1 /* CPP Library - lexical analysis.
2 Copyright (C) 2000, 2001, 2002, 2003, 2004, 2005, 2007, 2008, 2009
3 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 Broken out to separate file, Zack Weinberg, Mar 2000
9 This program is free software; you can redistribute it and/or modify it
10 under the terms of the GNU General Public License as published by the
11 Free Software Foundation; either version 3, or (at your option) any
14 This program is distributed in the hope that it will be useful,
15 but WITHOUT ANY WARRANTY; without even the implied warranty of
16 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17 GNU General Public License for more details.
19 You should have received a copy of the GNU General Public License
20 along with this program; see the file COPYING3. If not see
21 <http://www.gnu.org/licenses/>. */
38 enum spell_type category;
39 const unsigned char *name;
42 static const unsigned char *const digraph_spellings[] =
43 { UC"%:", UC"%:%:", UC"<:", UC":>", UC"<%", UC"%>" };
45 #define OP(e, s) { SPELL_OPERATOR, UC s },
46 #define TK(e, s) { SPELL_ ## s, UC #e },
47 static const struct token_spelling token_spellings[N_TTYPES] = { TTYPE_TABLE };
51 #define TOKEN_SPELL(token) (token_spellings[(token)->type].category)
52 #define TOKEN_NAME(token) (token_spellings[(token)->type].name)
54 static void add_line_note (cpp_buffer *, const uchar *, unsigned int);
55 static int skip_line_comment (cpp_reader *);
56 static void skip_whitespace (cpp_reader *, cppchar_t);
57 static void lex_string (cpp_reader *, cpp_token *, const uchar *);
58 static void save_comment (cpp_reader *, cpp_token *, const uchar *, cppchar_t);
59 static void store_comment (cpp_reader *, cpp_token *);
60 static void create_literal (cpp_reader *, cpp_token *, const uchar *,
61 unsigned int, enum cpp_ttype);
62 static bool warn_in_comment (cpp_reader *, _cpp_line_note *);
63 static int name_p (cpp_reader *, const cpp_string *);
64 static tokenrun *next_tokenrun (tokenrun *);
66 static _cpp_buff *new_buff (size_t);
71 Compares, the token TOKEN to the NUL-terminated string STRING.
72 TOKEN must be a CPP_NAME. Returns 1 for equal, 0 for unequal. */
74 cpp_ideq (const cpp_token *token, const char *string)
76 if (token->type != CPP_NAME)
79 return !ustrcmp (NODE_NAME (token->val.node), (const uchar *) string);
82 /* Record a note TYPE at byte POS into the current cleaned logical
85 add_line_note (cpp_buffer *buffer, const uchar *pos, unsigned int type)
87 if (buffer->notes_used == buffer->notes_cap)
89 buffer->notes_cap = buffer->notes_cap * 2 + 200;
90 buffer->notes = XRESIZEVEC (_cpp_line_note, buffer->notes,
94 buffer->notes[buffer->notes_used].pos = pos;
95 buffer->notes[buffer->notes_used].type = type;
99 /* Returns with a logical line that contains no escaped newlines or
100 trigraphs. This is a time-critical inner loop. */
102 _cpp_clean_line (cpp_reader *pfile)
108 buffer = pfile->buffer;
109 buffer->cur_note = buffer->notes_used = 0;
110 buffer->cur = buffer->line_base = buffer->next_line;
111 buffer->need_line = false;
112 s = buffer->next_line - 1;
114 if (!buffer->from_stage3)
116 const uchar *pbackslash = NULL;
118 /* Short circuit for the common case of an un-escaped line with
119 no trigraphs. The primary win here is by not writing any
120 data back to memory until we have to. */
124 if (__builtin_expect (c == '\n', false)
125 || __builtin_expect (c == '\r', false))
129 if (__builtin_expect (s == buffer->rlimit, false))
132 /* DOS line ending? */
133 if (__builtin_expect (c == '\r', false)
137 if (s == buffer->rlimit)
141 if (__builtin_expect (pbackslash == NULL, true))
144 /* Check for escaped newline. */
146 while (is_nvspace (p[-1]))
148 if (p - 1 != pbackslash)
151 /* Have an escaped newline; process it and proceed to
153 add_line_note (buffer, p - 1, p != d ? ' ' : '\\');
155 buffer->next_line = p - 1;
158 if (__builtin_expect (c == '\\', false))
160 else if (__builtin_expect (c == '?', false)
161 && __builtin_expect (s[1] == '?', false)
162 && _cpp_trigraph_map[s[2]])
164 /* Have a trigraph. We may or may not have to convert
165 it. Add a line note regardless, for -Wtrigraphs. */
166 add_line_note (buffer, s, s[2]);
167 if (CPP_OPTION (pfile, trigraphs))
169 /* We do, and that means we have to switch to the
172 *d = _cpp_trigraph_map[s[2]];
185 if (c == '\n' || c == '\r')
187 /* Handle DOS line endings. */
188 if (c == '\r' && s != buffer->rlimit && s[1] == '\n')
190 if (s == buffer->rlimit)
195 while (p != buffer->next_line && is_nvspace (p[-1]))
197 if (p == buffer->next_line || p[-1] != '\\')
200 add_line_note (buffer, p - 1, p != d ? ' ': '\\');
202 buffer->next_line = p - 1;
204 else if (c == '?' && s[1] == '?' && _cpp_trigraph_map[s[2]])
206 /* Add a note regardless, for the benefit of -Wtrigraphs. */
207 add_line_note (buffer, d, s[2]);
208 if (CPP_OPTION (pfile, trigraphs))
210 *d = _cpp_trigraph_map[s[2]];
220 while (*s != '\n' && *s != '\r');
223 /* Handle DOS line endings. */
224 if (*s == '\r' && s != buffer->rlimit && s[1] == '\n')
230 /* A sentinel note that should never be processed. */
231 add_line_note (buffer, d + 1, '\n');
232 buffer->next_line = s + 1;
235 /* Return true if the trigraph indicated by NOTE should be warned
236 about in a comment. */
238 warn_in_comment (cpp_reader *pfile, _cpp_line_note *note)
242 /* Within comments we don't warn about trigraphs, unless the
243 trigraph forms an escaped newline, as that may change
245 if (note->type != '/')
248 /* If -trigraphs, then this was an escaped newline iff the next note
250 if (CPP_OPTION (pfile, trigraphs))
251 return note[1].pos == note->pos;
253 /* Otherwise, see if this forms an escaped newline. */
255 while (is_nvspace (*p))
258 /* There might have been escaped newlines between the trigraph and the
259 newline we found. Hence the position test. */
260 return (*p == '\n' && p < note[1].pos);
263 /* Process the notes created by add_line_note as far as the current
266 _cpp_process_line_notes (cpp_reader *pfile, int in_comment)
268 cpp_buffer *buffer = pfile->buffer;
272 _cpp_line_note *note = &buffer->notes[buffer->cur_note];
275 if (note->pos > buffer->cur)
279 col = CPP_BUF_COLUMN (buffer, note->pos + 1);
281 if (note->type == '\\' || note->type == ' ')
283 if (note->type == ' ' && !in_comment)
284 cpp_error_with_line (pfile, CPP_DL_WARNING, pfile->line_table->highest_line, col,
285 "backslash and newline separated by space");
287 if (buffer->next_line > buffer->rlimit)
289 cpp_error_with_line (pfile, CPP_DL_PEDWARN, pfile->line_table->highest_line, col,
290 "backslash-newline at end of file");
291 /* Prevent "no newline at end of file" warning. */
292 buffer->next_line = buffer->rlimit;
295 buffer->line_base = note->pos;
296 CPP_INCREMENT_LINE (pfile, 0);
298 else if (_cpp_trigraph_map[note->type])
300 if (CPP_OPTION (pfile, warn_trigraphs)
301 && (!in_comment || warn_in_comment (pfile, note)))
303 if (CPP_OPTION (pfile, trigraphs))
304 cpp_error_with_line (pfile, CPP_DL_WARNING, pfile->line_table->highest_line, col,
305 "trigraph ??%c converted to %c",
307 (int) _cpp_trigraph_map[note->type]);
311 (pfile, CPP_DL_WARNING, pfile->line_table->highest_line, col,
312 "trigraph ??%c ignored, use -trigraphs to enable",
322 /* Skip a C-style block comment. We find the end of the comment by
323 seeing if an asterisk is before every '/' we encounter. Returns
324 nonzero if comment terminated by EOF, zero otherwise.
