]>
Commit | Line | Data |
---|---|---|
ec2bcbe7 | 1 | /* C preprocessor macro expansion for GDB. |
213516ef | 2 | Copyright (C) 2002-2023 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" | |
bf31fd38 | 21 | #include "gdbsupport/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 | |
ec2bcbe7 | 29 | |
ac3d4064 PA |
30 | /* A string type that we can use to refer to substrings of other |
31 | strings. */ | |
ec2bcbe7 | 32 | |
ac3d4064 | 33 | struct shared_macro_buffer |
ec2bcbe7 | 34 | { |
ac3d4064 PA |
35 | /* An array of characters. This buffer is a pointer into some |
36 | larger string and thus we can't assume in that the text is | |
ec2bcbe7 | 37 | null-terminated. */ |
ac3d4064 | 38 | const char *text; |
ec2bcbe7 JB |
39 | |
40 | /* The number of characters in the string. */ | |
41 | int len; | |
42 | ||
ec2bcbe7 JB |
43 | /* For detecting token splicing. |
44 | ||
45 | This is the index in TEXT of the first character of the token | |
46 | that abuts the end of TEXT. If TEXT contains no tokens, then we | |
47 | set this equal to LEN. If TEXT ends in whitespace, then there is | |
48 | no token abutting the end of TEXT (it's just whitespace), and | |
49 | again, we set this equal to LEN. We set this to -1 if we don't | |
50 | know the nature of TEXT. */ | |
1739cf24 | 51 | int last_token = -1; |
ec2bcbe7 JB |
52 | |
53 | /* If this buffer is holding the result from get_token, then this | |
54 | is non-zero if it is an identifier token, zero otherwise. */ | |
1739cf24 | 55 | int is_identifier = 0; |
ec2bcbe7 | 56 | |
ac3d4064 | 57 | shared_macro_buffer () |
1739cf24 | 58 | : text (NULL), |
ac3d4064 | 59 | len (0) |
1739cf24 | 60 | { |
1739cf24 | 61 | } |
abc9d0dc | 62 | |
1739cf24 TT |
63 | /* Set the macro buffer to refer to the LEN bytes at ADDR, as a |
64 | shared substring. */ | |
ac3d4064 | 65 | shared_macro_buffer (const char *addr, int len) |
1739cf24 TT |
66 | { |
67 | set_shared (addr, len); | |
68 | } | |
ec2bcbe7 | 69 | |
1739cf24 TT |
70 | /* Set the macro buffer to refer to the LEN bytes at ADDR, as a |
71 | shared substring. */ | |
72 | void set_shared (const char *addr, int len_) | |
73 | { | |
ac3d4064 | 74 | text = addr; |
1739cf24 | 75 | len = len_; |
1739cf24 | 76 | } |
ac3d4064 PA |
77 | }; |
78 | ||
79 | /* A string type that we can resize and quickly append to. */ | |
80 | ||
81 | struct growable_macro_buffer | |
82 | { | |
83 | /* An array of characters. The first LEN bytes are the real text, | |
84 | but there are SIZE bytes allocated to the array. */ | |
85 | char *text; | |
86 | ||
87 | /* The number of characters in the string. */ | |
88 | int len; | |
89 | ||
90 | /* The number of characters allocated to the string. */ | |
91 | int size; | |
92 | ||
93 | /* For detecting token splicing. | |
ec2bcbe7 | 94 | |
ac3d4064 PA |
95 | This is the index in TEXT of the first character of the token |
96 | that abuts the end of TEXT. If TEXT contains no tokens, then we | |
97 | set this equal to LEN. If TEXT ends in whitespace, then there is | |
98 | no token abutting the end of TEXT (it's just whitespace), and | |
99 | again, we set this equal to LEN. We set this to -1 if we don't | |
100 | know the nature of TEXT. */ | |
101 | int last_token = -1; | |
102 | ||
103 | /* Set the macro buffer to the empty string, guessing that its | |
104 | final contents will fit in N bytes. (It'll get resized if it | |
105 | doesn't, so the guess doesn't have to be right.) Allocate the | |
106 | initial storage with xmalloc. */ | |
107 | explicit growable_macro_buffer (int n) | |
108 | : len (0), | |
109 | size (n) | |
c530603c | 110 | { |
ac3d4064 PA |
111 | if (n > 0) |
112 | text = (char *) xmalloc (n); | |
113 | else | |
114 | text = NULL; | |
c530603c TV |
115 | } |
116 | ||
ac3d4064 PA |
117 | DISABLE_COPY_AND_ASSIGN (growable_macro_buffer); |
118 | ||
119 | ~growable_macro_buffer () | |
1739cf24 | 120 | { |
ac3d4064 | 121 | xfree (text); |
1739cf24 | 122 | } |
ec2bcbe7 | 123 | |
14d960c8 SM |
124 | /* Release the text of the buffer to the caller. */ |
125 | gdb::unique_xmalloc_ptr<char> release () | |
1739cf24 | 126 | { |
1739cf24 TT |
127 | gdb_assert (size); |
128 | char *result = text; | |
129 | text = NULL; | |
14d960c8 | 130 | return gdb::unique_xmalloc_ptr<char> (result); |
1739cf24 | 131 | } |
ec2bcbe7 | 132 | |
ac3d4064 | 133 | /* Resize the buffer to be at least N bytes long. */ |
1739cf24 TT |
134 | void resize_buffer (int n) |
135 | { | |
1739cf24 TT |
136 | if (size == 0) |
137 | size = n; | |
138 | else | |
139 | while (size <= n) | |
140 | size *= 2; | |
ec2bcbe7 | 141 | |
1739cf24 TT |
142 | text = (char *) xrealloc (text, size); |
143 | } | |
ec2bcbe7 | 144 | |
1739cf24 TT |
145 | /* Append the character C to the buffer. */ |
146 | void appendc (int c) | |
147 | { | |
148 | int new_len = len + 1; | |
ec2bcbe7 | 149 | |
1739cf24 TT |
150 | if (new_len > size) |
151 | resize_buffer (new_len); | |
ec2bcbe7 | 152 | |
1739cf24 TT |
153 | text[len] = c; |
154 | len = new_len; | |
155 | } | |
ec2bcbe7 | 156 | |
1739cf24 TT |
157 | /* Append the COUNT bytes at ADDR to the buffer. */ |
158 | void appendmem (const char *addr, int count) | |
159 | { | |
160 | int new_len = len + count; | |
ec2bcbe7 | 161 | |
1739cf24 TT |
162 | if (new_len > size) |
163 | resize_buffer (new_len); | |
ec2bcbe7 | 164 | |
1739cf24 TT |
165 | memcpy (text + len, addr, count); |
166 | len = new_len; | |
167 | } | |
168 | }; | |
ec2bcbe7 JB |
169 | |
170 | ||
171 | \f | |
172 | /* Recognizing preprocessor tokens. */ | |
173 | ||
174 | ||
d7d9f01e TT |
175 | int |
176 | macro_is_whitespace (int c) | |
ec2bcbe7 JB |
177 | { |
178 | return (c == ' ' | |
dda83cd7 SM |
179 | || c == '\t' |
180 | || c == '\n' | |
181 | || c == '\v' | |
182 | || c == '\f'); | |
ec2bcbe7 JB |
183 | } |
184 | ||
185 | ||
d7d9f01e TT |
186 | int |
187 | macro_is_digit (int c) | |
ec2bcbe7 JB |
188 | { |
189 | return ('0' <= c && c <= '9'); | |
190 | } | |
191 | ||
192 | ||
d7d9f01e TT |
193 | int |
194 | macro_is_identifier_nondigit (int c) | |
ec2bcbe7 JB |
195 | { |
196 | return (c == '_' | |
dda83cd7 SM |
197 | || ('a' <= c && c <= 'z') |
198 | || ('A' <= c && c <= 'Z')); | |
ec2bcbe7 JB |
199 | } |
200 | ||
201 | ||
202 | static void | |
ac3d4064 | 203 | set_token (shared_macro_buffer *tok, const char *start, const char *end) |
ec2bcbe7 | 204 | { |
1739cf24 | 205 | tok->set_shared (start, end - start); |
ec2bcbe7 JB |
206 | tok->last_token = 0; |
207 | ||
025bb325 | 208 | /* Presumed; get_identifier may overwrite this. */ |
ec2bcbe7 JB |
209 | tok->is_identifier = 0; |
210 | } | |
211 | ||
212 | ||
213 | static int | |
ac3d4064 | 214 | get_comment (shared_macro_buffer *tok, const char *p, const char *end) |
ec2bcbe7 JB |
215 | { |
216 | if (p + 2 > end) | |
