1 | /* Mainly the interface between cpplib and the C front ends. |
2 | Copyright (C) 1987-2023 Free Software Foundation, Inc. |
3 | |
4 | This file is part of GCC. |
5 | |
6 | GCC is free software; you can redistribute it and/or modify it under |
7 | the terms of the GNU General Public License as published by the Free |
8 | Software Foundation; either version 3, or (at your option) any later |
9 | version. |
10 | |
11 | GCC is distributed in the hope that it will be useful, but WITHOUT ANY |
12 | WARRANTY; without even the implied warranty of MERCHANTABILITY or |
13 | FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License |
14 | for more details. |
15 | |
16 | You should have received a copy of the GNU General Public License |
17 | along with GCC; see the file COPYING3. If not see |
18 | <http://www.gnu.org/licenses/>. */ |
19 | |
20 | #include "config.h" |
21 | #include "system.h" |
22 | #include "coretypes.h" |
23 | #include "target.h" |
24 | #include "c-common.h" |
25 | #include "timevar.h" |
26 | #include "stringpool.h" |
27 | #include "stor-layout.h" |
28 | #include "c-pragma.h" |
29 | #include "debug.h" |
30 | #include "flags.h" |
31 | #include "file-prefix-map.h" /* remap_macro_filename() */ |
32 | #include "langhooks.h" |
33 | #include "attribs.h" |
34 | |
35 | /* We may keep statistics about how long which files took to compile. */ |
36 | static int , body_time; |
37 | static splay_tree file_info_tree; |
38 | |
39 | int pending_lang_change; /* If we need to switch languages - C++ only */ |
40 | int ; /* depth in C headers - C++ only */ |
41 | |
42 | static tree interpret_integer (const cpp_token *, unsigned int, |
43 | enum overflow_type *); |
44 | static tree interpret_float (const cpp_token *, unsigned int, const char *, |
45 | enum overflow_type *); |
46 | static tree interpret_fixed (const cpp_token *, unsigned int); |
47 | static enum integer_type_kind narrowest_unsigned_type |
48 | (const widest_int &, unsigned int); |
49 | static enum integer_type_kind narrowest_signed_type |
50 | (const widest_int &, unsigned int); |
51 | static enum cpp_ttype lex_string (const cpp_token *, tree *, bool, bool); |
52 | static tree lex_charconst (const cpp_token *); |
53 | static void update_header_times (const char *); |
54 | static int dump_one_header (splay_tree_node, void *); |
55 | static void cb_line_change (cpp_reader *, const cpp_token *, int); |
56 | static void cb_ident (cpp_reader *, unsigned int, const cpp_string *); |
57 | static void cb_def_pragma (cpp_reader *, unsigned int); |
58 | static void cb_define (cpp_reader *, unsigned int, cpp_hashnode *); |
59 | static void cb_undef (cpp_reader *, unsigned int, cpp_hashnode *); |
60 | |
61 | void |
62 | init_c_lex (void) |
63 | { |
64 | struct c_fileinfo *toplevel; |
65 | |
66 | /* The get_fileinfo data structure must be initialized before |
67 | cpp_read_main_file is called. */ |
68 | toplevel = get_fileinfo ("<top level>" ); |
69 | if (flag_detailed_statistics) |
70 | { |
71 | header_time = 0; |
72 | body_time = get_run_time (); |
73 | toplevel->time = body_time; |
74 | } |
75 | |
76 | struct cpp_callbacks *cb = cpp_get_callbacks (parse_in); |
77 | |
78 | cb->line_change = cb_line_change; |
79 | cb->ident = cb_ident; |
80 | cb->def_pragma = cb_def_pragma; |
81 | cb->valid_pch = c_common_valid_pch; |
82 | cb->read_pch = c_common_read_pch; |
83 | cb->has_attribute = c_common_has_attribute; |
84 | cb->has_builtin = c_common_has_builtin; |
85 | cb->get_source_date_epoch = cb_get_source_date_epoch; |
86 | cb->get_suggestion = cb_get_suggestion; |
87 | cb->remap_filename = remap_macro_filename; |
88 | |
89 | /* Set the debug callbacks if we can use them. */ |
90 | if ((debug_info_level == DINFO_LEVEL_VERBOSE |
91 | && dwarf_debuginfo_p ()) |
92 | || flag_dump_go_spec != NULL) |
93 | { |
94 | cb->define = cb_define; |
95 | cb->undef = cb_undef; |
96 | } |
97 | } |
98 | |
99 | struct c_fileinfo * |
100 | get_fileinfo (const char *name) |
101 | { |
102 | splay_tree_node n; |
103 | struct c_fileinfo *fi; |
104 | |
105 | if (!file_info_tree) |
106 | file_info_tree = splay_tree_new (splay_tree_compare_strings, |
107 | 0, |
108 | splay_tree_delete_pointers); |
109 | |
110 | n = splay_tree_lookup (file_info_tree, (splay_tree_key) name); |
111 | if (n) |
112 | return (struct c_fileinfo *) n->value; |
113 | |
114 | fi = XNEW (struct c_fileinfo); |
115 | fi->time = 0; |
116 | fi->interface_only = 0; |
117 | fi->interface_unknown = 1; |
118 | splay_tree_insert (file_info_tree, (splay_tree_key) name, |
119 | (splay_tree_value) fi); |
120 | return fi; |
121 | } |
122 | |
123 | static void |
124 | (const char *name) |
125 | { |
126 | /* Changing files again. This means currently collected time |
127 | is charged against header time, and body time starts back at 0. */ |
128 | if (flag_detailed_statistics) |
129 | { |
130 | int this_time = get_run_time (); |
131 | struct c_fileinfo *file = get_fileinfo (name); |
132 | header_time += this_time - body_time; |
133 | file->time += this_time - body_time; |
134 | body_time = this_time; |
135 | } |
136 | } |
137 | |
138 | static int |
139 | (splay_tree_node n, void * ARG_UNUSED (dummy)) |
140 | { |
141 | print_time ((const char *) n->key, |
142 | ((struct c_fileinfo *) n->value)->time); |
143 | return 0; |
144 | } |
145 | |
146 | void |
147 | dump_time_statistics (void) |
148 | { |
149 | struct c_fileinfo *file = get_fileinfo (LOCATION_FILE (input_location)); |
150 | int this_time = get_run_time (); |
151 | file->time += this_time - body_time; |
152 | |
153 | fprintf (stderr, format: "\n******\n" ); |
154 | print_time ("header files (total)" , header_time); |
155 | print_time ("main file (total)" , this_time - body_time); |
156 | fprintf (stderr, format: "ratio = %g : 1\n" , |
157 | (double) header_time / (double) (this_time - body_time)); |
158 | fprintf (stderr, format: "\n******\n" ); |
159 | |
160 | splay_tree_foreach (file_info_tree, dump_one_header, 0); |
161 | } |
162 | |
163 | static void |
164 | cb_ident (cpp_reader * ARG_UNUSED (pfile), |
165 | unsigned int ARG_UNUSED (line), |
166 | const cpp_string * ARG_UNUSED (str)) |
167 | { |
168 | if (!flag_no_ident) |
169 | { |
170 | /* Convert escapes in the string. */ |
171 | cpp_string cstr = { .len: 0, .text: 0 }; |
172 | if (cpp_interpret_string (pfile, str, 1, &cstr, CPP_STRING)) |
173 | { |
174 | targetm.asm_out.output_ident ((const char *) cstr.text); |
175 | free (CONST_CAST (unsigned char *, cstr.text)); |
176 | } |
177 | } |
178 | } |
179 | |
180 | /* Called at the start of every non-empty line. TOKEN is the first |
181 | lexed token on the line. Used for diagnostic line numbers. */ |
182 | static void |
183 | cb_line_change (cpp_reader * ARG_UNUSED (pfile), const cpp_token *token, |
184 | int parsing_args) |
185 | { |
186 | if (token->type != CPP_EOF && !parsing_args) |
187 | input_location = token->src_loc; |
188 | } |
189 | |
190 | void |
191 | fe_file_change (const line_map_ordinary *new_map) |
192 | { |
193 | if (new_map == NULL) |
194 | return; |
195 | |
196 | if (new_map->reason == LC_ENTER) |
197 | { |
198 | /* Don't stack the main buffer on the input stack; |
199 | we already did in compile_file. */ |
200 | if (!MAIN_FILE_P (ord_map: new_map)) |
201 | { |