326 Buffer->cur points to the initial asterisk of the comment. */
328 _cpp_skip_block_comment (cpp_reader *pfile)
330 cpp_buffer *buffer = pfile->buffer;
331 const uchar *cur = buffer->cur;
340 /* People like decorating comments with '*', so check for '/'
341 instead for efficiency. */
349 /* Warn about potential nested comments, but not if the '/'
350 comes immediately before the true comment delimiter.
351 Don't bother to get it right across escaped newlines. */
352 if (CPP_OPTION (pfile, warn_comments)
353 && cur[0] == '*' && cur[1] != '/')
356 cpp_error_with_line (pfile, CPP_DL_WARNING,
357 pfile->line_table->highest_line, CPP_BUF_COL (buffer),
358 "\"/*\" within comment");
364 buffer->cur = cur - 1;
365 _cpp_process_line_notes (pfile, true);
366 if (buffer->next_line >= buffer->rlimit)
368 _cpp_clean_line (pfile);
370 cols = buffer->next_line - buffer->line_base;
371 CPP_INCREMENT_LINE (pfile, cols);
378 _cpp_process_line_notes (pfile, true);
382 /* Skip a C++ line comment, leaving buffer->cur pointing to the
383 terminating newline. Handles escaped newlines. Returns nonzero
384 if a multiline comment. */
386 skip_line_comment (cpp_reader *pfile)
388 cpp_buffer *buffer = pfile->buffer;
389 source_location orig_line = pfile->line_table->highest_line;
391 while (*buffer->cur != '\n')
394 _cpp_process_line_notes (pfile, true);
395 return orig_line != pfile->line_table->highest_line;
398 /* Skips whitespace, saving the next non-whitespace character. */
400 skip_whitespace (cpp_reader *pfile, cppchar_t c)
402 cpp_buffer *buffer = pfile->buffer;
403 bool saw_NUL = false;
407 /* Horizontal space always OK. */
408 if (c == ' ' || c == '\t')
410 /* Just \f \v or \0 left. */
413 else if (pfile->state.in_directive && CPP_PEDANTIC (pfile))
414 cpp_error_with_line (pfile, CPP_DL_PEDWARN, pfile->line_table->highest_line,
415 CPP_BUF_COL (buffer),
416 "%s in preprocessing directive",
417 c == '\f' ? "form feed" : "vertical tab");
421 /* We only want non-vertical space, i.e. ' ' \t \f \v \0. */
422 while (is_nvspace (c));
425 cpp_error (pfile, CPP_DL_WARNING, "null character(s) ignored");
430 /* See if the characters of a number token are valid in a name (no
433 name_p (cpp_reader *pfile, const cpp_string *string)
437 for (i = 0; i < string->len; i++)
438 if (!is_idchar (string->text[i]))
444 /* After parsing an identifier or other sequence, produce a warning about
445 sequences not in NFC/NFKC. */
447 warn_about_normalization (cpp_reader *pfile,
448 const cpp_token *token,
449 const struct normalize_state *s)
451 if (CPP_OPTION (pfile, warn_normalize) < NORMALIZE_STATE_RESULT (s)
452 && !pfile->state.skipping)
454 /* Make sure that the token is printed using UCNs, even
455 if we'd otherwise happily print UTF-8. */
456 unsigned char *buf = XNEWVEC (unsigned char, cpp_token_len (token));
459 sz = cpp_spell_token (pfile, token, buf, false) - buf;
460 if (NORMALIZE_STATE_RESULT (s) == normalized_C)
461 cpp_error_with_line (pfile, CPP_DL_WARNING, token->src_loc, 0,
462 "`%.*s' is not in NFKC", (int) sz, buf);
464 cpp_error_with_line (pfile, CPP_DL_WARNING, token->src_loc, 0,
465 "`%.*s' is not in NFC", (int) sz, buf);
469 /* Returns TRUE if the sequence starting at buffer->cur is invalid in
470 an identifier. FIRST is TRUE if this starts an identifier. */
472 forms_identifier_p (cpp_reader *pfile, int first,
473 struct normalize_state *state)
475 cpp_buffer *buffer = pfile->buffer;
477 if (*buffer->cur == '$')
479 if (!CPP_OPTION (pfile, dollars_in_ident))
483 if (CPP_OPTION (pfile, warn_dollars) && !pfile->state.skipping)
485 CPP_OPTION (pfile, warn_dollars) = 0;
486 cpp_error (pfile, CPP_DL_PEDWARN, "'$' in identifier or number");
492 /* Is this a syntactically valid UCN? */
493 if (CPP_OPTION (pfile, extended_identifiers)
494 && *buffer->cur == '\\'
495 && (buffer->cur[1] == 'u' || buffer->cur[1] == 'U'))
498 if (_cpp_valid_ucn (pfile, &buffer->cur, buffer->rlimit, 1 + !first,
507 /* Helper function to get the cpp_hashnode of the identifier BASE. */
508 static cpp_hashnode *
509 lex_identifier_intern (cpp_reader *pfile, const uchar *base)
511 cpp_hashnode *result;
514 unsigned int hash = HT_HASHSTEP (0, *base);
517 while (ISIDNUM (*cur))
519 hash = HT_HASHSTEP (hash, *cur);
523 hash = HT_HASHFINISH (hash, len);
524 result = CPP_HASHNODE (ht_lookup_with_hash (pfile->hash_table,
525 base, len, hash, HT_ALLOC));
527 /* Rarely, identifiers require diagnostics when lexed. */
528 if (__builtin_expect ((result->flags & NODE_DIAGNOSTIC)
529 && !pfile->state.skipping, 0))
531 /* It is allowed to poison the same identifier twice. */
532 if ((result->flags & NODE_POISONED) && !pfile->state.poisoned_ok)
533 cpp_error (pfile, CPP_DL_ERROR, "attempt to use poisoned \"%s\"",
536 /* Constraint 6.10.3.5: __VA_ARGS__ should only appear in the
537 replacement list of a variadic macro. */
538 if (result == pfile->spec_nodes.n__VA_ARGS__
539 && !pfile->state.va_args_ok)
540 cpp_error (pfile, CPP_DL_PEDWARN,
541 "__VA_ARGS__ can only appear in the expansion"
542 " of a C99 variadic macro");
548 /* Get the cpp_hashnode of an identifier specified by NAME in
549 the current cpp_reader object. If none is found, NULL is returned. */
551 _cpp_lex_identifier (cpp_reader *pfile, const char *name)
553 cpp_hashnode *result;
554 result = lex_identifier_intern (pfile, (uchar *) name);
558 /* Lex an identifier starting at BUFFER->CUR - 1. */
559 static cpp_hashnode *
560 lex_identifier (cpp_reader *pfile, const uchar *base, bool starts_ucn,
561 struct normalize_state *nst)
563 cpp_hashnode *result;
566 unsigned int hash = HT_HASHSTEP (0, *base);
568 cur = pfile->buffer->cur;
570 while (ISIDNUM (*cur))
572 hash = HT_HASHSTEP (hash, *cur);
575 pfile->buffer->cur = cur;
576 if (starts_ucn || forms_identifier_p (pfile, false, nst))
