/* CPP Library - lexical analysis.
- Copyright (C) 2000 Free Software Foundation, Inc.
+ Copyright (C) 2000, 2001, 2002 Free Software Foundation, Inc.
Contributed by Per Bothner, 1994-95.
Based on CCCP program by Paul Rubin, June 1986
Adapted to ANSI C, Richard Stallman, Jan 1987
along with this program; if not, write to the Free Software
Foundation, 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. */
-/* This lexer works with a single pass of the file. Recently I
- re-wrote it to minimize the places where we step backwards in the
- input stream, to make future changes to support multi-byte
- character sets fairly straight-forward.
-
- There is now only one routine where we do step backwards:
- skip_escaped_newlines. This routine could probably also be changed
- so that it doesn't need to step back. One possibility is to use a
- trick similar to that used in lex_period and lex_percent. Two
- extra characters might be needed, but skip_escaped_newlines itself
- would probably be the only place that needs to be aware of that,
- and changes to the remaining routines would probably only be needed
- if they process a backslash. */
-
#include "config.h"
#include "system.h"
#include "cpplib.h"
#include "cpphash.h"
-#include "symcat.h"
+
+/* MULTIBYTE_CHARS support only works for native compilers.
+ ??? Ideally what we want is to model widechar support after
+ the current floating point support. */
+#ifdef CROSS_COMPILE
+#undef MULTIBYTE_CHARS
+#endif
+
+#ifdef MULTIBYTE_CHARS
+#include "mbchar.h"
+#include <locale.h>
+#endif
/* Tokens with SPELL_STRING store their spelling in the token list,
and it's length in the token->val.name.len. */
SPELL_OPERATOR = 0,
SPELL_CHAR,
SPELL_IDENT,
+ SPELL_NUMBER,
SPELL_STRING,
SPELL_NONE
};
const unsigned char *name;
};
-const unsigned char *digraph_spellings [] = {U"%:", U"%:%:", U"<:",
- U":>", U"<%", U"%>"};
+static const unsigned char *const digraph_spellings[] =
+{ U"%:", U"%:%:", U"<:", U":>", U"<%", U"%>" };
#define OP(e, s) { SPELL_OPERATOR, U s },
#define TK(e, s) { s, U STRINGX (e) },
-const struct token_spelling token_spellings [N_TTYPES] = {TTYPE_TABLE };
+static const struct token_spelling token_spellings[N_TTYPES] = { TTYPE_TABLE };
#undef OP
#undef TK
#define TOKEN_SPELL(token) (token_spellings[(token)->type].category)
#define TOKEN_NAME(token) (token_spellings[(token)->type].name)
+#define BACKUP() do {buffer->cur = buffer->backup_to;} while (0)
-static cppchar_t handle_newline PARAMS ((cpp_buffer *, cppchar_t));
-static cppchar_t skip_escaped_newlines PARAMS ((cpp_buffer *, cppchar_t));
-static cppchar_t get_effective_char PARAMS ((cpp_buffer *));
+static void handle_newline PARAMS ((cpp_reader *));
+static cppchar_t skip_escaped_newlines PARAMS ((cpp_reader *));
+static cppchar_t get_effective_char PARAMS ((cpp_reader *));
static int skip_block_comment PARAMS ((cpp_reader *));
static int skip_line_comment PARAMS ((cpp_reader *));
static void adjust_column PARAMS ((cpp_reader *));
-static void skip_whitespace PARAMS ((cpp_reader *, cppchar_t));
-static cpp_hashnode *parse_identifier PARAMS ((cpp_reader *, cppchar_t));
+static int skip_whitespace PARAMS ((cpp_reader *, cppchar_t));
+static cpp_hashnode *parse_identifier PARAMS ((cpp_reader *));
+static cpp_hashnode *parse_identifier_slow PARAMS ((cpp_reader *,
+ const U_CHAR *));
static void parse_number PARAMS ((cpp_reader *, cpp_string *, cppchar_t, int));
static int unescaped_terminator_p PARAMS ((cpp_reader *, const U_CHAR *));
static void parse_string PARAMS ((cpp_reader *, cpp_token *, cppchar_t));
static void unterminated PARAMS ((cpp_reader *, int));
-static int trigraph_ok PARAMS ((cpp_reader *, cppchar_t));
+static bool trigraph_p PARAMS ((cpp_reader *));
static void save_comment PARAMS ((cpp_reader *, cpp_token *, const U_CHAR *));
-static void lex_percent PARAMS ((cpp_buffer *, cpp_token *));
-static void lex_dot PARAMS ((cpp_reader *, cpp_token *));
static int name_p PARAMS ((cpp_reader *, const cpp_string *));
+static int maybe_read_ucs PARAMS ((cpp_reader *, const unsigned char **,
+ const unsigned char *, unsigned int *));
+static tokenrun *next_tokenrun PARAMS ((tokenrun *));
-static cpp_chunk *new_chunk PARAMS ((unsigned int));
-static int chunk_suitable PARAMS ((cpp_pool *, cpp_chunk *, unsigned int));
+static unsigned int hex_digit_value PARAMS ((unsigned int));
+static _cpp_buff *new_buff PARAMS ((size_t));
/* Utility routine:
Compares, the token TOKEN to the NUL-terminated string STRING.
TOKEN must be a CPP_NAME. Returns 1 for equal, 0 for unequal. */
-
int
cpp_ideq (token, string)
const cpp_token *token;
if (token->type != CPP_NAME)
return 0;
- return !ustrcmp (token->val.node->name, (const U_CHAR *) string);
+ return !ustrcmp (NODE_NAME (token->val.node), (const U_CHAR *) string);
}
-/* Call when meeting a newline. Returns the character after the newline
- (or carriage-return newline combination), or EOF. */
-static cppchar_t
-handle_newline (buffer, newline_char)
- cpp_buffer *buffer;
- cppchar_t newline_char;
+/* Call when meeting a newline, assumed to be in buffer->cur[-1].