202 | location_t included_at = linemap_included_from (ord_map: new_map); |
203 | int line = 0; |
204 | if (included_at > BUILTINS_LOCATION) |
205 | line = SOURCE_LINE (ord_map: new_map - 1, loc: included_at); |
206 | |
207 | input_location = new_map->start_location; |
208 | (*debug_hooks->start_source_file) (line, LINEMAP_FILE (ord_map: new_map)); |
209 | #ifdef SYSTEM_IMPLICIT_EXTERN_C |
210 | if (c_header_level) |
211 | ++c_header_level; |
212 | else if (LINEMAP_SYSP (new_map) == 2) |
213 | { |
214 | c_header_level = 1; |
215 | ++pending_lang_change; |
216 | } |
217 | #endif |
218 | } |
219 | } |
220 | else if (new_map->reason == LC_LEAVE) |
221 | { |
222 | #ifdef SYSTEM_IMPLICIT_EXTERN_C |
223 | if (c_header_level && --c_header_level == 0) |
224 | { |
225 | if (LINEMAP_SYSP (new_map) == 2) |
226 | warning (0, "badly nested C headers from preprocessor" ); |
227 | --pending_lang_change; |
228 | } |
229 | #endif |
230 | input_location = new_map->start_location; |
231 | |
232 | (*debug_hooks->end_source_file) (LINEMAP_LINE (ord_map: new_map)); |
233 | } |
234 | |
235 | update_header_times (name: LINEMAP_FILE (ord_map: new_map)); |
236 | input_location = new_map->start_location; |
237 | } |
238 | |
239 | static void |
240 | cb_def_pragma (cpp_reader *pfile, location_t loc) |
241 | { |
242 | /* Issue a warning message if we have been asked to do so. Ignore |
243 | unknown pragmas in system headers unless an explicit |
244 | -Wunknown-pragmas has been given. */ |
245 | if (warn_unknown_pragmas > in_system_header_at (loc: input_location)) |
246 | { |
247 | const unsigned char *space, *name; |
248 | const cpp_token *s; |
249 | location_t fe_loc = loc; |
250 | |
251 | space = name = (const unsigned char *) "" ; |
252 | |
253 | /* N.B. It's fine to call cpp_get_token () directly here (rather than our |
254 | local wrapper get_token ()), because this callback is not used with |
255 | flag_preprocess_only==true. */ |
256 | s = cpp_get_token (pfile); |
257 | if (s->type != CPP_EOF) |
258 | { |
259 | space = cpp_token_as_text (pfile, s); |
260 | s = cpp_get_token (pfile); |
261 | if (s->type == CPP_NAME) |
262 | name = cpp_token_as_text (pfile, s); |
263 | } |
264 | |
265 | warning_at (fe_loc, OPT_Wunknown_pragmas, "ignoring %<#pragma %s %s%>" , |
266 | space, name); |
267 | } |
268 | } |
269 | |
270 | /* #define callback for DWARF and DWARF2 debug info. */ |
271 | static void |
272 | cb_define (cpp_reader *pfile, location_t loc, cpp_hashnode *node) |
273 | { |
274 | const struct line_map *map = linemap_lookup (line_table, loc); |
275 | (*debug_hooks->define) (SOURCE_LINE (ord_map: linemap_check_ordinary (map), loc), |
276 | (const char *) cpp_macro_definition (pfile, node)); |
277 | } |
278 | |
279 | /* #undef callback for DWARF and DWARF2 debug info. */ |
280 | static void |
281 | cb_undef (cpp_reader *pfile, location_t loc, cpp_hashnode *node) |
282 | { |
283 | if (lang_hooks.preprocess_undef) |
284 | lang_hooks.preprocess_undef (pfile, loc, node); |
285 | |
286 | const struct line_map *map = linemap_lookup (line_table, loc); |
287 | (*debug_hooks->undef) (SOURCE_LINE (ord_map: linemap_check_ordinary (map), loc), |
288 | (const char *) NODE_NAME (node)); |
289 | } |
290 | |
291 | /* Wrapper around cpp_get_token_with_location to stream the token to the |
292 | preprocessor so it can output it. This is necessary with |
293 | flag_preprocess_only if we are obtaining tokens here instead of from the loop |
294 | in c-ppoutput.cc, such as while processing a #pragma. */ |
295 | |
296 | static const cpp_token * |
297 | get_token (cpp_reader *pfile, location_t *loc = nullptr) |
298 | { |
299 | if (flag_preprocess_only) |
300 | { |
301 | location_t x; |
302 | if (!loc) |
303 | loc = &x; |
304 | const auto tok = cpp_get_token_with_location (pfile, loc); |
305 | c_pp_stream_token (pfile, tok, loc: *loc); |
306 | return tok; |
307 | } |
308 | else |
309 | return cpp_get_token_with_location (pfile, loc); |
310 | } |
311 | |
312 | /* Wrapper around cpp_get_token to skip CPP_PADDING tokens |
313 | and not consume CPP_EOF. This does not perform the optional |
314 | streaming in preprocess_only mode, so is suitable to be used |
315 | when processing builtin expansions such as c_common_has_attribute. */ |
316 | |
317 | static const cpp_token * |
318 | get_token_no_padding (cpp_reader *pfile) |
319 | { |
320 | for (;;) |
321 | { |
322 | const cpp_token *ret = cpp_peek_token (pfile, 0); |
323 | if (ret->type == CPP_EOF) |
324 | return ret; |
325 | ret = cpp_get_token (pfile); |
326 | if (ret->type != CPP_PADDING) |
327 | return ret; |
328 | } |
329 | } |
330 | |
331 | /* Callback for has_attribute. */ |
332 | int |
333 | c_common_has_attribute (cpp_reader *pfile, bool std_syntax) |
334 | { |
335 | int result = 0; |
336 | tree attr_name = NULL_TREE; |
337 | const cpp_token *token; |
338 | |
339 | token = get_token_no_padding (pfile); |
340 | if (token->type != CPP_OPEN_PAREN) |
341 | { |
342 | cpp_error (pfile, CPP_DL_ERROR, |
343 | msgid: "missing '(' after \"__has_attribute\"" ); |
344 | return 0; |
345 | } |
346 | token = get_token_no_padding (pfile); |
347 | if (token->type == CPP_NAME) |
348 | { |
349 | attr_name = get_identifier ((const char *) |
350 | cpp_token_as_text (pfile, token)); |
351 | attr_name = canonicalize_attr_name (attr_name); |
352 | bool have_scope = false; |
353 | int idx = 0; |
354 | const cpp_token *nxt_token; |
355 | do |
356 | nxt_token = cpp_peek_token (pfile, idx++); |
357 | while (nxt_token->type == CPP_PADDING); |
358 | if (nxt_token->type == CPP_SCOPE) |
359 | { |
360 | have_scope = true; |
361 | get_token_no_padding (pfile); // Eat scope. |
362 | nxt_token = get_token_no_padding (pfile); |
363 | if (nxt_token->type == CPP_NAME) |
364 | { |
365 | tree attr_ns = attr_name; |
366 | tree attr_id |
367 | = get_identifier ((const char *) |
368 | cpp_token_as_text (pfile, nxt_token)); |
369 | attr_id = canonicalize_attr_name (attr_name: attr_id); |
370 | /* OpenMP attributes need special handling. */ |
371 | if ((flag_openmp || flag_openmp_simd) |
372 | && is_attribute_p (attr_name: "omp" , ident: attr_ns) |
373 | && (is_attribute_p (attr_name: "directive" , ident: attr_id) |
374 | || is_attribute_p (attr_name: "sequence" , ident: attr_id) |
375 | || is_attribute_p (attr_name: "decl" , ident: attr_id))) |
376 | result = 1; |
377 | if (result) |
378 | attr_name = NULL_TREE; |
379 | else |
380 | attr_name = build_tree_list (attr_ns, attr_id); |
381 | } |
382 | else |
383 | { |
384 | cpp_error (pfile, CPP_DL_ERROR, |
385 | msgid: "attribute identifier required after scope" ); |
386 | attr_name = NULL_TREE; |
387 | } |
388 | } |
389 | else |
390 | { |
391 | /* Some standard attributes need special handling. */ |
392 | if (c_dialect_cxx ()) |
393 | { |
394 | if (is_attribute_p (attr_name: "noreturn" , ident: attr_name)) |
395 | result = 200809; |
396 | else if (is_attribute_p (attr_name: "deprecated" , ident: attr_name)) |
397 | result = 201309; |
398 | else if (is_attribute_p (attr_name: "maybe_unused" , ident: attr_name) |
399 | || is_attribute_p (attr_name: "fallthrough" , ident: attr_name)) |
400 | result = 201603; |
401 | else if (is_attribute_p (attr_name: "no_unique_address" , ident: attr_name) |
402 | || is_attribute_p (attr_name: "likely" , ident: attr_name) |
