]>
Commit | Line | Data |
---|---|---|
45b966db ZW |
1 | /* CPP Library - lexical analysis. |
2 | Copyright (C) 2000 Free Software Foundation, Inc. | |
3 | Contributed by Per Bothner, 1994-95. | |
4 | Based on CCCP program by Paul Rubin, June 1986 | |
5 | Adapted to ANSI C, Richard Stallman, Jan 1987 | |
6 | Broken out to separate file, Zack Weinberg, Mar 2000 | |
c5a04734 | 7 | Single-pass line tokenization by Neil Booth, April 2000 |
45b966db ZW |
8 | |
9 | This program is free software; you can redistribute it and/or modify it | |
10 | under the terms of the GNU General Public License as published by the | |
11 | Free Software Foundation; either version 2, or (at your option) any | |
12 | later version. | |
13 | ||
14 | This program is distributed in the hope that it will be useful, | |
15 | but WITHOUT ANY WARRANTY; without even the implied warranty of | |
16 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
17 | GNU General Public License for more details. | |
18 | ||
19 | You should have received a copy of the GNU General Public License | |
20 | along with this program; if not, write to the Free Software | |
21 | Foundation, 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. */ | |
22 | ||
041c3194 ZW |
23 | /* |
24 | ||
25 | Cleanups to do:- | |
26 | ||
041c3194 ZW |
27 | o Get use of digraphs in sync with the standard reqd on the command line. |
28 | o -dM and with _cpp_dump_list: too many \n output. | |
29 | o Put a printer object in cpp_reader? | |
30 | o Check line numbers assigned to all errors. | |
31 | o Replace strncmp with memcmp almost everywhere. | |
32 | o lex_line's use of cur_token, flags and list->token_used is a bit opaque. | |
041c3194 ZW |
33 | o Convert do_ functions to return void. Kaveh thinks its OK; and said he'll |
34 | give it a run when we've got some code. | |
041c3194 ZW |
35 | o Distinguish integers, floats, and 'other' pp-numbers. |
36 | o Store ints and char constants as binary values. | |
37 | o New command-line assertion syntax. | |
041c3194 ZW |
38 | o Work towards functions in cpperror.c taking a message level parameter. |
39 | If we do this, merge the common code of do_warning and do_error. | |
40 | o Comment all functions, and describe macro expansion algorithm. | |
41 | o Move as much out of header files as possible. | |
42 | o Remove single quote pairs `', and some '', from diagnostics. | |
43 | o Correct pastability test for CPP_NAME and CPP_NUMBER. | |
44 | ||
45 | */ | |
46 | ||
45b966db ZW |
47 | #include "config.h" |
48 | #include "system.h" | |
49 | #include "intl.h" | |
50 | #include "cpplib.h" | |
51 | #include "cpphash.h" | |
041c3194 | 52 | #include "symcat.h" |
45b966db | 53 | |
041c3194 ZW |
54 | #define auto_expand_name_space(list) \ |
55 | _cpp_expand_name_space ((list), 1 + (list)->name_cap / 2) | |
f2d5f0cc ZW |
56 | static void safe_fwrite PARAMS ((cpp_reader *, const U_CHAR *, |
57 | size_t, FILE *)); | |
041c3194 | 58 | static void dump_param_spelling PARAMS ((cpp_reader *, const cpp_toklist *, |
1368ee70 | 59 | unsigned int)); |
041c3194 | 60 | static void output_line_command PARAMS ((cpp_reader *, cpp_printer *, |
9e62c811 | 61 | unsigned int)); |
f2d5f0cc | 62 | |
041c3194 ZW |
63 | static void process_directive PARAMS ((cpp_reader *, const cpp_token *)); |
64 | static unsigned char *trigraph_replace PARAMS ((cpp_reader *, unsigned char *, | |
b8f41010 NB |
65 | unsigned char *)); |
66 | static const unsigned char *backslash_start PARAMS ((cpp_reader *, | |
67 | const unsigned char *)); | |
041c3194 ZW |
68 | static int skip_block_comment PARAMS ((cpp_reader *)); |
69 | static int skip_line_comment PARAMS ((cpp_reader *)); | |
b8f41010 | 70 | static void skip_whitespace PARAMS ((cpp_reader *, int)); |
bfb9dc7f ZW |
71 | static const U_CHAR *parse_name PARAMS ((cpp_reader *, cpp_token *, |
72 | const U_CHAR *, const U_CHAR *)); | |
73 | static void parse_number PARAMS ((cpp_reader *, cpp_toklist *, cpp_string *)); | |
041c3194 ZW |
74 | static void parse_string PARAMS ((cpp_reader *, cpp_toklist *, cpp_token *, |
75 | unsigned int)); | |
b8f41010 | 76 | static int trigraph_ok PARAMS ((cpp_reader *, const unsigned char *)); |
f624ffa7 NB |
77 | static void save_comment PARAMS ((cpp_toklist *, cpp_token *, |
78 | const unsigned char *, | |
ad265aa4 | 79 | unsigned int, unsigned int)); |
041c3194 ZW |
80 | static void lex_line PARAMS ((cpp_reader *, cpp_toklist *)); |
81 | static int lex_next PARAMS ((cpp_reader *, int)); | |
82 | static int is_macro_disabled PARAMS ((cpp_reader *, const cpp_toklist *, | |
83 | const cpp_token *)); | |
b8f41010 | 84 | |
041c3194 ZW |
85 | static cpp_token *stringify_arg PARAMS ((cpp_reader *, const cpp_token *)); |
86 | static void expand_context_stack PARAMS ((cpp_reader *)); | |
d1d9a6bd | 87 | static unsigned char * spell_token PARAMS ((cpp_reader *, const cpp_token *, |
041c3194 ZW |
88 | unsigned char *)); |
89 | static void output_token PARAMS ((cpp_reader *, const cpp_token *, | |
90 | const cpp_token *)); | |
b8f41010 NB |
91 | typedef unsigned int (* speller) PARAMS ((unsigned char *, cpp_toklist *, |
92 | cpp_token *)); | |
041c3194 ZW |
93 | static cpp_token *make_string_token PARAMS ((cpp_token *, const U_CHAR *, |
94 | unsigned int)); | |
95 | static cpp_token *alloc_number_token PARAMS ((cpp_reader *, int number)); | |
96 | static const cpp_token *special_symbol PARAMS ((cpp_reader *, cpp_hashnode *, | |
97 | const cpp_token *)); | |
98 | static cpp_token *duplicate_token PARAMS ((cpp_reader *, const cpp_token *)); | |
99 | static const cpp_token *maybe_paste_with_next PARAMS ((cpp_reader *, | |
100 | const cpp_token *)); | |
101 | static enum cpp_ttype can_paste PARAMS ((cpp_reader *, const cpp_token *, | |
102 | const cpp_token *, int *)); | |
103 | static unsigned int prevent_macro_expansion PARAMS ((cpp_reader *)); | |
104 | static void restore_macro_expansion PARAMS ((cpp_reader *, unsigned int)); | |
105 | static cpp_token *get_temp_token PARAMS ((cpp_reader *)); | |
106 | static void release_temp_tokens PARAMS ((cpp_reader *)); | |
107 | static U_CHAR * quote_string PARAMS ((U_CHAR *, const U_CHAR *, unsigned int)); | |
108 | static void process_directive PARAMS ((cpp_reader *, const cpp_token *)); | |
b8f41010 | 109 | |
bfb9dc7f ZW |
110 | #define INIT_TOKEN_STR(list, token) \ |
111 | do {(token)->val.str.len = 0; \ | |
112 | (token)->val.str.text = (list)->namebuf + (list)->name_used; \ | |
d1d9a6bd | 113 | } while (0) |
b8f41010 | 114 | |
041c3194 ZW |
115 | #define VALID_SIGN(c, prevc) \ |
116 | (((c) == '+' || (c) == '-') && \ | |
117 | ((prevc) == 'e' || (prevc) == 'E' \ | |
118 | || (((prevc) == 'p' || (prevc) == 'P') && !CPP_OPTION (pfile, c89)))) | |
119 | ||
b8f41010 NB |
120 | /* Handle LF, CR, CR-LF and LF-CR style newlines. Assumes next |
121 | character, if any, is in buffer. */ | |
041c3194 | 122 | |
b8f41010 | 123 | #define handle_newline(cur, limit, c) \ |
041c3194 | 124 | do { \ |
b8f41010 NB |
125 | if ((cur) < (limit) && *(cur) == '\r' + '\n' - c) \ |
126 | (cur)++; \ | |
041c3194 ZW |
127 | pfile->buffer->lineno++; \ |
128 | pfile->buffer->line_base = (cur); \ | |
6ab3e7dd | 129 | pfile->col_adjust = 0; \ |
041c3194 | 130 | } while (0) |
b8f41010 | 131 | |
041c3194 | 132 | #define IMMED_TOKEN() (!(cur_token->flags & PREV_WHITE)) |
b8f41010 NB |
133 | #define PREV_TOKEN_TYPE (cur_token[-1].type) |
134 | ||
f617b8e2 NB |
135 | #define PUSH_TOKEN(ttype) cur_token++->type = ttype |
136 | #define REVISE_TOKEN(ttype) cur_token[-1].type = ttype | |
137 | #define BACKUP_TOKEN(ttype) (--cur_token)->type = ttype | |
138 | #define BACKUP_DIGRAPH(ttype) do { \ | |
139 | BACKUP_TOKEN(ttype); cur_token->flags |= DIGRAPH;} while (0) | |
140 | ||
141 | /* An upper bound on the number of bytes needed to spell a token, | |
142 | including preceding whitespace. */ | |
bfb9dc7f ZW |
143 | #define TOKEN_SPELL(token) token_spellings[(token)->type].type |
144 | #define TOKEN_LEN(token) (5 + (TOKEN_SPELL(token) == SPELL_STRING \ | |
145 | ? (token)->val.str.len \ | |
146 | : (TOKEN_SPELL(token) == SPELL_IDENT \ | |
147 | ? (token)->val.node->length \ | |
148 | : 0))) | |
f617b8e2 | 149 | |
f617b8e2 | 150 | #define T(e, s) {SPELL_OPERATOR, (const U_CHAR *) s}, |
5d7ee2fa NB |
151 | #define I(e, s) {SPELL_IDENT, s}, |
152 | #define S(e, s) {SPELL_STRING, s}, | |
b8f41010 NB |
153 | #define C(e, s) {SPELL_CHAR, s}, |
154 | #define N(e, s) {SPELL_NONE, s}, | |
b8f41010 | 155 | |
041c3194 ZW |
156 | const struct token_spelling |
157 | token_spellings [N_TTYPES + 1] = {TTYPE_TABLE {0, 0} }; | |
158 | ||
159 | #undef T | |
160 | #undef I | |
161 | #undef S | |
162 | #undef C | |
163 | #undef N | |
164 | ||
165 | /* For debugging: the internal names of the tokens. */ | |
cf00a885 ZW |
166 | #define T(e, s) U STRINGX(e) + 4, |
167 | #define I(e, s) U STRINGX(e) + 4, | |
168 | #define S(e, s) U STRINGX(e) + 4, | |
169 | #define C(e, s) U STRINGX(e) + 4, | |
170 | #define N(e, s) U STRINGX(e) + 4, | |
041c3194 | 171 | |
cf00a885 | 172 | const U_CHAR *const token_names[N_TTYPES] = { TTYPE_TABLE }; |
b8f41010 NB |
173 | |
174 | #undef T | |
5d7ee2fa NB |
175 | #undef I |
176 | #undef S | |
b8f41010 NB |
177 | #undef C |
178 | #undef N | |
b8f41010 | 179 | |
041c3194 ZW |
180 | /* The following table is used by trigraph_ok/trigraph_replace. If we |
181 | have designated initializers, it can be constant data; otherwise, | |
182 | it is set up at runtime by _cpp_init_input_buffer. */ | |
183 | ||
184 | #if (GCC_VERSION >= 2007) | |
185 | #define init_trigraph_map() /* nothing */ | |
186 | #define TRIGRAPH_MAP \ | |
187 | __extension__ static const U_CHAR trigraph_map[UCHAR_MAX + 1] = { | |
188 | #define END }; | |
189 | #define s(p, v) [p] = v, | |
190 | #else | |
191 | #define TRIGRAPH_MAP static U_CHAR trigraph_map[UCHAR_MAX + 1] = { 0 }; \ | |
192 | static void init_trigraph_map PARAMS ((void)) { \ | |
193 | unsigned char *x = trigraph_map; | |
194 | #define END } | |
195 | #define s(p, v) x[p] = v; | |
196 | #endif | |
197 | ||
198 | TRIGRAPH_MAP | |
199 | s('=', '#') s(')', ']') s('!', '|') | |
200 | s('(', '[') s('\'', '^') s('>', '}') | |
201 | s('/', '\\') s('<', '{') s('-', '~') | |
202 | END | |
203 | ||
204 | #undef TRIGRAPH_MAP | |
205 | #undef END | |
206 | #undef s | |
207 | ||
45b966db ZW |
208 | /* Re-allocates PFILE->token_buffer so it will hold at least N more chars. */ |
209 | ||
210 | void | |
211 | _cpp_grow_token_buffer (pfile, n) | |
212 | cpp_reader *pfile; | |
213 | long n; | |
214 | { | |
215 | long old_written = CPP_WRITTEN (pfile); | |
216 | pfile->token_buffer_size = n + 2 * pfile->token_buffer_size; | |
217 | pfile->token_buffer = (U_CHAR *) | |
218 | xrealloc(pfile->token_buffer, pfile->token_buffer_size); | |
219 | CPP_SET_WRITTEN (pfile, old_written); | |
220 | } | |
221 | ||
f2d5f0cc ZW |
222 | /* Deal with the annoying semantics of fwrite. */ |
223 | static void | |
224 | safe_fwrite (pfile, buf, len, fp) | |
225 | cpp_reader *pfile; | |
226 | const U_CHAR *buf; | |
227 | size_t len; | |
228 | FILE *fp; | |
229 | { | |
230 | size_t count; | |
45b966db | 231 | |
f2d5f0cc ZW |
232 | while (len) |
233 | { | |
234 | count = fwrite (buf, 1, len, fp); | |
235 | if (count == 0) | |
236 | goto error; | |
237 | len -= count; | |
238 | buf += count; | |
239 | } | |
240 | return; | |
241 | ||
242 | error: | |
243 | cpp_notice_from_errno (pfile, CPP_OPTION (pfile, out_fname)); | |
244 | } | |
245 | ||
246 | /* Notify the compiler proper that the current line number has jumped, | |
247 | or the current file name has changed. */ | |
248 | ||
249 | static void | |
1368ee70 | 250 | output_line_command (pfile, print, line) |
45b966db | 251 | cpp_reader *pfile; |
f2d5f0cc | 252 | cpp_printer *print; |
1368ee70 | 253 | unsigned int line; |
45b966db | 254 | { |
041c3194 | 255 | cpp_buffer *ip = CPP_BUFFER (pfile); |
f2d5f0cc ZW |
256 | enum { same = 0, enter, leave, rname } change; |
257 | static const char * const codes[] = { "", " 1", " 2", "" }; | |
258 | ||
041c3194 ZW |
259 | if (line == 0) |
260 | return; | |
261 | ||
262 | /* End the previous line of text. */ | |
263 | if (pfile->need_newline) | |
264 | putc ('\n', print->outf); | |
265 | pfile->need_newline = 0; | |
266 | ||
f2d5f0cc ZW |
267 | if (CPP_OPTION (pfile, no_line_commands)) |
268 | return; | |
269 | ||
041c3194 ZW |
270 | /* If ip is null, we've been called from cpp_finish, and they just |
271 | needed the final flush and trailing newline. */ | |
272 | if (!ip) | |
273 | return; | |
274 | ||
fb753f88 | 275 | if (pfile->include_depth == print->last_id) |
54bef41d | 276 | { |
fb753f88 JJ |
277 | /* Determine whether the current filename has changed, and if so, |
278 | how. 'nominal_fname' values are unique, so they can be compared | |
279 | by comparing pointers. */ | |
280 | if (ip->nominal_fname == print->last_fname) | |
281 | change = same; | |
282 | else | |
0e500c78 | 283 | change = rname; |
fb753f88 JJ |
284 | } |
285 | else | |
286 | { | |
287 | if (pfile->include_depth > print->last_id) | |
288 | change = enter; | |
f2d5f0cc | 289 | else |
fb753f88 JJ |
290 | change = leave; |
291 | print->last_id = pfile->include_depth; | |
45b966db | 292 | } |
fb753f88 JJ |
293 | print->last_fname = ip->nominal_fname; |
294 | ||
f2d5f0cc ZW |
295 | /* If the current file has not changed, we can output a few newlines |
296 | instead if we want to increase the line number by a small amount. | |
297 | We cannot do this if print->lineno is zero, because that means we | |
298 | haven't output any line commands yet. (The very first line | |
299 | command output is a `same_file' command.) */ | |
041c3194 | 300 | if (change == same && print->lineno > 0 |
f2d5f0cc | 301 | && line >= print->lineno && line < print->lineno + 8) |
45b966db | 302 | { |
f2d5f0cc | 303 | while (line > print->lineno) |
45b966db | 304 | { |
f2d5f0cc ZW |
305 | putc ('\n', print->outf); |
306 | print->lineno++; | |
45b966db | 307 | } |
f2d5f0cc | 308 | return; |
45b966db | 309 | } |
f2d5f0cc ZW |
310 | |
311 | #ifndef NO_IMPLICIT_EXTERN_C | |
312 | if (CPP_OPTION (pfile, cplusplus)) | |
313 | fprintf (print->outf, "# %u \"%s\"%s%s%s\n", line, ip->nominal_fname, | |
314 | codes[change], | |
c31a6508 ZW |
315 | ip->inc->sysp ? " 3" : "", |
316 | (ip->inc->sysp == 2) ? " 4" : ""); | |
f2d5f0cc ZW |
317 | else |
318 | #endif | |
319 | fprintf (print->outf, "# %u \"%s\"%s%s\n", line, ip->nominal_fname, | |
320 | codes[change], | |
c31a6508 | 321 | ip->inc->sysp ? " 3" : ""); |
f2d5f0cc ZW |
322 | print->lineno = line; |
323 | } | |
324 | ||
325 | /* Write the contents of the token_buffer to the output stream, and | |
326 | clear the token_buffer. Also handles generating line commands and | |
327 | keeping track of file transitions. */ | |
328 | ||
329 | void | |
041c3194 | 330 | cpp_output_tokens (pfile, print, line) |
f2d5f0cc ZW |
331 | cpp_reader *pfile; |
332 | cpp_printer *print; | |
041c3194 | 333 | unsigned int line; |
f2d5f0cc | 334 | { |
f6fab919 ZW |
335 | if (CPP_WRITTEN (pfile) - print->written) |
336 | { | |
f6fab919 ZW |
337 | safe_fwrite (pfile, pfile->token_buffer, |
338 | CPP_WRITTEN (pfile) - print->written, print->outf); | |
041c3194 ZW |
339 | pfile->need_newline = 1; |
340 | if (print->lineno) | |
341 | print->lineno++; | |
45b966db | 342 | |
041c3194 | 343 | CPP_SET_WRITTEN (pfile, print->written); |
f2d5f0cc | 344 | } |
041c3194 | 345 | output_line_command (pfile, print, line); |
45b966db ZW |
346 | } |
347 | ||
c56c2073 | 348 | /* Scan until CPP_BUFFER (PFILE) is exhausted, discarding output. */ |
f2d5f0cc ZW |
349 | |
350 | void | |
351 | cpp_scan_buffer_nooutput (pfile) | |
352 | cpp_reader *pfile; | |
353 | { | |
f2d5f0cc | 354 | unsigned int old_written = CPP_WRITTEN (pfile); |
041c3194 ZW |
355 | cpp_buffer *stop = CPP_PREV_BUFFER (CPP_BUFFER (pfile)); |
356 | ||
f2d5f0cc ZW |
357 | for (;;) |
358 | { | |
041c3194 ZW |
359 | /* In no-output mode, we can ignore everything but directives. */ |
360 | const cpp_token *token = cpp_get_token (pfile); | |
361 | if (token->type == CPP_EOF) | |
362 | { | |
363 | cpp_pop_buffer (pfile); | |
364 | if (CPP_BUFFER (pfile) == stop) | |
365 | break; | |
366 | } | |
367 | _cpp_skip_rest_of_line (pfile); | |
f2d5f0cc ZW |
368 | } |
369 | CPP_SET_WRITTEN (pfile, old_written); | |
370 | } | |
371 | ||
c56c2073 | 372 | /* Scan until CPP_BUFFER (pfile) is exhausted, writing output to PRINT. */ |
f2d5f0cc ZW |
373 | |
374 | void | |
375 | cpp_scan_buffer (pfile, print) | |
376 | cpp_reader *pfile; | |
377 | cpp_printer *print; | |
378 | { | |
c56c2073 | 379 | cpp_buffer *stop = CPP_PREV_BUFFER (CPP_BUFFER (pfile)); |