+ Returns with buffer->cur pointing to the character immediately
+ following the newline (combination). */
+static void
+handle_newline (pfile)
+ cpp_reader *pfile;
{
- cppchar_t next = EOF;
-
- buffer->col_adjust = 0;
- buffer->lineno++;
- buffer->line_base = buffer->cur;
+ cpp_buffer *buffer = pfile->buffer;
- /* Handle CR-LF and LF-CR combinations, get the next character. */
- if (buffer->cur < buffer->rlimit)
- {
- next = *buffer->cur++;
- if (next + newline_char == '\r' + '\n')
- {
- buffer->line_base = buffer->cur;
- if (buffer->cur < buffer->rlimit)
- next = *buffer->cur++;
- else
- next = EOF;
- }
- }
+ /* Handle CR-LF and LF-CR. Most other implementations (e.g. java)
+ only accept CR-LF; maybe we should fall back to that behaviour? */
+ if (buffer->cur[-1] + buffer->cur[0] == '\r' + '\n')
+ buffer->cur++;
- buffer->read_ahead = next;
- return next;
+ buffer->line_base = buffer->cur;
+ buffer->col_adjust = 0;
+ pfile->line++;
}
-/* Subroutine of skip_escaped_newlines; called when a trigraph is
- encountered. It warns if necessary, and returns true if the
- trigraph should be honoured. FROM_CHAR is the third character of a
- trigraph, and presumed to be the previous character for position
- reporting. */
-static int
-trigraph_ok (pfile, from_char)
+/* Subroutine of skip_escaped_newlines; called when a 3-character
+ sequence beginning with "??" is encountered. buffer->cur points to
+ the second '?'.
+
+ Warn if necessary, and returns true if the sequence forms a
+ trigraph and the trigraph should be honoured. */
+static bool
+trigraph_p (pfile)
cpp_reader *pfile;
- cppchar_t from_char;
{
- int accept = CPP_OPTION (pfile, trigraphs);
-
+ cpp_buffer *buffer = pfile->buffer;
+ cppchar_t from_char = buffer->cur[1];
+ bool accept;
+
+ if (!_cpp_trigraph_map[from_char])
+ return false;
+
+ accept = CPP_OPTION (pfile, trigraphs);
+
/* Don't warn about trigraphs in comments. */
if (CPP_OPTION (pfile, warn_trigraphs) && !pfile->state.lexing_comment)
{
- cpp_buffer *buffer = pfile->buffer;
if (accept)
- cpp_warning_with_line (pfile, buffer->lineno, CPP_BUF_COL (buffer) - 2,
+ cpp_warning_with_line (pfile, pfile->line, CPP_BUF_COL (buffer) - 1,
"trigraph ??%c converted to %c",
(int) from_char,
(int) _cpp_trigraph_map[from_char]);
else if (buffer->cur != buffer->last_Wtrigraphs)
{
buffer->last_Wtrigraphs = buffer->cur;
- cpp_warning_with_line (pfile, buffer->lineno,
- CPP_BUF_COL (buffer) - 2,
+ cpp_warning_with_line (pfile, pfile->line,
+ CPP_BUF_COL (buffer) - 1,
"trigraph ??%c ignored", (int) from_char);
}
}
return accept;
}
-/* Assumes local variables buffer and result. */
-#define ACCEPT_CHAR(t) \
- do { result->type = t; buffer->read_ahead = EOF; } while (0)
-
-/* When we move to multibyte character sets, add to these something
- that saves and restores the state of the multibyte conversion
- library. This probably involves saving and restoring a "cookie".
- In the case of glibc it is an 8-byte structure, so is not a high
- overhead operation. In any case, it's out of the fast path. */
-#define SAVE_STATE() do { saved_cur = buffer->cur; } while (0)
-#define RESTORE_STATE() do { buffer->cur = saved_cur; } while (0)
-
-/* Skips any escaped newlines introduced by NEXT, which is either a
- '?' or a '\\'. Returns the next character, which will also have
- been placed in buffer->read_ahead. This routine performs
- preprocessing stages 1 and 2 of the ISO C standard. */
+/* Skips any escaped newlines introduced by '?' or a '\\', assumed to
+ lie in buffer->cur[-1]. Returns the next byte, which will be in
+ buffer->cur[-1]. This routine performs preprocessing stages 1 and
+ 2 of the ISO C standard. */
static cppchar_t
-skip_escaped_newlines (buffer, next)
- cpp_buffer *buffer;
- cppchar_t next;
+skip_escaped_newlines (pfile)
+ cpp_reader *pfile;
{
+ cpp_buffer *buffer = pfile->buffer;
+ cppchar_t next = buffer->cur[-1];
+
/* Only do this if we apply stages 1 and 2. */
if (!buffer->from_stage3)
{
- cppchar_t next1;
const unsigned char *saved_cur;
- int space;
+ cppchar_t next1;
do
{
- if (buffer->cur == buffer->rlimit)
- break;
-
- SAVE_STATE ();
if (next == '?')
{
- next1 = *buffer->cur++;
- if (next1 != '?' || buffer->cur == buffer->rlimit)
- {
- RESTORE_STATE ();
- break;
- }
-
- next1 = *buffer->cur++;
- if (!_cpp_trigraph_map[next1]
- || !trigraph_ok (buffer->pfile, next1))
- {
- RESTORE_STATE ();
- break;
- }
-
- /* We have a full trigraph here. */
- next = _cpp_trigraph_map[next1];
- if (next != '\\' || buffer->cur == buffer->rlimit)
+ if (buffer->cur[0] != '?' || !trigraph_p (pfile))
break;
- SAVE_STATE ();
- }
- /* We have a backslash, and room for at least one more character. */
- space = 0;
- do
- {
- next1 = *buffer->cur++;
- if (!is_nvspace (next1))
+ /* Translate the trigraph. */
+ next = _cpp_trigraph_map[buffer->cur[1]];
+ buffer->cur += 2;
+ if (next != '\\')
break;
- space = 1;
}
- while (buffer->cur < buffer->rlimit);
+
+ if (buffer->cur == buffer->rlimit)
+ break;
+
+ /* We have a backslash, and room for at least one more
+ character. Skip horizontal whitespace. */
+ saved_cur = buffer->cur;
+ do
+ next1 = *buffer->cur++;
+ while (is_nvspace (next1) && buffer->cur < buffer->rlimit);
if (!is_vspace (next1))
{
- RESTORE_STATE ();
+ buffer->cur = saved_cur;
break;
}
- if (space && !buffer->pfile->state.lexing_comment)
- cpp_warning (buffer->pfile,
- "backslash and newline separated by space");
+ if (saved_cur != buffer->cur - 1
+ && !pfile->state.lexing_comment)
+ cpp_warning (pfile, "backslash and newline separated by space");
- next = handle_newline (buffer, next1);
- if (next == EOF)
- cpp_pedwarn (buffer->pfile, "backslash-newline at end of file");
+ handle_newline (pfile);
+ buffer->backup_to = buffer->cur;
+ if (buffer->cur == buffer->rlimit)
+ {
+ cpp_pedwarn (pfile, "backslash-newline at end of file");
+ next = EOF;
+ }
+ else
+ next = *buffer->cur++;
}
while (next == '\\' || next == '?');
}
