1 /* CPP Library - lexical analysis.
2 Copyright (C) 2000, 2001, 2002, 2003, 2004, 2005 Free Software Foundation, Inc.
3 Contributed by Per Bothner, 1994-95.
4 Based on CCCP program by Paul Rubin, June 1986
5 Adapted to ANSI C, Richard Stallman, Jan 1987
6 Broken out to separate file, Zack Weinberg, Mar 2000
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. */
37 enum spell_type category;
38 const unsigned char *name;
41 static const unsigned char *const digraph_spellings[] =
42 { U"%:", U"%:%:", U"<:", U":>", U"<%", U"%>" };
44 #define OP(e, s) { SPELL_OPERATOR, U s },
45 #define TK(e, s) { SPELL_ ## s, U #e },
46 static const struct token_spelling token_spellings[N_TTYPES] = { TTYPE_TABLE };
50 #define TOKEN_SPELL(token) (token_spellings[(token)->type].category)
51 #define TOKEN_NAME(token) (token_spellings[(token)->type].name)
53 static void add_line_note (cpp_buffer *, const uchar *, unsigned int);
54 static int skip_line_comment (cpp_reader *);
55 static void skip_whitespace (cpp_reader *, cppchar_t);
56 static void lex_string (cpp_reader *, cpp_token *, const uchar *);
57 static void save_comment (cpp_reader *, cpp_token *, const uchar *, cppchar_t);
58 static void create_literal (cpp_reader *, cpp_token *, const uchar *,
59 unsigned int, enum cpp_ttype);
60 static bool warn_in_comment (cpp_reader *, _cpp_line_note *);
61 static int name_p (cpp_reader *, const cpp_string *);
62 static tokenrun *next_tokenrun (tokenrun *);
64 static _cpp_buff *new_buff (size_t);
69 Compares, the token TOKEN to the NUL-terminated string STRING.
70 TOKEN must be a CPP_NAME. Returns 1 for equal, 0 for unequal. */
72 cpp_ideq (const cpp_token *token, const char *string)
74 if (token->type != CPP_NAME)
77 return !ustrcmp (NODE_NAME (token->val.node), (const uchar *) string);
80 /* Record a note TYPE at byte POS into the current cleaned logical
83 add_line_note (cpp_buffer *buffer, const uchar *pos, unsigned int type)
85 if (buffer->notes_used == buffer->notes_cap)
87 buffer->notes_cap = buffer->notes_cap * 2 + 200;
88 buffer->notes = XRESIZEVEC (_cpp_line_note, buffer->notes,
92 buffer->notes[buffer->notes_used].pos = pos;
93 buffer->notes[buffer->notes_used].type = type;
97 /* Returns with a logical line that contains no escaped newlines or
98 trigraphs. This is a time-critical inner loop. */
100 _cpp_clean_line (cpp_reader *pfile)
106 buffer = pfile->buffer;
107 buffer->cur_note = buffer->notes_used = 0;
108 buffer->cur = buffer->line_base = buffer->next_line;
109 buffer->need_line = false;
110 s = buffer->next_line - 1;
112 if (!buffer->from_stage3)
114 const uchar *pbackslash = NULL;
116 /* Short circuit for the common case of an un-escaped line with
117 no trigraphs. The primary win here is by not writing any
118 data back to memory until we have to. */
122 if (__builtin_expect (c == '\n', false)
123 || __builtin_expect (c == '\r', false))
127 if (__builtin_expect (s == buffer->rlimit, false))
130 /* DOS line ending? */
131 if (__builtin_expect (c == '\r', false)
135 if (s == buffer->rlimit)
139 if (__builtin_expect (pbackslash == NULL, true))
142 /* Check for escaped newline. */
144 while (is_nvspace (p[-1]))
146 if (p - 1 != pbackslash)
149 /* Have an escaped newline; process it and proceed to
151 add_line_note (buffer, p - 1, p != d ? ' ' : '\\');
153 buffer->next_line = p - 1;
156 if (__builtin_expect (c == '\\', false))
158 else if (__builtin_expect (c == '?', false)
159 && __builtin_expect (s[1] == '?', false)
160 && _cpp_trigraph_map[s[2]])
162 /* Have a trigraph. We may or may not have to convert
163 it. Add a line note regardless, for -Wtrigraphs. */
164 add_line_note (buffer, s, s[2]);
165 if (CPP_OPTION (pfile, trigraphs))
167 /* We do, and that means we have to switch to the
170 *d = _cpp_trigraph_map[s[2]];
183 if (c == '\n' || c == '\r')
185 /* Handle DOS line endings. */
186 if (c == '\r' && s != buffer->rlimit && s[1] == '\n')
188 if (s == buffer->rlimit)
193 while (p != buffer->next_line && is_nvspace (p[-1]))
195 if (p == buffer->next_line || p[-1] != '\\')
198 add_line_note (buffer, p - 1, p != d ? ' ': '\\');
200 buffer->next_line = p - 1;
202 else if (c == '?' && s[1] == '?' && _cpp_trigraph_map[s[2]])
204 /* Add a note regardless, for the benefit of -Wtrigraphs. */
205 add_line_note (buffer, d, s[2]);
206 if (CPP_OPTION (pfile, trigraphs))
208 *d = _cpp_trigraph_map[s[2]];
218 while (*s != '\n' && *s != '\r');
221 /* Handle DOS line endings. */
222 if (*s == '\r' && s != buffer->rlimit && s[1] == '\n')
228 /* A sentinel note that should never be processed. */
229 add_line_note (buffer, d + 1, '\n');
230 buffer->next_line = s + 1;
233 /* Return true if the trigraph indicated by NOTE should be warned
234 about in a comment. */
236 warn_in_comment (cpp_reader *pfile, _cpp_line_note *note)
240 /* Within comments we don't warn about trigraphs, unless the
241 trigraph forms an escaped newline, as that may change
243 if (note->type != '/')
246 /* If -trigraphs, then this was an escaped newline iff the next note
248 if (CPP_OPTION (pfile, trigraphs))
249 return note[1].pos == note->pos;
251 /* Otherwise, see if this forms an escaped newline. */
