1 /* Definitions for CPP library.
2 Copyright (C) 1995, 96-99, 2000 Free Software Foundation, Inc.
3 Written by Per Bothner, 1994-95.
5 This program is free software; you can redistribute it and/or modify it
6 under the terms of the GNU General Public License as published by the
7 Free Software Foundation; either version 2, or (at your option) any
10 This program is distributed in the hope that it will be useful,
11 but WITHOUT ANY WARRANTY; without even the implied warranty of
12 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 GNU General Public License for more details.
15 You should have received a copy of the GNU General Public License
16 along with this program; if not, write to the Free Software
17 Foundation, 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
19 In other words, you are welcome to use, share and improve this program.
20 You are forbidden to forbid anyone else to use, share and improve
21 what you give them. Help stamp out software-hoarding! */
22 #ifndef __GCC_CPPLIB__
23 #define __GCC_CPPLIB__
25 #include <sys/types.h>
31 typedef struct cpp_reader cpp_reader;
32 typedef struct cpp_buffer cpp_buffer;
33 typedef struct cpp_options cpp_options;
34 typedef struct cpp_printer cpp_printer;
35 typedef struct cpp_token cpp_token;
36 typedef struct cpp_toklist cpp_toklist;
37 typedef struct cpp_name cpp_name;
39 /* The first two groups, apart from '=', can appear in preprocessor
40 expressions. This allows a lookup table to be implemented in
43 The first group, to CPP_LAST_EQ, can be immediately followed by an
44 '='. The lexer needs operators ending in '=', like ">>=", to be in
45 the same order as their counterparts without the '=', like ">>". */
47 /* Positions in the table. */
48 #define CPP_LAST_EQ CPP_LSHIFT
49 #define CPP_FIRST_DIGRAPH CPP_HASH
54 T(CPP_GREATER, ">") /* compare */ \
56 T(CPP_PLUS, "+") /* math */ \
61 T(CPP_AND, "&") /* bit ops */ \
68 T(CPP_AND_AND, "&&") /* logical */ \
72 T(CPP_COMMA, ",") /* grouping */ \
73 T(CPP_OPEN_PAREN, "(") \
74 T(CPP_CLOSE_PAREN, ")") \
75 T(CPP_EQ_EQ, "==") /* compare */ \
77 T(CPP_GREATER_EQ, ">=") \
78 T(CPP_LESS_EQ, "<=") \
80 T(CPP_PLUS_EQ, "+=") /* math */ \
81 T(CPP_MINUS_EQ, "-=") \
82 T(CPP_MULT_EQ, "*=") \
85 T(CPP_AND_EQ, "&=") /* bit ops */ \
88 T(CPP_COMPL_EQ, "~=") \
89 T(CPP_RSHIFT_EQ, ">>=") \
90 T(CPP_LSHIFT_EQ, "<<=") \
91 /* Digraphs together, beginning with CPP_FIRST_DIGRAPH. */ \
92 T(CPP_HASH, "#") /* digraphs */ \
94 T(CPP_OPEN_SQUARE, "[") \
95 T(CPP_CLOSE_SQUARE, "]") \
96 T(CPP_OPEN_BRACE, "{") \
97 T(CPP_CLOSE_BRACE, "}") \
98 /* The remainder of the punctuation. Order is not significant. */ \
99 T(CPP_SEMICOLON, ";") /* structure */ \
100 T(CPP_ELLIPSIS, "...") \
101 T(CPP_BACKSLASH, "\\") \
102 T(CPP_PLUS_PLUS, "++") /* increment */ \
103 T(CPP_MINUS_MINUS, "--") \
104 T(CPP_DEREF, "->") /* accessors */ \
107 T(CPP_DEREF_STAR, "->*") \
108 T(CPP_DOT_STAR, ".*") \
109 T(CPP_MIN, "<?") /* extension */ \
111 H(CPP_OTHER, spell_other) /* stray punctuation */ \
113 H(CPP_NAME, spell_name) /* word */ \
114 N(CPP_INT, 0) /* 23 */ \
115 N(CPP_FLOAT, 0) /* 3.14159 */ \
116 H(CPP_NUMBER, spell_name) /* 34_be+ta */ \
117 H(CPP_CHAR, spell_char) /* 'char' */ \
118 H(CPP_WCHAR, spell_char) /* L'char' */ \
119 H(CPP_STRING, spell_string) /* "string" */ \
120 H(CPP_WSTRING, spell_string) /* L"string" */ \
122 H(CPP_C_COMMENT, spell_comment) /* Only if output comments. */ \
123 H(CPP_CPP_COMMENT, spell_comment) /* Only if output comments. */ \
