2 Copyright (C) 1986, 1987, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998,
3 1999, 2000, 2001, 2002, 2003, 2004, 2005 Free Software Foundation, Inc.
4 Contributed by Per Bothner, 1994-95.
5 Based on CCCP program by Paul Rubin, June 1986
6 Adapted to ANSI C, Richard Stallman, Jan 1987
8 This program is free software; you can redistribute it and/or modify it
9 under the terms of the GNU General Public License as published by the
10 Free Software Foundation; either version 2, or (at your option) any
13 This program is distributed in the hope that it will be useful,
14 but WITHOUT ANY WARRANTY; without even the implied warranty of
15 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 GNU General Public License for more details.
18 You should have received a copy of the GNU General Public License
19 along with this program; if not, write to the Free Software
20 Foundation, 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA. */
27 #include "localedir.h"
29 static void init_library (void);
30 static void mark_named_operators (cpp_reader *);
31 static void read_original_filename (cpp_reader *);
32 static void read_original_directory (cpp_reader *);
33 static void post_options (cpp_reader *);
35 /* If we have designated initializers (GCC >2.7) these tables can be
36 initialized, constant data. Otherwise, they have to be filled in at
38 #if HAVE_DESIGNATED_INITIALIZERS
40 #define init_trigraph_map() /* Nothing. */
41 #define TRIGRAPH_MAP \
42 __extension__ const uchar _cpp_trigraph_map[UCHAR_MAX + 1] = {
45 #define s(p, v) [p] = v,
49 #define TRIGRAPH_MAP uchar _cpp_trigraph_map[UCHAR_MAX + 1] = { 0 }; \
50 static void init_trigraph_map (void) { \
51 unsigned char *x = _cpp_trigraph_map;
54 #define s(p, v) x[p] = v;
59 s('=', '#') s(')', ']') s('!', '|')
60 s('(', '[') s('\'', '^') s('>', '}')
61 s('/', '\\') s('<', '{') s('-', '~')
68 /* A set of booleans indicating what CPP features each source language
74 char extended_numbers;
75 char extended_identifiers;
77 char cplusplus_comments;
81 static const struct lang_flags lang_defaults[] =
82 { /* c99 c++ xnum xid std // digr */
83 /* GNUC89 */ { 0, 0, 1, 0, 0, 1, 1 },
84 /* GNUC99 */ { 1, 0, 1, 0, 0, 1, 1 },
85 /* STDC89 */ { 0, 0, 0, 0, 1, 0, 0 },
86 /* STDC94 */ { 0, 0, 0, 0, 1, 0, 1 },
87 /* STDC99 */ { 1, 0, 1, 0, 1, 1, 1 },
88 /* GNUCXX */ { 0, 1, 1, 0, 0, 1, 1 },
89 /* CXX98 */ { 0, 1, 1, 0, 1, 1, 1 },
90 /* ASM */ { 0, 0, 1, 0, 0, 1, 0 }
91 /* xid should be 1 for GNUC99, STDC99, GNUCXX and CXX98 when no
92 longer experimental (when all uses of identifiers in the compiler
93 have been audited for correct handling of extended
97 /* Sets internal flags correctly for a given language. */
99 cpp_set_lang (cpp_reader *pfile, enum c_lang lang)
101 const struct lang_flags *l = &lang_defaults[(int) lang];
103 CPP_OPTION (pfile, lang) = lang;
105 CPP_OPTION (pfile, c99) = l->c99;
106 CPP_OPTION (pfile, cplusplus) = l->cplusplus;
107 CPP_OPTION (pfile, extended_numbers) = l->extended_numbers;
108 CPP_OPTION (pfile, extended_identifiers) = l->extended_identifiers;
109 CPP_OPTION (pfile, std) = l->std;
110 CPP_OPTION (pfile, trigraphs) = l->std;
111 CPP_OPTION (pfile, cplusplus_comments) = l->cplusplus_comments;
112 CPP_OPTION (pfile, digraphs) = l->digraphs;
115 /* Initialize library global state. */
119 static int initialized = 0;
125 /* Set up the trigraph map. This doesn't need to do anything if
126 we were compiled with a compiler that supports C99 designated
128 init_trigraph_map ();
131 (void) bindtextdomain (PACKAGE, LOCALEDIR);
136 /* Initialize a cpp_reader structure. */
138 cpp_create_reader (enum c_lang lang, hash_table *table,
139 struct line_maps *line_table)
143 /* Initialize this instance of the library if it hasn't been already. */
146 pfile = XCNEW (cpp_reader);
148 cpp_set_lang (pfile, lang);
149 /* APPLE LOCAL begin -Wnewline-eof 2001-08-23 --sts */
150 /* Suppress warnings about missing newlines at ends of files. */
151 CPP_OPTION (pfile, warn_newline_at_eof) = 0;
