1/* CPP Library.
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
7
8This program is free software; you can redistribute it and/or modify it
9under the terms of the GNU General Public License as published by the
10Free Software Foundation; either version 2, or (at your option) any
11later version.
12
13This program is distributed in the hope that it will be useful,
14but WITHOUT ANY WARRANTY; without even the implied warranty of
15MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
16GNU General Public License for more details.
17
18You should have received a copy of the GNU General Public License
19along with this program; if not, write to the Free Software
20Foundation, 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.  */
21
22#include "config.h"
23#include "system.h"
24#include "cpplib.h"
25#include "internal.h"
26#include "mkdeps.h"
27#include "localedir.h"
28
29static void init_library (void);
30static void mark_named_operators (cpp_reader *);
31static void read_original_filename (cpp_reader *);
32static void read_original_directory (cpp_reader *);
33static void post_options (cpp_reader *);
34
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
37   runtime.  */
38#if HAVE_DESIGNATED_INITIALIZERS
39
40#define init_trigraph_map()  /* Nothing.  */
41#define TRIGRAPH_MAP \
42__extension__ const uchar _cpp_trigraph_map[UCHAR_MAX + 1] = {
43
44#define END };
45#define s(p, v) [p] = v,
46
47#else
48
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;
52
53#define END }
54#define s(p, v) x[p] = v;
55
56#endif
57
58TRIGRAPH_MAP
59  s('=', '#')	s(')', ']')	s('!', '|')
60  s('(', '[')	s('\'', '^')	s('>', '}')
61  s('/', '\\')	s('<', '{')	s('-', '~')
62END
63
64#undef s
65#undef END
66#undef TRIGRAPH_MAP
67
68/* A set of booleans indicating what CPP features each source language
69   requires.  */
70struct lang_flags
71{
72  char c99;
73  char cplusplus;
74  char extended_numbers;
75  char extended_identifiers;
76  char std;
77  char cplusplus_comments;
78  char digraphs;
79};
80
81static 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
94     identifiers).  */
95};
96
97/* Sets internal flags correctly for a given language.  */
98void
99cpp_set_lang (cpp_reader *pfile, enum c_lang lang)
100{
101  const struct lang_flags *l = &lang_defaults[(int) lang];
102
103  CPP_OPTION (pfile, lang) = lang;
104
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;
113}
114
115/* Initialize library global state.  */
116static void
117init_library (void)
118{
119  static int initialized = 0;
120
121  if (! initialized)
122    {
123      initialized = 1;
124
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
127	 initializers.  */
128      init_trigraph_map ();
129
130#ifdef ENABLE_NLS
131       (void) bindtextdomain (PACKAGE, LOCALEDIR);
132#endif
133    }
134}
135
136/* Initialize a cpp_reader structure.  */
137cpp_reader *
138cpp_create_reader (enum c_lang lang, hash_table *table,
139		   struct line_maps *line_table)
140{
141  cpp_reader *pfile;
142
143  /* Initialize this instance of the library if it hasn't been already.  */
144  init_library ();
145
146  pfile = XCNEW (cpp_reader);
147
148  cpp_set_lang (pfile, lang);
149  CPP_OPTION (pfile, warn_multichar) = 1;
150  CPP_OPTION (pfile, discard_comments) = 1;
151  CPP_OPTION (pfile, discard_comments_in_macro_exp) = 1;
152  CPP_OPTION (pfile, show_column) = 1;
153  CPP_OPTION (pfile, tabstop) = 8;
154  CPP_OPTION (pfile, operator_names) = 1;
155  CPP_OPTION (pfile, warn_trigraphs) = 2;
156  CPP_OPTION (pfile, warn_endif_labels) = 1;
157  CPP_OPTION (pfile, warn_deprecated) = 1;
158  CPP_OPTION (pfile, warn_long_long) = !CPP_OPTION (pfile, c99);
159  CPP_OPTION (pfile, dollars_in_ident) = 1;
160  CPP_OPTION (pfile, warn_dollars) = 1;
161  CPP_OPTION (pfile, warn_variadic_macros) = 1;
162  CPP_OPTION (pfile, warn_normalize) = normalized_C;
163
164  /* Default CPP arithmetic to something sensible for the host for the
165     benefit of dumb users like fix-header.  */