- buffer->read_ahead = next;
return next;
}
/* Obtain the next character, after trigraph conversion and skipping
- an arbitrary string of escaped newlines. The common case of no
- trigraphs or escaped newlines falls through quickly. */
+ an arbitrarily long string of escaped newlines. The common case of
+ no trigraphs or escaped newlines falls through quickly. On return,
+ buffer->backup_to points to where to return to if the character is
+ not to be processed. */
static cppchar_t
-get_effective_char (buffer)
- cpp_buffer *buffer;
+get_effective_char (pfile)
+ cpp_reader *pfile;
{
- cppchar_t next = EOF;
+ cppchar_t next;
+ cpp_buffer *buffer = pfile->buffer;
- if (buffer->cur < buffer->rlimit)
- {
- next = *buffer->cur++;
-
- /* '?' can introduce trigraphs (and therefore backslash); '\\'
- can introduce escaped newlines, which we want to skip, or
- UCNs, which, depending upon lexer state, we will handle in
- the future. */
- if (next == '?' || next == '\\')
- next = skip_escaped_newlines (buffer, next);
- }
+ buffer->backup_to = buffer->cur;
+ next = *buffer->cur++;
+ if (__builtin_expect (next == '?' || next == '\\', 0))
+ next = skip_escaped_newlines (pfile);
- buffer->read_ahead = next;
- return next;
+ return next;
}
/* Skip a C-style block comment. We find the end of the comment by
{
prevc = c, c = *buffer->cur++;
- next_char:
/* FIXME: For speed, create a new character class of characters
of interest inside block comments. */
if (c == '?' || c == '\\')
- c = skip_escaped_newlines (buffer, c);
+ c = skip_escaped_newlines (pfile);
/* People like decorating comments with '*', so check for '/'
instead for efficiency. */
break;
/* Warn about potential nested comments, but not if the '/'
- comes immediately before the true comment delimeter.
+ comes immediately before the true comment delimiter.
Don't bother to get it right across escaped newlines. */
if (CPP_OPTION (pfile, warn_comments)
- && buffer->cur != buffer->rlimit)
- {
- prevc = c, c = *buffer->cur++;
- if (c == '*' && buffer->cur != buffer->rlimit)
- {
- prevc = c, c = *buffer->cur++;
- if (c != '/')
- cpp_warning_with_line (pfile, CPP_BUF_LINE (buffer),
- CPP_BUF_COL (buffer),
- "\"/*\" within comment");
- }
- goto next_char;
- }
+ && buffer->cur[0] == '*' && buffer->cur[1] != '/')
+ cpp_warning_with_line (pfile,
+ pfile->line, CPP_BUF_COL (buffer),
+ "\"/*\" within comment");
}
else if (is_vspace (c))
- {
- prevc = c, c = handle_newline (buffer, c);
- goto next_char;
- }
+ handle_newline (pfile);
else if (c == '\t')
adjust_column (pfile);
}
pfile->state.lexing_comment = 0;
- buffer->read_ahead = EOF;
return c != '/' || prevc != '*';
}
-/* Skip a C++ line comment. Handles escaped newlines. Returns
- non-zero if a multiline comment. The following new line, if any,
- is left in buffer->read_ahead. */
+/* Skip a C++ line comment, leaving buffer->cur pointing to the
+ terminating newline. Handles escaped newlines. Returns non-zero
+ if a multiline comment. */
static int
skip_line_comment (pfile)
cpp_reader *pfile;
{
cpp_buffer *buffer = pfile->buffer;
- unsigned int orig_lineno = buffer->lineno;
+ unsigned int orig_line = pfile->line;
cppchar_t c;
pfile->state.lexing_comment = 1;
do
{
- c = EOF;
if (buffer->cur == buffer->rlimit)
- break;
+ goto at_eof;
c = *buffer->cur++;
if (c == '?' || c == '\\')
- c = skip_escaped_newlines (buffer, c);
+ c = skip_escaped_newlines (pfile);
}
while (!is_vspace (c));
+ /* Step back over the newline, except at EOF. */
+ buffer->cur--;
+ at_eof:
+
pfile->state.lexing_comment = 0;
- buffer->read_ahead = c; /* Leave any newline for caller. */
- return orig_lineno != buffer->lineno;
+ return orig_line != pfile->line;
}
/* pfile->buffer->cur is one beyond the \t character. Update
/* Skips whitespace, saving the next non-whitespace character.
Adjusts pfile->col_adjust to account for tabs. Without this,
tokens might be assigned an incorrect column. */
-static void
+static int
skip_whitespace (pfile, c)
cpp_reader *pfile;
cppchar_t c;
/* Just \f \v or \0 left. */
else if (c == '\0')
{
+ if (buffer->cur - 1 == buffer->rlimit)
+ return 0;
if (!warned)
{
cpp_warning (pfile, "null character(s) ignored");
}
}
else if (pfile->state.in_directive && CPP_PEDANTIC (pfile))
- cpp_pedwarn_with_line (pfile, CPP_BUF_LINE (buffer),
+ cpp_pedwarn_with_line (pfile, pfile->line,
CPP_BUF_COL (buffer),
"%s in preprocessing directive",
c == '\f' ? "form feed" : "vertical tab");
- c = EOF;
- if (buffer->cur == buffer->rlimit)
- break;
c = *buffer->cur++;
}
- /* We only want non-vertical space, i.e. ' ' \t \f \v \0. */
+ /* We only want non-vertical space, i.e. ' ' \t \f \v \0. */
while (is_nvspace (c));
- /* Remember the next character. */
- buffer->read_ahead = c;
+ buffer->cur--;
+ return 1;
}
/* See if the characters of a number token are valid in a name (no
return 1;
}
-/* Parse an identifier, skipping embedded backslash-newlines.
- Calculate the hash value of the token while parsing, for improved
- performance. The hashing algorithm *must* match cpp_lookup(). */
-
+/* Parse an identifier, skipping embedded backslash-newlines. This is
+ a critical inner loop. The common case is an identifier which has
+ not been split by backslash-newline, does not contain a dollar
+ sign, and has already been scanned (roughly 10:1 ratio of
+ seen:unseen identifiers in normal code; the distribution is
+ Poisson-like). Second most common case is a new identifier, not
+ split and no dollar sign. The other possibilities are rare and
+ have been relegated to parse_identifier_slow. */
static cpp_hashnode *
-parse_identifier (pfile, c)
+parse_identifier (pfile)
cpp_reader *pfile;
- cppchar_t c;
{
cpp_hashnode *result;
- cpp_buffer *buffer = pfile->buffer;
- unsigned char *dest, *limit;
- unsigned int r = 0, saw_dollar = 0;
+ const U_CHAR *cur;
- dest = POOL_FRONT (&pfile->ident_pool);
- limit = POOL_LIMIT (&pfile->ident_pool);
+ /* Fast-path loop. Skim over a normal identifier.
