Commit | Line | Data |
---|---|---|
ec2bcbe7 | 1 | /* C preprocessor macro expansion for GDB. |
b811d2c2 | 2 | Copyright (C) 2002-2020 Free Software Foundation, Inc. |
ec2bcbe7 JB |
3 | Contributed by Red Hat, Inc. |
4 | ||
5 | This file is part of GDB. | |
6 | ||
7 | This program is free software; you can redistribute it and/or modify | |
8 | it under the terms of the GNU General Public License as published by | |
a9762ec7 | 9 | the Free Software Foundation; either version 3 of the License, or |
ec2bcbe7 JB |
10 | (at your option) any 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 | |
a9762ec7 | 18 | along with this program. If not, see <http://www.gnu.org/licenses/>. */ |
ec2bcbe7 JB |
19 | |
20 | #include "defs.h" | |
04ea0df1 | 21 | #include "gdb_obstack.h" |
ec2bcbe7 JB |
22 | #include "macrotab.h" |
23 | #include "macroexp.h" | |
211d5b1c | 24 | #include "macroscope.h" |
6c7a06a3 | 25 | #include "c-lang.h" |
ec2bcbe7 JB |
26 | |
27 | ||
28 | \f | |
29 | /* A resizeable, substringable string type. */ | |
30 | ||
31 | ||
32 | /* A string type that we can resize, quickly append to, and use to | |
33 | refer to substrings of other strings. */ | |
34 | struct macro_buffer | |
35 | { | |
36 | /* An array of characters. The first LEN bytes are the real text, | |
37 | but there are SIZE bytes allocated to the array. If SIZE is | |
38 | zero, then this doesn't point to a malloc'ed block. If SHARED is | |
39 | non-zero, then this buffer is actually a pointer into some larger | |
40 | string, and we shouldn't append characters to it, etc. Because | |
41 | of sharing, we can't assume in general that the text is | |
42 | null-terminated. */ | |
43 | char *text; | |
44 | ||
45 | /* The number of characters in the string. */ | |
46 | int len; | |
47 | ||
48 | /* The number of characters allocated to the string. If SHARED is | |
49 | non-zero, this is meaningless; in this case, we set it to zero so | |
50 | that any "do we have room to append something?" tests will fail, | |
51 | so we don't always have to check SHARED before using this field. */ | |
52 | int size; | |
53 | ||
54 | /* Zero if TEXT can be safely realloc'ed (i.e., it's its own malloc | |
55 | block). Non-zero if TEXT is actually pointing into the middle of | |
b38ef47f PA |
56 | some other block, or to a string literal, and we shouldn't |
57 | reallocate it. */ | |
58 | bool shared; | |
ec2bcbe7 JB |
59 | |
60 | /* For detecting token splicing. | |
61 | ||
62 | This is the index in TEXT of the first character of the token | |
63 | that abuts the end of TEXT. If TEXT contains no tokens, then we | |
64 | set this equal to LEN. If TEXT ends in whitespace, then there is | |
65 | no token abutting the end of TEXT (it's just whitespace), and | |
66 | again, we set this equal to LEN. We set this to -1 if we don't | |
67 | know the nature of TEXT. */ | |
1739cf24 | 68 | int last_token = -1; |
ec2bcbe7 JB |
69 | |
70 | /* If this buffer is holding the result from get_token, then this | |
71 | is non-zero if it is an identifier token, zero otherwise. */ | |
1739cf24 | 72 | int is_identifier = 0; |
ec2bcbe7 JB |
73 | |
74 | ||
1739cf24 TT |
75 | macro_buffer () |
76 | : text (NULL), | |
77 | len (0), | |
78 | size (0), | |
79 | shared (false) | |
80 | { | |
81 | } | |
ec2bcbe7 | 82 | |
1739cf24 TT |
83 | /* Set the macro buffer to the empty string, guessing that its |
84 | final contents will fit in N bytes. (It'll get resized if it | |
85 | doesn't, so the guess doesn't have to be right.) Allocate the | |
86 | initial storage with xmalloc. */ | |
87 | explicit macro_buffer (int n) | |
88 | : len (0), | |
89 | size (n), | |
90 | shared (false) | |
91 | { | |
92 | if (n > 0) | |
93 | text = (char *) xmalloc (n); | |
94 | else | |
95 | text = NULL; | |
96 | } | |
abc9d0dc | 97 | |
1739cf24 TT |
98 | /* Set the macro buffer to refer to the LEN bytes at ADDR, as a |
99 | shared substring. */ | |
100 | macro_buffer (const char *addr, int len) | |
101 | { | |
102 | set_shared (addr, len); | |
103 | } | |
ec2bcbe7 | 104 | |
1739cf24 TT |
105 | /* Set the macro buffer to refer to the LEN bytes at ADDR, as a |
106 | shared substring. */ | |
107 | void set_shared (const char *addr, int len_) | |
108 | { | |
109 | text = (char *) addr; | |
110 | len = len_; | |
111 | size = 0; | |
112 | shared = true; | |
113 | } | |
ec2bcbe7 | 114 | |
c530603c TV |
115 | macro_buffer& operator= (const macro_buffer &src) |
116 | { | |
117 | gdb_assert (src.shared); | |
118 | gdb_assert (shared); | |
119 | set_shared (src.text, src.len); | |
120 | last_token = src.last_token; | |
121 | is_identifier = src.is_identifier; | |
122 | return *this; | |
123 | } | |
124 | ||
1739cf24 TT |
125 | ~macro_buffer () |
126 | { | |
127 | if (! shared && size) | |
128 | xfree (text); | |
129 | } | |
ec2bcbe7 | 130 | |
14d960c8 SM |
131 | /* Release the text of the buffer to the caller. */ |
132 | gdb::unique_xmalloc_ptr<char> release () | |
1739cf24 TT |
133 | { |
134 | gdb_assert (! shared); | |
135 | gdb_assert (size); | |
136 | char *result = text; | |
137 | text = NULL; | |
14d960c8 | 138 | return gdb::unique_xmalloc_ptr<char> (result); |
1739cf24 | 139 | } |
ec2bcbe7 | 140 | |
1739cf24 TT |
141 | /* Resize the buffer to be at least N bytes long. Raise an error if |
142 | the buffer shouldn't be resized. */ | |
143 | void resize_buffer (int n) | |
144 | { | |
145 | /* We shouldn't be trying to resize shared strings. */ | |
146 | gdb_assert (! shared); | |
ec2bcbe7 | 147 | |
1739cf24 TT |
148 | if (size == 0) |
149 | size = n; | |
150 | else | |
151 | while (size <= n) | |
152 | size *= 2; | |
ec2bcbe7 | 153 | |
1739cf24 TT |
154 | text = (char *) xrealloc (text, size); |
155 | } | |
ec2bcbe7 | 156 | |
1739cf24 TT |
157 | /* Append the character C to the buffer. */ |
158 | void appendc (int c) | |
159 | { | |
160 | int new_len = len + 1; | |
ec2bcbe7 | 161 | |
1739cf24 TT |
162 | if (new_len > size) |
163 | resize_buffer (new_len); | |
ec2bcbe7 | 164 | |
1739cf24 TT |
165 | text[len] = c; |
166 | len = new_len; | |
167 | } | |
ec2bcbe7 | 168 | |
1739cf24 TT |
169 | /* Append the COUNT bytes at ADDR to the buffer. */ |
170 | void appendmem (const char *addr, int count) | |
171 | { | |
172 | int new_len = len + count; | |
ec2bcbe7 | 173 | |
1739cf24 TT |
174 | if (new_len > size) |
175 | resize_buffer (new_len); | |
ec2bcbe7 | 176 | |
1739cf24 TT |
177 | memcpy (text + len, addr, count); |
178 | len = new_len; | |
179 | } | |
180 | }; | |
ec2bcbe7 JB |
181 | |
182 | ||
183 | \f | |
184 | /* Recognizing preprocessor tokens. */ | |
185 | ||
186 | ||
d7d9f01e TT |
187 | int |
188 | macro_is_whitespace (int c) | |
ec2bcbe7 JB |
189 | { |
190 | return (c == ' ' | |
191 | || c == '\t' | |
192 | || c == '\n' | |
193 | || c == '\v' | |
194 | || c == '\f'); | |