253 while (is_nvspace (*p))
256 /* There might have been escaped newlines between the trigraph and the
257 newline we found. Hence the position test. */
258 return (*p == '\n' && p < note[1].pos);
261 /* Process the notes created by add_line_note as far as the current
264 _cpp_process_line_notes (cpp_reader *pfile, int in_comment)
266 cpp_buffer *buffer = pfile->buffer;
270 _cpp_line_note *note = &buffer->notes[buffer->cur_note];
273 if (note->pos > buffer->cur)
277 col = CPP_BUF_COLUMN (buffer, note->pos + 1);
279 if (note->type == '\\' || note->type == ' ')
281 if (note->type == ' ' && !in_comment)
282 cpp_error_with_line (pfile, CPP_DL_WARNING, pfile->line_table->highest_line, col,
283 "backslash and newline separated by space");
285 if (buffer->next_line > buffer->rlimit)
287 cpp_error_with_line (pfile, CPP_DL_PEDWARN, pfile->line_table->highest_line, col,
288 "backslash-newline at end of file");
289 /* Prevent "no newline at end of file" warning. */
290 buffer->next_line = buffer->rlimit;
293 buffer->line_base = note->pos;
294 CPP_INCREMENT_LINE (pfile, 0);
296 else if (_cpp_trigraph_map[note->type])
298 if (CPP_OPTION (pfile, warn_trigraphs)
299 && (!in_comment || warn_in_comment (pfile, note)))
301 if (CPP_OPTION (pfile, trigraphs))
302 cpp_error_with_line (pfile, CPP_DL_WARNING, pfile->line_table->highest_line, col,
303 "trigraph ??%c converted to %c",
305 (int) _cpp_trigraph_map[note->type]);
309 (pfile, CPP_DL_WARNING, pfile->line_table->highest_line, col,
310 "trigraph ??%c ignored, use -trigraphs to enable",
320 /* Skip a C-style block comment. We find the end of the comment by
321 seeing if an asterisk is before every '/' we encounter. Returns
322 nonzero if comment terminated by EOF, zero otherwise.
324 Buffer->cur points to the initial asterisk of the comment. */
326 _cpp_skip_block_comment (cpp_reader *pfile)
328 cpp_buffer *buffer = pfile->buffer;
329 const uchar *cur = buffer->cur;
338 /* People like decorating comments with '*', so check for '/'
339 instead for efficiency. */
347 /* Warn about potential nested comments, but not if the '/'
348 comes immediately before the true comment delimiter.
349 Don't bother to get it right across escaped newlines. */
350 if (CPP_OPTION (pfile, warn_comments)
351 && cur[0] == '*' && cur[1] != '/')
354 cpp_error_with_line (pfile, CPP_DL_WARNING,
355 pfile->line_table->highest_line, CPP_BUF_COL (buffer),
356 "\"/*\" within comment");
362 buffer->cur = cur - 1;
363 _cpp_process_line_notes (pfile, true);
364 if (buffer->next_line >= buffer->rlimit)
366 _cpp_clean_line (pfile);
368 cols = buffer->next_line - buffer->line_base;
369 CPP_INCREMENT_LINE (pfile, cols);
376 _cpp_process_line_notes (pfile, true);
380 /* Skip a C++ line comment, leaving buffer->cur pointing to the
381 terminating newline. Handles escaped newlines. Returns nonzero
382 if a multiline comment. */
384 skip_line_comment (cpp_reader *pfile)
386 cpp_buffer *buffer = pfile->buffer;
387 unsigned int orig_line = pfile->line_table->highest_line;
389 while (*buffer->cur != '\n')
392 _cpp_process_line_notes (pfile, true);
393 return orig_line != pfile->line_table->highest_line;
396 /* Skips whitespace, saving the next non-whitespace character. */
398 skip_whitespace (cpp_reader *pfile, cppchar_t c)
400 cpp_buffer *buffer = pfile->buffer;
401 bool saw_NUL = false;
405 /* Horizontal space always OK. */
406 if (c == ' ' || c == '\t')
408 /* Just \f \v or \0 left. */
411 else if (pfile->state.in_directive && CPP_PEDANTIC (pfile))
412 cpp_error_with_line (pfile, CPP_DL_PEDWARN, pfile->line_table->highest_line,
413 CPP_BUF_COL (buffer),
414 "%s in preprocessing directive",
415 c == '\f' ? "form feed" : "vertical tab");
419 /* We only want non-vertical space, i.e. ' ' \t \f \v \0. */
420 while (is_nvspace (c));
423 cpp_error (pfile, CPP_DL_WARNING, "null character(s) ignored");
428 /* See if the characters of a number token are valid in a name (no
431 name_p (cpp_reader *pfile, const cpp_string *string)
435 for (i = 0; i < string->len; i++)
436 if (!is_idchar (string->text[i]))
442 /* After parsing an identifier or other sequence, produce a warning about
443 sequences not in NFC/NFKC. */
445 warn_about_normalization (cpp_reader *pfile,
446 const cpp_token *token,
447 const struct normalize_state *s)
449 if (CPP_OPTION (pfile, warn_normalize) < NORMALIZE_STATE_RESULT (s)
450 && !pfile->state.skipping)
452 /* Make sure that the token is printed using UCNs, even
453 if we'd otherwise happily print UTF-8. */
454 unsigned char *buf = XNEWVEC (unsigned char, cpp_token_len (token));
457 sz = cpp_spell_token (pfile, token, buf, false) - buf;
458 if (NORMALIZE_STATE_RESULT (s) == normalized_C)
459 cpp_error_with_line (pfile, CPP_DL_WARNING, token->src_loc, 0,
460 "`%.*s' is not in NFKC", (int) sz, buf);
462 cpp_error_with_line (pfile, CPP_DL_WARNING, token->src_loc, 0,
463 "`%.*s' is not in NFC", (int) sz, buf);
467 /* Returns TRUE if the sequence starting at buffer->cur is invalid in
468 an identifier. FIRST is TRUE if this starts an identifier. */
470 forms_identifier_p (cpp_reader *pfile, int first,
471 struct normalize_state *state)
473 cpp_buffer *buffer = pfile->buffer;