+ N.B. ISIDNUM does not include $. */
+ cur = pfile->buffer->cur;
+ while (ISIDNUM (*cur))
+ cur++;
- do
+ /* Check for slow-path cases. */
+ if (*cur == '?' || *cur == '\\' || *cur == '$')
+ result = parse_identifier_slow (pfile, cur);
+ else
{
- do
- {
- /* Need room for terminating null. */
- if (dest + 1 >= limit)
- limit = _cpp_next_chunk (&pfile->ident_pool, 0, &dest);
+ const U_CHAR *base = pfile->buffer->cur - 1;
+ result = (cpp_hashnode *)
+ ht_lookup (pfile->hash_table, base, cur - base, HT_ALLOC);
+ pfile->buffer->cur = cur;
+ }
- *dest++ = c;
- r = HASHSTEP (r, c);
+ /* Rarely, identifiers require diagnostics when lexed.
+ XXX Has to be forced out of the fast path. */
+ if (__builtin_expect ((result->flags & NODE_DIAGNOSTIC)
+ && !pfile->state.skipping, 0))
+ {
+ /* It is allowed to poison the same identifier twice. */
+ if ((result->flags & NODE_POISONED) && !pfile->state.poisoned_ok)
+ cpp_error (pfile, "attempt to use poisoned \"%s\"",
+ NODE_NAME (result));
- if (c == '$')
- saw_dollar++;
+ /* Constraint 6.10.3.5: __VA_ARGS__ should only appear in the
+ replacement list of a variadic macro. */
+ if (result == pfile->spec_nodes.n__VA_ARGS__
+ && !pfile->state.va_args_ok)
+ cpp_pedwarn (pfile,
+ "__VA_ARGS__ can only appear in the expansion of a C99 variadic macro");
+ }
- c = EOF;
- if (buffer->cur == buffer->rlimit)
- break;
+ return result;
+}
- c = *buffer->cur++;
- }
- while (is_idchar (c));
+/* Slow path. This handles identifiers which have been split, and
+ identifiers which contain dollar signs. The part of the identifier
+ from PFILE->buffer->cur-1 to CUR has already been scanned. */
+static cpp_hashnode *
+parse_identifier_slow (pfile, cur)
+ cpp_reader *pfile;
+ const U_CHAR *cur;
+{
+ cpp_buffer *buffer = pfile->buffer;
+ const U_CHAR *base = buffer->cur - 1;
+ struct obstack *stack = &pfile->hash_table->stack;
+ unsigned int c, saw_dollar = 0, len;
+
+ /* Copy the part of the token which is known to be okay. */
+ obstack_grow (stack, base, cur - base);
+
+ /* Now process the part which isn't. We are looking at one of
+ '$', '\\', or '?' on entry to this loop. */
+ c = *cur++;
+ buffer->cur = cur;
+ do
+ {
+ while (is_idchar (c))
+ {
+ obstack_1grow (stack, c);
+
+ if (c == '$')
+ saw_dollar++;
+
+ c = *buffer->cur++;
+ }
/* Potential escaped newline? */
+ buffer->backup_to = buffer->cur - 1;
if (c != '?' && c != '\\')
- break;
- c = skip_escaped_newlines (buffer, c);
+ break;
+ c = skip_escaped_newlines (pfile);
}
while (is_idchar (c));
- /* Remember the next character. */
- buffer->read_ahead = c;
+ /* Step back over the unwanted char. */
+ BACKUP ();
- /* $ is not a identifier character in the standard, but is commonly
+ /* $ is not an identifier character in the standard, but is commonly
accepted as an extension. Don't warn about it in skipped
conditional blocks. */
- if (saw_dollar && CPP_PEDANTIC (pfile) && ! pfile->skipping)
+ if (saw_dollar && CPP_PEDANTIC (pfile) && ! pfile->state.skipping)
cpp_pedwarn (pfile, "'$' character(s) in identifier");
/* Identifiers are null-terminated. */
- *dest = '\0';
-
- /* This routine commits the memory if necessary. */
- result = _cpp_lookup_with_hash (pfile,
- dest - POOL_FRONT (&pfile->ident_pool), r);
-
- /* Some identifiers require diagnostics when lexed. */
- if (result->flags & NODE_DIAGNOSTIC && !pfile->skipping)
- {
- /* It is allowed to poison the same identifier twice. */
- if ((result->flags & NODE_POISONED) && !pfile->state.poisoned_ok)
- cpp_error (pfile, "attempt to use poisoned \"%s\"", result->name);
-
- /* Constraint 6.10.3.5: __VA_ARGS__ should only appear in the
- replacement list of a variadic macro. */
- if (result == pfile->spec_nodes.n__VA_ARGS__
- && !pfile->state.va_args_ok)
- cpp_pedwarn (pfile, "__VA_ARGS__ can only appear in the expansion of a C99 variadic macro");
- }
+ len = obstack_object_size (stack);
+ obstack_1grow (stack, '\0');
- return result;
+ return (cpp_hashnode *)
+ ht_lookup (pfile->hash_table, obstack_finish (stack), len, HT_ALLOCED);
}
-/* Parse a number, skipping embedded backslash-newlines. */
+/* Parse a number, beginning with character C, skipping embedded
+ backslash-newlines. LEADING_PERIOD is non-zero if there was a "."