195 | } | |
196 | ||
197 | ||
d7d9f01e TT |
198 | int |
199 | macro_is_digit (int c) | |
ec2bcbe7 JB |
200 | { |
201 | return ('0' <= c && c <= '9'); | |
202 | } | |
203 | ||
204 | ||
d7d9f01e TT |
205 | int |
206 | macro_is_identifier_nondigit (int c) | |
ec2bcbe7 JB |
207 | { |
208 | return (c == '_' | |
209 | || ('a' <= c && c <= 'z') | |
210 | || ('A' <= c && c <= 'Z')); | |
211 | } | |
212 | ||
213 | ||
214 | static void | |
215 | set_token (struct macro_buffer *tok, char *start, char *end) | |
216 | { | |
1739cf24 | 217 | tok->set_shared (start, end - start); |
ec2bcbe7 JB |
218 | tok->last_token = 0; |
219 | ||
025bb325 | 220 | /* Presumed; get_identifier may overwrite this. */ |
ec2bcbe7 JB |
221 | tok->is_identifier = 0; |
222 | } | |
223 | ||
224 | ||
225 | static int | |
226 | get_comment (struct macro_buffer *tok, char *p, char *end) | |
227 | { | |
228 | if (p + 2 > end) | |
229 | return 0; | |
230 | else if (p[0] == '/' | |
231 | && p[1] == '*') | |
232 | { | |
233 | char *tok_start = p; | |
234 | ||
235 | p += 2; | |
236 | ||
237 | for (; p < end; p++) | |
238 | if (p + 2 <= end | |
239 | && p[0] == '*' | |
240 | && p[1] == '/') | |
241 | { | |
242 | p += 2; | |
243 | set_token (tok, tok_start, p); | |
244 | return 1; | |
245 | } | |
246 | ||
8a3fe4f8 | 247 | error (_("Unterminated comment in macro expansion.")); |
ec2bcbe7 JB |
248 | } |
249 | else if (p[0] == '/' | |
250 | && p[1] == '/') | |
251 | { | |
252 | char *tok_start = p; | |
253 | ||
254 | p += 2; | |
255 | for (; p < end; p++) | |
256 | if (*p == '\n') | |
257 | break; | |
258 | ||
259 | set_token (tok, tok_start, p); | |
260 | return 1; | |
261 | } | |
262 | else | |
263 | return 0; | |
264 | } | |
265 | ||
266 | ||
267 | static int | |
268 | get_identifier (struct macro_buffer *tok, char *p, char *end) | |
269 | { | |
270 | if (p < end | |
d7d9f01e | 271 | && macro_is_identifier_nondigit (*p)) |
ec2bcbe7 JB |
272 | { |
273 | char *tok_start = p; | |
274 | ||
275 | while (p < end | |
d7d9f01e TT |
276 | && (macro_is_identifier_nondigit (*p) |
277 | || macro_is_digit (*p))) | |
ec2bcbe7 JB |
278 | p++; |
279 | ||
280 | set_token (tok, tok_start, p); | |
281 | tok->is_identifier = 1; | |
282 | return 1; | |
283 | } | |
284 | else | |
285 | return 0; | |
286 | } | |
287 | ||
288 | ||
289 | static int | |
290 | get_pp_number (struct macro_buffer *tok, char *p, char *end) | |
291 | { | |
292 | if (p < end | |
d7d9f01e | 293 | && (macro_is_digit (*p) |
17c8aaf5 TT |
294 | || (*p == '.' |
295 | && p + 2 <= end | |
296 | && macro_is_digit (p[1])))) | |
ec2bcbe7 JB |
297 | { |
298 | char *tok_start = p; | |
299 | ||
300 | while (p < end) | |
301 | { | |
17c8aaf5 TT |
302 | if (p + 2 <= end |
303 | && strchr ("eEpP", *p) | |
304 | && (p[1] == '+' || p[1] == '-')) | |
ec2bcbe7 | 305 | p += 2; |
17c8aaf5 TT |
306 | else if (macro_is_digit (*p) |
307 | || macro_is_identifier_nondigit (*p) | |
308 | || *p == '.') | |
309 | p++; | |
ec2bcbe7 JB |
310 | else |
311 | break; | |
312 | } | |
313 | ||
314 | set_token (tok, tok_start, p); | |
315 | return 1; | |
316 | } | |
317 | else | |
318 | return 0; | |
319 | } | |
320 | ||
321 | ||
322 | ||
323 | /* If the text starting at P going up to (but not including) END | |
324 | starts with a character constant, set *TOK to point to that | |
325 | character constant, and return 1. Otherwise, return zero. | |
326 | Signal an error if it contains a malformed or incomplete character | |
327 | constant. */ | |
328 | static int | |
329 | get_character_constant (struct macro_buffer *tok, char *p, char *end) | |
330 | { | |
331 | /* ISO/IEC 9899:1999 (E) Section 6.4.4.4 paragraph 1 | |
332 | But of course, what really matters is that we handle it the same | |
333 | way GDB's C/C++ lexer does. So we call parse_escape in utils.c | |
334 | to handle escape sequences. */ | |
335 | if ((p + 1 <= end && *p == '\'') | |
6c7a06a3 TT |
336 | || (p + 2 <= end |
337 | && (p[0] == 'L' || p[0] == 'u' || p[0] == 'U') | |
338 | && p[1] == '\'')) | |
ec2bcbe7 JB |
339 | { |
340 | char *tok_start = p; | |
6c7a06a3 | 341 | int char_count = 0; |
ec2bcbe7 JB |
342 | |
343 | if (*p == '\'') | |
344 | p++; | |
6c7a06a3 | 345 | else if (*p == 'L' || *p == 'u' || *p == 'U') |
ec2bcbe7 JB |
346 | p += 2; |
347 | else | |
f3574227 | 348 | gdb_assert_not_reached ("unexpected character constant"); |
ec2bcbe7 | 349 | |
ec2bcbe7 JB |
350 | for (;;) |
351 | { | |
352 | if (p >= end) | |
8a3fe4f8 | 353 | error (_("Unmatched single quote.")); |
ec2bcbe7 JB |
354 | else if (*p == '\'') |
355 | { | |
6c7a06a3 | 356 | if (!char_count) |
8a3fe4f8 AC |
357 | error (_("A character constant must contain at least one " |
358 | "character.")); | |
ec2bcbe7 JB |
359 | p++; |
360 | break; | |
361 | } | |
362 | else if (*p == '\\') | |
363 | { | |
d7561cbb KS |
364 | const char *s, *o; |
365 | ||
366 | s = o = ++p; | |
367 | char_count += c_parse_escape (&s, NULL); | |
368 | p += s - o; | |
ec2bcbe7 JB |
369 | } |
370 | else | |
6c7a06a3 TT |
371 | { |
372 | p++; | |
373 | char_count++; | |
374 | } | |
ec2bcbe7 JB |
375 | } |
376 | ||
377 | set_token (tok, tok_start, p); | |
378 | return 1; | |
379 | } | |
380 | else | |
381 | return 0; | |
382 | } | |
383 | ||
384 | ||
385 | /* If the text starting at P going up to (but not including) END | |
386 | starts with a string literal, set *TOK to point to that string | |
387 | literal, and return 1. Otherwise, return zero. Signal an error if | |
388 | it contains a malformed or incomplete string literal. */ | |
389 | static int | |
390 | get_string_literal (struct macro_buffer *tok, char *p, char *end) | |
391 | { | |
392 | if ((p + 1 <= end | |
6c7a06a3 | 393 | && *p == '"') |
ec2bcbe7 | 394 | || (p + 2 <= end |
6c7a06a3 TT |
395 | && (p[0] == 'L' || p[0] == 'u' || p[0] == 'U') |
396 | && p[1] == '"')) | |
ec2bcbe7 JB |
397 | { |
398 | char *tok_start = p; | |
399 | ||
6c7a06a3 | 400 | if (*p == '"') |
ec2bcbe7 | 401 | p++; |
6c7a06a3 | 402 | else if (*p == 'L' || *p == 'u' || *p == 'U') |
ec2bcbe7 JB |
403 | p += 2; |
404 | else | |
f3574227 | 405 | gdb_assert_not_reached ("unexpected string literal"); |
ec2bcbe7 JB |
406 | |
407 | for (;;) | |
408 | { | |
409 | if (p >= end) | |
8a3fe4f8 | 410 | error (_("Unterminated string in expression.")); |
6c7a06a3 | 411 | else if (*p == '"') |
ec2bcbe7 JB |
412 | { |
413 | p++; | |
414 | break; | |
415 | } | |
416 | else if (*p == '\n') | |
8a3fe4f8 AC |
417 | error (_("Newline characters may not appear in string " |
418 | "constants.")); | |
ec2bcbe7 JB |
419 | else if (*p == '\\') |
420 | { | |
d7561cbb KS |
421 | const char *s, *o; |
422 | ||
423 | s = o = ++p; | |
424 | c_parse_escape (&s, NULL); | |
425 | p += s - o; | |
ec2bcbe7 JB |
426 | } |
427 | else | |
428 | p++; | |
429 | } | |
430 | ||