041c3194 | 380 | const cpp_token *token, *prev = 0; |
f2d5f0cc ZW |
381 | |
382 | for (;;) | |
383 | { | |
384 | token = cpp_get_token (pfile); | |
041c3194 | 385 | if (token->type == CPP_EOF) |
f2d5f0cc | 386 | { |
041c3194 ZW |
387 | cpp_pop_buffer (pfile); |
388 | if (CPP_BUFFER (pfile) == stop) | |
f2d5f0cc | 389 | return; |
041c3194 ZW |
390 | cpp_output_tokens (pfile, print, CPP_BUF_LINE (CPP_BUFFER (pfile))); |
391 | prev = 0; | |
392 | continue; | |
393 | } | |
394 | ||
395 | if (token->flags & BOL) | |
396 | { | |
397 | cpp_output_tokens (pfile, print, pfile->token_list.line); | |
398 | prev = 0; | |
f2d5f0cc | 399 | } |
041c3194 ZW |
400 | |
401 | output_token (pfile, token, prev); | |
402 | prev = token; | |
f2d5f0cc ZW |
403 | } |
404 | } | |
405 | ||
041c3194 ZW |
406 | /* Helper routine used by parse_include, which can't see spell_token. |
407 | Reinterpret the current line as an h-char-sequence (< ... >); we are | |
408 | looking at the first token after the <. */ | |
409 | const cpp_token * | |
410 | _cpp_glue_header_name (pfile) | |
45b966db ZW |
411 | cpp_reader *pfile; |
412 | { | |
041c3194 ZW |
413 | unsigned int written = CPP_WRITTEN (pfile); |
414 | const cpp_token *t; | |
415 | cpp_token *hdr; | |
416 | U_CHAR *buf; | |
417 | size_t len; | |
418 | ||
419 | for (;;) | |
420 | { | |
421 | t = cpp_get_token (pfile); | |
422 | if (t->type == CPP_GREATER || t->type == CPP_EOF) | |
423 | break; | |
424 | ||
425 | CPP_RESERVE (pfile, TOKEN_LEN (t)); | |
426 | if (t->flags & PREV_WHITE) | |
427 | CPP_PUTC_Q (pfile, ' '); | |
428 | pfile->limit = spell_token (pfile, t, pfile->limit); | |
429 | } | |
430 | ||
431 | if (t->type == CPP_EOF) | |
432 | cpp_error (pfile, "missing terminating > character"); | |
45b966db | 433 | |
041c3194 ZW |
434 | len = CPP_WRITTEN (pfile) - written; |
435 | buf = xmalloc (len); | |
436 | memcpy (buf, pfile->token_buffer + written, len); | |
437 | CPP_SET_WRITTEN (pfile, written); | |
438 | ||
439 | hdr = get_temp_token (pfile); | |
440 | hdr->type = CPP_HEADER_NAME; | |
441 | hdr->flags = 0; | |
bfb9dc7f ZW |
442 | hdr->val.str.text = buf; |
443 | hdr->val.str.len = len; | |
041c3194 | 444 | return hdr; |
45b966db ZW |
445 | } |
446 | ||
1368ee70 ZW |
447 | /* Token-buffer helper functions. */ |
448 | ||
d1d9a6bd NB |
449 | /* Expand a token list's string space. It is *vital* that |
450 | list->tokens_used is correct, to get pointer fix-up right. */ | |
041c3194 ZW |
451 | void |
452 | _cpp_expand_name_space (list, len) | |
1368ee70 | 453 | cpp_toklist *list; |
c5a04734 ZW |
454 | unsigned int len; |
455 | { | |
f617b8e2 | 456 | const U_CHAR *old_namebuf; |
f617b8e2 NB |
457 | |
458 | old_namebuf = list->namebuf; | |
c5a04734 ZW |
459 | list->name_cap += len; |
460 | list->namebuf = (unsigned char *) xrealloc (list->namebuf, list->name_cap); | |
f617b8e2 NB |
461 | |
462 | /* Fix up token text pointers. */ | |
79f50f2a | 463 | if (list->namebuf != old_namebuf) |
f617b8e2 NB |
464 | { |
465 | unsigned int i; | |
466 | ||
467 | for (i = 0; i < list->tokens_used; i++) | |
bfb9dc7f ZW |
468 | if (token_spellings[list->tokens[i].type].type == SPELL_STRING) |
469 | list->tokens[i].val.str.text += (list->namebuf - old_namebuf); | |
f617b8e2 | 470 | } |
1368ee70 ZW |
471 | } |
472 | ||
041c3194 ZW |
473 | /* If there is not enough room for LEN more characters, expand the |
474 | list by just enough to have room for LEN characters. */ | |
475 | void | |
476 | _cpp_reserve_name_space (list, len) | |
477 | cpp_toklist *list; | |
478 | unsigned int len; | |
479 | { | |
480 | unsigned int room = list->name_cap - list->name_used; | |
481 | ||
482 | if (room < len) | |
483 | _cpp_expand_name_space (list, len - room); | |
484 | } | |
485 | ||
1368ee70 | 486 | /* Expand the number of tokens in a list. */ |
d1d9a6bd NB |
487 | void |
488 | _cpp_expand_token_space (list, count) | |
1368ee70 | 489 | cpp_toklist *list; |
d1d9a6bd | 490 | unsigned int count; |
1368ee70 | 491 | { |
d1d9a6bd NB |
492 | unsigned int n; |
493 | ||
494 | list->tokens_cap += count; | |
495 | n = list->tokens_cap; | |
15dad1d9 | 496 | if (list->flags & LIST_OFFSET) |
d1d9a6bd | 497 | list->tokens--, n++; |
1368ee70 | 498 | list->tokens = (cpp_token *) |
d1d9a6bd | 499 | xrealloc (list->tokens, n * sizeof (cpp_token)); |
15dad1d9 ZW |
500 | if (list->flags & LIST_OFFSET) |
501 | list->tokens++; /* Skip the dummy. */ | |
1368ee70 ZW |
502 | } |
503 | ||
d1d9a6bd NB |
504 | /* Initialize a token list. If flags is DUMMY_TOKEN, we allocate |
505 | an extra token in front of the token list, as this allows the lexer | |
506 | to always peek at the previous token without worrying about | |
507 | underflowing the list, and some initial space. Otherwise, no | |
508 | token- or name-space is allocated, and there is no dummy token. */ | |
15dad1d9 | 509 | void |
d1d9a6bd | 510 | _cpp_init_toklist (list, flags) |
1368ee70 | 511 | cpp_toklist *list; |
d1d9a6bd | 512 | int flags; |
1368ee70 | 513 | { |
d1d9a6bd NB |
514 | if (flags == NO_DUMMY_TOKEN) |
515 | { | |
516 | list->tokens_cap = 0; | |
041c3194 | 517 | list->tokens = 0; |
d1d9a6bd | 518 | list->name_cap = 0; |
041c3194 | 519 | list->namebuf = 0; |
d1d9a6bd NB |
520 | list->flags = 0; |
521 | } | |
522 | else | |
523 | { | |
524 | /* Initialize token space. Put a dummy token before the start | |
525 | that will fail matches. */ | |
526 | list->tokens_cap = 256; /* 4K's worth. */ | |
527 | list->tokens = (cpp_token *) | |
528 | xmalloc ((list->tokens_cap + 1) * sizeof (cpp_token)); | |
529 | list->tokens[0].type = CPP_EOF; | |
530 | list->tokens++; | |
531 | ||
532 | /* Initialize name space. */ | |
533 | list->name_cap = 1024; | |
041c3194 | 534 | list->namebuf = (unsigned char *) xmalloc (list->name_cap); |
d1d9a6bd NB |
535 | list->flags = LIST_OFFSET; |
536 | } | |
15dad1d9 | 537 | |
15dad1d9 ZW |
538 | _cpp_clear_toklist (list); |
539 | } | |
1368ee70 | 540 | |
15dad1d9 ZW |
541 | /* Clear a token list. */ |
542 | void | |
543 | _cpp_clear_toklist (list) | |
544 | cpp_toklist *list; | |
545 | { | |
c5a04734 ZW |
546 | list->tokens_used = 0; |
547 | list->name_used = 0; | |
041c3194 ZW |
548 | list->directive = 0; |
549 | list->paramc = 0; | |
550 | list->params_len = 0; | |
15dad1d9 ZW |
551 | list->flags &= LIST_OFFSET; /* clear all but that one */ |
552 | } | |
553 | ||
554 | /* Free a token list. Does not free the list itself, which may be | |
555 | embedded in a larger structure. */ | |
556 | void | |
557 | _cpp_free_toklist (list) | |
041c3194 | 558 | const cpp_toklist *list; |
15dad1d9 | 559 | { |
15dad1d9 ZW |
560 | if (list->flags & LIST_OFFSET) |
561 | free (list->tokens - 1); /* Backup over dummy token. */ | |
562 | else | |
563 | free (list->tokens); | |
564 | free (list->namebuf); | |
1368ee70 ZW |
565 | } |
566 | ||
15dad1d9 ZW |
567 | /* Compare two tokens. */ |
568 | int | |
569 | _cpp_equiv_tokens (a, b) | |
570 | const cpp_token *a, *b; | |
571 | { | |
041c3194 ZW |
572 | if (a->type == b->type && a->flags == b->flags) |
573 | switch (token_spellings[a->type].type) | |
574 | { | |
575 | default: /* Keep compiler happy. */ | |
576 | case SPELL_OPERATOR: | |
577 | return 1; | |
578 | case SPELL_CHAR: | |
579 | case SPELL_NONE: | |
580 | return a->val.aux == b->val.aux; /* arg_no or character. */ | |
581 | case SPELL_IDENT: | |
bfb9dc7f | 582 | return a->val.node == b->val.node; |
041c3194 | 583 | case SPELL_STRING: |
bfb9dc7f ZW |
584 | return (a->val.str.len == b->val.str.len |
585 | && !memcmp (a->val.str.text, b->val.str.text, | |
586 | a->val.str.len)); | |
041c3194 | 587 | } |
15dad1d9 | 588 | |
041c3194 | 589 | return 0; |
15dad1d9 ZW |
590 | } |
591 | ||
592 | /* Compare two token lists. */ | |
593 | int | |
594 | _cpp_equiv_toklists (a, b) | |
595 | const cpp_toklist *a, *b; | |
596 | { | |
597 | unsigned int i; | |
598 | ||
041c3194 ZW |
599 | if (a->tokens_used != b->tokens_used |
600 | || a->flags != b->flags | |
601 | || a->paramc != b->paramc) | |
15dad1d9 ZW |
602 | return 0; |
603 | ||
604 | for (i = 0; i < a->tokens_used; i++) | |
605 | if (! _cpp_equiv_tokens (&a->tokens[i], &b->tokens[i])) | |
606 | return 0; | |
607 | return 1; | |
608 | } | |
609 | ||
041c3194 | 610 | /* Utility routine: |
9e62c811 | 611 | |
bfb9dc7f ZW |
612 | Compares, the token TOKEN to the NUL-terminated string STRING. |
613 | TOKEN must be a CPP_NAME. Returns 1 for equal, 0 for unequal. */ | |
15dad1d9 | 614 | |
041c3194 | 615 | int |
bfb9dc7f ZW |
616 | cpp_ideq (token, string) |
617 | const cpp_token *token; | |
041c3194 ZW |
618 | const char *string; |
619 | { | |
bfb9dc7f | 620 | if (token->type != CPP_NAME) |
041c3194 | 621 | return 0; |
bfb9dc7f ZW |
622 | |
623 | return !ustrcmp (token->val.node->name, (const U_CHAR *)string); | |
15dad1d9 | 624 | } |
1368ee70 | 625 | |
041c3194 | 626 | /* Lexing algorithm. |
45b966db | 627 | |
041c3194 ZW |
628 | The original lexer in cpplib was made up of two passes: a first pass |
629 | that replaced trigraphs and deleted esacped newlines, and a second | |
630 | pass that tokenized the result of the first pass. Tokenisation was | |
631 | performed by peeking at the next character in the input stream. For | |
632 | example, if the input stream contained "!=", the handler for the ! | |
633 | character would peek at the next character, and if it were a '=' | |
634 | would skip over it, and return a "!=" token, otherwise it would | |
635 | return just the "!" token. | |
61474454 | 636 | |
041c3194 ZW |
637 | To implement a single-pass lexer, this peeking ahead is unworkable. |
638 | An arbitrary number of escaped newlines, and trigraphs (in particular | |
639 | ??/ which translates to the escape \), could separate the '!' and '=' | |
640 | in the input stream, yet the next token is still a "!=". | |
61474454 | 641 | |
041c3194 ZW |
642 | Suppose instead that we lex by one logical line at a time, producing |
643 | a token list or stack for each logical line, and when seeing the '!' | |
644 | push a CPP_NOT token on the list. Then if the '!' is part of a | |
645 | longer token ("!=") we know we must see the remainder of the token by | |
646 | the time we reach the end of the logical line. Thus we can have the | |
647 | '=' handler look at the previous token (at the end of the list / top | |
648 | of the stack) and see if it is a "!" token, and if so, instead of | |
649 | pushing a "=" token revise the existing token to be a "!=" token. | |
61474454 | 650 | |
041c3194 ZW |
651 | This works in the presence of escaped newlines, because the '\' would |
652 | have been pushed on the top of the stack as a CPP_BACKSLASH. The | |
653 | newline ('\n' or '\r') handler looks at the token at the top of the | |
654 | stack to see if it is a CPP_BACKSLASH, and if so discards both. | |
655 | Otherwise it pushes the newline (CPP_VSPACE) token as normal. Hence | |
656 | the '=' handler would never see any intervening escaped newlines. | |
45b966db | 657 | |
041c3194 ZW |
658 | To make trigraphs work in this context, as in precedence trigraphs |
659 | are highest and converted before anything else, the '?' handler does | |
660 | lookahead to see if it is a trigraph, and if so skips the trigraph | |
661 | and pushes the token it represents onto the top of the stack. This | |
662 | also works in the particular case of a CPP_BACKSLASH trigraph. | |
45b966db | 663 | |
041c3194 ZW |
664 | To the preprocessor, whitespace is only significant to the point of |
665 | knowing whether whitespace precedes a particular token. For example, | |
666 | the '=' handler needs to know whether there was whitespace between it | |
667 | and a "!" token on the top of the stack, to make the token conversion | |
668 | decision correctly. So each token has a PREV_WHITE flag to | |
669 | indicate this - the standard permits consecutive whitespace to be | |
670 | regarded as a single space. The compiler front ends are not | |
671 | interested in whitespace at all; they just require a token stream. | |
672 | Another place where whitespace is significant to the preprocessor is | |
673 | a #define statment - if there is whitespace between the macro name | |
674 | and an initial "(" token the macro is "object-like", otherwise it is | |
675 | a function-like macro that takes arguments. | |
676 | ||
677 | However, all is not rosy. Parsing of identifiers, numbers, comments | |
678 | and strings becomes trickier because of the possibility of raw | |
679 | trigraphs and escaped newlines in the input stream. | |
680 | ||
681 | The trigraphs are three consecutive characters beginning with two | |
682 | question marks. A question mark is not valid as part of a number or | |
683 | identifier, so parsing of a number or identifier terminates normally | |
684 | upon reaching it, returning to the mainloop which handles the | |
685 | trigraph just like it would in any other position. Similarly for the | |
686 | backslash of a backslash-newline combination. So we just need the | |
687 | escaped-newline dropper in the mainloop to check if the token on the | |
688 | top of the stack after dropping the escaped newline is a number or | |
689 | identifier, and if so to continue the processing it as if nothing had | |
690 | happened. | |
691 | ||
692 | For strings, we replace trigraphs whenever we reach a quote or | |
693 | newline, because there might be a backslash trigraph escaping them. | |
694 | We need to be careful that we start trigraph replacing from where we | |
695 | left off previously, because it is possible for a first scan to leave | |
696 | "fake" trigraphs that a second scan would pick up as real (e.g. the | |
697 | sequence "????/\n=" would find a fake ??= trigraph after removing the | |
698 | escaped newline.) | |
699 | ||
700 | For line comments, on reaching a newline we scan the previous | |
701 | character(s) to see if it escaped, and continue if it is. Block | |
702 | comments ignore everything and just focus on finding the comment | |
703 | termination mark. The only difficult thing, and it is surprisingly | |
704 | tricky, is checking if an asterisk precedes the final slash since | |
705 | they could be separated by escaped newlines. If the preprocessor is | |
706 | invoked with the output comments option, we don't bother removing | |
707 | escaped newlines and replacing trigraphs for output. | |
708 | ||
709 | Finally, numbers can begin with a period, which is pushed initially | |
710 | as a CPP_DOT token in its own right. The digit handler checks if the | |
711 | previous token was a CPP_DOT not separated by whitespace, and if so | |
712 | pops it off the stack and pushes a period into the number's buffer | |
713 | before calling the number parser. | |
45b966db | 714 | |
041c3194 ZW |
715 | */ |
716 | ||
717 | static const unsigned char *digraph_spellings [] = {U"%:", U"%:%:", U"<:", | |
718 | U":>", U"<%", U"%>"}; | |
45b966db | 719 | |
041c3194 ZW |
720 | /* Call when a trigraph is encountered. It warns if necessary, and |
721 | returns true if the trigraph should be honoured. END is the third | |
722 | character of a trigraph in the input stream. */ | |
45b966db | 723 | static int |
041c3194 | 724 | trigraph_ok (pfile, end) |
45b966db | 725 | cpp_reader *pfile; |
041c3194 | 726 | const unsigned char *end; |
45b966db | 727 | { |
041c3194 ZW |
728 | int accept = CPP_OPTION (pfile, trigraphs); |
729 | ||
730 | if (CPP_OPTION (pfile, warn_trigraphs)) | |
45b966db | 731 | { |
041c3194 ZW |
732 | unsigned int col = end - 1 - pfile->buffer->line_base; |
733 | if (accept) | |
734 | cpp_warning_with_line (pfile, pfile->buffer->lineno, col, | |
735 | "trigraph ??%c converted to %c", | |
736 | (int) *end, (int) trigraph_map[*end]); | |
45b966db | 737 | else |
041c3194 ZW |
738 | cpp_warning_with_line (pfile, pfile->buffer->lineno, col, |
739 | "trigraph ??%c ignored", (int) *end); | |
45b966db | 740 | } |
041c3194 | 741 | return accept; |
45b966db ZW |
742 | } |
743 | ||
041c3194 ZW |
744 | /* Scan a string for trigraphs, warning or replacing them inline as |
745 | appropriate. When parsing a string, we must call this routine | |
746 | before processing a newline character (if trigraphs are enabled), | |
747 | since the newline might be escaped by a preceding backslash | |
748 | trigraph sequence. Returns a pointer to the end of the name after | |
749 | replacement. */ | |
750 | ||
751 | static unsigned char * | |
752 | trigraph_replace (pfile, src, limit) | |
45b966db | 753 | cpp_reader *pfile; |
041c3194 ZW |
754 | unsigned char *src; |
755 | unsigned char *limit; | |
45b966db | 756 | { |
041c3194 ZW |
