]>
Commit | Line | Data |
---|---|---|
6de1e2a9 | 1 | /* Part of CPP library. (Macro handling.) |
4283012f | 2 | Copyright (C) 1986, 87, 89, 92-95, 1996, 1998 Free Software Foundation, Inc. |
7f2935c7 | 3 | Written by Per Bothner, 1994. |
38e01259 | 4 | Based on CCCP program by Paul Rubin, June 1986 |
7f2935c7 PB |
5 | Adapted to ANSI C, Richard Stallman, Jan 1987 |
6 | ||
7 | This program is free software; you can redistribute it and/or modify it | |
8 | under the terms of the GNU General Public License as published by the | |
9 | Free Software Foundation; either version 2, or (at your option) any | |
10 | later version. | |
11 | ||
12 | This program is distributed in the hope that it will be useful, | |
13 | but WITHOUT ANY WARRANTY; without even the implied warranty of | |
14 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
15 | GNU General Public License for more details. | |
16 | ||
17 | You should have received a copy of the GNU General Public License | |
18 | along with this program; if not, write to the Free Software | |
940d9d63 | 19 | Foundation, 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. |
7f2935c7 PB |
20 | |
21 | In other words, you are welcome to use, share and improve this program. | |
22 | You are forbidden to forbid anyone else to use, share and improve | |
23 | what you give them. Help stamp out software-hoarding! */ | |
24 | ||
b04cd507 KG |
25 | #include "config.h" |
26 | #include "system.h" | |
7f2935c7 PB |
27 | #include "cpplib.h" |
28 | #include "cpphash.h" | |
34ca9541 | 29 | #undef abort |
7f2935c7 | 30 | |
6de1e2a9 ZW |
31 | static int comp_def_part PARAMS ((int, U_CHAR *, int, U_CHAR *, |
32 | int, int)); | |
33 | static int change_newlines PARAMS ((U_CHAR *, int)); | |
34 | static void push_macro_expansion PARAMS ((cpp_reader *, | |
35 | U_CHAR *, int, HASHNODE *)); | |
36 | static int unsafe_chars PARAMS ((int, int)); | |
37 | ||
38 | #define SKIP_WHITE_SPACE(p) do { while (is_hor_space[*p]) p++; } while (0) | |
39 | #define CPP_IS_MACRO_BUFFER(PBUF) ((PBUF)->data != NULL) | |
40 | #define FORWARD(N) CPP_FORWARD (CPP_BUFFER (pfile), (N)) | |
41 | ||
42 | extern char *version_string; | |
43 | ||
44 | /* The arglist structure is built by create_definition to tell | |
45 | collect_expansion where the argument names begin. That | |
46 | is, for a define like "#define f(x,y,z) foo+x-bar*y", the arglist | |
47 | would contain pointers to the strings x, y, and z. | |
48 | collect_expansion would then build a DEFINITION node, | |
49 | with reflist nodes pointing to the places x, y, and z had | |
50 | appeared. So the arglist is just convenience data passed | |
51 | between these two routines. It is not kept around after | |
52 | the current #define has been processed and entered into the | |
53 | hash table. */ | |
54 | ||
55 | struct arglist | |
56 | { | |
57 | struct arglist *next; | |
58 | U_CHAR *name; | |
59 | int length; | |
60 | int argno; | |
61 | char rest_args; | |
62 | }; | |
63 | ||
64 | /* This structure represents one parsed argument in a macro call. | |
65 | `raw' points to the argument text as written (`raw_length' is its length). | |
66 | `expanded' points to the argument's macro-expansion | |
67 | (its length is `expand_length'). | |
68 | `stringified_length' is the length the argument would have | |
69 | if stringified. | |
70 | `use_count' is the number of times this macro arg is substituted | |
71 | into the macro. If the actual use count exceeds 10, | |
72 | the value stored is 10. */ | |
73 | ||
74 | /* raw and expanded are relative to ARG_BASE */ | |
75 | #define ARG_BASE ((pfile)->token_buffer) | |
76 | ||
77 | struct argdata | |
78 | { | |
79 | /* Strings relative to pfile->token_buffer */ | |
80 | long raw, expanded, stringified; | |
81 | int raw_length, expand_length; | |
82 | int stringified_length; | |
83 | char newlines; | |
84 | char use_count; | |
85 | }; | |
86 | ||
87 | ||
0f41302f MS |
88 | /* Return hash function on name. must be compatible with the one |
89 | computed a step at a time, elsewhere */ | |
90 | ||
7f2935c7 PB |
91 | int |
92 | hashf (name, len, hashsize) | |
93 | register const U_CHAR *name; | |
94 | register int len; | |
95 | int hashsize; | |
96 | { | |
97 | register int r = 0; | |
98 | ||
99 | while (len--) | |
100 | r = HASHSTEP (r, *name++); | |
101 | ||
102 | return MAKE_POS (r) % hashsize; | |
103 | } | |
104 | ||
38e01259 | 105 | /* Find the most recent hash node for name "name" (ending with first |
122ae89b | 106 | non-identifier char) installed by cpp_install |
0f41302f MS |
107 | |
108 | If LEN is >= 0, it is the length of the name. | |
109 | Otherwise, compute the length by scanning the entire name. | |
110 | ||
111 | If HASH is >= 0, it is the precomputed hash code. | |
112 | Otherwise, compute the hash code. */ | |
113 | ||
7f2935c7 PB |
114 | HASHNODE * |
115 | cpp_lookup (pfile, name, len, hash) | |
d6f4ec51 | 116 | cpp_reader *pfile ATTRIBUTE_UNUSED; |
7f2935c7 PB |
117 | const U_CHAR *name; |
118 | int len; | |
119 | int hash; | |
120 | { | |
121 | register const U_CHAR *bp; | |
122 | register HASHNODE *bucket; | |
123 | ||
124 | if (len < 0) | |
125 | { | |
6de1e2a9 | 126 | for (bp = name; is_idchar[*bp]; bp++); |
7f2935c7 PB |
127 | len = bp - name; |
128 | } | |
129 | ||
130 | if (hash < 0) | |
131 | hash = hashf (name, len, HASHSIZE); | |
132 | ||
122ae89b | 133 | bucket = pfile->hashtab[hash]; |
6de1e2a9 ZW |
134 | while (bucket) |
135 | { | |
136 | if (bucket->length == len && strncmp (bucket->name, name, len) == 0) | |
137 | return bucket; | |
138 | bucket = bucket->next; | |
139 | } | |
0f41302f | 140 | return (HASHNODE *) 0; |
7f2935c7 PB |
141 | } |
142 | ||
143 | /* | |
144 | * Delete a hash node. Some weirdness to free junk from macros. | |
145 | * More such weirdness will have to be added if you define more hash | |
146 | * types that need it. | |
147 | */ | |
148 | ||
149 | /* Note that the DEFINITION of a macro is removed from the hash table | |
150 | but its storage is not freed. This would be a storage leak | |
151 | except that it is not reasonable to keep undefining and redefining | |
152 | large numbers of macros many times. | |
153 | In any case, this is necessary, because a macro can be #undef'd | |
154 | in the middle of reading the arguments to a call to it. | |
155 | If #undef freed the DEFINITION, that would crash. */ | |
156 | ||
157 | void | |
158 | delete_macro (hp) | |
159 | HASHNODE *hp; | |
160 | { | |
161 | ||
162 | if (hp->prev != NULL) | |
163 | hp->prev->next = hp->next; | |
164 | if (hp->next != NULL) | |
165 | hp->next->prev = hp->prev; | |
166 | ||
167 | /* make sure that the bucket chain header that | |
0f41302f | 168 | the deleted guy was on points to the right thing afterwards. */ |
7f2935c7 PB |
169 | if (hp == *hp->bucket_hdr) |
170 | *hp->bucket_hdr = hp->next; | |
171 | ||
896fc322 PB |
172 | if (hp->type == T_MACRO) |
173 | { | |
174 | DEFINITION *d = hp->value.defn; | |
175 | struct reflist *ap, *nextap; | |
176 | ||
177 | for (ap = d->pattern; ap != NULL; ap = nextap) | |