217 | return 0; | |
218 | else if (p[0] == '/' | |
dda83cd7 | 219 | && p[1] == '*') |
ec2bcbe7 | 220 | { |
ac3d4064 | 221 | const char *tok_start = p; |
ec2bcbe7 JB |
222 | |
223 | p += 2; | |
224 | ||
225 | for (; p < end; p++) | |
dda83cd7 SM |
226 | if (p + 2 <= end |
227 | && p[0] == '*' | |
228 | && p[1] == '/') | |
229 | { | |
230 | p += 2; | |
231 | set_token (tok, tok_start, p); | |
232 | return 1; | |
233 | } | |
ec2bcbe7 | 234 | |
8a3fe4f8 | 235 | error (_("Unterminated comment in macro expansion.")); |
ec2bcbe7 JB |
236 | } |
237 | else if (p[0] == '/' | |
dda83cd7 | 238 | && p[1] == '/') |
ec2bcbe7 | 239 | { |
ac3d4064 | 240 | const char *tok_start = p; |
ec2bcbe7 JB |
241 | |
242 | p += 2; | |
243 | for (; p < end; p++) | |
dda83cd7 SM |
244 | if (*p == '\n') |
245 | break; | |
ec2bcbe7 JB |
246 | |
247 | set_token (tok, tok_start, p); | |
248 | return 1; | |
249 | } | |
250 | else | |
251 | return 0; | |
252 | } | |
253 | ||
254 | ||
255 | static int | |
ac3d4064 | 256 | get_identifier (shared_macro_buffer *tok, const char *p, const char *end) |
ec2bcbe7 JB |
257 | { |
258 | if (p < end | |
d7d9f01e | 259 | && macro_is_identifier_nondigit (*p)) |
ec2bcbe7 | 260 | { |
ac3d4064 | 261 | const char *tok_start = p; |
ec2bcbe7 JB |
262 | |
263 | while (p < end | |
dda83cd7 SM |
264 | && (macro_is_identifier_nondigit (*p) |
265 | || macro_is_digit (*p))) | |
266 | p++; | |
ec2bcbe7 JB |
267 | |
268 | set_token (tok, tok_start, p); | |
269 | tok->is_identifier = 1; | |
270 | return 1; | |
271 | } | |
272 | else | |
273 | return 0; | |
274 | } | |
275 | ||
276 | ||
277 | static int | |
ac3d4064 | 278 | get_pp_number (shared_macro_buffer *tok, const char *p, const char *end) |
ec2bcbe7 JB |
279 | { |
280 | if (p < end | |
d7d9f01e | 281 | && (macro_is_digit (*p) |
dda83cd7 | 282 | || (*p == '.' |
17c8aaf5 TT |
283 | && p + 2 <= end |
284 | && macro_is_digit (p[1])))) | |
ec2bcbe7 | 285 | { |
ac3d4064 | 286 | const char *tok_start = p; |
ec2bcbe7 JB |
287 | |
288 | while (p < end) | |
dda83cd7 | 289 | { |
17c8aaf5 TT |
290 | if (p + 2 <= end |
291 | && strchr ("eEpP", *p) | |
292 | && (p[1] == '+' || p[1] == '-')) | |
dda83cd7 SM |
293 | p += 2; |
294 | else if (macro_is_digit (*p) | |
17c8aaf5 TT |
295 | || macro_is_identifier_nondigit (*p) |
296 | || *p == '.') | |
dda83cd7 SM |
297 | p++; |
298 | else | |
299 | break; | |
300 | } | |
ec2bcbe7 JB |
301 | |
302 | set_token (tok, tok_start, p); | |
303 | return 1; | |
304 | } | |
305 | else | |
306 | return 0; | |
307 | } | |
308 | ||
309 | ||
310 | ||
311 | /* If the text starting at P going up to (but not including) END | |
312 | starts with a character constant, set *TOK to point to that | |
313 | character constant, and return 1. Otherwise, return zero. | |
314 | Signal an error if it contains a malformed or incomplete character | |
315 | constant. */ | |
316 | static int | |
ac3d4064 PA |
317 | get_character_constant (shared_macro_buffer *tok, |
318 | const char *p, const char *end) | |
ec2bcbe7 JB |
319 | { |
320 | /* ISO/IEC 9899:1999 (E) Section 6.4.4.4 paragraph 1 | |
321 | But of course, what really matters is that we handle it the same | |
322 | way GDB's C/C++ lexer does. So we call parse_escape in utils.c | |
323 | to handle escape sequences. */ | |
324 | if ((p + 1 <= end && *p == '\'') | |
6c7a06a3 TT |
325 | || (p + 2 <= end |
326 | && (p[0] == 'L' || p[0] == 'u' || p[0] == 'U') | |
327 | && p[1] == '\'')) | |
ec2bcbe7 | 328 | { |
ac3d4064 | 329 | const char *tok_start = p; |
6c7a06a3 | 330 | int char_count = 0; |
ec2bcbe7 JB |
331 | |
332 | if (*p == '\'') | |
dda83cd7 | 333 | p++; |
6c7a06a3 | 334 | else if (*p == 'L' || *p == 'u' || *p == 'U') |
dda83cd7 | 335 | p += 2; |
ec2bcbe7 | 336 | else |
dda83cd7 | 337 | gdb_assert_not_reached ("unexpected character constant"); |
ec2bcbe7 | 338 | |
ec2bcbe7 | 339 | for (;;) |
dda83cd7 SM |
340 | { |
341 | if (p >= end) | |
342 | error (_("Unmatched single quote.")); | |
343 | else if (*p == '\'') | |
344 | { | |
345 | if (!char_count) | |
346 | error (_("A character constant must contain at least one " | |
347 | "character.")); | |
348 | p++; | |
349 | break; | |
350 | } | |
351 | else if (*p == '\\') | |
352 | { | |
d7561cbb KS |
353 | const char *s, *o; |
354 | ||
355 | s = o = ++p; | |
356 | char_count += c_parse_escape (&s, NULL); | |
357 | p += s - o; | |
dda83cd7 SM |
358 | } |
359 | else | |
6c7a06a3 TT |
360 | { |
361 | p++; | |
362 | char_count++; | |
363 | } | |
dda83cd7 | 364 | } |
ec2bcbe7 JB |
365 | |
366 | set_token (tok, tok_start, p); | |
367 | return 1; | |
368 | } | |
369 | else | |
370 | return 0; | |
371 | } | |
372 | ||
373 | ||
374 | /* If the text starting at P going up to (but not including) END | |
375 | starts with a string literal, set *TOK to point to that string | |
376 | literal, and return 1. Otherwise, return zero. Signal an error if | |
377 | it contains a malformed or incomplete string literal. */ | |
378 | static int | |
ac3d4064 | 379 | get_string_literal (shared_macro_buffer *tok, const char *p, const char *end) |
ec2bcbe7 JB |
380 | { |
381 | if ((p + 1 <= end | |
6c7a06a3 | 382 | && *p == '"') |
ec2bcbe7 | 383 | || (p + 2 <= end |
dda83cd7 SM |
384 | && (p[0] == 'L' || p[0] == 'u' || p[0] == 'U') |
385 | && p[1] == '"')) | |
ec2bcbe7 | 386 | { |
ac3d4064 | 387 | const char *tok_start = p; |
ec2bcbe7 | 388 | |
6c7a06a3 | 389 | if (*p == '"') |
dda83cd7 | 390 | p++; |
6c7a06a3 | 391 | else if (*p == 'L' || *p == 'u' || *p == 'U') |
dda83cd7 | 392 | p += 2; |
ec2bcbe7 | 393 | else |
dda83cd7 | 394 | gdb_assert_not_reached ("unexpected string literal"); |
ec2bcbe7 JB |
395 | |
396 | for (;;) | |
dda83cd7 SM |
397 | { |
398 | if (p >= end) | |
399 | error (_("Unterminated string in expression.")); | |
400 | else if (*p == '"') | |
401 | { | |
402 | p++; | |
403 | break; | |
404 | } | |
405 | else if (*p == '\n') | |
406 | error (_("Newline characters may not appear in string " | |
407 | "constants.")); | |
408 | else if (*p == '\\') | |
409 | { | |
d7561cbb KS |
410 | const char *s, *o; |
411 | ||
412 | s = o = ++p; | |
413 | c_parse_escape (&s, NULL); | |
414 | p += s - o; | |
dda83cd7 SM |
415 | } |
416 | else | |
417 | p++; | |
418 | } | |
ec2bcbe7 JB |
419 | |
420 | set_token (tok, tok_start, p); | |
421 | return 1; | |
422 | } | |
423 | else | |
424 | return 0; | |
425 | } | |
426 | ||
427 | ||
428 | static int | |
ac3d4064 | 429 | get_punctuator (shared_macro_buffer *tok, const char *p, const char *end) |
ec2bcbe7 JB |
430 | { |
431 | /* Here, speed is much less important than correctness and clarity. */ | |
432 | ||
ccb3ac8a TT |
433 | /* ISO/IEC 9899:1999 (E) Section 6.4.6 Paragraph 1. |
434 | Note that this table is ordered in a special way. A punctuator | |