+ before C. Place the result in NUMBER. */
static void
parse_number (pfile, number, c, leading_period)
cpp_reader *pfile;
int leading_period;
{
cpp_buffer *buffer = pfile->buffer;
- cpp_pool *pool = &pfile->ident_pool;
unsigned char *dest, *limit;
- dest = POOL_FRONT (pool);
- limit = POOL_LIMIT (pool);
+ dest = BUFF_FRONT (pfile->u_buff);
+ limit = BUFF_LIMIT (pfile->u_buff);
/* Place a leading period. */
if (leading_period)
{
- if (dest >= limit)
- limit = _cpp_next_chunk (pool, 0, &dest);
+ if (dest == limit)
+ {
+ _cpp_extend_buff (pfile, &pfile->u_buff, 1);
+ dest = BUFF_FRONT (pfile->u_buff);
+ limit = BUFF_LIMIT (pfile->u_buff);
+ }
*dest++ = '.';
}
do
{
/* Need room for terminating null. */
- if (dest + 1 >= limit)
- limit = _cpp_next_chunk (pool, 0, &dest);
+ if ((size_t) (limit - dest) < 2)
+ {
+ size_t len_so_far = dest - BUFF_FRONT (pfile->u_buff);
+ _cpp_extend_buff (pfile, &pfile->u_buff, 2);
+ dest = BUFF_FRONT (pfile->u_buff) + len_so_far;
+ limit = BUFF_LIMIT (pfile->u_buff);
+ }
*dest++ = c;
- c = EOF;
- if (buffer->cur == buffer->rlimit)
- break;
-
c = *buffer->cur++;
}
while (is_numchar (c) || c == '.' || VALID_SIGN (c, dest[-1]));
/* Potential escaped newline? */
+ buffer->backup_to = buffer->cur - 1;
if (c != '?' && c != '\\')
break;
- c = skip_escaped_newlines (buffer, c);
+ c = skip_escaped_newlines (pfile);
}
while (is_numchar (c) || c == '.' || VALID_SIGN (c, dest[-1]));
- /* Remember the next character. */
- buffer->read_ahead = c;
+ /* Step back over the unwanted char. */
+ BACKUP ();
/* Null-terminate the number. */
*dest = '\0';
- number->text = POOL_FRONT (pool);
+ number->text = BUFF_FRONT (pfile->u_buff);
number->len = dest - number->text;
- POOL_COMMIT (pool, number->len + 1);
+ BUFF_FRONT (pfile->u_buff) = dest + 1;
}
/* Subroutine of parse_string. Emits error for unterminated strings. */
{
cpp_error (pfile, "missing terminating %c character", term);
- if (term == '\"' && pfile->mlstring_pos.line
- && pfile->mlstring_pos.line != pfile->lexer_pos.line)
+ if (term == '\"' && pfile->mls_line && pfile->mls_line != pfile->line)
{
- cpp_error_with_line (pfile, pfile->mlstring_pos.line,
- pfile->mlstring_pos.col,
+ cpp_error_with_line (pfile, pfile->mls_line, pfile->mls_col,
"possible start of unterminated string literal");
- pfile->mlstring_pos.line = 0;
+ pfile->mls_line = 0;
}
}
if (pfile->state.angled_headers)
return 1;
- start = POOL_FRONT (&pfile->ident_pool);
+ start = BUFF_FRONT (pfile->u_buff);
/* An odd number of consecutive backslashes represents an escaped
terminator. */
}
/* Parses a string, character constant, or angle-bracketed header file
- name. Handles embedded trigraphs and escaped newlines.
+ name. Handles embedded trigraphs and escaped newlines. The stored
+ string is guaranteed NUL-terminated, but it is not guaranteed that
+ this is the first NUL since embedded NULs are preserved.
+ Multi-line strings are allowed, but they are deprecated.
- Multi-line strings are allowed, but they are deprecated within
- directives. */
+ When this function returns, buffer->cur points to the next
+ character to be processed. */
static void
parse_string (pfile, token, terminator)
cpp_reader *pfile;
cppchar_t terminator;
{
cpp_buffer *buffer = pfile->buffer;
- cpp_pool *pool = &pfile->ident_pool;
unsigned char *dest, *limit;
cppchar_t c;
- unsigned int nulls = 0;
+ bool warned_nulls = false, warned_multi = false;
- dest = POOL_FRONT (pool);
- limit = POOL_LIMIT (pool);
+ dest = BUFF_FRONT (pfile->u_buff);
+ limit = BUFF_LIMIT (pfile->u_buff);
for (;;)
{
- if (buffer->cur == buffer->rlimit)
+ /* We need room for another char, possibly the terminating NUL. */
+ if ((size_t) (limit - dest) < 1)
{
- c = EOF;
- unterminated (pfile, terminator);
- break;
+ size_t len_so_far = dest - BUFF_FRONT (pfile->u_buff);
+ _cpp_extend_buff (pfile, &pfile->u_buff, 2);
+ dest = BUFF_FRONT (pfile->u_buff) + len_so_far;
+ limit = BUFF_LIMIT (pfile->u_buff);
}
- c = *buffer->cur++;
- have_char:
/* Handle trigraphs, escaped newlines etc. */
+ c = *buffer->cur++;
if (c == '?' || c == '\\')
- c = skip_escaped_newlines (buffer, c);
+ c = skip_escaped_newlines (pfile);
- if (c == terminator && unescaped_terminator_p (pfile, dest))
+ if (c == terminator)
{
- c = EOF;
- break;
+ if (unescaped_terminator_p (pfile, dest))
+ break;
}
else if (is_vspace (c))
{
character literals at end of line. This is a kludge
around not knowing where comments are. */
if (CPP_OPTION (pfile, lang) == CLK_ASM && terminator != '>')
- break;
+ {
+ buffer->cur--;
+ break;
+ }
/* Character constants and header names may not extend over
multiple lines. In Standard C, neither may strings.