178 | { | |
179 | nextap = ap->next; | |
180 | free (ap); | |
181 | } | |
182 | if (d->nargs >= 0) | |
183 | free (d->args.argnames); | |
184 | free (d); | |
7f2935c7 | 185 | } |
896fc322 | 186 | |
7f2935c7 PB |
187 | free (hp); |
188 | } | |
0f41302f MS |
189 | |
190 | /* Install a name in the main hash table, even if it is already there. | |
5dfa4da1 ZW |
191 | Name stops with first non alphanumeric, except leading '#'. |
192 | Caller must check against redefinition if that is desired. | |
122ae89b | 193 | delete_macro () removes things installed by cpp_install () in fifo order. |
0f41302f MS |
194 | this is important because of the `defined' special symbol used |
195 | in #if, and also if pushdef/popdef directives are ever implemented. | |
196 | ||
197 | If LEN is >= 0, it is the length of the name. | |
198 | Otherwise, compute the length by scanning the entire name. | |
199 | ||
200 | If HASH is >= 0, it is the precomputed hash code. | |
201 | Otherwise, compute the hash code. */ | |
202 | ||
7f2935c7 | 203 | HASHNODE * |
122ae89b ZW |
204 | cpp_install (pfile, name, len, type, value, hash) |
205 | cpp_reader *pfile; | |
7f2935c7 PB |
206 | U_CHAR *name; |
207 | int len; | |
208 | enum node_type type; | |
5dfa4da1 | 209 | const char *value; |
7f2935c7 PB |
210 | int hash; |
211 | { | |
212 | register HASHNODE *hp; | |
213 | register int i, bucket; | |
7061aa5a | 214 | register U_CHAR *p; |
7f2935c7 | 215 | |
6de1e2a9 ZW |
216 | if (len < 0) |
217 | { | |
218 | p = name; | |
219 | while (is_idchar[*p]) | |
220 | p++; | |
221 | len = p - name; | |
222 | } | |
7f2935c7 PB |
223 | |
224 | if (hash < 0) | |
225 | hash = hashf (name, len, HASHSIZE); | |
226 | ||
227 | i = sizeof (HASHNODE) + len + 1; | |
228 | hp = (HASHNODE *) xmalloc (i); | |
229 | bucket = hash; | |
122ae89b ZW |
230 | hp->bucket_hdr = &pfile->hashtab[bucket]; |
231 | hp->next = pfile->hashtab[bucket]; | |
232 | pfile->hashtab[bucket] = hp; | |
7f2935c7 PB |
233 | hp->prev = NULL; |
234 | if (hp->next != NULL) | |
235 | hp->next->prev = hp; | |
236 | hp->type = type; | |
237 | hp->length = len; | |
5dfa4da1 | 238 | hp->value.cpval = value; |
7f2935c7 | 239 | hp->name = ((U_CHAR *) hp) + sizeof (HASHNODE); |
7061aa5a | 240 | bcopy (name, hp->name, len); |
7f2935c7 PB |
241 | hp->name[len] = 0; |
242 | return hp; | |
243 | } | |
896fc322 | 244 | |
6de1e2a9 ZW |
245 | static int |
246 | macro_cleanup (pbuf, pfile) | |
247 | cpp_buffer *pbuf; | |
248 | cpp_reader *pfile ATTRIBUTE_UNUSED; | |
249 | { | |
250 | HASHNODE *macro = (HASHNODE *) pbuf->data; | |
251 | if (macro->type == T_DISABLED) | |
252 | macro->type = T_MACRO; | |
253 | if (macro->type != T_MACRO || pbuf->buf != macro->value.defn->expansion) | |
254 | free (pbuf->buf); | |
255 | return 0; | |
256 | } | |
257 | ||
258 | ||
259 | /* Read a replacement list for a macro with parameters. | |
260 | Build the DEFINITION structure. | |
261 | Reads characters of text starting at BUF until END. | |
262 | ARGLIST specifies the formal parameters to look for | |
263 | in the text of the definition; NARGS is the number of args | |
264 | in that list, or -1 for a macro name that wants no argument list. | |
265 | MACRONAME is the macro name itself (so we can avoid recursive expansion) | |
266 | and NAMELEN is its length in characters. | |
267 | ||
268 | Note that comments, backslash-newlines, and leading white space | |
269 | have already been deleted from the argument. */ | |
270 | ||
271 | static DEFINITION * | |
272 | collect_expansion (pfile, buf, limit, nargs, arglist) | |
273 | cpp_reader *pfile; | |
274 | U_CHAR *buf, *limit; | |
275 | int nargs; | |
276 | struct arglist *arglist; | |
277 | { | |
278 | DEFINITION *defn; | |
279 | register U_CHAR *p, *lastp, *exp_p; | |
280 | struct reflist *endpat = NULL; | |
281 | /* Pointer to first nonspace after last ## seen. */ | |
282 | U_CHAR *concat = 0; | |
283 | /* Pointer to first nonspace after last single-# seen. */ | |
284 | U_CHAR *stringify = 0; | |
285 | int maxsize; | |
286 | int expected_delimiter = '\0'; | |
287 | ||
288 | /* Scan thru the replacement list, ignoring comments and quoted | |
289 | strings, picking up on the macro calls. It does a linear search | |
290 | thru the arg list on every potential symbol. Profiling might say | |
291 | that something smarter should happen. */ | |
292 | ||
293 | if (limit < buf) | |
34ca9541 ZW |
294 | { |
295 | cpp_fatal (pfile, "internal error: limit < buf in collect_expansion"); | |
296 | limit = buf; /* treat it like a null defn */ | |
297 | } | |
6de1e2a9 ZW |
298 | |
299 | /* Find the beginning of the trailing whitespace. */ | |
300 | p = buf; | |
301 | while (p < limit && is_space[limit[-1]]) | |
302 | limit--; | |
303 | ||
304 | /* Allocate space for the text in the macro definition. | |
305 | Leading and trailing whitespace chars need 2 bytes each. | |
306 | Each other input char may or may not need 1 byte, | |
307 | so this is an upper bound. The extra 5 are for invented | |
308 | leading and trailing newline-marker and final null. */ | |
309 | maxsize = (sizeof (DEFINITION) | |
310 | + (limit - p) + 5); | |
6de1e2a9 ZW |
311 | defn = (DEFINITION *) xcalloc (1, maxsize); |
312 | ||
313 | defn->nargs = nargs; | |
314 | exp_p = defn->expansion = (U_CHAR *) defn + sizeof (DEFINITION); | |
315 | lastp = exp_p; | |
316 | ||
317 | p = buf; | |
318 | ||
319 | /* Add one initial space escape-marker to prevent accidental | |
320 | token-pasting (often removed by macroexpand). */ | |
ed45de98 | 321 | *exp_p++ = '\r'; |
6de1e2a9 ZW |
322 | *exp_p++ = ' '; |
323 | ||
324 | if (limit - p >= 2 && p[0] == '#' && p[1] == '#') | |
325 | { | |
326 | cpp_error (pfile, "`##' at start of macro definition"); | |
327 | p += 2; | |
328 | } | |
329 | ||
330 | /* Process the main body of the definition. */ | |
331 | while (p < limit) | |
332 | { | |
333 | int skipped_arg = 0; | |
334 | register U_CHAR c = *p++; | |
335 | ||
336 | *exp_p++ = c; | |
337 | ||
338 | if (!CPP_TRADITIONAL (pfile)) | |
339 | { | |
340 | switch (c) | |
341 | { | |
342 | case '\'': | |
343 | case '\"': | |
344 | if (expected_delimiter != '\0') | |
345 | { | |
346 | if (c == expected_delimiter) | |
347 | expected_delimiter = '\0'; | |
348 | } | |
349 | else | |
350 | expected_delimiter = c; | |
351 | break; | |
352 | ||
353 | case '\\': | |
354 | if (p < limit && expected_delimiter) | |
355 | { | |
356 | /* In a string, backslash goes through | |
357 | and makes next char ordinary. */ | |
358 | *exp_p++ = *p++; | |
359 | } | |
360 | break; | |
361 | ||
6de1e2a9 ZW |
362 | case '#': |
363 | /* # is ordinary inside a string. */ | |
364 | if (expected_delimiter) | |
365 | break; | |
366 | if (p < limit && *p == '#') | |
367 | { | |
368 | /* ##: concatenate preceding and following tokens. */ | |
369 | /* Take out the first #, discard preceding whitespace. */ | |
370 | exp_p--; | |
371 | while (exp_p > lastp && is_hor_space[exp_p[-1]]) | |
372 | --exp_p; | |
373 | /* Skip the second #. */ | |
374 | p++; | |
375 | /* Discard following whitespace. */ | |