435 | which is a prefix of another punctuator must appear after its | |
436 | "extension". Otherwise, the wrong token will be returned. */ | |
ec2bcbe7 | 437 | static const char * const punctuators[] = { |
ccb3ac8a TT |
438 | "[", "]", "(", ")", "{", "}", "?", ";", ",", "~", |
439 | "...", ".", | |
440 | "->", "--", "-=", "-", | |
441 | "++", "+=", "+", | |
442 | "*=", "*", | |
443 | "!=", "!", | |
444 | "&&", "&=", "&", | |
445 | "/=", "/", | |
446 | "%>", "%:%:", "%:", "%=", "%", | |
447 | "^=", "^", | |
448 | "##", "#", | |
449 | ":>", ":", | |
450 | "||", "|=", "|", | |
451 | "<<=", "<<", "<=", "<:", "<%", "<", | |
452 | ">>=", ">>", ">=", ">", | |
453 | "==", "=", | |
ec2bcbe7 JB |
454 | 0 |
455 | }; | |
456 | ||
457 | int i; | |
458 | ||
459 | if (p + 1 <= end) | |
460 | { | |
461 | for (i = 0; punctuators[i]; i++) | |
dda83cd7 SM |
462 | { |
463 | const char *punctuator = punctuators[i]; | |
464 | ||
465 | if (p[0] == punctuator[0]) | |
466 | { | |
467 | int len = strlen (punctuator); | |
468 | ||
469 | if (p + len <= end | |
470 | && ! memcmp (p, punctuator, len)) | |
471 | { | |
472 | set_token (tok, p, p + len); | |
473 | return 1; | |
474 | } | |
475 | } | |
476 | } | |
ec2bcbe7 JB |
477 | } |
478 | ||
479 | return 0; | |
480 | } | |
481 | ||
482 | ||
483 | /* Peel the next preprocessor token off of SRC, and put it in TOK. | |
484 | Mutate TOK to refer to the first token in SRC, and mutate SRC to | |
ac3d4064 PA |
485 | refer to the text after that token. The resulting TOK will point |
486 | into the same string SRC does. Initialize TOK's last_token field. | |
487 | Return non-zero if we succeed, or 0 if we didn't find any more | |
488 | tokens in SRC. */ | |
489 | ||
ec2bcbe7 | 490 | static int |
ac3d4064 | 491 | get_token (shared_macro_buffer *tok, shared_macro_buffer *src) |
ec2bcbe7 | 492 | { |
ac3d4064 PA |
493 | const char *p = src->text; |
494 | const char *end = p + src->len; | |
ec2bcbe7 JB |
495 | |
496 | /* From the ISO C standard, ISO/IEC 9899:1999 (E), section 6.4: | |
497 | ||
498 | preprocessing-token: | |
dda83cd7 SM |
499 | header-name |
500 | identifier | |
501 | pp-number | |
502 | character-constant | |
503 | string-literal | |
504 | punctuator | |
505 | each non-white-space character that cannot be one of the above | |
ec2bcbe7 JB |
506 | |
507 | We don't have to deal with header-name tokens, since those can | |
508 | only occur after a #include, which we will never see. */ | |
509 | ||
510 | while (p < end) | |
d7d9f01e | 511 | if (macro_is_whitespace (*p)) |
ec2bcbe7 JB |
512 | p++; |
513 | else if (get_comment (tok, p, end)) | |
514 | p += tok->len; | |
515 | else if (get_pp_number (tok, p, end) | |
dda83cd7 SM |
516 | || get_character_constant (tok, p, end) |
517 | || get_string_literal (tok, p, end) | |
518 | /* Note: the grammar in the standard seems to be | |
519 | ambiguous: L'x' can be either a wide character | |
520 | constant, or an identifier followed by a normal | |
521 | character constant. By trying `get_identifier' after | |
522 | we try get_character_constant and get_string_literal, | |
523 | we give the wide character syntax precedence. Now, | |
524 | since GDB doesn't handle wide character constants | |
525 | anyway, is this the right thing to do? */ | |
526 | || get_identifier (tok, p, end) | |
527 | || get_punctuator (tok, p, end)) | |
ec2bcbe7 | 528 | { |
dda83cd7 SM |
529 | /* How many characters did we consume, including whitespace? */ |
530 | int consumed = p - src->text + tok->len; | |
b8d56208 | 531 | |
dda83cd7 SM |
532 | src->text += consumed; |
533 | src->len -= consumed; | |
534 | return 1; | |
ec2bcbe7 JB |
535 | } |
536 | else | |
537 | { | |
dda83cd7 SM |
538 | /* We have found a "non-whitespace character that cannot be |
539 | one of the above." Make a token out of it. */ | |
540 | int consumed; | |
541 | ||
542 | set_token (tok, p, p + 1); | |
543 | consumed = p - src->text + tok->len; | |
544 | src->text += consumed; | |
545 | src->len -= consumed; | |
546 | return 1; | |
ec2bcbe7 JB |
547 | } |
548 | ||
549 | return 0; | |
550 | } | |
551 | ||
552 | ||
553 | \f | |
554 | /* Appending token strings, with and without splicing */ | |
555 | ||
556 | ||
557 | /* Append the macro buffer SRC to the end of DEST, and ensure that | |
558 | doing so doesn't splice the token at the end of SRC with the token | |
559 | at the beginning of DEST. SRC and DEST must have their last_token | |
560 | fields set. Upon return, DEST's last_token field is set correctly. | |
561 | ||
562 | For example: | |
563 | ||
564 | If DEST is "(" and SRC is "y", then we can return with | |
565 | DEST set to "(y" --- we've simply appended the two buffers. | |
566 | ||
567 | However, if DEST is "x" and SRC is "y", then we must not return | |
568 | with DEST set to "xy" --- that would splice the two tokens "x" and | |
569 | "y" together to make a single token "xy". However, it would be | |
570 | fine to return with DEST set to "x y". Similarly, "<" and "<" must | |
571 | yield "< <", not "<<", etc. */ | |
572 | static void | |
ac3d4064 PA |
573 | append_tokens_without_splicing (growable_macro_buffer *dest, |
574 | shared_macro_buffer *src) | |
ec2bcbe7 JB |
575 | { |
576 | int original_dest_len = dest->len; | |
ac3d4064 | 577 | shared_macro_buffer dest_tail, new_token; |
ec2bcbe7 JB |
578 | |
579 | gdb_assert (src->last_token != -1); | |
580 | gdb_assert (dest->last_token != -1); | |
581 | ||
582 | /* First, just try appending the two, and call get_token to see if | |
583 | we got a splice. */ | |
1739cf24 | 584 | dest->appendmem (src->text, src->len); |
ec2bcbe7 JB |
585 | |
586 | /* If DEST originally had no token abutting its end, then we can't | |
587 | have spliced anything, so we're done. */ | |
588 | if (dest->last_token == original_dest_len) | |
589 | { | |
590 | dest->last_token = original_dest_len + src->last_token; | |
591 | return; | |
592 | } | |
593 | ||
594 | /* Set DEST_TAIL to point to the last token in DEST, followed by | |
595 | all the stuff we just appended. */ | |
1739cf24 TT |
596 | dest_tail.set_shared (dest->text + dest->last_token, |
597 | dest->len - dest->last_token); | |
ec2bcbe7 JB |
598 | |
599 | /* Re-parse DEST's last token. We know that DEST used to contain | |
600 | at least one token, so if it doesn't contain any after the | |
601 | append, then we must have spliced "/" and "*" or "/" and "/" to | |
602 | make a comment start. (Just for the record, I got this right | |
603 | the first time. This is not a bug fix.) */ | |
604 | if (get_token (&new_token, &dest_tail) | |
605 | && (new_token.text + new_token.len | |
dda83cd7 | 606 | == dest->text + original_dest_len)) |
ec2bcbe7 JB |
607 | { |
608 | /* No splice, so we're done. */ | |
609 | dest->last_token = original_dest_len + src->last_token; | |
