | token. This represents an argument passed to the macro, with the | |
1960 | argument number stored in the token's AUX field. The argument should | |
1961 | be substituted, this is achieved by pushing an "argument context". An | |
1962 | argument context is just refers to the tokens forming the argument, | |
1963 | which are obtained directly from the macro context. The STRINGIFY | |
1964 | flag on a CPP_MACRO_ARG token indicates that the argument should be | |
1965 | stringified. | |
1966 | ||
1967 | Here's a few simple rules the context stack obeys:- | |
1968 | ||
1969 | 1) The lex_line token list is always context zero. | |
1970 | ||
1971 | 2) Context 1, if it exists, must be a macro context. | |
1972 | ||
1973 | 3) An argument context can only appear above a macro context. | |
1974 | ||
1975 | 4) A macro context can appear above the base context, another macro | |
1976 | context, or an argument context. | |
1977 | ||
1978 | 5) These imply that the minimal level of an argument context is 2. | |
1979 | ||
1980 | The only tricky thing left is ensuring that macros are enabled and | |
1981 | disabled correctly. The algorithm controls macro expansion by the | |
1982 | level of the context a token is taken from in the context stack. If a | |
1983 | token is taken from a level equal to no_expand_level (a member of | |
1984 | struct cpp_reader), no expansion is performed. | |
1985 | ||
1986 | When popping a context off the stack, if no_expand_level equals the | |
1987 | level of the popped context, it is reduced by one to match the new | |
1988 | context level, so that expansion is still disabled. It does not | |
1989 | increase if a context is pushed, though. It starts out life as | |
1990 | UINT_MAX, which has the effect that initially macro expansion is | |
1991 | enabled. I explain how this mechanism works below. | |
1992 | ||
1993 | The standard requires:- | |
1994 | ||
1995 | 1) Arguments to be fully expanded before substitution. | |
1996 | ||
1997 | 2) Stringified arguments to not be expanded, nor the tokens | |
1998 | immediately surrounding a ## operator. | |
1999 | ||
2000 | 3) Continual rescanning until there are no more macros left to | |
2001 | replace. | |
2002 | ||
2003 | 4) Once a macro has been expanded in stage 1) or 3), it cannot be | |
2004 | expanded again during later rescans. This prevents infinite | |
2005 | recursion. | |
2006 | ||
2007 | The first thing to observe is that stage 3) is mostly redundant. | |
2008 | Since a macro is disabled once it has been expanded, how can a rescan | |
2009 | find an unexpanded macro name? There are only two cases where this is | |
2010 | possible:- | |
2011 | ||
2012 | a) If the macro name results from a token paste operation. | |
2013 | ||
2014 | b) If the macro in question is a function-like macro that hasn't | |
2015 | already been expanded because previously there was not the required | |
2016 | '(' token immediately following it. This is only possible when an | |
2017 | argument is substituted, and after substitution the last token of | |
2018 | the argument can bind with a parenthesis appearing in the tokens | |
2019 | following the substitution. Note that if the '(' appears within the | |
2020 | argument, the ')' must too, as expanding macro arguments cannot | |
2021 | "suck in" tokens outside the argument. | |
2022 | ||
2023 | So we tackle this as follows. When parsing the macro invocation for | |
2024 | arguments, we record the tokens forming each argument as a list of | |
2025 | pointers to those tokens. We do not expand any tokens that are "raw", | |
2026 | i.e. directly from the macro invocation, but other tokens that come | |
2027 | from (nested) argument substitution are fully expanded. | |
2028 | ||
2029 | This is achieved by setting the no_expand_level to that of the macro | |
2030 | invocation. A CPP_MACRO_ARG token never appears in the list of tokens | |
2031 | forming an argument, because parse_args (indirectly) calls | |
2032 | get_raw_token which automatically pushes argument contexts and traces | |
2033 | into them. Since these contexts are at a higher level than the | |
2034 | no_expand_level, they get fully macro expanded. | |
2035 | ||
2036 | "Raw" and non-raw tokens are separated in arguments by null pointers, | |
2037 | with the policy that the initial state of an argument is raw. If the | |
2038 | first token is not raw, it should be preceded by a null pointer. When | |
2039 | tracing through the tokens of an argument context, each time | |
2040 | get_raw_token encounters a null pointer, it toggles the flag | |
2041 | CONTEXT_RAW. | |
2042 | ||
2043 | This flag, when set, indicates to is_macro_disabled that we are | |
2044 | reading raw tokens which should be macro-expanded. Similarly, if | |
2045 | clear, is_macro_disabled suppresses re-expansion. | |
2046 | ||
2047 | It's probably time for an example. | |
2048 | ||
2049 | #define hash # | |
2050 | #define str(x) #x | |
2051 | #define xstr(y) str(y hash) | |
2052 | str(hash) // "hash" | |
2053 | xstr(hash) // "# hash" | |
2054 | ||
2055 | In the invocation of str, parse_args turns off macro expansion and so | |
2056 | parses the argument as <hash>. This is the only token (pointer) | |
2057 | passed as the argument to str. Since <hash> is raw there is no need | |
2058 | for an initial null pointer. stringify_arg is called from | |
2059 | get_raw_token when tracing through the expansion of str, since the | |
2060 | argument has the STRINGIFY flag set. stringify_arg turns off | |
2061 | macro_expansion by setting the no_expand_level to that of the argument | |
2062 | context. Thus it gets the token <hash> and stringifies it to "hash" | |
2063 | correctly. | |
2064 | ||
2065 | Similary xstr is passed <hash>. However, when parse_args is parsing | |
2066 | the invocation of str() in xstr's expansion, get_raw_token encounters | |
2067 | a CPP_MACRO_ARG token for y. Transparently to parse_args, it pushes | |
2068 | an argument context, and enters the tokens of the argument, | |
2069 | i.e. <hash>. This is at a higher context level than parse_args | |
2070 | disabled, and so is_macro_disabled permits expansion of it and a macro | |
2071 | context is pushed on top of the argument context. This contains the | |
2072 | <#> token, and the end result is that <hash> is macro expanded. | |
2073 | However, after popping off the argument context, the <hash> of xstr's | |
2074 | expansion does not get macro expanded because we're back at the | |
2075 | no_expand_level. The end result is that the argument passed to str is | |
2076 | <NULL> <#> <NULL> <hash>. Note the nulls - policy is we start off | |
2077 | raw, <#> is not raw, but then <hash> is. | |
2078 | ||
2079 | */ | |
ab12a39c | 2080 | |
f80e83a9 | 2081 | |
2082 | /* Free the storage allocated for macro arguments. */ | |
2083 | static void | |
2084 | free_macro_args (args) | |
2085 | macro_args *args; | |
6060326b | 2086 | { |
f80e83a9 | 2087 | if (args->tokens) |
deb356cf | 2088 | free ((PTR) args->tokens); |
f80e83a9 | 2089 | free (args->ends); |
2090 | free (args); | |
6060326b | 2091 | } |
2092 | ||
f80e83a9 | 2093 | /* Determines if a macro has been already used (and is therefore |
2094 | disabled). */ | |
6060326b | 2095 | static int |
f80e83a9 | 2096 | is_macro_disabled (pfile, expansion, token) |
6060326b | 2097 | cpp_reader *pfile; |
f80e83a9 | 2098 | const cpp_toklist *expansion; |
2099 | const cpp_token *token; | |
6060326b | 2100 | { |