124 H(CPP_CHILL_COMMENT, spell_comment) /* Only if output comments. */ \
125 N(CPP_MACRO_ARG, 0) /* Macro argument. */ \
126 N(CPP_SUBLIST, 0) /* Sublist. */ \
127 E(CPP_VSPACE, "\n") /* End of line. */ \
128 N(CPP_EOF, 0) /* End of file. */ \
129 N(CPP_HEADER_NAME, 0) /* <stdio.h> in #include */ \
130 N(CPP_ASSERTION, 0) /* (...) in #assert */ \
132 /* Obsolete - will be removed when no code uses them still. */ \
133 H(CPP_COMMENT, 0) /* Only if output comments. */ \
134 N(CPP_HSPACE, 0) /* Horizontal white space. */ \
135 N(CPP_POP, 0) /* End of buffer. */ \
136 N(CPP_DIRECTIVE, 0) /* #define and the like */ \
137 N(CPP_MACRO, 0) /* Like a NAME, but expanded. */
153 /* Payload of a NAME, NUMBER, FLOAT, STRING, or COMMENT token. */
157 unsigned int offset; /* from list->namebuf */
160 #define TOK_NAME(list, token) ((list)->namebuf + (token)->val.name.offset)
162 /* Flags for the cpp_token structure. */
163 #define PREV_WHITESPACE 1 /* If whitespace before this token. */
164 #define DIGRAPH 2 /* If it was a digraph. */
165 #define UNSIGNED_INT 4 /* If int preprocessing token unsigned. */
167 /* A preprocessing token.
168 This has been carefully packed and should occupy 16 bytes on
169 both 32- and 64-bit hosts. */
172 unsigned short col; /* starting column of this token */
173 #ifdef ENUM_BITFIELDS_ARE_UNSIGNED
174 enum cpp_ttype type : CHAR_BIT; /* node type */
178 unsigned char flags; /* flags - see above */
179 unsigned int aux; /* CPP_OTHER character. Hash of a
180 NAME, or something - see uses
184 struct cpp_name name; /* a string */
185 HOST_WIDEST_INT integer; /* an integer */
189 /* Directive flags. */
190 #define SYNTAX_INCLUDE (1 << 8)
191 #define SYNTAX_ASSERT (1 << 9)
193 typedef int (*directive_handler) PARAMS ((cpp_reader *));
194 typedef int (*parse_cleanup_t) PARAMS ((cpp_buffer *, cpp_reader *));
198 cpp_token *tokens; /* actual tokens as an array */
199 unsigned int tokens_used; /* tokens used */
200 unsigned int tokens_cap; /* tokens allocated */
202 unsigned char *namebuf; /* names buffer */
203 unsigned int name_used; /* _bytes_ used */
204 unsigned int name_cap; /* _bytes_ allocated */
206 unsigned int line; /* starting line number */
208 /* Comment copying. */
209 cpp_token *comments; /* comment tokens. */
210 unsigned int comments_used; /* comment tokens used. */
211 unsigned int comments_cap; /* comment token capacity. */
213 /* Only used if tokens[0].type == CPP_DIRECTIVE. This is the
214 handler to call after lexing the rest of this line. The flags
215 indicate whether the rest of the line gets special treatment
216 during lexing (#include, #if, #assert, #unassert). */
217 directive_handler dir_handler;
218 unsigned short dir_flags;
223 const unsigned char *cur; /* current position */
224 const unsigned char *rlimit; /* end of valid data */
225 const unsigned char *buf; /* entire buffer */
226 const unsigned char *line_base; /* start of current line */
227 const unsigned char *mark; /* Saved position for lengthy backtrack. */
229 struct cpp_buffer *prev;
231 /* Filename specified with #line command. */
232 const char *nominal_fname;
233 /* Actual directory of this file, used only for "" includes */
234 struct file_name_list *actual_dir;
236 /* Pointer into the include hash table. Used for include_next and
237 to record control macros. */
240 parse_cleanup_t cleanup;
242 /* If the buffer is the expansion of a macro, this points to the
243 macro's hash table entry. */
244 struct hashnode *macro;
246 /* Value of if_stack at start of this file.