152 /* APPLE LOCAL end -Wnewline-eof 2001-08-23 --sts */
153 CPP_OPTION (pfile, warn_multichar) = 1;
154 CPP_OPTION (pfile, discard_comments) = 1;
155 CPP_OPTION (pfile, discard_comments_in_macro_exp) = 1;
156 CPP_OPTION (pfile, show_column) = 1;
157 CPP_OPTION (pfile, tabstop) = 8;
158 CPP_OPTION (pfile, operator_names) = 1;
159 CPP_OPTION (pfile, warn_trigraphs) = 2;
160 CPP_OPTION (pfile, warn_endif_labels) = 1;
161 CPP_OPTION (pfile, warn_deprecated) = 1;
162 CPP_OPTION (pfile, warn_long_long) = !CPP_OPTION (pfile, c99);
163 CPP_OPTION (pfile, dollars_in_ident) = 1;
164 CPP_OPTION (pfile, warn_dollars) = 1;
165 CPP_OPTION (pfile, warn_variadic_macros) = 1;
166 CPP_OPTION (pfile, warn_normalize) = normalized_C;
168 /* Default CPP arithmetic to something sensible for the host for the
169 benefit of dumb users like fix-header. */
170 CPP_OPTION (pfile, precision) = CHAR_BIT * sizeof (long);
171 CPP_OPTION (pfile, char_precision) = CHAR_BIT;
172 CPP_OPTION (pfile, wchar_precision) = CHAR_BIT * sizeof (int);
173 CPP_OPTION (pfile, int_precision) = CHAR_BIT * sizeof (int);
174 CPP_OPTION (pfile, unsigned_char) = 0;
175 CPP_OPTION (pfile, unsigned_wchar) = 1;
176 CPP_OPTION (pfile, bytes_big_endian) = 1; /* does not matter */
178 /* Default to no charset conversion. */
179 CPP_OPTION (pfile, narrow_charset) = _cpp_default_encoding ();
180 CPP_OPTION (pfile, wide_charset) = 0;
182 /* Default the input character set to UTF-8. */
183 CPP_OPTION (pfile, input_charset) = _cpp_default_encoding ();
185 /* A fake empty "directory" used as the starting point for files
186 looked up without a search path. Name cannot be '/' because we
187 don't want to prepend anything at all to filenames using it. All
188 other entries are correct zero-initialized. */
189 pfile->no_search_path.name = (char *) "";
191 /* Initialize the line map. */
192 pfile->line_table = line_table;
194 /* Initialize lexer state. */
195 pfile->state.save_comments = ! CPP_OPTION (pfile, discard_comments);
197 /* Set up static tokens. */
198 pfile->avoid_paste.type = CPP_PADDING;
199 pfile->avoid_paste.val.source = NULL;
200 pfile->eof.type = CPP_EOF;
201 pfile->eof.flags = 0;
203 /* Create a token buffer for the lexer. */
204 _cpp_init_tokenrun (&pfile->base_run, 250);
205 pfile->cur_run = &pfile->base_run;
206 pfile->cur_token = pfile->base_run.base;
208 /* Initialize the base context. */
209 pfile->context = &pfile->base_context;
210 pfile->base_context.macro = 0;
211 pfile->base_context.prev = pfile->base_context.next = 0;
213 /* Aligned and unaligned storage. */
214 pfile->a_buff = _cpp_get_buff (pfile, 0);
215 pfile->u_buff = _cpp_get_buff (pfile, 0);
217 /* The expression parser stack. */
218 _cpp_expand_op_stack (pfile);
220 /* Initialize the buffer obstack. */
221 _obstack_begin (&pfile->buffer_ob, 0, 0,
222 (void *(*) (long)) xmalloc,
223 (void (*) (void *)) free);
225 _cpp_init_files (pfile);
227 _cpp_init_hashtable (pfile, table);
232 /* Free resources used by PFILE. Accessing PFILE after this function
233 returns leads to undefined behavior. Returns the error count. */
235 cpp_destroy (cpp_reader *pfile)
237 cpp_context *context, *contextn;
238 tokenrun *run, *runn;
240 free (pfile->op_stack);
242 while (CPP_BUFFER (pfile) != NULL)
243 _cpp_pop_buffer (pfile);
246 free (pfile->out.base);
248 if (pfile->macro_buffer)
250 free (pfile->macro_buffer);
251 pfile->macro_buffer = NULL;
252 pfile->macro_buffer_len = 0;
256 deps_free (pfile->deps);
257 obstack_free (&pfile->buffer_ob, 0);
259 _cpp_destroy_hashtable (pfile);
260 _cpp_cleanup_files (pfile);
261 _cpp_destroy_iconv (pfile);
263 _cpp_free_buff (pfile->a_buff);
264 _cpp_free_buff (pfile->u_buff);
265 _cpp_free_buff (pfile->free_buffs);
267 for (run = &pfile->base_run; run; run = runn)
271 if (run != &pfile->base_run)
275 for (context = pfile->base_context.next; context; context = contextn)
277 contextn = context->next;
284 /* This structure defines one built-in identifier. A node will be
285 entered in the hash table under the name NAME, with value VALUE.