431 | set_token (tok, tok_start, p); | |
432 | return 1; | |
433 | } | |
434 | else | |
435 | return 0; | |
436 | } | |
437 | ||
438 | ||
439 | static int | |
440 | get_punctuator (struct macro_buffer *tok, char *p, char *end) | |
441 | { | |
442 | /* Here, speed is much less important than correctness and clarity. */ | |
443 | ||
ccb3ac8a TT |
444 | /* ISO/IEC 9899:1999 (E) Section 6.4.6 Paragraph 1. |
445 | Note that this table is ordered in a special way. A punctuator | |
446 | which is a prefix of another punctuator must appear after its | |
447 | "extension". Otherwise, the wrong token will be returned. */ | |
ec2bcbe7 | 448 | static const char * const punctuators[] = { |
ccb3ac8a TT |
449 | "[", "]", "(", ")", "{", "}", "?", ";", ",", "~", |
450 | "...", ".", | |
451 | "->", "--", "-=", "-", | |
452 | "++", "+=", "+", | |
453 | "*=", "*", | |
454 | "!=", "!", | |
455 | "&&", "&=", "&", | |
456 | "/=", "/", | |
457 | "%>", "%:%:", "%:", "%=", "%", | |
458 | "^=", "^", | |
459 | "##", "#", | |
460 | ":>", ":", | |
461 | "||", "|=", "|", | |
462 | "<<=", "<<", "<=", "<:", "<%", "<", | |
463 | ">>=", ">>", ">=", ">", | |
464 | "==", "=", | |
ec2bcbe7 JB |
465 | 0 |
466 | }; | |
467 | ||
468 | int i; | |
469 | ||
470 | if (p + 1 <= end) | |
471 | { | |
472 | for (i = 0; punctuators[i]; i++) | |
473 | { | |
474 | const char *punctuator = punctuators[i]; | |
475 | ||
476 | if (p[0] == punctuator[0]) | |
477 | { | |
478 | int len = strlen (punctuator); | |
479 | ||
480 | if (p + len <= end | |
481 | && ! memcmp (p, punctuator, len)) | |
482 | { | |
483 | set_token (tok, p, p + len); | |
484 | return 1; | |
485 | } | |
486 | } | |
487 | } | |
488 | } | |
489 | ||
490 | return 0; | |
491 | } | |
492 | ||
493 | ||
494 | /* Peel the next preprocessor token off of SRC, and put it in TOK. | |
495 | Mutate TOK to refer to the first token in SRC, and mutate SRC to | |
496 | refer to the text after that token. SRC must be a shared buffer; | |
497 | the resulting TOK will be shared, pointing into the same string SRC | |
498 | does. Initialize TOK's last_token field. Return non-zero if we | |
499 | succeed, or 0 if we didn't find any more tokens in SRC. */ | |
500 | static int | |
501 | get_token (struct macro_buffer *tok, | |
502 | struct macro_buffer *src) | |
503 | { | |
504 | char *p = src->text; | |
505 | char *end = p + src->len; | |
506 | ||
507 | gdb_assert (src->shared); | |
508 | ||
509 | /* From the ISO C standard, ISO/IEC 9899:1999 (E), section 6.4: | |
510 | ||
511 | preprocessing-token: | |
512 | header-name | |
513 | identifier | |
514 | pp-number | |
515 | character-constant | |
516 | string-literal | |
517 | punctuator | |
518 | each non-white-space character that cannot be one of the above | |
519 | ||
520 | We don't have to deal with header-name tokens, since those can | |
521 | only occur after a #include, which we will never see. */ | |
522 | ||
523 | while (p < end) | |
d7d9f01e | 524 | if (macro_is_whitespace (*p)) |
ec2bcbe7 JB |
525 | p++; |
526 | else if (get_comment (tok, p, end)) | |
527 | p += tok->len; | |
528 | else if (get_pp_number (tok, p, end) | |
529 | || get_character_constant (tok, p, end) | |
530 | || get_string_literal (tok, p, end) | |
531 | /* Note: the grammar in the standard seems to be | |
532 | ambiguous: L'x' can be either a wide character | |
533 | constant, or an identifier followed by a normal | |
534 | character constant. By trying `get_identifier' after | |
535 | we try get_character_constant and get_string_literal, | |
536 | we give the wide character syntax precedence. Now, | |
537 | since GDB doesn't handle wide character constants | |
538 | anyway, is this the right thing to do? */ | |
539 | || get_identifier (tok, p, end) | |
540 | || get_punctuator (tok, p, end)) | |
541 | { | |
542 | /* How many characters did we consume, including whitespace? */ | |
543 | int consumed = p - src->text + tok->len; | |
b8d56208 | 544 | |
ec2bcbe7 JB |
545 | src->text += consumed; |
546 | src->len -= consumed; | |
547 | return 1; | |
548 | } | |
549 | else | |
550 | { | |
551 | /* We have found a "non-whitespace character that cannot be | |
552 | one of the above." Make a token out of it. */ | |
553 | int consumed; | |
554 | ||
555 | set_token (tok, p, p + 1); | |
556 | consumed = p - src->text + tok->len; | |
557 | src->text += consumed; | |
558 | src->len -= consumed; | |
559 | return 1; | |
560 | } | |
561 | ||
562 | return 0; | |
563 | } | |
564 | ||
565 | ||
566 | \f | |
567 | /* Appending token strings, with and without splicing */ | |
568 | ||
569 | ||
570 | /* Append the macro buffer SRC to the end of DEST, and ensure that | |
571 | doing so doesn't splice the token at the end of SRC with the token | |
572 | at the beginning of DEST. SRC and DEST must have their last_token | |
573 | fields set. Upon return, DEST's last_token field is set correctly. | |
574 | ||
575 | For example: | |
576 | ||
577 | If DEST is "(" and SRC is "y", then we can return with | |
578 | DEST set to "(y" --- we've simply appended the two buffers. | |
579 | ||
580 | However, if DEST is "x" and SRC is "y", then we must not return | |
581 | with DEST set to "xy" --- that would splice the two tokens "x" and | |
582 | "y" together to make a single token "xy". However, it would be | |
583 | fine to return with DEST set to "x y". Similarly, "<" and "<" must | |
584 | yield "< <", not "<<", etc. */ | |
585 | static void | |
586 | append_tokens_without_splicing (struct macro_buffer *dest, | |
587 | struct macro_buffer *src) | |
588 | { | |
589 | int original_dest_len = dest->len; | |
590 | struct macro_buffer dest_tail, new_token; | |
591 | ||
592 | gdb_assert (src->last_token != -1); | |
593 | gdb_assert (dest->last_token != -1); | |
594 | ||
595 | /* First, just try appending the two, and call get_token to see if | |
596 | we got a splice. */ | |
1739cf24 | 597 | dest->appendmem (src->text, src->len); |
ec2bcbe7 JB |
598 | |
599 | /* If DEST originally had no token abutting its end, then we can't | |
600 | have spliced anything, so we're done. */ | |
601 | if (dest->last_token == original_dest_len) | |
602 | { | |
603 | dest->last_token = original_dest_len + src->last_token; | |
604 | return; | |
605 | } | |
606 | ||
607 | /* Set DEST_TAIL to point to the last token in DEST, followed by | |
608 | all the stuff we just appended. */ | |
1739cf24 TT |
609 | dest_tail.set_shared (dest->text + dest->last_token, |
610 | dest->len - dest->last_token); | |
ec2bcbe7 JB |
611 | |
612 | /* Re-parse DEST's last token. We know that DEST used to contain | |
613 | at least one token, so if it doesn't contain any after the | |
614 | append, then we must have spliced "/" and "*" or "/" and "/" to | |
615 | make a comment start. (Just for the record, I got this right | |