403 | || is_attribute_p (attr_name: "unlikely" , ident: attr_name)) |
404 | result = 201803; |
405 | else if (is_attribute_p (attr_name: "nodiscard" , ident: attr_name)) |
406 | result = 201907; |
407 | else if (is_attribute_p (attr_name: "assume" , ident: attr_name)) |
408 | result = 202207; |
409 | else if (is_attribute_p (attr_name: "init_priority" , ident: attr_name)) |
410 | { |
411 | /* The (non-standard) init_priority attribute is always |
412 | included in the attribute table, but we don't want to |
413 | advertise the attribute unless the target actually |
414 | supports init priorities. */ |
415 | result = SUPPORTS_INIT_PRIORITY ? 1 : 0; |
416 | attr_name = NULL_TREE; |
417 | } |
418 | } |
419 | else |
420 | { |
421 | if (is_attribute_p (attr_name: "deprecated" , ident: attr_name) |
422 | || is_attribute_p (attr_name: "fallthrough" , ident: attr_name) |
423 | || is_attribute_p (attr_name: "maybe_unused" , ident: attr_name) |
424 | || is_attribute_p (attr_name: "nodiscard" , ident: attr_name) |
425 | || is_attribute_p (attr_name: "noreturn" , ident: attr_name) |
426 | || is_attribute_p (attr_name: "_Noreturn" , ident: attr_name)) |
427 | result = 202311; |
428 | } |
429 | if (result) |
430 | attr_name = NULL_TREE; |
431 | } |
432 | if (attr_name && (have_scope || !std_syntax)) |
433 | { |
434 | init_attributes (); |
435 | const struct attribute_spec *attr = lookup_attribute_spec (attr_name); |
436 | if (attr) |
437 | result = 1; |
438 | } |
439 | } |
440 | else |
441 | { |
442 | cpp_error (pfile, CPP_DL_ERROR, |
443 | msgid: "macro \"__has_attribute\" requires an identifier" ); |
444 | return 0; |
445 | } |
446 | |
447 | if (get_token_no_padding (pfile)->type != CPP_CLOSE_PAREN) |
448 | cpp_error (pfile, CPP_DL_ERROR, |
449 | msgid: "missing ')' after \"__has_attribute\"" ); |
450 | |
451 | return result; |
452 | } |
453 | |
454 | /* Callback for has_builtin. */ |
455 | |
456 | int |
457 | c_common_has_builtin (cpp_reader *pfile) |
458 | { |
459 | const cpp_token *token = get_token_no_padding (pfile); |
460 | if (token->type != CPP_OPEN_PAREN) |
461 | { |
462 | cpp_error (pfile, CPP_DL_ERROR, |
463 | msgid: "missing '(' after \"__has_builtin\"" ); |
464 | return 0; |
465 | } |
466 | |
467 | const char *name = "" ; |
468 | token = get_token_no_padding (pfile); |
469 | if (token->type == CPP_NAME) |
470 | { |
471 | name = (const char *) cpp_token_as_text (pfile, token); |
472 | token = get_token_no_padding (pfile); |
473 | if (token->type != CPP_CLOSE_PAREN) |
474 | { |
475 | cpp_error (pfile, CPP_DL_ERROR, |
476 | msgid: "expected ')' after \"%s\"" , name); |
477 | name = "" ; |
478 | } |
479 | } |
480 | else |
481 | { |
482 | cpp_error (pfile, CPP_DL_ERROR, |
483 | msgid: "macro \"__has_builtin\" requires an identifier" ); |
484 | if (token->type == CPP_CLOSE_PAREN) |
485 | return 0; |
486 | } |
487 | |
488 | /* Consume tokens up to the closing parenthesis, including any nested |
489 | pairs of parentheses, to avoid confusing redundant errors. */ |
490 | for (unsigned nparen = 1; ; token = get_token_no_padding (pfile)) |
491 | { |
492 | if (token->type == CPP_OPEN_PAREN) |
493 | ++nparen; |
494 | else if (token->type == CPP_CLOSE_PAREN) |
495 | --nparen; |
496 | else if (token->type == CPP_EOF) |
497 | break; |
498 | if (!nparen) |
499 | break; |
500 | } |
501 | |
502 | return names_builtin_p (name); |
503 | } |
504 | |
505 | |
506 | /* Read a token and return its type. Fill *VALUE with its value, if |
507 | applicable. Fill *CPP_FLAGS with the token's flags, if it is |
508 | non-NULL. */ |
509 | |
510 | enum cpp_ttype |
511 | c_lex_with_flags (tree *value, location_t *loc, unsigned char *cpp_flags, |
512 | int lex_flags) |
513 | { |
514 | const cpp_token *tok; |
515 | enum cpp_ttype type; |
516 | unsigned char add_flags = 0; |
517 | enum overflow_type overflow = OT_NONE; |
518 | |
519 | timevar_push (tv: TV_CPP); |
520 | retry: |
521 | tok = get_token (pfile: parse_in, loc); |
522 | type = tok->type; |
523 | |
524 | retry_after_at: |
525 | switch (type) |
526 | { |
527 | case CPP_PADDING: |
528 | goto retry; |
529 | |
530 | case CPP_NAME: |
531 | *value = HT_IDENT_TO_GCC_IDENT (HT_NODE (tok->val.node.node)); |
532 | break; |
533 | |
534 | case CPP_NUMBER: |
535 | { |
536 | const char *suffix = NULL; |
537 | unsigned int flags = cpp_classify_number (parse_in, tok, &suffix, *loc); |
538 | |
539 | switch (flags & CPP_N_CATEGORY) |
540 | { |
541 | case CPP_N_INVALID: |
542 | /* cpplib has issued an error. */ |
543 | *value = error_mark_node; |
544 | break; |
545 | |
546 | case CPP_N_INTEGER: |
547 | /* C++ uses '0' to mark virtual functions as pure. |
548 | Set PURE_ZERO to pass this information to the C++ parser. */ |
549 | if (tok->val.str.len == 1 && *tok->val.str.text == '0') |
550 | add_flags = PURE_ZERO | DECIMAL_INT; |
551 | else if ((flags & CPP_N_INTEGER) && (flags & CPP_N_DECIMAL)) |
552 | /* -Wxor-used-as-pow is only active for LHS of ^ expressed |
553 | as a decimal integer. */ |
554 | add_flags = DECIMAL_INT; |
555 | *value = interpret_integer (tok, flags, &overflow); |
556 | break; |
557 | |
558 | case CPP_N_FLOATING: |
559 | *value = interpret_float (tok, flags, suffix, &overflow); |
560 | break; |
561 | |
562 | default: |
563 | gcc_unreachable (); |
564 | } |
565 | |
566 | if (flags & CPP_N_USERDEF) |
567 | { |
568 | char *str; |
569 | tree literal; |
570 | tree suffix_id = get_identifier (suffix); |
571 | int len = tok->val.str.len - strlen (s: suffix); |
572 | /* If this is going to be used as a C string to pass to a |
573 | raw literal operator, we need to add a trailing NUL. */ |
574 | tree num_string = build_string (len + 1, |
575 | (const char *) tok->val.str.text); |
576 | TREE_TYPE (num_string) = char_array_type_node; |
577 | num_string = fix_string_type (num_string); |
578 | str = CONST_CAST (char *, TREE_STRING_POINTER (num_string)); |
579 | str[len] = '\0'; |
580 | literal = build_userdef_literal (suffix_id, value: *value, overflow, |
581 | num_string); |
582 | *value = literal; |
583 | } |
584 | } |
585 | break; |
586 | |
587 | case CPP_ATSIGN: |
588 | /* An @ may give the next token special significance in Objective-C. */ |
589 | if (c_dialect_objc ()) |
590 | { |
591 | location_t atloc = *loc; |
592 | location_t newloc; |
593 | |
594 | retry_at: |
595 | tok = get_token (pfile: parse_in, loc: &newloc); |
596 | type = tok->type; |
597 | switch (type) |
598 | { |
599 | case CPP_PADDING: |
600 | goto retry_at; |
601 | |
602 | case CPP_STRING: |
603 | case CPP_WSTRING: |
604 | case CPP_STRING16: |
605 | case CPP_STRING32: |
606 | case CPP_UTF8STRING: |
607 | type = lex_string (tok, value, true, true); |
608 | break; |
609 | |
610 | case CPP_NAME: |
611 | *value = HT_IDENT_TO_GCC_IDENT (HT_NODE (tok->val.node.node)); |
612 | if (OBJC_IS_AT_KEYWORD (C_RID_CODE (*value)) |
613 | || OBJC_IS_CXX_KEYWORD (C_RID_CODE (*value))) |
614 | { |
615 | type = CPP_AT_NAME; |
616 | /* Note the complication: if we found an OBJC_CXX |
617 | keyword, for example, 'class', we will be |
618 | returning a token of type CPP_AT_NAME and rid |
619 | code RID_CLASS (not RID_AT_CLASS). The language |
620 | parser needs to convert that to RID_AT_CLASS. |
621 | However, we've now spliced the '@' together with the |