247 Used to prohibit unmatched #endif (etc) in an include file. */
248 struct if_stack *if_stack;
250 /* Line number at line_base (above). */
253 /* True if this is a header file included using <FILENAME>. */
254 char system_header_p;
256 /* True if end-of-file has already been hit once in this buffer. */
259 /* True if buffer contains escape sequences.
260 Currently there are two kinds:
261 "\r-" means following identifier should not be macro-expanded.
262 "\r " means a token-separator. This turns into " " in final output
263 if not stringizing and needed to separate tokens; otherwise nothing.
264 Any other two-character sequence beginning with \r is an error.
266 If this is NOT set, then \r is a one-character escape meaning backslash
267 newline. This is guaranteed not to occur in the middle of a token.
268 The two interpretations of \r do not conflict, because the two-character
269 escapes are used only in macro buffers, and backslash-newline is removed
270 from macro expansion text in collect_expansion and/or macarg. */
273 /* Used by the C++ frontend to implement redirected input (such as for
274 default argument and/or template parsing). */
277 /* True if we have already warned about C++ comments in this file.
278 The warning happens only for C89 extended mode with -pedantic on,
279 or for -Wtraditional, and only once per file (otherwise it would
280 be far too noisy). */
281 char warned_cplusplus_comments;
283 /* True if this buffer's data is mmapped. */
287 struct file_name_map_list;
290 /* Maximum nesting of cpp_buffers. We use a static limit, partly for
291 efficiency, and partly to limit runaway recursion. */
292 #define CPP_STACK_MAX 200
294 /* Values for opts.dump_macros.
295 dump_only means inhibit output of the preprocessed text
296 and instead output the definitions of all user-defined
297 macros in a form suitable for use as input to cpp.
298 dump_names means pass #define and the macro name through to output.
299 dump_definitions means pass the whole definition (plus #define) through
301 enum { dump_none = 0, dump_only, dump_names, dump_definitions };
303 /* This structure is nested inside struct cpp_reader, and
304 carries all the options visible to the command line. */
307 /* Name of input and output files. */
308 const char *in_fname;
309 const char *out_fname;
311 /* Pending options - -D, -U, -A, -I, -ixxx. */
312 struct cpp_pending *pending;
314 /* File name which deps are being written to. This is 0 if deps are
315 being written to stdout. */
316 const char *deps_file;
318 /* Target-name to write with the dependency information. */
321 /* Search paths for include files. */
322 struct file_name_list *quote_include; /* First dir to search for "file" */
323 struct file_name_list *bracket_include;/* First dir to search for <file> */
325 /* Map between header names and file names, used only on DOS where
326 file names are limited in length. */
327 struct file_name_map_list *map_list;
329 /* Directory prefix that should replace `/usr/lib/gcc-lib/TARGET/VERSION'
330 in the standard include file directories. */
331 const char *include_prefix;
332 unsigned int include_prefix_len;
334 /* Non-0 means -v, so print the full set of include dirs. */
335 unsigned char verbose;
337 /* Nonzero means use extra default include directories for C++. */
338 unsigned char cplusplus;
340 /* Nonzero means handle cplusplus style comments */
341 unsigned char cplusplus_comments;
343 /* Nonzero means handle #import, for objective C. */
346 /* Nonzero means this is an assembly file, so ignore unrecognized
347 directives and the "# 33" form of #line, both of which are
348 probably comments. Also, permit unbalanced ' strings (again,
349 likely to be in comments). */
350 unsigned char lang_asm;
352 /* Nonzero means this is Fortran, and we don't know where the
353 comments are, so permit unbalanced ' strings. Unlike lang_asm,
354 this does not ignore unrecognized directives. */
355 unsigned char lang_fortran;