475 if (*buffer->cur == '$')
477 if (!CPP_OPTION (pfile, dollars_in_ident))
481 if (CPP_OPTION (pfile, warn_dollars) && !pfile->state.skipping)
483 CPP_OPTION (pfile, warn_dollars) = 0;
484 cpp_error (pfile, CPP_DL_PEDWARN, "'$' in identifier or number");
490 /* Is this a syntactically valid UCN? */
491 if (CPP_OPTION (pfile, extended_identifiers)
492 && *buffer->cur == '\\'
493 && (buffer->cur[1] == 'u' || buffer->cur[1] == 'U'))
496 if (_cpp_valid_ucn (pfile, &buffer->cur, buffer->rlimit, 1 + !first,
505 /* Lex an identifier starting at BUFFER->CUR - 1. */
506 static cpp_hashnode *
507 lex_identifier (cpp_reader *pfile, const uchar *base, bool starts_ucn,
508 struct normalize_state *nst)
510 cpp_hashnode *result;
513 unsigned int hash = HT_HASHSTEP (0, *base);
515 cur = pfile->buffer->cur;
517 while (ISIDNUM (*cur))
519 hash = HT_HASHSTEP (hash, *cur);
522 pfile->buffer->cur = cur;
523 if (starts_ucn || forms_identifier_p (pfile, false, nst))
525 /* Slower version for identifiers containing UCNs (or $). */
527 while (ISIDNUM (*pfile->buffer->cur))
529 pfile->buffer->cur++;
530 NORMALIZE_STATE_UPDATE_IDNUM (nst);
532 } while (forms_identifier_p (pfile, false, nst));
533 result = _cpp_interpret_identifier (pfile, base,
534 pfile->buffer->cur - base);
539 hash = HT_HASHFINISH (hash, len);
541 result = (cpp_hashnode *)
542 ht_lookup_with_hash (pfile->hash_table, base, len, hash, HT_ALLOC);
545 /* Rarely, identifiers require diagnostics when lexed. */
546 if (__builtin_expect ((result->flags & NODE_DIAGNOSTIC)
547 && !pfile->state.skipping, 0))
549 /* It is allowed to poison the same identifier twice. */
550 if ((result->flags & NODE_POISONED) && !pfile->state.poisoned_ok)
551 cpp_error (pfile, CPP_DL_ERROR, "attempt to use poisoned \"%s\"",
554 /* Constraint 6.10.3.5: __VA_ARGS__ should only appear in the
555 replacement list of a variadic macro. */
556 if (result == pfile->spec_nodes.n__VA_ARGS__
557 && !pfile->state.va_args_ok)
558 cpp_error (pfile, CPP_DL_PEDWARN,
559 "__VA_ARGS__ can only appear in the expansion"
560 " of a C99 variadic macro");
566 /* Lex a number to NUMBER starting at BUFFER->CUR - 1. */
568 lex_number (cpp_reader *pfile, cpp_string *number,
569 struct normalize_state *nst)
575 base = pfile->buffer->cur - 1;
578 cur = pfile->buffer->cur;
580 /* N.B. ISIDNUM does not include $. */
581 while (ISIDNUM (*cur) || *cur == '.' || VALID_SIGN (*cur, cur[-1]))
584 NORMALIZE_STATE_UPDATE_IDNUM (nst);
587 pfile->buffer->cur = cur;
589 while (forms_identifier_p (pfile, false, nst));
591 number->len = cur - base;
592 dest = _cpp_unaligned_alloc (pfile, number->len + 1);
593 memcpy (dest, base, number->len);
594 dest[number->len] = '\0';
598 /* Create a token of type TYPE with a literal spelling. */
600 create_literal (cpp_reader *pfile, cpp_token *token, const uchar *base,
601 unsigned int len, enum cpp_ttype type)
603 uchar *dest = _cpp_unaligned_alloc (pfile, len + 1);
605 memcpy (dest, base, len);
608 token->val.str.len = len;
609 token->val.str.text = dest;
612 /* Lexes a string, character constant, or angle-bracketed header file
613 name. The stored string contains the spelling, including opening
614 quote and leading any leading 'L'. It returns the type of the
615 literal, or CPP_OTHER if it was not properly terminated.
617 The spelling is NUL-terminated, but it is not guaranteed that this
618 is the first NUL since embedded NULs are preserved. */
620 lex_string (cpp_reader *pfile, cpp_token *token, const uchar *base)
622 bool saw_NUL = false;
624 cppchar_t terminator;
629 if (terminator == 'L')
631 if (terminator == '\"')
632 type = *base == 'L' ? CPP_WSTRING: CPP_STRING;
633 else if (terminator == '\'')
634 type = *base == 'L' ? CPP_WCHAR: CPP_CHAR;
636 terminator = '>', type = CPP_HEADER_NAME;
640 cppchar_t c = *cur++;
642 /* In #include-style directives, terminators are not escapable. */
643 if (c == '\\' && !pfile->state.angled_headers && *cur != '\n')
645 else if (c == terminator)
657 if (saw_NUL && !pfile->state.skipping)
658 cpp_error (pfile, CPP_DL_WARNING,
659 "null character(s) preserved in literal");
661 /* APPLE LOCAL begin #error with unmatched quotes 5607574 */
662 if (type == CPP_OTHER
663 && CPP_OPTION (pfile, lang) != CLK_ASM
664 && !pfile->state.in_diagnostic
665 && !pfile->state.skipping)
666 /* APPLE LOCAL end #error with unmatched quotes 5607574 */
667 cpp_error (pfile, CPP_DL_PEDWARN, "missing terminating %c character",
670 pfile->buffer->cur = cur;
671 create_literal (pfile, token, base, cur - base, type);
674 /* The stored comment includes the comment start and any terminator. */
676 save_comment (cpp_reader *pfile, cpp_token *token, const unsigned char *from,
679 unsigned char *buffer;
680 unsigned int len, clen;
682 len = pfile->buffer->cur - from + 1; /* + 1 for the initial '/'. */
684 /* C++ comments probably (not definitely) have moved past a new
685 line, which we don't want to save in the comment. */
686 if (is_vspace (pfile->buffer->cur[-1]))
689 /* If we are currently in a directive, then we need to store all
690 C++ comments as C comments internally, and so we need to
691 allocate a little extra space in that case.