622 | keyword that follows; Adjust the location so that we |
623 | get a source range covering the composite. |
624 | */ |
625 | *loc = make_location (caret: atloc, start: atloc, finish: newloc); |
626 | break; |
627 | } |
628 | /* FALLTHROUGH */ |
629 | |
630 | default: |
631 | /* ... or not. */ |
632 | error_at (atloc, "stray %<@%> in program" ); |
633 | *loc = newloc; |
634 | goto retry_after_at; |
635 | } |
636 | break; |
637 | } |
638 | |
639 | /* FALLTHROUGH */ |
640 | case CPP_HASH: |
641 | case CPP_PASTE: |
642 | { |
643 | unsigned char name[8]; |
644 | |
645 | *cpp_spell_token (parse_in, tok, name, true) = 0; |
646 | |
647 | error_at (*loc, "stray %qs in program" , name); |
648 | } |
649 | |
650 | goto retry; |
651 | |
652 | case CPP_OTHER: |
653 | { |
654 | cppchar_t c = tok->val.str.text[0]; |
655 | |
656 | if (c == '"' || c == '\'') |
657 | error_at (*loc, "missing terminating %c character" , (int) c); |
658 | else if (ISGRAPH (c)) |
659 | error_at (*loc, "stray %qc in program" , (int) c); |
660 | else |
661 | { |
662 | rich_location rich_loc (line_table, *loc); |
663 | rich_loc.set_escape_on_output (true); |
664 | error_at (&rich_loc, "stray %<\\%o%> in program" , (int) c); |
665 | } |
666 | } |
667 | goto retry; |
668 | |
669 | case CPP_CHAR_USERDEF: |
670 | case CPP_WCHAR_USERDEF: |
671 | case CPP_CHAR16_USERDEF: |
672 | case CPP_CHAR32_USERDEF: |
673 | case CPP_UTF8CHAR_USERDEF: |
674 | { |
675 | tree literal; |
676 | cpp_token temp_tok = *tok; |
677 | const char *suffix = cpp_get_userdef_suffix (tok); |
678 | temp_tok.val.str.len -= strlen (s: suffix); |
679 | temp_tok.type = cpp_userdef_char_remove_type (type); |
680 | literal = build_userdef_literal (get_identifier (suffix), |
681 | value: lex_charconst (&temp_tok), |
682 | overflow: OT_NONE, NULL_TREE); |
683 | *value = literal; |
684 | } |
685 | break; |
686 | |
687 | case CPP_CHAR: |
688 | case CPP_WCHAR: |
689 | case CPP_CHAR16: |
690 | case CPP_CHAR32: |
691 | case CPP_UTF8CHAR: |
692 | *value = lex_charconst (tok); |
693 | break; |
694 | |
695 | case CPP_STRING_USERDEF: |
696 | case CPP_WSTRING_USERDEF: |
697 | case CPP_STRING16_USERDEF: |
698 | case CPP_STRING32_USERDEF: |
699 | case CPP_UTF8STRING_USERDEF: |
700 | { |
701 | tree literal, string; |
702 | const char *suffix = cpp_get_userdef_suffix (tok); |
703 | string = build_string (tok->val.str.len - strlen (s: suffix), |
704 | (const char *) tok->val.str.text); |
705 | literal = build_userdef_literal (get_identifier (suffix), |
706 | value: string, overflow: OT_NONE, NULL_TREE); |
707 | *value = literal; |
708 | } |
709 | break; |
710 | |
711 | case CPP_STRING: |
712 | case CPP_WSTRING: |
713 | case CPP_STRING16: |
714 | case CPP_STRING32: |
715 | case CPP_UTF8STRING: |
716 | if ((lex_flags & C_LEX_STRING_NO_JOIN) == 0) |
717 | { |
718 | type = lex_string (tok, value, false, |
719 | (lex_flags & C_LEX_STRING_NO_TRANSLATE) == 0); |
720 | break; |
721 | } |
722 | *value = build_string (tok->val.str.len, (const char *) tok->val.str.text); |
723 | break; |
724 | |
725 | case CPP_PRAGMA: |
726 | *value = build_int_cst (integer_type_node, tok->val.pragma); |
727 | break; |
728 | |
729 | case CPP_HEADER_NAME: |
730 | *value = build_string (tok->val.str.len, (const char *)tok->val.str.text); |
731 | break; |
732 | |
733 | /* This token should not be visible outside cpplib. */ |
734 | case CPP_MACRO_ARG: |
735 | gcc_unreachable (); |
736 | |
737 | /* CPP_COMMENT will appear when compiling with -C. Ignore, except |
738 | when it is a FALLTHROUGH comment, in that case set |
739 | PREV_FALLTHROUGH flag on the next non-comment token. */ |
740 | case CPP_COMMENT: |
741 | if (tok->flags & PREV_FALLTHROUGH) |
742 | { |
743 | do |
744 | { |
745 | tok = get_token (pfile: parse_in, loc); |
746 | type = tok->type; |
747 | } |
748 | while (type == CPP_PADDING || type == CPP_COMMENT); |
749 | add_flags |= PREV_FALLTHROUGH; |
750 | goto retry_after_at; |
751 | } |
752 | goto retry; |
753 | |
754 | default: |
755 | *value = NULL_TREE; |
756 | break; |
757 | } |
758 | |
759 | if (cpp_flags) |
760 | *cpp_flags = tok->flags | add_flags; |
761 | |
762 | timevar_pop (tv: TV_CPP); |
763 | |
764 | return type; |
765 | } |
766 | |
767 | /* Returns the narrowest C-visible unsigned type, starting with the |
768 | minimum specified by FLAGS, that can fit HIGH:LOW, or itk_none if |
769 | there isn't one. */ |
770 | |
771 | static enum integer_type_kind |
772 | narrowest_unsigned_type (const widest_int &val, unsigned int flags) |
773 | { |
774 | int itk; |
775 | |
776 | if ((flags & CPP_N_WIDTH) == CPP_N_SMALL) |
777 | itk = itk_unsigned_int; |
778 | else if ((flags & CPP_N_WIDTH) == CPP_N_MEDIUM) |
779 | itk = itk_unsigned_long; |
780 | else |
781 | itk = itk_unsigned_long_long; |
782 | |
783 | for (; itk < itk_none; itk += 2 /* skip unsigned types */) |
784 | { |
785 | tree upper; |
786 | |
787 | if (integer_types[itk] == NULL_TREE) |
788 | continue; |
789 | upper = TYPE_MAX_VALUE (integer_types[itk]); |
790 | |
791 | if (wi::geu_p (x: wi::to_widest (t: upper), y: val)) |
792 | return (enum integer_type_kind) itk; |
793 | } |
794 | |
795 | return itk_none; |
796 | } |
797 | |
798 | /* Ditto, but narrowest signed type. */ |
799 | static enum integer_type_kind |
800 | narrowest_signed_type (const widest_int &val, unsigned int flags) |
801 | { |
802 | int itk; |
803 | |
804 | if ((flags & CPP_N_WIDTH) == CPP_N_SMALL) |
805 | itk = itk_int; |
806 | else if ((flags & CPP_N_WIDTH) == CPP_N_MEDIUM) |
807 | itk = itk_long; |
808 | else |
809 | itk = itk_long_long; |
810 | |
811 | for (; itk < itk_none; itk += 2 /* skip signed types */) |
812 | { |
813 | tree upper; |
814 | |
815 | if (integer_types[itk] == NULL_TREE) |
816 | continue; |
817 | upper = TYPE_MAX_VALUE (integer_types[itk]); |
818 | |
819 | if (wi::geu_p (x: wi::to_widest (t: upper), y: val)) |
820 | return (enum integer_type_kind) itk; |
821 | } |
822 | |
823 | return itk_none; |
824 | } |
825 | |
826 | /* Interpret TOKEN, an integer with FLAGS as classified by cpplib. */ |
827 | static tree |
828 | interpret_integer (const cpp_token *token, unsigned int flags, |
829 | enum overflow_type *overflow) |
830 | { |
831 | tree value, type; |
832 | enum integer_type_kind itk; |
833 | cpp_num integer; |
834 | HOST_WIDE_INT ival[3]; |
835 | |
836 | *overflow = OT_NONE; |
837 | |
838 | if (UNLIKELY (flags & CPP_N_BITINT)) |
839 | { |
840 | unsigned int suffix_len = 2 + ((flags & CPP_N_UNSIGNED) ? 1 : 0); |
841 | int max_bits_per_digit = 4; // ceil (log2 (10)) |
842 | unsigned int prefix_len = 0; |
843 | bool hex = false; |
844 | const int bitint_maxwidth = WIDE_INT_MAX_PRECISION - 1; |
845 | if ((flags & CPP_N_RADIX) == CPP_N_OCTAL) |
846 | { |
847 | max_bits_per_digit = 3; |
848 | prefix_len = 1; |
849 | } |
850 | else if ((flags & CPP_N_RADIX) == CPP_N_HEX) |
851 | { |
852 | max_bits_per_digit = 4; |
853 | prefix_len = 2; |
854 | hex = true; |
855 | } |
856 | else if ((flags & CPP_N_RADIX) == CPP_N_BINARY) |
857 | { |
858 | max_bits_per_digit = 1; |
859 | prefix_len = 2; |
860 | } |
861 | int max_digits |
862 | = TYPE_PRECISION (intmax_type_node) >> max_bits_per_digit; |
863 | const int max_buf = 128; |
864 | if (max_digits > max_buf) |
865 | max_digits = max_buf; |
866 | |
867 | widest_int wval; |