616 | the first time. This is not a bug fix.) */ | |
617 | if (get_token (&new_token, &dest_tail) | |
618 | && (new_token.text + new_token.len | |
619 | == dest->text + original_dest_len)) | |
620 | { | |
621 | /* No splice, so we're done. */ | |
622 | dest->last_token = original_dest_len + src->last_token; | |
623 | return; | |
624 | } | |
625 | ||
626 | /* Okay, a simple append caused a splice. Let's chop dest back to | |
627 | its original length and try again, but separate the texts with a | |
628 | space. */ | |
629 | dest->len = original_dest_len; | |
1739cf24 TT |
630 | dest->appendc (' '); |
631 | dest->appendmem (src->text, src->len); | |
ec2bcbe7 | 632 | |
1739cf24 TT |
633 | dest_tail.set_shared (dest->text + dest->last_token, |
634 | dest->len - dest->last_token); | |
ec2bcbe7 JB |
635 | |
636 | /* Try to re-parse DEST's last token, as above. */ | |
637 | if (get_token (&new_token, &dest_tail) | |
638 | && (new_token.text + new_token.len | |
639 | == dest->text + original_dest_len)) | |
640 | { | |
641 | /* No splice, so we're done. */ | |
642 | dest->last_token = original_dest_len + 1 + src->last_token; | |
643 | return; | |
644 | } | |
645 | ||
646 | /* As far as I know, there's no case where inserting a space isn't | |
647 | enough to prevent a splice. */ | |
648 | internal_error (__FILE__, __LINE__, | |
e2e0b3e5 | 649 | _("unable to avoid splicing tokens during macro expansion")); |
ec2bcbe7 JB |
650 | } |
651 | ||
2fae03e8 TT |
652 | /* Stringify an argument, and insert it into DEST. ARG is the text to |
653 | stringify; it is LEN bytes long. */ | |
654 | ||
655 | static void | |
abc9d0dc | 656 | stringify (struct macro_buffer *dest, const char *arg, int len) |
2fae03e8 TT |
657 | { |
658 | /* Trim initial whitespace from ARG. */ | |
659 | while (len > 0 && macro_is_whitespace (*arg)) | |
660 | { | |
661 | ++arg; | |
662 | --len; | |
663 | } | |
664 | ||
665 | /* Trim trailing whitespace from ARG. */ | |
666 | while (len > 0 && macro_is_whitespace (arg[len - 1])) | |
667 | --len; | |
668 | ||
669 | /* Insert the string. */ | |
1739cf24 | 670 | dest->appendc ('"'); |
2fae03e8 TT |
671 | while (len > 0) |
672 | { | |
673 | /* We could try to handle strange cases here, like control | |
674 | characters, but there doesn't seem to be much point. */ | |
675 | if (macro_is_whitespace (*arg)) | |
676 | { | |
677 | /* Replace a sequence of whitespace with a single space. */ | |
1739cf24 | 678 | dest->appendc (' '); |
2fae03e8 TT |
679 | while (len > 1 && macro_is_whitespace (arg[1])) |
680 | { | |
681 | ++arg; | |
682 | --len; | |
683 | } | |
684 | } | |
685 | else if (*arg == '\\' || *arg == '"') | |
686 | { | |
1739cf24 TT |
687 | dest->appendc ('\\'); |
688 | dest->appendc (*arg); | |
2fae03e8 TT |
689 | } |
690 | else | |
1739cf24 | 691 | dest->appendc (*arg); |
2fae03e8 TT |
692 | ++arg; |
693 | --len; | |
694 | } | |
1739cf24 | 695 | dest->appendc ('"'); |
2fae03e8 TT |
696 | dest->last_token = dest->len; |
697 | } | |
ec2bcbe7 | 698 | |
abc9d0dc TT |
699 | /* See macroexp.h. */ |
700 | ||
a36158ec | 701 | gdb::unique_xmalloc_ptr<char> |
abc9d0dc TT |
702 | macro_stringify (const char *str) |
703 | { | |
abc9d0dc | 704 | int len = strlen (str); |
1739cf24 | 705 | struct macro_buffer buffer (len); |
abc9d0dc | 706 | |
abc9d0dc | 707 | stringify (&buffer, str, len); |
1739cf24 | 708 | buffer.appendc ('\0'); |
abc9d0dc | 709 | |
a36158ec | 710 | return buffer.release (); |
abc9d0dc TT |
711 | } |
712 | ||
ec2bcbe7 JB |
713 | \f |
714 | /* Expanding macros! */ | |
715 | ||
716 | ||
717 | /* A singly-linked list of the names of the macros we are currently | |
718 | expanding --- for detecting expansion loops. */ | |
719 | struct macro_name_list { | |
720 | const char *name; | |
721 | struct macro_name_list *next; | |
722 | }; | |
723 | ||
724 | ||
725 | /* Return non-zero if we are currently expanding the macro named NAME, | |
726 | according to LIST; otherwise, return zero. | |
727 | ||
728 | You know, it would be possible to get rid of all the NO_LOOP | |
729 | arguments to these functions by simply generating a new lookup | |
730 | function and baton which refuses to find the definition for a | |
731 | particular macro, and otherwise delegates the decision to another | |
732 | function/baton pair. But that makes the linked list of excluded | |
733 | macros chained through untyped baton pointers, which will make it | |
025bb325 | 734 | harder to debug. :( */ |
ec2bcbe7 JB |
735 | static int |
736 | currently_rescanning (struct macro_name_list *list, const char *name) | |
737 | { | |
738 | for (; list; list = list->next) | |
a86bc61c | 739 | if (strcmp (name, list->name) == 0) |
ec2bcbe7 JB |
740 | return 1; |
741 | ||
742 | return 0; | |
743 | } | |
744 | ||
745 | ||
746 | /* Gather the arguments to a macro expansion. | |
747 | ||
748 | NAME is the name of the macro being invoked. (It's only used for | |
749 | printing error messages.) | |
750 | ||
751 | Assume that SRC is the text of the macro invocation immediately | |
752 | following the macro name. For example, if we're processing the | |
753 | text foo(bar, baz), then NAME would be foo and SRC will be (bar, | |
754 | baz). | |
755 | ||
756 | If SRC doesn't start with an open paren ( token at all, return | |
1739cf24 | 757 | false, leave SRC unchanged, and don't set *ARGS_PTR to anything. |
ec2bcbe7 JB |
758 | |
759 | If SRC doesn't contain a properly terminated argument list, then | |
760 | raise an error. | |
1739cf24 | 761 | |
2fae03e8 TT |
762 | For a variadic macro, NARGS holds the number of formal arguments to |
763 | the macro. For a GNU-style variadic macro, this should be the | |
764 | number of named arguments. For a non-variadic macro, NARGS should | |
765 | be -1. | |
ec2bcbe7 | 766 | |
1739cf24 TT |
767 | Otherwise, return true and set *ARGS_PTR to a vector of macro |
768 | buffers referring to the argument texts. The macro buffers share | |
769 | their text with SRC, and their last_token fields are initialized. | |
ec2bcbe7 JB |
770 | |
771 | NOTE WELL: if SRC starts with a open paren ( token followed | |
772 | immediately by a close paren ) token (e.g., the invocation looks | |
773 | like "foo()"), we treat that as one argument, which happens to be | |
774 | the empty list of tokens. The caller should keep in mind that such | |
775 | a sequence of tokens is a valid way to invoke one-parameter | |
776 | function-like macros, but also a valid way to invoke zero-parameter | |
777 | function-like macros. Eeew. | |
778 | ||
779 | Consume the tokens from SRC; after this call, SRC contains the text | |
780 | following the invocation. */ | |
781 | ||
1739cf24 TT |
782 | static bool |
783 | gather_arguments (const char *name, struct macro_buffer *src, int nargs, | |
784 | std::vector<struct macro_buffer> *args_ptr) | |