757 | unsigned char *dest; |
758 | ||
759 | /* Starting with src[1], find two consecutive '?'. The case of no | |
760 | trigraphs is streamlined. */ | |
761 | ||
043afb2a | 762 | for (src++; src + 1 < limit; src += 2) |
041c3194 ZW |
763 | { |
764 | if (src[0] != '?') | |
765 | continue; | |
766 | ||
767 | /* Make src point to the 1st (NOT 2nd) of two consecutive '?'s. */ | |
768 | if (src[-1] == '?') | |
769 | src--; | |
770 | else if (src + 2 == limit || src[1] != '?') | |
771 | continue; | |
45b966db | 772 | |
041c3194 ZW |
773 | /* Check if it really is a trigraph. */ |
774 | if (trigraph_map[src[2]] == 0) | |
775 | continue; | |
45b966db | 776 | |
041c3194 ZW |
777 | dest = src; |
778 | goto trigraph_found; | |
779 | } | |
780 | return limit; | |
45b966db | 781 | |
041c3194 ZW |
782 | /* Now we have a trigraph, we need to scan the remaining buffer, and |
783 | copy-shifting its contents left if replacement is enabled. */ | |
784 | for (; src + 2 < limit; dest++, src++) | |
785 | if ((*dest = *src) == '?' && src[1] == '?' && trigraph_map[src[2]]) | |
786 | { | |
787 | trigraph_found: | |
788 | src += 2; | |
789 | if (trigraph_ok (pfile, pfile->buffer->cur - (limit - src))) | |
790 | *dest = trigraph_map[*src]; | |
791 | } | |
792 | ||
793 | /* Copy remaining (at most 2) characters. */ | |
794 | while (src < limit) | |
795 | *dest++ = *src++; | |
796 | return dest; | |
45b966db ZW |
797 | } |
798 | ||
041c3194 ZW |
799 | /* If CUR is a backslash or the end of a trigraphed backslash, return |
800 | a pointer to its beginning, otherwise NULL. We don't read beyond | |
801 | the buffer start, because there is the start of the comment in the | |
802 | buffer. */ | |
803 | static const unsigned char * | |
804 | backslash_start (pfile, cur) | |
64aaf407 | 805 | cpp_reader *pfile; |
041c3194 | 806 | const unsigned char *cur; |
64aaf407 | 807 | { |
041c3194 ZW |
808 | if (cur[0] == '\\') |
809 | return cur; | |
810 | if (cur[0] == '/' && cur[-1] == '?' && cur[-2] == '?' | |
811 | && trigraph_ok (pfile, cur)) | |
812 | return cur - 2; | |
813 | return 0; | |
64aaf407 NB |
814 | } |
815 | ||
041c3194 ZW |
816 | /* Skip a C-style block comment. This is probably the trickiest |
817 | handler. We find the end of the comment by seeing if an asterisk | |
818 | is before every '/' we encounter. The nasty complication is that a | |
819 | previous asterisk may be separated by one or more escaped newlines. | |
820 | Returns non-zero if comment terminated by EOF, zero otherwise. */ | |
821 | static int | |
822 | skip_block_comment (pfile) | |
45b966db ZW |
823 | cpp_reader *pfile; |
824 | { | |
041c3194 ZW |
825 | cpp_buffer *buffer = pfile->buffer; |
826 | const unsigned char *char_after_star = 0; | |
827 | register const unsigned char *cur = buffer->cur; | |
828 | int seen_eof = 0; | |
829 | ||
830 | /* Inner loop would think the comment has ended if the first comment | |
831 | character is a '/'. Avoid this and keep the inner loop clean by | |
832 | skipping such a character. */ | |
833 | if (cur < buffer->rlimit && cur[0] == '/') | |
834 | cur++; | |
64aaf407 | 835 | |
041c3194 | 836 | for (; cur < buffer->rlimit; ) |
45b966db | 837 | { |
041c3194 ZW |
838 | unsigned char c = *cur++; |
839 | ||
840 | /* People like decorating comments with '*', so check for | |
841 | '/' instead for efficiency. */ | |
842 | if (c == '/') | |
45b966db | 843 | { |
041c3194 ZW |
844 | if (cur[-2] == '*' || cur - 1 == char_after_star) |
845 | goto out; | |
846 | ||
847 | /* Warn about potential nested comments, but not when | |
848 | the final character inside the comment is a '/'. | |
849 | Don't bother to get it right across escaped newlines. */ | |
850 | if (CPP_OPTION (pfile, warn_comments) && cur + 1 < buffer->rlimit | |
851 | && cur[0] == '*' && cur[1] != '/') | |
45b966db | 852 | { |
041c3194 ZW |
853 | buffer->cur = cur; |
854 | cpp_warning (pfile, "'/*' within comment"); | |
45b966db | 855 | } |
45b966db | 856 | } |
91fcd158 | 857 | else if (is_vspace (c)) |
45b966db | 858 | { |
041c3194 ZW |
859 | const unsigned char* bslash = backslash_start (pfile, cur - 2); |
860 | ||
861 | handle_newline (cur, buffer->rlimit, c); | |
862 | /* Work correctly if there is an asterisk before an | |
863 | arbirtrarily long sequence of escaped newlines. */ | |
864 | if (bslash && (bslash[-1] == '*' || bslash == char_after_star)) | |
865 | char_after_star = cur; | |
866 | else | |
867 | char_after_star = 0; | |
45b966db | 868 | } |
45b966db | 869 | } |
041c3194 ZW |
870 | seen_eof = 1; |
871 | ||
64aaf407 | 872 | out: |
041c3194 ZW |
873 | buffer->cur = cur; |
874 | return seen_eof; | |
45b966db ZW |
875 | } |
876 | ||
041c3194 ZW |
877 | /* Skip a C++ or Chill line comment. Handles escaped newlines. |
878 | Returns non-zero if a multiline comment. */ | |
879 | static int | |
880 | skip_line_comment (pfile) | |
45b966db ZW |
881 | cpp_reader *pfile; |
882 | { | |
041c3194 ZW |
883 | cpp_buffer *buffer = pfile->buffer; |
884 | register const unsigned char *cur = buffer->cur; | |
885 | int multiline = 0; | |
886 | ||
887 | for (; cur < buffer->rlimit; ) | |
888 | { | |
889 | unsigned char c = *cur++; | |
890 | ||
91fcd158 | 891 | if (is_vspace (c)) |
45b966db | 892 | { |
041c3194 ZW |
893 | /* Check for a (trigaph?) backslash escaping the newline. */ |
894 | if (!backslash_start (pfile, cur - 2)) | |
895 | goto out; | |
896 | multiline = 1; | |
897 | handle_newline (cur, buffer->rlimit, c); | |
898 | } | |
899 | } | |
900 | cur++; | |
45b966db | 901 | |
041c3194 ZW |
902 | out: |
903 | buffer->cur = cur - 1; /* Leave newline for caller. */ | |
904 | return multiline; | |
905 | } | |
45b966db | 906 | |
041c3194 ZW |
907 | /* Skips whitespace, stopping at next non-whitespace character. |
908 | Adjusts pfile->col_adjust to account for tabs. This enables tokens | |
909 | to be assigned the correct column. */ | |
910 | static void | |
911 | skip_whitespace (pfile, in_directive) | |
912 | cpp_reader *pfile; | |
913 | int in_directive; | |
914 | { | |
915 | cpp_buffer *buffer = pfile->buffer; | |
91fcd158 | 916 | unsigned short warned = 0; |
45b966db | 917 | |
91fcd158 NB |
918 | /* We only want non-vertical space, i.e. ' ' \t \f \v \0. */ |
919 | while (buffer->cur < buffer->rlimit) | |
041c3194 | 920 | { |
91fcd158 | 921 | unsigned char c = *buffer->cur; |
45b966db | 922 | |
91fcd158 NB |
923 | if (!is_nvspace (c)) |
924 | break; | |
925 | ||
926 | buffer->cur++; | |
927 | /* Horizontal space always OK. */ | |
928 | if (c == ' ') | |
929 | continue; | |
930 | else if (c == '\t') | |
931 | pfile->col_adjust += CPP_OPTION (pfile, tabstop) - 1 | |
932 | - (CPP_BUF_COL (buffer) - 1) % CPP_OPTION(pfile, tabstop); | |
933 | /* Must be \f \v or \0. */ | |
934 | else if (c == '\0') | |
041c3194 | 935 | { |
91fcd158 NB |
936 | if (!warned) |
937 | cpp_warning_with_line (pfile, CPP_BUF_LINE (buffer), | |
938 | CPP_BUF_COL (buffer), | |
939 | "embedded null character ignored"); | |
940 | warned = 1; | |
45b966db | 941 | } |
041c3194 | 942 | else if (in_directive && CPP_PEDANTIC (pfile)) |
91fcd158 NB |
943 | cpp_pedwarn_with_line (pfile, CPP_BUF_LINE (buffer), |
944 | CPP_BUF_COL (buffer), | |
945 | "%s in preprocessing directive", | |
946 | c == '\f' ? "form feed" : "vertical tab"); | |
45b966db | 947 | } |
041c3194 | 948 | } |
45b966db | 949 | |
041c3194 | 950 | /* Parse (append) an identifier. */ |
bfb9dc7f ZW |
951 | static inline const U_CHAR * |
952 | parse_name (pfile, tok, cur, rlimit) | |
45b966db | 953 | cpp_reader *pfile; |
bfb9dc7f ZW |
954 | cpp_token *tok; |
955 | const U_CHAR *cur, *rlimit; | |
45b966db | 956 | { |
bfb9dc7f ZW |
957 | const U_CHAR *name = cur; |
958 | unsigned int len; | |
041c3194 | 959 | |
bfb9dc7f | 960 | while (cur < rlimit) |
041c3194 | 961 | { |
bfb9dc7f ZW |
962 | if (! is_idchar (*cur)) |
963 | break; | |
e5ec2402 ZW |
964 | /* $ is not a legal identifier character in the standard, but is |
965 | commonly accepted as an extension. Don't warn about it in | |
966 | skipped conditional blocks. */ | |
bfb9dc7f | 967 | if (*cur == '$' && CPP_PEDANTIC (pfile) && ! pfile->skipping) |
041c3194 | 968 | { |
bfb9dc7f | 969 | CPP_BUFFER (pfile)->cur = cur; |
041c3194 ZW |
970 | cpp_pedwarn (pfile, "'$' character in identifier"); |
971 | } | |
bfb9dc7f | 972 | cur++; |
041c3194 | 973 | } |
bfb9dc7f | 974 | len = cur - name; |
45b966db | 975 | |
bfb9dc7f | 976 | if (tok->val.node) |
041c3194 | 977 | { |
bfb9dc7f ZW |
978 | unsigned int oldlen = tok->val.node->length; |
979 | U_CHAR *newname = alloca (oldlen + len); | |
980 | memcpy (newname, tok->val.node->name, oldlen); | |
981 | memcpy (newname + oldlen, name, len); | |
982 | len += oldlen; | |
983 | name = newname; | |
041c3194 ZW |
984 | } |
985 | ||
bfb9dc7f ZW |
986 | tok->val.node = cpp_lookup (pfile, name, len); |
987 | return cur; | |
45b966db ZW |
988 | } |
989 | ||
041c3194 | 990 | /* Parse (append) a number. */ |
45b966db | 991 | static void |
041c3194 | 992 | parse_number (pfile, list, name) |
45b966db | 993 | cpp_reader *pfile; |
041c3194 | 994 | cpp_toklist *list; |
bfb9dc7f | 995 | cpp_string *name; |
45b966db | 996 | { |
041c3194 ZW |
997 | const unsigned char *name_limit; |
998 | unsigned char *namebuf; | |
999 | cpp_buffer *buffer = pfile->buffer; | |
1000 | register const unsigned char *cur = buffer->cur; | |
45b966db | 1001 | |
041c3194 ZW |
1002 | expanded: |
1003 | name_limit = list->namebuf + list->name_cap; | |
1004 | namebuf = list->namebuf + list->name_used; | |
45b966db | 1005 | |
041c3194 ZW |
1006 | for (; cur < buffer->rlimit && namebuf < name_limit; ) |
1007 | { | |
1008 | unsigned char c = *namebuf = *cur; /* Copy a single char. */ | |
45b966db | 1009 | |
041c3194 ZW |
1010 | /* Perhaps we should accept '$' here if we accept it for |
1011 | identifiers. We know namebuf[-1] is safe, because for c to | |
1012 | be a sign we must have pushed at least one character. */ | |
1013 | if (!is_numchar (c) && c != '.' && ! VALID_SIGN (c, namebuf[-1])) | |
1014 | goto out; | |
45b966db | 1015 | |
041c3194 ZW |
1016 | namebuf++; |
1017 | cur++; | |
45b966db | 1018 | } |
64aaf407 | 1019 | |
041c3194 ZW |
1020 | /* Run out of name space? */ |
1021 | if (cur < buffer->rlimit) | |
1022 | { | |
1023 | list->name_used = namebuf - list->namebuf; | |
1024 | auto_expand_name_space (list); | |
1025 | goto expanded; | |
1026 | } | |
1027 | ||
64aaf407 | 1028 | out: |
041c3194 ZW |
1029 | buffer->cur = cur; |
1030 | name->len = namebuf - name->text; | |
1031 | list->name_used = namebuf - list->namebuf; | |
45b966db ZW |
1032 | } |
1033 | ||
041c3194 | 1034 | /* Places a string terminated by an unescaped TERMINATOR into a |
bfb9dc7f | 1035 | cpp_string, which should be expandable and thus at the top of the |
041c3194 ZW |
1036 | list's stack. Handles embedded trigraphs, if necessary, and |
1037 | escaped newlines. | |
45b966db | 1038 | |
041c3194 ZW |
1039 | Can be used for character constants (terminator = '\''), string |
1040 | constants ('"') and angled headers ('>'). Multi-line strings are | |
1041 | allowed, except for within directives. */ | |
45b966db | 1042 | |
041c3194 ZW |
1043 | static void |
1044 | parse_string (pfile, list, token, terminator) | |
45b966db | 1045 | cpp_reader *pfile; |
041c3194 ZW |
1046 | cpp_toklist *list; |
1047 | cpp_token *token; | |
1048 | unsigned int terminator; | |
45b966db | 1049 | { |
041c3194 | 1050 | cpp_buffer *buffer = pfile->buffer; |
bfb9dc7f | 1051 | cpp_string *name = &token->val.str; |
041c3194 ZW |
1052 | register const unsigned char *cur = buffer->cur; |
1053 | const unsigned char *name_limit; | |
1054 | unsigned char *namebuf; | |
1055 | unsigned int null_count = 0; | |
1056 | unsigned int trigraphed = list->name_used; | |
45b966db | 1057 | |
041c3194 ZW |
1058 | expanded: |
1059 | name_limit = list->namebuf + list->name_cap; | |
1060 | namebuf = list->namebuf + list->name_used; | |
f2d5f0cc | 1061 | |
041c3194 | 1062 | for (; cur < buffer->rlimit && namebuf < name_limit; ) |
45b966db | 1063 | { |
041c3194 | 1064 | unsigned int c = *namebuf++ = *cur++; /* Copy a single char. */ |
45b966db | 1065 | |
041c3194 ZW |
1066 | if (c == '\0') |
1067 | null_count++; | |
91fcd158 | 1068 | else if (c == terminator || is_vspace (c)) |
45b966db | 1069 | { |
041c3194 ZW |
1070 | /* Needed for trigraph_replace and multiline string warning. */ |
1071 | buffer->cur = cur; | |
5eec0563 | 1072 | |
041c3194 ZW |
1073 | /* Scan for trigraphs before checking if backslash-escaped. */ |
1074 | if ((CPP_OPTION (pfile, trigraphs) | |
1075 | || CPP_OPTION (pfile, warn_trigraphs)) | |
1076 | && namebuf - (list->namebuf + trigraphed) >= 3) | |
45b966db | 1077 | { |
041c3194 ZW |
1078 | namebuf = trigraph_replace (pfile, list->namebuf + trigraphed, |
1079 | namebuf); | |
1080 | /* The test above guarantees trigraphed will be positive. */ | |
1081 | trigraphed = namebuf - list->namebuf - 2; | |
45b966db | 1082 | } |
45b966db | 1083 | |
041c3194 | 1084 | namebuf--; /* Drop the newline / terminator from the name. */ |
91fcd158 | 1085 | if (is_vspace (c)) |
45b966db | 1086 | { |
041c3194 ZW |
1087 | /* Drop a backslash newline, and continue. */ |
1088 | if (namebuf[-1] == '\\') | |
45b966db | 1089 | { |
041c3194 ZW |
1090 | handle_newline (cur, buffer->rlimit, c); |
1091 | namebuf--; | |
1092 | continue; | |
45b966db | 1093 | } |
45b966db | 1094 | |
041c3194 | 1095 | cur--; |
45b966db | 1096 | |
041c3194 ZW |
1097 | /* In Fortran and assembly language, silently terminate |
1098 | strings of either variety at end of line. This is a | |
1099 | kludge around not knowing where comments are in these | |
1100 | languages. */ | |
1101 | if (CPP_OPTION (pfile, lang_fortran) | |
1102 | || CPP_OPTION (pfile, lang_asm)) | |
1103 | goto out; | |
64aaf407 | 1104 | |
041c3194 ZW |
1105 | /* Character constants, headers and asserts may not |
1106 | extend over multiple lines. In Standard C, neither | |
1107 | may strings. We accept multiline strings as an | |
1108 | extension. (Even in directives - otherwise, glibc's | |
1109 | longlong.h breaks.) */ | |
1110 | if (terminator != '"') | |
1111 | goto unterminated; | |
1112 | ||
1113 | cur++; /* Move forwards again. */ | |
45b966db | 1114 | |
041c3194 ZW |
1115 | if (pfile->multiline_string_line == 0) |
1116 | { | |
1117 | pfile->multiline_string_line = token->line; | |
1118 | pfile->multiline_string_column = token->col; | |
1119 | if (CPP_PEDANTIC (pfile)) | |
1120 | cpp_pedwarn (pfile, "multi-line string constant"); | |
1121 | } | |
1122 | ||
1123 | *namebuf++ = '\n'; | |
1124 | handle_newline (cur, buffer->rlimit, c); | |
45b966db ZW |
1125 | } |
1126 | else | |
1127 | { | |
041c3194 | 1128 | unsigned char *temp; |
45b966db | 1129 | |
041c3194 ZW |
1130 | /* An odd number of consecutive backslashes represents |
1131 | an escaped terminator. */ | |
1132 | temp = namebuf - 1; | |
1133 | while (temp >= name->text && *temp == '\\') | |
1134 | temp--; | |
45b966db | 1135 | |
041c3194 ZW |
1136 | if ((namebuf - temp) & 1) |
1137 | goto out; | |
1138 | namebuf++; | |
1139 | } | |
45b966db | 1140 | } |
ff2b53ef | 1141 | } |
45b966db | 1142 | |
041c3194 ZW |
1143 | /* Run out of name space? */ |
1144 | if (cur < buffer->rlimit) | |
45b966db | 1145 | { |
041c3194 ZW |
1146 | list->name_used = namebuf - list->namebuf; |
1147 | auto_expand_name_space (list); | |
1148 | goto expanded; | |
1149 | } | |
45b966db | 1150 | |
041c3194 ZW |
1151 | /* We may not have trigraph-replaced the input for this code path, |
1152 | but as the input is in error by being unterminated we don't | |
1153 | bother. Prevent warnings about no newlines at EOF. */ | |
91fcd158 | 1154 | if (is_vspace (cur[-1])) |
041c3194 | 1155 | cur--; |
45b966db | 1156 | |
041c3194 ZW |
1157 | unterminated: |
1158 | cpp_error (pfile, "missing terminating %c character", (int) terminator); | |
476f2869 | 1159 | |
041c3194 ZW |
1160 | if (terminator == '\"' && pfile->multiline_string_line != list->line |
1161 | && pfile->multiline_string_line != 0) | |
1162 | { | |
1163 | cpp_error_with_line (pfile, pfile->multiline_string_line, | |
1164 | pfile->multiline_string_column, | |
1165 | "possible start of unterminated string literal"); | |
1166 | pfile->multiline_string_line = 0; | |
45b966db | 1167 | } |
041c3194 ZW |
1168 | |
1169 | out: | |
1170 | buffer->cur = cur; | |
1171 | name->len = namebuf - name->text; | |
1172 | list->name_used = namebuf - list->namebuf; | |
45b966db | 1173 | |
041c3194 ZW |
1174 | if (null_count > 0) |
1175 | cpp_warning (pfile, (null_count > 1 ? "null characters preserved" | |
1176 | : "null character preserved")); | |
45b966db ZW |
1177 | } |
1178 | ||