868 | unsigned int prec; |
869 | gcc_checking_assert (token->val.str.len > prefix_len + suffix_len |
870 | || token->val.str.len == 1 + suffix_len); |
871 | if (token->val.str.len - (prefix_len + suffix_len) |
872 | <= (unsigned) max_digits) |
873 | { |
874 | integer = cpp_interpret_integer (parse_in, token, |
875 | (flags & CPP_N_RADIX) |
876 | | CPP_N_UNSIGNED); |
877 | ival[0] = integer.low; |
878 | ival[1] = integer.high; |
879 | ival[2] = 0; |
880 | wval = widest_int::from_array (val: ival, len: 3); |
881 | } |
882 | else |
883 | { |
884 | unsigned char buf[3 + max_buf]; |
885 | memcpy (dest: buf, src: token->val.str.text, n: prefix_len); |
886 | wval = 0U; |
887 | const unsigned char *p = token->val.str.text + prefix_len; |
888 | cpp_token tok = *token; |
889 | tok.val.str.text = buf; |
890 | if (!prefix_len) |
891 | max_digits = 19; |
892 | do |
893 | { |
894 | unsigned char *q = buf + prefix_len; |
895 | do |
896 | { |
897 | unsigned char c = *p++; |
898 | if (ISDIGIT (c) || (hex && ISXDIGIT (c))) |
899 | { |
900 | *q++ = c; |
901 | if (q == buf + prefix_len + max_digits) |
902 | break; |
903 | } |
904 | else if (c != '\'') |
905 | { |
906 | --p; |
907 | break; |
908 | } |
909 | } |
910 | while (1); |
911 | if (q == buf + prefix_len) |
912 | break; |
913 | else |
914 | { |
915 | wi::overflow_type wioverflow; |
916 | *q = '\0'; |
917 | tok.val.str.len = q - buf; |
918 | if (wval == 0) |
919 | ; |
920 | else if (prefix_len) |
921 | { |
922 | prec = wi::min_precision (x: wval, sgn: UNSIGNED); |
923 | unsigned HOST_WIDE_INT shift |
924 | = (tok.val.str.len - prefix_len) * max_bits_per_digit; |
925 | if (prec + shift > bitint_maxwidth) |
926 | goto bitint_overflow; |
927 | wval = wi::lshift (x: wval, y: shift); |
928 | } |
929 | else |
930 | { |
931 | static unsigned HOST_WIDE_INT tens[] |
932 | = { 1U, 10U, 100U, 1000U, |
933 | HOST_WIDE_INT_UC (10000), |
934 | HOST_WIDE_INT_UC (100000), |
935 | HOST_WIDE_INT_UC (1000000), |
936 | HOST_WIDE_INT_UC (10000000), |
937 | HOST_WIDE_INT_UC (100000000), |
938 | HOST_WIDE_INT_UC (1000000000), |
939 | HOST_WIDE_INT_UC (10000000000), |
940 | HOST_WIDE_INT_UC (100000000000), |
941 | HOST_WIDE_INT_UC (1000000000000), |
942 | HOST_WIDE_INT_UC (10000000000000), |
943 | HOST_WIDE_INT_UC (100000000000000), |
944 | HOST_WIDE_INT_UC (1000000000000000), |
945 | HOST_WIDE_INT_UC (10000000000000000), |
946 | HOST_WIDE_INT_UC (100000000000000000), |
947 | HOST_WIDE_INT_UC (1000000000000000000), |
948 | HOST_WIDE_INT_UC (10000000000000000000) }; |
949 | widest_int ten = tens[q - buf]; |
950 | wval = wi::umul (x: wval, y: ten, overflow: &wioverflow); |
951 | if (wioverflow) |
952 | goto bitint_overflow; |
953 | } |
954 | integer = cpp_interpret_integer (parse_in, &tok, |
955 | (flags & CPP_N_RADIX) |
956 | | CPP_N_UNSIGNED); |
957 | ival[0] = integer.low; |
958 | ival[1] = integer.high; |
959 | ival[2] = 0; |
960 | if (prefix_len) |
961 | wval = wval + widest_int::from_array (val: ival, len: 3); |
962 | else |
963 | { |
964 | widest_int addend = widest_int::from_array (val: ival, len: 3); |
965 | wval = wi::add (x: wval, y: addend, sgn: UNSIGNED, overflow: &wioverflow); |
966 | if (wioverflow) |
967 | goto bitint_overflow; |
968 | } |
969 | } |
970 | } |
971 | while (1); |
972 | } |
973 | |
974 | prec = wi::min_precision (x: wval, sgn: UNSIGNED); |
975 | if (prec == 0) |
976 | prec = 1; |
977 | if ((flags & CPP_N_UNSIGNED) == 0) |
978 | ++prec; |
979 | if (prec > bitint_maxwidth) |
980 | { |
981 | bitint_overflow: |
982 | if ((flags & CPP_N_UNSIGNED) != 0) |
983 | error ("integer constant is too large for " |
984 | "%<unsigned _BitInt(%d)%> type" , bitint_maxwidth); |
985 | else |
986 | error ("integer constant is too large for " |
987 | "%<_BitInt(%d)%> type" , bitint_maxwidth); |
988 | return integer_zero_node; |
989 | } |
990 | |
991 | struct bitint_info info; |
992 | if (!targetm.c.bitint_type_info (prec, &info)) |
993 | { |
994 | sorry ("%<_BitInt(%d)%> is not supported on this target" , prec); |
995 | return integer_zero_node; |
996 | } |
997 | |
998 | type = build_bitint_type (prec, (flags & CPP_N_UNSIGNED) != 0); |
999 | return wide_int_to_tree (type, cst: wval); |
1000 | } |
1001 | |
1002 | integer = cpp_interpret_integer (parse_in, token, flags); |
1003 | if (integer.overflow) |
1004 | *overflow = OT_OVERFLOW; |
1005 | |
1006 | ival[0] = integer.low; |
1007 | ival[1] = integer.high; |
1008 | ival[2] = 0; |
1009 | widest_int wval = widest_int::from_array (val: ival, len: 3); |
1010 | |
1011 | /* The type of a constant with a U suffix is straightforward. */ |
1012 | if (flags & CPP_N_UNSIGNED) |
1013 | itk = narrowest_unsigned_type (val: wval, flags); |
1014 | else |
1015 | { |
1016 | /* The type of a potentially-signed integer constant varies |
1017 | depending on the base it's in, the standard in use, and the |
1018 | length suffixes. */ |
1019 | enum integer_type_kind itk_u |
1020 | = narrowest_unsigned_type (val: wval, flags); |
1021 | enum integer_type_kind itk_s |
1022 | = narrowest_signed_type (val: wval, flags); |
1023 | |
1024 | /* In both C89 and C99, octal and hex constants may be signed or |
1025 | unsigned, whichever fits tighter. We do not warn about this |
1026 | choice differing from the traditional choice, as the constant |
1027 | is probably a bit pattern and either way will work. */ |
1028 | if ((flags & CPP_N_RADIX) != CPP_N_DECIMAL) |
1029 | itk = MIN (itk_u, itk_s); |
1030 | else |
1031 | { |
1032 | /* In C99, decimal constants are always signed. |
1033 | In C89, decimal constants that don't fit in long have |
1034 | undefined behavior; we try to make them unsigned long. |
1035 | In GCC's extended C89, that last is true of decimal |
1036 | constants that don't fit in long long, too. */ |
1037 | |
1038 | itk = itk_s; |
1039 | if (itk_s > itk_u && itk_s > itk_long) |
1040 | { |
1041 | if (!flag_isoc99) |
1042 | { |
1043 | if (itk_u < itk_unsigned_long) |
1044 | itk_u = itk_unsigned_long; |
1045 | itk = itk_u; |
1046 | warning (0, "this decimal constant is unsigned only in ISO C90" ); |
1047 | } |
1048 | else |
1049 | warning (OPT_Wtraditional, |
1050 | "this decimal constant would be unsigned in ISO C90" ); |
1051 | } |
1052 | } |
1053 | } |
1054 | |
1055 | if (itk == itk_none) |
1056 | /* cpplib has already issued a warning for overflow. */ |
1057 | type = ((flags & CPP_N_UNSIGNED) |
1058 | ? widest_unsigned_literal_type_node |
1059 | : widest_integer_literal_type_node); |
1060 | else if (flags & CPP_N_SIZE_T) |
1061 | { |
1062 | /* itk refers to fundamental types not aliased size types. */ |
1063 | if (flags & CPP_N_UNSIGNED) |
1064 | type = size_type_node; |
1065 | else |
1066 | type = signed_size_type_node; |
1067 | } |
1068 | else |
1069 | { |
1070 | type = integer_types[itk]; |
1071 | if (itk > itk_unsigned_long |
1072 | && (flags & CPP_N_WIDTH) != CPP_N_LARGE) |
1073 | emit_diagnostic |
1074 | ((c_dialect_cxx () ? cxx_dialect == cxx98 : !flag_isoc99) |
1075 | ? DK_PEDWARN : DK_WARNING, |
1076 | input_location, OPT_Wlong_long, |
1077 | (flags & CPP_N_UNSIGNED) |
1078 | ? "integer constant is too large for %<unsigned long%> type" |
1079 | : "integer constant is too large for %<long%> type" ); |
1080 | } |
1081 | |
1082 | value = wide_int_to_tree (type, cst: wval); |