287 There are two tables of these. builtin_array holds all the
288 "builtin" macros: these are handled by builtin_macro() in
289 macro.c. Builtin is somewhat of a misnomer -- the property of
290 interest is that these macros require special code to compute their
291 expansions. The value is a "builtin_type" enumerator.
293 operator_array holds the C++ named operators. These are keywords
294 which act as aliases for punctuators. In C++, they cannot be
295 altered through #define, and #if recognizes them as operators. In
296 C, these are not entered into the hash table at all (but see
297 <iso646.h>). The value is a token-type enumerator. */
302 unsigned short value;
305 #define B(n, t) { DSC(n), t }
306 static const struct builtin builtin_array[] =
308 B("__TIMESTAMP__", BT_TIMESTAMP),
309 B("__TIME__", BT_TIME),
310 B("__DATE__", BT_DATE),
311 B("__FILE__", BT_FILE),
312 B("__BASE_FILE__", BT_BASE_FILE),
313 B("__LINE__", BT_SPECLINE),
314 B("__INCLUDE_LEVEL__", BT_INCLUDE_LEVEL),
315 B("__COUNTER__", BT_COUNTER),
316 /* Keep builtins not used for -traditional-cpp at the end, and
317 update init_builtins() if any more are added. */
318 B("_Pragma", BT_PRAGMA),
319 B("__STDC__", BT_STDC),
322 static const struct builtin operator_array[] =
324 B("and", CPP_AND_AND),
325 B("and_eq", CPP_AND_EQ),
326 B("bitand", CPP_AND),
328 B("compl", CPP_COMPL),
330 B("not_eq", CPP_NOT_EQ),
332 B("or_eq", CPP_OR_EQ),
334 B("xor_eq", CPP_XOR_EQ)
338 /* Mark the C++ named operators in the hash table. */
340 mark_named_operators (cpp_reader *pfile)
342 const struct builtin *b;
344 for (b = operator_array;
345 b < (operator_array + ARRAY_SIZE (operator_array));
348 cpp_hashnode *hp = cpp_lookup (pfile, b->name, b->len);
349 hp->flags |= NODE_OPERATOR;
350 hp->is_directive = 0;
351 hp->directive_index = b->value;
356 cpp_init_special_builtins (cpp_reader *pfile)
358 const struct builtin *b;
359 size_t n = ARRAY_SIZE (builtin_array);
361 if (CPP_OPTION (pfile, traditional))
363 else if (! CPP_OPTION (pfile, stdc_0_in_system_headers)
364 || CPP_OPTION (pfile, std))
367 for (b = builtin_array; b < builtin_array + n; b++)
369 cpp_hashnode *hp = cpp_lookup (pfile, b->name, b->len);
371 hp->flags |= NODE_BUILTIN | NODE_WARN;
372 hp->value.builtin = (enum builtin_type) b->value;
376 /* Read the builtins table above and enter them, and language-specific
377 macros, into the hash table. HOSTED is true if this is a hosted
380 cpp_init_builtins (cpp_reader *pfile, int hosted)
382 cpp_init_special_builtins (pfile);
384 if (!CPP_OPTION (pfile, traditional)
385 && (! CPP_OPTION (pfile, stdc_0_in_system_headers)
386 || CPP_OPTION (pfile, std)))
387 _cpp_define_builtin (pfile, "__STDC__ 1");
389 if (CPP_OPTION (pfile, cplusplus))
390 _cpp_define_builtin (pfile, "__cplusplus 1");
391 else if (CPP_OPTION (pfile, lang) == CLK_ASM)
392 _cpp_define_builtin (pfile, "__ASSEMBLER__ 1");
393 else if (CPP_OPTION (pfile, lang) == CLK_STDC94)
394 _cpp_define_builtin (pfile, "__STDC_VERSION__ 199409L");
395 else if (CPP_OPTION (pfile, c99))
396 _cpp_define_builtin (pfile, "__STDC_VERSION__ 199901L");
399 _cpp_define_builtin (pfile, "__STDC_HOSTED__ 1");
401 _cpp_define_builtin (pfile, "__STDC_HOSTED__ 0");
403 if (CPP_OPTION (pfile, objc))