578 /* Slower version for identifiers containing UCNs (or $). */
580 while (ISIDNUM (*pfile->buffer->cur))
582 pfile->buffer->cur++;
583 NORMALIZE_STATE_UPDATE_IDNUM (nst);
585 } while (forms_identifier_p (pfile, false, nst));
586 result = _cpp_interpret_identifier (pfile, base,
587 pfile->buffer->cur - base);
592 hash = HT_HASHFINISH (hash, len);
594 result = CPP_HASHNODE (ht_lookup_with_hash (pfile->hash_table,
595 base, len, hash, HT_ALLOC));
598 /* Rarely, identifiers require diagnostics when lexed. */
599 if (__builtin_expect ((result->flags & NODE_DIAGNOSTIC)
600 && !pfile->state.skipping, 0))
602 /* It is allowed to poison the same identifier twice. */
603 if ((result->flags & NODE_POISONED) && !pfile->state.poisoned_ok)
604 cpp_error (pfile, CPP_DL_ERROR, "attempt to use poisoned \"%s\"",
607 /* Constraint 6.10.3.5: __VA_ARGS__ should only appear in the
608 replacement list of a variadic macro. */
609 if (result == pfile->spec_nodes.n__VA_ARGS__
610 && !pfile->state.va_args_ok)
611 cpp_error (pfile, CPP_DL_PEDWARN,
612 "__VA_ARGS__ can only appear in the expansion"
613 " of a C99 variadic macro");
619 /* Lex a number to NUMBER starting at BUFFER->CUR - 1. */
621 lex_number (cpp_reader *pfile, cpp_string *number,
622 struct normalize_state *nst)
628 base = pfile->buffer->cur - 1;
631 cur = pfile->buffer->cur;
633 /* N.B. ISIDNUM does not include $. */
634 while (ISIDNUM (*cur) || *cur == '.' || VALID_SIGN (*cur, cur[-1]))
637 NORMALIZE_STATE_UPDATE_IDNUM (nst);
640 pfile->buffer->cur = cur;
642 while (forms_identifier_p (pfile, false, nst));
644 number->len = cur - base;
645 dest = _cpp_unaligned_alloc (pfile, number->len + 1);
646 memcpy (dest, base, number->len);
647 dest[number->len] = '\0';
651 /* Create a token of type TYPE with a literal spelling. */
653 create_literal (cpp_reader *pfile, cpp_token *token, const uchar *base,
654 unsigned int len, enum cpp_ttype type)
656 uchar *dest = _cpp_unaligned_alloc (pfile, len + 1);
658 memcpy (dest, base, len);
661 token->val.str.len = len;
662 token->val.str.text = dest;
665 /* Lexes a string, character constant, or angle-bracketed header file
666 name. The stored string contains the spelling, including opening
667 quote and leading any leading 'L', 'u' or 'U'. It returns the type
668 of the literal, or CPP_OTHER if it was not properly terminated, or
669 CPP_LESS for an unterminated header name which must be relexed as
672 The spelling is NUL-terminated, but it is not guaranteed that this
673 is the first NUL since embedded NULs are preserved. */
675 lex_string (cpp_reader *pfile, cpp_token *token, const uchar *base)
677 bool saw_NUL = false;
679 cppchar_t terminator;
684 if (terminator == 'L' || terminator == 'u' || terminator == 'U')
686 if (terminator == '\"')
687 type = (*base == 'L' ? CPP_WSTRING :
688 *base == 'U' ? CPP_STRING32 :
689 *base == 'u' ? CPP_STRING16 : CPP_STRING);
690 else if (terminator == '\'')
691 type = (*base == 'L' ? CPP_WCHAR :
692 *base == 'U' ? CPP_CHAR32 :
693 *base == 'u' ? CPP_CHAR16 : CPP_CHAR);
695 terminator = '>', type = CPP_HEADER_NAME;
699 cppchar_t c = *cur++;
701 /* In #include-style directives, terminators are not escapable. */
702 if (c == '\\' && !pfile->state.angled_headers && *cur != '\n')
704 else if (c == terminator)
709 /* Unmatched quotes always yield undefined behavior, but
710 greedy lexing means that what appears to be an unterminated
711 header name may actually be a legitimate sequence of tokens. */
712 if (terminator == '>')
714 token->type = CPP_LESS;
724 if (saw_NUL && !pfile->state.skipping)
725 cpp_error (pfile, CPP_DL_WARNING,
726 "null character(s) preserved in literal");
728 if (type == CPP_OTHER && CPP_OPTION (pfile, lang) != CLK_ASM)
729 cpp_error (pfile, CPP_DL_PEDWARN, "missing terminating %c character",
732 pfile->buffer->cur = cur;
733 create_literal (pfile, token, base, cur - base, type);
736 /* Return the comment table. The client may not make any assumption
737 about the ordering of the table. */
739 cpp_get_comments (cpp_reader *pfile)
741 return &pfile->comments;
744 /* Append a comment to the end of the comment table. */
746 store_comment (cpp_reader *pfile, cpp_token *token)
750 if (pfile->comments.allocated == 0)
752 pfile->comments.allocated = 256;
753 pfile->comments.entries = (cpp_comment *) xmalloc
754 (pfile->comments.allocated * sizeof (cpp_comment));
757 if (pfile->comments.count == pfile->comments.allocated)
759 pfile->comments.allocated *= 2;
760 pfile->comments.entries = (cpp_comment *) xrealloc
761 (pfile->comments.entries,
762 pfile->comments.allocated * sizeof (cpp_comment));
765 len = token->val.str.len;
767 /* Copy comment. Note, token may not be NULL terminated. */
768 pfile->comments.entries[pfile->comments.count].comment =
769 (char *) xmalloc (sizeof (char) * (len + 1));
770 memcpy (pfile->comments.entries[pfile->comments.count].comment,
771 token->val.str.text, len);
772 pfile->comments.entries[pfile->comments.count].comment[len] = '\0';
774 /* Set source location. */
775 pfile->comments.entries[pfile->comments.count].sloc = token->src_loc;
777 /* Increment the count of entries in the comment table. */
778 pfile->comments.count++;
781 /* The stored comment includes the comment start and any terminator. */
783 save_comment (cpp_reader *pfile, cpp_token *token, const unsigned char *from,
786 unsigned char *buffer;
787 unsigned int len, clen;
789 len = pfile->buffer->cur - from + 1; /* + 1 for the initial '/'. */
791 /* C++ comments probably (not definitely) have moved past a new
792 line, which we don't want to save in the comment. */
793 if (is_vspace (pfile->buffer->cur[-1]))
796 /* If we are currently in a directive, then we need to store all
797 C++ comments as C comments internally, and so we need to
798 allocate a little extra space in that case.