376 | SKIP_WHITE_SPACE (p); | |
377 | concat = p; | |
378 | if (p == limit) | |
379 | cpp_error (pfile, "`##' at end of macro definition"); | |
380 | } | |
381 | else if (nargs >= 0) | |
382 | { | |
383 | /* Single #: stringify following argument ref. | |
384 | Don't leave the # in the expansion. */ | |
385 | exp_p--; | |
386 | SKIP_WHITE_SPACE (p); | |
387 | if (p == limit || !is_idstart[*p] | |
388 | || (*p == 'L' && p + 1 < limit && (p[1] == '\'' || | |
389 | p[1] == '"'))) | |
390 | cpp_error (pfile, | |
391 | "`#' operator is not followed by a macro argument name"); | |
392 | else | |
393 | stringify = p; | |
394 | } | |
395 | break; | |
396 | } | |
397 | } | |
398 | else | |
399 | { | |
400 | /* In -traditional mode, recognize arguments inside strings and | |
401 | character constants, and ignore special properties of #. | |
402 | Arguments inside strings are considered "stringified", but no | |
403 | extra quote marks are supplied. */ | |
404 | switch (c) | |
405 | { | |
406 | case '\'': | |
407 | case '\"': | |
408 | if (expected_delimiter != '\0') | |
409 | { | |
410 | if (c == expected_delimiter) | |
411 | expected_delimiter = '\0'; | |
412 | } | |
413 | else | |
414 | expected_delimiter = c; | |
415 | break; | |
416 | ||
417 | case '\\': | |
418 | /* Backslash quotes delimiters and itself, | |
419 | but not macro args. */ | |
420 | if (expected_delimiter != 0 && p < limit | |
421 | && (*p == expected_delimiter || *p == '\\')) | |
422 | { | |
423 | *exp_p++ = *p++; | |
424 | continue; | |
425 | } | |
426 | break; | |
427 | ||
428 | case '/': | |
429 | if (expected_delimiter != '\0') | |
430 | /* No comments inside strings. */ | |
431 | break; | |
432 | if (*p == '*') | |
433 | { | |
434 | /* If we find a comment that wasn't removed by | |
435 | handle_directive, this must be -traditional. | |
436 | So replace the comment with nothing at all. */ | |
437 | exp_p--; | |
438 | p += 1; | |
439 | while (p < limit && !(p[-2] == '*' && p[-1] == '/')) | |
440 | p++; | |
441 | #if 0 | |
442 | /* Mark this as a concatenation-point, | |
443 | as if it had been ##. */ | |
444 | concat = p; | |
445 | #endif | |
446 | } | |
447 | break; | |
448 | } | |
449 | } | |
450 | ||
451 | /* Handle the start of a symbol. */ | |
452 | if (is_idchar[c] && nargs > 0) | |
453 | { | |
454 | U_CHAR *id_beg = p - 1; | |
455 | int id_len; | |
456 | ||
457 | --exp_p; | |
458 | while (p != limit && is_idchar[*p]) | |
459 | p++; | |
460 | id_len = p - id_beg; | |
461 | ||
462 | if (is_idstart[c] | |
463 | && !(id_len == 1 && c == 'L' && (*p == '\'' || *p == '"'))) | |
464 | { | |
465 | register struct arglist *arg; | |
466 | ||
467 | for (arg = arglist; arg != NULL; arg = arg->next) | |
468 | { | |
469 | struct reflist *tpat; | |
470 | ||
471 | if (arg->name[0] == c | |
472 | && arg->length == id_len | |
473 | && strncmp (arg->name, id_beg, id_len) == 0) | |
474 | { | |
475 | if (expected_delimiter && CPP_OPTIONS | |
476 | (pfile)->warn_stringify) | |
477 | { | |
478 | if (CPP_TRADITIONAL (pfile)) | |
479 | { | |
480 | cpp_warning (pfile, | |
481 | "macro argument `%.*s' is stringified.", | |
482 | id_len, arg->name); | |
483 | } | |
484 | else | |
485 | { | |
486 | cpp_warning (pfile, | |
487 | "macro arg `%.*s' would be stringified with -traditional.", | |
488 | id_len, arg->name); | |
489 | } | |
490 | } | |
491 | /* If ANSI, don't actually substitute | |
492 | inside a string. */ | |
493 | if (!CPP_TRADITIONAL (pfile) && expected_delimiter) | |
494 | break; | |
495 | /* make a pat node for this arg and append it | |
496 | to the end of the pat list */ | |
497 | tpat = (struct reflist *) | |
498 | xmalloc (sizeof (struct reflist)); | |
499 | tpat->next = NULL; | |
500 | tpat->raw_before = concat == id_beg; | |
501 | tpat->raw_after = 0; | |
502 | tpat->rest_args = arg->rest_args; | |
503 | tpat->stringify = (CPP_TRADITIONAL (pfile) | |
504 | ? expected_delimiter != '\0' | |
505 | : stringify == id_beg); | |
506 | ||
507 | if (endpat == NULL) | |
508 | defn->pattern = tpat; | |
509 | else | |
510 | endpat->next = tpat; | |
511 | endpat = tpat; | |
512 | ||
513 | tpat->argno = arg->argno; | |
514 | tpat->nchars = exp_p - lastp; | |
515 | { | |
516 | register U_CHAR *p1 = p; | |
517 | SKIP_WHITE_SPACE (p1); | |
518 | if (p1 + 2 <= limit && p1[0] == '#' && p1[1] == '#') | |
519 | tpat->raw_after = 1; | |
520 | } | |
521 | lastp = exp_p; | |
522 | skipped_arg = 1; | |
523 | break; | |
524 | } | |
525 | } | |
526 | } | |
527 | ||
528 | /* If this was not a macro arg, copy it into the expansion. */ | |
529 | if (!skipped_arg) | |
530 | { | |
531 | register U_CHAR *lim1 = p; | |
532 | p = id_beg; | |
533 | while (p != lim1) | |
534 | *exp_p++ = *p++; | |
535 | if (stringify == id_beg) | |
536 | cpp_error (pfile, | |
537 | "`#' operator should be followed by a macro argument name"); | |
538 | } | |
539 | } | |
540 | } | |
541 | ||
542 | if (!CPP_TRADITIONAL (pfile) && expected_delimiter == 0) | |
543 | { | |
ed45de98 | 544 | /* If ANSI, put in a "\r " marker to prevent token pasting. |
6de1e2a9 ZW |
545 | But not if "inside a string" (which in ANSI mode |
546 | happens only for -D option). */ | |
ed45de98 | 547 | *exp_p++ = '\r'; |
6de1e2a9 ZW |
548 | *exp_p++ = ' '; |
549 | } | |
550 | ||
551 | *exp_p = '\0'; | |
552 | ||
553 | defn->length = exp_p - defn->expansion; | |
554 | ||
555 | /* Crash now if we overrun the allocated size. */ | |
556 | if (defn->length + 1 > maxsize) | |
557 | abort (); | |
558 | ||
559 | #if 0 | |
560 | /* This isn't worth the time it takes. */ | |
561 | /* give back excess storage */ | |
562 | defn->expansion = (U_CHAR *) xrealloc (defn->expansion, defn->length + 1); | |
563 | #endif | |
564 | ||
565 | return defn; | |
566 | } | |
567 | ||
568 | /* | |
569 | * special extension string that can be added to the last macro argument to | |
570 | * allow it to absorb the "rest" of the arguments when expanded. Ex: | |
571 | * #define wow(a, b...) process (b, a, b) | |
572 | * { wow (1, 2, 3); } -> { process (2, 3, 1, 2, 3); } | |
573 | * { wow (one, two); } -> { process (two, one, two); } | |
574 | * if this "rest_arg" is used with the concat token '##' and if it is not | |
575 | * supplied then the token attached to with ## will not be outputted. Ex: | |
576 | * #define wow (a, b...) process (b ## , a, ## b) | |
577 | * { wow (1, 2); } -> { process (2, 1, 2); } | |
578 | * { wow (one); } -> { process (one); { | |
579 | */ | |
580 | static char rest_extension[] = "..."; | |
581 | #define REST_EXTENSION_LENGTH (sizeof (rest_extension) - 1) | |
582 | ||
583 | /* Create a DEFINITION node from a #define directive. Arguments are | |
584 | as for do_define. */ | |
585 | ||
586 | MACRODEF | |
587 | create_definition (buf, limit, pfile, predefinition) | |
588 | U_CHAR *buf, *limit; | |
589 | cpp_reader *pfile; | |
590 | int predefinition; | |
591 | { | |
592 | U_CHAR *bp; /* temp ptr into input buffer */ | |
593 | U_CHAR *symname; /* remember where symbol name starts */ | |
594 | int sym_length; /* and how long it is */ | |