357 /* Nonzero means handle CHILL comment syntax and output CHILL string
358 delimiters for __DATE__ etc. */
361 /* Nonzero means don't copy comments into the output file. */
362 unsigned char discard_comments;
364 /* Nonzero means process the ANSI trigraph sequences. */
365 unsigned char trigraphs;
367 /* Nonzero means print the names of included files rather than the
368 preprocessed output. 1 means just the #include "...", 2 means
369 #include <...> as well. */
370 unsigned char print_deps;
372 /* Nonzero if missing .h files in -M output are assumed to be
373 generated files and not errors. */
374 unsigned char print_deps_missing_files;
376 /* If true, fopen (deps_file, "a") else fopen (deps_file, "w"). */
377 unsigned char print_deps_append;
379 /* Nonzero means print names of header files (-H). */
380 unsigned char print_include_names;
382 /* Nonzero means cpp_pedwarn causes a hard error. */
383 unsigned char pedantic_errors;
385 /* Nonzero means don't print warning messages. */
386 unsigned char inhibit_warnings;
388 /* Nonzero means don't print error messages. Has no option to
389 select it, but can be set by a user of cpplib (e.g. fix-header). */
390 unsigned char inhibit_errors;
392 /* Nonzero means warn if slash-star appears in a comment. */
393 unsigned char warn_comments;
395 /* Nonzero means warn if there are any trigraphs. */
396 unsigned char warn_trigraphs;
398 /* Nonzero means warn if #import is used. */
399 unsigned char warn_import;
401 /* Nonzero means warn if a macro argument is (or would be)
402 stringified with -traditional, and warn about directives
403 with the # indented from the beginning of the line. */
404 unsigned char warn_traditional;
406 /* Nonzero means turn warnings into errors. */
407 unsigned char warnings_are_errors;
409 /* Nonzero causes output not to be done, but directives such as
410 #define that have side effects are still obeyed. */
411 unsigned char no_output;
413 /* Nonzero means we should look for header.gcc files that remap file
417 /* Nonzero means don't output line number information. */
418 unsigned char no_line_commands;
420 /* Nonzero means -I- has been seen, so don't look for #include "foo"
421 the source-file directory. */
422 unsigned char ignore_srcdir;
424 /* Zero means dollar signs are punctuation. */
425 unsigned char dollars_in_ident;
427 /* Nonzero means try to imitate old fashioned non-ANSI preprocessor. */
428 unsigned char traditional;
430 /* Nonzero means warn if undefined identifiers are evaluated in an #if. */
431 unsigned char warn_undef;
433 /* Nonzero for the 1989 C Standard, including corrigenda and amendments. */
436 /* Nonzero for the 1999 C Standard, including corrigenda and amendments. */
439 /* Nonzero means give all the error messages the ANSI standard requires. */
440 unsigned char pedantic;
442 /* Nonzero means we're looking at already preprocessed code, so don't
443 bother trying to do macro expansion and whatnot. */
444 unsigned char preprocessed;
446 /* Nonzero disables all the standard directories for headers. */
447 unsigned char no_standard_includes;
449 /* Nonzero disables the C++-specific standard directories for headers. */
450 unsigned char no_standard_cplusplus_includes;
452 /* Nonzero means dump macros in some fashion - see above. */
453 unsigned char dump_macros;
455 /* Nonzero means pass all #define and #undef directives which we
456 actually process through to the output stream. This feature is
457 used primarily to allow cc1 to record the #defines and #undefs
458 for the sake of debuggers which understand about preprocessor
459 macros, but it may also be useful with -E to figure out how
460 symbols are defined, and where they are defined. */
461 unsigned char debug_output;
463 /* Nonzero means pass #include lines through to the output. */
464 unsigned char dump_includes;
466 /* Print column number in error messages. */
467 unsigned char show_column;
471 /* A cpp_reader encapsulates the "state" of a pre-processor run.