1083 | |
1084 | /* Convert imaginary to a complex type. */ |
1085 | if (flags & CPP_N_IMAGINARY) |
1086 | value = build_complex (NULL_TREE, build_int_cst (type, 0), value); |
1087 | |
1088 | return value; |
1089 | } |
1090 | |
1091 | /* Interpret TOKEN, a floating point number with FLAGS as classified |
1092 | by cpplib. For C++11 SUFFIX may contain a user-defined literal suffix. */ |
1093 | static tree |
1094 | interpret_float (const cpp_token *token, unsigned int flags, |
1095 | const char *suffix, enum overflow_type *overflow) |
1096 | { |
1097 | tree type; |
1098 | tree const_type; |
1099 | tree value; |
1100 | REAL_VALUE_TYPE real; |
1101 | REAL_VALUE_TYPE real_trunc; |
1102 | char *copy; |
1103 | size_t copylen; |
1104 | |
1105 | *overflow = OT_NONE; |
1106 | |
1107 | /* Default (no suffix) depends on whether the FLOAT_CONST_DECIMAL64 |
1108 | pragma has been used and is either double or _Decimal64. Types |
1109 | that are not allowed with decimal float default to double. */ |
1110 | if (flags & CPP_N_DEFAULT) |
1111 | { |
1112 | flags ^= CPP_N_DEFAULT; |
1113 | flags |= CPP_N_MEDIUM; |
1114 | |
1115 | if (((flags & CPP_N_HEX) == 0) && ((flags & CPP_N_IMAGINARY) == 0)) |
1116 | { |
1117 | warning (OPT_Wunsuffixed_float_constants, |
1118 | "unsuffixed floating constant" ); |
1119 | if (float_const_decimal64_p ()) |
1120 | flags |= CPP_N_DFLOAT; |
1121 | } |
1122 | } |
1123 | |
1124 | /* Decode _Fract and _Accum. */ |
1125 | if (flags & CPP_N_FRACT || flags & CPP_N_ACCUM) |
1126 | return interpret_fixed (token, flags); |
1127 | |
1128 | /* Decode type based on width and properties. */ |
1129 | if (flags & CPP_N_DFLOAT) |
1130 | if (!targetm.decimal_float_supported_p ()) |
1131 | { |
1132 | error ("decimal floating-point not supported for this target" ); |
1133 | return error_mark_node; |
1134 | } |
1135 | else if ((flags & CPP_N_WIDTH) == CPP_N_LARGE) |
1136 | type = dfloat128_type_node; |
1137 | else if ((flags & CPP_N_WIDTH) == CPP_N_SMALL) |
1138 | type = dfloat32_type_node; |
1139 | else |
1140 | type = dfloat64_type_node; |
1141 | else |
1142 | if (flags & CPP_N_WIDTH_MD) |
1143 | { |
1144 | char suffix; |
1145 | machine_mode mode; |
1146 | |
1147 | if ((flags & CPP_N_WIDTH_MD) == CPP_N_MD_W) |
1148 | suffix = 'w'; |
1149 | else |
1150 | suffix = 'q'; |
1151 | |
1152 | mode = targetm.c.mode_for_suffix (suffix); |
1153 | if (mode == VOIDmode) |
1154 | { |
1155 | error ("unsupported non-standard suffix on floating constant" ); |
1156 | |
1157 | return error_mark_node; |
1158 | } |
1159 | else |
1160 | pedwarn (input_location, OPT_Wpedantic, "non-standard suffix on floating constant" ); |
1161 | |
1162 | type = c_common_type_for_mode (mode, 0); |
1163 | /* For Q suffix, prefer float128t_type_node (__float128) type |
1164 | over float128_type_node (_Float128) type if they are distinct. */ |
1165 | if (type == float128_type_node && float128t_type_node) |
1166 | type = float128t_type_node; |
1167 | gcc_assert (type); |
1168 | } |
1169 | else if ((flags & (CPP_N_FLOATN | CPP_N_FLOATNX)) != 0) |
1170 | { |
1171 | unsigned int n = (flags & CPP_N_WIDTH_FLOATN_NX) >> CPP_FLOATN_SHIFT; |
1172 | bool extended = (flags & CPP_N_FLOATNX) != 0; |
1173 | type = NULL_TREE; |
1174 | for (int i = 0; i < NUM_FLOATN_NX_TYPES; i++) |
1175 | if (floatn_nx_types[i].n == (int) n |
1176 | && floatn_nx_types[i].extended == extended) |
1177 | { |
1178 | type = FLOATN_NX_TYPE_NODE (i); |
1179 | break; |
1180 | } |
1181 | if (type == NULL_TREE) |
1182 | { |
1183 | error ("unsupported non-standard suffix on floating constant" ); |
1184 | return error_mark_node; |
1185 | } |
1186 | else if (!c_dialect_cxx ()) |
1187 | { |
1188 | if (warn_c11_c23_compat > 0) |
1189 | { |
1190 | if (pedantic && !flag_isoc23) |
1191 | pedwarn (input_location, OPT_Wc11_c23_compat, |
1192 | "non-standard suffix on floating constant " |
1193 | "before C23" ); |
1194 | else |
1195 | warning (OPT_Wc11_c23_compat, |
1196 | "non-standard suffix on floating constant " |
1197 | "before C23" ); |
1198 | } |
1199 | else if (warn_c11_c23_compat != 0 && pedantic && !flag_isoc23) |
1200 | pedwarn (input_location, OPT_Wpedantic, |
1201 | "non-standard suffix on floating constant " |
1202 | "before C23" ); |
1203 | } |
1204 | else if (!extended) |
1205 | { |
1206 | if (cxx_dialect < cxx23) |
1207 | pedwarn (input_location, OPT_Wpedantic, |
1208 | "%<f%d%> or %<F%d%> suffix on floating constant only " |
1209 | "available with %<-std=c++2b%> or %<-std=gnu++2b%>" , |
1210 | n, n); |
1211 | } |
1212 | else |
1213 | pedwarn (input_location, OPT_Wpedantic, |
1214 | "non-standard suffix on floating constant" ); |
1215 | } |
1216 | else if ((flags & CPP_N_BFLOAT16) != 0) |
1217 | { |
1218 | type = bfloat16_type_node; |
1219 | if (type == NULL_TREE) |
1220 | { |
1221 | error ("unsupported non-standard suffix on floating constant" ); |
1222 | return error_mark_node; |
1223 | } |
1224 | if (!c_dialect_cxx ()) |
1225 | pedwarn (input_location, OPT_Wpedantic, |
1226 | "non-standard suffix on floating constant" ); |
1227 | else if (cxx_dialect < cxx23) |
1228 | pedwarn (input_location, OPT_Wpedantic, |
1229 | "%<bf16%> or %<BF16%> suffix on floating constant only " |
1230 | "available with %<-std=c++2b%> or %<-std=gnu++2b%>" ); |
1231 | } |
1232 | else if ((flags & CPP_N_WIDTH) == CPP_N_LARGE) |
1233 | type = long_double_type_node; |
1234 | else if ((flags & CPP_N_WIDTH) == CPP_N_SMALL |
1235 | || flag_single_precision_constant) |
1236 | type = float_type_node; |
1237 | else |
1238 | type = double_type_node; |
1239 | |
1240 | const_type = excess_precision_type (type); |
1241 | if (!const_type) |
1242 | const_type = type; |
1243 | |
1244 | /* Copy the constant to a nul-terminated buffer. If the constant |
1245 | has any suffixes, cut them off; REAL_VALUE_ATOF/ REAL_VALUE_HTOF |
1246 | can't handle them. */ |
1247 | copylen = token->val.str.len; |
1248 | if (flags & CPP_N_USERDEF) |
1249 | copylen -= strlen (s: suffix); |
1250 | else if (flags & CPP_N_DFLOAT) |
1251 | copylen -= 2; |
1252 | else |
1253 | { |
1254 | if ((flags & CPP_N_WIDTH) != CPP_N_MEDIUM) |
1255 | /* Must be an F or L or machine defined suffix. */ |
1256 | copylen--; |
1257 | if (flags & CPP_N_IMAGINARY) |
1258 | /* I or J suffix. */ |
1259 | copylen--; |
1260 | if (flags & CPP_N_FLOATNX) |
1261 | copylen--; |
1262 | if (flags & (CPP_N_FLOATN | CPP_N_FLOATNX)) |
1263 | { |
1264 | unsigned int n = (flags & CPP_N_WIDTH_FLOATN_NX) >> CPP_FLOATN_SHIFT; |
1265 | while (n > 0) |
1266 | { |
1267 | copylen--; |
1268 | n /= 10; |
1269 | } |
1270 | } |
1271 | } |
1272 | |
1273 | copy = (char *) alloca (copylen + 1); |
1274 | if (c_dialect_cxx () ? cxx_dialect > cxx11 : flag_isoc23) |
1275 | { |
1276 | size_t maxlen = 0; |
1277 | for (size_t i = 0; i < copylen; ++i) |
1278 | if (token->val.str.text[i] != '\'') |
1279 | copy[maxlen++] = token->val.str.text[i]; |
1280 | copy[maxlen] = '\0'; |
1281 | } |
1282 | else |
1283 | { |
1284 | memcpy (dest: copy, src: token->val.str.text, n: copylen); |
1285 | copy[copylen] = '\0'; |
1286 | } |
1287 | |
1288 | real_from_string3 (&real, copy, TYPE_MODE (const_type)); |
1289 | if (const_type != type) |
1290 | /* Diagnosing if the result of converting the value with excess |
1291 | precision to the semantic type would overflow (with associated |