404 _cpp_define_builtin (pfile, "__OBJC__ 1");
407 /* Sanity-checks are dependent on command-line options, so it is
408 called as a subroutine of cpp_read_main_file (). */
410 static void sanity_checks (cpp_reader *);
411 static void sanity_checks (cpp_reader *pfile)
414 size_t max_precision = 2 * CHAR_BIT * sizeof (cpp_num_part);
416 /* Sanity checks for assumptions about CPP arithmetic and target
417 type precisions made by cpplib. */
420 cpp_error (pfile, CPP_DL_ICE, "cppchar_t must be an unsigned type");
422 if (CPP_OPTION (pfile, precision) > max_precision)
423 cpp_error (pfile, CPP_DL_ICE,
424 "preprocessor arithmetic has maximum precision of %lu bits;"
425 " target requires %lu bits",
426 (unsigned long) max_precision,
427 (unsigned long) CPP_OPTION (pfile, precision));
429 if (CPP_OPTION (pfile, precision) < CPP_OPTION (pfile, int_precision))
430 cpp_error (pfile, CPP_DL_ICE,
431 "CPP arithmetic must be at least as precise as a target int");
433 if (CPP_OPTION (pfile, char_precision) < 8)
434 cpp_error (pfile, CPP_DL_ICE, "target char is less than 8 bits wide");
436 if (CPP_OPTION (pfile, wchar_precision) < CPP_OPTION (pfile, char_precision))
437 cpp_error (pfile, CPP_DL_ICE,
438 "target wchar_t is narrower than target char");
440 if (CPP_OPTION (pfile, int_precision) < CPP_OPTION (pfile, char_precision))
441 cpp_error (pfile, CPP_DL_ICE,
442 "target int is narrower than target char");
444 /* This is assumed in eval_token() and could be fixed if necessary. */
445 if (sizeof (cppchar_t) > sizeof (cpp_num_part))
446 cpp_error (pfile, CPP_DL_ICE,
447 "CPP half-integer narrower than CPP character");
449 if (CPP_OPTION (pfile, wchar_precision) > BITS_PER_CPPCHAR_T)
450 cpp_error (pfile, CPP_DL_ICE,
451 "CPP on this host cannot handle wide character constants over"
452 " %lu bits, but the target requires %lu bits",
453 (unsigned long) BITS_PER_CPPCHAR_T,
454 (unsigned long) CPP_OPTION (pfile, wchar_precision));
457 # define sanity_checks(PFILE)
460 /* This is called after options have been parsed, and partially
463 cpp_post_options (cpp_reader *pfile)
465 sanity_checks (pfile);
467 post_options (pfile);
469 /* Mark named operators before handling command line macros. */
470 if (CPP_OPTION (pfile, cplusplus) && CPP_OPTION (pfile, operator_names))
471 mark_named_operators (pfile);
474 /* Setup for processing input from the file named FNAME, or stdin if
475 it is the empty string. Return the original filename
476 on success (e.g. foo.i->foo.c), or NULL on failure. */
478 cpp_read_main_file (cpp_reader *pfile, const char *fname)
480 if (CPP_OPTION (pfile, deps.style) != DEPS_NONE)
483 pfile->deps = deps_init ();
485 /* Set the default target (if there is none already). */
486 deps_add_default_target (pfile->deps, fname);
490 = _cpp_find_file (pfile, fname, &pfile->no_search_path, false, 0);
491 if (_cpp_find_failed (pfile->main_file))
494 _cpp_stack_file (pfile, pfile->main_file, false);
496 /* For foo.i, read the original filename foo.c now, for the benefit
497 of the front ends. */
498 if (CPP_OPTION (pfile, preprocessed))
500 read_original_filename (pfile);
501 fname = pfile->line_table->maps[pfile->line_table->used-1].to_file;
506 /* For preprocessed files, if the first tokens are of the form # NUM.