693 Note that the only time we encounter a directive here is
694 when we are saving comments in a "#define". */
695 clen = (pfile->state.in_directive && type == '/') ? len + 2 : len;
697 buffer = _cpp_unaligned_alloc (pfile, clen);
699 token->type = CPP_COMMENT;
700 token->val.str.len = clen;
701 token->val.str.text = buffer;
704 memcpy (buffer + 1, from, len - 1);
706 /* Finish conversion to a C comment, if necessary. */
707 if (pfile->state.in_directive && type == '/')
710 buffer[clen - 2] = '*';
711 buffer[clen - 1] = '/';
715 /* Allocate COUNT tokens for RUN. */
717 _cpp_init_tokenrun (tokenrun *run, unsigned int count)
719 run->base = XNEWVEC (cpp_token, count);
720 run->limit = run->base + count;
724 /* Returns the next tokenrun, or creates one if there is none. */
726 next_tokenrun (tokenrun *run)
728 if (run->next == NULL)
730 run->next = XNEW (tokenrun);
731 run->next->prev = run;
732 _cpp_init_tokenrun (run->next, 250);
738 /* Allocate a single token that is invalidated at the same time as the
739 rest of the tokens on the line. Has its line and col set to the
740 same as the last lexed token, so that diagnostics appear in the
743 _cpp_temp_token (cpp_reader *pfile)
745 cpp_token *old, *result;
747 old = pfile->cur_token - 1;
748 if (pfile->cur_token == pfile->cur_run->limit)
750 pfile->cur_run = next_tokenrun (pfile->cur_run);
751 pfile->cur_token = pfile->cur_run->base;
754 result = pfile->cur_token++;
755 result->src_loc = old->src_loc;
759 /* Lex a token into RESULT (external interface). Takes care of issues
760 like directive handling, token lookahead, multiple include
761 optimization and skipping. */
763 _cpp_lex_token (cpp_reader *pfile)
769 if (pfile->cur_token == pfile->cur_run->limit)
771 pfile->cur_run = next_tokenrun (pfile->cur_run);
772 pfile->cur_token = pfile->cur_run->base;
774 /* We assume that the current token is somewhere in the current
776 if (pfile->cur_token < pfile->cur_run->base
777 || pfile->cur_token >= pfile->cur_run->limit)
780 if (pfile->lookaheads)
783 result = pfile->cur_token++;
786 result = _cpp_lex_direct (pfile);
788 if (result->flags & BOL)
790 /* Is this a directive. If _cpp_handle_directive returns
791 false, it is an assembler #. */
792 if (result->type == CPP_HASH
793 /* 6.10.3 p 11: Directives in a list of macro arguments
794 gives undefined behavior. This implementation
795 handles the directive as normal. */
796 && pfile->state.parsing_args != 1)
798 if (_cpp_handle_directive (pfile, result->flags & PREV_WHITE))
800 if (pfile->directive_result.type == CPP_PADDING)
802 result = &pfile->directive_result;
805 else if (pfile->state.in_deferred_pragma)
806 result = &pfile->directive_result;
808 if (pfile->cb.line_change && !pfile->state.skipping)
809 pfile->cb.line_change (pfile, result, pfile->state.parsing_args);
812 /* We don't skip tokens in directives. */
813 if (pfile->state.in_directive || pfile->state.in_deferred_pragma)
816 /* Outside a directive, invalidate controlling macros. At file
817 EOF, _cpp_lex_direct takes care of popping the buffer, so we never
818 get here and MI optimization works. */
819 pfile->mi_valid = false;
821 if (!pfile->state.skipping || result->type == CPP_EOF)
828 /* Returns true if a fresh line has been loaded. */
830 _cpp_get_fresh_line (cpp_reader *pfile)
834 /* We can't get a new line until we leave the current directive. */
835 if (pfile->state.in_directive)
840 cpp_buffer *buffer = pfile->buffer;
842 if (!buffer->need_line)
845 if (buffer->next_line < buffer->rlimit)
847 _cpp_clean_line (pfile);
851 /* First, get out of parsing arguments state. */
852 if (pfile->state.parsing_args)
855 /* End of buffer. Non-empty files should end in a newline. */
856 if (buffer->buf != buffer->rlimit
857 && buffer->next_line > buffer->rlimit
858 && !buffer->from_stage3)
860 /* Clip to buffer size. */
861 buffer->next_line = buffer->rlimit;
862 /* APPLE LOCAL begin suppress no newline warning. */
863 if ( CPP_OPTION (pfile, warn_newline_at_eof))
865 cpp_error_with_line (pfile, CPP_DL_PEDWARN, pfile->line_table->highest_line,
866 CPP_BUF_COLUMN (buffer, buffer->cur),
867 "no newline at end of file");
869 /* APPLE LOCAL end suppress no newline warning. */
872 return_at_eof = buffer->return_at_eof;
873 _cpp_pop_buffer (pfile);
874 if (pfile->buffer == NULL || return_at_eof)
879 #define IF_NEXT_IS(CHAR, THEN_TYPE, ELSE_TYPE) \
882 result->type = ELSE_TYPE; \
883 if (*buffer->cur == CHAR) \
884 buffer->cur++, result->type = THEN_TYPE; \
888 /* Lex a token into pfile->cur_token, which is also incremented, to
889 get diagnostics pointing to the correct location.
891 Does not handle issues such as token lookahead, multiple-include
892 optimization, directives, skipping etc. This function is only
893 suitable for use by _cpp_lex_token, and in special cases like
894 lex_expansion_token which doesn't care for any of these issues.
896 When meeting a newline, returns CPP_EOF if parsing a directive,
897 otherwise returns to the start of the token buffer if permissible.