ec2bcbe7 JB |
785 | { |
786 | struct macro_buffer tok; | |
1739cf24 | 787 | std::vector<struct macro_buffer> args; |
ec2bcbe7 JB |
788 | |
789 | /* Does SRC start with an opening paren token? Read from a copy of | |
790 | SRC, so SRC itself is unaffected if we don't find an opening | |
791 | paren. */ | |
792 | { | |
1739cf24 | 793 | struct macro_buffer temp (src->text, src->len); |
ec2bcbe7 JB |
794 | |
795 | if (! get_token (&tok, &temp) | |
796 | || tok.len != 1 | |
797 | || tok.text[0] != '(') | |
1739cf24 | 798 | return false; |
ec2bcbe7 JB |
799 | } |
800 | ||
801 | /* Consume SRC's opening paren. */ | |
802 | get_token (&tok, src); | |
803 | ||
ec2bcbe7 JB |
804 | for (;;) |
805 | { | |
806 | struct macro_buffer *arg; | |
807 | int depth; | |
808 | ||
ec2bcbe7 | 809 | /* Initialize the next argument. */ |
1739cf24 TT |
810 | args.emplace_back (); |
811 | arg = &args.back (); | |
ec2bcbe7 JB |
812 | set_token (arg, src->text, src->text); |
813 | ||
814 | /* Gather the argument's tokens. */ | |
815 | depth = 0; | |
816 | for (;;) | |
817 | { | |
ec2bcbe7 | 818 | if (! get_token (&tok, src)) |
8a3fe4f8 | 819 | error (_("Malformed argument list for macro `%s'."), name); |
1739cf24 | 820 | |
ec2bcbe7 JB |
821 | /* Is tok an opening paren? */ |
822 | if (tok.len == 1 && tok.text[0] == '(') | |
823 | depth++; | |
824 | ||
825 | /* Is tok is a closing paren? */ | |
826 | else if (tok.len == 1 && tok.text[0] == ')') | |
827 | { | |
828 | /* If it's a closing paren at the top level, then that's | |
829 | the end of the argument list. */ | |
830 | if (depth == 0) | |
831 | { | |
2fae03e8 TT |
832 | /* In the varargs case, the last argument may be |
833 | missing. Add an empty argument in this case. */ | |
1739cf24 | 834 | if (nargs != -1 && args.size () == nargs - 1) |
2fae03e8 | 835 | { |
1739cf24 TT |
836 | args.emplace_back (); |
837 | arg = &args.back (); | |
2fae03e8 TT |
838 | set_token (arg, src->text, src->text); |
839 | } | |
840 | ||
1739cf24 TT |
841 | *args_ptr = std::move (args); |
842 | return true; | |
ec2bcbe7 JB |
843 | } |
844 | ||
845 | depth--; | |
846 | } | |
847 | ||
848 | /* If tok is a comma at top level, then that's the end of | |
2fae03e8 TT |
849 | the current argument. However, if we are handling a |
850 | variadic macro and we are computing the last argument, we | |
851 | want to include the comma and remaining tokens. */ | |
852 | else if (tok.len == 1 && tok.text[0] == ',' && depth == 0 | |
1739cf24 | 853 | && (nargs == -1 || args.size () < nargs)) |
ec2bcbe7 JB |
854 | break; |
855 | ||
856 | /* Extend the current argument to enclose this token. If | |
857 | this is the current argument's first token, leave out any | |
858 | leading whitespace, just for aesthetics. */ | |
859 | if (arg->len == 0) | |
860 | { | |
861 | arg->text = tok.text; | |
862 | arg->len = tok.len; | |
863 | arg->last_token = 0; | |
864 | } | |
865 | else | |
866 | { | |
867 | arg->len = (tok.text + tok.len) - arg->text; | |
868 | arg->last_token = tok.text - arg->text; | |
869 | } | |
870 | } | |
871 | } | |
872 | } | |
873 | ||
874 | ||
875 | /* The `expand' and `substitute_args' functions both invoke `scan' | |
876 | recursively, so we need a forward declaration somewhere. */ | |
877 | static void scan (struct macro_buffer *dest, | |
878 | struct macro_buffer *src, | |
879 | struct macro_name_list *no_loop, | |
211d5b1c | 880 | const macro_scope &scope); |
ec2bcbe7 | 881 | |
2fae03e8 TT |
882 | /* A helper function for substitute_args. |
883 | ||
884 | ARGV is a vector of all the arguments; ARGC is the number of | |
885 | arguments. IS_VARARGS is true if the macro being substituted is a | |
886 | varargs macro; in this case VA_ARG_NAME is the name of the | |
887 | "variable" argument. VA_ARG_NAME is ignored if IS_VARARGS is | |
888 | false. | |
889 | ||
890 | If the token TOK is the name of a parameter, return the parameter's | |
891 | index. If TOK is not an argument, return -1. */ | |
892 | ||
893 | static int | |
894 | find_parameter (const struct macro_buffer *tok, | |
895 | int is_varargs, const struct macro_buffer *va_arg_name, | |
896 | int argc, const char * const *argv) | |
897 | { | |
898 | int i; | |
899 | ||
900 | if (! tok->is_identifier) | |
901 | return -1; | |
902 | ||
903 | for (i = 0; i < argc; ++i) | |
3e43a32a MS |
904 | if (tok->len == strlen (argv[i]) |
905 | && !memcmp (tok->text, argv[i], tok->len)) | |
2fae03e8 TT |
906 | return i; |
907 | ||
908 | if (is_varargs && tok->len == va_arg_name->len | |
909 | && ! memcmp (tok->text, va_arg_name->text, tok->len)) | |
910 | return argc - 1; | |
911 | ||
912 | return -1; | |
913 | } | |
914 | ||
a9bbfbd8 TT |
915 | /* Helper function for substitute_args that gets the next token and |
916 | updates the passed-in state variables. */ | |
917 | ||
918 | static void | |
919 | get_next_token_for_substitution (struct macro_buffer *replacement_list, | |
920 | struct macro_buffer *token, | |
921 | char **start, | |
922 | struct macro_buffer *lookahead, | |
923 | char **lookahead_start, | |
924 | int *lookahead_valid, | |
925 | bool *keep_going) | |
926 | { | |
927 | if (!*lookahead_valid) | |
928 | *keep_going = false; | |
929 | else | |
930 | { | |
931 | *keep_going = true; | |
932 | *token = *lookahead; | |
933 | *start = *lookahead_start; | |
934 | *lookahead_start = replacement_list->text; | |
935 | *lookahead_valid = get_token (lookahead, replacement_list); | |
936 | } | |
937 | } | |
938 | ||
ec2bcbe7 | 939 | /* Given the macro definition DEF, being invoked with the actual |
1739cf24 | 940 | arguments given by ARGV, substitute the arguments into the |
ec2bcbe7 JB |
941 | replacement list, and store the result in DEST. |
942 | ||
2fae03e8 TT |
943 | IS_VARARGS should be true if DEF is a varargs macro. In this case, |
944 | VA_ARG_NAME should be the name of the "variable" argument -- either | |
945 | __VA_ARGS__ for c99-style varargs, or the final argument name, for | |
946 | GNU-style varargs. If IS_VARARGS is false, this parameter is | |
947 | ignored. | |
948 | ||
ec2bcbe7 JB |
949 | If it is necessary to expand macro invocations in one of the |
950 | arguments, use LOOKUP_FUNC and LOOKUP_BATON to find the macro | |
951 | definitions, and don't expand invocations of the macros listed in | |
952 | NO_LOOP. */ | |
2fae03e8 | 953 | |
ec2bcbe7 | 954 | static void |
1739cf24 | 955 | substitute_args (struct macro_buffer *dest, |
ec2bcbe7 | 956 | struct macro_definition *def, |
2fae03e8 | 957 | int is_varargs, const struct macro_buffer *va_arg_name, |
1739cf24 | 958 | const std::vector<struct macro_buffer> &argv, |
ec2bcbe7 | 959 | struct macro_name_list *no_loop, |
211d5b1c | 960 | const macro_scope &scope) |
ec2bcbe7 | 961 | { |
2fae03e8 TT |
962 | /* The token we are currently considering. */ |