507 handle the directive so we know the original file name. This will
508 generate file_change callbacks, which the front ends must handle
509 appropriately given their state of initialization. */
511 read_original_filename (cpp_reader *pfile)
513 const cpp_token *token, *token1;
515 /* Lex ahead; if the first tokens are of the form # NUM, then
516 process the directive, otherwise back up. */
517 token = _cpp_lex_direct (pfile);
518 if (token->type == CPP_HASH)
520 pfile->state.in_directive = 1;
521 token1 = _cpp_lex_direct (pfile);
522 _cpp_backup_tokens (pfile, 1);
523 pfile->state.in_directive = 0;
525 /* If it's a #line directive, handle it. */
526 if (token1->type == CPP_NUMBER)
528 _cpp_handle_directive (pfile, token->flags & PREV_WHITE);
529 read_original_directory (pfile);
534 /* Backup as if nothing happened. */
535 _cpp_backup_tokens (pfile, 1);
538 /* For preprocessed files, if the tokens following the first filename
539 line is of the form # <line> "/path/name//", handle the
540 directive so we know the original current directory. */
542 read_original_directory (cpp_reader *pfile)
544 const cpp_token *hash, *token;
546 /* Lex ahead; if the first tokens are of the form # NUM, then
547 process the directive, otherwise back up. */
548 hash = _cpp_lex_direct (pfile);
549 if (hash->type != CPP_HASH)
551 _cpp_backup_tokens (pfile, 1);
555 token = _cpp_lex_direct (pfile);
557 if (token->type != CPP_NUMBER)
559 _cpp_backup_tokens (pfile, 2);
563 token = _cpp_lex_direct (pfile);
565 if (token->type != CPP_STRING
566 || ! (token->val.str.len >= 5
567 && token->val.str.text[token->val.str.len-2] == '/'
568 && token->val.str.text[token->val.str.len-3] == '/'))
570 _cpp_backup_tokens (pfile, 3);
574 if (pfile->cb.dir_change)
576 char *debugdir = (char *) alloca (token->val.str.len - 3);
578 memcpy (debugdir, (const char *) token->val.str.text + 1,
579 token->val.str.len - 4);
580 debugdir[token->val.str.len - 4] = '\0';
582 pfile->cb.dir_change (pfile, debugdir);
586 /* This is called at the end of preprocessing. It pops the last
587 buffer and writes dependency output, and returns the number of
590 Maybe it should also reset state, such that you could call
591 cpp_start_read with a new filename to restart processing. */
593 cpp_finish (cpp_reader *pfile, FILE *deps_stream)
595 /* Warn about unused macros before popping the final buffer. */
596 if (CPP_OPTION (pfile, warn_unused_macros))
597 cpp_forall_identifiers (pfile, _cpp_warn_if_unused_macro, NULL);
599 /* lex.c leaves the final buffer on the stack. This it so that
600 it returns an unending stream of CPP_EOFs to the client. If we
601 popped the buffer, we'd dereference a NULL buffer pointer and
602 segfault. It's nice to allow the client to do worry-free excess
603 cpp_get_token calls. */
604 while (pfile->buffer)
605 _cpp_pop_buffer (pfile);
607 /* Don't write the deps file if there are errors. */
608 if (CPP_OPTION (pfile, deps.style) != DEPS_NONE
609 && deps_stream && pfile->errors == 0)
611 deps_write (pfile->deps, deps_stream, 72);
613 if (CPP_OPTION (pfile, deps.phony_targets))
614 deps_phony_targets (pfile->deps, deps_stream);
617 /* Report on headers that could use multiple include guards. */
618 if (CPP_OPTION (pfile, print_include_names))
619 _cpp_report_missing_guards (pfile);
621 return pfile->errors;
625 post_options (cpp_reader *pfile)
627 /* -Wtraditional is not useful in C++ mode. */
628 if (CPP_OPTION (pfile, cplusplus))
629 CPP_OPTION (pfile, warn_traditional) = 0;
631 /* Permanently disable macro expansion if we are rescanning
632 preprocessed text. Read preprocesed source in ISO mode. */
633 if (CPP_OPTION (pfile, preprocessed))
635 if (!CPP_OPTION (pfile, directives_only))
636 pfile->state.prevent_expansion = 1;
637 CPP_OPTION (pfile, traditional) = 0;
640 if (CPP_OPTION (pfile, warn_trigraphs) == 2)
641 CPP_OPTION (pfile, warn_trigraphs) = !CPP_OPTION (pfile, trigraphs);
643 if (CPP_OPTION (pfile, traditional))
645 CPP_OPTION (pfile, cplusplus_comments) = 0;
647 /* Traditional CPP does not accurately track column information. */
648 CPP_OPTION (pfile, show_column) = 0;
649 CPP_OPTION (pfile, trigraphs) = 0;
650 CPP_OPTION (pfile, warn_trigraphs) = 0;