595 | int rest_args = 0; | |
596 | long line, col; | |
597 | char *file = CPP_BUFFER (pfile) ? CPP_BUFFER (pfile)->nominal_fname : ""; | |
598 | DEFINITION *defn; | |
599 | int arglengths = 0; /* Accumulate lengths of arg names | |
600 | plus number of args. */ | |
601 | MACRODEF mdef; | |
602 | cpp_buf_line_and_col (CPP_BUFFER (pfile), &line, &col); | |
603 | ||
604 | bp = buf; | |
605 | ||
606 | while (is_hor_space[*bp]) | |
607 | bp++; | |
608 | ||
609 | symname = bp; /* remember where it starts */ | |
610 | ||
611 | sym_length = check_macro_name (pfile, bp, 0); | |
612 | bp += sym_length; | |
613 | ||
614 | /* Lossage will occur if identifiers or control keywords are broken | |
615 | across lines using backslash. This is not the right place to take | |
616 | care of that. */ | |
617 | ||
618 | if (*bp == '(') | |
619 | { | |
620 | struct arglist *arg_ptrs = NULL; | |
621 | int argno = 0; | |
622 | ||
623 | bp++; /* skip '(' */ | |
624 | SKIP_WHITE_SPACE (bp); | |
625 | ||
626 | /* Loop over macro argument names. */ | |
627 | while (*bp != ')') | |
628 | { | |
629 | struct arglist *temp; | |
630 | ||
631 | temp = (struct arglist *) alloca (sizeof (struct arglist)); | |
632 | temp->name = bp; | |
633 | temp->next = arg_ptrs; | |
634 | temp->argno = argno++; | |
635 | temp->rest_args = 0; | |
636 | arg_ptrs = temp; | |
637 | ||
638 | if (rest_args) | |
639 | cpp_pedwarn (pfile, "another parameter follows `%s'", | |
640 | rest_extension); | |
641 | ||
642 | if (!is_idstart[*bp]) | |
643 | cpp_pedwarn (pfile, "invalid character in macro parameter name"); | |
644 | ||
645 | /* Find the end of the arg name. */ | |
646 | while (is_idchar[*bp]) | |
647 | { | |
648 | bp++; | |
649 | /* do we have a "special" rest-args extension here? */ | |
650 | if ((size_t) (limit - bp) > REST_EXTENSION_LENGTH | |
651 | && !strncmp (rest_extension, bp, REST_EXTENSION_LENGTH)) | |
652 | { | |
653 | rest_args = 1; | |
654 | temp->rest_args = 1; | |
655 | break; | |
656 | } | |
657 | } | |
658 | temp->length = bp - temp->name; | |
659 | if (rest_args == 1) | |
660 | bp += REST_EXTENSION_LENGTH; | |
661 | arglengths += temp->length + 2; | |
662 | SKIP_WHITE_SPACE (bp); | |
663 | if (temp->length == 0 || (*bp != ',' && *bp != ')')) | |
664 | { | |
665 | cpp_error (pfile, | |
666 | "badly punctuated parameter list in `#define'"); | |
667 | goto nope; | |
668 | } | |
669 | if (*bp == ',') | |
670 | { | |
671 | bp++; | |
672 | SKIP_WHITE_SPACE (bp); | |
673 | } | |
674 | if (bp >= limit) | |
675 | { | |
676 | cpp_error (pfile, "unterminated parameter list in `#define'"); | |
677 | goto nope; | |
678 | } | |
679 | { | |
680 | struct arglist *otemp; | |
681 | ||
682 | for (otemp = temp->next; otemp != NULL; otemp = otemp->next) | |
683 | if (temp->length == otemp->length | |
684 | && strncmp (temp->name, otemp->name, temp->length) == 0) | |
685 | { | |
686 | U_CHAR *name; | |
687 | ||
688 | name = (U_CHAR *) alloca (temp->length + 1); | |
689 | (void) strncpy (name, temp->name, temp->length); | |
690 | name[temp->length] = '\0'; | |
691 | cpp_error (pfile, | |
692 | "duplicate argument name `%s' in `#define'", | |
693 | name); | |
694 | goto nope; | |
695 | } | |
696 | } | |
697 | } | |
698 | ||
699 | ++bp; /* skip paren */ | |
700 | SKIP_WHITE_SPACE (bp); | |
701 | /* now everything from bp before limit is the definition. */ | |
702 | defn = collect_expansion (pfile, bp, limit, argno, arg_ptrs); | |
703 | defn->rest_args = rest_args; | |
704 | ||
705 | /* Now set defn->args.argnames to the result of concatenating | |
706 | the argument names in reverse order | |
707 | with comma-space between them. */ | |
708 | defn->args.argnames = (U_CHAR *) xmalloc (arglengths + 1); | |
709 | { | |
710 | struct arglist *temp; | |
711 | int i = 0; | |
712 | for (temp = arg_ptrs; temp; temp = temp->next) | |
713 | { | |
714 | bcopy (temp->name, &defn->args.argnames[i], temp->length); | |
715 | i += temp->length; | |
716 | if (temp->next != 0) | |
717 | { | |
718 | defn->args.argnames[i++] = ','; | |
719 | defn->args.argnames[i++] = ' '; | |
720 | } | |
721 | } | |
722 | defn->args.argnames[i] = 0; | |
723 | } | |
724 | } | |
725 | else | |
726 | { | |
727 | /* Simple expansion or empty definition. */ | |
728 | ||
729 | if (bp < limit) | |
730 | { | |
731 | if (is_hor_space[*bp]) | |
732 | { | |
733 | bp++; | |
734 | SKIP_WHITE_SPACE (bp); | |
735 | } | |
736 | else | |
737 | /* Per C9x, missing white space after the name in a #define | |
738 | of an object-like macro is always a constraint violation. */ | |
739 | cpp_pedwarn (pfile, | |
740 | "missing white space after `#define %.*s'", | |
741 | sym_length, symname); | |
742 | } | |
743 | /* now everything from bp before limit is the definition. */ | |
744 | defn = collect_expansion (pfile, bp, limit, -1, NULL_PTR); | |
745 | defn->args.argnames = (U_CHAR *) ""; | |
746 | } | |
747 | ||
748 | defn->line = line; | |
749 | defn->file = file; | |
750 | ||
751 | /* OP is null if this is a predefinition */ | |
752 | defn->predefined = predefinition; | |
753 | mdef.defn = defn; | |
754 | mdef.symnam = symname; | |
755 | mdef.symlen = sym_length; | |
756 | ||
757 | return mdef; | |
758 | ||
759 | nope: | |
760 | mdef.defn = 0; | |
761 | return mdef; | |
762 | } | |
763 | ||
764 | /* | |
765 | * Parse a macro argument and append the info on PFILE's token_buffer. | |
766 | * REST_ARGS means to absorb the rest of the args. | |
767 | * Return nonzero to indicate a syntax error. | |
768 | */ | |
769 | ||
770 | static enum cpp_token | |
771 | macarg (pfile, rest_args) | |
772 | cpp_reader *pfile; | |
773 | int rest_args; | |
774 | { | |
775 | int paren = 0; | |
776 | enum cpp_token token; | |
777 | char save_put_out_comments = CPP_OPTIONS (pfile)->put_out_comments; | |
778 | CPP_OPTIONS (pfile)->put_out_comments = 0; | |
779 | ||
780 | /* Try to parse as much of the argument as exists at this | |
781 | input stack level. */ | |
782 | pfile->no_macro_expand++; | |
3fdc651f | 783 | CPP_OPTIONS (pfile)->no_line_commands++; |
6de1e2a9 ZW |
784 | for (;;) |
785 | { | |
786 | token = cpp_get_token (pfile); | |
787 | switch (token) | |
788 | { | |
789 | case CPP_EOF: | |
790 | goto done; | |
791 | case CPP_POP: | |
792 | /* If we've hit end of file, it's an error (reported by caller). | |
793 | Ditto if it's the end of cpp_expand_to_buffer text. | |
794 | If we've hit end of macro, just continue. */ | |
795 | if (!CPP_IS_MACRO_BUFFER (CPP_BUFFER (pfile))) | |
796 | goto done; | |
797 | break; | |
798 | case CPP_LPAREN: | |
799 | paren++; | |
800 | break; | |
801 | case CPP_RPAREN: | |
802 | if (--paren < 0) | |
803 | goto found; | |
804 | break; | |
805 | case CPP_COMMA: | |
806 | /* if we've returned to lowest level and | |
807 | we aren't absorbing all args */ | |
808 | if (paren == 0 && rest_args == 0) | |
809 | goto found; | |
810 | break; | |
811 | found: | |
812 | /* Remove ',' or ')' from argument buffer. */ | |
813 | CPP_ADJUST_WRITTEN (pfile, -1); | |
814 | goto done; | |
815 | default:; | |
816 | } | |