041c3194 ZW |
1179 | /* The character TYPE helps us distinguish comment types: '*' = C |
1180 | style, '-' = Chill-style and '/' = C++ style. For code simplicity, | |
1181 | the stored comment includes the comment start and any terminator. */ | |
1182 | ||
1183 | #define COMMENT_START_LEN 2 | |
9e62c811 | 1184 | static void |
041c3194 ZW |
1185 | save_comment (list, token, from, len, type) |
1186 | cpp_toklist *list; | |
1187 | cpp_token *token; | |
1188 | const unsigned char *from; | |
9e62c811 | 1189 | unsigned int len; |
041c3194 | 1190 | unsigned int type; |
9e62c811 | 1191 | { |
041c3194 ZW |
1192 | unsigned char *buffer; |
1193 | ||
1194 | len += COMMENT_START_LEN; | |
9e62c811 | 1195 | |
041c3194 ZW |
1196 | if (list->name_used + len > list->name_cap) |
1197 | _cpp_expand_name_space (list, len); | |
9e62c811 | 1198 | |
bfb9dc7f | 1199 | INIT_TOKEN_STR (list, token); |
041c3194 | 1200 | token->type = CPP_COMMENT; |
bfb9dc7f | 1201 | token->val.str.len = len; |
45b966db | 1202 | |
041c3194 ZW |
1203 | buffer = list->namebuf + list->name_used; |
1204 | list->name_used += len; | |
45b966db | 1205 | |
041c3194 ZW |
1206 | /* Copy the comment. */ |
1207 | if (type == '*') | |
45b966db | 1208 | { |
041c3194 ZW |
1209 | *buffer++ = '/'; |
1210 | *buffer++ = '*'; | |
45b966db | 1211 | } |
041c3194 | 1212 | else |
ea4a453b | 1213 | { |
041c3194 ZW |
1214 | *buffer++ = type; |
1215 | *buffer++ = type; | |
ea4a453b | 1216 | } |
041c3194 | 1217 | memcpy (buffer, from, len - COMMENT_START_LEN); |
45b966db ZW |
1218 | } |
1219 | ||
041c3194 ZW |
1220 | /* |
1221 | * The tokenizer's main loop. Returns a token list, representing a | |
1222 | * logical line in the input file. On EOF after some tokens have | |
1223 | * been processed, we return immediately. Then in next call, or if | |
1224 | * EOF occurred at the beginning of a logical line, a single CPP_EOF | |
1225 | * token is placed in the list. | |
1226 | * | |
1227 | * Implementation relies almost entirely on lookback, rather than | |
1228 | * looking forwards. This means that tokenization requires just | |
1229 | * a single pass of the file, even in the presence of trigraphs and | |
1230 | * escaped newlines, providing significant performance benefits. | |
1231 | * Trigraph overhead is negligible if they are disabled, and low | |
1232 | * even when enabled. | |
1233 | */ | |
1234 | ||
91fcd158 | 1235 | #define KNOWN_DIRECTIVE() (list->directive != 0) |
041c3194 ZW |
1236 | #define MIGHT_BE_DIRECTIVE() \ |
1237 | (cur_token == &list->tokens[first_token + 1] && cur_token[-1].type == CPP_HASH) | |
45b966db | 1238 | |
041c3194 ZW |
1239 | static void |
1240 | lex_line (pfile, list) | |
45b966db | 1241 | cpp_reader *pfile; |
041c3194 | 1242 | cpp_toklist *list; |
45b966db | 1243 | { |
041c3194 ZW |
1244 | cpp_token *cur_token, *token_limit, *first; |
1245 | cpp_buffer *buffer = pfile->buffer; | |
1246 | const unsigned char *cur = buffer->cur; | |
1247 | unsigned char flags = 0; | |
1248 | unsigned int first_token = list->tokens_used; | |
45b966db | 1249 | |
041c3194 ZW |
1250 | if (!(list->flags & LIST_OFFSET)) |
1251 | (abort) (); | |
1252 | ||
1253 | list->file = buffer->nominal_fname; | |
1254 | list->line = CPP_BUF_LINE (buffer); | |
1255 | pfile->col_adjust = 0; | |
1256 | pfile->in_lex_line = 1; | |
1257 | if (cur == buffer->buf) | |
1258 | list->flags |= BEG_OF_FILE; | |
1259 | ||
1260 | expanded: | |
1261 | token_limit = list->tokens + list->tokens_cap; | |
1262 | cur_token = list->tokens + list->tokens_used; | |
45b966db | 1263 | |
041c3194 | 1264 | for (; cur < buffer->rlimit && cur_token < token_limit;) |
45b966db | 1265 | { |
041c3194 | 1266 | unsigned char c; |
45b966db | 1267 | |
91fcd158 NB |
1268 | /* Optimize non-vertical whitespace skipping; most tokens are |
1269 | probably separated by whitespace. (' ' '\t' '\v' '\f' '\0'). */ | |
1270 | c = *cur; | |
1271 | if (is_nvspace (c)) | |
45b966db | 1272 | { |
91fcd158 NB |
1273 | buffer->cur = cur; |
1274 | skip_whitespace (pfile, (list->tokens[first_token].type == CPP_HASH | |
1275 | && cur_token > &list->tokens[first_token])); | |
041c3194 | 1276 | cur = buffer->cur; |
ff2b53ef | 1277 | |
041c3194 ZW |
1278 | flags = PREV_WHITE; |
1279 | if (cur == buffer->rlimit) | |
1280 | break; | |
91fcd158 | 1281 | c = *cur; |
45b966db | 1282 | } |
91fcd158 | 1283 | cur++; |
45b966db | 1284 | |
041c3194 ZW |
1285 | /* Initialize current token. CPP_EOF will not be fixed up by |
1286 | expand_name_space. */ | |
1287 | list->tokens_used = cur_token - list->tokens + 1; | |
1288 | cur_token->type = CPP_EOF; | |
1289 | cur_token->col = CPP_BUF_COLUMN (buffer, cur); | |
1290 | cur_token->line = CPP_BUF_LINE (buffer); | |
1291 | cur_token->flags = flags; | |
1292 | flags = 0; | |
46d07497 | 1293 | |
041c3194 ZW |
1294 | switch (c) |
1295 | { | |
1296 | case '0': case '1': case '2': case '3': case '4': | |
1297 | case '5': case '6': case '7': case '8': case '9': | |
1298 | { | |
1299 | int prev_dot; | |
46d07497 | 1300 | |
041c3194 ZW |
1301 | cur--; /* Backup character. */ |
1302 | prev_dot = PREV_TOKEN_TYPE == CPP_DOT && IMMED_TOKEN (); | |
1303 | if (prev_dot) | |
1304 | cur_token--; | |
bfb9dc7f | 1305 | INIT_TOKEN_STR (list, cur_token); |
041c3194 ZW |
1306 | /* Prepend an immediately previous CPP_DOT token. */ |
1307 | if (prev_dot) | |
1308 | { | |
1309 | if (list->name_cap == list->name_used) | |
1310 | auto_expand_name_space (list); | |
46d07497 | 1311 | |
bfb9dc7f | 1312 | cur_token->val.str.len = 1; |
041c3194 ZW |
1313 | list->namebuf[list->name_used++] = '.'; |
1314 | } | |
46d07497 | 1315 | |
041c3194 ZW |
1316 | continue_number: |
1317 | cur_token->type = CPP_NUMBER; /* Before parse_number. */ | |
1318 | buffer->cur = cur; | |
bfb9dc7f | 1319 | parse_number (pfile, list, &cur_token->val.str); |
041c3194 ZW |
1320 | cur = buffer->cur; |
1321 | } | |
1322 | /* Check for # 123 form of #line. */ | |
1323 | if (MIGHT_BE_DIRECTIVE ()) | |
1324 | list->directive = _cpp_check_linemarker (pfile, cur_token, | |
1325 | !(cur_token[-1].flags | |
1326 | & PREV_WHITE)); | |
1327 | cur_token++; | |
1328 | break; | |
46d07497 | 1329 | |
041c3194 ZW |
1330 | letter: |
1331 | case '_': | |
1332 | case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': | |
1333 | case 'g': case 'h': case 'i': case 'j': case 'k': case 'l': | |
1334 | case 'm': case 'n': case 'o': case 'p': case 'q': case 'r': | |
1335 | case 's': case 't': case 'u': case 'v': case 'w': case 'x': | |
1336 | case 'y': case 'z': | |
1337 | case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': | |
1338 | case 'G': case 'H': case 'I': case 'J': case 'K': case 'L': | |
1339 | case 'M': case 'N': case 'O': case 'P': case 'Q': case 'R': | |
1340 | case 'S': case 'T': case 'U': case 'V': case 'W': case 'X': | |
1341 | case 'Y': case 'Z': | |
1342 | cur--; /* Backup character. */ | |
bfb9dc7f | 1343 | cur_token->val.node = 0; |
041c3194 ZW |
1344 | cur_token->type = CPP_NAME; /* Identifier, macro etc. */ |
1345 | ||
1346 | continue_name: | |
bfb9dc7f | 1347 | cur = parse_name (pfile, cur_token, cur, buffer->rlimit); |
45b966db | 1348 | |
041c3194 ZW |
1349 | if (MIGHT_BE_DIRECTIVE ()) |
1350 | list->directive = _cpp_check_directive (pfile, cur_token, | |
1351 | !(list->tokens[0].flags | |
1352 | & PREV_WHITE)); | |
1353 | cur_token++; | |
1354 | break; | |
f8f769ea | 1355 | |
041c3194 ZW |
1356 | case '\'': |
1357 | /* Character constants are not recognized when processing Fortran, | |
1358 | or if -traditional. */ | |
1359 | if (CPP_OPTION (pfile, lang_fortran) || CPP_TRADITIONAL (pfile)) | |
1360 | goto other; | |
45b966db | 1361 | |
041c3194 ZW |
1362 | /* Fall through. */ |
1363 | case '\"': | |
1364 | /* Traditionally, escaped strings are not strings. */ | |
1365 | if (CPP_TRADITIONAL (pfile) && IMMED_TOKEN () | |
1366 | && PREV_TOKEN_TYPE == CPP_BACKSLASH) | |
1367 | goto other; | |
04e3ec78 | 1368 | |
041c3194 ZW |
1369 | cur_token->type = c == '\'' ? CPP_CHAR : CPP_STRING; |
1370 | /* Do we have a wide string? */ | |
1371 | if (cur_token[-1].type == CPP_NAME && IMMED_TOKEN () | |
bfb9dc7f | 1372 | && cur_token[-1].val.node == pfile->spec_nodes->n_L |
041c3194 | 1373 | && !CPP_TRADITIONAL (pfile)) |
04e3ec78 | 1374 | { |
041c3194 | 1375 | (--cur_token)->type = (c == '\'' ? CPP_WCHAR : CPP_WSTRING); |
04e3ec78 | 1376 | } |
45b966db | 1377 | |
041c3194 ZW |
1378 | do_parse_string: |
1379 | /* Here c is one of ' " or >. */ | |
bfb9dc7f | 1380 | INIT_TOKEN_STR (list, cur_token); |
041c3194 ZW |
1381 | buffer->cur = cur; |
1382 | parse_string (pfile, list, cur_token, c); | |
1383 | cur = buffer->cur; | |
1384 | cur_token++; | |
1385 | break; | |
45b966db | 1386 | |
041c3194 ZW |
1387 | case '/': |
1388 | cur_token->type = CPP_DIV; | |
1389 | if (IMMED_TOKEN ()) | |
1390 | { | |
1391 | if (PREV_TOKEN_TYPE == CPP_DIV) | |
1392 | { | |
1393 | /* We silently allow C++ comments in system headers, | |
1394 | irrespective of conformance mode, because lots of | |
1395 | broken systems do that and trying to clean it up | |
1396 | in fixincludes is a nightmare. */ | |
1397 | if (CPP_IN_SYSTEM_HEADER (pfile)) | |
1398 | goto do_line_comment; | |
1399 | else if (CPP_OPTION (pfile, cplusplus_comments)) | |
1400 | { | |
1401 | if (CPP_OPTION (pfile, c89) && CPP_PEDANTIC (pfile) | |
1402 | && ! buffer->warned_cplusplus_comments) | |
1403 | { | |
1404 | buffer->cur = cur; | |
1405 | cpp_pedwarn (pfile, | |
1406 | "C++ style comments are not allowed in ISO C89"); | |
1407 | cpp_pedwarn (pfile, | |
1408 | "(this will be reported only once per input file)"); | |
1409 | buffer->warned_cplusplus_comments = 1; | |
1410 | } | |
1411 | do_line_comment: | |
1412 | buffer->cur = cur; | |
1413 | #if 0 /* Leave until new lexer in place. */ | |
1414 | if (cur[-2] != c) | |
1415 | cpp_warning (pfile, | |
1416 | "comment start split across lines"); | |
1417 | #endif | |
1418 | if (skip_line_comment (pfile)) | |
1419 | cpp_warning (pfile, "multi-line comment"); | |
f8f769ea | 1420 | |
041c3194 ZW |
1421 | /* Back-up to first '-' or '/'. */ |
1422 | cur_token--; | |
1423 | if (!CPP_OPTION (pfile, discard_comments) | |
91fcd158 | 1424 | && (!KNOWN_DIRECTIVE() |
041c3194 ZW |
1425 | || (list->directive->flags & COMMENTS))) |
1426 | save_comment (list, cur_token++, cur, | |
1427 | buffer->cur - cur, c); | |
1428 | else if (!CPP_OPTION (pfile, traditional)) | |
1429 | flags = PREV_WHITE; | |
1430 | ||
1431 | cur = buffer->cur; | |
1432 | break; | |
1433 | } | |
1434 | } | |
1435 | } | |
1436 | cur_token++; | |
1437 | break; | |
1438 | ||
1439 | case '*': | |
1440 | cur_token->type = CPP_MULT; | |
1441 | if (IMMED_TOKEN ()) | |
45b966db | 1442 | { |
041c3194 ZW |
1443 | if (PREV_TOKEN_TYPE == CPP_DIV) |
1444 | { | |
1445 | buffer->cur = cur; | |
1446 | #if 0 /* Leave until new lexer in place. */ | |
1447 | if (cur[-2] != '/') | |
1448 | cpp_warning (pfile, | |
1449 | "comment start '/*' split across lines"); | |
1450 | #endif | |
1451 | if (skip_block_comment (pfile)) | |
1452 | cpp_error_with_line (pfile, list->line, cur_token[-1].col, | |
1453 | "unterminated comment"); | |
1454 | #if 0 /* Leave until new lexer in place. */ | |
1455 | else if (buffer->cur[-2] != '*') | |
1456 | cpp_warning (pfile, | |
1457 | "comment end '*/' split across lines"); | |
1458 | #endif | |
1459 | /* Back up to opening '/'. */ | |
1460 | cur_token--; | |
1461 | if (!CPP_OPTION (pfile, discard_comments) | |
91fcd158 | 1462 | && (!KNOWN_DIRECTIVE() |
041c3194 ZW |
1463 | || (list->directive->flags & COMMENTS))) |
1464 | save_comment (list, cur_token++, cur, | |
1465 | buffer->cur - cur, c); | |
1466 | else if (!CPP_OPTION (pfile, traditional)) | |
1467 | flags = PREV_WHITE; | |
1468 | ||
1469 | cur = buffer->cur; | |
1470 | break; | |
1471 | } | |
1472 | else if (CPP_OPTION (pfile, cplusplus)) | |
1473 | { | |
1474 | /* In C++, there are .* and ->* operators. */ | |
1475 | if (PREV_TOKEN_TYPE == CPP_DEREF) | |
1476 | BACKUP_TOKEN (CPP_DEREF_STAR); | |
1477 | else if (PREV_TOKEN_TYPE == CPP_DOT) | |
1478 | BACKUP_TOKEN (CPP_DOT_STAR); | |
1479 | } | |
45b966db | 1480 | } |
041c3194 | 1481 | cur_token++; |
f8f769ea | 1482 | break; |
45b966db | 1483 | |
041c3194 ZW |
1484 | case '\n': |
1485 | case '\r': | |
1486 | handle_newline (cur, buffer->rlimit, c); | |
1487 | if (PREV_TOKEN_TYPE == CPP_BACKSLASH) | |
45b966db | 1488 | { |
041c3194 | 1489 | if (IMMED_TOKEN ()) |
45b966db | 1490 | { |
041c3194 ZW |
1491 | /* Remove the escaped newline. Then continue to process |
1492 | any interrupted name or number. */ | |
1493 | cur_token--; | |
1494 | /* Backslash-newline may not be immediately followed by | |
1495 | EOF (C99 5.1.1.2). */ | |
1496 | if (cur >= buffer->rlimit) | |
1497 | { | |
1498 | cpp_pedwarn (pfile, "backslash-newline at end of file"); | |
1499 | break; | |
1500 | } | |
1501 | if (IMMED_TOKEN ()) | |
1502 | { | |
1503 | cur_token--; | |
1504 | if (cur_token->type == CPP_NAME) | |
1505 | goto continue_name; | |
1506 | else if (cur_token->type == CPP_NUMBER) | |
1507 | goto continue_number; | |
1508 | cur_token++; | |
1509 | } | |
1510 | /* Remember whitespace setting. */ | |
1511 | flags = cur_token->flags; | |
f8f769ea | 1512 | break; |
45b966db | 1513 | } |
041c3194 ZW |
1514 | else |
1515 | { | |
1516 | buffer->cur = cur; | |
1517 | cpp_warning (pfile, | |
1518 | "backslash and newline separated by space"); | |
1519 | } | |
1520 | } | |
1521 | else if (MIGHT_BE_DIRECTIVE ()) | |
1522 | { | |
1523 | /* "Null directive." C99 6.10.7: A preprocessing | |
1524 | directive of the form # <new-line> has no effect. | |
1525 | ||
1526 | But it is still a directive, and therefore disappears | |
1527 | from the output. */ | |
1528 | cur_token--; | |
1529 | if (cur_token->flags & PREV_WHITE) | |
45b966db | 1530 | { |
041c3194 ZW |
1531 | if (CPP_WTRADITIONAL (pfile)) |
1532 | cpp_warning (pfile, | |
1533 | "K+R C ignores #\\n with the # indented"); | |
1534 | if (CPP_TRADITIONAL (pfile)) | |
1535 | cur_token++; | |
45b966db | 1536 | } |
f8f769ea | 1537 | } |
45b966db | 1538 | |
041c3194 ZW |
1539 | /* Skip vertical space until we have at least one token to |
1540 | return. */ | |
1541 | if (cur_token != &list->tokens[first_token]) | |
1542 | goto out; | |
1543 | list->line = CPP_BUF_LINE (buffer); | |
f8f769ea | 1544 | break; |
04e3ec78 | 1545 | |
041c3194 ZW |
1546 | case '-': |
1547 | if (IMMED_TOKEN () && PREV_TOKEN_TYPE == CPP_MINUS) | |
1548 | { | |
1549 | if (CPP_OPTION (pfile, chill)) | |
1550 | goto do_line_comment; | |
1551 | REVISE_TOKEN (CPP_MINUS_MINUS); | |
1552 | } | |
1553 | else | |
1554 | PUSH_TOKEN (CPP_MINUS); | |
1555 | break; | |
45b966db | 1556 | |
041c3194 ZW |
1557 | make_hash: |
1558 | case '#': | |
1559 | /* The digraph flag checking ensures that ## and %:%: | |
1560 | are interpreted as CPP_PASTE, but #%: and %:# are not. */ | |
1561 | if (PREV_TOKEN_TYPE == CPP_HASH && IMMED_TOKEN () | |
1562 | && ((cur_token->flags ^ cur_token[-1].flags) & DIGRAPH) == 0) | |
1563 | REVISE_TOKEN (CPP_PASTE); | |
1564 | else | |
1565 | PUSH_TOKEN (CPP_HASH); | |
1566 | break; | |
04e3ec78 | 1567 | |
041c3194 ZW |
1568 | case ':': |
1569 | cur_token->type = CPP_COLON; | |
1570 | if (IMMED_TOKEN ()) | |
1571 | { | |
1572 | if (PREV_TOKEN_TYPE == CPP_COLON | |
1573 | && CPP_OPTION (pfile, cplusplus)) | |
1574 | BACKUP_TOKEN (CPP_SCOPE); | |
1575 | /* Digraph: "<:" is a '[' */ | |
1576 | else if (PREV_TOKEN_TYPE == CPP_LESS) | |
1577 | BACKUP_DIGRAPH (CPP_OPEN_SQUARE); | |
1578 | /* Digraph: "%:" is a '#' */ | |
1579 | else if (PREV_TOKEN_TYPE == CPP_MOD) | |
1580 | { | |
1581 | (--cur_token)->flags |= DIGRAPH; | |
1582 | goto make_hash; | |
1583 | } | |
1584 | } | |
1585 | cur_token++; | |
1586 | break; | |
f8f769ea | 1587 | |
041c3194 ZW |
1588 | case '&': |
1589 | if (IMMED_TOKEN () && PREV_TOKEN_TYPE == CPP_AND) | |
1590 | REVISE_TOKEN (CPP_AND_AND); | |
1591 | else | |
1592 | PUSH_TOKEN (CPP_AND); | |
f8f769ea | 1593 | break; |
45b966db | 1594 | |
041c3194 ZW |
1595 | make_or: |
1596 | case '|': | |
1597 | if (IMMED_TOKEN () && PREV_TOKEN_TYPE == CPP_OR) | |
1598 | REVISE_TOKEN (CPP_OR_OR); | |
1599 | else | |
1600 | PUSH_TOKEN (CPP_OR); | |
1601 | break; | |
45b966db | 1602 | |
041c3194 ZW |
1603 | case '+': |
1604 | if (IMMED_TOKEN () && PREV_TOKEN_TYPE == CPP_PLUS) | |
1605 | REVISE_TOKEN (CPP_PLUS_PLUS); | |
1606 | else | |
1607 | PUSH_TOKEN (CPP_PLUS); | |
1608 | break; | |
45b966db | 1609 | |
041c3194 ZW |
1610 | case '=': |
1611 | /* This relies on equidistance of "?=" and "?" tokens. */ | |
1612 | if (IMMED_TOKEN () && PREV_TOKEN_TYPE <= CPP_LAST_EQ) | |