898 Returns the location of the lexed token. */
900 _cpp_lex_direct (cpp_reader *pfile)
904 const unsigned char *comment_start;
905 cpp_token *result = pfile->cur_token++;
909 buffer = pfile->buffer;
910 if (buffer->need_line)
912 if (pfile->state.in_deferred_pragma)
914 result->type = CPP_PRAGMA_EOL;
915 pfile->state.in_deferred_pragma = false;
916 if (!pfile->state.pragma_allow_expansion)
917 pfile->state.prevent_expansion--;
920 if (!_cpp_get_fresh_line (pfile))
922 result->type = CPP_EOF;
923 if (!pfile->state.in_directive)
925 /* Tell the compiler the line number of the EOF token. */
926 result->src_loc = pfile->line_table->highest_line;
931 if (!pfile->keep_tokens)
933 pfile->cur_run = &pfile->base_run;
934 result = pfile->base_run.base;
935 pfile->cur_token = result + 1;
938 if (pfile->state.parsing_args == 2)
939 result->flags |= PREV_WHITE;
941 buffer = pfile->buffer;
943 result->src_loc = pfile->line_table->highest_line;
946 if (buffer->cur >= buffer->notes[buffer->cur_note].pos
947 && !pfile->overlaid_buffer)
949 _cpp_process_line_notes (pfile, false);
950 result->src_loc = pfile->line_table->highest_line;
954 LINEMAP_POSITION_FOR_COLUMN (result->src_loc, pfile->line_table,
955 CPP_BUF_COLUMN (buffer, buffer->cur));
959 case ' ': case '\t': case '\f': case '\v': case '\0':
960 result->flags |= PREV_WHITE;
961 skip_whitespace (pfile, c);
965 if (buffer->cur < buffer->rlimit)
966 CPP_INCREMENT_LINE (pfile, 0);
967 buffer->need_line = true;
970 case '0': case '1': case '2': case '3': case '4':
971 case '5': case '6': case '7': case '8': case '9':
973 struct normalize_state nst = INITIAL_NORMALIZE_STATE;
974 result->type = CPP_NUMBER;
975 lex_number (pfile, &result->val.str, &nst);
976 warn_about_normalization (pfile, result, &nst);
981 /* 'L' may introduce wide characters or strings. */
982 if (*buffer->cur == '\'' || *buffer->cur == '"')
984 lex_string (pfile, result, buffer->cur - 1);
990 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
991 case 'g': case 'h': case 'i': case 'j': case 'k': case 'l':
992 case 'm': case 'n': case 'o': case 'p': case 'q': case 'r':
993 case 's': case 't': case 'u': case 'v': case 'w': case 'x':
995 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
996 case 'G': case 'H': case 'I': case 'J': case 'K':
997 case 'M': case 'N': case 'O': case 'P': case 'Q': case 'R':
998 case 'S': case 'T': case 'U': case 'V': case 'W': case 'X':
1000 result->type = CPP_NAME;
1002 struct normalize_state nst = INITIAL_NORMALIZE_STATE;
1003 result->val.node = lex_identifier (pfile, buffer->cur - 1, false,
1005 warn_about_normalization (pfile, result, &nst);
1008 /* Convert named operators to their proper types. */
1009 if (result->val.node->flags & NODE_OPERATOR)
1011 result->flags |= NAMED_OP;
1012 result->type = (enum cpp_ttype) result->val.node->directive_index;
1018 lex_string (pfile, result, buffer->cur - 1);
1022 /* A potential block or line comment. */
1023 comment_start = buffer->cur;
1028 if (_cpp_skip_block_comment (pfile))
1029 cpp_error (pfile, CPP_DL_ERROR, "unterminated comment");
1031 else if (c == '/' && (CPP_OPTION (pfile, cplusplus_comments)
1032 || cpp_in_system_header (pfile)))
1034 /* Warn about comments only if pedantically GNUC89, and not
1035 in system headers. */
1036 if (CPP_OPTION (pfile, lang) == CLK_GNUC89 && CPP_PEDANTIC (pfile)
1037 && ! buffer->warned_cplusplus_comments)
1039 cpp_error (pfile, CPP_DL_PEDWARN,
1040 "C++ style comments are not allowed in ISO C90");
1041 cpp_error (pfile, CPP_DL_PEDWARN,
1042 "(this will be reported only once per input file)");
1043 buffer->warned_cplusplus_comments = 1;
1046 if (skip_line_comment (pfile) && CPP_OPTION (pfile, warn_comments))
1047 cpp_error (pfile, CPP_DL_WARNING, "multi-line comment");
1052 result->type = CPP_DIV_EQ;
1057 result->type = CPP_DIV;
1061 if (!pfile->state.save_comments)
1063 result->flags |= PREV_WHITE;
1064 goto update_tokens_line;
1067 /* Save the comment as a token in its own right. */
1068 save_comment (pfile, result, comment_start, c);
1072 if (pfile->state.angled_headers)
1074 lex_string (pfile, result, buffer->cur - 1);
1078 result->type = CPP_LESS;
1079 if (*buffer->cur == '=')
1080 buffer->cur++, result->type = CPP_LESS_EQ;
1081 else if (*buffer->cur == '<')
1084 IF_NEXT_IS ('=', CPP_LSHIFT_EQ, CPP_LSHIFT);
1086 else if (CPP_OPTION (pfile, digraphs))
1088 if (*buffer->cur == ':')
1091 result->flags |= DIGRAPH;
1092 result->type = CPP_OPEN_SQUARE;
1094 else if (*buffer->cur == '%')
1097 result->flags |= DIGRAPH;
1098 result->type = CPP_OPEN_BRACE;
1104 result->type = CPP_GREATER;
1105 if (*buffer->cur == '=')
1106 buffer->cur++, result->type = CPP_GREATER_EQ;
1107 else if (*buffer->cur == '>')
1110 IF_NEXT_IS ('=', CPP_RSHIFT_EQ, CPP_RSHIFT);
1115 result->type = CPP_MOD;
1116 if (*buffer->cur == '=')
1117 buffer->cur++, result->type = CPP_MOD_EQ;
1118 else if (CPP_OPTION (pfile, digraphs))
1120 if (*buffer->cur == ':')
1123 result->flags |= DIGRAPH;
1124 result->type = CPP_HASH;
1125 if (*buffer->cur == '%' && buffer->cur[1] == ':')
1126 buffer->cur += 2, result->type = CPP_PASTE;
1128 else if (*buffer->cur == '>')
1131 result->flags |= DIGRAPH;
1132 result->type = CPP_CLOSE_BRACE;
1138 result->type = CPP_DOT;
1139 if (ISDIGIT (*buffer->cur))
1141 struct normalize_state nst = INITIAL_NORMALIZE_STATE;
1142 result->type = CPP_NUMBER;
1143 lex_number (pfile, &result->val.str, &nst);
1144 warn_about_normalization (pfile, result, &nst);
1146 else if (*buffer->cur == '.' && buffer->cur[1] == '.')