f80e83a9 | 2101 | cpp_context *context = CURRENT_CONTEXT (pfile); |
2102 | ||
2103 | /* Arguments on either side of ## are inserted in place without | |
2104 | macro expansion (6.10.3.3.2). Conceptually, any macro expansion | |
2105 | occurs during a later rescan pass. The effect is that we expand | |
2106 | iff we would as part of the macro's expansion list, so we should | |
2107 | drop to the macro's context. */ | |
2108 | if (IS_ARG_CONTEXT (context)) | |
6060326b | 2109 | { |
f80e83a9 | 2110 | if (token->flags & PASTED) |
2111 | context--; | |
2112 | else if (!(context->flags & CONTEXT_RAW)) | |
2113 | return 1; | |
2114 | else if (context->flags & (CONTEXT_PASTEL | CONTEXT_PASTER)) | |
2115 | context--; | |
6060326b | 2116 | } |
6060326b | 2117 | |
f80e83a9 | 2118 | /* Have we already used this macro? */ |
2119 | while (context->level > 0) | |
6060326b | 2120 | { |
f80e83a9 | 2121 | if (!IS_ARG_CONTEXT (context) && context->u.list == expansion) |
2122 | return 1; | |
2123 | /* Raw argument tokens are judged based on the token list they | |
2124 | came from. */ | |
2125 | if (context->flags & CONTEXT_RAW) | |
2126 | context = pfile->contexts + context->level; | |
2127 | else | |
2128 | context--; | |
2129 | } | |
6060326b | 2130 | |
f80e83a9 | 2131 | /* Function-like macros may be disabled if the '(' is not in the |
2132 | current context. We check this without disrupting the context | |
2133 | stack. */ | |
2134 | if (expansion->paramc >= 0) | |
2135 | { | |
2136 | const cpp_token *next; | |
2137 | unsigned int prev_nme; | |
6060326b | 2138 | |
f80e83a9 | 2139 | context = CURRENT_CONTEXT (pfile); |
2140 | /* Drop down any contexts we're at the end of: the '(' may | |
2141 | appear in lower macro expansions, or in the rest of the file. */ | |
2142 | while (context->posn == context->count && context > pfile->contexts) | |
2143 | { | |
2144 | context--; | |
2145 | /* If we matched, we are disabled, as we appear in the | |
2146 | expansion of each macro we meet. */ | |
2147 | if (!IS_ARG_CONTEXT (context) && context->u.list == expansion) | |
2148 | return 1; | |
2149 | } | |
6060326b | 2150 | |
f80e83a9 | 2151 | prev_nme = pfile->no_expand_level; |
2152 | pfile->no_expand_level = context - pfile->contexts; | |
deb356cf | 2153 | next = _cpp_get_token (pfile); |
f80e83a9 | 2154 | restore_macro_expansion (pfile, prev_nme); |
2155 | if (next->type != CPP_OPEN_PAREN) | |
2156 | { | |
2157 | _cpp_push_token (pfile, next); | |
1449850f | 2158 | if (CPP_WTRADITIONAL (pfile)) |
f80e83a9 | 2159 | cpp_warning (pfile, |
31674461 | 2160 | "function macro %s must be used with arguments in traditional C", |
2161 | token->val.node->name); | |
f80e83a9 | 2162 | return 1; |
2163 | } | |
6060326b | 2164 | } |
6060326b | 2165 | |
f80e83a9 | 2166 | return 0; |
6060326b | 2167 | } |
2168 | ||
f80e83a9 | 2169 | /* Add a token to the set of tokens forming the arguments to the macro |
2170 | being parsed in parse_args. */ | |
2171 | static void | |
2172 | save_token (args, token) | |
2173 | macro_args *args; | |
2174 | const cpp_token *token; | |
6060326b | 2175 | { |
f80e83a9 | 2176 | if (args->used == args->capacity) |
2177 | { | |
2178 | args->capacity += args->capacity + 100; | |
2179 | args->tokens = (const cpp_token **) | |
deb356cf | 2180 | xrealloc ((PTR) args->tokens, |
2181 | args->capacity * sizeof (const cpp_token *)); | |
f80e83a9 | 2182 | } |
2183 | args->tokens[args->used++] = token; | |
6060326b | 2184 | } |
2185 | ||
f80e83a9 | 2186 | /* Take and save raw tokens until we finish one argument. Empty |
2187 | arguments are saved as a single CPP_PLACEMARKER token. */ | |
2188 | static const cpp_token * | |
2189 | parse_arg (pfile, var_args, paren_context, args, pcount) | |
6060326b | 2190 | cpp_reader *pfile; |
f80e83a9 | 2191 | int var_args; |
2192 | unsigned int paren_context; | |
2193 | macro_args *args; | |
2194 | unsigned int *pcount; | |
6060326b | 2195 | { |
f80e83a9 | 2196 | const cpp_token *token; |
2197 | unsigned int paren = 0, count = 0; | |
2198 | int raw, was_raw = 1; | |
6060326b | 2199 | |
f80e83a9 | 2200 | for (count = 0;; count++) |
6060326b | 2201 | { |
deb356cf | 2202 | token = _cpp_get_token (pfile); |
6060326b | 2203 | |
f80e83a9 | 2204 | switch (token->type) |
6060326b | 2205 | { |
f80e83a9 | 2206 | default: |
2207 | break; | |
6060326b | 2208 | |
f80e83a9 | 2209 | case CPP_OPEN_PAREN: |
2210 | paren++; | |
2211 | break; | |
2212 | ||
2213 | case CPP_CLOSE_PAREN: | |
2214 | if (paren-- != 0) | |
2215 | break; | |
2216 | goto out; | |
2217 | ||
2218 | case CPP_COMMA: | |
2219 | /* Commas are not terminators within parantheses or var_args. */ | |
2220 | if (paren || var_args) | |
2221 | break; | |
2222 | goto out; | |
2223 | ||
2224 | case CPP_EOF: /* Error reported by caller. */ | |
2225 | goto out; | |
6060326b | 2226 | } |
6060326b | 2227 | |
f80e83a9 | 2228 | raw = pfile->cur_context <= paren_context; |
2229 | if (raw != was_raw) | |
2230 | { | |
2231 | was_raw = raw; | |
2232 | save_token (args, 0); | |
2233 | count++; | |
6060326b | 2234 | } |
f80e83a9 | 2235 | save_token (args, token); |
6060326b | 2236 | } |
6060326b | 2237 | |
2238 | out: | |
f80e83a9 | 2239 | if (count == 0) |
2240 | { | |
2241 | /* Duplicate the placemarker. Then we can set its flags and | |
2242 | position and safely be using more than one. */ | |
2243 | save_token (args, duplicate_token (pfile, &placemarker_token)); | |
2244 | count++; | |
2245 | } | |
2246 | ||
2247 | *pcount = count; | |
2248 | return token; | |
6060326b | 2249 | } |
2250 | ||
f80e83a9 | 2251 | /* This macro returns true if the argument starting at offset O of arglist |
2252 | A is empty - that is, it's either a single PLACEMARKER token, or a null | |
2253 | pointer followed by a PLACEMARKER. */ | |
2254 | ||
2255 | #define empty_argument(A, O) \ | |
2256 | ((A)->tokens[O] ? (A)->tokens[O]->type == CPP_PLACEMARKER \ | |
2257 | : (A)->tokens[(O)+1]->type == CPP_PLACEMARKER) | |
2258 | ||
2259 | /* Parse the arguments making up a macro invocation. Nested arguments | |
2260 | are automatically macro expanded, but immediate macros are not | |
2261 | expanded; this enables e.g. operator # to work correctly. Returns | |
2262 | non-zero on error. */ | |
6060326b | 2263 | static int |
f80e83a9 | 2264 | parse_args (pfile, hp, args) |
6060326b | 2265 | cpp_reader *pfile; |
f80e83a9 | 2266 | cpp_hashnode *hp; |
2267 | macro_args *args; | |
6060326b | 2268 | { |
f80e83a9 | 2269 | const cpp_token *token; |
2270 | const cpp_toklist *macro; | |
2271 | unsigned int total = 0; | |
2272 | unsigned int paren_context = pfile->cur_context; | |
2273 | int argc = 0; | |
2274 | ||
2275 | macro = hp->value.expansion; | |
2276 | do | |
6060326b | 2277 | { |
f80e83a9 | 2278 | unsigned int count; |
6060326b | 2279 | |
f80e83a9 | 2280 | token = parse_arg (pfile, (argc + 1 == macro->paramc |
2281 | && (macro->flags & VAR_ARGS)), | |
2282 | paren_context, args, &count); | |
2283 | if (argc < macro->paramc) | |
6060326b | 2284 | { |
f80e83a9 | 2285 | total += count; |
2286 | args->ends[argc] = total; | |
6060326b | 2287 | } |
f80e83a9 | 2288 | argc++; |
6060326b | 2289 | } |
f80e83a9 | 2290 | while (token->type != CPP_CLOSE_PAREN && token->type != CPP_EOF); |
6060326b | 2291 | |
f80e83a9 | 2292 | if (token->type == CPP_EOF) |
2293 | { | |
31674461 | 2294 | cpp_error(pfile, "unterminated argument list for macro \"%s\"", hp->name); |
f80e83a9 | 2295 | return 1; |
2296 | } | |
2297 | else if (argc < macro->paramc) | |
2298 | { | |