if (terminator != '"' || pfile->state.angled_headers)
{
unterminated (pfile, terminator);
+ buffer->cur--;
break;
}
- if (pfile->mlstring_pos.line == 0)
+ if (!warned_multi)
+ {
+ warned_multi = true;
+ cpp_pedwarn (pfile, "multi-line string literals are deprecated");
+ }
+
+ if (pfile->mls_line == 0)
{
- pfile->mlstring_pos = pfile->lexer_pos;
- if (CPP_PEDANTIC (pfile))
- cpp_pedwarn (pfile, "multi-line string constant");
+ pfile->mls_line = token->line;
+ pfile->mls_col = token->col;
}
- handle_newline (buffer, c); /* Stores to read_ahead. */
+ handle_newline (pfile);
c = '\n';
}
else if (c == '\0')
{
- if (nulls++ == 0)
- cpp_warning (pfile, "null character(s) preserved in literal");
+ if (buffer->cur - 1 == buffer->rlimit)
+ {
+ unterminated (pfile, terminator);
+ buffer->cur--;
+ break;
+ }
+ if (!warned_nulls)
+ {
+ warned_nulls = true;
+ cpp_warning (pfile, "null character(s) preserved in literal");
+ }
}
- /* No terminating null for strings - they could contain nulls. */
- if (dest >= limit)
- limit = _cpp_next_chunk (pool, 0, &dest);
*dest++ = c;
-
- /* If we had a new line, the next character is in read_ahead. */
- if (c != '\n')
- continue;
- c = buffer->read_ahead;
- if (c != EOF)
- goto have_char;
}
- /* Remember the next character. */
- buffer->read_ahead = c;
+ *dest = '\0';
- token->val.str.text = POOL_FRONT (pool);
- token->val.str.len = dest - token->val.str.text;
- POOL_COMMIT (pool, token->val.str.len);
+ token->val.str.text = BUFF_FRONT (pfile->u_buff);
+ token->val.str.len = dest - BUFF_FRONT (pfile->u_buff);
+ BUFF_FRONT (pfile->u_buff) = dest + 1;
}
/* The stored comment includes the comment start and any terminator. */
unsigned int len;
len = pfile->buffer->cur - from + 1; /* + 1 for the initial '/'. */
+
/* C++ comments probably (not definitely) have moved past a new
line, which we don't want to save in the comment. */
- if (pfile->buffer->read_ahead != EOF)
+ if (is_vspace (pfile->buffer->cur[-1]))
len--;
- buffer = _cpp_pool_alloc (&pfile->ident_pool, len);
+ buffer = _cpp_unaligned_alloc (pfile, len);
token->type = CPP_COMMENT;
token->val.str.len = len;
memcpy (buffer + 1, from, len - 1);
}
-/* Subroutine of lex_token to handle '%'. A little tricky, since we
- want to avoid stepping back when lexing %:%X. */
-static void
-lex_percent (buffer, result)
- cpp_buffer *buffer;
- cpp_token *result;
+/* Allocate COUNT tokens for RUN. */
+void
+_cpp_init_tokenrun (run, count)
+ tokenrun *run;
+ unsigned int count;
{
- cppchar_t c;
+ run->base = xnewvec (cpp_token, count);
+ run->limit = run->base + count;
+ run->next = NULL;
+}
- result->type = CPP_MOD;
- /* Parsing %:%X could leave an extra character. */
- if (buffer->extra_char == EOF)
- c = get_effective_char (buffer);
- else
+/* Returns the next tokenrun, or creates one if there is none. */
+static tokenrun *
+next_tokenrun (run)
+ tokenrun *run;
+{
+ if (run->next == NULL)
{
- c = buffer->read_ahead = buffer->extra_char;
- buffer->extra_char = EOF;
+ run->next = xnew (tokenrun);
+ run->next->prev = run;
+ _cpp_init_tokenrun (run->next, 250);
}
- if (c == '=')
- ACCEPT_CHAR (CPP_MOD_EQ);
- else if (CPP_OPTION (buffer->pfile, digraphs))
- {
- if (c == ':')
- {
- result->flags |= DIGRAPH;
- ACCEPT_CHAR (CPP_HASH);
- if (get_effective_char (buffer) == '%')
- {
- buffer->extra_char = get_effective_char (buffer);
- if (buffer->extra_char == ':')
- {
- buffer->extra_char = EOF;
- ACCEPT_CHAR (CPP_PASTE);
- }
- else
- /* We'll catch the extra_char when we're called back. */
- buffer->read_ahead = '%';
- }
- }
- else if (c == '>')
- {
- result->flags |= DIGRAPH;
- ACCEPT_CHAR (CPP_CLOSE_BRACE);
- }
- }
+ return run->next;
}
-/* Subroutine of lex_token to handle '.'. This is tricky, since we
- want to avoid stepping back when lexing '...' or '.123'. In the
- latter case we should also set a flag for parse_number. */
-static void
-lex_dot (pfile, result)
+/* Allocate a single token that is invalidated at the same time as the
+ rest of the tokens on the line. Has its line and col set to the
+ same as the last lexed token, so that diagnostics appear in the
+ right place. */
+cpp_token *
+_cpp_temp_token (pfile)
cpp_reader *pfile;
- cpp_token *result;
{
- cpp_buffer *buffer = pfile->buffer;
- cppchar_t c;
+ cpp_token *old, *result;
- /* Parsing ..X could leave an extra character. */
- if (buffer->extra_char == EOF)
- c = get_effective_char (buffer);
- else
+ old = pfile->cur_token - 1;
+ if (pfile->cur_token == pfile->cur_run->limit)
{
- c = buffer->read_ahead = buffer->extra_char;
- buffer->extra_char = EOF;
+ pfile->cur_run = next_tokenrun (pfile->cur_run);
+ pfile->cur_token = pfile->cur_run->base;
}
- /* All known character sets have 0...9 contiguous. */
- if (c >= '0' && c <= '9')
- {
- result->type = CPP_NUMBER;
- parse_number (pfile, &result->val.str, c, 1);
- }
- else
+ result = pfile->cur_token++;
+ result->line = old->line;
+ result->col = old->col;
+ return result;
+}
+
+/* Lex a token into RESULT (external interface). Takes care of issues
+ like directive handling, token lookahead, multiple include
+ optimization and skipping. */
+const cpp_token *
+_cpp_lex_token (pfile)
+ cpp_reader *pfile;
+{
+ cpp_token *result;
+
+ for (;;)
{
- result->type = CPP_DOT;
- if (c == '.')
+ if (pfile->cur_token == pfile->cur_run->limit)
{
- buffer->extra_char = get_effective_char (buffer);
- if (buffer->extra_char == '.')
- {
- buffer->extra_char = EOF;
- ACCEPT_CHAR (CPP_ELLIPSIS);
- }
- else
- /* We'll catch the extra_char when we're called back. */
- buffer->read_ahead = '.';
+ pfile->cur_run = next_tokenrun (pfile->cur_run);
+ pfile->cur_token = pfile->cur_run->base;
}
- else if (c == '*' && CPP_OPTION (pfile, cplusplus))
- ACCEPT_CHAR (CPP_DOT_STAR);
+
+ if (pfile->lookaheads)
+ {
+ pfile->lookaheads--;
+ result = pfile->cur_token++;
+ }
+ else
+ result = _cpp_lex_direct (pfile);
+
+ if (result->flags & BOL)
+ {
+ /* Is this a directive. If _cpp_handle_directive returns
+ false, it is an assembler #. */
+ if (result->type == CPP_HASH
+ && !pfile->state.parsing_args
+ && _cpp_handle_directive (pfile, result->flags & PREV_WHITE))
+ continue;
+ if (pfile->cb.line_change && !pfile->state.skipping)
+ (*pfile->cb.line_change)(pfile, result, pfile->state.parsing_args);
+ }
+
+ /* We don't skip tokens in directives. */
+ if (pfile->state.in_directive)
+ break;
+
+ /* Outside a directive, invalidate controlling macros. At file
+ EOF, _cpp_lex_direct takes care of popping the buffer, so we never
+ get here and MI optimisation works. */
+ pfile->mi_valid = false;
+
+ if (!pfile->state.skipping || result->type == CPP_EOF)
+ break;
}
+
+ return result;
}
-void
-_cpp_lex_token (pfile, result)
+#define IF_NEXT_IS(CHAR, THEN_TYPE, ELSE_TYPE) \
+ do { \
+ if (get_effective_char (pfile) == CHAR) \
+ result->type = THEN_TYPE; \
+ else \
+ { \
+ BACKUP (); \
+ result->type = ELSE_TYPE; \
+ } \
+ } while (0)
+
+/* Lex a token into pfile->cur_token, which is also incremented, to
+ get diagnostics pointing to the correct location.