610 | return; | |
611 | } | |
612 | ||
613 | /* Okay, a simple append caused a splice. Let's chop dest back to | |
614 | its original length and try again, but separate the texts with a | |
615 | space. */ | |
616 | dest->len = original_dest_len; | |
1739cf24 TT |
617 | dest->appendc (' '); |
618 | dest->appendmem (src->text, src->len); | |
ec2bcbe7 | 619 | |
1739cf24 TT |
620 | dest_tail.set_shared (dest->text + dest->last_token, |
621 | dest->len - dest->last_token); | |
ec2bcbe7 JB |
622 | |
623 | /* Try to re-parse DEST's last token, as above. */ | |
624 | if (get_token (&new_token, &dest_tail) | |
625 | && (new_token.text + new_token.len | |
dda83cd7 | 626 | == dest->text + original_dest_len)) |
ec2bcbe7 JB |
627 | { |
628 | /* No splice, so we're done. */ | |
629 | dest->last_token = original_dest_len + 1 + src->last_token; | |
630 | return; | |
631 | } | |
632 | ||
633 | /* As far as I know, there's no case where inserting a space isn't | |
634 | enough to prevent a splice. */ | |
f34652de | 635 | internal_error (_("unable to avoid splicing tokens during macro expansion")); |
ec2bcbe7 JB |
636 | } |
637 | ||
2fae03e8 TT |
638 | /* Stringify an argument, and insert it into DEST. ARG is the text to |
639 | stringify; it is LEN bytes long. */ | |
640 | ||
641 | static void | |
ac3d4064 | 642 | stringify (growable_macro_buffer *dest, const char *arg, int len) |
2fae03e8 TT |
643 | { |
644 | /* Trim initial whitespace from ARG. */ | |
645 | while (len > 0 && macro_is_whitespace (*arg)) | |
646 | { | |
647 | ++arg; | |
648 | --len; | |
649 | } | |
650 | ||
651 | /* Trim trailing whitespace from ARG. */ | |
652 | while (len > 0 && macro_is_whitespace (arg[len - 1])) | |
653 | --len; | |
654 | ||
655 | /* Insert the string. */ | |
1739cf24 | 656 | dest->appendc ('"'); |
2fae03e8 TT |
657 | while (len > 0) |
658 | { | |
659 | /* We could try to handle strange cases here, like control | |
660 | characters, but there doesn't seem to be much point. */ | |
661 | if (macro_is_whitespace (*arg)) | |
662 | { | |
663 | /* Replace a sequence of whitespace with a single space. */ | |
1739cf24 | 664 | dest->appendc (' '); |
2fae03e8 TT |
665 | while (len > 1 && macro_is_whitespace (arg[1])) |
666 | { | |
667 | ++arg; | |
668 | --len; | |
669 | } | |
670 | } | |
671 | else if (*arg == '\\' || *arg == '"') | |
672 | { | |
1739cf24 TT |
673 | dest->appendc ('\\'); |
674 | dest->appendc (*arg); | |
2fae03e8 TT |
675 | } |
676 | else | |
1739cf24 | 677 | dest->appendc (*arg); |
2fae03e8 TT |
678 | ++arg; |
679 | --len; | |
680 | } | |
1739cf24 | 681 | dest->appendc ('"'); |
2fae03e8 TT |
682 | dest->last_token = dest->len; |
683 | } | |
ec2bcbe7 | 684 | |
abc9d0dc TT |
685 | /* See macroexp.h. */ |
686 | ||
a36158ec | 687 | gdb::unique_xmalloc_ptr<char> |
abc9d0dc TT |
688 | macro_stringify (const char *str) |
689 | { | |
abc9d0dc | 690 | int len = strlen (str); |
ac3d4064 | 691 | growable_macro_buffer buffer (len); |
abc9d0dc | 692 | |
abc9d0dc | 693 | stringify (&buffer, str, len); |
1739cf24 | 694 | buffer.appendc ('\0'); |
abc9d0dc | 695 | |
a36158ec | 696 | return buffer.release (); |
abc9d0dc TT |
697 | } |
698 | ||
ec2bcbe7 JB |
699 | \f |
700 | /* Expanding macros! */ | |
701 | ||
702 | ||
703 | /* A singly-linked list of the names of the macros we are currently | |
704 | expanding --- for detecting expansion loops. */ | |
705 | struct macro_name_list { | |
706 | const char *name; | |
707 | struct macro_name_list *next; | |
708 | }; | |
709 | ||
710 | ||
711 | /* Return non-zero if we are currently expanding the macro named NAME, | |
712 | according to LIST; otherwise, return zero. | |
713 | ||
714 | You know, it would be possible to get rid of all the NO_LOOP | |
715 | arguments to these functions by simply generating a new lookup | |
716 | function and baton which refuses to find the definition for a | |
717 | particular macro, and otherwise delegates the decision to another | |
718 | function/baton pair. But that makes the linked list of excluded | |
719 | macros chained through untyped baton pointers, which will make it | |
025bb325 | 720 | harder to debug. :( */ |
ec2bcbe7 JB |
721 | static int |
722 | currently_rescanning (struct macro_name_list *list, const char *name) | |
723 | { | |
724 | for (; list; list = list->next) | |
a86bc61c | 725 | if (strcmp (name, list->name) == 0) |
ec2bcbe7 JB |
726 | return 1; |
727 | ||
728 | return 0; | |
729 | } | |
730 | ||
731 | ||
732 | /* Gather the arguments to a macro expansion. | |
733 | ||
734 | NAME is the name of the macro being invoked. (It's only used for | |
735 | printing error messages.) | |
736 | ||
737 | Assume that SRC is the text of the macro invocation immediately | |
738 | following the macro name. For example, if we're processing the | |
739 | text foo(bar, baz), then NAME would be foo and SRC will be (bar, | |
740 | baz). | |
741 | ||
742 | If SRC doesn't start with an open paren ( token at all, return | |
1739cf24 | 743 | false, leave SRC unchanged, and don't set *ARGS_PTR to anything. |
ec2bcbe7 JB |
744 | |
745 | If SRC doesn't contain a properly terminated argument list, then | |
746 | raise an error. | |
1739cf24 | 747 | |
2fae03e8 TT |
748 | For a variadic macro, NARGS holds the number of formal arguments to |
749 | the macro. For a GNU-style variadic macro, this should be the | |
750 | number of named arguments. For a non-variadic macro, NARGS should | |
751 | be -1. | |
ec2bcbe7 | 752 | |
1739cf24 TT |
753 | Otherwise, return true and set *ARGS_PTR to a vector of macro |
754 | buffers referring to the argument texts. The macro buffers share | |
755 | their text with SRC, and their last_token fields are initialized. | |
ec2bcbe7 JB |
756 | |
757 | NOTE WELL: if SRC starts with a open paren ( token followed | |
758 | immediately by a close paren ) token (e.g., the invocation looks | |
759 | like "foo()"), we treat that as one argument, which happens to be | |
760 | the empty list of tokens. The caller should keep in mind that such | |
761 | a sequence of tokens is a valid way to invoke one-parameter | |
762 | function-like macros, but also a valid way to invoke zero-parameter | |
763 | function-like macros. Eeew. | |
764 | ||
765 | Consume the tokens from SRC; after this call, SRC contains the text | |
766 | following the invocation. */ | |
767 | ||
1739cf24 | 768 | static bool |
ac3d4064 PA |
769 | gather_arguments (const char *name, shared_macro_buffer *src, int nargs, |
770 | std::vector<shared_macro_buffer> *args_ptr) | |
ec2bcbe7 | 771 | { |
ac3d4064 PA |
772 | shared_macro_buffer tok; |
773 | std::vector<shared_macro_buffer> args; | |
ec2bcbe7 JB |
774 | |
775 | /* Does SRC start with an opening paren token? Read from a copy of | |