2299 | /* A rest argument is allowed to not appear in the invocation at all. | |
2300 | e.g. #define debug(format, args...) ... | |
2301 | debug("string"); | |
2302 | This is exactly the same as if the rest argument had received no | |
57ba19ba | 2303 | tokens - debug("string",); This extension is deprecated. */ |
35677230 | 2304 | |
2305 | if (argc + 1 == macro->paramc && (macro->flags & VAR_ARGS)) | |
f80e83a9 | 2306 | { |
2307 | /* Duplicate the placemarker. Then we can set its flags and | |
2308 | position and safely be using more than one. */ | |
35677230 | 2309 | cpp_token *pm = duplicate_token (pfile, &placemarker_token); |
2310 | pm->flags = VOID_REST; | |
2311 | save_token (args, pm); | |
f80e83a9 | 2312 | args->ends[argc] = total + 1; |
35677230 | 2313 | |
2314 | if (CPP_OPTION (pfile, c99) && CPP_PEDANTIC (pfile)) | |
2315 | cpp_pedwarn (pfile, "ISO C99 requires rest arguments to be used"); | |
2316 | ||
f80e83a9 | 2317 | return 0; |
2318 | } | |
2319 | else | |
2320 | { | |
31674461 | 2321 | cpp_error (pfile, "not enough arguments for macro \"%s\"", hp->name); |
f80e83a9 | 2322 | return 1; |
2323 | } | |
2324 | } | |
2325 | /* An empty argument to an empty function-like macro is fine. */ | |
2326 | else if (argc > macro->paramc | |
2327 | && !(macro->paramc == 0 && argc == 1 && empty_argument (args, 0))) | |
2328 | { | |
31674461 | 2329 | cpp_error (pfile, "too many arguments for macro \"%s\"", hp->name); |
f80e83a9 | 2330 | return 1; |
2331 | } | |
6060326b | 2332 | |
f80e83a9 | 2333 | return 0; |
2334 | } | |
2335 | ||
2336 | /* Adds backslashes before all backslashes and double quotes appearing | |
2337 | in strings. Non-printable characters are converted to octal. */ | |
2338 | static U_CHAR * | |
2339 | quote_string (dest, src, len) | |
2340 | U_CHAR *dest; | |
2341 | const U_CHAR *src; | |
2342 | unsigned int len; | |
2343 | { | |
2344 | while (len--) | |
6060326b | 2345 | { |
f80e83a9 | 2346 | U_CHAR c = *src++; |
6060326b | 2347 | |
f80e83a9 | 2348 | if (c == '\\' || c == '"') |
3b304865 | 2349 | { |
f80e83a9 | 2350 | *dest++ = '\\'; |
2351 | *dest++ = c; | |
2352 | } | |
2353 | else | |
2354 | { | |
2355 | if (ISPRINT (c)) | |
2356 | *dest++ = c; | |
2357 | else | |
2358 | { | |
2359 | sprintf ((char *) dest, "\\%03o", c); | |
2360 | dest += 4; | |
2361 | } | |
3b304865 | 2362 | } |
6060326b | 2363 | } |
6060326b | 2364 | |
f80e83a9 | 2365 | return dest; |
6060326b | 2366 | } |
2367 | ||
f80e83a9 | 2368 | /* Allocates a buffer to hold a token's TEXT, and converts TOKEN to a |
2369 | CPP_STRING token containing TEXT in quoted form. */ | |
2370 | static cpp_token * | |
2371 | make_string_token (token, text, len) | |
2372 | cpp_token *token; | |
2373 | const U_CHAR *text; | |
2374 | unsigned int len; | |
2375 | { | |
2376 | U_CHAR *buf; | |
2377 | ||
2378 | buf = (U_CHAR *) xmalloc (len * 4); | |
2379 | token->type = CPP_STRING; | |
2380 | token->flags = 0; | |
76faa4c0 | 2381 | token->val.str.text = buf; |
2382 | token->val.str.len = quote_string (buf, text, len) - buf; | |
f80e83a9 | 2383 | return token; |
2384 | } | |
2385 | ||
2386 | /* Allocates and converts a temporary token to a CPP_NUMBER token, | |
2387 | evaluating to NUMBER. */ | |
2388 | static cpp_token * | |
2389 | alloc_number_token (pfile, number) | |
6060326b | 2390 | cpp_reader *pfile; |
f80e83a9 | 2391 | int number; |
6060326b | 2392 | { |
f80e83a9 | 2393 | cpp_token *result; |
2394 | char *buf; | |
2395 | ||
2396 | result = get_temp_token (pfile); | |
2397 | buf = xmalloc (20); | |
2398 | sprintf (buf, "%d", number); | |
2399 | ||
2400 | result->type = CPP_NUMBER; | |
2401 | result->flags = 0; | |
76faa4c0 | 2402 | result->val.str.text = (U_CHAR *) buf; |
2403 | result->val.str.len = strlen (buf); | |
f80e83a9 | 2404 | return result; |
2405 | } | |
6060326b | 2406 | |
f80e83a9 | 2407 | /* Returns a temporary token from the temporary token store of PFILE. */ |
2408 | static cpp_token * | |
2409 | get_temp_token (pfile) | |
2410 | cpp_reader *pfile; | |
2411 | { | |
2412 | if (pfile->temp_used == pfile->temp_alloced) | |
2413 | { | |
2414 | if (pfile->temp_used == pfile->temp_cap) | |
2415 | { | |
2416 | pfile->temp_cap += pfile->temp_cap + 20; | |
2417 | pfile->temp_tokens = (cpp_token **) xrealloc | |
2418 | (pfile->temp_tokens, pfile->temp_cap * sizeof (cpp_token *)); | |
2419 | } | |
2420 | pfile->temp_tokens[pfile->temp_alloced++] = (cpp_token *) xmalloc | |
2421 | (sizeof (cpp_token)); | |
2422 | } | |
6060326b | 2423 | |
f80e83a9 | 2424 | return pfile->temp_tokens[pfile->temp_used++]; |
2425 | } | |
2426 | ||
2427 | /* Release (not free) for re-use the temporary tokens of PFILE. */ | |
2428 | static void | |
2429 | release_temp_tokens (pfile) | |
2430 | cpp_reader *pfile; | |
2431 | { | |
2432 | while (pfile->temp_used) | |
6060326b | 2433 | { |
f80e83a9 | 2434 | cpp_token *token = pfile->temp_tokens[--pfile->temp_used]; |
6060326b | 2435 | |
7e842f95 | 2436 | if (TOKEN_SPELL (token) == SPELL_STRING) |
6060326b | 2437 | { |
76faa4c0 | 2438 | free ((char *) token->val.str.text); |
2439 | token->val.str.text = 0; | |
6060326b | 2440 | } |
2441 | } | |
f80e83a9 | 2442 | } |
6060326b | 2443 | |
f80e83a9 | 2444 | /* Free all of PFILE's dynamically-allocated temporary tokens. */ |
2445 | void | |
2446 | _cpp_free_temp_tokens (pfile) | |
2447 | cpp_reader *pfile; | |
2448 | { | |
2449 | if (pfile->temp_tokens) | |
6060326b | 2450 | { |
f80e83a9 | 2451 | /* It is possible, though unlikely (looking for '(' of a funlike |
2452 | macro into EOF), that we haven't released the tokens yet. */ | |
2453 | release_temp_tokens (pfile); | |
2454 | while (pfile->temp_alloced) | |
2455 | free (pfile->temp_tokens[--pfile->temp_alloced]); | |
2456 | free (pfile->temp_tokens); | |
6060326b | 2457 | } |
2458 | ||
f80e83a9 | 2459 | if (pfile->date) |
2460 | { | |
76faa4c0 | 2461 | free ((char *) pfile->date->val.str.text); |
f80e83a9 | 2462 | free (pfile->date); |
76faa4c0 | 2463 | free ((char *) pfile->time->val.str.text); |
f80e83a9 | 2464 | free (pfile->time); |
2465 | } | |
6060326b | 2466 | } |
2467 | ||
f80e83a9 | 2468 | /* Copy TOKEN into a temporary token from PFILE's store. */ |
2469 | static cpp_token * | |
2470 | duplicate_token (pfile, token) | |
2471 | cpp_reader *pfile; | |
2472 | const cpp_token *token; | |
2473 | { | |
2474 | cpp_token *result = get_temp_token (pfile); | |
6060326b | 2475 | |
f80e83a9 | 2476 | *result = *token; |
7e842f95 | 2477 | if (TOKEN_SPELL (token) == SPELL_STRING) |
f80e83a9 | 2478 | { |
76faa4c0 | 2479 | U_CHAR *buff = (U_CHAR *) xmalloc (token->val.str.len); |
2480 | memcpy (buff, token->val.str.text, token->val.str.len); | |
2481 | result->val.str.text = buff; | |
f80e83a9 | 2482 | } |
2483 | return result; | |
2484 | } | |
6060326b | 2485 | |
f80e83a9 | 2486 | /* Determine whether two tokens can be pasted together, and if so, |
2487 | what the resulting token is. Returns CPP_EOF if the tokens cannot | |
2488 | be pasted, or the appropriate type for the merged token if they | |
2489 | can. */ | |
2490 | static enum cpp_ttype | |
2491 | can_paste (pfile, token1, token2, digraph) | |
2492 | cpp_reader * pfile; | |
2493 | const cpp_token *token1, *token2; | |
2494 | int* digraph; | |
6060326b | 2495 | { |
f80e83a9 | 2496 | enum cpp_ttype a = token1->type, b = token2->type; |
2497 | int cxx = CPP_OPTION (pfile, cplusplus); | |
6060326b | 2498 | |
31674461 | 2499 | /* Treat named operators as if they were ordinary NAMEs. */ |
2500 | if (token1->flags & NAMED_OP) | |