+
+ Does not handle issues such as token lookahead, multiple-include
+ optimisation, directives, skipping etc. This function is only
+ suitable for use by _cpp_lex_token, and in special cases like
+ lex_expansion_token which doesn't care for any of these issues.
+
+ When meeting a newline, returns CPP_EOF if parsing a directive,
+ otherwise returns to the start of the token buffer if permissible.
+ Returns the location of the lexed token. */
+cpp_token *
+_cpp_lex_direct (pfile)
cpp_reader *pfile;
- cpp_token *result;
{
cppchar_t c;
cpp_buffer *buffer;
const unsigned char *comment_start;
- unsigned char bol;
+ cpp_token *result = pfile->cur_token++;
- skip:
- bol = pfile->state.next_bol;
- done_directive:
+ fresh_line:
buffer = pfile->buffer;
- pfile->state.next_bol = 0;
- result->flags = 0;
- next_char:
- pfile->lexer_pos.line = buffer->lineno;
- next_char2:
- pfile->lexer_pos.col = CPP_BUF_COLUMN (buffer, buffer->cur);
-
- c = buffer->read_ahead;
- if (c == EOF && buffer->cur < buffer->rlimit)
- {
- c = *buffer->cur++;
- pfile->lexer_pos.col++;
- }
+ result->flags = buffer->saved_flags;
+ buffer->saved_flags = 0;
+ update_tokens_line:
+ result->line = pfile->line;
- do_switch:
- buffer->read_ahead = EOF;
+ skipped_white:
+ c = *buffer->cur++;
+ result->col = CPP_BUF_COLUMN (buffer, buffer->cur);
+
+ trigraph:
switch (c)
{
- case EOF:
- /* Non-empty files should end in a newline. Ignore for command
- line and _Pragma buffers. */
- if (pfile->lexer_pos.col != 0 && !buffer->from_stage3)
- cpp_pedwarn (pfile, "no newline at end of file");
- pfile->state.next_bol = 1;
- pfile->skipping = 0; /* In case missing #endif. */
- result->type = CPP_EOF;
- /* Don't do MI optimisation. */
- return;
-
case ' ': case '\t': case '\f': case '\v': case '\0':
- skip_whitespace (pfile, c);
result->flags |= PREV_WHITE;
- goto next_char2;
+ if (skip_whitespace (pfile, c))
+ goto skipped_white;
- case '\n': case '\r':
- if (!pfile->state.in_directive)
+ /* EOF. */
+ buffer->cur--;
+ buffer->saved_flags = BOL;
+ if (!pfile->state.parsing_args && !pfile->state.in_directive)
{
- handle_newline (buffer, c);
- bol = 1;
- pfile->lexer_pos.output_line = buffer->lineno;
- /* This is a new line, so clear any white space flag.
- Newlines in arguments are white space (6.10.3.10);
- parse_arg takes care of that. */
- result->flags &= ~PREV_WHITE;
- goto next_char;
+ if (buffer->cur != buffer->line_base)
+ {
+ /* Non-empty files should end in a newline. Don't warn
+ for command line and _Pragma buffers. */
+ if (!buffer->from_stage3)
+ cpp_pedwarn (pfile, "no newline at end of file");
+ handle_newline (pfile);
+ }
+
+ /* Don't pop the last buffer. */
+ if (buffer->prev)
+ {
+ unsigned char stop = buffer->return_at_eof;
+
+ _cpp_pop_buffer (pfile);
+ if (!stop)
+ goto fresh_line;
+ }
}
+ result->type = CPP_EOF;
+ break;
- /* Don't let directives spill over to the next line. */
- buffer->read_ahead = c;
- pfile->state.next_bol = 1;
+ case '\n': case '\r':
+ handle_newline (pfile);
+ buffer->saved_flags = BOL;
+ if (! pfile->state.in_directive)
+ {
+ if (pfile->state.parsing_args == 2)
+ buffer->saved_flags |= PREV_WHITE;
+ if (!pfile->keep_tokens)
+ {
+ pfile->cur_run = &pfile->base_run;
+ result = pfile->base_run.base;
+ pfile->cur_token = result + 1;
+ }
+ goto fresh_line;
+ }
result->type = CPP_EOF;
- /* Don't break; pfile->skipping might be true. */
- return;
+ break;
case '?':
case '\\':
/* These could start an escaped newline, or '?' a trigraph. Let
skip_escaped_newlines do all the work. */
{
- unsigned int lineno = buffer->lineno;
+ unsigned int line = pfile->line;
- c = skip_escaped_newlines (buffer, c);
- if (lineno != buffer->lineno)
- /* We had at least one escaped newline of some sort, and the
- next character is in buffer->read_ahead. Update the
- token's line and column. */
- goto next_char;
+ c = skip_escaped_newlines (pfile);
+ if (line != pfile->line)
+ {
+ buffer->cur--;
+ /* We had at least one escaped newline of some sort.
+ Update the token's line and column. */
+ goto update_tokens_line;
+ }
+ }
- /* We are either the original '?' or '\\', or a trigraph. */
+ /* We are either the original '?' or '\\', or a trigraph. */
+ if (c == '?')
result->type = CPP_QUERY;
- buffer->read_ahead = EOF;
- if (c == '\\')
- goto random_char;
- else if (c != '?')