963 | struct macro_buffer tok; | |
964 | /* The replacement list's pointer from just before TOK was lexed. */ | |
965 | char *original_rl_start; | |
966 | /* We have a single lookahead token to handle token splicing. */ | |
967 | struct macro_buffer lookahead; | |
968 | /* The lookahead token might not be valid. */ | |
969 | int lookahead_valid; | |
970 | /* The replacement list's pointer from just before LOOKAHEAD was | |
971 | lexed. */ | |
972 | char *lookahead_rl_start; | |
ec2bcbe7 | 973 | |
1739cf24 TT |
974 | /* A macro buffer for the macro's replacement list. */ |
975 | struct macro_buffer replacement_list (def->replacement, | |
976 | strlen (def->replacement)); | |
ec2bcbe7 JB |
977 | |
978 | gdb_assert (dest->len == 0); | |
979 | dest->last_token = 0; | |
980 | ||
2fae03e8 TT |
981 | original_rl_start = replacement_list.text; |
982 | if (! get_token (&tok, &replacement_list)) | |
983 | return; | |
984 | lookahead_rl_start = replacement_list.text; | |
985 | lookahead_valid = get_token (&lookahead, &replacement_list); | |
986 | ||
a9bbfbd8 TT |
987 | /* __VA_OPT__ state variable. The states are: |
988 | 0 - nothing happening | |
989 | 1 - saw __VA_OPT__ | |
990 | >= 2 in __VA_OPT__, the value encodes the parenthesis depth. */ | |
991 | unsigned vaopt_state = 0; | |
992 | ||
993 | for (bool keep_going = true; | |
994 | keep_going; | |
995 | get_next_token_for_substitution (&replacement_list, | |
996 | &tok, | |
997 | &original_rl_start, | |
998 | &lookahead, | |
999 | &lookahead_rl_start, | |
1000 | &lookahead_valid, | |
1001 | &keep_going)) | |
ec2bcbe7 | 1002 | { |
a9bbfbd8 TT |
1003 | bool token_is_vaopt = (tok.len == 10 |
1004 | && strncmp (tok.text, "__VA_OPT__", 10) == 0); | |
1005 | ||
1006 | if (vaopt_state > 0) | |
1007 | { | |
1008 | if (token_is_vaopt) | |
1009 | error (_("__VA_OPT__ cannot appear inside __VA_OPT__")); | |
1010 | else if (tok.len == 1 && tok.text[0] == '(') | |
1011 | { | |
1012 | ++vaopt_state; | |
1013 | /* We just entered __VA_OPT__, so don't emit this | |
1014 | token. */ | |
1015 | continue; | |
1016 | } | |
1017 | else if (vaopt_state == 1) | |
1018 | error (_("__VA_OPT__ must be followed by an open parenthesis")); | |
1019 | else if (tok.len == 1 && tok.text[0] == ')') | |
1020 | { | |
1021 | --vaopt_state; | |
1022 | if (vaopt_state == 1) | |
1023 | { | |
1024 | /* Done with __VA_OPT__. */ | |
1025 | vaopt_state = 0; | |
1026 | /* Don't emit. */ | |
1027 | continue; | |
1028 | } | |
1029 | } | |
1030 | ||
1031 | /* If __VA_ARGS__ is empty, then drop the contents of | |
1032 | __VA_OPT__. */ | |
1739cf24 | 1033 | if (argv.back ().len == 0) |
a9bbfbd8 TT |
1034 | continue; |
1035 | } | |
1036 | else if (token_is_vaopt) | |
1037 | { | |
1038 | if (!is_varargs) | |
1039 | error (_("__VA_OPT__ is only valid in a variadic macro")); | |
1040 | vaopt_state = 1; | |
1041 | /* Don't emit this token. */ | |
1042 | continue; | |
1043 | } | |
1044 | ||
ec2bcbe7 JB |
1045 | /* Just for aesthetics. If we skipped some whitespace, copy |
1046 | that to DEST. */ | |
1047 | if (tok.text > original_rl_start) | |
1048 | { | |
1739cf24 | 1049 | dest->appendmem (original_rl_start, tok.text - original_rl_start); |
ec2bcbe7 JB |
1050 | dest->last_token = dest->len; |
1051 | } | |
1052 | ||
1053 | /* Is this token the stringification operator? */ | |
1054 | if (tok.len == 1 | |
1055 | && tok.text[0] == '#') | |
2fae03e8 TT |
1056 | { |
1057 | int arg; | |
ec2bcbe7 | 1058 | |
2fae03e8 TT |
1059 | if (!lookahead_valid) |
1060 | error (_("Stringification operator requires an argument.")); | |
ec2bcbe7 | 1061 | |
2fae03e8 TT |
1062 | arg = find_parameter (&lookahead, is_varargs, va_arg_name, |
1063 | def->argc, def->argv); | |
1064 | if (arg == -1) | |
1065 | error (_("Argument to stringification operator must name " | |
1066 | "a macro parameter.")); | |
ec2bcbe7 | 1067 | |
2fae03e8 TT |
1068 | stringify (dest, argv[arg].text, argv[arg].len); |
1069 | ||
1070 | /* Read one token and let the loop iteration code handle the | |
1071 | rest. */ | |
1072 | lookahead_rl_start = replacement_list.text; | |
1073 | lookahead_valid = get_token (&lookahead, &replacement_list); | |
1074 | } | |
1075 | /* Is this token the splicing operator? */ | |
1076 | else if (tok.len == 2 | |
1077 | && tok.text[0] == '#' | |
1078 | && tok.text[1] == '#') | |
1079 | error (_("Stray splicing operator")); | |
1080 | /* Is the next token the splicing operator? */ | |
1081 | else if (lookahead_valid | |
1082 | && lookahead.len == 2 | |
1083 | && lookahead.text[0] == '#' | |
1084 | && lookahead.text[1] == '#') | |
1085 | { | |
308d96ed | 1086 | int finished = 0; |
2fae03e8 TT |
1087 | int prev_was_comma = 0; |
1088 | ||
1089 | /* Note that GCC warns if the result of splicing is not a | |
1090 | token. In the debugger there doesn't seem to be much | |
1091 | benefit from doing this. */ | |
1092 | ||
1093 | /* Insert the first token. */ | |
1094 | if (tok.len == 1 && tok.text[0] == ',') | |
1095 | prev_was_comma = 1; | |
1096 | else | |
1097 | { | |
1098 | int arg = find_parameter (&tok, is_varargs, va_arg_name, | |
1099 | def->argc, def->argv); | |
b8d56208 | 1100 | |
2fae03e8 | 1101 | if (arg != -1) |
1739cf24 | 1102 | dest->appendmem (argv[arg].text, argv[arg].len); |
2fae03e8 | 1103 | else |
1739cf24 | 1104 | dest->appendmem (tok.text, tok.len); |
2fae03e8 TT |
1105 | } |
1106 | ||
1107 | /* Apply a possible sequence of ## operators. */ | |
1108 | for (;;) | |
1109 | { | |
1110 | if (! get_token (&tok, &replacement_list)) | |
1111 | error (_("Splicing operator at end of macro")); | |
1112 | ||
1113 | /* Handle a comma before a ##. If we are handling | |
1114 | varargs, and the token on the right hand side is the | |
1115 | varargs marker, and the final argument is empty or | |
1116 | missing, then drop the comma. This is a GNU | |
1117 | extension. There is one ambiguous case here, | |
1118 | involving pedantic behavior with an empty argument, | |
1119 | but we settle that in favor of GNU-style (GCC uses an | |
1120 | option). If we aren't dealing with varargs, we | |
1121 | simply insert the comma. */ | |
1122 | if (prev_was_comma) | |
1123 | { | |
1124 | if (! (is_varargs | |
1125 | && tok.len == va_arg_name->len | |
1126 | && !memcmp (tok.text, va_arg_name->text, tok.len) | |
1739cf24 TT |
1127 | && argv.back ().len == 0)) |
1128 | dest->appendmem (",", 1); | |
2fae03e8 TT |
1129 | prev_was_comma = 0; |
1130 | } | |
1131 | ||
1132 | /* Insert the token. If it is a parameter, insert the | |
1133 | argument. If it is a comma, treat it specially. */ | |
1134 | if (tok.len == 1 && tok.text[0] == ',') | |
1135 | prev_was_comma = 1; | |
1136 | else | |
1137 | { | |
1138 | int arg = find_parameter (&tok, is_varargs, va_arg_name, | |
1139 | def->argc, def->argv); | |