2501 | a = CPP_NAME; | |
2502 | if (token2->flags & NAMED_OP) | |
2503 | b = CPP_NAME; | |
2504 | ||
f80e83a9 | 2505 | if (a <= CPP_LAST_EQ && b == CPP_EQ) |
2506 | return a + (CPP_EQ_EQ - CPP_EQ); | |
6060326b | 2507 | |
f80e83a9 | 2508 | switch (a) |
6060326b | 2509 | { |
f80e83a9 | 2510 | case CPP_GREATER: |
2511 | if (b == a) return CPP_RSHIFT; | |
2512 | if (b == CPP_QUERY && cxx) return CPP_MAX; | |
2513 | if (b == CPP_GREATER_EQ) return CPP_RSHIFT_EQ; | |
2514 | break; | |
2515 | case CPP_LESS: | |
2516 | if (b == a) return CPP_LSHIFT; | |
2517 | if (b == CPP_QUERY && cxx) return CPP_MIN; | |
2518 | if (b == CPP_LESS_EQ) return CPP_LSHIFT_EQ; | |
27fdc0b6 | 2519 | if (CPP_OPTION (pfile, digraphs)) |
2520 | { | |
2521 | if (b == CPP_COLON) | |
2522 | {*digraph = 1; return CPP_OPEN_SQUARE;} /* <: digraph */ | |
2523 | if (b == CPP_MOD) | |
2524 | {*digraph = 1; return CPP_OPEN_BRACE;} /* <% digraph */ | |
2525 | } | |
f80e83a9 | 2526 | break; |
6060326b | 2527 | |
f80e83a9 | 2528 | case CPP_PLUS: if (b == a) return CPP_PLUS_PLUS; break; |
2529 | case CPP_AND: if (b == a) return CPP_AND_AND; break; | |
2530 | case CPP_OR: if (b == a) return CPP_OR_OR; break; | |
6060326b | 2531 | |
f80e83a9 | 2532 | case CPP_MINUS: |
2533 | if (b == a) return CPP_MINUS_MINUS; | |
2534 | if (b == CPP_GREATER) return CPP_DEREF; | |
2535 | break; | |
2536 | case CPP_COLON: | |
2537 | if (b == a && cxx) return CPP_SCOPE; | |
27fdc0b6 | 2538 | if (b == CPP_GREATER && CPP_OPTION (pfile, digraphs)) |
f80e83a9 | 2539 | {*digraph = 1; return CPP_CLOSE_SQUARE;} /* :> digraph */ |
2540 | break; | |
2541 | ||
2542 | case CPP_MOD: | |
27fdc0b6 | 2543 | if (CPP_OPTION (pfile, digraphs)) |
2544 | { | |
2545 | if (b == CPP_GREATER) | |
2546 | {*digraph = 1; return CPP_CLOSE_BRACE;} /* %> digraph */ | |
2547 | if (b == CPP_COLON) | |
2548 | {*digraph = 1; return CPP_HASH;} /* %: digraph */ | |
2549 | } | |
f80e83a9 | 2550 | break; |
2551 | case CPP_DEREF: | |
2552 | if (b == CPP_MULT && cxx) return CPP_DEREF_STAR; | |
2553 | break; | |
2554 | case CPP_DOT: | |
2555 | if (b == CPP_MULT && cxx) return CPP_DOT_STAR; | |
2556 | if (b == CPP_NUMBER) return CPP_NUMBER; | |
2557 | break; | |
2558 | ||
2559 | case CPP_HASH: | |
2560 | if (b == a && (token1->flags & DIGRAPH) == (token2->flags & DIGRAPH)) | |
2561 | /* %:%: digraph */ | |
2562 | {*digraph = (token1->flags & DIGRAPH); return CPP_PASTE;} | |
2563 | break; | |
2564 | ||
2565 | case CPP_NAME: | |
2566 | if (b == CPP_NAME) return CPP_NAME; | |
2567 | if (b == CPP_NUMBER | |
76faa4c0 | 2568 | && is_numstart(token2->val.str.text[0])) return CPP_NAME; |
f80e83a9 | 2569 | if (b == CPP_CHAR |
76faa4c0 | 2570 | && token1->val.node == pfile->spec_nodes->n_L) return CPP_WCHAR; |
f80e83a9 | 2571 | if (b == CPP_STRING |
76faa4c0 | 2572 | && token1->val.node == pfile->spec_nodes->n_L) return CPP_WSTRING; |
f80e83a9 | 2573 | break; |
2574 | ||
2575 | case CPP_NUMBER: | |
2576 | if (b == CPP_NUMBER) return CPP_NUMBER; | |
2577 | if (b == CPP_NAME) return CPP_NUMBER; | |
2578 | if (b == CPP_DOT) return CPP_NUMBER; | |
2579 | /* Numbers cannot have length zero, so this is safe. */ | |
2580 | if ((b == CPP_PLUS || b == CPP_MINUS) | |
76faa4c0 | 2581 | && VALID_SIGN ('+', token1->val.str.text[token1->val.str.len - 1])) |
f80e83a9 | 2582 | return CPP_NUMBER; |
2583 | break; | |
2584 | ||
71aa9da4 | 2585 | case CPP_OTHER: |
2586 | if (CPP_OPTION (pfile, objc) && token1->val.aux == '@') | |
2587 | { | |
2588 | if (b == CPP_NAME) return CPP_NAME; | |
2589 | if (b == CPP_STRING) return CPP_OSTRING; | |
2590 | } | |
2591 | ||
f80e83a9 | 2592 | default: |
2593 | break; | |
6060326b | 2594 | } |
2595 | ||
f80e83a9 | 2596 | return CPP_EOF; |
2597 | } | |
2598 | ||
2599 | /* Check if TOKEN is to be ##-pasted with the token after it. */ | |
2600 | static const cpp_token * | |
2601 | maybe_paste_with_next (pfile, token) | |
2602 | cpp_reader *pfile; | |
2603 | const cpp_token *token; | |
2604 | { | |
2605 | cpp_token *pasted; | |
2606 | const cpp_token *second; | |
2607 | cpp_context *context = CURRENT_CONTEXT (pfile); | |
2608 | ||
2609 | /* Is this token on the LHS of ## ? */ | |
f80e83a9 | 2610 | |
deb356cf | 2611 | while ((token->flags & PASTE_LEFT) |
2612 | || ((context->flags & CONTEXT_PASTEL) | |
2613 | && context->posn == context->count)) | |
6060326b | 2614 | { |
deb356cf | 2615 | /* Suppress macro expansion for next token, but don't conflict |
2616 | with the other method of suppression. If it is an argument, | |
2617 | macro expansion within the argument will still occur. */ | |
2618 | pfile->paste_level = pfile->cur_context; | |
2619 | second = _cpp_get_token (pfile); | |
2620 | pfile->paste_level = 0; | |
2621 | ||
2622 | /* Ignore placemarker argument tokens (cannot be from an empty | |
2623 | macro since macros are not expanded). */ | |
2624 | if (token->type == CPP_PLACEMARKER) | |
2625 | pasted = duplicate_token (pfile, second); | |
2626 | else if (second->type == CPP_PLACEMARKER) | |
f80e83a9 | 2627 | { |
e0a859f1 | 2628 | /* GCC has special extended semantics for , ## b where b is |
2629 | a varargs parameter: the comma disappears if b was given | |
2630 | no actual arguments (not merely if b is an empty | |
2631 | argument). */ | |
2632 | if (token->type == CPP_COMMA && second->flags & VOID_REST) | |
35677230 | 2633 | pasted = duplicate_token (pfile, second); |
deb356cf | 2634 | else |
2635 | pasted = duplicate_token (pfile, token); | |
f80e83a9 | 2636 | } |
2637 | else | |
f80e83a9 | 2638 | { |
deb356cf | 2639 | int digraph = 0; |
2640 | enum cpp_ttype type = can_paste (pfile, token, second, &digraph); | |
6060326b | 2641 | |
deb356cf | 2642 | if (type == CPP_EOF) |
2643 | { | |
2644 | if (CPP_OPTION (pfile, warn_paste)) | |
e0a859f1 | 2645 | { |
2646 | /* Do not complain about , ## <whatever> if | |
2647 | <whatever> came from a variable argument, because | |
2648 | the author probably intended the ## to trigger | |
2649 | the special extended semantics (see above). */ | |
2650 | if (token->type == CPP_COMMA | |
2651 | && IS_ARG_CONTEXT (CURRENT_CONTEXT (pfile)) | |
58fe658a | 2652 | && ON_REST_ARG (CURRENT_CONTEXT (pfile) - 1)) |
e0a859f1 | 2653 | /* no warning */; |
2654 | else | |
2655 | cpp_warning (pfile, | |
deb356cf | 2656 | "pasting would not give a valid preprocessing token"); |
e0a859f1 | 2657 | } |
deb356cf | 2658 | _cpp_push_token (pfile, second); |
58fe658a | 2659 | /* A short term hack to safely clear the PASTE_LEFT flag. */ |
2660 | pasted = duplicate_token (pfile, token); | |
2661 | pasted->flags &= ~PASTE_LEFT; | |
2662 | return pasted; | |
deb356cf | 2663 | } |
6060326b | 2664 | |
deb356cf | 2665 | if (type == CPP_NAME || type == CPP_NUMBER) |
2666 | { | |
2667 | /* Join spellings. */ | |
2668 | U_CHAR *buf, *end; | |
2669 | ||
2670 | pasted = get_temp_token (pfile); | |
2671 | buf = (U_CHAR *) alloca (TOKEN_LEN (token) + TOKEN_LEN (second)); | |
2672 | end = spell_token (pfile, token, buf); | |
2673 | end = spell_token (pfile, second, end); | |
2674 | *end = '\0'; | |
f80e83a9 | 2675 | |
deb356cf | 2676 | if (type == CPP_NAME) |
2677 | pasted->val.node = cpp_lookup (pfile, buf, end - buf); | |
2678 | else | |
2679 | { | |
2680 | pasted->val.str.text = uxstrdup (buf); | |
2681 | pasted->val.str.len = end - buf; | |
2682 | } | |
2683 | } | |
71aa9da4 | 2684 | else if (type == CPP_WCHAR || type == CPP_WSTRING |
2685 | || type == CPP_OSTRING) | |
deb356cf | 2686 | pasted = duplicate_token (pfile, second); |