776 | SRC, so SRC itself is unaffected if we don't find an opening | |
777 | paren. */ | |
778 | { | |
ac3d4064 | 779 | shared_macro_buffer temp (src->text, src->len); |
ec2bcbe7 JB |
780 | |
781 | if (! get_token (&tok, &temp) | |
dda83cd7 SM |
782 | || tok.len != 1 |
783 | || tok.text[0] != '(') | |
1739cf24 | 784 | return false; |
ec2bcbe7 JB |
785 | } |
786 | ||
787 | /* Consume SRC's opening paren. */ | |
788 | get_token (&tok, src); | |
789 | ||
ec2bcbe7 JB |
790 | for (;;) |
791 | { | |
ac3d4064 | 792 | shared_macro_buffer *arg; |
ec2bcbe7 JB |
793 | int depth; |
794 | ||
ec2bcbe7 | 795 | /* Initialize the next argument. */ |
1739cf24 TT |
796 | args.emplace_back (); |
797 | arg = &args.back (); | |
ec2bcbe7 JB |
798 | set_token (arg, src->text, src->text); |
799 | ||
800 | /* Gather the argument's tokens. */ | |
801 | depth = 0; | |
802 | for (;;) | |
dda83cd7 SM |
803 | { |
804 | if (! get_token (&tok, src)) | |
805 | error (_("Malformed argument list for macro `%s'."), name); | |
806 | ||
807 | /* Is tok an opening paren? */ | |
808 | if (tok.len == 1 && tok.text[0] == '(') | |
809 | depth++; | |
810 | ||
811 | /* Is tok is a closing paren? */ | |
812 | else if (tok.len == 1 && tok.text[0] == ')') | |
813 | { | |
814 | /* If it's a closing paren at the top level, then that's | |
815 | the end of the argument list. */ | |
816 | if (depth == 0) | |
817 | { | |
2fae03e8 TT |
818 | /* In the varargs case, the last argument may be |
819 | missing. Add an empty argument in this case. */ | |
1739cf24 | 820 | if (nargs != -1 && args.size () == nargs - 1) |
2fae03e8 | 821 | { |
1739cf24 TT |
822 | args.emplace_back (); |
823 | arg = &args.back (); | |
2fae03e8 TT |
824 | set_token (arg, src->text, src->text); |
825 | } | |
826 | ||
1739cf24 TT |
827 | *args_ptr = std::move (args); |
828 | return true; | |
dda83cd7 | 829 | } |
ec2bcbe7 | 830 | |
dda83cd7 SM |
831 | depth--; |
832 | } | |
ec2bcbe7 | 833 | |
dda83cd7 SM |
834 | /* If tok is a comma at top level, then that's the end of |
835 | the current argument. However, if we are handling a | |
836 | variadic macro and we are computing the last argument, we | |
837 | want to include the comma and remaining tokens. */ | |
838 | else if (tok.len == 1 && tok.text[0] == ',' && depth == 0 | |
1739cf24 | 839 | && (nargs == -1 || args.size () < nargs)) |
dda83cd7 SM |
840 | break; |
841 | ||
842 | /* Extend the current argument to enclose this token. If | |
843 | this is the current argument's first token, leave out any | |
844 | leading whitespace, just for aesthetics. */ | |
845 | if (arg->len == 0) | |
846 | { | |
847 | arg->text = tok.text; | |
848 | arg->len = tok.len; | |
849 | arg->last_token = 0; | |
850 | } | |
851 | else | |
852 | { | |
853 | arg->len = (tok.text + tok.len) - arg->text; | |
854 | arg->last_token = tok.text - arg->text; | |
855 | } | |
856 | } | |
ec2bcbe7 JB |
857 | } |
858 | } | |
859 | ||
860 | ||
861 | /* The `expand' and `substitute_args' functions both invoke `scan' | |
862 | recursively, so we need a forward declaration somewhere. */ | |
ac3d4064 PA |
863 | static void scan (growable_macro_buffer *dest, |
864 | shared_macro_buffer *src, | |
dda83cd7 | 865 | struct macro_name_list *no_loop, |
211d5b1c | 866 | const macro_scope &scope); |
ec2bcbe7 | 867 | |
2fae03e8 TT |
868 | /* A helper function for substitute_args. |
869 | ||
870 | ARGV is a vector of all the arguments; ARGC is the number of | |
871 | arguments. IS_VARARGS is true if the macro being substituted is a | |
872 | varargs macro; in this case VA_ARG_NAME is the name of the | |
873 | "variable" argument. VA_ARG_NAME is ignored if IS_VARARGS is | |
874 | false. | |
875 | ||
876 | If the token TOK is the name of a parameter, return the parameter's | |
877 | index. If TOK is not an argument, return -1. */ | |
878 | ||
879 | static int | |
ac3d4064 PA |
880 | find_parameter (const shared_macro_buffer *tok, |
881 | int is_varargs, const shared_macro_buffer *va_arg_name, | |
2fae03e8 TT |
882 | int argc, const char * const *argv) |
883 | { | |
884 | int i; | |
885 | ||
886 | if (! tok->is_identifier) | |
887 | return -1; | |
888 | ||
889 | for (i = 0; i < argc; ++i) | |
3e43a32a MS |
890 | if (tok->len == strlen (argv[i]) |
891 | && !memcmp (tok->text, argv[i], tok->len)) | |
2fae03e8 TT |
892 | return i; |
893 | ||
894 | if (is_varargs && tok->len == va_arg_name->len | |
895 | && ! memcmp (tok->text, va_arg_name->text, tok->len)) | |
896 | return argc - 1; | |
897 | ||
898 | return -1; | |
899 | } | |
900 | ||
a9bbfbd8 TT |
901 | /* Helper function for substitute_args that gets the next token and |
902 | updates the passed-in state variables. */ | |
903 | ||
904 | static void | |
ac3d4064 PA |
905 | get_next_token_for_substitution (shared_macro_buffer *replacement_list, |
906 | shared_macro_buffer *token, | |
907 | const char **start, | |
908 | shared_macro_buffer *lookahead, | |
909 | const char **lookahead_start, | |
a9bbfbd8 TT |
910 | int *lookahead_valid, |
911 | bool *keep_going) | |
912 | { | |
913 | if (!*lookahead_valid) | |
914 | *keep_going = false; | |
915 | else | |
916 | { | |
917 | *keep_going = true; | |
918 | *token = *lookahead; | |
919 | *start = *lookahead_start; | |
920 | *lookahead_start = replacement_list->text; | |
921 | *lookahead_valid = get_token (lookahead, replacement_list); | |
922 | } | |
923 | } | |
924 | ||
ec2bcbe7 | 925 | /* Given the macro definition DEF, being invoked with the actual |
1739cf24 | 926 | arguments given by ARGV, substitute the arguments into the |
ec2bcbe7 JB |
927 | replacement list, and store the result in DEST. |
928 | ||
2fae03e8 TT |
929 | IS_VARARGS should be true if DEF is a varargs macro. In this case, |
930 | VA_ARG_NAME should be the name of the "variable" argument -- either | |
931 | __VA_ARGS__ for c99-style varargs, or the final argument name, for | |
932 | GNU-style varargs. If IS_VARARGS is false, this parameter is | |
933 | ignored. | |
934 | ||
ec2bcbe7 JB |
935 | If it is necessary to expand macro invocations in one of the |
936 | arguments, use LOOKUP_FUNC and LOOKUP_BATON to find the macro | |
937 | definitions, and don't expand invocations of the macros listed in | |
938 | NO_LOOP. */ | |
2fae03e8 | 939 | |
ec2bcbe7 | 940 | static void |
ac3d4064 | 941 | substitute_args (growable_macro_buffer *dest, |
dda83cd7 | 942 | struct macro_definition *def, |
ac3d4064 PA |
943 | int is_varargs, const shared_macro_buffer *va_arg_name, |
944 | const std::vector<shared_macro_buffer> &argv, | |
dda83cd7 | 945 | struct macro_name_list *no_loop, |
211d5b1c | 946 | const macro_scope &scope) |
ec2bcbe7 | 947 | { |
2fae03e8 | 948 | /* The token we are currently considering. */ |
ac3d4064 | 949 | shared_macro_buffer tok; |
2fae03e8 | 950 | /* The replacement list's pointer from just before TOK was lexed. */ |