1147 buffer->cur += 2, result->type = CPP_ELLIPSIS;
1148 else if (*buffer->cur == '*' && CPP_OPTION (pfile, cplusplus))
1149 buffer->cur++, result->type = CPP_DOT_STAR;
1153 result->type = CPP_PLUS;
1154 if (*buffer->cur == '+')
1155 buffer->cur++, result->type = CPP_PLUS_PLUS;
1156 else if (*buffer->cur == '=')
1157 buffer->cur++, result->type = CPP_PLUS_EQ;
1161 result->type = CPP_MINUS;
1162 if (*buffer->cur == '>')
1165 result->type = CPP_DEREF;
1166 if (*buffer->cur == '*' && CPP_OPTION (pfile, cplusplus))
1167 buffer->cur++, result->type = CPP_DEREF_STAR;
1169 else if (*buffer->cur == '-')
1170 buffer->cur++, result->type = CPP_MINUS_MINUS;
1171 else if (*buffer->cur == '=')
1172 buffer->cur++, result->type = CPP_MINUS_EQ;
1176 result->type = CPP_AND;
1177 if (*buffer->cur == '&')
1178 buffer->cur++, result->type = CPP_AND_AND;
1179 else if (*buffer->cur == '=')
1180 buffer->cur++, result->type = CPP_AND_EQ;
1184 result->type = CPP_OR;
1185 if (*buffer->cur == '|')
1186 buffer->cur++, result->type = CPP_OR_OR;
1187 else if (*buffer->cur == '=')
1188 buffer->cur++, result->type = CPP_OR_EQ;
1192 result->type = CPP_COLON;
1193 if (*buffer->cur == ':' && CPP_OPTION (pfile, cplusplus))
1194 buffer->cur++, result->type = CPP_SCOPE;
1195 else if (*buffer->cur == '>' && CPP_OPTION (pfile, digraphs))
1198 result->flags |= DIGRAPH;
1199 result->type = CPP_CLOSE_SQUARE;
1203 case '*': IF_NEXT_IS ('=', CPP_MULT_EQ, CPP_MULT); break;
1204 case '=': IF_NEXT_IS ('=', CPP_EQ_EQ, CPP_EQ); break;
1205 case '!': IF_NEXT_IS ('=', CPP_NOT_EQ, CPP_NOT); break;
1206 case '^': IF_NEXT_IS ('=', CPP_XOR_EQ, CPP_XOR); break;
1207 case '#': IF_NEXT_IS ('#', CPP_PASTE, CPP_HASH); break;
1209 case '?': result->type = CPP_QUERY; break;
1210 case '~': result->type = CPP_COMPL; break;
1211 case ',': result->type = CPP_COMMA; break;
1212 case '(': result->type = CPP_OPEN_PAREN; break;
1213 case ')': result->type = CPP_CLOSE_PAREN; break;
1214 case '[': result->type = CPP_OPEN_SQUARE; break;
1215 case ']': result->type = CPP_CLOSE_SQUARE; break;
1216 case '{': result->type = CPP_OPEN_BRACE; break;
1217 case '}': result->type = CPP_CLOSE_BRACE; break;
1218 case ';': result->type = CPP_SEMICOLON; break;
1220 /* @ is a punctuator in Objective-C. */
1221 case '@': result->type = CPP_ATSIGN; break;
1226 const uchar *base = --buffer->cur;
1227 struct normalize_state nst = INITIAL_NORMALIZE_STATE;
1229 if (forms_identifier_p (pfile, true, &nst))
1231 result->type = CPP_NAME;
1232 result->val.node = lex_identifier (pfile, base, true, &nst);
1233 warn_about_normalization (pfile, result, &nst);
1240 create_literal (pfile, result, buffer->cur - 1, 1, CPP_OTHER);
1247 /* An upper bound on the number of bytes needed to spell TOKEN.
1248 Does not include preceding whitespace. */
1250 cpp_token_len (const cpp_token *token)
1254 switch (TOKEN_SPELL (token))
1256 default: len = 4; break;
1257 case SPELL_LITERAL: len = token->val.str.len; break;
1258 case SPELL_IDENT: len = NODE_LEN (token->val.node) * 10; break;
1264 /* Parse UTF-8 out of NAMEP and place a \U escape in BUFFER.
1265 Return the number of bytes read out of NAME. (There are always
1266 10 bytes written to BUFFER.) */
1269 utf8_to_ucn (unsigned char *buffer, const unsigned char *name)
1275 unsigned long utf32;
1277 /* Compute the length of the UTF-8 sequence. */
1278 for (t = *name; t & 0x80; t <<= 1)
1281 utf32 = *name & (0x7F >> ucn_len);
1282 for (ucn_len_c = 1; ucn_len_c < ucn_len; ucn_len_c++)
1284 utf32 = (utf32 << 6) | (*++name & 0x3F);
1286 /* Ill-formed UTF-8. */
1287 if ((*name & ~0x3F) != 0x80)
1293 for (j = 7; j >= 0; j--)
1294 *buffer++ = "0123456789abcdef"[(utf32 >> (4 * j)) & 0xF];
1299 /* Write the spelling of a token TOKEN to BUFFER. The buffer must
1300 already contain the enough space to hold the token's spelling.
1301 Returns a pointer to the character after the last character written.
1302 FORSTRING is true if this is to be the spelling after translation
1303 phase 1 (this is different for UCNs).