817 | } | |
818 | ||
819 | done: | |
820 | CPP_OPTIONS (pfile)->put_out_comments = save_put_out_comments; | |
3fdc651f | 821 | CPP_OPTIONS (pfile)->no_line_commands--; |
6de1e2a9 ZW |
822 | pfile->no_macro_expand--; |
823 | ||
824 | return token; | |
825 | } | |
826 | \f | |
827 | /* Turn newlines to spaces in the string of length LENGTH at START, | |
828 | except inside of string constants. | |
829 | The string is copied into itself with its beginning staying fixed. */ | |
830 | ||
831 | static int | |
832 | change_newlines (start, length) | |
833 | U_CHAR *start; | |
834 | int length; | |
835 | { | |
836 | register U_CHAR *ibp; | |
837 | register U_CHAR *obp; | |
838 | register U_CHAR *limit; | |
839 | register int c; | |
840 | ||
841 | ibp = start; | |
842 | limit = start + length; | |
843 | obp = start; | |
844 | ||
845 | while (ibp < limit) | |
846 | { | |
847 | *obp++ = c = *ibp++; | |
848 | switch (c) | |
849 | { | |
850 | ||
851 | case '\'': | |
852 | case '\"': | |
853 | /* Notice and skip strings, so that we don't | |
854 | delete newlines in them. */ | |
855 | { | |
856 | int quotec = c; | |
857 | while (ibp < limit) | |
858 | { | |
859 | *obp++ = c = *ibp++; | |
860 | if (c == quotec) | |
861 | break; | |
862 | if (c == '\n' && quotec == '\'') | |
863 | break; | |
864 | } | |
865 | } | |
866 | break; | |
867 | } | |
868 | } | |
869 | ||
870 | return obp - start; | |
871 | } | |
872 | \f | |
873 | ||
874 | static struct tm * | |
875 | timestamp (pfile) | |
876 | cpp_reader *pfile; | |
877 | { | |
878 | if (!pfile->timebuf) | |
879 | { | |
880 | time_t t = time ((time_t *) 0); | |
881 | pfile->timebuf = localtime (&t); | |
882 | } | |
883 | return pfile->timebuf; | |
884 | } | |
885 | ||
886 | static char *monthnames[] = | |
887 | { | |
888 | "Jan", "Feb", "Mar", "Apr", "May", "Jun", | |
889 | "Jul", "Aug", "Sep", "Oct", "Nov", "Dec", | |
890 | }; | |
891 | ||
892 | /* | |
893 | * expand things like __FILE__. Place the expansion into the output | |
894 | * buffer *without* rescanning. | |
895 | */ | |
896 | ||
897 | static void | |
898 | special_symbol (hp, pfile) | |
899 | HASHNODE *hp; | |
900 | cpp_reader *pfile; | |
901 | { | |
902 | const char *buf; | |
903 | int len; | |
904 | cpp_buffer *ip; | |
905 | ||
906 | switch (hp->type) | |
907 | { | |
908 | case T_FILE: | |
909 | case T_BASE_FILE: | |
910 | { | |
911 | ip = CPP_BUFFER (pfile); | |
912 | if (hp->type == T_BASE_FILE) | |
913 | { | |
914 | while (CPP_PREV_BUFFER (ip) != CPP_NULL_BUFFER (pfile)) | |
915 | ip = CPP_PREV_BUFFER (ip); | |
916 | } | |
917 | else | |
918 | { | |
919 | ip = CPP_BUFFER (pfile); | |
920 | while (!ip->nominal_fname && ip != CPP_NULL_BUFFER (pfile)) | |
921 | ip = CPP_PREV_BUFFER (ip); | |
922 | } | |
923 | ||
924 | buf = ip->nominal_fname; | |
925 | ||
926 | if (!buf) | |
927 | buf = ""; | |
928 | CPP_RESERVE (pfile, 3 + 4 * strlen (buf)); | |
929 | quote_string (pfile, buf); | |
930 | return; | |
931 | } | |
932 | ||
933 | case T_INCLUDE_LEVEL: | |
934 | { | |
935 | int true_indepth = 0; | |
936 | ip = CPP_BUFFER (pfile); | |
937 | for (; ip != CPP_NULL_BUFFER (pfile); ip = CPP_PREV_BUFFER (ip)) | |
938 | if (ip->fname != NULL) | |
939 | true_indepth++; | |
940 | ||
941 | CPP_RESERVE (pfile, 10); | |
942 | sprintf (CPP_PWRITTEN (pfile), "%d", true_indepth); | |
943 | CPP_ADJUST_WRITTEN (pfile, strlen (CPP_PWRITTEN (pfile))); | |
944 | return; | |
945 | } | |
946 | ||
947 | case T_VERSION: | |
948 | len = strlen (version_string); | |
949 | CPP_RESERVE (pfile, 3 + len); | |
950 | CPP_PUTC_Q (pfile, '"'); | |
951 | CPP_PUTS_Q (pfile, version_string, len); | |
952 | CPP_PUTC_Q (pfile, '"'); | |
953 | CPP_NUL_TERMINATE_Q (pfile); | |
954 | return; | |
955 | ||
956 | case T_CONST: | |
957 | buf = hp->value.cpval; | |
958 | if (!buf) | |
959 | return; | |
960 | if (*buf == '\0') | |
ed45de98 | 961 | buf = "\r "; |
6de1e2a9 ZW |
962 | |
963 | len = strlen (buf); | |
964 | CPP_RESERVE (pfile, len + 1); | |
965 | CPP_PUTS_Q (pfile, buf, len); | |
966 | CPP_NUL_TERMINATE_Q (pfile); | |
967 | return; | |
968 | ||
969 | case T_STDC: | |
970 | CPP_RESERVE (pfile, 2); | |
971 | #ifdef STDC_0_IN_SYSTEM_HEADERS | |
972 | ip = CPP_BUFFER (pfile); | |
973 | while (!ip->nominal_fname && ip != CPP_NULL_BUFFER (pfile)) | |
974 | ip = CPP_PREV_BUFFER (ip); | |
975 | if (ip->system_header_p | |
976 | && !cpp_lookup (pfile, (U_CHAR *) "__STRICT_ANSI__", 15, -1)) | |
977 | CPP_PUTC_Q (pfile, '0'); | |
978 | else | |
979 | #endif | |
980 | CPP_PUTC_Q (pfile, '1'); | |
981 | CPP_NUL_TERMINATE_Q (pfile); | |
982 | return; | |
983 | ||
984 | case T_SPECLINE: | |
985 | { | |
986 | long line; | |
5e4df1ae | 987 | cpp_buf_line_and_col (cpp_file_buffer (pfile), &line, NULL); |
6de1e2a9 ZW |
988 | |
989 | CPP_RESERVE (pfile, 10); | |
990 | sprintf (CPP_PWRITTEN (pfile), "%ld", line); | |
991 | CPP_ADJUST_WRITTEN (pfile, strlen (CPP_PWRITTEN (pfile))); | |
992 | return; | |
993 | } | |
994 | ||
995 | case T_DATE: | |
996 | case T_TIME: | |
997 | { | |
998 | struct tm *timebuf; | |
999 | ||
1000 | CPP_RESERVE (pfile, 20); | |
1001 | timebuf = timestamp (pfile); | |
1002 | if (hp->type == T_DATE) | |
1003 | sprintf (CPP_PWRITTEN (pfile), "\"%s %2d %4d\"", | |
1004 | monthnames[timebuf->tm_mon], | |
1005 | timebuf->tm_mday, timebuf->tm_year + 1900); | |
1006 | else | |
1007 | sprintf (CPP_PWRITTEN (pfile), "\"%02d:%02d:%02d\"", | |
1008 | timebuf->tm_hour, timebuf->tm_min, timebuf->tm_sec); | |
1009 | ||
1010 | CPP_ADJUST_WRITTEN (pfile, strlen (CPP_PWRITTEN (pfile))); | |
1011 | return; | |
1012 | } | |
1013 | ||
1014 | default: | |
1015 | cpp_fatal (pfile, "cpplib internal error: invalid special hash type"); | |
1016 | return; | |
1017 | } | |
6de1e2a9 ZW |
1018 | } |
1019 | ||
1020 | /* Expand a macro call. | |
1021 | HP points to the symbol that is the macro being called. | |
1022 | Put the result of expansion onto the input stack | |
1023 | so that subsequent input by our caller will use it. | |
1024 | ||
1025 | If macro wants arguments, caller has already verified that | |
1026 | an argument list follows; arguments come from the input stack. */ | |
1027 | ||
1028 | void | |
1029 | macroexpand (pfile, hp) | |
1030 | cpp_reader *pfile; | |
1031 | HASHNODE *hp; | |
1032 | { | |
1033 | int nargs; | |
1034 | DEFINITION *defn; | |
1035 | register U_CHAR *xbuf; | |
1036 | long start_line, start_column; | |
1037 | int xbuf_len; | |
1038 | struct argdata *args; | |
1039 | long old_written = CPP_WRITTEN (pfile); | |
1040 | #if 0 | |
1041 | int start_line = instack[indepth].lineno; | |
1042 | #endif | |
1043 | int rest_args, rest_zero; | |
1044 | register int i; | |
1045 | ||
1046 | #if 0 | |
1047 | /* This macro is being used inside a #if, which means it must be */ | |
1048 | /* recorded as a precondition. */ | |
1049 | if (pcp_inside_if && pcp_outfile && defn->predefined) | |
1050 | dump_single_macro (hp, pcp_outfile); | |
1051 | #endif | |
1052 | ||
1053 | cpp_buf_line_and_col (cpp_file_buffer (pfile), &start_line, &start_column); | |
1054 | ||