472 Applying cpp_get_token repeatedly yields a stream of pre-processor
473 tokens. Usually, there is only one cpp_reader object active. */
477 /* Top of buffer stack. */
480 /* Token list used by get_directive_token. */
481 cpp_toklist directbuf;
483 /* A buffer used for both for cpp_get_token's output, and also internally. */
484 unsigned char *token_buffer;
485 /* Allocated size of token_buffer. CPP_RESERVE allocates space. */
486 unsigned int token_buffer_size;
487 /* End of the written part of token_buffer. */
488 unsigned char *limit;
490 /* Error counter for exit code */
493 /* Line where a newline was first seen in a string constant. */
494 unsigned int multiline_string_line;
496 /* Current depth in #include directives that use <...>. */
497 unsigned int system_include_depth;
499 /* Current depth of buffer stack. */
500 unsigned int buffer_stack_depth;
502 /* Hash table of macros and assertions. See cpphash.c */
503 struct htab *hashtab;
505 /* Hash table of other included files. See cppfiles.c */
506 struct htab *all_include_files;
508 /* Chain of `actual directory' file_name_list entries,
510 struct file_name_list *actual_dirs;
512 /* Current maximum length of directory names in the search path
513 for include files. (Altered as we get more of them.) */
514 unsigned int max_include_len;
516 struct if_stack *if_stack;
517 const unsigned char *potential_control_macro;
519 /* Buffer of -M output. */
522 /* A buffer used only by read_and_prescan (in cppfiles.c), which is
523 allocated once per cpp_reader object to keep it off the stack. */
524 unsigned char *input_buffer;
525 size_t input_buffer_len;
527 /* User visible options. */
528 struct cpp_options opts;
530 /* Nonzero means we have printed (while error reporting) a list of
531 containing files that matches the current status. */
532 unsigned char input_stack_listing_current;
534 /* If non-zero, macros are not expanded. */
535 unsigned char no_macro_expand;
537 /* If non-zero, directives cause a hard error. Used when parsing
539 unsigned char no_directives;
541 /* We're printed a warning recommending against using #import. */
542 unsigned char import_warning;
544 /* If true, characters between '<' and '>' are a single (string) token. */
545 unsigned char parsing_include_directive;
547 /* If true, # introduces an assertion (see do_assert) */
548 unsigned char parsing_if_directive;
550 /* If true, # and ## are the STRINGIZE and TOKPASTE operators */
551 unsigned char parsing_define_directive;
553 /* True if escape sequences (as described for has_escapes in
554 parse_buffer) should be emitted. */
555 unsigned char output_escapes;
557 /* 0: Have seen non-white-space on this line.
558 1: Only seen white space so far on this line.
559 2: Only seen white space so far in this file. */
560 unsigned char only_seen_white;
562 /* True after cpp_start_read completes. Used to inhibit some
563 warnings while parsing the command line. */
564 unsigned char done_initializing;
567 /* struct cpp_printer encapsulates state used to convert the stream of
568 tokens coming from cpp_get_token back into a text file. Not
569 everyone wants to do that, hence we separate the function. */
573 FILE *outf; /* stream to write to */
574 const char *last_fname; /* previous file name */
575 unsigned int last_bsd; /* did we just push? */
576 unsigned int lineno; /* line currently being written */
577 unsigned int written; /* low water mark in token buffer */
580 #define CPP_FATAL_LIMIT 1000
581 /* True if we have seen a "fatal" error. */
582 #define CPP_FATAL_ERRORS(READER) ((READER)->errors >= CPP_FATAL_LIMIT)
584 /* Macros for manipulating the token_buffer. */