800 Note that the only time we encounter a directive here is
801 when we are saving comments in a "#define". */
802 clen = (pfile->state.in_directive && type == '/') ? len + 2 : len;
804 buffer = _cpp_unaligned_alloc (pfile, clen);
806 token->type = CPP_COMMENT;
807 token->val.str.len = clen;
808 token->val.str.text = buffer;
811 memcpy (buffer + 1, from, len - 1);
813 /* Finish conversion to a C comment, if necessary. */
814 if (pfile->state.in_directive && type == '/')
817 buffer[clen - 2] = '*';
818 buffer[clen - 1] = '/';
821 /* Finally store this comment for use by clients of libcpp. */
822 store_comment (pfile, token);
825 /* Allocate COUNT tokens for RUN. */
827 _cpp_init_tokenrun (tokenrun *run, unsigned int count)
829 run->base = XNEWVEC (cpp_token, count);
830 run->limit = run->base + count;
834 /* Returns the next tokenrun, or creates one if there is none. */
836 next_tokenrun (tokenrun *run)
838 if (run->next == NULL)
840 run->next = XNEW (tokenrun);
841 run->next->prev = run;
842 _cpp_init_tokenrun (run->next, 250);
848 /* Look ahead in the input stream. */
850 cpp_peek_token (cpp_reader *pfile, int index)
852 cpp_context *context = pfile->context;
853 const cpp_token *peektok;
856 /* First, scan through any pending cpp_context objects. */
857 while (context->prev)
859 ptrdiff_t sz = (context->direct_p
860 ? LAST (context).token - FIRST (context).token
861 : LAST (context).ptoken - FIRST (context).ptoken);
863 if (index < (int) sz)
864 return (context->direct_p
865 ? FIRST (context).token + index
866 : *(FIRST (context).ptoken + index));
869 context = context->prev;
872 /* We will have to read some new tokens after all (and do so
873 without invalidating preceding tokens). */
875 pfile->keep_tokens++;
879 peektok = _cpp_lex_token (pfile);
880 if (peektok->type == CPP_EOF)
885 _cpp_backup_tokens_direct (pfile, count + 1);
886 pfile->keep_tokens--;
891 /* Allocate a single token that is invalidated at the same time as the
892 rest of the tokens on the line. Has its line and col set to the
893 same as the last lexed token, so that diagnostics appear in the
896 _cpp_temp_token (cpp_reader *pfile)
898 cpp_token *old, *result;
899 ptrdiff_t sz = pfile->cur_run->limit - pfile->cur_token;
900 ptrdiff_t la = (ptrdiff_t) pfile->lookaheads;
902 old = pfile->cur_token - 1;
903 /* Any pre-existing lookaheads must not be clobbered. */
908 tokenrun *next = next_tokenrun (pfile->cur_run);
911 memmove (next->base + 1, next->base,
912 (la - sz) * sizeof (cpp_token));
914 next->base[0] = pfile->cur_run->limit[-1];
918 memmove (pfile->cur_token + 1, pfile->cur_token,
919 MIN (la, sz - 1) * sizeof (cpp_token));
922 if (!sz && pfile->cur_token == pfile->cur_run->limit)
924 pfile->cur_run = next_tokenrun (pfile->cur_run);
925 pfile->cur_token = pfile->cur_run->base;
928 result = pfile->cur_token++;
929 result->src_loc = old->src_loc;
933 /* Lex a token into RESULT (external interface). Takes care of issues
934 like directive handling, token lookahead, multiple include
935 optimization and skipping. */
937 _cpp_lex_token (cpp_reader *pfile)
943 if (pfile->cur_token == pfile->cur_run->limit)
945 pfile->cur_run = next_tokenrun (pfile->cur_run);
946 pfile->cur_token = pfile->cur_run->base;
948 /* We assume that the current token is somewhere in the current
950 if (pfile->cur_token < pfile->cur_run->base
951 || pfile->cur_token >= pfile->cur_run->limit)
954 if (pfile->lookaheads)
957 result = pfile->cur_token++;
960 result = _cpp_lex_direct (pfile);
962 if (result->flags & BOL)
964 /* Is this a directive. If _cpp_handle_directive returns
965 false, it is an assembler #. */
966 if (result->type == CPP_HASH
967 /* 6.10.3 p 11: Directives in a list of macro arguments
968 gives undefined behavior. This implementation
969 handles the directive as normal. */
970 && pfile->state.parsing_args != 1)
972 if (_cpp_handle_directive (pfile, result->flags & PREV_WHITE))
974 if (pfile->directive_result.type == CPP_PADDING)
976 result = &pfile->directive_result;
979 else if (pfile->state.in_deferred_pragma)
980 result = &pfile->directive_result;
982 if (pfile->cb.line_change && !pfile->state.skipping)
983 pfile->cb.line_change (pfile, result, pfile->state.parsing_args);
986 /* We don't skip tokens in directives. */
987 if (pfile->state.in_directive || pfile->state.in_deferred_pragma)
990 /* Outside a directive, invalidate controlling macros. At file
991 EOF, _cpp_lex_direct takes care of popping the buffer, so we never
992 get here and MI optimization works. */
993 pfile->mi_valid = false;
995 if (!pfile->state.skipping || result->type == CPP_EOF)
1002 /* Returns true if a fresh line has been loaded. */
1004 _cpp_get_fresh_line (cpp_reader *pfile)
1008 /* We can't get a new line until we leave the current directive. */
1009 if (pfile->state.in_directive)
1014 cpp_buffer *buffer = pfile->buffer;
1016 if (!buffer->need_line)
1019 if (buffer->next_line < buffer->rlimit)
1021 _cpp_clean_line (pfile);
1025 /* First, get out of parsing arguments state. */
1026 if (pfile->state.parsing_args)
1029 /* End of buffer. Non-empty files should end in a newline. */
1030 if (buffer->buf != buffer->rlimit
1031 && buffer->next_line > buffer->rlimit
1032 && !buffer->from_stage3)
1034 /* Clip to buffer size. */
1035 buffer->next_line = buffer->rlimit;
1038 return_at_eof = buffer->return_at_eof;
1039 _cpp_pop_buffer (pfile);
1040 if (pfile->buffer == NULL || return_at_eof)
1045 #define IF_NEXT_IS(CHAR, THEN_TYPE, ELSE_TYPE) \
1048 result->type = ELSE_TYPE; \
1049 if (*buffer->cur == CHAR) \
1050 buffer->cur++, result->type = THEN_TYPE; \
1054 /* Lex a token into pfile->cur_token, which is also incremented, to
1055 get diagnostics pointing to the correct location.