76faa4c0 | 2687 | else |
2688 | { | |
deb356cf | 2689 | pasted = get_temp_token (pfile); |
2690 | pasted->val.integer = 0; | |
76faa4c0 | 2691 | } |
deb356cf | 2692 | |
2693 | pasted->type = type; | |
2694 | pasted->flags = digraph ? DIGRAPH : 0; | |
31674461 | 2695 | |
2696 | if (type == CPP_NAME && pasted->val.node->type == T_OPERATOR) | |
2697 | { | |
2698 | pasted->type = pasted->val.node->value.code; | |
2699 | pasted->flags |= NAMED_OP; | |
2700 | } | |
f80e83a9 | 2701 | } |
2702 | ||
deb356cf | 2703 | /* The pasted token gets the whitespace flags and position of the |
2704 | first token, the PASTE_LEFT flag of the second token, plus the | |
2705 | PASTED flag to indicate it is the result of a paste. However, we | |
2706 | want to preserve the DIGRAPH flag. */ | |
2707 | pasted->flags &= ~(PREV_WHITE | BOL | PASTE_LEFT); | |
2708 | pasted->flags |= ((token->flags & (PREV_WHITE | BOL)) | |
2709 | | (second->flags & PASTE_LEFT) | PASTED); | |
2710 | pasted->col = token->col; | |
2711 | pasted->line = token->line; | |
2712 | ||
2713 | /* See if there is another token to be pasted onto the one we just | |
2714 | constructed. */ | |
2715 | token = pasted; | |
2716 | context = CURRENT_CONTEXT (pfile); | |
2717 | /* and loop */ | |
f80e83a9 | 2718 | } |
deb356cf | 2719 | return token; |
f80e83a9 | 2720 | } |
2721 | ||
2722 | /* Convert a token sequence to a single string token according to the | |
2723 | rules of the ISO C #-operator. */ | |
2724 | #define INIT_SIZE 200 | |
2725 | static cpp_token * | |
2726 | stringify_arg (pfile, token) | |
6060326b | 2727 | cpp_reader *pfile; |
f80e83a9 | 2728 | const cpp_token *token; |
6060326b | 2729 | { |
f80e83a9 | 2730 | cpp_token *result; |
2731 | unsigned char *main_buf; | |
2732 | unsigned int prev_value, backslash_count = 0; | |
2733 | unsigned int buf_used = 0, whitespace = 0, buf_cap = INIT_SIZE; | |
6060326b | 2734 | |
deb356cf | 2735 | push_arg_context (pfile, token); |
f80e83a9 | 2736 | prev_value = prevent_macro_expansion (pfile); |
2737 | main_buf = (unsigned char *) xmalloc (buf_cap); | |
6060326b | 2738 | |
f80e83a9 | 2739 | result = get_temp_token (pfile); |
2740 | ASSIGN_FLAGS_AND_POS (result, token); | |
2741 | ||
deb356cf | 2742 | for (; (token = _cpp_get_token (pfile))->type != CPP_EOF; ) |
6060326b | 2743 | { |
f80e83a9 | 2744 | int escape; |
2745 | unsigned char *buf; | |
2746 | unsigned int len = TOKEN_LEN (token); | |
6060326b | 2747 | |
7e842f95 | 2748 | if (token->type == CPP_PLACEMARKER) |
2749 | continue; | |
2750 | ||
f80e83a9 | 2751 | escape = (token->type == CPP_STRING || token->type == CPP_WSTRING |
2752 | || token->type == CPP_CHAR || token->type == CPP_WCHAR); | |
2753 | if (escape) | |
2754 | len *= 4 + 1; | |
2755 | ||
2756 | if (buf_used + len > buf_cap) | |
6060326b | 2757 | { |
f80e83a9 | 2758 | buf_cap = buf_used + len + INIT_SIZE; |
2759 | main_buf = xrealloc (main_buf, buf_cap); | |
2760 | } | |
6060326b | 2761 | |
f80e83a9 | 2762 | if (whitespace && (token->flags & PREV_WHITE)) |
2763 | main_buf[buf_used++] = ' '; | |
6060326b | 2764 | |
f80e83a9 | 2765 | if (escape) |
2766 | buf = (unsigned char *) xmalloc (len); | |
2767 | else | |
2768 | buf = main_buf + buf_used; | |
2769 | ||
2770 | len = spell_token (pfile, token, buf) - buf; | |
2771 | if (escape) | |
2772 | { | |
2773 | buf_used = quote_string (&main_buf[buf_used], buf, len) - main_buf; | |
2774 | free (buf); | |
2775 | } | |
2776 | else | |
2777 | buf_used += len; | |
6060326b | 2778 | |
f80e83a9 | 2779 | whitespace = 1; |
2780 | if (token->type == CPP_BACKSLASH) | |
2781 | backslash_count++; | |
2782 | else | |
2783 | backslash_count = 0; | |
2784 | } | |
6060326b | 2785 | |
f80e83a9 | 2786 | /* Ignore the final \ of invalid string literals. */ |
2787 | if (backslash_count & 1) | |
2788 | { | |
2789 | cpp_warning (pfile, "invalid string literal, ignoring final '\\'"); | |
2790 | buf_used--; | |
2791 | } | |
6060326b | 2792 | |
f80e83a9 | 2793 | result->type = CPP_STRING; |
76faa4c0 | 2794 | result->val.str.text = main_buf; |
2795 | result->val.str.len = buf_used; | |
f80e83a9 | 2796 | restore_macro_expansion (pfile, prev_value); |
2797 | return result; | |
2798 | } | |
6060326b | 2799 | |
f80e83a9 | 2800 | /* Allocate more room on the context stack of PFILE. */ |
2801 | static void | |
2802 | expand_context_stack (pfile) | |
2803 | cpp_reader *pfile; | |
2804 | { | |
2805 | pfile->context_cap += pfile->context_cap + 20; | |
2806 | pfile->contexts = (cpp_context *) | |
2807 | xrealloc (pfile->contexts, pfile->context_cap * sizeof (cpp_context)); | |
2808 | } | |
6060326b | 2809 | |
f80e83a9 | 2810 | /* Push the context of macro NODE onto the context stack. TOKEN is |
2811 | the CPP_NAME token invoking the macro. */ | |
deb356cf | 2812 | static int |
2813 | push_macro_context (pfile, token) | |
f80e83a9 | 2814 | cpp_reader *pfile; |
f80e83a9 | 2815 | const cpp_token *token; |
2816 | { | |
2817 | unsigned char orig_flags; | |
2818 | macro_args *args; | |
2819 | cpp_context *context; | |
deb356cf | 2820 | cpp_hashnode *node = token->val.node; |
6060326b | 2821 | |
f80e83a9 | 2822 | /* Token's flags may change when parsing args containing a nested |
2823 | invocation of this macro. */ | |
2824 | orig_flags = token->flags & (PREV_WHITE | BOL); | |
2825 | args = 0; | |
2826 | if (node->value.expansion->paramc >= 0) | |
6060326b | 2827 | { |
f80e83a9 | 2828 | unsigned int error, prev_nme; |
2829 | ||
2830 | /* Allocate room for the argument contexts, and parse them. */ | |
2831 | args = (macro_args *) xmalloc (sizeof (macro_args)); | |
2832 | args->ends = (unsigned int *) | |
2833 | xmalloc (node->value.expansion->paramc * sizeof (unsigned int)); | |
2834 | args->tokens = 0; | |
2835 | args->capacity = 0; | |
2836 | args->used = 0; | |
f80e83a9 | 2837 | |
2838 | prev_nme = prevent_macro_expansion (pfile); | |
2839 | pfile->args = args; | |
2840 | error = parse_args (pfile, node, args); | |
2841 | pfile->args = 0; | |
2842 | restore_macro_expansion (pfile, prev_nme); | |
2843 | if (error) | |
2844 | { | |
2845 | free_macro_args (args); | |
deb356cf | 2846 | return 1; |
f80e83a9 | 2847 | } |
e2ab8477 | 2848 | /* Set the level after the call to parse_args. */ |
2849 | args->level = pfile->cur_context; | |
6060326b | 2850 | } |
6060326b | 2851 | |
f80e83a9 | 2852 | /* Now push its context. */ |
2853 | pfile->cur_context++; | |
2854 | if (pfile->cur_context == pfile->context_cap) | |
2855 | expand_context_stack (pfile); | |
2856 | ||
2857 | context = CURRENT_CONTEXT (pfile); | |
2858 | context->u.list = node->value.expansion; | |
2859 | context->args = args; | |
2860 | context->posn = 0; | |
2861 | context->count = context->u.list->tokens_used; | |
2862 | context->level = pfile->cur_context; | |
2863 | context->flags = 0; | |
2864 | context->pushed_token = 0; | |
2865 | ||
2866 | /* Set the flags of the first token. We know there must | |
2867 | be one, empty macros are a single placemarker token. */ | |
2868 | MODIFY_FLAGS_AND_POS (&context->u.list->tokens[0], token, orig_flags); | |
2869 | ||
deb356cf | 2870 | return 0; |
6060326b | 2871 | } |
2872 | ||
f80e83a9 | 2873 | /* Push an argument to the current macro onto the context stack. |
2874 | TOKEN is the MACRO_ARG token representing the argument expansion. */ | |
deb356cf | 2875 | static void |
f80e83a9 | 2876 | push_arg_context (pfile, token) |
2877 | cpp_reader *pfile; | |
2878 | const cpp_token *token; | |
6060326b | 2879 | { |
f80e83a9 | 2880 | cpp_context *context; |