1055 | /* Check for and handle special symbols. */ | |
1056 | if (hp->type != T_MACRO) | |
1057 | { | |
1058 | special_symbol (hp, pfile); | |
1059 | xbuf_len = CPP_WRITTEN (pfile) - old_written; | |
1060 | xbuf = (U_CHAR *) xmalloc (xbuf_len + 1); | |
1061 | CPP_SET_WRITTEN (pfile, old_written); | |
1062 | bcopy (CPP_PWRITTEN (pfile), xbuf, xbuf_len + 1); | |
1063 | push_macro_expansion (pfile, xbuf, xbuf_len, hp); | |
1064 | CPP_BUFFER (pfile)->has_escapes = 1; | |
1065 | return; | |
1066 | } | |
1067 | ||
1068 | defn = hp->value.defn; | |
1069 | nargs = defn->nargs; | |
1070 | pfile->output_escapes++; | |
1071 | ||
1072 | if (nargs >= 0) | |
1073 | { | |
099a9dd0 | 1074 | enum cpp_token token = CPP_EOF; |
6de1e2a9 ZW |
1075 | |
1076 | args = (struct argdata *) alloca ((nargs + 1) * sizeof (struct argdata)); | |
1077 | ||
1078 | for (i = 0; i < nargs; i++) | |
1079 | { | |
1080 | args[i].raw = args[i].expanded = 0; | |
1081 | args[i].raw_length = 0; | |
1082 | args[i].expand_length = args[i].stringified_length = -1; | |
1083 | args[i].use_count = 0; | |
1084 | } | |
1085 | ||
1086 | /* Parse all the macro args that are supplied. I counts them. | |
1087 | The first NARGS args are stored in ARGS. | |
1088 | The rest are discarded. If rest_args is set then we assume | |
1089 | macarg absorbed the rest of the args. */ | |
1090 | i = 0; | |
1091 | rest_args = 0; | |
1092 | rest_args = 0; | |
1093 | FORWARD (1); /* Discard open-parenthesis before first arg. */ | |
1094 | do | |
1095 | { | |
1096 | if (rest_args) | |
1097 | continue; | |
1098 | if (i < nargs || (nargs == 0 && i == 0)) | |
1099 | { | |
1100 | /* if we are working on last arg which absorbs rest of args... */ | |
1101 | if (i == nargs - 1 && defn->rest_args) | |
1102 | rest_args = 1; | |
1103 | args[i].raw = CPP_WRITTEN (pfile); | |
1104 | token = macarg (pfile, rest_args); | |
1105 | args[i].raw_length = CPP_WRITTEN (pfile) - args[i].raw; | |
1106 | args[i].newlines = 0; /* FIXME */ | |
1107 | } | |
1108 | else | |
1109 | token = macarg (pfile, 0); | |
1110 | if (token == CPP_EOF || token == CPP_POP) | |
1111 | { | |
1112 | cpp_error_with_line (pfile, start_line, start_column, | |
1113 | "unterminated macro call"); | |
1114 | return; | |
1115 | } | |
1116 | i++; | |
1117 | } | |
1118 | while (token == CPP_COMMA); | |
1119 | ||
1120 | /* If we got one arg but it was just whitespace, call that 0 args. */ | |
1121 | if (i == 1) | |
1122 | { | |
1123 | register U_CHAR *bp = ARG_BASE + args[0].raw; | |
1124 | register U_CHAR *lim = bp + args[0].raw_length; | |
1125 | /* cpp.texi says for foo ( ) we provide one argument. | |
1126 | However, if foo wants just 0 arguments, treat this as 0. */ | |
1127 | if (nargs == 0) | |
1128 | while (bp != lim && is_space[*bp]) | |
1129 | bp++; | |
1130 | if (bp == lim) | |
1131 | i = 0; | |
1132 | } | |
1133 | ||
1134 | /* Don't output an error message if we have already output one for | |
1135 | a parse error above. */ | |
1136 | rest_zero = 0; | |
1137 | if (nargs == 0 && i > 0) | |
1138 | { | |
1139 | cpp_error (pfile, "arguments given to macro `%s'", hp->name); | |
1140 | } | |
1141 | else if (i < nargs) | |
1142 | { | |
1143 | /* traditional C allows foo() if foo wants one argument. */ | |
1144 | if (nargs == 1 && i == 0 && CPP_TRADITIONAL (pfile)) | |
1145 | ; | |
1146 | /* the rest args token is allowed to absorb 0 tokens */ | |
1147 | else if (i == nargs - 1 && defn->rest_args) | |
1148 | rest_zero = 1; | |
1149 | else if (i == 0) | |
1150 | cpp_error (pfile, "macro `%s' used without args", hp->name); | |
1151 | else if (i == 1) | |
1152 | cpp_error (pfile, "macro `%s' used with just one arg", hp->name); | |
1153 | else | |
1154 | cpp_error (pfile, "macro `%s' used with only %d args", | |
1155 | hp->name, i); | |
1156 | } | |
1157 | else if (i > nargs) | |
1158 | { | |
1159 | cpp_error (pfile, | |
1160 | "macro `%s' used with too many (%d) args", hp->name, i); | |
1161 | } | |
1162 | } | |
1163 | ||
1164 | /* If macro wants zero args, we parsed the arglist for checking only. | |
1165 | Read directly from the macro definition. */ | |
1166 | if (nargs <= 0) | |
1167 | { | |
1168 | xbuf = defn->expansion; | |
1169 | xbuf_len = defn->length; | |
1170 | } | |
1171 | else | |
1172 | { | |
1173 | register U_CHAR *exp = defn->expansion; | |
1174 | register int offset; /* offset in expansion, | |
1175 | copied a piece at a time */ | |
1176 | register int totlen; /* total amount of exp buffer filled so far */ | |
1177 | ||
1178 | register struct reflist *ap, *last_ap; | |
1179 | ||
1180 | /* Macro really takes args. Compute the expansion of this call. */ | |
1181 | ||
1182 | /* Compute length in characters of the macro's expansion. | |
1183 | Also count number of times each arg is used. */ | |
1184 | xbuf_len = defn->length; | |
1185 | for (ap = defn->pattern; ap != NULL; ap = ap->next) | |
1186 | { | |
1187 | if (ap->stringify) | |
1188 | { | |
1189 | register struct argdata *arg = &args[ap->argno]; | |
1190 | /* Stringify if it hasn't already been */ | |
1191 | if (arg->stringified_length < 0) | |
1192 | { | |
1193 | int arglen = arg->raw_length; | |
1194 | int escaped = 0; | |
1195 | int in_string = 0; | |
1196 | int c; | |
1197 | /* Initially need_space is -1. Otherwise, 1 means the | |
1198 | previous character was a space, but we suppressed it; | |
1199 | 0 means the previous character was a non-space. */ | |
1200 | int need_space = -1; | |
1201 | i = 0; | |
1202 | arg->stringified = CPP_WRITTEN (pfile); | |
1203 | if (!CPP_TRADITIONAL (pfile)) | |
1204 | CPP_PUTC (pfile, '\"'); /* insert beginning quote */ | |
1205 | for (; i < arglen; i++) | |
1206 | { | |
1207 | c = (ARG_BASE + arg->raw)[i]; | |
1208 | ||
1209 | if (!in_string) | |
1210 | { | |
1211 | /* Internal sequences of whitespace are | |
1212 | replaced by one space except within | |
1213 | a string or char token. */ | |
1214 | if (is_space[c]) | |
1215 | { | |
1216 | if (CPP_WRITTEN (pfile) > (unsigned) arg->stringified | |
ed45de98 | 1217 | && (CPP_PWRITTEN (pfile))[-1] == '\r') |
6de1e2a9 | 1218 | { |
ed45de98 | 1219 | /* "\r " escape markers are removed */ |
6de1e2a9 ZW |
1220 | CPP_ADJUST_WRITTEN (pfile, -1); |
1221 | continue; | |
1222 | } | |
1223 | if (need_space == 0) | |
1224 | need_space = 1; | |
1225 | continue; | |
1226 | } | |
1227 | else if (need_space > 0) | |
1228 | CPP_PUTC (pfile, ' '); | |
1229 | need_space = 0; | |
1230 | } | |
1231 | ||
1232 | if (escaped) | |
1233 | escaped = 0; | |
1234 | else | |
1235 | { | |
1236 | if (c == '\\') | |
1237 | escaped = 1; | |
1238 | if (in_string) | |
1239 | { | |
1240 | if (c == in_string) | |
1241 | in_string = 0; | |
1242 | } | |
1243 | else if (c == '\"' || c == '\'') | |
1244 | in_string = c; | |
1245 | } | |
1246 | ||
1247 | /* Escape these chars */ | |
1248 | if (c == '\"' || (in_string && c == '\\')) | |
1249 | CPP_PUTC (pfile, '\\'); | |
1250 | if (ISPRINT (c)) | |
1251 | CPP_PUTC (pfile, c); | |
1252 | else | |
1253 | { | |
1254 | CPP_RESERVE (pfile, 4); | |
1255 | sprintf ((char *) CPP_PWRITTEN (pfile), "\\%03o", | |