166  CPP_OPTION (pfile, precision) = CHAR_BIT * sizeof (long);
167  CPP_OPTION (pfile, char_precision) = CHAR_BIT;
168  CPP_OPTION (pfile, wchar_precision) = CHAR_BIT * sizeof (int);
169  CPP_OPTION (pfile, int_precision) = CHAR_BIT * sizeof (int);
170  CPP_OPTION (pfile, unsigned_char) = 0;
171  CPP_OPTION (pfile, unsigned_wchar) = 1;
172  CPP_OPTION (pfile, bytes_big_endian) = 1;  /* does not matter */
173
174  /* Default to no charset conversion.  */
175  CPP_OPTION (pfile, narrow_charset) = _cpp_default_encoding ();
176  CPP_OPTION (pfile, wide_charset) = 0;
177
178  /* Default the input character set to UTF-8.  */
179  CPP_OPTION (pfile, input_charset) = _cpp_default_encoding ();
180
181  /* A fake empty "directory" used as the starting point for files
182     looked up without a search path.  Name cannot be '/' because we
183     don't want to prepend anything at all to filenames using it.  All
184     other entries are correct zero-initialized.  */
185  pfile->no_search_path.name = (char *) "";
186
187  /* Initialize the line map.  */
188  pfile->line_table = line_table;
189
190  /* Initialize lexer state.  */
191  pfile->state.save_comments = ! CPP_OPTION (pfile, discard_comments);
192
193  /* Set up static tokens.  */
194  pfile->avoid_paste.type = CPP_PADDING;
195  pfile->avoid_paste.val.source = NULL;
196  pfile->eof.type = CPP_EOF;
197  pfile->eof.flags = 0;
198
199  /* Create a token buffer for the lexer.  */
200  _cpp_init_tokenrun (&pfile->base_run, 250);
201  pfile->cur_run = &pfile->base_run;
202  pfile->cur_token = pfile->base_run.base;
203
204  /* Initialize the base context.  */
205  pfile->context = &pfile->base_context;
206  pfile->base_context.macro = 0;
207  pfile->base_context.prev = pfile->base_context.next = 0;
208
209  /* Aligned and unaligned storage.  */
210  pfile->a_buff = _cpp_get_buff (pfile, 0);
211  pfile->u_buff = _cpp_get_buff (pfile, 0);
212
213  /* The expression parser stack.  */
214  _cpp_expand_op_stack (pfile);
215
216  /* Initialize the buffer obstack.  */
217  _obstack_begin (&pfile->buffer_ob, 0, 0,
218		  (void *(*) (long)) xmalloc,
219		  (void (*) (void *)) free);
220
221  _cpp_init_files (pfile);
222
223  _cpp_init_hashtable (pfile, table);
224
225  return pfile;
226}
227
228/* Free resources used by PFILE.  Accessing PFILE after this function
229   returns leads to undefined behavior.  Returns the error count.  */
230void
231cpp_destroy (cpp_reader *pfile)
232{
233  cpp_context *context, *contextn;
234  tokenrun *run, *runn;
235
236  free (pfile->op_stack);
237
238  while (CPP_BUFFER (pfile) != NULL)
239    _cpp_pop_buffer (pfile);
240
241  if (pfile->out.base)
242    free (pfile->out.base);
243
244  if (pfile->macro_buffer)
245    {
246      free (pfile->macro_buffer);
247      pfile->macro_buffer = NULL;
248      pfile->macro_buffer_len = 0;
249    }
250
251  if (pfile->deps)
252    deps_free (pfile->deps);
253  obstack_free (&pfile->buffer_ob, 0);
254
255  _cpp_destroy_hashtable (pfile);
256  _cpp_cleanup_files (pfile);
257  _cpp_destroy_iconv (pfile);
258
259  _cpp_free_buff (pfile->a_buff);
260  _cpp_free_buff (pfile->u_buff);
261  _cpp_free_buff (pfile->free_buffs);
262
263  for (run = &pfile->base_run; run; run = runn)
264    {
265      runn = run->next;
266      free (run->base);
267      if (run != &pfile->base_run)
268	free (run);
269    }
270
271  for (context = pfile->base_context.next; context; context = contextn)
272    {
273      contextn = context->next;
274      free (context);
275    }
276
277  free (pfile);
278}
279
280/* This structure defines one built-in identifier.  A node will be
281   entered in the hash table under the name NAME, with value VALUE.
282
283   There are two tables of these.  builtin_array holds all the
284   "builtin" macros: these are handled by builtin_macro() in
285   macro.c.  Builtin is somewhat of a misnomer -- the property of
286   interest is that these macros require special code to compute their
287   expansions.  The value is a "builtin_type" enumerator.