ac3d4064 | 951 | const char *original_rl_start; |
2fae03e8 | 952 | /* We have a single lookahead token to handle token splicing. */ |
ac3d4064 | 953 | shared_macro_buffer lookahead; |
2fae03e8 TT |
954 | /* The lookahead token might not be valid. */ |
955 | int lookahead_valid; | |
956 | /* The replacement list's pointer from just before LOOKAHEAD was | |
957 | lexed. */ | |
ac3d4064 | 958 | const char *lookahead_rl_start; |
ec2bcbe7 | 959 | |
1739cf24 | 960 | /* A macro buffer for the macro's replacement list. */ |
ac3d4064 | 961 | shared_macro_buffer replacement_list (def->replacement, |
1739cf24 | 962 | strlen (def->replacement)); |
ec2bcbe7 JB |
963 | |
964 | gdb_assert (dest->len == 0); | |
965 | dest->last_token = 0; | |
966 | ||
2fae03e8 TT |
967 | original_rl_start = replacement_list.text; |
968 | if (! get_token (&tok, &replacement_list)) | |
969 | return; | |
970 | lookahead_rl_start = replacement_list.text; | |
971 | lookahead_valid = get_token (&lookahead, &replacement_list); | |
972 | ||
a9bbfbd8 TT |
973 | /* __VA_OPT__ state variable. The states are: |
974 | 0 - nothing happening | |
975 | 1 - saw __VA_OPT__ | |
976 | >= 2 in __VA_OPT__, the value encodes the parenthesis depth. */ | |
977 | unsigned vaopt_state = 0; | |
978 | ||
979 | for (bool keep_going = true; | |
980 | keep_going; | |
981 | get_next_token_for_substitution (&replacement_list, | |
982 | &tok, | |
983 | &original_rl_start, | |
984 | &lookahead, | |
985 | &lookahead_rl_start, | |
986 | &lookahead_valid, | |
987 | &keep_going)) | |
ec2bcbe7 | 988 | { |
a9bbfbd8 | 989 | bool token_is_vaopt = (tok.len == 10 |
733f5eea | 990 | && startswith (tok.text, "__VA_OPT__")); |
a9bbfbd8 TT |
991 | |
992 | if (vaopt_state > 0) | |
993 | { | |
994 | if (token_is_vaopt) | |
995 | error (_("__VA_OPT__ cannot appear inside __VA_OPT__")); | |
996 | else if (tok.len == 1 && tok.text[0] == '(') | |
997 | { | |
998 | ++vaopt_state; | |
999 | /* We just entered __VA_OPT__, so don't emit this | |
1000 | token. */ | |
1001 | continue; | |
1002 | } | |
1003 | else if (vaopt_state == 1) | |
1004 | error (_("__VA_OPT__ must be followed by an open parenthesis")); | |
1005 | else if (tok.len == 1 && tok.text[0] == ')') | |
1006 | { | |
1007 | --vaopt_state; | |
1008 | if (vaopt_state == 1) | |
1009 | { | |
1010 | /* Done with __VA_OPT__. */ | |
1011 | vaopt_state = 0; | |
1012 | /* Don't emit. */ | |
1013 | continue; | |
1014 | } | |
1015 | } | |
1016 | ||
1017 | /* If __VA_ARGS__ is empty, then drop the contents of | |
1018 | __VA_OPT__. */ | |
1739cf24 | 1019 | if (argv.back ().len == 0) |
a9bbfbd8 TT |
1020 | continue; |
1021 | } | |
1022 | else if (token_is_vaopt) | |
1023 | { | |
1024 | if (!is_varargs) | |
1025 | error (_("__VA_OPT__ is only valid in a variadic macro")); | |
1026 | vaopt_state = 1; | |
1027 | /* Don't emit this token. */ | |
1028 | continue; | |
1029 | } | |
1030 | ||
ec2bcbe7 | 1031 | /* Just for aesthetics. If we skipped some whitespace, copy |
dda83cd7 | 1032 | that to DEST. */ |
ec2bcbe7 | 1033 | if (tok.text > original_rl_start) |
dda83cd7 SM |
1034 | { |
1035 | dest->appendmem (original_rl_start, tok.text - original_rl_start); | |
1036 | dest->last_token = dest->len; | |
1037 | } | |
ec2bcbe7 JB |
1038 | |
1039 | /* Is this token the stringification operator? */ | |
1040 | if (tok.len == 1 | |
dda83cd7 | 1041 | && tok.text[0] == '#') |
2fae03e8 TT |
1042 | { |
1043 | int arg; | |
ec2bcbe7 | 1044 | |
2fae03e8 TT |
1045 | if (!lookahead_valid) |
1046 | error (_("Stringification operator requires an argument.")); | |
ec2bcbe7 | 1047 | |
2fae03e8 TT |
1048 | arg = find_parameter (&lookahead, is_varargs, va_arg_name, |
1049 | def->argc, def->argv); | |
1050 | if (arg == -1) | |
1051 | error (_("Argument to stringification operator must name " | |
1052 | "a macro parameter.")); | |
ec2bcbe7 | 1053 | |
2fae03e8 TT |
1054 | stringify (dest, argv[arg].text, argv[arg].len); |
1055 | ||
1056 | /* Read one token and let the loop iteration code handle the | |
1057 | rest. */ | |
1058 | lookahead_rl_start = replacement_list.text; | |
1059 | lookahead_valid = get_token (&lookahead, &replacement_list); | |
1060 | } | |
1061 | /* Is this token the splicing operator? */ | |
1062 | else if (tok.len == 2 | |
1063 | && tok.text[0] == '#' | |
1064 | && tok.text[1] == '#') | |
1065 | error (_("Stray splicing operator")); | |
1066 | /* Is the next token the splicing operator? */ | |
1067 | else if (lookahead_valid | |
1068 | && lookahead.len == 2 | |
1069 | && lookahead.text[0] == '#' | |
1070 | && lookahead.text[1] == '#') | |
1071 | { | |
308d96ed | 1072 | int finished = 0; |
2fae03e8 TT |
1073 | int prev_was_comma = 0; |
1074 | ||
1075 | /* Note that GCC warns if the result of splicing is not a | |
1076 | token. In the debugger there doesn't seem to be much | |
1077 | benefit from doing this. */ | |
1078 | ||
1079 | /* Insert the first token. */ | |
1080 | if (tok.len == 1 && tok.text[0] == ',') | |
1081 | prev_was_comma = 1; | |
1082 | else | |
1083 | { | |
1084 | int arg = find_parameter (&tok, is_varargs, va_arg_name, | |
1085 | def->argc, def->argv); | |
b8d56208 | 1086 | |
2fae03e8 | 1087 | if (arg != -1) |
1739cf24 | 1088 | dest->appendmem (argv[arg].text, argv[arg].len); |
2fae03e8 | 1089 | else |
1739cf24 | 1090 | dest->appendmem (tok.text, tok.len); |
2fae03e8 TT |
1091 | } |
1092 | ||
1093 | /* Apply a possible sequence of ## operators. */ | |
1094 | for (;;) | |
1095 | { | |
1096 | if (! get_token (&tok, &replacement_list)) | |
1097 | error (_("Splicing operator at end of macro")); | |
1098 | ||
1099 | /* Handle a comma before a ##. If we are handling | |
1100 | varargs, and the token on the right hand side is the | |
1101 | varargs marker, and the final argument is empty or | |
1102 | missing, then drop the comma. This is a GNU | |
1103 | extension. There is one ambiguous case here, | |
1104 | involving pedantic behavior with an empty argument, | |
1105 | but we settle that in favor of GNU-style (GCC uses an | |
1106 | option). If we aren't dealing with varargs, we | |
1107 | simply insert the comma. */ | |
1108 | if (prev_was_comma) | |
1109 | { | |
1110 | if (! (is_varargs | |
1111 | && tok.len == va_arg_name->len | |
1112 | && !memcmp (tok.text, va_arg_name->text, tok.len) | |
1739cf24 TT |
1113 | && argv.back ().len == 0)) |
1114 | dest->appendmem (",", 1); | |
2fae03e8 TT |
1115 | prev_was_comma = 0; |
1116 | } | |
1117 | ||
1118 | /* Insert the token. If it is a parameter, insert the | |
1119 | argument. If it is a comma, treat it specially. */ | |
1120 | if (tok.len == 1 && tok.text[0] == ',') | |
1121 | prev_was_comma = 1; | |
1122 | else | |
1123 | { | |
1124 | int arg = find_parameter (&tok, is_varargs, va_arg_name, | |