1057 Does not handle issues such as token lookahead, multiple-include
1058 optimization, directives, skipping etc. This function is only
1059 suitable for use by _cpp_lex_token, and in special cases like
1060 lex_expansion_token which doesn't care for any of these issues.
1062 When meeting a newline, returns CPP_EOF if parsing a directive,
1063 otherwise returns to the start of the token buffer if permissible.
1064 Returns the location of the lexed token. */
1066 _cpp_lex_direct (cpp_reader *pfile)
1070 const unsigned char *comment_start;
1071 cpp_token *result = pfile->cur_token++;
1075 buffer = pfile->buffer;
1076 if (buffer->need_line)
1078 if (pfile->state.in_deferred_pragma)
1080 result->type = CPP_PRAGMA_EOL;
1081 pfile->state.in_deferred_pragma = false;
1082 if (!pfile->state.pragma_allow_expansion)
1083 pfile->state.prevent_expansion--;
1086 if (!_cpp_get_fresh_line (pfile))
1088 result->type = CPP_EOF;
1089 if (!pfile->state.in_directive)
1091 /* Tell the compiler the line number of the EOF token. */
1092 result->src_loc = pfile->line_table->highest_line;
1093 result->flags = BOL;
1097 if (!pfile->keep_tokens)
1099 pfile->cur_run = &pfile->base_run;
1100 result = pfile->base_run.base;
1101 pfile->cur_token = result + 1;
1103 result->flags = BOL;
1104 if (pfile->state.parsing_args == 2)
1105 result->flags |= PREV_WHITE;
1107 buffer = pfile->buffer;
1109 result->src_loc = pfile->line_table->highest_line;
1112 if (buffer->cur >= buffer->notes[buffer->cur_note].pos
1113 && !pfile->overlaid_buffer)
1115 _cpp_process_line_notes (pfile, false);
1116 result->src_loc = pfile->line_table->highest_line;
1120 LINEMAP_POSITION_FOR_COLUMN (result->src_loc, pfile->line_table,
1121 CPP_BUF_COLUMN (buffer, buffer->cur));
1125 case ' ': case '\t': case '\f': case '\v': case '\0':
1126 result->flags |= PREV_WHITE;
1127 skip_whitespace (pfile, c);
1131 if (buffer->cur < buffer->rlimit)
1132 CPP_INCREMENT_LINE (pfile, 0);
1133 buffer->need_line = true;
1136 case '0': case '1': case '2': case '3': case '4':
1137 case '5': case '6': case '7': case '8': case '9':
1139 struct normalize_state nst = INITIAL_NORMALIZE_STATE;
1140 result->type = CPP_NUMBER;
1141 lex_number (pfile, &result->val.str, &nst);
1142 warn_about_normalization (pfile, result, &nst);
1149 /* 'L', 'u' or 'U' may introduce wide characters or strings. */
1150 if (c == 'L' || CPP_OPTION (pfile, uliterals))
1152 if (*buffer->cur == '\'' || *buffer->cur == '"')
1154 lex_string (pfile, result, buffer->cur - 1);
1161 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
1162 case 'g': case 'h': case 'i': case 'j': case 'k': case 'l':
1163 case 'm': case 'n': case 'o': case 'p': case 'q': case 'r':
1164 case 's': case 't': case 'v': case 'w': case 'x':
1166 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
1167 case 'G': case 'H': case 'I': case 'J': case 'K':
1168 case 'M': case 'N': case 'O': case 'P': case 'Q': case 'R':
1169 case 'S': case 'T': case 'V': case 'W': case 'X':
1171 result->type = CPP_NAME;
1173 struct normalize_state nst = INITIAL_NORMALIZE_STATE;
1174 result->val.node = lex_identifier (pfile, buffer->cur - 1, false,
1176 warn_about_normalization (pfile, result, &nst);
1179 /* Convert named operators to their proper types. */
1180 if (result->val.node->flags & NODE_OPERATOR)
1182 result->flags |= NAMED_OP;
1183 result->type = (enum cpp_ttype) result->val.node->directive_index;
1189 lex_string (pfile, result, buffer->cur - 1);
1193 /* A potential block or line comment. */
1194 comment_start = buffer->cur;
1199 if (_cpp_skip_block_comment (pfile))
1200 cpp_error (pfile, CPP_DL_ERROR, "unterminated comment");
1202 else if (c == '/' && (CPP_OPTION (pfile, cplusplus_comments)
1203 || cpp_in_system_header (pfile)))
1205 /* Warn about comments only if pedantically GNUC89, and not
1206 in system headers. */
1207 if (CPP_OPTION (pfile, lang) == CLK_GNUC89 && CPP_PEDANTIC (pfile)
1208 && ! buffer->warned_cplusplus_comments)
1210 cpp_error (pfile, CPP_DL_PEDWARN,
1211 "C++ style comments are not allowed in ISO C90");
1212 cpp_error (pfile, CPP_DL_PEDWARN,
1213 "(this will be reported only once per input file)");
1214 buffer->warned_cplusplus_comments = 1;
1217 if (skip_line_comment (pfile) && CPP_OPTION (pfile, warn_comments))
1218 cpp_error (pfile, CPP_DL_WARNING, "multi-line comment");
1223 result->type = CPP_DIV_EQ;
1228 result->type = CPP_DIV;
1232 if (!pfile->state.save_comments)
1234 result->flags |= PREV_WHITE;
1235 goto update_tokens_line;
1238 /* Save the comment as a token in its own right. */
1239 save_comment (pfile, result, comment_start, c);
1243 if (pfile->state.angled_headers)
1245 lex_string (pfile, result, buffer->cur - 1);
1246 if (result->type != CPP_LESS)
1250 result->type = CPP_LESS;
1251 if (*buffer->cur == '=')
1252 buffer->cur++, result->type = CPP_LESS_EQ;
1253 else if (*buffer->cur == '<')
1256 IF_NEXT_IS ('=', CPP_LSHIFT_EQ, CPP_LSHIFT);
1258 else if (CPP_OPTION (pfile, digraphs))
1260 if (*buffer->cur == ':')
1263 result->flags |= DIGRAPH;
1264 result->type = CPP_OPEN_SQUARE;
1266 else if (*buffer->cur == '%')
1269 result->flags |= DIGRAPH;
1270 result->type = CPP_OPEN_BRACE;
1276 result->type = CPP_GREATER;
1277 if (*buffer->cur == '=')
1278 buffer->cur++, result->type = CPP_GREATER_EQ;
1279 else if (*buffer->cur == '>')
1282 IF_NEXT_IS ('=', CPP_RSHIFT_EQ, CPP_RSHIFT);
1287 result->type = CPP_MOD;
1288 if (*buffer->cur == '=')
1289 buffer->cur++, result->type = CPP_MOD_EQ;
1290 else if (CPP_OPTION (pfile, digraphs))
1292 if (*buffer->cur == ':')
1295 result->flags |= DIGRAPH;
1296 result->type = CPP_HASH;
1297 if (*buffer->cur == '%' && buffer->cur[1] == ':')
1298 buffer->cur += 2, result->type = CPP_PASTE;
1300 else if (*buffer->cur == '>')
1303 result->flags |= DIGRAPH;
1304 result->type = CPP_CLOSE_BRACE;
1310 result->type = CPP_DOT;
1311 if (ISDIGIT (*buffer->cur))
1313 struct normalize_state nst = INITIAL_NORMALIZE_STATE;
1314 result->type = CPP_NUMBER;
1315 lex_number (pfile, &result->val.str, &nst);
1316 warn_about_normalization (pfile, result, &nst);
1318 else if (*buffer->cur == '.' && buffer->cur[1] == '.')