1256 | (unsigned int) c); | |
1257 | CPP_ADJUST_WRITTEN (pfile, 4); | |
1258 | } | |
1259 | } | |
1260 | if (!CPP_TRADITIONAL (pfile)) | |
1261 | CPP_PUTC (pfile, '\"'); /* insert ending quote */ | |
1262 | arg->stringified_length | |
1263 | = CPP_WRITTEN (pfile) - arg->stringified; | |
1264 | } | |
1265 | xbuf_len += args[ap->argno].stringified_length; | |
1266 | } | |
1267 | else if (ap->raw_before || ap->raw_after || CPP_TRADITIONAL (pfile)) | |
1268 | /* Add 4 for two newline-space markers to prevent | |
1269 | token concatenation. */ | |
1270 | xbuf_len += args[ap->argno].raw_length + 4; | |
1271 | else | |
1272 | { | |
1273 | /* We have an ordinary (expanded) occurrence of the arg. | |
1274 | So compute its expansion, if we have not already. */ | |
1275 | if (args[ap->argno].expand_length < 0) | |
1276 | { | |
1277 | args[ap->argno].expanded = CPP_WRITTEN (pfile); | |
1278 | cpp_expand_to_buffer (pfile, | |
1279 | ARG_BASE + args[ap->argno].raw, | |
1280 | args[ap->argno].raw_length); | |
1281 | ||
1282 | args[ap->argno].expand_length | |
1283 | = CPP_WRITTEN (pfile) - args[ap->argno].expanded; | |
1284 | } | |
1285 | ||
1286 | /* Add 4 for two newline-space markers to prevent | |
1287 | token concatenation. */ | |
1288 | xbuf_len += args[ap->argno].expand_length + 4; | |
1289 | } | |
1290 | if (args[ap->argno].use_count < 10) | |
1291 | args[ap->argno].use_count++; | |
1292 | } | |
1293 | ||
1294 | xbuf = (U_CHAR *) xmalloc (xbuf_len + 1); | |
1295 | ||
1296 | /* Generate in XBUF the complete expansion | |
1297 | with arguments substituted in. | |
1298 | TOTLEN is the total size generated so far. | |
1299 | OFFSET is the index in the definition | |
1300 | of where we are copying from. */ | |
1301 | offset = totlen = 0; | |
1302 | for (last_ap = NULL, ap = defn->pattern; ap != NULL; | |
1303 | last_ap = ap, ap = ap->next) | |
1304 | { | |
1305 | register struct argdata *arg = &args[ap->argno]; | |
1306 | int count_before = totlen; | |
1307 | ||
1308 | /* Add chars to XBUF. */ | |
1309 | for (i = 0; i < ap->nchars; i++, offset++) | |
1310 | xbuf[totlen++] = exp[offset]; | |
1311 | ||
1312 | /* If followed by an empty rest arg with concatenation, | |
1313 | delete the last run of nonwhite chars. */ | |
1314 | if (rest_zero && totlen > count_before | |
1315 | && ((ap->rest_args && ap->raw_before) | |
1316 | || (last_ap != NULL && last_ap->rest_args | |
1317 | && last_ap->raw_after))) | |
1318 | { | |
1319 | /* Delete final whitespace. */ | |
1320 | while (totlen > count_before && is_space[xbuf[totlen - 1]]) | |
1321 | totlen--; | |
1322 | ||
1323 | /* Delete the nonwhites before them. */ | |
1324 | while (totlen > count_before && !is_space[xbuf[totlen - 1]]) | |
1325 | totlen--; | |
1326 | } | |
1327 | ||
1328 | if (ap->stringify != 0) | |
1329 | { | |
1330 | bcopy (ARG_BASE + arg->stringified, | |
1331 | xbuf + totlen, arg->stringified_length); | |
1332 | totlen += arg->stringified_length; | |
1333 | } | |
1334 | else if (ap->raw_before || ap->raw_after || CPP_TRADITIONAL (pfile)) | |
1335 | { | |
1336 | U_CHAR *p1 = ARG_BASE + arg->raw; | |
1337 | U_CHAR *l1 = p1 + arg->raw_length; | |
1338 | if (ap->raw_before) | |
1339 | { | |
1340 | while (p1 != l1 && is_space[*p1]) | |
1341 | p1++; | |
1342 | while (p1 != l1 && is_idchar[*p1]) | |
1343 | xbuf[totlen++] = *p1++; | |
1344 | } | |
1345 | if (ap->raw_after) | |
1346 | { | |
1347 | /* Arg is concatenated after: delete trailing whitespace, | |
1348 | whitespace markers, and no-reexpansion markers. */ | |
1349 | while (p1 != l1) | |
1350 | { | |
1351 | if (is_space[l1[-1]]) | |
1352 | l1--; | |
ed45de98 ZW |
1353 | else if (l1[-1] == '\r') |
1354 | l1--; | |
6de1e2a9 ZW |
1355 | else if (l1[-1] == '-') |
1356 | { | |
ed45de98 | 1357 | if (l1 != p1 + 1 && l1[-2] == '\r') |
6de1e2a9 ZW |
1358 | l1 -= 2; |
1359 | else | |
1360 | break; | |
1361 | } | |
1362 | else | |
1363 | break; | |
1364 | } | |
1365 | } | |
1366 | ||
1367 | /* Delete any no-reexpansion marker that precedes | |
1368 | an identifier at the beginning of the argument. */ | |
ed45de98 | 1369 | if (p1[0] == '\r' && p1[1] == '-') |
6de1e2a9 ZW |
1370 | p1 += 2; |
1371 | ||
1372 | bcopy (p1, xbuf + totlen, l1 - p1); | |
1373 | totlen += l1 - p1; | |
1374 | } | |
1375 | else | |
1376 | { | |
1377 | U_CHAR *expanded = ARG_BASE + arg->expanded; | |
1378 | if (!ap->raw_before && totlen > 0 && arg->expand_length | |
1379 | && !CPP_TRADITIONAL (pfile) | |
1380 | && unsafe_chars (xbuf[totlen - 1], expanded[0])) | |
1381 | { | |
ed45de98 | 1382 | xbuf[totlen++] = '\r'; |
6de1e2a9 ZW |
1383 | xbuf[totlen++] = ' '; |
1384 | } | |
1385 | ||
1386 | bcopy (expanded, xbuf + totlen, arg->expand_length); | |
1387 | totlen += arg->expand_length; | |
1388 | ||
1389 | if (!ap->raw_after && totlen > 0 && offset < defn->length | |
1390 | && !CPP_TRADITIONAL (pfile) | |
1391 | && unsafe_chars (xbuf[totlen - 1], exp[offset])) | |
1392 | { | |
ed45de98 | 1393 | xbuf[totlen++] = '\r'; |
6de1e2a9 ZW |
1394 | xbuf[totlen++] = ' '; |
1395 | } | |
1396 | ||
1397 | /* If a macro argument with newlines is used multiple times, | |
1398 | then only expand the newlines once. This avoids creating | |
1399 | output lines which don't correspond to any input line, | |
1400 | which confuses gdb and gcov. */ | |
1401 | if (arg->use_count > 1 && arg->newlines > 0) | |
1402 | { | |
1403 | /* Don't bother doing change_newlines for subsequent | |
1404 | uses of arg. */ | |
1405 | arg->use_count = 1; | |
1406 | arg->expand_length | |
1407 | = change_newlines (expanded, arg->expand_length); | |
1408 | } | |
1409 | } | |
1410 | ||
1411 | if (totlen > xbuf_len) | |
34ca9541 ZW |
1412 | { |
1413 | cpp_fatal (pfile, "internal_error: buffer overrun in macroexpand"); | |
1414 | return; | |
1415 | } | |
6de1e2a9 ZW |
1416 | } |
1417 | ||
1418 | /* if there is anything left of the definition | |
1419 | after handling the arg list, copy that in too. */ | |
1420 | ||
1421 | for (i = offset; i < defn->length; i++) | |
1422 | { | |
1423 | /* if we've reached the end of the macro */ | |
1424 | if (exp[i] == ')') | |
1425 | rest_zero = 0; | |
1426 | if (!(rest_zero && last_ap != NULL && last_ap->rest_args | |
1427 | && last_ap->raw_after)) | |
1428 | xbuf[totlen++] = exp[i]; | |
1429 | } | |
1430 | ||
1431 | xbuf[totlen] = 0; | |
1432 | xbuf_len = totlen; | |
1433 | ||
1434 | } | |
1435 | ||
1436 | pfile->output_escapes--; | |
1437 | ||
1438 | /* Now put the expansion on the input stack | |
1439 | so our caller will commence reading from it. */ | |
1440 | push_macro_expansion (pfile, xbuf, xbuf_len, hp); | |
1441 | CPP_BUFFER (pfile)->has_escapes = 1; | |
1442 | ||
1443 | /* Pop the space we've used in the token_buffer for argument expansion. */ | |
1444 | CPP_SET_WRITTEN (pfile, old_written); | |
1445 | ||
1446 | /* Recursive macro use sometimes works traditionally. | |
1447 | #define foo(x,y) bar (x (y,0), y) | |
1448 | foo (foo, baz) */ | |
1449 | ||
1450 | if (!CPP_TRADITIONAL (pfile)) | |
1451 | hp->type = T_DISABLED; | |
1452 | } | |
1453 | ||