1292 | double rounding) is more appropriate than diagnosing if the |
1293 | result of converting the string directly to the semantic type |
1294 | would overflow. */ |
1295 | real_convert (&real_trunc, TYPE_MODE (type), &real); |
1296 | |
1297 | /* Both C and C++ require a diagnostic for a floating constant |
1298 | outside the range of representable values of its type. Since we |
1299 | have __builtin_inf* to produce an infinity, this is now a |
1300 | mandatory pedwarn if the target does not support infinities. */ |
1301 | if (REAL_VALUE_ISINF (real) |
1302 | || (const_type != type && REAL_VALUE_ISINF (real_trunc))) |
1303 | { |
1304 | *overflow = OT_OVERFLOW; |
1305 | if (!(flags & CPP_N_USERDEF)) |
1306 | { |
1307 | if (!MODE_HAS_INFINITIES (TYPE_MODE (type))) |
1308 | pedwarn (input_location, 0, |
1309 | "floating constant exceeds range of %qT" , type); |
1310 | else |
1311 | warning (OPT_Woverflow, |
1312 | "floating constant exceeds range of %qT" , type); |
1313 | } |
1314 | } |
1315 | /* We also give a warning if the value underflows. */ |
1316 | else if (real_equal (&real, &dconst0) |
1317 | || (const_type != type |
1318 | && real_equal (&real_trunc, &dconst0))) |
1319 | { |
1320 | REAL_VALUE_TYPE realvoidmode; |
1321 | int oflow = real_from_string (&realvoidmode, copy); |
1322 | *overflow = (oflow == 0 ? OT_NONE |
1323 | : (oflow < 0 ? OT_UNDERFLOW : OT_OVERFLOW)); |
1324 | if (!(flags & CPP_N_USERDEF)) |
1325 | { |
1326 | if (oflow < 0 || !real_equal (&realvoidmode, &dconst0)) |
1327 | warning (OPT_Woverflow, "floating constant truncated to zero" ); |
1328 | } |
1329 | } |
1330 | |
1331 | /* Create a node with determined type and value. */ |
1332 | value = build_real (const_type, real); |
1333 | if (flags & CPP_N_IMAGINARY) |
1334 | { |
1335 | value = build_complex (NULL_TREE, |
1336 | fold_convert (const_type, |
1337 | integer_zero_node), value); |
1338 | if (type != const_type) |
1339 | { |
1340 | const_type = TREE_TYPE (value); |
1341 | type = build_complex_type (type); |
1342 | } |
1343 | } |
1344 | |
1345 | if (type != const_type) |
1346 | value = build1_loc (loc: token->src_loc, code: EXCESS_PRECISION_EXPR, type, arg1: value); |
1347 | |
1348 | return value; |
1349 | } |
1350 | |
1351 | /* Interpret TOKEN, a fixed-point number with FLAGS as classified |
1352 | by cpplib. */ |
1353 | |
1354 | static tree |
1355 | interpret_fixed (const cpp_token *token, unsigned int flags) |
1356 | { |
1357 | tree type; |
1358 | tree value; |
1359 | FIXED_VALUE_TYPE fixed; |
1360 | char *copy; |
1361 | size_t copylen; |
1362 | |
1363 | copylen = token->val.str.len; |
1364 | |
1365 | if (flags & CPP_N_FRACT) /* _Fract. */ |
1366 | { |
1367 | if (flags & CPP_N_UNSIGNED) /* Unsigned _Fract. */ |
1368 | { |
1369 | if ((flags & CPP_N_WIDTH) == CPP_N_LARGE) |
1370 | { |
1371 | type = unsigned_long_long_fract_type_node; |
1372 | copylen -= 4; |
1373 | } |
1374 | else if ((flags & CPP_N_WIDTH) == CPP_N_MEDIUM) |
1375 | { |
1376 | type = unsigned_long_fract_type_node; |
1377 | copylen -= 3; |
1378 | } |
1379 | else if ((flags & CPP_N_WIDTH) == CPP_N_SMALL) |
1380 | { |
1381 | type = unsigned_short_fract_type_node; |
1382 | copylen -= 3; |
1383 | } |
1384 | else |
1385 | { |
1386 | type = unsigned_fract_type_node; |
1387 | copylen -= 2; |
1388 | } |
1389 | } |
1390 | else /* Signed _Fract. */ |
1391 | { |
1392 | if ((flags & CPP_N_WIDTH) == CPP_N_LARGE) |
1393 | { |
1394 | type = long_long_fract_type_node; |
1395 | copylen -= 3; |
1396 | } |
1397 | else if ((flags & CPP_N_WIDTH) == CPP_N_MEDIUM) |
1398 | { |
1399 | type = long_fract_type_node; |
1400 | copylen -= 2; |
1401 | } |
1402 | else if ((flags & CPP_N_WIDTH) == CPP_N_SMALL) |
1403 | { |
1404 | type = short_fract_type_node; |
1405 | copylen -= 2; |
1406 | } |
1407 | else |
1408 | { |
1409 | type = fract_type_node; |
1410 | copylen --; |
1411 | } |
1412 | } |
1413 | } |
1414 | else /* _Accum. */ |
1415 | { |
1416 | if (flags & CPP_N_UNSIGNED) /* Unsigned _Accum. */ |
1417 | { |
1418 | if ((flags & CPP_N_WIDTH) == CPP_N_LARGE) |
1419 | { |
1420 | type = unsigned_long_long_accum_type_node; |
1421 | copylen -= 4; |
1422 | } |
1423 | else if ((flags & CPP_N_WIDTH) == CPP_N_MEDIUM) |
1424 | { |
1425 | type = unsigned_long_accum_type_node; |
1426 | copylen -= 3; |
1427 | } |
1428 | else if ((flags & CPP_N_WIDTH) == CPP_N_SMALL) |
1429 | { |
1430 | type = unsigned_short_accum_type_node; |
1431 | copylen -= 3; |
1432 | } |
1433 | else |
1434 | { |
1435 | type = unsigned_accum_type_node; |
1436 | copylen -= 2; |
1437 | } |
1438 | } |
1439 | else /* Signed _Accum. */ |
1440 | { |
1441 | if ((flags & CPP_N_WIDTH) == CPP_N_LARGE) |
1442 | { |
1443 | type = long_long_accum_type_node; |
1444 | copylen -= 3; |
1445 | } |
1446 | else if ((flags & CPP_N_WIDTH) == CPP_N_MEDIUM) |
1447 | { |
1448 | type = long_accum_type_node; |
1449 | copylen -= 2; |
1450 | } |
1451 | else if ((flags & CPP_N_WIDTH) == CPP_N_SMALL) |
1452 | { |
1453 | type = short_accum_type_node; |
1454 | copylen -= 2; |
1455 | } |
1456 | else |
1457 | { |
1458 | type = accum_type_node; |
1459 | copylen --; |
1460 | } |
1461 | } |
1462 | } |
1463 | |
1464 | copy = (char *) alloca (copylen + 1); |
1465 | memcpy (dest: copy, src: token->val.str.text, n: copylen); |
1466 | copy[copylen] = '\0'; |
1467 | |
1468 | fixed_from_string (&fixed, copy, SCALAR_TYPE_MODE (type)); |
1469 | |
1470 | /* Create a node with determined type and value. */ |
1471 | value = build_fixed (type, fixed); |
1472 | |
1473 | return value; |
1474 | } |
1475 | |
1476 | /* Convert a series of STRING, WSTRING, STRING16, STRING32 and/or |
1477 | UTF8STRING tokens into a tree, performing string constant |
1478 | concatenation. TOK is the first of these. VALP is the location to |
1479 | write the string into. OBJC_STRING indicates whether an '@' token |
1480 | preceded the incoming token (in that case, the strings can either |
1481 | be ObjC strings, preceded by a single '@', or normal strings, not |
1482 | preceded by '@'. The result will be a CPP_OBJC_STRING). Returns |
1483 | the CPP token type of the result (CPP_STRING, CPP_WSTRING, |
1484 | CPP_STRING32, CPP_STRING16, CPP_UTF8STRING, or CPP_OBJC_STRING). |
1485 | |
1486 | This is unfortunately more work than it should be. If any of the |
1487 | strings in the series has an L prefix, the result is a wide string |
1488 | (6.4.5p4). Whether or not the result is a wide string affects the |
1489 | meaning of octal and hexadecimal escapes (6.4.4.4p6,9). But escape |
1490 | sequences do not continue across the boundary between two strings in |
1491 | a series (6.4.5p7), so we must not lose the boundaries. Therefore |
1492 | cpp_interpret_string takes a vector of cpp_string structures, which |
1493 | we must arrange to provide. */ |
1494 | |
1495 | static enum cpp_ttype |
1496 | lex_string (const cpp_token *tok, tree *valp, bool objc_string, bool translate) |
1497 | { |
1498 | tree value; |
1499 | size_t concats = 0; |
1500 | struct obstack str_ob; |
1501 | struct obstack loc_ob; |
1502 | cpp_string istr; |
1503 | enum cpp_ttype type = tok->type; |
1504 | |
1505 | /* Try to avoid the overhead of creating and destroying an obstack |
1506 | for the common case of just one string. */ |