2881 | macro_args *args; | |
2882 | ||
2883 | pfile->cur_context++; | |
2884 | if (pfile->cur_context == pfile->context_cap) | |
2885 | expand_context_stack (pfile); | |
2886 | ||
2887 | context = CURRENT_CONTEXT (pfile); | |
2888 | args = context[-1].args; | |
2889 | ||
2890 | context->count = token->val.aux ? args->ends[token->val.aux - 1]: 0; | |
2891 | context->u.arg = args->tokens + context->count; | |
2892 | context->count = args->ends[token->val.aux] - context->count; | |
2893 | context->args = 0; | |
2894 | context->posn = 0; | |
2895 | context->level = args->level; | |
2896 | context->flags = CONTEXT_ARG | CONTEXT_RAW; | |
2897 | context->pushed_token = 0; | |
2898 | ||
2899 | /* Set the flags of the first token. There is one. */ | |
2900 | { | |
2901 | const cpp_token *first = context->u.arg[0]; | |
2902 | if (!first) | |
2903 | first = context->u.arg[1]; | |
6060326b | 2904 | |
f80e83a9 | 2905 | MODIFY_FLAGS_AND_POS ((cpp_token *) first, token, |
2906 | token->flags & (PREV_WHITE | BOL)); | |
2907 | } | |
6060326b | 2908 | |
f80e83a9 | 2909 | if (token->flags & PASTE_LEFT) |
2910 | context->flags |= CONTEXT_PASTEL; | |
2911 | if (pfile->paste_level) | |
2912 | context->flags |= CONTEXT_PASTER; | |
6060326b | 2913 | } |
2914 | ||
f80e83a9 | 2915 | /* "Unget" a token. It is effectively inserted in the token queue and |
2916 | will be returned by the next call to get_raw_token. */ | |
6060326b | 2917 | void |
f80e83a9 | 2918 | _cpp_push_token (pfile, token) |
6060326b | 2919 | cpp_reader *pfile; |
f80e83a9 | 2920 | const cpp_token *token; |
6060326b | 2921 | { |
f80e83a9 | 2922 | cpp_context *context = CURRENT_CONTEXT (pfile); |
7e842f95 | 2923 | |
2924 | if (context->posn > 0) | |
2925 | { | |
2926 | const cpp_token *prev; | |
2927 | if (IS_ARG_CONTEXT (context)) | |
2928 | prev = context->u.arg[context->posn - 1]; | |
2929 | else | |
2930 | prev = &context->u.list->tokens[context->posn - 1]; | |
2931 | ||
2932 | if (prev == token) | |
2933 | { | |
2934 | context->posn--; | |
2935 | return; | |
2936 | } | |
2937 | } | |
2938 | ||
f80e83a9 | 2939 | if (context->pushed_token) |
2940 | cpp_ice (pfile, "two tokens pushed in a row"); | |
2941 | if (token->type != CPP_EOF) | |
2942 | context->pushed_token = token; | |
2943 | /* Don't push back a directive's CPP_EOF, step back instead. */ | |
2944 | else if (pfile->cur_context == 0) | |
2945 | pfile->contexts[0].posn--; | |
2946 | } | |
6060326b | 2947 | |
f80e83a9 | 2948 | /* Handle a preprocessing directive. TOKEN is the CPP_HASH token |
2949 | introducing the directive. */ | |
2950 | static void | |
2951 | process_directive (pfile, token) | |
2952 | cpp_reader *pfile; | |
2953 | const cpp_token *token; | |
2954 | { | |
2955 | const struct directive *d = pfile->token_list.directive; | |
2956 | int prev_nme = 0; | |
2957 | ||
2958 | /* Skip over the directive name. */ | |
2959 | if (token[1].type == CPP_NAME) | |
2960 | _cpp_get_raw_token (pfile); | |
2961 | else if (token[1].type != CPP_NUMBER) | |
7e842f95 | 2962 | cpp_ice (pfile, "directive begins with %s?!", TOKEN_NAME (token)); |
f80e83a9 | 2963 | |
f80e83a9 | 2964 | if (! (d->flags & EXPAND)) |
2965 | prev_nme = prevent_macro_expansion (pfile); | |
2966 | (void) (*d->handler) (pfile); | |
2967 | if (! (d->flags & EXPAND)) | |
2968 | restore_macro_expansion (pfile, prev_nme); | |
2969 | _cpp_skip_rest_of_line (pfile); | |
2970 | } | |
6060326b | 2971 | |
f80e83a9 | 2972 | /* The external interface to return the next token. All macro |
2973 | expansion and directive processing is handled internally, the | |
2974 | caller only ever sees the output after preprocessing. */ | |
2975 | const cpp_token * | |
2976 | cpp_get_token (pfile) | |
2977 | cpp_reader *pfile; | |
2978 | { | |
2979 | const cpp_token *token; | |
deb356cf | 2980 | /* Loop till we hit a non-directive, non-placemarker token. */ |
f80e83a9 | 2981 | for (;;) |
2982 | { | |
deb356cf | 2983 | token = _cpp_get_token (pfile); |
2984 | ||
2985 | if (token->type == CPP_PLACEMARKER) | |
2986 | continue; | |
2987 | ||
2988 | if (token->type == CPP_HASH && token->flags & BOL | |
f80e83a9 | 2989 | && pfile->token_list.directive) |
6060326b | 2990 | { |
f80e83a9 | 2991 | process_directive (pfile, token); |
2992 | continue; | |
6060326b | 2993 | } |
2994 | ||
deb356cf | 2995 | return token; |
2996 | } | |
2997 | } | |
2998 | ||
2999 | /* The internal interface to return the next token. There are two | |
3000 | differences between the internal and external interfaces: the | |
3001 | internal interface may return a PLACEMARKER token, and it does not | |
3002 | process directives. */ | |
3003 | const cpp_token * | |
3004 | _cpp_get_token (pfile) | |
3005 | cpp_reader *pfile; | |
3006 | { | |
c2cce424 | 3007 | const cpp_token *token, *old_token; |
deb356cf | 3008 | cpp_hashnode *node; |
3009 | ||
3010 | /* Loop until we hit a non-macro token. */ | |
3011 | for (;;) | |
3012 | { | |
3013 | token = get_raw_token (pfile); | |
3014 | ||
f80e83a9 | 3015 | /* Short circuit EOF. */ |
3016 | if (token->type == CPP_EOF) | |
3017 | return token; | |
deb356cf | 3018 | |
3019 | /* If we are skipping... */ | |
3020 | if (pfile->skipping) | |
6060326b | 3021 | { |
deb356cf | 3022 | /* we still have to process directives, */ |
3023 | if (pfile->token_list.directive) | |
3024 | return token; | |
3025 | ||
3026 | /* but everything else is ignored. */ | |
f80e83a9 | 3027 | _cpp_skip_rest_of_line (pfile); |
3028 | continue; | |
3029 | } | |
6060326b | 3030 | |
deb356cf | 3031 | /* If there's a potential control macro and we get here, then that |
3032 | #ifndef didn't cover the entire file and its argument shouldn't | |
3033 | be taken as a control macro. */ | |
3034 | pfile->potential_control_macro = 0; | |
6060326b | 3035 | |
db54da8f | 3036 | /* If we are rescanning preprocessed input, no macro expansion or |
3037 | token pasting may occur. */ | |
3038 | if (CPP_OPTION (pfile, preprocessed)) | |
3039 | return token; | |
3040 | ||
c2cce424 | 3041 | old_token = token; |
3042 | ||
deb356cf | 3043 | /* See if there's a token to paste with this one. */ |
3044 | if (!pfile->paste_level) | |
3045 | token = maybe_paste_with_next (pfile, token); | |
6060326b | 3046 | |
deb356cf | 3047 | /* If it isn't a macro, return it now. */ |
31674461 | 3048 | if (token->type != CPP_NAME || token->val.node->type == T_VOID) |
deb356cf | 3049 | return token; |
6060326b | 3050 | |
31674461 | 3051 | /* Is macro expansion disabled in general, or are we in the |
c2cce424 | 3052 | middle of a token paste, or was this token just pasted? |
3053 | (Note we don't check token->flags & PASTED, because that | |
3054 | counts tokens that were pasted at some point in the past, | |
3055 | we're only interested in tokens that were pasted by this call | |
3056 | to maybe_paste_with_next.) */ | |
3057 | if (pfile->no_expand_level == pfile->cur_context | |
3058 | || pfile->paste_level | |
3059 | || (token != old_token | |
3060 | && pfile->no_expand_level + 1 == pfile->cur_context)) | |
deb356cf | 3061 | return token; |
c2cce424 | 3062 | |
deb356cf | 3063 | node = token->val.node; |
3064 | if (node->type != T_MACRO) | |
3065 | return special_symbol (pfile, node, token); | |
6060326b | 3066 | |
f80e83a9 | 3067 | if (is_macro_disabled (pfile, node->value.expansion, token)) |
3068 | return token; | |
6060326b | 3069 | |
deb356cf | 3070 | if (push_macro_context (pfile, token)) |
3071 | return token; | |
3072 | /* else loop */ | |