b8d56208 | 1140 | |
2fae03e8 | 1141 | if (arg != -1) |
1739cf24 | 1142 | dest->appendmem (argv[arg].text, argv[arg].len); |
2fae03e8 | 1143 | else |
1739cf24 | 1144 | dest->appendmem (tok.text, tok.len); |
2fae03e8 TT |
1145 | } |
1146 | ||
1147 | /* Now read another token. If it is another splice, we | |
1148 | loop. */ | |
1149 | original_rl_start = replacement_list.text; | |
1150 | if (! get_token (&tok, &replacement_list)) | |
1151 | { | |
1152 | finished = 1; | |
1153 | break; | |
1154 | } | |
1155 | ||
1156 | if (! (tok.len == 2 | |
1157 | && tok.text[0] == '#' | |
1158 | && tok.text[1] == '#')) | |
1159 | break; | |
1160 | } | |
1161 | ||
1162 | if (prev_was_comma) | |
1163 | { | |
1164 | /* We saw a comma. Insert it now. */ | |
1739cf24 | 1165 | dest->appendmem (",", 1); |
2fae03e8 TT |
1166 | } |
1167 | ||
1168 | dest->last_token = dest->len; | |
1169 | if (finished) | |
1170 | lookahead_valid = 0; | |
1171 | else | |
1172 | { | |
1173 | /* Set up for the loop iterator. */ | |
1174 | lookahead = tok; | |
1175 | lookahead_rl_start = original_rl_start; | |
1176 | lookahead_valid = 1; | |
1177 | } | |
1178 | } | |
1179 | else | |
1180 | { | |
1181 | /* Is this token an identifier? */ | |
1182 | int substituted = 0; | |
1183 | int arg = find_parameter (&tok, is_varargs, va_arg_name, | |
1184 | def->argc, def->argv); | |
1185 | ||
1186 | if (arg != -1) | |
1187 | { | |
2fae03e8 TT |
1188 | /* Expand any macro invocations in the argument text, |
1189 | and append the result to dest. Remember that scan | |
1190 | mutates its source, so we need to scan a new buffer | |
1191 | referring to the argument's text, not the argument | |
1192 | itself. */ | |
1739cf24 | 1193 | struct macro_buffer arg_src (argv[arg].text, argv[arg].len); |
211d5b1c | 1194 | scan (dest, &arg_src, no_loop, scope); |
2fae03e8 TT |
1195 | substituted = 1; |
1196 | } | |
1197 | ||
1198 | /* If it wasn't a parameter, then just copy it across. */ | |
1199 | if (! substituted) | |
1200 | append_tokens_without_splicing (dest, &tok); | |
1201 | } | |
ec2bcbe7 | 1202 | } |
a9bbfbd8 TT |
1203 | |
1204 | if (vaopt_state > 0) | |
1205 | error (_("Unterminated __VA_OPT__")); | |
ec2bcbe7 JB |
1206 | } |
1207 | ||
1208 | ||
1209 | /* Expand a call to a macro named ID, whose definition is DEF. Append | |
1210 | its expansion to DEST. SRC is the input text following the ID | |
1211 | token. We are currently rescanning the expansions of the macros | |
1212 | named in NO_LOOP; don't re-expand them. Use LOOKUP_FUNC and | |
025bb325 | 1213 | LOOKUP_BATON to find definitions for any nested macro references. |
ec2bcbe7 JB |
1214 | |
1215 | Return 1 if we decided to expand it, zero otherwise. (If it's a | |
1216 | function-like macro name that isn't followed by an argument list, | |
1217 | we don't expand it.) If we return zero, leave SRC unchanged. */ | |
1218 | static int | |
1219 | expand (const char *id, | |
1220 | struct macro_definition *def, | |
1221 | struct macro_buffer *dest, | |
1222 | struct macro_buffer *src, | |
1223 | struct macro_name_list *no_loop, | |
211d5b1c | 1224 | const macro_scope &scope) |
ec2bcbe7 JB |
1225 | { |
1226 | struct macro_name_list new_no_loop; | |
1227 | ||
1228 | /* Create a new node to be added to the front of the no-expand list. | |
1229 | This list is appropriate for re-scanning replacement lists, but | |
1230 | it is *not* appropriate for scanning macro arguments; invocations | |
1231 | of the macro whose arguments we are gathering *do* get expanded | |
1232 | there. */ | |
1233 | new_no_loop.name = id; | |
1234 | new_no_loop.next = no_loop; | |
1235 | ||
1236 | /* What kind of macro are we expanding? */ | |
1237 | if (def->kind == macro_object_like) | |
1238 | { | |
1739cf24 TT |
1239 | struct macro_buffer replacement_list (def->replacement, |
1240 | strlen (def->replacement)); | |
ec2bcbe7 | 1241 | |
211d5b1c | 1242 | scan (dest, &replacement_list, &new_no_loop, scope); |
ec2bcbe7 JB |
1243 | return 1; |
1244 | } | |
1245 | else if (def->kind == macro_function_like) | |
1246 | { | |
1739cf24 | 1247 | struct macro_buffer va_arg_name; |
2fae03e8 TT |
1248 | int is_varargs = 0; |
1249 | ||
1250 | if (def->argc >= 1) | |
1251 | { | |
1252 | if (strcmp (def->argv[def->argc - 1], "...") == 0) | |
1253 | { | |
1254 | /* In C99-style varargs, substitution is done using | |
1255 | __VA_ARGS__. */ | |
1739cf24 | 1256 | va_arg_name.set_shared ("__VA_ARGS__", strlen ("__VA_ARGS__")); |
2fae03e8 TT |
1257 | is_varargs = 1; |
1258 | } | |
1259 | else | |
1260 | { | |
1261 | int len = strlen (def->argv[def->argc - 1]); | |
b8d56208 | 1262 | |
2fae03e8 TT |
1263 | if (len > 3 |
1264 | && strcmp (def->argv[def->argc - 1] + len - 3, "...") == 0) | |
1265 | { | |
1266 | /* In GNU-style varargs, the name of the | |
1267 | substitution parameter is the name of the formal | |
1268 | argument without the "...". */ | |
1739cf24 | 1269 | va_arg_name.set_shared (def->argv[def->argc - 1], len - 3); |
2fae03e8 TT |
1270 | is_varargs = 1; |
1271 | } | |
1272 | } | |
1273 | } | |
ec2bcbe7 | 1274 | |
1739cf24 | 1275 | std::vector<struct macro_buffer> argv; |
ec2bcbe7 JB |
1276 | /* If we couldn't find any argument list, then we don't expand |
1277 | this macro. */ | |
1739cf24 TT |
1278 | if (!gather_arguments (id, src, is_varargs ? def->argc : -1, |
1279 | &argv)) | |
1280 | return 0; | |
ec2bcbe7 JB |
1281 | |
1282 | /* Check that we're passing an acceptable number of arguments for | |
1283 | this macro. */ | |
1739cf24 | 1284 | if (argv.size () != def->argc) |
ec2bcbe7 | 1285 | { |
1739cf24 | 1286 | if (is_varargs && argv.size () >= def->argc - 1) |
2fae03e8 TT |
1287 | { |
1288 | /* Ok. */ | |
1289 | } | |
ec2bcbe7 JB |
1290 | /* Remember that a sequence of tokens like "foo()" is a |
1291 | valid invocation of a macro expecting either zero or one | |
1292 | arguments. */ | |
1739cf24 | 1293 | else if (! (argv.size () == 1 |
2fae03e8 TT |
1294 | && argv[0].len == 0 |
1295 | && def->argc == 0)) | |
8a3fe4f8 AC |
1296 | error (_("Wrong number of arguments to macro `%s' " |
1297 | "(expected %d, got %d)."), | |
1739cf24 | 1298 | id, def->argc, int (argv.size ())); |
ec2bcbe7 JB |
1299 | } |
1300 | ||
1301 | /* Note that we don't expand macro invocations in the arguments | |
1302 | yet --- we let subst_args take care of that. Parameters that | |
1303 | appear as operands of the stringifying operator "#" or the | |
1304 | splicing operator "##" don't get macro references expanded, | |
1305 | so we can't really tell whether it's appropriate to macro- | |
1306 | expand an argument until we see how it's being used. */ | |
1739cf24 | 1307 | struct macro_buffer substituted (0); |
2fae03e8 | 1308 | substitute_args (&substituted, def, is_varargs, &va_arg_name, |
211d5b1c | 1309 | argv, no_loop, scope); |