1613 | REVISE_TOKEN (PREV_TOKEN_TYPE + (CPP_EQ_EQ - CPP_EQ)); | |
1614 | else | |
1615 | PUSH_TOKEN (CPP_EQ); | |
1616 | break; | |
45b966db | 1617 | |
041c3194 ZW |
1618 | case '>': |
1619 | cur_token->type = CPP_GREATER; | |
1620 | if (IMMED_TOKEN ()) | |
1621 | { | |
1622 | if (PREV_TOKEN_TYPE == CPP_GREATER) | |
1623 | BACKUP_TOKEN (CPP_RSHIFT); | |
1624 | else if (PREV_TOKEN_TYPE == CPP_MINUS) | |
1625 | BACKUP_TOKEN (CPP_DEREF); | |
1626 | /* Digraph: ":>" is a ']' */ | |
1627 | else if (PREV_TOKEN_TYPE == CPP_COLON) | |
1628 | BACKUP_DIGRAPH (CPP_CLOSE_SQUARE); | |
1629 | /* Digraph: "%>" is a '}' */ | |
1630 | else if (PREV_TOKEN_TYPE == CPP_MOD) | |
1631 | BACKUP_DIGRAPH (CPP_CLOSE_BRACE); | |
1632 | } | |
1633 | cur_token++; | |
1634 | break; | |
1635 | ||
1636 | case '<': | |
1637 | if (IMMED_TOKEN () && PREV_TOKEN_TYPE == CPP_LESS) | |
1638 | { | |
1639 | REVISE_TOKEN (CPP_LSHIFT); | |
1640 | break; | |
1641 | } | |
1642 | /* Is this the beginning of a header name? */ | |
91fcd158 | 1643 | if (KNOWN_DIRECTIVE () && (list->directive->flags & INCL)) |
041c3194 ZW |
1644 | { |
1645 | c = '>'; /* Terminator. */ | |
1646 | cur_token->type = CPP_HEADER_NAME; | |
1647 | goto do_parse_string; | |
1648 | } | |
1649 | PUSH_TOKEN (CPP_LESS); | |
1650 | break; | |
45b966db | 1651 | |
041c3194 ZW |
1652 | case '%': |
1653 | /* Digraph: "<%" is a '{' */ | |
1654 | cur_token->type = CPP_MOD; | |
1655 | if (IMMED_TOKEN () && PREV_TOKEN_TYPE == CPP_LESS) | |
1656 | BACKUP_DIGRAPH (CPP_OPEN_BRACE); | |
1657 | cur_token++; | |
1658 | break; | |
04e3ec78 | 1659 | |
041c3194 ZW |
1660 | case '?': |
1661 | if (cur + 1 < buffer->rlimit && *cur == '?' | |
1662 | && trigraph_map[cur[1]] && trigraph_ok (pfile, cur + 1)) | |
1663 | { | |
1664 | /* Handle trigraph. */ | |
1665 | cur++; | |
1666 | switch (*cur++) | |
1667 | { | |
1668 | case '(': goto make_open_square; | |
1669 | case ')': goto make_close_square; | |
1670 | case '<': goto make_open_brace; | |
1671 | case '>': goto make_close_brace; | |
1672 | case '=': goto make_hash; | |
1673 | case '!': goto make_or; | |
1674 | case '-': goto make_complement; | |
1675 | case '/': goto make_backslash; | |
1676 | case '\'': goto make_xor; | |
1677 | } | |
1678 | } | |
1679 | if (IMMED_TOKEN () && CPP_OPTION (pfile, cplusplus)) | |
1680 | { | |
1681 | /* GNU C++ defines <? and >? operators. */ | |
1682 | if (PREV_TOKEN_TYPE == CPP_LESS) | |
1683 | { | |
1684 | REVISE_TOKEN (CPP_MIN); | |
1685 | break; | |
1686 | } | |
1687 | else if (PREV_TOKEN_TYPE == CPP_GREATER) | |
1688 | { | |
1689 | REVISE_TOKEN (CPP_MAX); | |
1690 | break; | |
1691 | } | |
1692 | } | |
1693 | PUSH_TOKEN (CPP_QUERY); | |
1694 | break; | |
45b966db | 1695 | |
041c3194 ZW |
1696 | case '.': |
1697 | if (PREV_TOKEN_TYPE == CPP_DOT && cur_token[-2].type == CPP_DOT | |
1698 | && IMMED_TOKEN () | |
1699 | && !(cur_token[-1].flags & PREV_WHITE)) | |
1700 | { | |
1701 | cur_token -= 2; | |
1702 | PUSH_TOKEN (CPP_ELLIPSIS); | |
1703 | } | |
1704 | else | |
1705 | PUSH_TOKEN (CPP_DOT); | |
1706 | break; | |
c5a04734 | 1707 | |
041c3194 ZW |
1708 | make_complement: |
1709 | case '~': PUSH_TOKEN (CPP_COMPL); break; | |
1710 | make_xor: | |
1711 | case '^': PUSH_TOKEN (CPP_XOR); break; | |
1712 | make_open_brace: | |
1713 | case '{': PUSH_TOKEN (CPP_OPEN_BRACE); break; | |
1714 | make_close_brace: | |
1715 | case '}': PUSH_TOKEN (CPP_CLOSE_BRACE); break; | |
1716 | make_open_square: | |
1717 | case '[': PUSH_TOKEN (CPP_OPEN_SQUARE); break; | |
1718 | make_close_square: | |
1719 | case ']': PUSH_TOKEN (CPP_CLOSE_SQUARE); break; | |
1720 | make_backslash: | |
1721 | case '\\': PUSH_TOKEN (CPP_BACKSLASH); break; | |
1722 | case '!': PUSH_TOKEN (CPP_NOT); break; | |
1723 | case ',': PUSH_TOKEN (CPP_COMMA); break; | |
1724 | case ';': PUSH_TOKEN (CPP_SEMICOLON); break; | |
1725 | case '(': PUSH_TOKEN (CPP_OPEN_PAREN); break; | |
1726 | case ')': PUSH_TOKEN (CPP_CLOSE_PAREN); break; | |
6d2c2047 | 1727 | |
041c3194 ZW |
1728 | case '$': |
1729 | if (CPP_OPTION (pfile, dollars_in_ident)) | |
1730 | goto letter; | |
1731 | /* Fall through */ | |
1732 | other: | |
1733 | default: | |
1734 | cur_token->val.aux = c; | |
1735 | PUSH_TOKEN (CPP_OTHER); | |
1736 | break; | |
1737 | } | |
1738 | } | |
6d2c2047 | 1739 | |
041c3194 ZW |
1740 | /* Run out of token space? */ |
1741 | if (cur_token == token_limit) | |
1742 | { | |
1743 | list->tokens_used = cur_token - list->tokens; | |
1744 | _cpp_expand_token_space (list, 256); | |
1745 | goto expanded; | |
1746 | } | |
6d2c2047 | 1747 | |
041c3194 ZW |
1748 | cur_token->flags = flags; |
1749 | if (cur_token == &list->tokens[first_token] && pfile->done_initializing) | |
1750 | { | |
91fcd158 | 1751 | if (cur > buffer->buf && !is_vspace (cur[-1])) |
041c3194 ZW |
1752 | cpp_pedwarn_with_line (pfile, CPP_BUF_LINE (buffer), |
1753 | CPP_BUF_COLUMN (buffer, cur), | |
1754 | "no newline at end of file"); | |
1755 | cur_token++->type = CPP_EOF; | |
1756 | } | |
6d2c2047 | 1757 | |
041c3194 ZW |
1758 | out: |
1759 | /* All tokens are allocated, so the memory location is fixed. */ | |
1760 | first = &list->tokens[first_token]; | |
1761 | ||
1762 | /* Don't complain about the null directive, nor directives in | |
1763 | assembly source: we don't know where the comments are, and # may | |
1764 | introduce assembler pseudo-ops. Don't complain about invalid | |
1765 | directives in skipped conditional groups (6.10 p4). */ | |
1766 | if (first->type == CPP_HASH && list->directive == 0 && !pfile->skipping | |
1767 | && cur_token > first + 1 && !CPP_OPTION (pfile, lang_asm)) | |
1768 | { | |
1769 | if (first[1].type == CPP_NAME) | |
1770 | cpp_error (pfile, "invalid preprocessing directive #%.*s", | |
bfb9dc7f | 1771 | (int) first[1].val.node->length, first[1].val.node->name); |
041c3194 ZW |
1772 | else |
1773 | cpp_error (pfile, "invalid preprocessing directive"); | |
1774 | } | |
1775 | ||
91fcd158 NB |
1776 | /* Put EOF at end of known directives. This covers "directives do |
1777 | not extend beyond the end of the line (description 6.10 part 2)". */ | |
1778 | if (KNOWN_DIRECTIVE () || !pfile->done_initializing) | |
041c3194 ZW |
1779 | { |
1780 | pfile->first_directive_token = first; | |
1781 | cur_token++->type = CPP_EOF; | |
1782 | } | |
1783 | ||
91fcd158 NB |
1784 | /* Directives, known or not, always start a new line. */ |
1785 | if (first_token == 0 || list->tokens[first_token].type == CPP_HASH) | |
041c3194 | 1786 | first->flags |= BOL; |
6d2c2047 | 1787 | else |
041c3194 ZW |
1788 | /* 6.10.3.10: Within the sequence of preprocessing tokens making |
1789 | up the invocation of a function-like macro, new line is | |
1790 | considered a normal white-space character. */ | |
1791 | first->flags |= PREV_WHITE; | |
1792 | ||
1793 | buffer->cur = cur; | |
1794 | list->tokens_used = cur_token - list->tokens; | |
1795 | pfile->in_lex_line = 0; | |
6d2c2047 ZW |
1796 | } |
1797 | ||
041c3194 ZW |
1798 | /* Write the spelling of a token TOKEN, with any appropriate |
1799 | whitespace before it, to the token_buffer. PREV is the previous | |
1800 | token, which is used to determine if we need to shove in an extra | |
1801 | space in order to avoid accidental token paste. */ | |
1802 | static void | |
1803 | output_token (pfile, token, prev) | |
1804 | cpp_reader *pfile; | |
1805 | const cpp_token *token, *prev; | |
1806 | { | |
1807 | int dummy; | |
c5a04734 | 1808 | |
041c3194 ZW |
1809 | if (token->col && (token->flags & BOL)) |
1810 | { | |
1811 | /* Supply enough whitespace to put this token in its original | |
1812 | column. Don't bother trying to reconstruct tabs; we can't | |
1813 | get it right in general, and nothing ought to care. (Yes, | |
1814 | some things do care; the fault lies with them.) */ | |
1815 | unsigned char *buffer; | |
1816 | unsigned int spaces = token->col - 1; | |
1817 | ||
1818 | CPP_RESERVE (pfile, token->col); | |
1819 | buffer = pfile->limit; | |
1820 | ||
1821 | while (spaces--) | |
1822 | *buffer++ = ' '; | |
1823 | pfile->limit = buffer; | |
1824 | } | |
1825 | else if (token->flags & PREV_WHITE) | |
1826 | CPP_PUTC (pfile, ' '); | |
1827 | /* Check for and prevent accidental token pasting, in ANSI mode. */ | |
d6d5f795 | 1828 | |
041c3194 ZW |
1829 | else if (!CPP_TRADITIONAL (pfile) && prev) |
1830 | { | |
1831 | if (can_paste (pfile, prev, token, &dummy) != CPP_EOF) | |
1832 | CPP_PUTC (pfile, ' '); | |
1833 | /* can_paste catches most of the accidental paste cases, but not all. | |
1834 | Consider a + ++b - if there is not a space between the + and ++, it | |
1835 | will be misparsed as a++ + b. */ | |
1836 | else if ((prev->type == CPP_PLUS && token->type == CPP_PLUS_PLUS) | |
1837 | || (prev->type == CPP_MINUS && token->type == CPP_MINUS_MINUS)) | |
1838 | CPP_PUTC (pfile, ' '); | |
1839 | } | |
d6d5f795 | 1840 | |
041c3194 ZW |
1841 | CPP_RESERVE (pfile, TOKEN_LEN (token)); |
1842 | pfile->limit = spell_token (pfile, token, pfile->limit); | |
1843 | } | |
d6d5f795 | 1844 | |
041c3194 | 1845 | /* Write the spelling of a token TOKEN to BUFFER. The buffer must |
cf00a885 ZW |
1846 | already contain the enough space to hold the token's spelling. |
1847 | Returns a pointer to the character after the last character | |
1848 | written. */ | |
d6d5f795 | 1849 | |
041c3194 ZW |
1850 | static unsigned char * |
1851 | spell_token (pfile, token, buffer) | |
1852 | cpp_reader *pfile; /* Would be nice to be rid of this... */ | |
1853 | const cpp_token *token; | |
1854 | unsigned char *buffer; | |
1855 | { | |
1856 | switch (token_spellings[token->type].type) | |
1857 | { | |
1858 | case SPELL_OPERATOR: | |
1859 | { | |
1860 | const unsigned char *spelling; | |
1861 | unsigned char c; | |
d6d5f795 | 1862 | |
041c3194 ZW |
1863 | if (token->flags & DIGRAPH) |
1864 | spelling = digraph_spellings[token->type - CPP_FIRST_DIGRAPH]; | |
1865 | else | |
1866 | spelling = token_spellings[token->type].spelling; | |
1867 | ||
1868 | while ((c = *spelling++) != '\0') | |
1869 | *buffer++ = c; | |
1870 | } | |
1871 | break; | |
d6d5f795 | 1872 | |
041c3194 | 1873 | case SPELL_IDENT: |
bfb9dc7f ZW |
1874 | memcpy (buffer, token->val.node->name, token->val.node->length); |
1875 | buffer += token->val.node->length; | |
041c3194 | 1876 | break; |
d6d5f795 | 1877 | |
041c3194 ZW |
1878 | case SPELL_STRING: |
1879 | { | |
041c3194 ZW |
1880 | if (token->type == CPP_WSTRING || token->type == CPP_WCHAR) |
1881 | *buffer++ = 'L'; | |
bfb9dc7f | 1882 | |
041c3194 | 1883 | if (token->type == CPP_STRING || token->type == CPP_WSTRING) |
bfb9dc7f ZW |
1884 | *buffer++ = '"'; |
1885 | if (token->type == CPP_CHAR || token->type == CPP_WCHAR) | |
1886 | *buffer++ = '\''; | |
1887 | ||
1888 | memcpy (buffer, token->val.str.text, token->val.str.len); | |
1889 | buffer += token->val.str.len; | |
1890 | ||
1891 | if (token->type == CPP_STRING || token->type == CPP_WSTRING) | |
1892 | *buffer++ = '"'; | |
1893 | if (token->type == CPP_CHAR || token->type == CPP_WCHAR) | |
1894 | *buffer++ = '\''; | |
041c3194 ZW |
1895 | } |
1896 | break; | |
d6d5f795 | 1897 | |
041c3194 ZW |
1898 | case SPELL_CHAR: |
1899 | *buffer++ = token->val.aux; | |
1900 | break; | |
d6d5f795 | 1901 | |
041c3194 ZW |
1902 | case SPELL_NONE: |
1903 | cpp_ice (pfile, "Unspellable token %s", token_names[token->type]); | |
1904 | break; | |
1905 | } | |
d6d5f795 | 1906 | |
041c3194 ZW |
1907 | return buffer; |
1908 | } | |
d6d5f795 | 1909 | |
cf00a885 ZW |
1910 | /* Return the spelling of a token known to be an operator. |
1911 | Does not distinguish digraphs from their counterparts. */ | |
1912 | const unsigned char * | |
1913 | _cpp_spell_operator (type) | |
1914 | enum cpp_ttype type; | |
1915 | { | |
1916 | if (token_spellings[type].type == SPELL_OPERATOR) | |
1917 | return token_spellings[type].spelling; | |
1918 | else | |
1919 | return token_names[type]; | |
1920 | } | |
1921 | ||
1922 | ||
041c3194 | 1923 | /* Macro expansion algorithm. TODO. */ |
d6d5f795 | 1924 | |
7de9cc38 KG |
1925 | static const cpp_token placemarker_token = {0, 0, CPP_PLACEMARKER, 0 UNION_INIT_ZERO}; |
1926 | static const cpp_token eof_token = {0, 0, CPP_EOF, 0 UNION_INIT_ZERO}; | |
c5a04734 | 1927 | |
041c3194 ZW |
1928 | #define IS_ARG_CONTEXT(c) ((c)->flags & CONTEXT_ARG) |
1929 | #define CURRENT_CONTEXT(pfile) ((pfile)->contexts + (pfile)->cur_context) | |
1930 | ||
1931 | /* Flags for cpp_context. */ | |
1932 | #define CONTEXT_PASTEL (1 << 0) /* An argument context on LHS of ##. */ | |
1933 | #define CONTEXT_PASTER (1 << 1) /* An argument context on RHS of ##. */ | |
1934 | #define CONTEXT_RAW (1 << 2) /* If argument tokens already expanded. */ | |
1935 | #define CONTEXT_ARG (1 << 3) /* If an argument context. */ | |
1936 | ||
1937 | #define ASSIGN_FLAGS_AND_POS(d, s) \ | |
1938 | do {(d)->flags = (s)->flags & (PREV_WHITE | BOL | PASTE_LEFT); \ | |
1939 | if ((d)->flags & BOL) {(d)->col = (s)->col; (d)->line = (s)->line;} \ | |
1940 | } while (0) | |
1941 | ||
1942 | /* f is flags, just consisting of PREV_WHITE | BOL. */ | |
1943 | #define MODIFY_FLAGS_AND_POS(d, s, f) \ | |
1944 | do {(d)->flags &= ~(PREV_WHITE | BOL); (d)->flags |= (f); \ | |
1945 | if ((f) & BOL) {(d)->col = (s)->col; (d)->line = (s)->line;} \ | |
1946 | } while (0) | |
1947 | ||
1948 | typedef struct cpp_context cpp_context; | |
1949 | struct cpp_context | |
1950 | { | |
1951 | union | |
1952 | { | |
1953 | const cpp_toklist *list; /* Used for macro contexts only. */ | |
1954 | const cpp_token **arg; /* Used for arg contexts only. */ | |
1955 | } u; | |
1956 | ||
1957 | /* Pushed token to be returned by next call to cpp_get_token. */ | |
1958 | const cpp_token *pushed_token; | |
1959 | ||
1960 | struct macro_args *args; /* 0 for arguments and object-like macros. */ | |
1961 | unsigned short posn; /* Current posn, index into u. */ | |
1962 | unsigned short count; /* No. of tokens in u. */ | |
1963 | unsigned short level; | |
1964 | unsigned char flags; | |
1965 | }; | |
1966 | ||
1967 | typedef struct macro_args macro_args; | |
1968 | struct macro_args | |
1969 | { | |
1970 | unsigned int *ends; | |
1971 | const cpp_token **tokens; | |
1972 | unsigned int capacity; | |
1973 | unsigned int used; | |
1974 | unsigned short level; | |
1975 | }; | |
1976 | ||
1977 | static const cpp_token *get_raw_token PARAMS ((cpp_reader *)); | |
1978 | static const cpp_token *parse_arg PARAMS ((cpp_reader *, int, unsigned int, | |
1979 | macro_args *, unsigned int *)); | |
1980 | static int parse_args PARAMS ((cpp_reader *, cpp_hashnode *, macro_args *)); | |
1981 | static void save_token PARAMS ((macro_args *, const cpp_token *)); | |
1982 | static const cpp_token *push_arg_context PARAMS ((cpp_reader *, | |
1983 | const cpp_token *)); | |
1984 | static int do_pop_context PARAMS ((cpp_reader *)); | |
1985 | static const cpp_token *pop_context PARAMS ((cpp_reader *)); | |
1986 | static const cpp_token *push_macro_context PARAMS ((cpp_reader *, | |
1987 | cpp_hashnode *, | |
1988 | const cpp_token *)); | |
1989 | static void free_macro_args PARAMS ((macro_args *)); | |
1990 | ||
1991 | /* Free the storage allocated for macro arguments. */ | |
1992 | static void | |
1993 | free_macro_args (args) | |
1994 | macro_args *args; | |
c5a04734 | 1995 | { |
041c3194 ZW |
1996 | if (args->tokens) |
1997 | free (args->tokens); | |
1998 | free (args->ends); | |
1999 | free (args); | |
c5a04734 ZW |
2000 | } |
2001 | ||
041c3194 ZW |
2002 | /* Determines if a macro has been already used (and is therefore |
2003 | disabled). */ | |
c5a04734 | 2004 | static int |
041c3194 | 2005 | is_macro_disabled (pfile, expansion, token) |
c5a04734 | 2006 | cpp_reader *pfile; |
041c3194 ZW |
2007 | const cpp_toklist *expansion; |
2008 | const cpp_token *token; | |
c5a04734 | 2009 | { |
041c3194 ZW |
2010 | cpp_context *context = CURRENT_CONTEXT (pfile); |
2011 | ||
cf00a885 ZW |
2012 | /* Don't expand anything if this file has already been preprocessed. */ |
2013 | if (CPP_OPTION (pfile, preprocessed)) | |
2014 | return 1; | |
2015 | ||
041c3194 ZW |
2016 | /* Arguments on either side of ## are inserted in place without |
2017 | macro expansion (6.10.3.3.2). Conceptually, any macro expansion | |
2018 | occurs during a later rescan pass. The effect is that we expand | |
2019 | iff we would as part of the macro's expansion list, so we should | |
2020 | drop to the macro's context. */ | |
2021 | if (IS_ARG_CONTEXT (context)) | |
c5a04734 | 2022 | { |
041c3194 ZW |
2023 | if (token->flags & PASTED) |
2024 | context--; | |
2025 | else if (!(context->flags & CONTEXT_RAW)) | |
2026 | return 1; | |
2027 | else if (context->flags & (CONTEXT_PASTEL | CONTEXT_PASTER)) | |