288
289   operator_array holds the C++ named operators.  These are keywords
290   which act as aliases for punctuators.  In C++, they cannot be
291   altered through #define, and #if recognizes them as operators.  In
292   C, these are not entered into the hash table at all (but see
293   <iso646.h>).  The value is a token-type enumerator.  */
294struct builtin
295{
296  const uchar *name;
297  unsigned short len;
298  unsigned short value;
299};
300
301#define B(n, t)    { DSC(n), t }
302static const struct builtin builtin_array[] =
303{
304  B("__TIMESTAMP__",	 BT_TIMESTAMP),
305  B("__TIME__",		 BT_TIME),
306  B("__DATE__",		 BT_DATE),
307  B("__FILE__",		 BT_FILE),
308  B("__BASE_FILE__",	 BT_BASE_FILE),
309  B("__LINE__",		 BT_SPECLINE),
310  B("__INCLUDE_LEVEL__", BT_INCLUDE_LEVEL),
311  /* Keep builtins not used for -traditional-cpp at the end, and
312     update init_builtins() if any more are added.  */
313  B("_Pragma",		 BT_PRAGMA),
314  B("__STDC__",		 BT_STDC),
315};
316
317static const struct builtin operator_array[] =
318{
319  B("and",	CPP_AND_AND),
320  B("and_eq",	CPP_AND_EQ),
321  B("bitand",	CPP_AND),
322  B("bitor",	CPP_OR),
323  B("compl",	CPP_COMPL),
324  B("not",	CPP_NOT),
325  B("not_eq",	CPP_NOT_EQ),
326  B("or",	CPP_OR_OR),
327  B("or_eq",	CPP_OR_EQ),
328  B("xor",	CPP_XOR),
329  B("xor_eq",	CPP_XOR_EQ)
330};
331#undef B
332
333/* Mark the C++ named operators in the hash table.  */
334static void
335mark_named_operators (cpp_reader *pfile)
336{
337  const struct builtin *b;
338
339  for (b = operator_array;
340       b < (operator_array + ARRAY_SIZE (operator_array));
341       b++)
342    {
343      cpp_hashnode *hp = cpp_lookup (pfile, b->name, b->len);
344      hp->flags |= NODE_OPERATOR;
345      hp->is_directive = 0;
346      hp->directive_index = b->value;
347    }
348}
349
350/* Read the builtins table above and enter them, and language-specific
351   macros, into the hash table.  HOSTED is true if this is a hosted
352   environment.  */
353void
354cpp_init_builtins (cpp_reader *pfile, int hosted)
355{
356  const struct builtin *b;
357  size_t n = ARRAY_SIZE (builtin_array);
358
359  if (CPP_OPTION (pfile, traditional))
360    n -= 2;
361  else if (! CPP_OPTION (pfile, stdc_0_in_system_headers)
362	   || CPP_OPTION (pfile, std))
363    {
364      n--;
365      _cpp_define_builtin (pfile, "__STDC__ 1");
366    }
367
368  for (b = builtin_array; b < builtin_array + n; b++)
369    {
370      cpp_hashnode *hp = cpp_lookup (pfile, b->name, b->len);
371      hp->type = NT_MACRO;
372      hp->flags |= NODE_BUILTIN | NODE_WARN;
373      hp->value.builtin = (enum builtin_type) b->value;
374    }
375
376  if (CPP_OPTION (pfile, cplusplus))
377    _cpp_define_builtin (pfile, "__cplusplus 1");
378  else if (CPP_OPTION (pfile, lang) == CLK_ASM)
379    _cpp_define_builtin (pfile, "__ASSEMBLER__ 1");
380  else if (CPP_OPTION (pfile, lang) == CLK_STDC94)
381    _cpp_define_builtin (pfile, "__STDC_VERSION__ 199409L");
382  else if (CPP_OPTION (pfile, c99))
383    _cpp_define_builtin (pfile, "__STDC_VERSION__ 199901L");
384
385  if (hosted)
386    _cpp_define_builtin (pfile, "__STDC_HOSTED__ 1");
387  else
388    _cpp_define_builtin (pfile, "__STDC_HOSTED__ 0");
389
390  if (CPP_OPTION (pfile, objc))
391    _cpp_define_builtin (pfile, "__OBJC__ 1");
392}
393
394/* Sanity-checks are dependent on command-line options, so it is
395   called as a subroutine of cpp_read_main_file ().  */
396#if ENABLE_CHECKING
397static void sanity_checks (cpp_reader *);