1454 | /* Return 1 iff a token ending in C1 followed directly by a token C2 | |
1455 | could cause mis-tokenization. */ | |
1456 | ||
1457 | static int | |
1458 | unsafe_chars (c1, c2) | |
1459 | int c1, c2; | |
1460 | { | |
1461 | switch (c1) | |
1462 | { | |
1463 | case '+': | |
1464 | case '-': | |
1465 | if (c2 == c1 || c2 == '=') | |
1466 | return 1; | |
1467 | goto letter; | |
1468 | ||
1469 | case '.': case '0': case '1': case '2': case '3': | |
1470 | case '4': case '5': case '6': case '7': case '8': | |
1471 | case '9': case 'e': case 'E': case 'p': case 'P': | |
1472 | if (c2 == '-' || c2 == '+') | |
1473 | return 1; /* could extend a pre-processing number */ | |
1474 | goto letter; | |
1475 | ||
1476 | case 'L': | |
1477 | if (c2 == '\'' || c2 == '\"') | |
1478 | return 1; /* Could turn into L"xxx" or L'xxx'. */ | |
1479 | goto letter; | |
1480 | ||
1481 | case '_': case 'a': case 'b': case 'c': case 'd': case 'f': | |
1482 | case 'g': case 'h': case 'i': case 'j': case 'k': case 'l': | |
1483 | case 'm': case 'n': case 'o': case 'q': case 'r': case 's': | |
1484 | case 't': case 'u': case 'v': case 'w': case 'x': case 'y': | |
1485 | case 'z': case 'A': case 'B': case 'C': case 'D': case 'F': | |
1486 | case 'G': case 'H': case 'I': case 'J': case 'K': case 'M': | |
1487 | case 'N': case 'O': case 'Q': case 'R': case 'S': case 'T': | |
1488 | case 'U': case 'V': case 'W': case 'X': case 'Y': case 'Z': | |
1489 | letter: | |
1490 | /* We're in the middle of either a name or a pre-processing number. */ | |
1491 | return (is_idchar[c2] || c2 == '.'); | |
1492 | ||
1493 | case '<': case '>': case '!': case '%': case '#': case ':': | |
1494 | case '^': case '&': case '|': case '*': case '/': case '=': | |
1495 | return (c2 == c1 || c2 == '='); | |
1496 | } | |
1497 | return 0; | |
1498 | } | |
1499 | ||
1500 | static void | |
1501 | push_macro_expansion (pfile, xbuf, xbuf_len, hp) | |
1502 | cpp_reader *pfile; | |
1503 | register U_CHAR *xbuf; | |
1504 | int xbuf_len; | |
1505 | HASHNODE *hp; | |
1506 | { | |
1507 | register cpp_buffer *mbuf = cpp_push_buffer (pfile, xbuf, xbuf_len); | |
1508 | if (mbuf == NULL) | |
1509 | return; | |
1510 | mbuf->cleanup = macro_cleanup; | |
1511 | mbuf->data = hp; | |
1512 | ||
ed45de98 | 1513 | /* The first chars of the expansion should be a "\r " added by |
6de1e2a9 ZW |
1514 | collect_expansion. This is to prevent accidental token-pasting |
1515 | between the text preceding the macro invocation, and the macro | |
1516 | expansion text. | |
1517 | ||
1518 | We would like to avoid adding unneeded spaces (for the sake of | |
1519 | tools that use cpp, such as imake). In some common cases we can | |
1520 | tell that it is safe to omit the space. | |
1521 | ||
1522 | The character before the macro invocation cannot have been an | |
1523 | idchar (or else it would have been pasted with the idchars of | |
1524 | the macro name). Therefore, if the first non-space character | |
1525 | of the expansion is an idchar, we do not need the extra space | |
1526 | to prevent token pasting. | |
1527 | ||
1528 | Also, we don't need the extra space if the first char is '(', | |
1529 | or some other (less common) characters. */ | |
1530 | ||
ed45de98 | 1531 | if (xbuf[0] == '\r' && xbuf[1] == ' ' |
6de1e2a9 ZW |
1532 | && (is_idchar[xbuf[2]] || xbuf[2] == '(' || xbuf[2] == '\'' |
1533 | || xbuf[2] == '\"')) | |
1534 | mbuf->cur += 2; | |
1535 | ||
1536 | /* Likewise, avoid the extra space at the end of the macro expansion | |
1537 | if this is safe. We can do a better job here since we can know | |
1538 | what the next char will be. */ | |
1539 | if (xbuf_len >= 3 | |
ed45de98 | 1540 | && mbuf->rlimit[-2] == '\r' |
6de1e2a9 ZW |
1541 | && mbuf->rlimit[-1] == ' ') |
1542 | { | |
1543 | int c1 = mbuf->rlimit[-3]; | |
1544 | int c2 = CPP_BUF_PEEK (CPP_PREV_BUFFER (CPP_BUFFER (pfile))); | |
1545 | if (c2 == EOF || !unsafe_chars (c1, c2)) | |
1546 | mbuf->rlimit -= 2; | |
1547 | } | |
1548 | } | |
1549 | ||
1550 | /* Return zero if two DEFINITIONs are isomorphic. */ | |
1551 | ||
1552 | int | |
1553 | compare_defs (pfile, d1, d2) | |
1554 | cpp_reader *pfile; | |
1555 | DEFINITION *d1, *d2; | |
1556 | { | |
1557 | register struct reflist *a1, *a2; | |
1558 | register U_CHAR *p1 = d1->expansion; | |
1559 | register U_CHAR *p2 = d2->expansion; | |
1560 | int first = 1; | |
1561 | ||
1562 | if (d1->nargs != d2->nargs) | |
1563 | return 1; | |
1564 | if (CPP_PEDANTIC (pfile) | |
1565 | && strcmp ((char *) d1->args.argnames, (char *) d2->args.argnames)) | |
1566 | return 1; | |
1567 | for (a1 = d1->pattern, a2 = d2->pattern; a1 && a2; | |
1568 | a1 = a1->next, a2 = a2->next) | |
1569 | { | |
1570 | if (!((a1->nchars == a2->nchars && !strncmp (p1, p2, a1->nchars)) | |
1571 | || !comp_def_part (first, p1, a1->nchars, p2, a2->nchars, 0)) | |
1572 | || a1->argno != a2->argno | |
1573 | || a1->stringify != a2->stringify | |
1574 | || a1->raw_before != a2->raw_before | |
1575 | || a1->raw_after != a2->raw_after) | |
1576 | return 1; | |
1577 | first = 0; | |
1578 | p1 += a1->nchars; | |
1579 | p2 += a2->nchars; | |
1580 | } | |
1581 | if (a1 != a2) | |
1582 | return 1; | |
1583 | ||
1584 | return comp_def_part (first, p1, d1->length - (p1 - d1->expansion), | |
1585 | p2, d2->length - (p2 - d2->expansion), 1); | |
1586 | } | |
1587 | ||
1588 | /* Return 1 if two parts of two macro definitions are effectively different. | |
1589 | One of the parts starts at BEG1 and has LEN1 chars; | |
1590 | the other has LEN2 chars at BEG2. | |
1591 | Any sequence of whitespace matches any other sequence of whitespace. | |
1592 | FIRST means these parts are the first of a macro definition; | |
1593 | so ignore leading whitespace entirely. | |
1594 | LAST means these parts are the last of a macro definition; | |
1595 | so ignore trailing whitespace entirely. */ | |
1596 | ||
1597 | static int | |
1598 | comp_def_part (first, beg1, len1, beg2, len2, last) | |
1599 | int first; | |
1600 | U_CHAR *beg1, *beg2; | |
1601 | int len1, len2; | |
1602 | int last; | |
1603 | { | |
1604 | register U_CHAR *end1 = beg1 + len1; | |
1605 | register U_CHAR *end2 = beg2 + len2; | |
1606 | if (first) | |
1607 | { | |
1608 | while (beg1 != end1 && is_space[*beg1]) | |
1609 | beg1++; | |
1610 | while (beg2 != end2 && is_space[*beg2]) | |
1611 | beg2++; | |
1612 | } | |
1613 | if (last) | |
1614 | { | |
1615 | while (beg1 != end1 && is_space[end1[-1]]) | |
1616 | end1--; | |
1617 | while (beg2 != end2 && is_space[end2[-1]]) | |
1618 | end2--; | |
1619 | } | |
1620 | while (beg1 != end1 && beg2 != end2) | |
1621 | { | |
1622 | if (is_space[*beg1] && is_space[*beg2]) | |
1623 | { | |
1624 | while (beg1 != end1 && is_space[*beg1]) | |
1625 | beg1++; | |
1626 | while (beg2 != end2 && is_space[*beg2]) | |
1627 | beg2++; | |
1628 | } | |
1629 | else if (*beg1 == *beg2) | |
1630 | { | |
1631 | beg1++; | |
1632 | beg2++; | |
1633 | } | |
1634 | else | |
1635 | break; | |
1636 | } | |
1637 | return (beg1 != end1) || (beg2 != end2); | |
1638 | } |