1319 buffer->cur += 2, result->type = CPP_ELLIPSIS;
1320 else if (*buffer->cur == '*' && CPP_OPTION (pfile, cplusplus))
1321 buffer->cur++, result->type = CPP_DOT_STAR;
1325 result->type = CPP_PLUS;
1326 if (*buffer->cur == '+')
1327 buffer->cur++, result->type = CPP_PLUS_PLUS;
1328 else if (*buffer->cur == '=')
1329 buffer->cur++, result->type = CPP_PLUS_EQ;
1333 result->type = CPP_MINUS;
1334 if (*buffer->cur == '>')
1337 result->type = CPP_DEREF;
1338 if (*buffer->cur == '*' && CPP_OPTION (pfile, cplusplus))
1339 buffer->cur++, result->type = CPP_DEREF_STAR;
1341 else if (*buffer->cur == '-')
1342 buffer->cur++, result->type = CPP_MINUS_MINUS;
1343 else if (*buffer->cur == '=')
1344 buffer->cur++, result->type = CPP_MINUS_EQ;
1348 result->type = CPP_AND;
1349 if (*buffer->cur == '&')
1350 buffer->cur++, result->type = CPP_AND_AND;
1351 else if (*buffer->cur == '=')
1352 buffer->cur++, result->type = CPP_AND_EQ;
1356 result->type = CPP_OR;
1357 if (*buffer->cur == '|')
1358 buffer->cur++, result->type = CPP_OR_OR;
1359 else if (*buffer->cur == '=')
1360 buffer->cur++, result->type = CPP_OR_EQ;
1364 result->type = CPP_COLON;
1365 if (*buffer->cur == ':' && CPP_OPTION (pfile, cplusplus))
1366 buffer->cur++, result->type = CPP_SCOPE;
1367 else if (*buffer->cur == '>' && CPP_OPTION (pfile, digraphs))
1370 result->flags |= DIGRAPH;
1371 result->type = CPP_CLOSE_SQUARE;
1375 case '*': IF_NEXT_IS ('=', CPP_MULT_EQ, CPP_MULT); break;
1376 case '=': IF_NEXT_IS ('=', CPP_EQ_EQ, CPP_EQ); break;
1377 case '!': IF_NEXT_IS ('=', CPP_NOT_EQ, CPP_NOT); break;
1378 case '^': IF_NEXT_IS ('=', CPP_XOR_EQ, CPP_XOR); break;
1379 case '#': IF_NEXT_IS ('#', CPP_PASTE, CPP_HASH); break;
1381 case '?': result->type = CPP_QUERY; break;
1382 case '~': result->type = CPP_COMPL; break;
1383 case ',': result->type = CPP_COMMA; break;
1384 case '(': result->type = CPP_OPEN_PAREN; break;
1385 case ')': result->type = CPP_CLOSE_PAREN; break;
1386 case '[': result->type = CPP_OPEN_SQUARE; break;
1387 case ']': result->type = CPP_CLOSE_SQUARE; break;
1388 case '{': result->type = CPP_OPEN_BRACE; break;
1389 case '}': result->type = CPP_CLOSE_BRACE; break;
1390 case ';': result->type = CPP_SEMICOLON; break;
1392 /* @ is a punctuator in Objective-C. */
1393 case '@': result->type = CPP_ATSIGN; break;
1398 const uchar *base = --buffer->cur;
1399 struct normalize_state nst = INITIAL_NORMALIZE_STATE;
1401 if (forms_identifier_p (pfile, true, &nst))
1403 result->type = CPP_NAME;
1404 result->val.node = lex_identifier (pfile, base, true, &nst);
1405 warn_about_normalization (pfile, result, &nst);
1412 create_literal (pfile, result, buffer->cur - 1, 1, CPP_OTHER);
1419 /* An upper bound on the number of bytes needed to spell TOKEN.
1420 Does not include preceding whitespace. */
1422 cpp_token_len (const cpp_token *token)
1426 switch (TOKEN_SPELL (token))
1428 default: len = 6; break;
1429 case SPELL_LITERAL: len = token->val.str.len; break;
1430 case SPELL_IDENT: len = NODE_LEN (token->val.node) * 10; break;
1436 /* Parse UTF-8 out of NAMEP and place a \U escape in BUFFER.
1437 Return the number of bytes read out of NAME. (There are always
1438 10 bytes written to BUFFER.) */
1441 utf8_to_ucn (unsigned char *buffer, const unsigned char *name)
1447 unsigned long utf32;
1449 /* Compute the length of the UTF-8 sequence. */
1450 for (t = *name; t & 0x80; t <<= 1)
1453 utf32 = *name & (0x7F >> ucn_len);
1454 for (ucn_len_c = 1; ucn_len_c < ucn_len; ucn_len_c++)
1456 utf32 = (utf32 << 6) | (*++name & 0x3F);
1458 /* Ill-formed UTF-8. */
1459 if ((*name & ~0x3F) != 0x80)
1465 for (j = 7; j >= 0; j--)
1466 *buffer++ = "0123456789abcdef"[(utf32 >> (4 * j)) & 0xF];
1471 /* Write the spelling of a token TOKEN to BUFFER. The buffer must
1472 already contain the enough space to hold the token's spelling.
1473 Returns a pointer to the character after the last character written.
1474 FORSTRING is true if this is to be the spelling after translation
1475 phase 1 (this is different for UCNs).