1304 FIXME: Would be nice if we didn't need the PFILE argument. */
1306 cpp_spell_token (cpp_reader *pfile, const cpp_token *token,
1307 unsigned char *buffer, bool forstring)
1309 switch (TOKEN_SPELL (token))
1311 case SPELL_OPERATOR:
1313 const unsigned char *spelling;
1316 if (token->flags & DIGRAPH)
1318 = digraph_spellings[(int) token->type - (int) CPP_FIRST_DIGRAPH];
1319 else if (token->flags & NAMED_OP)
1322 spelling = TOKEN_NAME (token);
1324 while ((c = *spelling++) != '\0')
1333 memcpy (buffer, NODE_NAME (token->val.node),
1334 NODE_LEN (token->val.node));
1335 buffer += NODE_LEN (token->val.node);
1340 const unsigned char * name = NODE_NAME (token->val.node);
1342 for (i = 0; i < NODE_LEN (token->val.node); i++)
1343 if (name[i] & ~0x7F)
1345 i += utf8_to_ucn (buffer, name + i) - 1;
1349 *buffer++ = NODE_NAME (token->val.node)[i];
1354 memcpy (buffer, token->val.str.text, token->val.str.len);
1355 buffer += token->val.str.len;
1359 cpp_error (pfile, CPP_DL_ICE,
1360 "unspellable token %s", TOKEN_NAME (token));
1367 /* Returns TOKEN spelt as a null-terminated string. The string is
1368 freed when the reader is destroyed. Useful for diagnostics. */
1370 cpp_token_as_text (cpp_reader *pfile, const cpp_token *token)
1372 unsigned int len = cpp_token_len (token) + 1;
1373 unsigned char *start = _cpp_unaligned_alloc (pfile, len), *end;
1375 end = cpp_spell_token (pfile, token, start, false);
1381 /* Used by C front ends, which really should move to using
1382 cpp_token_as_text. */
1384 cpp_type2name (enum cpp_ttype type)
1386 return (const char *) token_spellings[type].name;
1389 /* Writes the spelling of token to FP, without any preceding space.
1390 Separated from cpp_spell_token for efficiency - to avoid stdio
1391 double-buffering. */
1393 cpp_output_token (const cpp_token *token, FILE *fp)
1395 switch (TOKEN_SPELL (token))
1397 case SPELL_OPERATOR:
1399 const unsigned char *spelling;
1402 if (token->flags & DIGRAPH)
1404 = digraph_spellings[(int) token->type - (int) CPP_FIRST_DIGRAPH];
1405 else if (token->flags & NAMED_OP)
1408 spelling = TOKEN_NAME (token);
1413 while ((c = *++spelling) != '\0');
1421 const unsigned char * name = NODE_NAME (token->val.node);
1423 for (i = 0; i < NODE_LEN (token->val.node); i++)
1424 if (name[i] & ~0x7F)
1426 unsigned char buffer[10];
1427 i += utf8_to_ucn (buffer, name + i) - 1;
1428 fwrite (buffer, 1, 10, fp);
1431 fputc (NODE_NAME (token->val.node)[i], fp);
1436 fwrite (token->val.str.text, 1, token->val.str.len, fp);
1440 /* An error, most probably. */
1445 /* Compare two tokens. */
1447 _cpp_equiv_tokens (const cpp_token *a, const cpp_token *b)
1449 if (a->type == b->type && a->flags == b->flags)
1450 switch (TOKEN_SPELL (a))
1452 default: /* Keep compiler happy. */
1453 case SPELL_OPERATOR:
1456 return (a->type != CPP_MACRO_ARG || a->val.arg_no == b->val.arg_no);
1458 return a->val.node == b->val.node;
1460 return (a->val.str.len == b->val.str.len
1461 && !memcmp (a->val.str.text, b->val.str.text,
1468 /* Returns nonzero if a space should be inserted to avoid an
1469 accidental token paste for output. For simplicity, it is
1470 conservative, and occasionally advises a space where one is not
1471 needed, e.g. "." and ".2". */
1473 cpp_avoid_paste (cpp_reader *pfile, const cpp_token *token1,
1474 const cpp_token *token2)
1476 enum cpp_ttype a = token1->type, b = token2->type;
1479 if (token1->flags & NAMED_OP)
1481 if (token2->flags & NAMED_OP)
1485 if (token2->flags & DIGRAPH)
1486 c = digraph_spellings[(int) b - (int) CPP_FIRST_DIGRAPH][0];
1487 else if (token_spellings[b].category == SPELL_OPERATOR)
1488 c = token_spellings[b].name[0];
1490 /* Quickly get everything that can paste with an '='. */
1491 if ((int) a <= (int) CPP_LAST_EQ && c == '=')
1496 case CPP_GREATER: return c == '>';
1497 case CPP_LESS: return c == '<' || c == '%' || c == ':';
1498 case CPP_PLUS: return c == '+';
1499 case CPP_MINUS: return c == '-' || c == '>';
1500 case CPP_DIV: return c == '/' || c == '*'; /* Comments. */
1501 case CPP_MOD: return c == ':' || c == '>';
1502 case CPP_AND: return c == '&';
1503 case CPP_OR: return c == '|';
1504 case CPP_COLON: return c == ':' || c == '>';
1505 case CPP_DEREF: return c == '*';
1506 case CPP_DOT: return c == '.' || c == '%' || b == CPP_NUMBER;
1507 case CPP_HASH: return c == '#' || c == '%'; /* Digraph form. */
1508 case CPP_NAME: return ((b == CPP_NUMBER
1509 && name_p (pfile, &token2->val.str))
1511 || b == CPP_CHAR || b == CPP_STRING); /* L */
1512 case CPP_NUMBER: return (b == CPP_NUMBER || b == CPP_NAME
1513 || c == '.' || c == '+' || c == '-');
1515 case CPP_OTHER: return ((token1->val.str.text[0] == '\\'
1517 || (CPP_OPTION (pfile, objc)
1518 && token1->val.str.text[0] == '@'
1519 && (b == CPP_NAME || b == CPP_STRING)));
1526 /* Output all the remaining tokens on the current line, and a newline
1527 character, to FP. Leading whitespace is removed. If there are
1528 macros, special token padding is not performed. */
1530 cpp_output_line (cpp_reader *pfile, FILE *fp)
1532 const cpp_token *token;
1534 token = cpp_get_token (pfile);
1535 while (token->type != CPP_EOF)
1537 cpp_output_token (token, fp);
1538 token = cpp_get_token (pfile);
1539 if (token->flags & PREV_WHITE)
1546 /* Memory buffers. Changing these three constants can have a dramatic
1547 effect on performance. The values here are reasonable defaults,
1548 but might be tuned. If you adjust them, be sure to test across a
1549 range of uses of cpplib, including heavy nested function-like macro
1550 expansion. Also check the change in peak memory usage (NJAMD is a
1551 good tool for this). */
1552 #define MIN_BUFF_SIZE 8000
1553 #define BUFF_SIZE_UPPER_BOUND(MIN_SIZE) (MIN_BUFF_SIZE + (MIN_SIZE) * 3 / 2)
1554 #define EXTENDED_BUFF_SIZE(BUFF, MIN_EXTRA) \
1555 (MIN_EXTRA + ((BUFF)->limit - (BUFF)->cur) * 2)
1557 #if MIN_BUFF_SIZE > BUFF_SIZE_UPPER_BOUND (0)
1558 #error BUFF_SIZE_UPPER_BOUND must be at least as large as MIN_BUFF_SIZE!