398static void sanity_checks (cpp_reader *pfile)
399{
400  cppchar_t test = 0;
401  size_t max_precision = 2 * CHAR_BIT * sizeof (cpp_num_part);
402
403  /* Sanity checks for assumptions about CPP arithmetic and target
404     type precisions made by cpplib.  */
405  test--;
406  if (test < 1)
407    cpp_error (pfile, CPP_DL_ICE, "cppchar_t must be an unsigned type");
408
409  if (CPP_OPTION (pfile, precision) > max_precision)
410    cpp_error (pfile, CPP_DL_ICE,
411	       "preprocessor arithmetic has maximum precision of %lu bits;"
412	       " target requires %lu bits",
413	       (unsigned long) max_precision,
414	       (unsigned long) CPP_OPTION (pfile, precision));
415
416  if (CPP_OPTION (pfile, precision) < CPP_OPTION (pfile, int_precision))
417    cpp_error (pfile, CPP_DL_ICE,
418	       "CPP arithmetic must be at least as precise as a target int");
419
420  if (CPP_OPTION (pfile, char_precision) < 8)
421    cpp_error (pfile, CPP_DL_ICE, "target char is less than 8 bits wide");
422
423  if (CPP_OPTION (pfile, wchar_precision) < CPP_OPTION (pfile, char_precision))
424    cpp_error (pfile, CPP_DL_ICE,
425	       "target wchar_t is narrower than target char");
426
427  if (CPP_OPTION (pfile, int_precision) < CPP_OPTION (pfile, char_precision))
428    cpp_error (pfile, CPP_DL_ICE,
429	       "target int is narrower than target char");
430
431  /* This is assumed in eval_token() and could be fixed if necessary.  */
432  if (sizeof (cppchar_t) > sizeof (cpp_num_part))
433    cpp_error (pfile, CPP_DL_ICE,
434	       "CPP half-integer narrower than CPP character");
435
436  if (CPP_OPTION (pfile, wchar_precision) > BITS_PER_CPPCHAR_T)
437    cpp_error (pfile, CPP_DL_ICE,
438	       "CPP on this host cannot handle wide character constants over"
439	       " %lu bits, but the target requires %lu bits",
440	       (unsigned long) BITS_PER_CPPCHAR_T,
441	       (unsigned long) CPP_OPTION (pfile, wchar_precision));
442}
443#else
444# define sanity_checks(PFILE)
445#endif
446
447/* This is called after options have been parsed, and partially
448   processed.  */
449void
450cpp_post_options (cpp_reader *pfile)
451{
452  sanity_checks (pfile);
453
454  post_options (pfile);
455
456  /* Mark named operators before handling command line macros.  */
457  if (CPP_OPTION (pfile, cplusplus) && CPP_OPTION (pfile, operator_names))
458    mark_named_operators (pfile);
459}
460
461/* Setup for processing input from the file named FNAME, or stdin if
462   it is the empty string.  Return the original filename
463   on success (e.g. foo.i->foo.c), or NULL on failure.  */
464const char *
465cpp_read_main_file (cpp_reader *pfile, const char *fname)
466{
467  if (CPP_OPTION (pfile, deps.style) != DEPS_NONE)
468    {
469      if (!pfile->deps)
470	pfile->deps = deps_init ();
471
472      /* Set the default target (if there is none already).  */
473      deps_add_default_target (pfile->deps, fname);
474    }
475
476  pfile->main_file
477    = _cpp_find_file (pfile, fname, &pfile->no_search_path, false, 0);
478  if (_cpp_find_failed (pfile->main_file))
479    return NULL;
480
481  _cpp_stack_file (pfile, pfile->main_file, false);
482
483  /* For foo.i, read the original filename foo.c now, for the benefit
484     of the front ends.  */
485  if (CPP_OPTION (pfile, preprocessed))
486    {
487      read_original_filename (pfile);
488      fname = pfile->line_table->maps[pfile->line_table->used-1].to_file;
489    }
490  return fname;
491}
492
493/* For preprocessed files, if the first tokens are of the form # NUM.