f80e83a9 | 3073 | } |
f80e83a9 | 3074 | } |
6060326b | 3075 | |
f80e83a9 | 3076 | /* Returns the next raw token, i.e. without performing macro |
3077 | expansion. Argument contexts are automatically entered. */ | |
3078 | static const cpp_token * | |
3079 | get_raw_token (pfile) | |
3080 | cpp_reader *pfile; | |
3081 | { | |
3082 | const cpp_token *result; | |
deb356cf | 3083 | cpp_context *context; |
6060326b | 3084 | |
deb356cf | 3085 | for (;;) |
f80e83a9 | 3086 | { |
deb356cf | 3087 | context = CURRENT_CONTEXT (pfile); |
3088 | if (context->pushed_token) | |
f80e83a9 | 3089 | { |
deb356cf | 3090 | result = context->pushed_token; |
3091 | context->pushed_token = 0; | |
35677230 | 3092 | return result; /* Cannot be a CPP_MACRO_ARG */ |
deb356cf | 3093 | } |
3094 | else if (context->posn == context->count) | |
3095 | { | |
3096 | if (pop_context (pfile)) | |
3097 | return &eof_token; | |
3098 | continue; | |
3099 | } | |
35677230 | 3100 | else if (IS_ARG_CONTEXT (context)) |
deb356cf | 3101 | { |
35677230 | 3102 | result = context->u.arg[context->posn++]; |
3103 | if (result == 0) | |
6060326b | 3104 | { |
35677230 | 3105 | context->flags ^= CONTEXT_RAW; |
f80e83a9 | 3106 | result = context->u.arg[context->posn++]; |
6060326b | 3107 | } |
35677230 | 3108 | return result; /* Cannot be a CPP_MACRO_ARG */ |
f80e83a9 | 3109 | } |
6060326b | 3110 | |
35677230 | 3111 | result = &context->u.list->tokens[context->posn++]; |
3112 | ||
deb356cf | 3113 | if (result->type != CPP_MACRO_ARG) |
3114 | return result; | |
3115 | ||
3116 | if (result->flags & STRINGIFY_ARG) | |
3117 | return stringify_arg (pfile, result); | |
3118 | ||
3119 | push_arg_context (pfile, result); | |
3120 | } | |
f80e83a9 | 3121 | } |
6060326b | 3122 | |
f80e83a9 | 3123 | /* Internal interface to get the token without macro expanding. */ |
3124 | const cpp_token * | |
3125 | _cpp_get_raw_token (pfile) | |
3126 | cpp_reader *pfile; | |
3127 | { | |
3128 | int prev_nme = prevent_macro_expansion (pfile); | |
deb356cf | 3129 | const cpp_token *result = _cpp_get_token (pfile); |
f80e83a9 | 3130 | restore_macro_expansion (pfile, prev_nme); |
3131 | return result; | |
3132 | } | |
6060326b | 3133 | |
f80e83a9 | 3134 | /* A thin wrapper to lex_line. CLEAR is non-zero if the current token |
3135 | list should be overwritten, or zero if we need to append | |
3136 | (typically, if we are within the arguments to a macro, or looking | |
3137 | for the '(' to start a function-like macro invocation). */ | |
3138 | static int | |
3139 | lex_next (pfile, clear) | |
3140 | cpp_reader *pfile; | |
3141 | int clear; | |
3142 | { | |
3143 | cpp_toklist *list = &pfile->token_list; | |
3144 | const cpp_token *old_list = list->tokens; | |
3145 | unsigned int old_used = list->tokens_used; | |
6060326b | 3146 | |
f80e83a9 | 3147 | if (clear) |
6060326b | 3148 | { |
f80e83a9 | 3149 | /* Release all temporary tokens. */ |
3150 | _cpp_clear_toklist (list); | |
3151 | pfile->contexts[0].posn = 0; | |
3152 | if (pfile->temp_used) | |
3153 | release_temp_tokens (pfile); | |
6060326b | 3154 | } |
f80e83a9 | 3155 | lex_line (pfile, list); |
3156 | pfile->contexts[0].count = list->tokens_used; | |
6060326b | 3157 | |
f80e83a9 | 3158 | if (!clear && pfile->args) |
6060326b | 3159 | { |
f80e83a9 | 3160 | /* Fix up argument token pointers. */ |
3161 | if (old_list != list->tokens) | |
3162 | { | |
3163 | unsigned int i; | |
3164 | ||
3165 | for (i = 0; i < pfile->args->used; i++) | |
3166 | { | |
3167 | const cpp_token *token = pfile->args->tokens[i]; | |
3168 | if (token >= old_list && token < old_list + old_used) | |
3169 | pfile->args->tokens[i] = (const cpp_token *) | |
3170 | ((char *) token + ((char *) list->tokens - (char *) old_list)); | |
3171 | } | |
3172 | } | |
3173 | ||
3174 | /* 6.10.3 paragraph 11: If there are sequences of preprocessing | |
3175 | tokens within the list of arguments that would otherwise act as | |
3176 | preprocessing directives, the behavior is undefined. | |
3177 | ||
3178 | This implementation will report a hard error and treat the | |
3179 | 'sequence of preprocessing tokens' as part of the macro argument, | |
3180 | not a directive. | |
3181 | ||
3182 | Note if pfile->args == 0, we're OK since we're only inside a | |
3183 | macro argument after a '('. */ | |
3184 | if (list->directive) | |
3185 | { | |
3186 | cpp_error_with_line (pfile, list->tokens[old_used].line, | |
3187 | list->tokens[old_used].col, | |
3188 | "#%s may not be used inside a macro argument", | |
3189 | list->directive->name); | |
78719282 | 3190 | return 1; |
f80e83a9 | 3191 | } |
6060326b | 3192 | } |
3193 | ||
f80e83a9 | 3194 | return 0; |
6060326b | 3195 | } |
3196 | ||
deb356cf | 3197 | /* Pops a context off the context stack. If we're at the bottom, lexes |
3198 | the next logical line. Returns EOF if we're at the end of the | |
e0a859f1 | 3199 | argument list to the # operator, or we should not "overflow" |
f80e83a9 | 3200 | into the rest of the file (e.g. 6.10.3.1.1). */ |
3201 | static int | |
deb356cf | 3202 | pop_context (pfile) |
f80e83a9 | 3203 | cpp_reader *pfile; |
3204 | { | |
3205 | cpp_context *context; | |
89b05ef6 | 3206 | |
f80e83a9 | 3207 | if (pfile->cur_context == 0) |
deb356cf | 3208 | { |
3209 | /* If we are currently processing a directive, do not advance. 6.10 | |
3210 | paragraph 2: A new-line character ends the directive even if it | |
3211 | occurs within what would otherwise be an invocation of a | |
3212 | function-like macro. */ | |
3213 | if (pfile->token_list.directive) | |
3214 | return 1; | |
3215 | ||
3216 | return lex_next (pfile, pfile->no_expand_level == UINT_MAX); | |
3217 | } | |
f80e83a9 | 3218 | |
3219 | /* Argument contexts, when parsing args or handling # operator | |
3220 | return CPP_EOF at the end. */ | |
3221 | context = CURRENT_CONTEXT (pfile); | |
3222 | if (IS_ARG_CONTEXT (context) && pfile->cur_context == pfile->no_expand_level) | |
3223 | return 1; | |
3224 | ||
3225 | /* Free resources when leaving macro contexts. */ | |
3226 | if (context->args) | |
3227 | free_macro_args (context->args); | |
3228 | ||
3229 | if (pfile->cur_context == pfile->no_expand_level) | |
3230 | pfile->no_expand_level--; | |
3231 | pfile->cur_context--; | |
3232 | ||
3233 | return 0; | |
3234 | } | |
3235 | ||
f80e83a9 | 3236 | /* Turn off macro expansion at the current context level. */ |
3237 | static unsigned int | |
3238 | prevent_macro_expansion (pfile) | |
3239 | cpp_reader *pfile; | |
3240 | { | |
3241 | unsigned int prev_value = pfile->no_expand_level; | |
3242 | pfile->no_expand_level = pfile->cur_context; | |
3243 | return prev_value; | |
3244 | } | |
3245 | ||
3246 | /* Restore macro expansion to its previous state. */ | |
3247 | static void | |
3248 | restore_macro_expansion (pfile, prev_value) | |
3249 | cpp_reader *pfile; | |
3250 | unsigned int prev_value; | |
3251 | { | |
3252 | pfile->no_expand_level = prev_value; | |
3253 | } | |
3254 | ||
3255 | /* Used by cpperror.c to obtain the correct line and column to report | |
3256 | in a diagnostic. */ | |
3257 | unsigned int | |
3258 | _cpp_get_line (pfile, pcol) | |
3259 | cpp_reader *pfile; | |
3260 | unsigned int *pcol; | |
3261 | { | |
3262 | unsigned int index; | |
3263 | const cpp_token *cur_token; | |
3264 | ||
338fa5f7 | 3265 | if (pfile->state.in_lex_line) |
f80e83a9 | 3266 | index = pfile->token_list.tokens_used; |
3267 | else | |
c2cce424 | 3268 | { |
338fa5f7 | 3269 | index = pfile->contexts[0].posn; |
3270 | ||