1476 FIXME: Would be nice if we didn't need the PFILE argument. */
1478 cpp_spell_token (cpp_reader *pfile, const cpp_token *token,
1479 unsigned char *buffer, bool forstring)
1481 switch (TOKEN_SPELL (token))
1483 case SPELL_OPERATOR:
1485 const unsigned char *spelling;
1488 if (token->flags & DIGRAPH)
1490 = digraph_spellings[(int) token->type - (int) CPP_FIRST_DIGRAPH];
1491 else if (token->flags & NAMED_OP)
1494 spelling = TOKEN_NAME (token);
1496 while ((c = *spelling++) != '\0')
1505 memcpy (buffer, NODE_NAME (token->val.node),
1506 NODE_LEN (token->val.node));
1507 buffer += NODE_LEN (token->val.node);
1512 const unsigned char * name = NODE_NAME (token->val.node);
1514 for (i = 0; i < NODE_LEN (token->val.node); i++)
1515 if (name[i] & ~0x7F)
1517 i += utf8_to_ucn (buffer, name + i) - 1;
1521 *buffer++ = NODE_NAME (token->val.node)[i];
1526 memcpy (buffer, token->val.str.text, token->val.str.len);
1527 buffer += token->val.str.len;
1531 cpp_error (pfile, CPP_DL_ICE,
1532 "unspellable token %s", TOKEN_NAME (token));
1539 /* Returns TOKEN spelt as a null-terminated string. The string is
1540 freed when the reader is destroyed. Useful for diagnostics. */
1542 cpp_token_as_text (cpp_reader *pfile, const cpp_token *token)
1544 unsigned int len = cpp_token_len (token) + 1;
1545 unsigned char *start = _cpp_unaligned_alloc (pfile, len), *end;
1547 end = cpp_spell_token (pfile, token, start, false);
1553 /* Used by C front ends, which really should move to using
1554 cpp_token_as_text. */
1556 cpp_type2name (enum cpp_ttype type)
1558 return (const char *) token_spellings[type].name;
1561 /* Writes the spelling of token to FP, without any preceding space.
1562 Separated from cpp_spell_token for efficiency - to avoid stdio
1563 double-buffering. */
1565 cpp_output_token (const cpp_token *token, FILE *fp)
1567 switch (TOKEN_SPELL (token))
1569 case SPELL_OPERATOR:
1571 const unsigned char *spelling;
1574 if (token->flags & DIGRAPH)
1576 = digraph_spellings[(int) token->type - (int) CPP_FIRST_DIGRAPH];
1577 else if (token->flags & NAMED_OP)
1580 spelling = TOKEN_NAME (token);
1585 while ((c = *++spelling) != '\0');
1593 const unsigned char * name = NODE_NAME (token->val.node);
1595 for (i = 0; i < NODE_LEN (token->val.node); i++)
1596 if (name[i] & ~0x7F)
1598 unsigned char buffer[10];
1599 i += utf8_to_ucn (buffer, name + i) - 1;
1600 fwrite (buffer, 1, 10, fp);
1603 fputc (NODE_NAME (token->val.node)[i], fp);
1608 fwrite (token->val.str.text, 1, token->val.str.len, fp);
1612 /* An error, most probably. */
1617 /* Compare two tokens. */
1619 _cpp_equiv_tokens (const cpp_token *a, const cpp_token *b)
1621 if (a->type == b->type && a->flags == b->flags)
1622 switch (TOKEN_SPELL (a))
1624 default: /* Keep compiler happy. */
1625 case SPELL_OPERATOR:
1628 return (a->type != CPP_MACRO_ARG || a->val.arg_no == b->val.arg_no);
1630 return a->val.node == b->val.node;
1632 return (a->val.str.len == b->val.str.len
1633 && !memcmp (a->val.str.text, b->val.str.text,
1640 /* Returns nonzero if a space should be inserted to avoid an
1641 accidental token paste for output. For simplicity, it is
1642 conservative, and occasionally advises a space where one is not
1643 needed, e.g. "." and ".2". */
1645 cpp_avoid_paste (cpp_reader *pfile, const cpp_token *token1,
1646 const cpp_token *token2)
1648 enum cpp_ttype a = token1->type, b = token2->type;
1651 if (token1->flags & NAMED_OP)
1653 if (token2->flags & NAMED_OP)
1657 if (token2->flags & DIGRAPH)
1658 c = digraph_spellings[(int) b - (int) CPP_FIRST_DIGRAPH][0];
1659 else if (token_spellings[b].category == SPELL_OPERATOR)
1660 c = token_spellings[b].name[0];
1662 /* Quickly get everything that can paste with an '='. */
1663 if ((int) a <= (int) CPP_LAST_EQ && c == '=')
1668 case CPP_GREATER: return c == '>';
1669 case CPP_LESS: return c == '<' || c == '%' || c == ':';
1670 case CPP_PLUS: return c == '+';
1671 case CPP_MINUS: return c == '-' || c == '>';
1672 case CPP_DIV: return c == '/' || c == '*'; /* Comments. */
1673 case CPP_MOD: return c == ':' || c == '>';
1674 case CPP_AND: return c == '&';
1675 case CPP_OR: return c == '|';
1676 case CPP_COLON: return c == ':' || c == '>';
1677 case CPP_DEREF: return c == '*';
1678 case CPP_DOT: return c == '.' || c == '%' || b == CPP_NUMBER;
1679 case CPP_HASH: return c == '#' || c == '%'; /* Digraph form. */
1680 case CPP_NAME: return ((b == CPP_NUMBER
1681 && name_p (pfile, &token2->val.str))
1683 || b == CPP_CHAR || b == CPP_STRING); /* L */
1684 case CPP_NUMBER: return (b == CPP_NUMBER || b == CPP_NAME
1685 || c == '.' || c == '+' || c == '-');
1687 case CPP_OTHER: return ((token1->val.str.text[0] == '\\'
1689 || (CPP_OPTION (pfile, objc)
1690 && token1->val.str.text[0] == '@'
1691 && (b == CPP_NAME || b == CPP_STRING)));
1698 /* Output all the remaining tokens on the current line, and a newline
1699 character, to FP. Leading whitespace is removed. If there are
1700 macros, special token padding is not performed. */
1702 cpp_output_line (cpp_reader *pfile, FILE *fp)
1704 const cpp_token *token;
1706 token = cpp_get_token (pfile);
1707 while (token->type != CPP_EOF)
1709 cpp_output_token (token, fp);
1710 token = cpp_get_token (pfile);
1711 if (token->flags & PREV_WHITE)
1718 /* Return a string representation of all the remaining tokens on the
1719 current line. The result is allocated using xmalloc and must be
1720 freed by the caller. */
1722 cpp_output_line_to_string (cpp_reader *pfile, const unsigned char *dir_name)
1724 const cpp_token *token;
1725 unsigned int out = dir_name ? ustrlen (dir_name) : 0;
1726 unsigned int alloced = 120 + out;
1727 unsigned char *result = (unsigned char *) xmalloc (alloced);
1729 /* If DIR_NAME is empty, there are no initial contents. */
1732 sprintf ((char *) result, "#%s ", dir_name);
1736 token = cpp_get_token (pfile);
1737 while (token->type != CPP_EOF)
1739 unsigned char *last;
1740 /* Include room for a possible space and the terminating nul. */
1741 unsigned int len = cpp_token_len (token) + 2;
1743 if (out + len > alloced)
1746 if (out + len > alloced)
1747 alloced = out + len;
1748 result = (unsigned char *) xrealloc (result, alloced);
1751 last = cpp_spell_token (pfile, token, &result[out], 0);
1752 out = last - result;
1754 token = cpp_get_token (pfile);
1755 if (token->flags & PREV_WHITE)
1756 result[out++] = ' ';
1763 /* Memory buffers. Changing these three constants can have a dramatic
1764 effect on performance. The values here are reasonable defaults,
1765 but might be tuned. If you adjust them, be sure to test across a
1766 range of uses of cpplib, including heavy nested function-like macro
1767 expansion. Also check the change in peak memory usage (NJAMD is a
1768 good tool for this). */
1769 #define MIN_BUFF_SIZE 8000
1770 #define BUFF_SIZE_UPPER_BOUND(MIN_SIZE) (MIN_BUFF_SIZE + (MIN_SIZE) * 3 / 2)
1771 #define EXTENDED_BUFF_SIZE(BUFF, MIN_EXTRA) \
1772 (MIN_EXTRA + ((BUFF)->limit - (BUFF)->cur) * 2)
1774 #if MIN_BUFF_SIZE > BUFF_SIZE_UPPER_BOUND (0)
1775 #error BUFF_SIZE_UPPER_BOUND must be at least as large as MIN_BUFF_SIZE!