494   handle the directive so we know the original file name.  This will
495   generate file_change callbacks, which the front ends must handle
496   appropriately given their state of initialization.  */
497static void
498read_original_filename (cpp_reader *pfile)
499{
500  const cpp_token *token, *token1;
501
502  /* Lex ahead; if the first tokens are of the form # NUM, then
503     process the directive, otherwise back up.  */
504  token = _cpp_lex_direct (pfile);
505  if (token->type == CPP_HASH)
506    {
507      pfile->state.in_directive = 1;
508      token1 = _cpp_lex_direct (pfile);
509      _cpp_backup_tokens (pfile, 1);
510      pfile->state.in_directive = 0;
511
512      /* If it's a #line directive, handle it.  */
513      if (token1->type == CPP_NUMBER)
514	{
515	  _cpp_handle_directive (pfile, token->flags & PREV_WHITE);
516	  read_original_directory (pfile);
517	  return;
518	}
519    }
520
521  /* Backup as if nothing happened.  */
522  _cpp_backup_tokens (pfile, 1);
523}
524
525/* For preprocessed files, if the tokens following the first filename
526   line is of the form # <line> "/path/name//", handle the
527   directive so we know the original current directory.  */
528static void
529read_original_directory (cpp_reader *pfile)
530{
531  const cpp_token *hash, *token;
532
533  /* Lex ahead; if the first tokens are of the form # NUM, then
534     process the directive, otherwise back up.  */
535  hash = _cpp_lex_direct (pfile);
536  if (hash->type != CPP_HASH)
537    {
538      _cpp_backup_tokens (pfile, 1);
539      return;
540    }
541
542  token = _cpp_lex_direct (pfile);
543
544  if (token->type != CPP_NUMBER)
545    {
546      _cpp_backup_tokens (pfile, 2);
547      return;
548    }
549
550  token = _cpp_lex_direct (pfile);
551
552  if (token->type != CPP_STRING
553      || ! (token->val.str.len >= 5
554	    && token->val.str.text[token->val.str.len-2] == '/'
555	    && token->val.str.text[token->val.str.len-3] == '/'))
556    {
557      _cpp_backup_tokens (pfile, 3);
558      return;
559    }
560
561  if (pfile->cb.dir_change)
562    {
563      char *debugdir = (char *) alloca (token->val.str.len - 3);
564
565      memcpy (debugdir, (const char *) token->val.str.text + 1,
566	      token->val.str.len - 4);
567      debugdir[token->val.str.len - 4] = '\0';
568
569      pfile->cb.dir_change (pfile, debugdir);
570    }
571}
572
573/* This is called at the end of preprocessing.  It pops the last
574   buffer and writes dependency output, and returns the number of
575   errors.
576
577   Maybe it should also reset state, such that you could call
578   cpp_start_read with a new filename to restart processing.  */
579int
580cpp_finish (cpp_reader *pfile, FILE *deps_stream)
581{
582  /* Warn about unused macros before popping the final buffer.  */
583  if (CPP_OPTION (pfile, warn_unused_macros))
584    cpp_forall_identifiers (pfile, _cpp_warn_if_unused_macro, NULL);
585
586  /* lex.c leaves the final buffer on the stack.  This it so that
587     it returns an unending stream of CPP_EOFs to the client.  If we
588     popped the buffer, we'd dereference a NULL buffer pointer and
589     segfault.  It's nice to allow the client to do worry-free excess
590     cpp_get_token calls.  */
591  while (pfile->buffer)
592    _cpp_pop_buffer (pfile);
593
594  /* Don't write the deps file if there are errors.  */
595  if (CPP_OPTION (pfile, deps.style) != DEPS_NONE
596      && deps_stream && pfile->errors == 0)
597    {
598      deps_write (pfile->deps, deps_stream, 72);
599
600      if (CPP_OPTION (pfile, deps.phony_targets))
601	deps_phony_targets (pfile->deps, deps_stream);
602    }
603
604  /* Report on headers that could use multiple include guards.  */
605  if (CPP_OPTION (pfile, print_include_names))
606    _cpp_report_missing_guards (pfile);
607
608  return pfile->errors;
609}
610
611static void
612post_options (cpp_reader *pfile)
613{
614  /* -Wtraditional is not useful in C++ mode.  */
615  if (CPP_OPTION (pfile, cplusplus))
616    CPP_OPTION (pfile, warn_traditional) = 0;
617
618  /* Permanently disable macro expansion if we are rescanning
619     preprocessed text.  Read preprocesed source in ISO mode.  */
620  if (CPP_OPTION (pfile, preprocessed))
621    {
622      pfile->state.prevent_expansion = 1;
623      CPP_OPTION (pfile, traditional) = 0;
624    }
625
626  if (CPP_OPTION (pfile, warn_trigraphs) == 2)
627    CPP_OPTION (pfile, warn_trigraphs) = !CPP_OPTION (pfile, trigraphs);
628
629  if (CPP_OPTION (pfile, traditional))
630    {
631      CPP_OPTION (pfile, cplusplus_comments) = 0;
632
633      /* Traditional CPP does not accurately track column information.  */
634      CPP_OPTION (pfile, show_column) = 0;
635      CPP_OPTION (pfile, trigraphs) = 0;
636      CPP_OPTION (pfile, warn_trigraphs) = 0;
637    }
638}
639