3271 | if (index == 0) | |
3272 | { | |
3273 | if (pcol) | |
3274 | *pcol = 0; | |
3275 | return 0; | |
3276 | } | |
3277 | index--; | |
c2cce424 | 3278 | } |
3279 | ||
338fa5f7 | 3280 | cur_token = &pfile->token_list.tokens[index]; |
f80e83a9 | 3281 | if (pcol) |
3282 | *pcol = cur_token->col; | |
3283 | return cur_token->line; | |
3284 | } | |
3285 | ||
3286 | #define DSC(str) (const U_CHAR *)str, sizeof str - 1 | |
3287 | static const char * const monthnames[] = | |
3288 | { | |
3289 | "Jan", "Feb", "Mar", "Apr", "May", "Jun", | |
3290 | "Jul", "Aug", "Sep", "Oct", "Nov", "Dec", | |
3291 | }; | |
3292 | ||
3293 | /* Handle builtin macros like __FILE__. */ | |
3294 | static const cpp_token * | |
3295 | special_symbol (pfile, node, token) | |
3296 | cpp_reader *pfile; | |
3297 | cpp_hashnode *node; | |
d2efb5ed | 3298 | const cpp_token *token; |
89b05ef6 | 3299 | { |
f80e83a9 | 3300 | cpp_token *result; |
3301 | cpp_buffer *ip; | |
89b05ef6 | 3302 | |
f80e83a9 | 3303 | switch (node->type) |
89b05ef6 | 3304 | { |
f80e83a9 | 3305 | case T_FILE: |
3306 | case T_BASE_FILE: | |
89b05ef6 | 3307 | { |
f80e83a9 | 3308 | const char *file; |
89b05ef6 | 3309 | |
f80e83a9 | 3310 | ip = CPP_BUFFER (pfile); |
3311 | if (ip == 0) | |
3312 | file = ""; | |
89b05ef6 | 3313 | else |
f80e83a9 | 3314 | { |
3315 | if (node->type == T_BASE_FILE) | |
3316 | while (CPP_PREV_BUFFER (ip) != NULL) | |
3317 | ip = CPP_PREV_BUFFER (ip); | |
3318 | ||
3319 | file = ip->nominal_fname; | |
3320 | } | |
3321 | result = make_string_token (get_temp_token (pfile), (U_CHAR *) file, | |
3322 | strlen (file)); | |
3323 | } | |
3324 | break; | |
89b05ef6 | 3325 | |
f80e83a9 | 3326 | case T_INCLUDE_LEVEL: |
241e762e | 3327 | /* pfile->include_depth counts the primary source as level 1, |
3328 | but historically __INCLUDE_DEPTH__ has called the primary | |
3329 | source level 0. */ | |
3330 | result = alloc_number_token (pfile, pfile->include_depth - 1); | |
89b05ef6 | 3331 | break; |
3332 | ||
f80e83a9 | 3333 | case T_SPECLINE: |
3334 | /* If __LINE__ is embedded in a macro, it must expand to the | |
3335 | line of the macro's invocation, not its definition. | |
3336 | Otherwise things like assert() will not work properly. */ | |
3337 | result = alloc_number_token (pfile, _cpp_get_line (pfile, NULL)); | |
de3d9756 | 3338 | break; |
3339 | ||
f80e83a9 | 3340 | case T_STDC: |
89b05ef6 | 3341 | { |
f80e83a9 | 3342 | int stdc = 1; |
89b05ef6 | 3343 | |
f80e83a9 | 3344 | #ifdef STDC_0_IN_SYSTEM_HEADERS |
3345 | if (CPP_IN_SYSTEM_HEADER (pfile) | |
76faa4c0 | 3346 | && pfile->spec_nodes->n__STRICT_ANSI__->type == T_VOID) |
f80e83a9 | 3347 | stdc = 0; |
3348 | #endif | |
3349 | result = alloc_number_token (pfile, stdc); | |
89b05ef6 | 3350 | } |
3351 | break; | |
3352 | ||
f80e83a9 | 3353 | case T_DATE: |
3354 | case T_TIME: | |
3355 | if (pfile->date == 0) | |
3356 | { | |
3357 | /* Allocate __DATE__ and __TIME__ from permanent storage, | |
3358 | and save them in pfile so we don't have to do this again. | |
3359 | We don't generate these strings at init time because | |
3360 | time() and localtime() are very slow on some systems. */ | |
3361 | time_t tt = time (NULL); | |
3362 | struct tm *tb = localtime (&tt); | |
3363 | ||
3364 | pfile->date = make_string_token | |
3365 | ((cpp_token *) xmalloc (sizeof (cpp_token)), DSC("Oct 11 1347")); | |
3366 | pfile->time = make_string_token | |
3367 | ((cpp_token *) xmalloc (sizeof (cpp_token)), DSC("12:34:56")); | |
3368 | ||
76faa4c0 | 3369 | sprintf ((char *) pfile->date->val.str.text, "%s %2d %4d", |
f80e83a9 | 3370 | monthnames[tb->tm_mon], tb->tm_mday, tb->tm_year + 1900); |
76faa4c0 | 3371 | sprintf ((char *) pfile->time->val.str.text, "%02d:%02d:%02d", |
f80e83a9 | 3372 | tb->tm_hour, tb->tm_min, tb->tm_sec); |
3373 | } | |
3374 | result = node->type == T_DATE ? pfile->date: pfile->time; | |
89b05ef6 | 3375 | break; |
3376 | ||
f80e83a9 | 3377 | case T_POISON: |
76faa4c0 | 3378 | cpp_error (pfile, "attempt to use poisoned \"%s\"", node->name); |
f80e83a9 | 3379 | return token; |
3380 | ||
3381 | default: | |
3382 | cpp_ice (pfile, "invalid special hash type"); | |
3383 | return token; | |
89b05ef6 | 3384 | } |
3385 | ||
f80e83a9 | 3386 | ASSIGN_FLAGS_AND_POS (result, token); |
3387 | return result; | |
3388 | } | |
3389 | #undef DSC | |
3390 | ||
b3954366 | 3391 | /* Allocate pfile->input_buffer, and initialize _cpp_trigraph_map[] |
f80e83a9 | 3392 | if it hasn't happened already. */ |
3393 | ||
3394 | void | |
3395 | _cpp_init_input_buffer (pfile) | |
3396 | cpp_reader *pfile; | |
3397 | { | |
deb356cf | 3398 | cpp_context *base; |
3399 | ||
deb356cf | 3400 | _cpp_init_toklist (&pfile->token_list, DUMMY_TOKEN); |
3401 | pfile->no_expand_level = UINT_MAX; | |
f80e83a9 | 3402 | pfile->context_cap = 20; |
f80e83a9 | 3403 | pfile->cur_context = 0; |
f80e83a9 | 3404 | |
deb356cf | 3405 | pfile->contexts = (cpp_context *) |
3406 | xmalloc (pfile->context_cap * sizeof (cpp_context)); | |
f80e83a9 | 3407 | |
deb356cf | 3408 | /* Clear the base context. */ |
3409 | base = &pfile->contexts[0]; | |
3410 | base->u.list = &pfile->token_list; | |
3411 | base->posn = 0; | |
3412 | base->count = 0; | |
3413 | base->args = 0; | |
3414 | base->level = 0; | |
3415 | base->flags = 0; | |
3416 | base->pushed_token = 0; | |
f80e83a9 | 3417 | } |
3418 | ||
3419 | /* Moves to the end of the directive line, popping contexts as | |
3420 | necessary. */ | |
3421 | void | |
3422 | _cpp_skip_rest_of_line (pfile) | |
3423 | cpp_reader *pfile; | |
3424 | { | |
deb356cf | 3425 | /* Discard all stacked contexts. */ |
3426 | int i; | |
3427 | for (i = pfile->cur_context; i > 0; i--) | |
3428 | if (pfile->contexts[i].args) | |
3429 | free_macro_args (pfile->contexts[i].args); | |
3430 | ||
3431 | if (pfile->no_expand_level <= pfile->cur_context) | |
3432 | pfile->no_expand_level = 0; | |
3433 | pfile->cur_context = 0; | |
f80e83a9 | 3434 | |
deb356cf | 3435 | /* Clear the base context, and clear the directive pointer so that |
3436 | get_raw_token will advance to the next line. */ | |
3437 | pfile->contexts[0].count = 0; | |
3438 | pfile->contexts[0].posn = 0; | |
f80e83a9 | 3439 | pfile->token_list.directive = 0; |
6060326b | 3440 | } |
3441 | ||
f80e83a9 | 3442 | /* Directive handler wrapper used by the command line option |
3443 | processor. */ | |
3444 | void | |
a336277c | 3445 | _cpp_run_directive (pfile, dir, buf, count, name) |
f80e83a9 | 3446 | cpp_reader *pfile; |
3447 | const struct directive *dir; | |
3448 | const char *buf; | |
3449 | size_t count; | |
a336277c | 3450 | const char *name; |
f80e83a9 | 3451 | { |
3452 | if (cpp_push_buffer (pfile, (const U_CHAR *)buf, count) != NULL) | |
3453 | { | |
3454 | unsigned int prev_lvl = 0; | |
deb356cf | 3455 | |
a336277c | 3456 | if (name) |
3457 | CPP_BUFFER (pfile)->nominal_fname = name; | |
3458 | else | |
3459 | CPP_BUFFER (pfile)->nominal_fname = _("<command line>"); | |
3460 | CPP_BUFFER (pfile)->lineno = (unsigned int)-1; | |
3461 | ||
deb356cf | 3462 | /* Scan the line now, else prevent_macro_expansion won't work. */ |
3463 | lex_next (pfile, 1); | |
f80e83a9 | 3464 | if (! (dir->flags & EXPAND)) |
3465 | prev_lvl = prevent_macro_expansion (pfile); | |
3466 | ||
3467 | (void) (*dir->handler) (pfile); | |
3468 | ||
3469 | if (! (dir->flags & EXPAND)) | |
3470 | restore_macro_expansion (pfile, prev_lvl); | |
3471 | ||
3472 | _cpp_skip_rest_of_line (pfile); | |
3473 | cpp_pop_buffer (pfile); | |
3474 | } | |
3475 | } |