1561 /* Create a new allocation buffer. Place the control block at the end
1562 of the buffer, so that buffer overflows will cause immediate chaos. */
1564 new_buff (size_t len)
1567 unsigned char *base;
1569 if (len < MIN_BUFF_SIZE)
1570 len = MIN_BUFF_SIZE;
1571 len = CPP_ALIGN (len);
1573 base = XNEWVEC (unsigned char, len + sizeof (_cpp_buff));
1574 result = (_cpp_buff *) (base + len);
1575 result->base = base;
1577 result->limit = base + len;
1578 result->next = NULL;
1582 /* Place a chain of unwanted allocation buffers on the free list. */
1584 _cpp_release_buff (cpp_reader *pfile, _cpp_buff *buff)
1586 _cpp_buff *end = buff;
1590 end->next = pfile->free_buffs;
1591 pfile->free_buffs = buff;
1594 /* Return a free buffer of size at least MIN_SIZE. */
1596 _cpp_get_buff (cpp_reader *pfile, size_t min_size)
1598 _cpp_buff *result, **p;
1600 for (p = &pfile->free_buffs;; p = &(*p)->next)
1605 return new_buff (min_size);
1607 size = result->limit - result->base;
1608 /* Return a buffer that's big enough, but don't waste one that's
1610 if (size >= min_size && size <= BUFF_SIZE_UPPER_BOUND (min_size))
1615 result->next = NULL;
1616 result->cur = result->base;
1620 /* Creates a new buffer with enough space to hold the uncommitted
1621 remaining bytes of BUFF, and at least MIN_EXTRA more bytes. Copies
1622 the excess bytes to the new buffer. Chains the new buffer after
1623 BUFF, and returns the new buffer. */
1625 _cpp_append_extend_buff (cpp_reader *pfile, _cpp_buff *buff, size_t min_extra)
1627 size_t size = EXTENDED_BUFF_SIZE (buff, min_extra);
1628 _cpp_buff *new_buff = _cpp_get_buff (pfile, size);
1630 buff->next = new_buff;
1631 memcpy (new_buff->base, buff->cur, BUFF_ROOM (buff));
1635 /* Creates a new buffer with enough space to hold the uncommitted
1636 remaining bytes of the buffer pointed to by BUFF, and at least
1637 MIN_EXTRA more bytes. Copies the excess bytes to the new buffer.
1638 Chains the new buffer before the buffer pointed to by BUFF, and
1639 updates the pointer to point to the new buffer. */
1641 _cpp_extend_buff (cpp_reader *pfile, _cpp_buff **pbuff, size_t min_extra)
1643 _cpp_buff *new_buff, *old_buff = *pbuff;
1644 size_t size = EXTENDED_BUFF_SIZE (old_buff, min_extra);
1646 new_buff = _cpp_get_buff (pfile, size);
1647 memcpy (new_buff->base, old_buff->cur, BUFF_ROOM (old_buff));
1648 new_buff->next = old_buff;
1652 /* Free a chain of buffers starting at BUFF. */
1654 _cpp_free_buff (_cpp_buff *buff)
1658 for (; buff; buff = next)
1665 /* Allocate permanent, unaligned storage of length LEN. */
1667 _cpp_unaligned_alloc (cpp_reader *pfile, size_t len)
1669 _cpp_buff *buff = pfile->u_buff;
1670 unsigned char *result = buff->cur;
1672 if (len > (size_t) (buff->limit - result))
1674 buff = _cpp_get_buff (pfile, len);
1675 buff->next = pfile->u_buff;
1676 pfile->u_buff = buff;
1680 buff->cur = result + len;
1684 /* Allocate permanent, unaligned storage of length LEN from a_buff.
1685 That buffer is used for growing allocations when saving macro
1686 replacement lists in a #define, and when parsing an answer to an
1687 assertion in #assert, #unassert or #if (and therefore possibly
1688 whilst expanding macros). It therefore must not be used by any
1689 code that they might call: specifically the lexer and the guts of
1692 All existing other uses clearly fit this restriction: storing
1693 registered pragmas during initialization. */
1695 _cpp_aligned_alloc (cpp_reader *pfile, size_t len)
1697 _cpp_buff *buff = pfile->a_buff;
1698 unsigned char *result = buff->cur;
1700 if (len > (size_t) (buff->limit - result))
1702 buff = _cpp_get_buff (pfile, len);
1703 buff->next = pfile->a_buff;
1704 pfile->a_buff = buff;
1708 buff->cur = result + len;
1712 /* Say which field of TOK is in use. */
1714 enum cpp_token_fld_kind
1715 cpp_token_val_index (cpp_token *tok)
1717 switch (TOKEN_SPELL (tok))
1720 return CPP_TOKEN_FLD_NODE;
1722 return CPP_TOKEN_FLD_STR;
1724 if (tok->type == CPP_MACRO_ARG)
1725 return CPP_TOKEN_FLD_ARG_NO;
1726 else if (tok->type == CPP_PADDING)
1727 return CPP_TOKEN_FLD_SOURCE;
1728 else if (tok->type == CPP_PRAGMA)
1729 return CPP_TOKEN_FLD_PRAGMA;
1730 /* else fall through */
1732 return CPP_TOKEN_FLD_NONE;