1125 | def->argc, def->argv); | |
b8d56208 | 1126 | |
2fae03e8 | 1127 | if (arg != -1) |
1739cf24 | 1128 | dest->appendmem (argv[arg].text, argv[arg].len); |
2fae03e8 | 1129 | else |
1739cf24 | 1130 | dest->appendmem (tok.text, tok.len); |
2fae03e8 TT |
1131 | } |
1132 | ||
1133 | /* Now read another token. If it is another splice, we | |
1134 | loop. */ | |
1135 | original_rl_start = replacement_list.text; | |
1136 | if (! get_token (&tok, &replacement_list)) | |
1137 | { | |
1138 | finished = 1; | |
1139 | break; | |
1140 | } | |
1141 | ||
1142 | if (! (tok.len == 2 | |
1143 | && tok.text[0] == '#' | |
1144 | && tok.text[1] == '#')) | |
1145 | break; | |
1146 | } | |
1147 | ||
1148 | if (prev_was_comma) | |
1149 | { | |
1150 | /* We saw a comma. Insert it now. */ | |
1739cf24 | 1151 | dest->appendmem (",", 1); |
2fae03e8 TT |
1152 | } |
1153 | ||
dda83cd7 | 1154 | dest->last_token = dest->len; |
2fae03e8 TT |
1155 | if (finished) |
1156 | lookahead_valid = 0; | |
1157 | else | |
1158 | { | |
1159 | /* Set up for the loop iterator. */ | |
1160 | lookahead = tok; | |
1161 | lookahead_rl_start = original_rl_start; | |
1162 | lookahead_valid = 1; | |
1163 | } | |
1164 | } | |
1165 | else | |
1166 | { | |
1167 | /* Is this token an identifier? */ | |
1168 | int substituted = 0; | |
1169 | int arg = find_parameter (&tok, is_varargs, va_arg_name, | |
1170 | def->argc, def->argv); | |
1171 | ||
1172 | if (arg != -1) | |
1173 | { | |
2fae03e8 TT |
1174 | /* Expand any macro invocations in the argument text, |
1175 | and append the result to dest. Remember that scan | |
1176 | mutates its source, so we need to scan a new buffer | |
1177 | referring to the argument's text, not the argument | |
1178 | itself. */ | |
ac3d4064 | 1179 | shared_macro_buffer arg_src (argv[arg].text, argv[arg].len); |
211d5b1c | 1180 | scan (dest, &arg_src, no_loop, scope); |
2fae03e8 TT |
1181 | substituted = 1; |
1182 | } | |
1183 | ||
1184 | /* If it wasn't a parameter, then just copy it across. */ | |
1185 | if (! substituted) | |
1186 | append_tokens_without_splicing (dest, &tok); | |
1187 | } | |
ec2bcbe7 | 1188 | } |
a9bbfbd8 TT |
1189 | |
1190 | if (vaopt_state > 0) | |
1191 | error (_("Unterminated __VA_OPT__")); | |
ec2bcbe7 JB |
1192 | } |
1193 | ||
1194 | ||
1195 | /* Expand a call to a macro named ID, whose definition is DEF. Append | |
1196 | its expansion to DEST. SRC is the input text following the ID | |
1197 | token. We are currently rescanning the expansions of the macros | |
1198 | named in NO_LOOP; don't re-expand them. Use LOOKUP_FUNC and | |
025bb325 | 1199 | LOOKUP_BATON to find definitions for any nested macro references. |
ec2bcbe7 JB |
1200 | |
1201 | Return 1 if we decided to expand it, zero otherwise. (If it's a | |
1202 | function-like macro name that isn't followed by an argument list, | |
1203 | we don't expand it.) If we return zero, leave SRC unchanged. */ | |
1204 | static int | |
1205 | expand (const char *id, | |
ac3d4064 PA |
1206 | struct macro_definition *def, |
1207 | growable_macro_buffer *dest, | |
1208 | shared_macro_buffer *src, | |
dda83cd7 | 1209 | struct macro_name_list *no_loop, |
211d5b1c | 1210 | const macro_scope &scope) |
ec2bcbe7 JB |
1211 | { |
1212 | struct macro_name_list new_no_loop; | |
1213 | ||
1214 | /* Create a new node to be added to the front of the no-expand list. | |
1215 | This list is appropriate for re-scanning replacement lists, but | |
1216 | it is *not* appropriate for scanning macro arguments; invocations | |
1217 | of the macro whose arguments we are gathering *do* get expanded | |
1218 | there. */ | |
1219 | new_no_loop.name = id; | |
1220 | new_no_loop.next = no_loop; | |
1221 | ||
1222 | /* What kind of macro are we expanding? */ | |
1223 | if (def->kind == macro_object_like) | |
1224 | { | |
ac3d4064 | 1225 | shared_macro_buffer replacement_list (def->replacement, |
1739cf24 | 1226 | strlen (def->replacement)); |
ec2bcbe7 | 1227 | |
211d5b1c | 1228 | scan (dest, &replacement_list, &new_no_loop, scope); |
ec2bcbe7 JB |
1229 | return 1; |
1230 | } | |
1231 | else if (def->kind == macro_function_like) | |
1232 | { | |
ac3d4064 | 1233 | shared_macro_buffer va_arg_name; |
2fae03e8 TT |
1234 | int is_varargs = 0; |
1235 | ||
1236 | if (def->argc >= 1) | |
1237 | { | |
1238 | if (strcmp (def->argv[def->argc - 1], "...") == 0) | |
1239 | { | |
1240 | /* In C99-style varargs, substitution is done using | |
1241 | __VA_ARGS__. */ | |
1739cf24 | 1242 | va_arg_name.set_shared ("__VA_ARGS__", strlen ("__VA_ARGS__")); |
2fae03e8 TT |
1243 | is_varargs = 1; |
1244 | } | |
1245 | else | |
1246 | { | |
1247 | int len = strlen (def->argv[def->argc - 1]); | |
b8d56208 | 1248 | |
2fae03e8 TT |
1249 | if (len > 3 |
1250 | && strcmp (def->argv[def->argc - 1] + len - 3, "...") == 0) | |
1251 | { | |
1252 | /* In GNU-style varargs, the name of the | |
1253 | substitution parameter is the name of the formal | |
1254 | argument without the "...". */ | |
1739cf24 | 1255 | va_arg_name.set_shared (def->argv[def->argc - 1], len - 3); |
2fae03e8 TT |
1256 | is_varargs = 1; |
1257 | } | |
1258 | } | |
1259 | } | |
ec2bcbe7 | 1260 | |
ac3d4064 | 1261 | std::vector<shared_macro_buffer> argv; |
ec2bcbe7 | 1262 | /* If we couldn't find any argument list, then we don't expand |
dda83cd7 | 1263 | this macro. */ |
1739cf24 TT |
1264 | if (!gather_arguments (id, src, is_varargs ? def->argc : -1, |
1265 | &argv)) | |
1266 | return 0; | |
ec2bcbe7 JB |
1267 | |
1268 | /* Check that we're passing an acceptable number of arguments for | |
dda83cd7 | 1269 | this macro. */ |
1739cf24 | 1270 | if (argv.size () != def->argc) |
dda83cd7 | 1271 | { |
1739cf24 | 1272 | if (is_varargs && argv.size () >= def->argc - 1) |
2fae03e8 TT |
1273 | { |
1274 | /* Ok. */ | |
1275 | } | |
dda83cd7 SM |
1276 | /* Remember that a sequence of tokens like "foo()" is a |
1277 | valid invocation of a macro expecting either zero or one | |
1278 | arguments. */ | |
1279 | else if (! (argv.size () == 1 | |
2fae03e8 TT |
1280 | && argv[0].len == 0 |
1281 | && def->argc == 0)) | |
dda83cd7 SM |
1282 | error (_("Wrong number of arguments to macro `%s' " |
1283 | "(expected %d, got %d)."), | |
1284 | id, def->argc, int (argv.size ())); | |
1285 | } | |
ec2bcbe7 JB |
1286 | |
1287 | /* Note that we don't expand macro invocations in the arguments | |
dda83cd7 SM |
1288 | yet --- we let subst_args take care of that. Parameters that |
1289 | appear as operands of the stringifying operator "#" or the | |
1290 | splicing operator "##" don't get macro references expanded, | |
1291 | so we can't really tell whether it's appropriate to macro- | |
1292 | expand an argument until we see how it's being used. */ | |
ac3d4064 | 1293 | growable_macro_buffer substituted (0); |