1778 /* Create a new allocation buffer. Place the control block at the end
1779 of the buffer, so that buffer overflows will cause immediate chaos. */
1781 new_buff (size_t len)
1784 unsigned char *base;
1786 if (len < MIN_BUFF_SIZE)
1787 len = MIN_BUFF_SIZE;
1788 len = CPP_ALIGN (len);
1790 base = XNEWVEC (unsigned char, len + sizeof (_cpp_buff));
1791 result = (_cpp_buff *) (base + len);
1792 result->base = base;
1794 result->limit = base + len;
1795 result->next = NULL;
1799 /* Place a chain of unwanted allocation buffers on the free list. */
1801 _cpp_release_buff (cpp_reader *pfile, _cpp_buff *buff)
1803 _cpp_buff *end = buff;
1807 end->next = pfile->free_buffs;
1808 pfile->free_buffs = buff;
1811 /* Return a free buffer of size at least MIN_SIZE. */
1813 _cpp_get_buff (cpp_reader *pfile, size_t min_size)
1815 _cpp_buff *result, **p;
1817 for (p = &pfile->free_buffs;; p = &(*p)->next)
1822 return new_buff (min_size);
1824 size = result->limit - result->base;
1825 /* Return a buffer that's big enough, but don't waste one that's
1827 if (size >= min_size && size <= BUFF_SIZE_UPPER_BOUND (min_size))
1832 result->next = NULL;
1833 result->cur = result->base;
1837 /* Creates a new buffer with enough space to hold the uncommitted
1838 remaining bytes of BUFF, and at least MIN_EXTRA more bytes. Copies
1839 the excess bytes to the new buffer. Chains the new buffer after
1840 BUFF, and returns the new buffer. */
1842 _cpp_append_extend_buff (cpp_reader *pfile, _cpp_buff *buff, size_t min_extra)
1844 size_t size = EXTENDED_BUFF_SIZE (buff, min_extra);
1845 _cpp_buff *new_buff = _cpp_get_buff (pfile, size);
1847 buff->next = new_buff;
1848 memcpy (new_buff->base, buff->cur, BUFF_ROOM (buff));
1852 /* Creates a new buffer with enough space to hold the uncommitted
1853 remaining bytes of the buffer pointed to by BUFF, and at least
1854 MIN_EXTRA more bytes. Copies the excess bytes to the new buffer.
1855 Chains the new buffer before the buffer pointed to by BUFF, and
1856 updates the pointer to point to the new buffer. */
1858 _cpp_extend_buff (cpp_reader *pfile, _cpp_buff **pbuff, size_t min_extra)
1860 _cpp_buff *new_buff, *old_buff = *pbuff;
1861 size_t size = EXTENDED_BUFF_SIZE (old_buff, min_extra);
1863 new_buff = _cpp_get_buff (pfile, size);
1864 memcpy (new_buff->base, old_buff->cur, BUFF_ROOM (old_buff));
1865 new_buff->next = old_buff;
1869 /* Free a chain of buffers starting at BUFF. */
1871 _cpp_free_buff (_cpp_buff *buff)
1875 for (; buff; buff = next)
1882 /* Allocate permanent, unaligned storage of length LEN. */
1884 _cpp_unaligned_alloc (cpp_reader *pfile, size_t len)
1886 _cpp_buff *buff = pfile->u_buff;
1887 unsigned char *result = buff->cur;
1889 if (len > (size_t) (buff->limit - result))
1891 buff = _cpp_get_buff (pfile, len);
1892 buff->next = pfile->u_buff;
1893 pfile->u_buff = buff;
1897 buff->cur = result + len;
1901 /* Allocate permanent, unaligned storage of length LEN from a_buff.
1902 That buffer is used for growing allocations when saving macro
1903 replacement lists in a #define, and when parsing an answer to an
1904 assertion in #assert, #unassert or #if (and therefore possibly
1905 whilst expanding macros). It therefore must not be used by any
1906 code that they might call: specifically the lexer and the guts of
1909 All existing other uses clearly fit this restriction: storing
1910 registered pragmas during initialization. */
1912 _cpp_aligned_alloc (cpp_reader *pfile, size_t len)
1914 _cpp_buff *buff = pfile->a_buff;
1915 unsigned char *result = buff->cur;
1917 if (len > (size_t) (buff->limit - result))
1919 buff = _cpp_get_buff (pfile, len);
1920 buff->next = pfile->a_buff;
1921 pfile->a_buff = buff;
1925 buff->cur = result + len;
1929 /* Say which field of TOK is in use. */
1931 enum cpp_token_fld_kind
1932 cpp_token_val_index (cpp_token *tok)
1934 switch (TOKEN_SPELL (tok))
1937 return CPP_TOKEN_FLD_NODE;
1939 return CPP_TOKEN_FLD_STR;
1941 if (tok->type == CPP_MACRO_ARG)
1942 return CPP_TOKEN_FLD_ARG_NO;
1943 else if (tok->type == CPP_PADDING)
1944 return CPP_TOKEN_FLD_SOURCE;
1945 else if (tok->type == CPP_PRAGMA)
1946 return CPP_TOKEN_FLD_PRAGMA;
1947 /* else fall through */
1949 return CPP_TOKEN_FLD_NONE;