ec2bcbe7 JB |
1310 | |
1311 | /* Now `substituted' is the macro's replacement list, with all | |
1312 | argument values substituted into it properly. Re-scan it for | |
1313 | macro references, but don't expand invocations of this macro. | |
1314 | ||
1315 | We create a new buffer, `substituted_src', which points into | |
1316 | `substituted', and scan that. We can't scan `substituted' | |
1317 | itself, since the tokenization process moves the buffer's | |
1318 | text pointer around, and we still need to be able to find | |
1319 | `substituted's original text buffer after scanning it so we | |
1320 | can free it. */ | |
1739cf24 | 1321 | struct macro_buffer substituted_src (substituted.text, substituted.len); |
211d5b1c | 1322 | scan (dest, &substituted_src, &new_no_loop, scope); |
ec2bcbe7 | 1323 | |
ec2bcbe7 JB |
1324 | return 1; |
1325 | } | |
1326 | else | |
e2e0b3e5 | 1327 | internal_error (__FILE__, __LINE__, _("bad macro definition kind")); |
ec2bcbe7 JB |
1328 | } |
1329 | ||
1330 | ||
1331 | /* If the single token in SRC_FIRST followed by the tokens in SRC_REST | |
30baf67b | 1332 | constitute a macro invocation not forbidden in NO_LOOP, append its |
ec2bcbe7 JB |
1333 | expansion to DEST and return non-zero. Otherwise, return zero, and |
1334 | leave DEST unchanged. | |
1335 | ||
1336 | SRC_FIRST and SRC_REST must be shared buffers; DEST must not be one. | |
1337 | SRC_FIRST must be a string built by get_token. */ | |
1338 | static int | |
1339 | maybe_expand (struct macro_buffer *dest, | |
1340 | struct macro_buffer *src_first, | |
1341 | struct macro_buffer *src_rest, | |
1342 | struct macro_name_list *no_loop, | |
211d5b1c | 1343 | const macro_scope &scope) |
ec2bcbe7 JB |
1344 | { |
1345 | gdb_assert (src_first->shared); | |
1346 | gdb_assert (src_rest->shared); | |
1347 | gdb_assert (! dest->shared); | |
1348 | ||
1349 | /* Is this token an identifier? */ | |
1350 | if (src_first->is_identifier) | |
1351 | { | |
1352 | /* Make a null-terminated copy of it, since that's what our | |
1353 | lookup function expects. */ | |
0354904b | 1354 | std::string id (src_first->text, src_first->len); |
b8d56208 | 1355 | |
ec2bcbe7 JB |
1356 | /* If we're currently re-scanning the result of expanding |
1357 | this macro, don't expand it again. */ | |
0354904b | 1358 | if (! currently_rescanning (no_loop, id.c_str ())) |
ec2bcbe7 JB |
1359 | { |
1360 | /* Does this identifier have a macro definition in scope? */ | |
211d5b1c | 1361 | macro_definition *def = standard_macro_lookup (id.c_str (), scope); |
ec2bcbe7 | 1362 | |
211d5b1c | 1363 | if (def && expand (id.c_str (), def, dest, src_rest, no_loop, scope)) |
0354904b | 1364 | return 1; |
ec2bcbe7 | 1365 | } |
ec2bcbe7 JB |
1366 | } |
1367 | ||
1368 | return 0; | |
1369 | } | |
1370 | ||
1371 | ||
1372 | /* Expand macro references in SRC, appending the results to DEST. | |
1373 | Assume we are re-scanning the result of expanding the macros named | |
1374 | in NO_LOOP, and don't try to re-expand references to them. | |
1375 | ||
1376 | SRC must be a shared buffer; DEST must not be one. */ | |
1377 | static void | |
1378 | scan (struct macro_buffer *dest, | |
1379 | struct macro_buffer *src, | |
1380 | struct macro_name_list *no_loop, | |
211d5b1c | 1381 | const macro_scope &scope) |
ec2bcbe7 JB |
1382 | { |
1383 | gdb_assert (src->shared); | |
1384 | gdb_assert (! dest->shared); | |
1385 | ||
1386 | for (;;) | |
1387 | { | |
1388 | struct macro_buffer tok; | |
1389 | char *original_src_start = src->text; | |
1390 | ||
1391 | /* Find the next token in SRC. */ | |
1392 | if (! get_token (&tok, src)) | |
1393 | break; | |
1394 | ||
1395 | /* Just for aesthetics. If we skipped some whitespace, copy | |
1396 | that to DEST. */ | |
1397 | if (tok.text > original_src_start) | |
1398 | { | |
1739cf24 | 1399 | dest->appendmem (original_src_start, tok.text - original_src_start); |
ec2bcbe7 JB |
1400 | dest->last_token = dest->len; |
1401 | } | |
1402 | ||
211d5b1c | 1403 | if (! maybe_expand (dest, &tok, src, no_loop, scope)) |
ec2bcbe7 JB |
1404 | /* We didn't end up expanding tok as a macro reference, so |
1405 | simply append it to dest. */ | |
1406 | append_tokens_without_splicing (dest, &tok); | |
1407 | } | |
1408 | ||
1409 | /* Just for aesthetics. If there was any trailing whitespace in | |
1410 | src, copy it to dest. */ | |
1411 | if (src->len) | |
1412 | { | |
1739cf24 | 1413 | dest->appendmem (src->text, src->len); |
ec2bcbe7 JB |
1414 | dest->last_token = dest->len; |
1415 | } | |
1416 | } | |
1417 | ||
1418 | ||
f6c2623e | 1419 | gdb::unique_xmalloc_ptr<char> |
211d5b1c | 1420 | macro_expand (const char *source, const macro_scope &scope) |
ec2bcbe7 | 1421 | { |
1739cf24 | 1422 | struct macro_buffer src (source, strlen (source)); |
ec2bcbe7 | 1423 | |
1739cf24 | 1424 | struct macro_buffer dest (0); |
ec2bcbe7 | 1425 | dest.last_token = 0; |
ec2bcbe7 | 1426 | |
211d5b1c | 1427 | scan (&dest, &src, 0, scope); |
ec2bcbe7 | 1428 | |
1739cf24 | 1429 | dest.appendc ('\0'); |
ec2bcbe7 | 1430 | |
14d960c8 | 1431 | return dest.release (); |
ec2bcbe7 JB |
1432 | } |
1433 | ||
1434 | ||
f6c2623e | 1435 | gdb::unique_xmalloc_ptr<char> |
211d5b1c | 1436 | macro_expand_once (const char *source, const macro_scope &scope) |
ec2bcbe7 | 1437 | { |
8a3fe4f8 | 1438 | error (_("Expand-once not implemented yet.")); |
ec2bcbe7 JB |
1439 | } |
1440 | ||
14d960c8 | 1441 | gdb::unique_xmalloc_ptr<char> |
211d5b1c | 1442 | macro_expand_next (const char **lexptr, const macro_scope &scope) |
ec2bcbe7 | 1443 | { |
1739cf24 | 1444 | struct macro_buffer tok; |
ec2bcbe7 JB |
1445 | |
1446 | /* Set up SRC to refer to the input text, pointed to by *lexptr. */ | |
1739cf24 | 1447 | struct macro_buffer src (*lexptr, strlen (*lexptr)); |
ec2bcbe7 JB |
1448 | |
1449 | /* Set up DEST to receive the expansion, if there is one. */ | |
1739cf24 | 1450 | struct macro_buffer dest (0); |
ec2bcbe7 | 1451 | dest.last_token = 0; |
ec2bcbe7 JB |
1452 | |
1453 | /* Get the text's first preprocessing token. */ | |
1454 | if (! get_token (&tok, &src)) | |
14d960c8 | 1455 | return nullptr; |
ec2bcbe7 JB |
1456 | |
1457 | /* If it's a macro invocation, expand it. */ | |
211d5b1c | 1458 | if (maybe_expand (&dest, &tok, &src, 0, scope)) |
ec2bcbe7 JB |
1459 | { |
1460 | /* It was a macro invocation! Package up the expansion as a | |
1461 | null-terminated string and return it. Set *lexptr to the | |
1462 | start of the next token in the input. */ | |
1739cf24 | 1463 | dest.appendc ('\0'); |
ec2bcbe7 | 1464 | *lexptr = src.text; |
1739cf24 | 1465 | return dest.release (); |
ec2bcbe7 JB |
1466 | } |
1467 | else | |
1468 | { | |
1469 | /* It wasn't a macro invocation. */ | |
14d960c8 | 1470 | return nullptr; |
ec2bcbe7 JB |
1471 | } |
1472 | } |