586 /* Number of characters currently in PFILE's output buffer. */
587 #define CPP_WRITTEN(PFILE) ((size_t)((PFILE)->limit - (PFILE)->token_buffer))
588 #define CPP_PWRITTEN(PFILE) ((PFILE)->limit)
589 #define CPP_ADJUST_WRITTEN(PFILE,DELTA) ((PFILE)->limit += (DELTA))
590 #define CPP_SET_WRITTEN(PFILE,N) ((PFILE)->limit = (PFILE)->token_buffer + (N))
592 #define CPP_OPTION(PFILE, OPTION) ((PFILE)->opts.OPTION)
593 #define CPP_BUFFER(PFILE) ((PFILE)->buffer)
594 #define CPP_BUF_LINE(BUF) ((BUF)->lineno)
595 #define CPP_BUF_COL(BUF) ((BUF)->cur - (BUF)->line_base)
597 /* Name under which this program was invoked. */
598 extern const char *progname;
600 extern void _cpp_lex_file PARAMS((cpp_reader *));
601 extern int cpp_handle_options PARAMS ((cpp_reader *, int, char **));
602 extern enum cpp_ttype cpp_get_token PARAMS ((cpp_reader *));
603 extern enum cpp_ttype cpp_get_non_space_token PARAMS ((cpp_reader *));
605 extern void cpp_reader_init PARAMS ((cpp_reader *));
606 extern cpp_printer *cpp_printer_init PARAMS ((cpp_reader *, cpp_printer *));
607 extern int cpp_start_read PARAMS ((cpp_reader *, cpp_printer *, const char *));
608 extern void cpp_output_tokens PARAMS ((cpp_reader *, cpp_printer *));
609 extern void cpp_output_list PARAMS ((cpp_reader *, cpp_printer *,
610 const cpp_toklist *));
611 extern void cpp_finish PARAMS ((cpp_reader *, cpp_printer *));
612 extern void cpp_cleanup PARAMS ((cpp_reader *));
614 extern cpp_buffer *cpp_file_buffer PARAMS((cpp_reader *));
615 extern void cpp_define PARAMS ((cpp_reader *, const char *));
616 extern void cpp_assert PARAMS ((cpp_reader *, const char *));
617 extern void cpp_undef PARAMS ((cpp_reader *, const char *));
618 extern void cpp_unassert PARAMS ((cpp_reader *, const char *));
620 extern void cpp_free_token_list PARAMS ((cpp_toklist *));
622 /* N.B. The error-message-printer prototypes have not been nicely
623 formatted because exgettext needs to see 'msgid' on the same line
624 as the name of the function in order to work properly. Only the
625 string argument gets a name in an effort to keep the lines from
626 getting ridiculously oversized. */
628 extern void cpp_ice PARAMS ((cpp_reader *, const char *msgid, ...))
630 extern void cpp_fatal PARAMS ((cpp_reader *, const char *msgid, ...))
632 extern void cpp_error PARAMS ((cpp_reader *, const char *msgid, ...))
634 extern void cpp_warning PARAMS ((cpp_reader *, const char *msgid, ...))
636 extern void cpp_pedwarn PARAMS ((cpp_reader *, const char *msgid, ...))
638 extern void cpp_notice PARAMS ((cpp_reader *, const char *msgid, ...))
640 extern void cpp_error_with_line PARAMS ((cpp_reader *, int, int, const char *msgid, ...))
642 extern void cpp_warning_with_line PARAMS ((cpp_reader *, int, int, const char *msgid, ...))
644 extern void cpp_pedwarn_with_line PARAMS ((cpp_reader *, int, int, const char *msgid, ...))
646 extern void cpp_pedwarn_with_file_and_line PARAMS ((cpp_reader *, const char *, int, int, const char *msgid, ...))
648 extern void cpp_error_from_errno PARAMS ((cpp_reader *, const char *));
649 extern void cpp_notice_from_errno PARAMS ((cpp_reader *, const char *));
652 extern cpp_buffer *cpp_push_buffer PARAMS ((cpp_reader *,
653 const unsigned char *, long));
654 extern cpp_buffer *cpp_pop_buffer PARAMS ((cpp_reader *));
655 extern void cpp_scan_buffer PARAMS ((cpp_reader *, cpp_printer *));
656 extern void cpp_scan_buffer_nooutput PARAMS ((cpp_reader *));
661 extern int cpp_defined PARAMS ((cpp_reader *,
662 const unsigned char *, int));
665 extern int cpp_included PARAMS ((cpp_reader *, const char *));
666 extern int cpp_read_file PARAMS ((cpp_reader *, const char *));
671 #endif /* __GCC_CPPLIB__ */