2028 | context--; | |
c5a04734 | 2029 | } |
c5a04734 | 2030 | |
041c3194 ZW |
2031 | /* Have we already used this macro? */ |
2032 | while (context->level > 0) | |
c5a04734 | 2033 | { |
041c3194 ZW |
2034 | if (!IS_ARG_CONTEXT (context) && context->u.list == expansion) |
2035 | return 1; | |
2036 | /* Raw argument tokens are judged based on the token list they | |
2037 | came from. */ | |
2038 | if (context->flags & CONTEXT_RAW) | |
2039 | context = pfile->contexts + context->level; | |
2040 | else | |
2041 | context--; | |
2042 | } | |
c5a04734 | 2043 | |
041c3194 ZW |
2044 | /* Function-like macros may be disabled if the '(' is not in the |
2045 | current context. We check this without disrupting the context | |
2046 | stack. */ | |
2047 | if (expansion->paramc >= 0) | |
2048 | { | |
2049 | const cpp_token *next; | |
2050 | unsigned int prev_nme; | |
c5a04734 | 2051 | |
041c3194 ZW |
2052 | context = CURRENT_CONTEXT (pfile); |
2053 | /* Drop down any contexts we're at the end of: the '(' may | |
2054 | appear in lower macro expansions, or in the rest of the file. */ | |
2055 | while (context->posn == context->count && context > pfile->contexts) | |
2056 | { | |
2057 | context--; | |
2058 | /* If we matched, we are disabled, as we appear in the | |
2059 | expansion of each macro we meet. */ | |
2060 | if (!IS_ARG_CONTEXT (context) && context->u.list == expansion) | |
2061 | return 1; | |
2062 | } | |
c5a04734 | 2063 | |
041c3194 ZW |
2064 | prev_nme = pfile->no_expand_level; |
2065 | pfile->no_expand_level = context - pfile->contexts; | |
2066 | next = cpp_get_token (pfile); | |
2067 | restore_macro_expansion (pfile, prev_nme); | |
2068 | if (next->type != CPP_OPEN_PAREN) | |
2069 | { | |
2070 | _cpp_push_token (pfile, next); | |
2071 | if (CPP_OPTION (pfile, warn_traditional)) | |
2072 | cpp_warning (pfile, | |
2073 | "function macro %.*s must be used with arguments in traditional C", | |
bfb9dc7f | 2074 | (int) token->val.node->length, token->val.node->name); |
041c3194 ZW |
2075 | return 1; |
2076 | } | |
c5a04734 | 2077 | } |
c5a04734 | 2078 | |
041c3194 | 2079 | return 0; |
c5a04734 ZW |
2080 | } |
2081 | ||
041c3194 ZW |
2082 | /* Add a token to the set of tokens forming the arguments to the macro |
2083 | being parsed in parse_args. */ | |
2084 | static void | |
2085 | save_token (args, token) | |
2086 | macro_args *args; | |
2087 | const cpp_token *token; | |
c5a04734 | 2088 | { |
041c3194 ZW |
2089 | if (args->used == args->capacity) |
2090 | { | |
2091 | args->capacity += args->capacity + 100; | |
2092 | args->tokens = (const cpp_token **) | |
2093 | xrealloc (args->tokens, args->capacity * sizeof (const cpp_token *)); | |
2094 | } | |
2095 | args->tokens[args->used++] = token; | |
c5a04734 ZW |
2096 | } |
2097 | ||
041c3194 ZW |
2098 | /* Take and save raw tokens until we finish one argument. Empty |
2099 | arguments are saved as a single CPP_PLACEMARKER token. */ | |
2100 | static const cpp_token * | |
2101 | parse_arg (pfile, var_args, paren_context, args, pcount) | |
c5a04734 | 2102 | cpp_reader *pfile; |
041c3194 ZW |
2103 | int var_args; |
2104 | unsigned int paren_context; | |
2105 | macro_args *args; | |
2106 | unsigned int *pcount; | |
c5a04734 | 2107 | { |
041c3194 ZW |
2108 | const cpp_token *token; |
2109 | unsigned int paren = 0, count = 0; | |
2110 | int raw, was_raw = 1; | |
c5a04734 | 2111 | |
041c3194 | 2112 | for (count = 0;; count++) |
c5a04734 | 2113 | { |
041c3194 | 2114 | token = cpp_get_token (pfile); |
c5a04734 | 2115 | |
041c3194 | 2116 | switch (token->type) |
c5a04734 | 2117 | { |
041c3194 ZW |
2118 | default: |
2119 | break; | |
c5a04734 | 2120 | |
041c3194 ZW |
2121 | case CPP_OPEN_PAREN: |
2122 | paren++; | |
2123 | break; | |
2124 | ||
2125 | case CPP_CLOSE_PAREN: | |
2126 | if (paren-- != 0) | |
2127 | break; | |
2128 | goto out; | |
2129 | ||
2130 | case CPP_COMMA: | |
2131 | /* Commas are not terminators within parantheses or var_args. */ | |
2132 | if (paren || var_args) | |
2133 | break; | |
2134 | goto out; | |
2135 | ||
2136 | case CPP_EOF: /* Error reported by caller. */ | |
2137 | goto out; | |
c5a04734 | 2138 | } |
c5a04734 | 2139 | |
041c3194 ZW |
2140 | raw = pfile->cur_context <= paren_context; |
2141 | if (raw != was_raw) | |
2142 | { | |
2143 | was_raw = raw; | |
2144 | save_token (args, 0); | |
2145 | count++; | |
c5a04734 | 2146 | } |
041c3194 | 2147 | save_token (args, token); |
c5a04734 | 2148 | } |
c5a04734 ZW |
2149 | |
2150 | out: | |
041c3194 ZW |
2151 | if (count == 0) |
2152 | { | |
2153 | /* Duplicate the placemarker. Then we can set its flags and | |
2154 | position and safely be using more than one. */ | |
2155 | save_token (args, duplicate_token (pfile, &placemarker_token)); | |
2156 | count++; | |
2157 | } | |
2158 | ||
2159 | *pcount = count; | |
2160 | return token; | |
c5a04734 ZW |
2161 | } |
2162 | ||
041c3194 ZW |
2163 | /* This macro returns true if the argument starting at offset O of arglist |
2164 | A is empty - that is, it's either a single PLACEMARKER token, or a null | |
2165 | pointer followed by a PLACEMARKER. */ | |
2166 | ||
2167 | #define empty_argument(A, O) \ | |
2168 | ((A)->tokens[O] ? (A)->tokens[O]->type == CPP_PLACEMARKER \ | |
2169 | : (A)->tokens[(O)+1]->type == CPP_PLACEMARKER) | |
2170 | ||
2171 | /* Parse the arguments making up a macro invocation. Nested arguments | |
2172 | are automatically macro expanded, but immediate macros are not | |
2173 | expanded; this enables e.g. operator # to work correctly. Returns | |
2174 | non-zero on error. */ | |
c5a04734 | 2175 | static int |
041c3194 | 2176 | parse_args (pfile, hp, args) |
c5a04734 | 2177 | cpp_reader *pfile; |
041c3194 ZW |
2178 | cpp_hashnode *hp; |
2179 | macro_args *args; | |
c5a04734 | 2180 | { |
041c3194 ZW |
2181 | const cpp_token *token; |
2182 | const cpp_toklist *macro; | |
2183 | unsigned int total = 0; | |
2184 | unsigned int paren_context = pfile->cur_context; | |
2185 | int argc = 0; | |
2186 | ||
2187 | macro = hp->value.expansion; | |
2188 | do | |
c5a04734 | 2189 | { |
041c3194 | 2190 | unsigned int count; |
c5a04734 | 2191 | |
041c3194 ZW |
2192 | token = parse_arg (pfile, (argc + 1 == macro->paramc |
2193 | && (macro->flags & VAR_ARGS)), | |
2194 | paren_context, args, &count); | |
2195 | if (argc < macro->paramc) | |
c5a04734 | 2196 | { |
041c3194 ZW |
2197 | total += count; |
2198 | args->ends[argc] = total; | |
c5a04734 | 2199 | } |
041c3194 | 2200 | argc++; |
c5a04734 | 2201 | } |
041c3194 | 2202 | while (token->type != CPP_CLOSE_PAREN && token->type != CPP_EOF); |
c5a04734 | 2203 | |
041c3194 ZW |
2204 | if (token->type == CPP_EOF) |
2205 | { | |
2206 | cpp_error (pfile, "unterminated invocation of macro \"%.*s\"", | |
2207 | hp->length, hp->name); | |
2208 | return 1; | |
2209 | } | |
2210 | else if (argc < macro->paramc) | |
2211 | { | |
2212 | /* A rest argument is allowed to not appear in the invocation at all. | |
2213 | e.g. #define debug(format, args...) ... | |
2214 | debug("string"); | |
2215 | This is exactly the same as if the rest argument had received no | |
563dd08a | 2216 | tokens - debug("string",); This extension is deprecated. */ |
041c3194 | 2217 | |
563dd08a | 2218 | if (argc + 1 == macro->paramc && (macro->flags & GNU_REST_ARGS)) |
041c3194 ZW |
2219 | { |
2220 | /* Duplicate the placemarker. Then we can set its flags and | |
2221 | position and safely be using more than one. */ | |
2222 | save_token (args, duplicate_token (pfile, &placemarker_token)); | |
2223 | args->ends[argc] = total + 1; | |
2224 | return 0; | |
2225 | } | |
2226 | else | |
2227 | { | |
2228 | cpp_error (pfile, | |
2229 | "insufficient arguments in invocation of macro \"%.*s\"", | |
2230 | hp->length, hp->name); | |
2231 | return 1; | |
2232 | } | |
2233 | } | |
2234 | /* An empty argument to an empty function-like macro is fine. */ | |
2235 | else if (argc > macro->paramc | |
2236 | && !(macro->paramc == 0 && argc == 1 && empty_argument (args, 0))) | |
2237 | { | |
2238 | cpp_error (pfile, | |
2239 | "too many arguments in invocation of macro \"%.*s\"", | |
2240 | hp->length, hp->name); | |
2241 | return 1; | |
2242 | } | |
c5a04734 | 2243 | |
041c3194 ZW |
2244 | return 0; |
2245 | } | |
2246 | ||
2247 | /* Adds backslashes before all backslashes and double quotes appearing | |
2248 | in strings. Non-printable characters are converted to octal. */ | |
2249 | static U_CHAR * | |
2250 | quote_string (dest, src, len) | |
2251 | U_CHAR *dest; | |
2252 | const U_CHAR *src; | |
2253 | unsigned int len; | |
2254 | { | |
2255 | while (len--) | |
c5a04734 | 2256 | { |
041c3194 | 2257 | U_CHAR c = *src++; |
c5a04734 | 2258 | |
041c3194 | 2259 | if (c == '\\' || c == '"') |
6ab3e7dd | 2260 | { |
041c3194 ZW |
2261 | *dest++ = '\\'; |
2262 | *dest++ = c; | |
2263 | } | |
2264 | else | |
2265 | { | |
2266 | if (ISPRINT (c)) | |
2267 | *dest++ = c; | |
2268 | else | |
2269 | { | |
2270 | sprintf ((char *) dest, "\\%03o", c); | |
2271 | dest += 4; | |
2272 | } | |
6ab3e7dd | 2273 | } |
c5a04734 | 2274 | } |
c5a04734 | 2275 | |
041c3194 | 2276 | return dest; |
c5a04734 ZW |
2277 | } |
2278 | ||
041c3194 ZW |
2279 | /* Allocates a buffer to hold a token's TEXT, and converts TOKEN to a |
2280 | CPP_STRING token containing TEXT in quoted form. */ | |
2281 | static cpp_token * | |
2282 | make_string_token (token, text, len) | |
2283 | cpp_token *token; | |
2284 | const U_CHAR *text; | |
2285 | unsigned int len; | |
2286 | { | |
2287 | U_CHAR *buf; | |
2288 | ||
2289 | buf = (U_CHAR *) xmalloc (len * 4); | |
2290 | token->type = CPP_STRING; | |
2291 | token->flags = 0; | |
bfb9dc7f ZW |
2292 | token->val.str.text = buf; |
2293 | token->val.str.len = quote_string (buf, text, len) - buf; | |
041c3194 ZW |
2294 | return token; |
2295 | } | |
2296 | ||
2297 | /* Allocates and converts a temporary token to a CPP_NUMBER token, | |
2298 | evaluating to NUMBER. */ | |
2299 | static cpp_token * | |
2300 | alloc_number_token (pfile, number) | |
c5a04734 | 2301 | cpp_reader *pfile; |
041c3194 | 2302 | int number; |
c5a04734 | 2303 | { |
041c3194 ZW |
2304 | cpp_token *result; |
2305 | char *buf; | |
2306 | ||
2307 | result = get_temp_token (pfile); | |
2308 | buf = xmalloc (20); | |
2309 | sprintf (buf, "%d", number); | |
2310 | ||
2311 | result->type = CPP_NUMBER; | |
2312 | result->flags = 0; | |
bfb9dc7f ZW |
2313 | result->val.str.text = (U_CHAR *) buf; |
2314 | result->val.str.len = strlen (buf); | |
041c3194 ZW |
2315 | return result; |
2316 | } | |
c5a04734 | 2317 | |
041c3194 ZW |
2318 | /* Returns a temporary token from the temporary token store of PFILE. */ |
2319 | static cpp_token * | |
2320 | get_temp_token (pfile) | |
2321 | cpp_reader *pfile; | |
2322 | { | |
2323 | if (pfile->temp_used == pfile->temp_alloced) | |
2324 | { | |
2325 | if (pfile->temp_used == pfile->temp_cap) | |
2326 | { | |
2327 | pfile->temp_cap += pfile->temp_cap + 20; | |
2328 | pfile->temp_tokens = (cpp_token **) xrealloc | |
2329 | (pfile->temp_tokens, pfile->temp_cap * sizeof (cpp_token *)); | |
2330 | } | |
2331 | pfile->temp_tokens[pfile->temp_alloced++] = (cpp_token *) xmalloc | |
2332 | (sizeof (cpp_token)); | |
2333 | } | |
c5a04734 | 2334 | |
041c3194 ZW |
2335 | return pfile->temp_tokens[pfile->temp_used++]; |
2336 | } | |
2337 | ||
2338 | /* Release (not free) for re-use the temporary tokens of PFILE. */ | |
2339 | static void | |
2340 | release_temp_tokens (pfile) | |
2341 | cpp_reader *pfile; | |
2342 | { | |
2343 | while (pfile->temp_used) | |
c5a04734 | 2344 | { |
041c3194 | 2345 | cpp_token *token = pfile->temp_tokens[--pfile->temp_used]; |
c5a04734 | 2346 | |
bfb9dc7f | 2347 | if (token_spellings[token->type].type == SPELL_STRING) |
c5a04734 | 2348 | { |
bfb9dc7f ZW |
2349 | free ((char *) token->val.str.text); |
2350 | token->val.str.text = 0; | |
c5a04734 ZW |
2351 | } |
2352 | } | |
041c3194 | 2353 | } |
c5a04734 | 2354 | |
041c3194 ZW |
2355 | /* Free all of PFILE's dynamically-allocated temporary tokens. */ |
2356 | void | |
2357 | _cpp_free_temp_tokens (pfile) | |
2358 | cpp_reader *pfile; | |
2359 | { | |
2360 | if (pfile->temp_tokens) | |
c5a04734 | 2361 | { |
041c3194 ZW |
2362 | /* It is possible, though unlikely (looking for '(' of a funlike |
2363 | macro into EOF), that we haven't released the tokens yet. */ | |
2364 | release_temp_tokens (pfile); | |
2365 | while (pfile->temp_alloced) | |
2366 | free (pfile->temp_tokens[--pfile->temp_alloced]); | |
2367 | free (pfile->temp_tokens); | |
c5a04734 ZW |
2368 | } |
2369 | ||
041c3194 ZW |
2370 | if (pfile->date) |
2371 | { | |
bfb9dc7f | 2372 | free ((char *) pfile->date->val.str.text); |
041c3194 | 2373 | free (pfile->date); |
bfb9dc7f | 2374 | free ((char *) pfile->time->val.str.text); |
041c3194 ZW |
2375 | free (pfile->time); |
2376 | } | |
c5a04734 ZW |
2377 | } |
2378 | ||
041c3194 ZW |
2379 | /* Copy TOKEN into a temporary token from PFILE's store. */ |
2380 | static cpp_token * | |
2381 | duplicate_token (pfile, token) | |
2382 | cpp_reader *pfile; | |
2383 | const cpp_token *token; | |
2384 | { | |
2385 | cpp_token *result = get_temp_token (pfile); | |
c5a04734 | 2386 | |
041c3194 | 2387 | *result = *token; |
bfb9dc7f | 2388 | if (token_spellings[token->type].type == SPELL_STRING) |
041c3194 | 2389 | { |
bfb9dc7f ZW |
2390 | U_CHAR *buff = (U_CHAR *) xmalloc (token->val.str.len); |
2391 | memcpy (buff, token->val.str.text, token->val.str.len); | |
2392 | result->val.str.text = buff; | |
041c3194 ZW |
2393 | } |
2394 | return result; | |
2395 | } | |
c5a04734 | 2396 | |
041c3194 ZW |
2397 | /* Determine whether two tokens can be pasted together, and if so, |
2398 | what the resulting token is. Returns CPP_EOF if the tokens cannot | |
2399 | be pasted, or the appropriate type for the merged token if they | |
2400 | can. */ | |
2401 | static enum cpp_ttype | |
2402 | can_paste (pfile, token1, token2, digraph) | |
2403 | cpp_reader * pfile; | |
2404 | const cpp_token *token1, *token2; | |
2405 | int* digraph; | |
c5a04734 | 2406 | { |
041c3194 ZW |
2407 | enum cpp_ttype a = token1->type, b = token2->type; |
2408 | int cxx = CPP_OPTION (pfile, cplusplus); | |
c5a04734 | 2409 | |
041c3194 ZW |
2410 | if (a <= CPP_LAST_EQ && b == CPP_EQ) |
2411 | return a + (CPP_EQ_EQ - CPP_EQ); | |
c5a04734 | 2412 | |
041c3194 | 2413 | switch (a) |
c5a04734 | 2414 | { |
041c3194 ZW |
2415 | case CPP_GREATER: |
2416 | if (b == a) return CPP_RSHIFT; | |
2417 | if (b == CPP_QUERY && cxx) return CPP_MAX; | |
2418 | if (b == CPP_GREATER_EQ) return CPP_RSHIFT_EQ; | |
2419 | break; | |
2420 | case CPP_LESS: | |
2421 | if (b == a) return CPP_LSHIFT; | |
2422 | if (b == CPP_QUERY && cxx) return CPP_MIN; | |
2423 | if (b == CPP_LESS_EQ) return CPP_LSHIFT_EQ; | |
2424 | if (b == CPP_COLON) | |
2425 | {*digraph = 1; return CPP_OPEN_SQUARE;} /* <: digraph */ | |
2426 | if (b == CPP_MOD) | |
2427 | {*digraph = 1; return CPP_OPEN_BRACE;} /* <% digraph */ | |
2428 | break; | |
c5a04734 | 2429 | |
041c3194 ZW |
2430 | case CPP_PLUS: if (b == a) return CPP_PLUS_PLUS; break; |
2431 | case CPP_AND: if (b == a) return CPP_AND_AND; break; | |
2432 | case CPP_OR: if (b == a) return CPP_OR_OR; break; | |
c5a04734 | 2433 | |
041c3194 ZW |
2434 | case CPP_MINUS: |
2435 | if (b == a) return CPP_MINUS_MINUS; | |
2436 | if (b == CPP_GREATER) return CPP_DEREF; | |
2437 | break; | |
2438 | case CPP_COLON: | |
2439 | if (b == a && cxx) return CPP_SCOPE; | |
2440 | if (b == CPP_GREATER) | |
2441 | {*digraph = 1; return CPP_CLOSE_SQUARE;} /* :> digraph */ | |
2442 | break; | |
2443 | ||
2444 | case CPP_MOD: | |
2445 | if (b == CPP_GREATER) | |
2446 | {*digraph = 1; return CPP_CLOSE_BRACE;} /* %> digraph */ | |
2447 | if (b == CPP_COLON) | |
2448 | {*digraph = 1; return CPP_HASH;} /* %: digraph */ | |
2449 | break; | |
2450 | case CPP_DEREF: | |
2451 | if (b == CPP_MULT && cxx) return CPP_DEREF_STAR; | |
2452 | break; | |
2453 | case CPP_DOT: | |
2454 | if (b == CPP_MULT && cxx) return CPP_DOT_STAR; | |
2455 | if (b == CPP_NUMBER) return CPP_NUMBER; | |
2456 | break; | |
2457 | ||
2458 | case CPP_HASH: | |
2459 | if (b == a && (token1->flags & DIGRAPH) == (token2->flags & DIGRAPH)) | |
2460 | /* %:%: digraph */ | |
2461 | {*digraph = (token1->flags & DIGRAPH); return CPP_PASTE;} | |
2462 | break; | |
2463 | ||
2464 | case CPP_NAME: | |
2465 | if (b == CPP_NAME) return CPP_NAME; | |
2466 | if (b == CPP_NUMBER | |
bfb9dc7f | 2467 | && is_numstart(token2->val.str.text[0])) return CPP_NAME; |
041c3194 | 2468 | if (b == CPP_CHAR |
bfb9dc7f | 2469 | && token1->val.node == pfile->spec_nodes->n_L) return CPP_WCHAR; |