1507 | cpp_string str = tok->val.str; |
1508 | location_t init_loc = tok->src_loc; |
1509 | cpp_string *strs = &str; |
1510 | location_t *locs = NULL; |
1511 | |
1512 | /* objc_at_sign_was_seen is only used when doing Objective-C string |
1513 | concatenation. It is 'true' if we have seen an '@' before the |
1514 | current string, and 'false' if not. We must see exactly one or |
1515 | zero '@' before each string. */ |
1516 | bool objc_at_sign_was_seen = false; |
1517 | |
1518 | retry: |
1519 | tok = get_token (pfile: parse_in); |
1520 | switch (tok->type) |
1521 | { |
1522 | case CPP_PADDING: |
1523 | goto retry; |
1524 | case CPP_ATSIGN: |
1525 | if (objc_string) |
1526 | { |
1527 | if (objc_at_sign_was_seen) |
1528 | error ("repeated %<@%> before Objective-C string" ); |
1529 | |
1530 | objc_at_sign_was_seen = true; |
1531 | goto retry; |
1532 | } |
1533 | /* FALLTHROUGH */ |
1534 | |
1535 | default: |
1536 | break; |
1537 | |
1538 | case CPP_WSTRING: |
1539 | case CPP_STRING16: |
1540 | case CPP_STRING32: |
1541 | case CPP_UTF8STRING: |
1542 | if (type != tok->type) |
1543 | { |
1544 | if (type == CPP_STRING) |
1545 | type = tok->type; |
1546 | else |
1547 | error ("unsupported non-standard concatenation of string literals" ); |
1548 | } |
1549 | /* FALLTHROUGH */ |
1550 | |
1551 | case CPP_STRING: |
1552 | if (!concats) |
1553 | { |
1554 | gcc_obstack_init (&str_ob); |
1555 | gcc_obstack_init (&loc_ob); |
1556 | obstack_grow (&str_ob, &str, sizeof (cpp_string)); |
1557 | obstack_grow (&loc_ob, &init_loc, sizeof (location_t)); |
1558 | } |
1559 | |
1560 | concats++; |
1561 | obstack_grow (&str_ob, &tok->val.str, sizeof (cpp_string)); |
1562 | obstack_grow (&loc_ob, &tok->src_loc, sizeof (location_t)); |
1563 | |
1564 | if (objc_string) |
1565 | objc_at_sign_was_seen = false; |
1566 | goto retry; |
1567 | } |
1568 | |
1569 | /* It is an error if we saw a '@' with no following string. */ |
1570 | if (objc_at_sign_was_seen) |
1571 | error ("stray %<@%> in program" ); |
1572 | |
1573 | /* We have read one more token than we want. */ |
1574 | _cpp_backup_tokens (parse_in, 1); |
1575 | if (concats) |
1576 | { |
1577 | strs = XOBFINISH (&str_ob, cpp_string *); |
1578 | locs = XOBFINISH (&loc_ob, location_t *); |
1579 | } |
1580 | |
1581 | if (concats && !objc_string && !in_system_header_at (loc: input_location)) |
1582 | warning (OPT_Wtraditional, |
1583 | "traditional C rejects string constant concatenation" ); |
1584 | |
1585 | if ((translate |
1586 | ? cpp_interpret_string : cpp_interpret_string_notranslate) |
1587 | (parse_in, strs, concats + 1, &istr, type)) |
1588 | { |
1589 | value = build_string (istr.len, (const char *) istr.text); |
1590 | free (CONST_CAST (unsigned char *, istr.text)); |
1591 | if (concats) |
1592 | { |
1593 | gcc_assert (locs); |
1594 | gcc_assert (g_string_concat_db); |
1595 | g_string_concat_db->record_string_concatenation (num: concats + 1, locs); |
1596 | } |
1597 | } |
1598 | else |
1599 | { |
1600 | /* Callers cannot generally handle error_mark_node in this context, |
1601 | so return the empty string instead. cpp_interpret_string has |
1602 | issued an error. */ |
1603 | switch (type) |
1604 | { |
1605 | default: |
1606 | case CPP_STRING: |
1607 | case CPP_UTF8STRING: |
1608 | if (type == CPP_UTF8STRING && flag_char8_t) |
1609 | { |
1610 | value = build_string (TYPE_PRECISION (char8_type_node) |
1611 | / TYPE_PRECISION (char_type_node), |
1612 | "" ); /* char8_t is 8 bits */ |
1613 | } |
1614 | else |
1615 | value = build_string (1, "" ); |
1616 | break; |
1617 | case CPP_STRING16: |
1618 | value = build_string (TYPE_PRECISION (char16_type_node) |
1619 | / TYPE_PRECISION (char_type_node), |
1620 | "\0" ); /* char16_t is 16 bits */ |
1621 | break; |
1622 | case CPP_STRING32: |
1623 | value = build_string (TYPE_PRECISION (char32_type_node) |
1624 | / TYPE_PRECISION (char_type_node), |
1625 | "\0\0\0" ); /* char32_t is 32 bits */ |
1626 | break; |
1627 | case CPP_WSTRING: |
1628 | value = build_string (TYPE_PRECISION (wchar_type_node) |
1629 | / TYPE_PRECISION (char_type_node), |
1630 | "\0\0\0" ); /* widest supported wchar_t |
1631 | is 32 bits */ |
1632 | break; |
1633 | } |
1634 | } |
1635 | |
1636 | switch (type) |
1637 | { |
1638 | default: |
1639 | case CPP_STRING: |
1640 | TREE_TYPE (value) = char_array_type_node; |
1641 | break; |
1642 | case CPP_UTF8STRING: |
1643 | if (flag_char8_t) |
1644 | TREE_TYPE (value) = char8_array_type_node; |
1645 | else |
1646 | TREE_TYPE (value) = char_array_type_node; |
1647 | break; |
1648 | case CPP_STRING16: |
1649 | TREE_TYPE (value) = char16_array_type_node; |
1650 | break; |
1651 | case CPP_STRING32: |
1652 | TREE_TYPE (value) = char32_array_type_node; |
1653 | break; |
1654 | case CPP_WSTRING: |
1655 | TREE_TYPE (value) = wchar_array_type_node; |
1656 | } |
1657 | *valp = fix_string_type (value); |
1658 | |
1659 | if (concats) |
1660 | { |
1661 | obstack_free (&str_ob, 0); |
1662 | obstack_free (&loc_ob, 0); |
1663 | } |
1664 | |
1665 | return objc_string ? CPP_OBJC_STRING : type; |
1666 | } |
1667 | |
1668 | /* Converts a (possibly wide) character constant token into a tree. */ |
1669 | static tree |
1670 | lex_charconst (const cpp_token *token) |
1671 | { |
1672 | cppchar_t result; |
1673 | tree type, value; |
1674 | unsigned int chars_seen; |
1675 | int unsignedp = 0; |
1676 | |
1677 | result = cpp_interpret_charconst (parse_in, token, |
1678 | &chars_seen, &unsignedp); |
1679 | |
1680 | if (token->type == CPP_WCHAR) |
1681 | type = wchar_type_node; |
1682 | else if (token->type == CPP_CHAR32) |
1683 | type = char32_type_node; |
1684 | else if (token->type == CPP_CHAR16) |
1685 | type = char16_type_node; |
1686 | else if (token->type == CPP_UTF8CHAR) |
1687 | { |
1688 | if (flag_char8_t) |
1689 | type = char8_type_node; |
1690 | else |
1691 | type = char_type_node; |
1692 | } |
1693 | /* In C, a character constant has type 'int'. |
1694 | In C++ 'char', but multi-char charconsts have type 'int'. */ |
1695 | else if (!c_dialect_cxx () || chars_seen > 1) |
1696 | type = integer_type_node; |
1697 | else |
1698 | type = char_type_node; |
1699 | |
1700 | /* Cast to cppchar_signed_t to get correct sign-extension of RESULT |
1701 | before possibly widening to HOST_WIDE_INT for build_int_cst. */ |
1702 | if (unsignedp || (cppchar_signed_t) result >= 0) |
1703 | value = build_int_cst (type, result); |
1704 | else |
1705 | value = build_int_cst (type, (cppchar_signed_t) result); |
1706 | |
1707 | return value; |
1708 | } |
1709 | |
1710 | /* Helper function for c_parser_peek_conflict_marker |
1711 | and cp_lexer_peek_conflict_marker. |
1712 | Given a possible conflict marker token of kind TOK1_KIND |
1713 | consisting of a pair of characters, get the token kind for the |
1714 | standalone final character. */ |
1715 | |
1716 | enum cpp_ttype |
1717 | conflict_marker_get_final_tok_kind (enum cpp_ttype tok1_kind) |
1718 | { |
1719 | switch (tok1_kind) |
1720 | { |
1721 | default: gcc_unreachable (); |
1722 | case CPP_LSHIFT: |
1723 | /* "<<" and '<' */ |
1724 | return CPP_LESS; |
1725 | |
1726 | case CPP_EQ_EQ: |
1727 | /* "==" and '=' */ |
1728 | return CPP_EQ; |
1729 | |
1730 | case CPP_RSHIFT: |
1731 | /* ">>" and '>' */ |
1732 | return CPP_GREATER; |
1733 | } |
1734 | } |
1735 | |