2fae03e8 | 1294 | substitute_args (&substituted, def, is_varargs, &va_arg_name, |
211d5b1c | 1295 | argv, no_loop, scope); |
ec2bcbe7 JB |
1296 | |
1297 | /* Now `substituted' is the macro's replacement list, with all | |
ac3d4064 | 1298 | argument values substituted into it properly. Re-scan it for |
dda83cd7 SM |
1299 | macro references, but don't expand invocations of this macro. |
1300 | ||
1301 | We create a new buffer, `substituted_src', which points into | |
ac3d4064 | 1302 | `substituted', and scan that. We can't scan `substituted' |
dda83cd7 SM |
1303 | itself, since the tokenization process moves the buffer's |
1304 | text pointer around, and we still need to be able to find | |
1305 | `substituted's original text buffer after scanning it so we | |
1306 | can free it. */ | |
ac3d4064 | 1307 | shared_macro_buffer substituted_src (substituted.text, substituted.len); |
211d5b1c | 1308 | scan (dest, &substituted_src, &new_no_loop, scope); |
ec2bcbe7 | 1309 | |
ec2bcbe7 JB |
1310 | return 1; |
1311 | } | |
1312 | else | |
f34652de | 1313 | internal_error (_("bad macro definition kind")); |
ec2bcbe7 JB |
1314 | } |
1315 | ||
1316 | ||
1317 | /* If the single token in SRC_FIRST followed by the tokens in SRC_REST | |
30baf67b | 1318 | constitute a macro invocation not forbidden in NO_LOOP, append its |
ec2bcbe7 JB |
1319 | expansion to DEST and return non-zero. Otherwise, return zero, and |
1320 | leave DEST unchanged. | |
1321 | ||
ec2bcbe7 JB |
1322 | SRC_FIRST must be a string built by get_token. */ |
1323 | static int | |
ac3d4064 PA |
1324 | maybe_expand (growable_macro_buffer *dest, |
1325 | shared_macro_buffer *src_first, | |
1326 | shared_macro_buffer *src_rest, | |
dda83cd7 | 1327 | struct macro_name_list *no_loop, |
211d5b1c | 1328 | const macro_scope &scope) |
ec2bcbe7 | 1329 | { |
ec2bcbe7 JB |
1330 | /* Is this token an identifier? */ |
1331 | if (src_first->is_identifier) | |
1332 | { | |
1333 | /* Make a null-terminated copy of it, since that's what our | |
dda83cd7 | 1334 | lookup function expects. */ |
0354904b | 1335 | std::string id (src_first->text, src_first->len); |
b8d56208 | 1336 | |
ec2bcbe7 | 1337 | /* If we're currently re-scanning the result of expanding |
dda83cd7 | 1338 | this macro, don't expand it again. */ |
0354904b | 1339 | if (! currently_rescanning (no_loop, id.c_str ())) |
dda83cd7 SM |
1340 | { |
1341 | /* Does this identifier have a macro definition in scope? */ | |
1342 | macro_definition *def = standard_macro_lookup (id.c_str (), scope); | |
ec2bcbe7 | 1343 | |
dda83cd7 | 1344 | if (def && expand (id.c_str (), def, dest, src_rest, no_loop, scope)) |
0354904b | 1345 | return 1; |
dda83cd7 | 1346 | } |
ec2bcbe7 JB |
1347 | } |
1348 | ||
1349 | return 0; | |
1350 | } | |
1351 | ||
1352 | ||
1353 | /* Expand macro references in SRC, appending the results to DEST. | |
1354 | Assume we are re-scanning the result of expanding the macros named | |
ac3d4064 | 1355 | in NO_LOOP, and don't try to re-expand references to them. */ |
ec2bcbe7 | 1356 | |
ec2bcbe7 | 1357 | static void |
ac3d4064 PA |
1358 | scan (growable_macro_buffer *dest, |
1359 | shared_macro_buffer *src, | |
ec2bcbe7 | 1360 | struct macro_name_list *no_loop, |
211d5b1c | 1361 | const macro_scope &scope) |
ec2bcbe7 | 1362 | { |
ec2bcbe7 JB |
1363 | |
1364 | for (;;) | |
1365 | { | |
ac3d4064 PA |
1366 | shared_macro_buffer tok; |
1367 | const char *original_src_start = src->text; | |
ec2bcbe7 JB |
1368 | |
1369 | /* Find the next token in SRC. */ | |
1370 | if (! get_token (&tok, src)) | |
dda83cd7 | 1371 | break; |
ec2bcbe7 JB |
1372 | |
1373 | /* Just for aesthetics. If we skipped some whitespace, copy | |
dda83cd7 | 1374 | that to DEST. */ |
ec2bcbe7 | 1375 | if (tok.text > original_src_start) |
dda83cd7 SM |
1376 | { |
1377 | dest->appendmem (original_src_start, tok.text - original_src_start); | |
1378 | dest->last_token = dest->len; | |
1379 | } | |
ec2bcbe7 | 1380 | |
211d5b1c | 1381 | if (! maybe_expand (dest, &tok, src, no_loop, scope)) |
dda83cd7 SM |
1382 | /* We didn't end up expanding tok as a macro reference, so |
1383 | simply append it to dest. */ | |
1384 | append_tokens_without_splicing (dest, &tok); | |
ec2bcbe7 JB |
1385 | } |
1386 | ||
1387 | /* Just for aesthetics. If there was any trailing whitespace in | |
1388 | src, copy it to dest. */ | |
1389 | if (src->len) | |
1390 | { | |
1739cf24 | 1391 | dest->appendmem (src->text, src->len); |
ec2bcbe7 JB |
1392 | dest->last_token = dest->len; |
1393 | } | |
1394 | } | |
1395 | ||
1396 | ||
f6c2623e | 1397 | gdb::unique_xmalloc_ptr<char> |
211d5b1c | 1398 | macro_expand (const char *source, const macro_scope &scope) |
ec2bcbe7 | 1399 | { |
ac3d4064 | 1400 | shared_macro_buffer src (source, strlen (source)); |
ec2bcbe7 | 1401 | |
ac3d4064 | 1402 | growable_macro_buffer dest (0); |
ec2bcbe7 | 1403 | dest.last_token = 0; |
ec2bcbe7 | 1404 | |
211d5b1c | 1405 | scan (&dest, &src, 0, scope); |
ec2bcbe7 | 1406 | |
1739cf24 | 1407 | dest.appendc ('\0'); |
ec2bcbe7 | 1408 | |
14d960c8 | 1409 | return dest.release (); |
ec2bcbe7 JB |
1410 | } |
1411 | ||
1412 | ||
f6c2623e | 1413 | gdb::unique_xmalloc_ptr<char> |
211d5b1c | 1414 | macro_expand_once (const char *source, const macro_scope &scope) |
ec2bcbe7 | 1415 | { |
8a3fe4f8 | 1416 | error (_("Expand-once not implemented yet.")); |
ec2bcbe7 JB |
1417 | } |
1418 | ||
14d960c8 | 1419 | gdb::unique_xmalloc_ptr<char> |
211d5b1c | 1420 | macro_expand_next (const char **lexptr, const macro_scope &scope) |
ec2bcbe7 | 1421 | { |
ac3d4064 | 1422 | shared_macro_buffer tok; |
ec2bcbe7 JB |
1423 | |
1424 | /* Set up SRC to refer to the input text, pointed to by *lexptr. */ | |
ac3d4064 | 1425 | shared_macro_buffer src (*lexptr, strlen (*lexptr)); |
ec2bcbe7 JB |
1426 | |
1427 | /* Set up DEST to receive the expansion, if there is one. */ | |
ac3d4064 | 1428 | growable_macro_buffer dest (0); |
ec2bcbe7 | 1429 | dest.last_token = 0; |
ec2bcbe7 JB |
1430 | |
1431 | /* Get the text's first preprocessing token. */ | |
1432 | if (! get_token (&tok, &src)) | |
14d960c8 | 1433 | return nullptr; |
ec2bcbe7 JB |
1434 | |
1435 | /* If it's a macro invocation, expand it. */ | |
211d5b1c | 1436 | if (maybe_expand (&dest, &tok, &src, 0, scope)) |
ec2bcbe7 JB |
1437 | { |
1438 | /* It was a macro invocation! Package up the expansion as a | |
dda83cd7 SM |
1439 | null-terminated string and return it. Set *lexptr to the |
1440 | start of the next token in the input. */ | |
1739cf24 | 1441 | dest.appendc ('\0'); |
ec2bcbe7 | 1442 | *lexptr = src.text; |
1739cf24 | 1443 | return dest.release (); |
ec2bcbe7 JB |
1444 | } |
1445 | else | |
1446 | { | |
1447 | /* It wasn't a macro invocation. */ | |
14d960c8 | 1448 | return nullptr; |
ec2bcbe7 JB |
1449 | } |
1450 | } |