041c3194 | 2470 | if (b == CPP_STRING |
bfb9dc7f | 2471 | && token1->val.node == pfile->spec_nodes->n_L) return CPP_WSTRING; |
041c3194 ZW |
2472 | break; |
2473 | ||
2474 | case CPP_NUMBER: | |
2475 | if (b == CPP_NUMBER) return CPP_NUMBER; | |
2476 | if (b == CPP_NAME) return CPP_NUMBER; | |
2477 | if (b == CPP_DOT) return CPP_NUMBER; | |
2478 | /* Numbers cannot have length zero, so this is safe. */ | |
2479 | if ((b == CPP_PLUS || b == CPP_MINUS) | |
bfb9dc7f | 2480 | && VALID_SIGN ('+', token1->val.str.text[token1->val.str.len - 1])) |
041c3194 ZW |
2481 | return CPP_NUMBER; |
2482 | break; | |
2483 | ||
2484 | default: | |
2485 | break; | |
c5a04734 ZW |
2486 | } |
2487 | ||
041c3194 ZW |
2488 | return CPP_EOF; |
2489 | } | |
2490 | ||
2491 | /* Check if TOKEN is to be ##-pasted with the token after it. */ | |
2492 | static const cpp_token * | |
2493 | maybe_paste_with_next (pfile, token) | |
2494 | cpp_reader *pfile; | |
2495 | const cpp_token *token; | |
2496 | { | |
2497 | cpp_token *pasted; | |
2498 | const cpp_token *second; | |
2499 | cpp_context *context = CURRENT_CONTEXT (pfile); | |
2500 | ||
2501 | /* Is this token on the LHS of ## ? */ | |
2502 | if (!((context->flags & CONTEXT_PASTEL) && context->posn == context->count) | |
2503 | && !(token->flags & PASTE_LEFT)) | |
2504 | return token; | |
2505 | ||
2506 | /* Prevent recursion, and possibly pushing back more than one token. */ | |
2507 | if (pfile->paste_level) | |
2508 | return token; | |
2509 | ||
2510 | /* Suppress macro expansion for next token, but don't conflict with | |
2511 | the other method of suppression. If it is an argument, macro | |
2512 | expansion within the argument will still occur. */ | |
2513 | pfile->paste_level = pfile->cur_context; | |
2514 | second = cpp_get_token (pfile); | |
2515 | pfile->paste_level = 0; | |
2516 | ||
563dd08a NB |
2517 | /* Ignore placemarker argument tokens (cannot be from an empty macro |
2518 | since macros are not expanded). */ | |
041c3194 ZW |
2519 | if (token->type == CPP_PLACEMARKER) |
2520 | pasted = duplicate_token (pfile, second); | |
2521 | else if (second->type == CPP_PLACEMARKER) | |
c5a04734 | 2522 | { |
563dd08a | 2523 | cpp_context *mac_context = CURRENT_CONTEXT (pfile) - 1; |
041c3194 ZW |
2524 | /* GCC has special extended semantics for a ## b where b is a |
2525 | varargs parameter: a disappears if b consists of no tokens. | |
2526 | This extension is deprecated. */ | |
563dd08a NB |
2527 | if ((mac_context->u.list->flags & GNU_REST_ARGS) |
2528 | && (mac_context->u.list->tokens[mac_context->posn - 1].val.aux + 1 | |
2529 | == (unsigned) mac_context->u.list->paramc)) | |
041c3194 ZW |
2530 | { |
2531 | cpp_warning (pfile, "deprecated GNU ## extension used"); | |
2532 | pasted = duplicate_token (pfile, second); | |
2533 | } | |
2534 | else | |
2535 | pasted = duplicate_token (pfile, token); | |
c5a04734 | 2536 | } |
041c3194 ZW |
2537 | else |
2538 | { | |
2539 | int digraph = 0; | |
2540 | enum cpp_ttype type = can_paste (pfile, token, second, &digraph); | |
c5a04734 | 2541 | |
041c3194 ZW |
2542 | if (type == CPP_EOF) |
2543 | { | |
2544 | if (CPP_OPTION (pfile, warn_paste)) | |
2545 | cpp_warning (pfile, | |
2546 | "pasting would not give a valid preprocessing token"); | |
2547 | _cpp_push_token (pfile, second); | |
2548 | return token; | |
2549 | } | |
c5a04734 | 2550 | |
041c3194 ZW |
2551 | if (type == CPP_NAME || type == CPP_NUMBER) |
2552 | { | |
2553 | /* Join spellings. */ | |
bfb9dc7f | 2554 | U_CHAR *buf, *end; |
c5a04734 | 2555 | |
041c3194 | 2556 | pasted = get_temp_token (pfile); |
bfb9dc7f ZW |
2557 | buf = (U_CHAR *) alloca (TOKEN_LEN (token) + TOKEN_LEN (second)); |
2558 | end = spell_token (pfile, token, buf); | |
2559 | end = spell_token (pfile, second, end); | |
2560 | *end = '\0'; | |
041c3194 | 2561 | |
bfb9dc7f ZW |
2562 | if (type == CPP_NAME) |
2563 | pasted->val.node = cpp_lookup (pfile, buf, end - buf); | |
2564 | else | |
2565 | { | |
2566 | pasted->val.str.text = uxstrdup (buf); | |
2567 | pasted->val.str.len = end - buf; | |
2568 | } | |
041c3194 ZW |
2569 | } |
2570 | else if (type == CPP_WCHAR || type == CPP_WSTRING) | |
2571 | pasted = duplicate_token (pfile, second); | |
2572 | else | |
2573 | { | |
2574 | pasted = get_temp_token (pfile); | |
2575 | pasted->val.integer = 0; | |
2576 | } | |
2577 | ||
2578 | pasted->type = type; | |
bfb9dc7f | 2579 | pasted->flags = digraph ? DIGRAPH : 0; |
041c3194 ZW |
2580 | } |
2581 | ||
2582 | /* The pasted token gets the whitespace flags and position of the | |
2583 | first token, the PASTE_LEFT flag of the second token, plus the | |
2584 | PASTED flag to indicate it is the result of a paste. However, we | |
2585 | want to preserve the DIGRAPH flag. */ | |
2586 | pasted->flags &= ~(PREV_WHITE | BOL | PASTE_LEFT); | |
2587 | pasted->flags |= ((token->flags & (PREV_WHITE | BOL)) | |
2588 | | (second->flags & PASTE_LEFT) | PASTED); | |
2589 | pasted->col = token->col; | |
2590 | pasted->line = token->line; | |
2591 | ||
2592 | return maybe_paste_with_next (pfile, pasted); | |
2593 | } | |
2594 | ||
2595 | /* Convert a token sequence to a single string token according to the | |
2596 | rules of the ISO C #-operator. */ | |
2597 | #define INIT_SIZE 200 | |
2598 | static cpp_token * | |
2599 | stringify_arg (pfile, token) | |
c5a04734 | 2600 | cpp_reader *pfile; |
041c3194 | 2601 | const cpp_token *token; |
c5a04734 | 2602 | { |
041c3194 ZW |
2603 | cpp_token *result; |
2604 | unsigned char *main_buf; | |
2605 | unsigned int prev_value, backslash_count = 0; | |
2606 | unsigned int buf_used = 0, whitespace = 0, buf_cap = INIT_SIZE; | |
c5a04734 | 2607 | |
041c3194 ZW |
2608 | prev_value = prevent_macro_expansion (pfile); |
2609 | main_buf = (unsigned char *) xmalloc (buf_cap); | |
c5a04734 | 2610 | |
041c3194 ZW |
2611 | result = get_temp_token (pfile); |
2612 | ASSIGN_FLAGS_AND_POS (result, token); | |
2613 | ||
2614 | for (; (token = cpp_get_token (pfile))->type != CPP_EOF; ) | |
c5a04734 | 2615 | { |
041c3194 ZW |
2616 | int escape; |
2617 | unsigned char *buf; | |
2618 | unsigned int len = TOKEN_LEN (token); | |
c5a04734 | 2619 | |
041c3194 ZW |
2620 | escape = (token->type == CPP_STRING || token->type == CPP_WSTRING |
2621 | || token->type == CPP_CHAR || token->type == CPP_WCHAR); | |
2622 | if (escape) | |
2623 | len *= 4 + 1; | |
2624 | ||
2625 | if (buf_used + len > buf_cap) | |
c5a04734 | 2626 | { |
041c3194 ZW |
2627 | buf_cap = buf_used + len + INIT_SIZE; |
2628 | main_buf = xrealloc (main_buf, buf_cap); | |
2629 | } | |
c5a04734 | 2630 | |
041c3194 ZW |
2631 | if (whitespace && (token->flags & PREV_WHITE)) |
2632 | main_buf[buf_used++] = ' '; | |
c5a04734 | 2633 | |
041c3194 ZW |
2634 | if (escape) |
2635 | buf = (unsigned char *) xmalloc (len); | |
2636 | else | |
2637 | buf = main_buf + buf_used; | |
2638 | ||
2639 | len = spell_token (pfile, token, buf) - buf; | |
2640 | if (escape) | |
2641 | { | |
2642 | buf_used = quote_string (&main_buf[buf_used], buf, len) - main_buf; | |
2643 | free (buf); | |
2644 | } | |
2645 | else | |
2646 | buf_used += len; | |
c5a04734 | 2647 | |
041c3194 ZW |
2648 | whitespace = 1; |
2649 | if (token->type == CPP_BACKSLASH) | |
2650 | backslash_count++; | |
2651 | else | |
2652 | backslash_count = 0; | |
2653 | } | |
c5a04734 | 2654 | |
041c3194 ZW |
2655 | /* Ignore the final \ of invalid string literals. */ |
2656 | if (backslash_count & 1) | |
2657 | { | |
2658 | cpp_warning (pfile, "invalid string literal, ignoring final '\\'"); | |
2659 | buf_used--; | |
2660 | } | |
c5a04734 | 2661 | |
041c3194 | 2662 | result->type = CPP_STRING; |
bfb9dc7f ZW |
2663 | result->val.str.text = main_buf; |
2664 | result->val.str.len = buf_used; | |
041c3194 ZW |
2665 | restore_macro_expansion (pfile, prev_value); |
2666 | return result; | |
2667 | } | |
c5a04734 | 2668 | |
041c3194 ZW |
2669 | /* Allocate more room on the context stack of PFILE. */ |
2670 | static void | |
2671 | expand_context_stack (pfile) | |
2672 | cpp_reader *pfile; | |
2673 | { | |
2674 | pfile->context_cap += pfile->context_cap + 20; | |
2675 | pfile->contexts = (cpp_context *) | |
2676 | xrealloc (pfile->contexts, pfile->context_cap * sizeof (cpp_context)); | |
2677 | } | |
c5a04734 | 2678 | |
041c3194 ZW |
2679 | /* Push the context of macro NODE onto the context stack. TOKEN is |
2680 | the CPP_NAME token invoking the macro. */ | |
2681 | static const cpp_token * | |
2682 | push_macro_context (pfile, node, token) | |
2683 | cpp_reader *pfile; | |
2684 | cpp_hashnode *node; | |
2685 | const cpp_token *token; | |
2686 | { | |
2687 | unsigned char orig_flags; | |
2688 | macro_args *args; | |
2689 | cpp_context *context; | |
c5a04734 | 2690 | |
041c3194 | 2691 | if (pfile->cur_context > CPP_STACK_MAX) |
c5a04734 | 2692 | { |
041c3194 ZW |
2693 | cpp_error (pfile, "infinite macro recursion invoking '%s'", node->name); |
2694 | return token; | |
c5a04734 ZW |
2695 | } |
2696 | ||
041c3194 ZW |
2697 | /* Token's flags may change when parsing args containing a nested |
2698 | invocation of this macro. */ | |
2699 | orig_flags = token->flags & (PREV_WHITE | BOL); | |
2700 | args = 0; | |
2701 | if (node->value.expansion->paramc >= 0) | |
c5a04734 | 2702 | { |
041c3194 ZW |
2703 | unsigned int error, prev_nme; |
2704 | ||
2705 | /* Allocate room for the argument contexts, and parse them. */ | |
2706 | args = (macro_args *) xmalloc (sizeof (macro_args)); | |
2707 | args->ends = (unsigned int *) | |
2708 | xmalloc (node->value.expansion->paramc * sizeof (unsigned int)); | |
2709 | args->tokens = 0; | |
2710 | args->capacity = 0; | |
2711 | args->used = 0; | |
2712 | args->level = pfile->cur_context; | |
2713 | ||
2714 | prev_nme = prevent_macro_expansion (pfile); | |
2715 | pfile->args = args; | |
2716 | error = parse_args (pfile, node, args); | |
2717 | pfile->args = 0; | |
2718 | restore_macro_expansion (pfile, prev_nme); | |
2719 | if (error) | |
2720 | { | |
2721 | free_macro_args (args); | |
2722 | return token; | |
2723 | } | |
c5a04734 | 2724 | } |
c5a04734 | 2725 | |
041c3194 ZW |
2726 | /* Now push its context. */ |
2727 | pfile->cur_context++; | |
2728 | if (pfile->cur_context == pfile->context_cap) | |
2729 | expand_context_stack (pfile); | |
2730 | ||
2731 | context = CURRENT_CONTEXT (pfile); | |
2732 | context->u.list = node->value.expansion; | |
2733 | context->args = args; | |
2734 | context->posn = 0; | |
2735 | context->count = context->u.list->tokens_used; | |
2736 | context->level = pfile->cur_context; | |
2737 | context->flags = 0; | |
2738 | context->pushed_token = 0; | |
2739 | ||
2740 | /* Set the flags of the first token. We know there must | |
2741 | be one, empty macros are a single placemarker token. */ | |
2742 | MODIFY_FLAGS_AND_POS (&context->u.list->tokens[0], token, orig_flags); | |
2743 | ||
2744 | return cpp_get_token (pfile); | |
c5a04734 ZW |
2745 | } |
2746 | ||
041c3194 ZW |
2747 | /* Push an argument to the current macro onto the context stack. |
2748 | TOKEN is the MACRO_ARG token representing the argument expansion. */ | |
2749 | static const cpp_token * | |
2750 | push_arg_context (pfile, token) | |
2751 | cpp_reader *pfile; | |
2752 | const cpp_token *token; | |
c5a04734 | 2753 | { |
041c3194 ZW |
2754 | cpp_context *context; |
2755 | macro_args *args; | |
2756 | ||
2757 | pfile->cur_context++; | |
2758 | if (pfile->cur_context == pfile->context_cap) | |
2759 | expand_context_stack (pfile); | |
2760 | ||
2761 | context = CURRENT_CONTEXT (pfile); | |
2762 | args = context[-1].args; | |
2763 | ||
2764 | context->count = token->val.aux ? args->ends[token->val.aux - 1]: 0; | |
2765 | context->u.arg = args->tokens + context->count; | |
2766 | context->count = args->ends[token->val.aux] - context->count; | |
2767 | context->args = 0; | |
2768 | context->posn = 0; | |
2769 | context->level = args->level; | |
2770 | context->flags = CONTEXT_ARG | CONTEXT_RAW; | |
2771 | context->pushed_token = 0; | |
2772 | ||
2773 | /* Set the flags of the first token. There is one. */ | |
2774 | { | |
2775 | const cpp_token *first = context->u.arg[0]; | |
2776 | if (!first) | |
2777 | first = context->u.arg[1]; | |
c5a04734 | 2778 | |
041c3194 ZW |
2779 | MODIFY_FLAGS_AND_POS ((cpp_token *) first, token, |
2780 | token->flags & (PREV_WHITE | BOL)); | |
2781 | } | |
c5a04734 | 2782 | |
041c3194 ZW |
2783 | if (token->flags & STRINGIFY_ARG) |
2784 | return stringify_arg (pfile, token); | |
c5a04734 | 2785 | |
041c3194 ZW |
2786 | if (token->flags & PASTE_LEFT) |
2787 | context->flags |= CONTEXT_PASTEL; | |
2788 | if (pfile->paste_level) | |
2789 | context->flags |= CONTEXT_PASTER; | |
d1d9a6bd | 2790 | |
041c3194 | 2791 | return get_raw_token (pfile); |
c5a04734 ZW |
2792 | } |
2793 | ||
041c3194 ZW |
2794 | /* "Unget" a token. It is effectively inserted in the token queue and |
2795 | will be returned by the next call to get_raw_token. */ | |
c5a04734 | 2796 | void |
041c3194 | 2797 | _cpp_push_token (pfile, token) |
c5a04734 | 2798 | cpp_reader *pfile; |
041c3194 | 2799 | const cpp_token *token; |
c5a04734 | 2800 | { |
041c3194 ZW |
2801 | cpp_context *context = CURRENT_CONTEXT (pfile); |
2802 | if (context->pushed_token) | |
2803 | cpp_ice (pfile, "two tokens pushed in a row"); | |
2804 | if (token->type != CPP_EOF) | |
2805 | context->pushed_token = token; | |
2806 | /* Don't push back a directive's CPP_EOF, step back instead. */ | |
2807 | else if (pfile->cur_context == 0) | |
2808 | pfile->contexts[0].posn--; | |
2809 | } | |
c5a04734 | 2810 | |
041c3194 ZW |
2811 | /* Handle a preprocessing directive. TOKEN is the CPP_HASH token |
2812 | introducing the directive. */ | |
2813 | static void | |
2814 | process_directive (pfile, token) | |
2815 | cpp_reader *pfile; | |
2816 | const cpp_token *token; | |
2817 | { | |
2818 | const struct directive *d = pfile->token_list.directive; | |
2819 | int prev_nme = 0; | |
2820 | ||
2821 | /* Skip over the directive name. */ | |
2822 | if (token[1].type == CPP_NAME) | |
2823 | _cpp_get_raw_token (pfile); | |
2824 | else if (token[1].type != CPP_NUMBER) | |
2825 | cpp_ice (pfile, "directive begins with %s?!", | |
2826 | token_names[token[1].type]); | |
2827 | ||
2828 | /* Flush pending tokens at this point, in case the directive produces | |
2829 | output. XXX Directive output won't be visible to a direct caller of | |
2830 | cpp_get_token. */ | |
2831 | if (pfile->printer && CPP_WRITTEN (pfile) - pfile->printer->written) | |
2832 | cpp_output_tokens (pfile, pfile->printer, pfile->token_list.line); | |
2833 | ||
2834 | if (! (d->flags & EXPAND)) | |
2835 | prev_nme = prevent_macro_expansion (pfile); | |
2836 | (void) (*d->handler) (pfile); | |
2837 | if (! (d->flags & EXPAND)) | |
2838 | restore_macro_expansion (pfile, prev_nme); | |
2839 | _cpp_skip_rest_of_line (pfile); | |
2840 | } | |
c5a04734 | 2841 | |
041c3194 ZW |
2842 | /* The external interface to return the next token. All macro |
2843 | expansion and directive processing is handled internally, the | |
2844 | caller only ever sees the output after preprocessing. */ | |
2845 | const cpp_token * | |
2846 | cpp_get_token (pfile) | |
2847 | cpp_reader *pfile; | |
2848 | { | |
2849 | const cpp_token *token; | |
2850 | cpp_hashnode *node; | |
6ab3e7dd | 2851 | |
041c3194 ZW |
2852 | /* Loop till we hit a non-directive, non-skipped, non-placemarker token. */ |
2853 | for (;;) | |
2854 | { | |
2855 | token = get_raw_token (pfile); | |
2856 | if (token->flags & BOL && token->type == CPP_HASH | |
2857 | && pfile->token_list.directive) | |
c5a04734 | 2858 | { |
041c3194 ZW |
2859 | process_directive (pfile, token); |
2860 | continue; | |
c5a04734 ZW |
2861 | } |
2862 | ||
041c3194 ZW |
2863 | /* Short circuit EOF. */ |
2864 | if (token->type == CPP_EOF) | |
2865 | return token; | |
2866 | ||
2867 | if (pfile->skipping && ! pfile->token_list.directive) | |
c5a04734 | 2868 | { |
041c3194 ZW |
2869 | _cpp_skip_rest_of_line (pfile); |
2870 | continue; | |
2871 | } | |
2872 | break; | |
2873 | } | |
c5a04734 | 2874 | |
041c3194 ZW |
2875 | /* If there's a potential control macro and we get here, then that |
2876 | #ifndef didn't cover the entire file and its argument shouldn't | |
2877 | be taken as a control macro. */ | |
2878 | pfile->potential_control_macro = 0; | |
c5a04734 | 2879 | |
041c3194 | 2880 | token = maybe_paste_with_next (pfile, token); |
c5a04734 | 2881 | |
041c3194 ZW |
2882 | if (token->type != CPP_NAME) |
2883 | return token; | |
c5a04734 | 2884 | |
041c3194 ZW |
2885 | /* Is macro expansion disabled in general? */ |
2886 | if (pfile->no_expand_level == pfile->cur_context || pfile->paste_level) | |
2887 | return token; | |
2888 | ||
bfb9dc7f | 2889 | node = token->val.node; |
041c3194 ZW |
2890 | if (node->type == T_VOID) |
2891 | return token; | |
c5a04734 | 2892 | |
041c3194 ZW |
2893 | if (node->type == T_MACRO) |
2894 | { | |
2895 | if (is_macro_disabled (pfile, node->value.expansion, token)) | |
2896 | return token; | |
c5a04734 | 2897 | |
041c3194 ZW |
2898 | return push_macro_context (pfile, node, token); |
2899 | } | |
2900 | else | |
2901 | return special_symbol (pfile, node, token); | |
2902 | } | |
c5a04734 | 2903 | |
041c3194 ZW |
2904 | /* Returns the next raw token, i.e. without performing macro |
2905 | expansion. Argument contexts are automatically entered. */ | |
2906 | static const cpp_token * | |
2907 | get_raw_token (pfile) | |
2908 | cpp_reader *pfile; | |
2909 | { | |
2910 | const cpp_token *result; | |
2911 | cpp_context *context = CURRENT_CONTEXT (pfile); | |
c5a04734 | 2912 | |
041c3194 ZW |
2913 | if (context->pushed_token) |
2914 | { | |
2915 | result = context->pushed_token; | |
2916 | context->pushed_token = 0; | |
2917 | } | |
2918 | else if (context->posn == context->count) | |
2919 | result = pop_context (pfile); | |
2920 | else | |
2921 | { | |
2922 | if (IS_ARG_CONTEXT (context)) | |
2923 | { | |
2924 | result = context->u.arg[context->posn++]; | |
2925 | if (result == 0) | |
c5a04734 | 2926 | { |
041c3194 ZW |
2927 | context->flags ^= CONTEXT_RAW; |
2928 | result = context->u.arg[context->posn++]; | |
c5a04734 | 2929 | } |
041c3194 ZW |
2930 | return result; /* Cannot be a CPP_MACRO_ARG */ |
2931 | } | |
2932 | result = &context->u.list->tokens[context->posn++]; | |
2933 | } | |
c5a04734 | 2934 | |
041c3194 ZW |
2935 | if (result->type == CPP_MACRO_ARG) |
2936 | result = push_arg_context (pfile, result); | |
2937 | return result; | |
2938 | } | |
c5a04734 | 2939 | |
041c3194 ZW |
2940 | /* Internal interface to get the token without macro expanding. */ |
2941 | const cpp_token * | |
2942 | _cpp_get_raw_token (pfile) | |
2943 | cpp_reader *pfile; | |
2944 | { | |
2945 | int prev_nme = prevent_macro_expansion (pfile); | |
2946 | const cpp_token *result = cpp_get_token (pfile); | |
2947 | restore_macro_expansion (pfile, prev_nme); | |
2948 | return result; | |
2949 | } | |
c5a04734 | 2950 | |
041c3194 ZW |
2951 | /* A thin wrapper to lex_line. CLEAR is non-zero if the current token |
2952 | list should be overwritten, or zero if we need to append | |
2953 | (typically, if we are within the arguments to a macro, or looking | |
2954 | for the '(' to start a function-like macro invocation). */ | |
2955 | static int | |
2956 | lex_next (pfile, clear) | |
2957 | cpp_reader *pfile; | |
2958 | int clear; | |
2959 | { | |
2960 | cpp_toklist *list = &pfile->token_list; | |
2961 | const cpp_token *old_list = list->tokens; | |
2962 | unsigned int old_used = list->tokens_used; | |
c5a04734 | 2963 | |
91fcd158 NB |
2964 | /* If we are currently processing a directive, do not advance. 6.10 |
2965 | paragraph 2: A new-line character ends the directive even if it | |
2966 | occurs within what would otherwise be an invocation of a | |
2967 | function-like macro. | |
2968 | ||
2969 | It is possible that clear == 1 too; e.g. "#if funlike_macro (" | |
2970 | since parse_args swallowed the directive's EOF. */ | |
2971 | if (list->directive) | |
2972 | return 1; | |
2973 | ||
041c3194 | 2974 | if (clear) |
c5a04734 | 2975 | { |
041c3194 ZW |
2976 | /* Release all temporary tokens. */ |
2977 | _cpp_clear_toklist (list); | |
2978 | pfile->contexts[0].posn = 0; | |
2979 | if (pfile->temp_used) | |
2980 | release_temp_tokens (pfile); | |
c5a04734 | 2981 | } |
041c3194 ZW |
2982 | |
2983 | lex_line (pfile, list); | |
2984 | pfile->contexts[0].count = list->tokens_used; | |
c5a04734 | 2985 | |
041c3194 | 2986 | if (!clear && pfile->args) |
c5a04734 | 2987 | { |
041c3194 ZW |
2988 | /* Fix up argument token pointers. */ |
2989 | if (old_list != list->tokens) | |
2990 | { | |
2991 | unsigned int i; | |
2992 | ||
2993 | for (i = 0; i < pfile->args->used; i++) | |
2994 | { | |
2995 | const cpp_token *token = pfile->args->tokens[i]; | |
2996 | if (token >= old_list && token < old_list + old_used) | |
2997 | pfile->args->tokens[i] = (const cpp_token *) | |
2998 | ((char *) token + ((char *) list->tokens - (char *) old_list)); | |
2999 | } | |
3000 | } | |
3001 | ||
3002 | /* 6.10.3 paragraph 11: If there are sequences of preprocessing | |
3003 | tokens within the list of arguments that would otherwise act as | |
3004 | preprocessing directives, the behavior is undefined. | |
3005 | ||
3006 | This implementation will report a hard error and treat the | |
3007 | 'sequence of preprocessing tokens' as part of the macro argument, | |
3008 | not a directive. | |
3009 | ||
3010 | Note if pfile->args == 0, we're OK since we're only inside a | |
3011 | macro argument after a '('. */ | |
3012 | if (list->directive) | |
3013 | { | |
3014 | cpp_error_with_line (pfile, list->tokens[old_used].line, | |
3015 | list->tokens[old_used].col, | |
3016 | "#%s may not be used inside a macro argument", | |
3017 | list->directive->name); | |
91fcd158 | 3018 | return 1; |
041c3194 | 3019 | } |
c5a04734 ZW |
3020 | } |
3021 | ||
041c3194 | 3022 | return 0; |
c5a04734 ZW |
3023 | } |
3024 | ||
041c3194 ZW |
3025 | /* Pops a context of the context stack. If we're at the bottom, lexes |
3026 | the next logical line. Returns 1 if we're at the end of the | |
3027 | argument list to the # operator, or if it is illegal to "overflow" | |
3028 | into the rest of the file (e.g. 6.10.3.1.1). */ | |
3029 | static int | |
3030 | do_pop_context (pfile) | |
3031 | cpp_reader *pfile; | |
3032 | { | |
3033 | cpp_context *context; | |
3fef5b2b | 3034 | |
041c3194 ZW |
3035 | if (pfile->cur_context == 0) |
3036 | return lex_next (pfile, pfile->no_expand_level == UINT_MAX); | |
3037 | ||
3038 | /* Argument contexts, when parsing args or handling # operator | |
3039 | return CPP_EOF at the end. */ | |
3040 | context = CURRENT_CONTEXT (pfile); | |
3041 | if (IS_ARG_CONTEXT (context) && pfile->cur_context == pfile->no_expand_level) | |
3042 | return 1; | |
3043 | ||
3044 | /* Free resources when leaving macro contexts. */ | |
3045 | if (context->args) | |
3046 | free_macro_args (context->args); | |
3047 | ||
3048 | if (pfile->cur_context == pfile->no_expand_level) | |
3049 | pfile->no_expand_level--; | |
3050 | pfile->cur_context--; | |
3051 | ||
3052 | return 0; | |
3053 | } | |
3054 | ||
3055 | /* Move down the context stack, and return the next raw token. */ | |
3056 | static const cpp_token * | |
3057 | pop_context (pfile) | |
3058 | cpp_reader *pfile; | |
3059 | { | |
3060 | if (do_pop_context (pfile)) | |
3061 | return &eof_token; | |
3062 | return get_raw_token (pfile); | |
3063 | } | |
3064 | ||
3065 | /* Turn off macro expansion at the current context level. */ | |
3066 | static unsigned int | |
3067 | prevent_macro_expansion (pfile) | |
3068 | cpp_reader *pfile; | |
3069 | { | |
3070 | unsigned int prev_value = pfile->no_expand_level; | |
3071 | pfile->no_expand_level = pfile->cur_context; | |
3072 | return prev_value; | |
3073 | } | |
3074 | ||
3075 | /* Restore macro expansion to its previous state. */ | |
3076 | static void | |
3077 | restore_macro_expansion (pfile, prev_value) | |
3078 | cpp_reader *pfile; | |
3079 | unsigned int prev_value; | |
3080 | { | |
3081 | pfile->no_expand_level = prev_value; | |
3082 | } | |
3083 | ||
3084 | /* Used by cpperror.c to obtain the correct line and column to report | |
3085 | in a diagnostic. */ | |
3086 | unsigned int | |
3087 | _cpp_get_line (pfile, pcol) | |
3088 | cpp_reader *pfile; | |
3089 | unsigned int *pcol; | |
3090 | { | |
3091 | unsigned int index; | |
3092 | const cpp_token *cur_token; | |
3093 | ||
3094 | if (pfile->in_lex_line) | |
3095 | index = pfile->token_list.tokens_used; | |
3096 | else | |
3097 | index = pfile->contexts[0].posn; | |
3098 | ||
3099 | cur_token = &pfile->token_list.tokens[index - 1]; | |
3100 | if (pcol) | |
3101 | *pcol = cur_token->col; | |
3102 | return cur_token->line; | |
3103 | } | |
3104 | ||
3105 | #define DSC(str) (const U_CHAR *)str, sizeof str - 1 | |
3106 | static const char * const monthnames[] = | |
3107 | { | |
3108 | "Jan", "Feb", "Mar", "Apr", "May", "Jun", | |
3109 | "Jul", "Aug", "Sep", "Oct", "Nov", "Dec", | |
3110 | }; | |
3111 | ||
3112 | /* Handle builtin macros like __FILE__. */ | |
3113 | static const cpp_token * | |
3114 | special_symbol (pfile, node, token) | |
3115 | cpp_reader *pfile; | |
3116 | cpp_hashnode *node; | |
d1d9a6bd | 3117 | const cpp_token *token; |
3fef5b2b | 3118 | { |
041c3194 ZW |
3119 | cpp_token *result; |
3120 | cpp_buffer *ip; | |
3fef5b2b | 3121 | |
041c3194 | 3122 | switch (node->type) |
3fef5b2b | 3123 | { |
041c3194 ZW |
3124 | case T_FILE: |
3125 | case T_BASE_FILE: | |
3fef5b2b | 3126 | { |
041c3194 | 3127 | const char *file; |
3fef5b2b | 3128 | |
041c3194 ZW |
3129 | ip = CPP_BUFFER (pfile); |
3130 | if (ip == 0) | |
3131 | file = ""; | |
3fef5b2b | 3132 | else |
041c3194 ZW |
3133 | { |
3134 | if (node->type == T_BASE_FILE) | |
3135 | while (CPP_PREV_BUFFER (ip) != NULL) | |
3136 | ip = CPP_PREV_BUFFER (ip); | |
3137 | ||
3138 | file = ip->nominal_fname; | |
3139 | } | |
3140 | result = make_string_token (get_temp_token (pfile), (U_CHAR *) file, | |
3141 | strlen (file)); | |
3142 | } | |
3143 | break; | |
3fef5b2b | 3144 | |
041c3194 ZW |
3145 | case T_INCLUDE_LEVEL: |
3146 | { | |
3147 | int true_indepth = 0; | |
3148 | ||
3149 | /* Do not count the primary source file in the include level. */ | |
3150 | ip = CPP_PREV_BUFFER (CPP_BUFFER (pfile)); | |
3151 | while (ip) | |
3152 | { | |
3153 | true_indepth++; | |
3154 | ip = CPP_PREV_BUFFER (ip); | |
3155 | } | |
3156 | result = alloc_number_token (pfile, true_indepth); | |
3fef5b2b NB |
3157 | } |
3158 | break; | |
3159 | ||
041c3194 ZW |
3160 | case T_SPECLINE: |
3161 | /* If __LINE__ is embedded in a macro, it must expand to the | |
3162 | line of the macro's invocation, not its definition. | |
3163 | Otherwise things like assert() will not work properly. */ | |
3164 | result = alloc_number_token (pfile, _cpp_get_line (pfile, NULL)); | |
5d7ee2fa NB |
3165 | break; |
3166 | ||
041c3194 | 3167 | case T_STDC: |
3fef5b2b | 3168 | { |
041c3194 | 3169 | int stdc = 1; |
3fef5b2b | 3170 | |
041c3194 ZW |
3171 | #ifdef STDC_0_IN_SYSTEM_HEADERS |
3172 | if (CPP_IN_SYSTEM_HEADER (pfile) | |
bfb9dc7f | 3173 | && pfile->spec_nodes->n__STRICT_ANSI__->type == T_VOID) |
041c3194 ZW |
3174 | stdc = 0; |
3175 | #endif | |
3176 | result = alloc_number_token (pfile, stdc); | |
3fef5b2b NB |
3177 | } |
3178 | break; | |
3179 | ||
041c3194 ZW |
3180 | case T_DATE: |
3181 | case T_TIME: | |
3182 | if (pfile->date == 0) | |
3183 | { | |
3184 | /* Allocate __DATE__ and __TIME__ from permanent storage, | |
3185 | and save them in pfile so we don't have to do this again. | |
3186 | We don't generate these strings at init time because | |
3187 | time() and localtime() are very slow on some systems. */ | |
3188 | time_t tt = time (NULL); | |
3189 | struct tm *tb = localtime (&tt); | |
3190 | ||
3191 | pfile->date = make_string_token | |
3192 | ((cpp_token *) xmalloc (sizeof (cpp_token)), DSC("Oct 11 1347")); | |
3193 | pfile->time = make_string_token | |
3194 | ((cpp_token *) xmalloc (sizeof (cpp_token)), DSC("12:34:56")); | |
3195 | ||
bfb9dc7f | 3196 | sprintf ((char *) pfile->date->val.str.text, "%s %2d %4d", |
041c3194 | 3197 | monthnames[tb->tm_mon], tb->tm_mday, tb->tm_year + 1900); |
bfb9dc7f | 3198 | sprintf ((char *) pfile->time->val.str.text, "%02d:%02d:%02d", |
041c3194 ZW |
3199 | tb->tm_hour, tb->tm_min, tb->tm_sec); |
3200 | } | |
3201 | result = node->type == T_DATE ? pfile->date: pfile->time; | |
3fef5b2b NB |
3202 | break; |
3203 | ||
041c3194 | 3204 | case T_POISON: |
bfb9dc7f | 3205 | cpp_error (pfile, "attempt to use poisoned \"%s\"", node->name); |
041c3194 ZW |
3206 | return token; |
3207 | ||
3208 | default: | |
3209 | cpp_ice (pfile, "invalid special hash type"); | |
3210 | return token; | |
3fef5b2b NB |
3211 | } |
3212 | ||
041c3194 ZW |
3213 | ASSIGN_FLAGS_AND_POS (result, token); |
3214 | return result; | |
3215 | } | |
3216 | #undef DSC | |
3217 | ||
3218 | /* Dump the original user's spelling of argument index ARG_NO to the | |
3219 | macro whose expansion is LIST. */ | |
3220 | static void | |
3221 | dump_param_spelling (pfile, list, arg_no) | |
3222 | cpp_reader *pfile; | |
3223 | const cpp_toklist *list; | |
3224 | unsigned int arg_no; | |
3225 | { | |
3226 | const U_CHAR *param = list->namebuf; | |
3227 | ||
3228 | while (arg_no--) | |
3229 | param += ustrlen (param) + 1; | |
3230 | CPP_PUTS (pfile, param, ustrlen (param)); | |
3fef5b2b NB |
3231 | } |
3232 | ||
041c3194 | 3233 | /* Dump a token list to the output. */ |
c5a04734 | 3234 | void |
041c3194 ZW |
3235 | _cpp_dump_list (pfile, list, token, flush) |
3236 | cpp_reader *pfile; | |
3237 | const cpp_toklist *list; | |
3238 | const cpp_token *token; | |
3239 | int flush; | |
c5a04734 | 3240 | { |
041c3194 ZW |
3241 | const cpp_token *limit = list->tokens + list->tokens_used; |
3242 | const cpp_token *prev = 0; | |
c5a04734 | 3243 | |
041c3194 ZW |
3244 | /* Avoid the CPP_EOF. */ |
3245 | if (list->directive) | |
3246 | limit--; | |
c5a04734 | 3247 | |
041c3194 | 3248 | while (token < limit) |
c5a04734 | 3249 | { |
041c3194 ZW |
3250 | if (token->type == CPP_MACRO_ARG) |
3251 | { | |
3252 | if (token->flags & PREV_WHITE) | |
3253 | CPP_PUTC (pfile, ' '); | |
3254 | if (token->flags & STRINGIFY_ARG) | |
3255 | CPP_PUTC (pfile, '#'); | |
3256 | dump_param_spelling (pfile, list, token->val.aux); | |
3257 | } | |
c5a04734 | 3258 | else |
041c3194 ZW |
3259 | output_token (pfile, token, prev); |
3260 | if (token->flags & PASTE_LEFT) | |
3261 | CPP_PUTS (pfile, " ##", 3); | |
3262 | prev = token; | |
3263 | token++; | |
c5a04734 | 3264 | } |
041c3194 ZW |
3265 | |
3266 | if (flush && pfile->printer) | |
3267 | cpp_output_tokens (pfile, pfile->printer, pfile->token_list.line); | |
c5a04734 ZW |
3268 | } |
3269 | ||
041c3194 ZW |
3270 | /* Allocate pfile->input_buffer, and initialize trigraph_map[] |
3271 | if it hasn't happened already. */ | |
3272 | ||
3273 | void | |
3274 | _cpp_init_input_buffer (pfile) | |
3275 | cpp_reader *pfile; | |
3276 | { | |
3277 | init_trigraph_map (); | |
3278 | pfile->context_cap = 20; | |
3279 | pfile->contexts = (cpp_context *) | |
3280 | xmalloc (pfile->context_cap * sizeof (cpp_context)); | |
3281 | pfile->cur_context = 0; | |
3282 | pfile->contexts[0].u.list = &pfile->token_list; | |
3283 | ||
3284 | pfile->contexts[0].posn = 0; | |
3285 | pfile->contexts[0].count = 0; | |
3286 | pfile->no_expand_level = UINT_MAX; | |
3287 | ||
3288 | _cpp_init_toklist (&pfile->token_list, DUMMY_TOKEN); | |
3289 | } | |
3290 | ||
3291 | /* Moves to the end of the directive line, popping contexts as | |
3292 | necessary. */ | |
3293 | void | |
3294 | _cpp_skip_rest_of_line (pfile) | |
3295 | cpp_reader *pfile; | |
3296 | { | |
3297 | /* Get to base context. Clear parsing args and each contexts flags, | |
3298 | since these can cause pop_context to return without popping. */ | |
3299 | pfile->no_expand_level = UINT_MAX; | |
3300 | while (pfile->cur_context != 0) | |
c5a04734 | 3301 | { |
041c3194 ZW |
3302 | pfile->contexts[pfile->cur_context].flags = 0; |
3303 | do_pop_context (pfile); | |
c5a04734 | 3304 | } |
041c3194 ZW |
3305 | |
3306 | pfile->contexts[pfile->cur_context].count = 0; | |
3307 | pfile->contexts[pfile->cur_context].posn = 0; | |
3308 | pfile->token_list.directive = 0; | |
c5a04734 ZW |
3309 | } |
3310 | ||
041c3194 ZW |
3311 | /* Directive handler wrapper used by the command line option |
3312 | processor. */ | |
3313 | void | |
3314 | _cpp_run_directive (pfile, dir, buf, count) | |
3315 | cpp_reader *pfile; | |
3316 | const struct directive *dir; | |
3317 | const char *buf; | |
3318 | size_t count; | |
3319 | { | |
3320 | if (cpp_push_buffer (pfile, (const U_CHAR *)buf, count) != NULL) | |
3321 | { | |
3322 | unsigned int prev_lvl = 0; | |
3323 | /* scan the line now, else prevent_macro_expansion won't work */ | |
3324 | do_pop_context (pfile); | |
3325 | if (! (dir->flags & EXPAND)) | |
3326 | prev_lvl = prevent_macro_expansion (pfile); | |
3327 | ||
3328 | (void) (*dir->handler) (pfile); | |
3329 | ||
3330 | if (! (dir->flags & EXPAND)) | |
3331 | restore_macro_expansion (pfile, prev_lvl); | |
3332 | ||
3333 | _cpp_skip_rest_of_line (pfile); | |
3334 | cpp_pop_buffer (pfile); | |
3335 | } | |
3336 | } |