- goto do_switch;
- }
+ else if (c == '\\')
+ goto random_char;
+ else
+ goto trigraph;
break;
case '0': case '1': case '2': case '3': case '4':
parse_number (pfile, &result->val.str, c, 0);
break;
- case '$':
- if (!CPP_OPTION (pfile, dollars_in_ident))
- goto random_char;
- /* Fall through... */
+ case 'L':
+ /* 'L' may introduce wide characters or strings. */
+ {
+ const unsigned char *pos = buffer->cur;
+
+ c = get_effective_char (pfile);
+ if (c == '\'' || c == '"')
+ {
+ result->type = (c == '"' ? CPP_WSTRING: CPP_WCHAR);
+ parse_string (pfile, result, c);
+ break;
+ }
+ buffer->cur = pos;
+ }
+ /* Fall through. */
+ start_ident:
case '_':
case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
case 'g': case 'h': case 'i': case 'j': case 'k': case 'l':
case 's': case 't': case 'u': case 'v': case 'w': case 'x':
case 'y': case 'z':
case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
- case 'G': case 'H': case 'I': case 'J': case 'K': case 'L':
+ case 'G': case 'H': case 'I': case 'J': case 'K':
case 'M': case 'N': case 'O': case 'P': case 'Q': case 'R':
case 'S': case 'T': case 'U': case 'V': case 'W': case 'X':
case 'Y': case 'Z':
result->type = CPP_NAME;
- result->val.node = parse_identifier (pfile, c);
+ result->val.node = parse_identifier (pfile);
- /* 'L' may introduce wide characters or strings. */
- if (result->val.node == pfile->spec_nodes.n_L)
- {
- c = buffer->read_ahead; /* For make_string. */
- if (c == '\'' || c == '"')
- {
- ACCEPT_CHAR (c == '"' ? CPP_WSTRING: CPP_WCHAR);
- goto make_string;
- }
- }
/* Convert named operators to their proper types. */
- else if (result->val.node->flags & NODE_OPERATOR)
+ if (result->val.node->flags & NODE_OPERATOR)
{
result->flags |= NAMED_OP;
result->type = result->val.node->value.operator;
case '\'':
case '"':
result->type = c == '"' ? CPP_STRING: CPP_CHAR;
- make_string:
parse_string (pfile, result, c);
break;
case '/':
/* A potential block or line comment. */
comment_start = buffer->cur;
- result->type = CPP_DIV;
- c = get_effective_char (buffer);
- if (c == '=')
- ACCEPT_CHAR (CPP_DIV_EQ);
- if (c != '/' && c != '*')
- break;
+ c = get_effective_char (pfile);
if (c == '*')
{
if (skip_block_comment (pfile))
- cpp_error_with_line (pfile, pfile->lexer_pos.line,
- pfile->lexer_pos.col,
- "unterminated comment");
+ cpp_error (pfile, "unterminated comment");
}
- else
+ else if (c == '/' && (CPP_OPTION (pfile, cplusplus_comments)
+ || CPP_IN_SYSTEM_HEADER (pfile)))
{
- if (!CPP_OPTION (pfile, cplusplus_comments)
- && !CPP_IN_SYSTEM_HEADER (pfile))
- break;
-
/* Warn about comments only if pedantically GNUC89, and not
in system headers. */
if (CPP_OPTION (pfile, lang) == CLK_GNUC89 && CPP_PEDANTIC (pfile)
buffer->warned_cplusplus_comments = 1;
}
- /* Skip_line_comment updates buffer->read_ahead. */
- if (skip_line_comment (pfile))
- cpp_warning_with_line (pfile, pfile->lexer_pos.line,
- pfile->lexer_pos.col,
- "multi-line comment");
+ if (skip_line_comment (pfile) && CPP_OPTION (pfile, warn_comments))
+ cpp_warning (pfile, "multi-line comment");
+ }
+ else if (c == '=')
+ {
+ result->type = CPP_DIV_EQ;
+ break;
+ }
+ else
+ {
+ BACKUP ();
+ result->type = CPP_DIV;
+ break;
}
- /* Skipping the comment has updated buffer->read_ahead. */
if (!pfile->state.save_comments)
{
result->flags |= PREV_WHITE;
- goto next_char;
+ goto update_tokens_line;
}
/* Save the comment as a token in its own right. */
save_comment (pfile, result, comment_start);
- /* Don't do MI optimisation. */
- return;
+ break;
case '<':
if (pfile->state.angled_headers)
{
result->type = CPP_HEADER_NAME;
- c = '>'; /* terminator. */
- goto make_string;
+ parse_string (pfile, result, '>');
+ break;
}
- result->type = CPP_LESS;
- c = get_effective_char (buffer);
+ c = get_effective_char (pfile);
if (c == '=')
- ACCEPT_CHAR (CPP_LESS_EQ);
+ result->type = CPP_LESS_EQ;
else if (c == '<')
- {
- ACCEPT_CHAR (CPP_LSHIFT);
- if (get_effective_char (buffer) == '=')
- ACCEPT_CHAR (CPP_LSHIFT_EQ);
- }
+ IF_NEXT_IS ('=', CPP_LSHIFT_EQ, CPP_LSHIFT);
else if (c == '?' && CPP_OPTION (pfile, cplusplus))
- {
- ACCEPT_CHAR (CPP_MIN);
- if (get_effective_char (buffer) == '=')
- ACCEPT_CHAR (CPP_MIN_EQ);
- }
+ IF_NEXT_IS ('=', CPP_MIN_EQ, CPP_MIN);
else if (c == ':' && CPP_OPTION (pfile, digraphs))
{
- ACCEPT_CHAR (CPP_OPEN_SQUARE);
+ result->type = CPP_OPEN_SQUARE;
result->flags |= DIGRAPH;
}
else if (c == '%' && CPP_OPTION (pfile, digraphs))
{
- ACCEPT_CHAR (CPP_OPEN_BRACE);
+ result->type = CPP_OPEN_BRACE;
result->flags |= DIGRAPH;
}
+ else
+ {
+ BACKUP ();
+ result->type = CPP_LESS;
+ }
break;
case '>':
- result->type = CPP_GREATER;
- c = get_effective_char (buffer);
+ c = get_effective_char (pfile);
if (c == '=')
- ACCEPT_CHAR (CPP_GREATER_EQ);
+ result->type = CPP_GREATER_EQ;
else if (c == '>')
- {
- ACCEPT_CHAR (CPP_RSHIFT);
- if (get_effective_char (buffer) == '=')
- ACCEPT_CHAR (CPP_RSHIFT_EQ);
- }
+ IF_NEXT_IS ('=', CPP_RSHIFT_EQ, CPP_RSHIFT);
else if (c == '?' && CPP_OPTION (pfile, cplusplus))
+ IF_NEXT_IS ('=', CPP_MAX_EQ, CPP_MAX);
+ else
{
- ACCEPT_CHAR (CPP_MAX);
- if (get_effective_char (buffer) == '=')
- ACCEPT_CHAR (CPP_MAX_EQ);
+ BACKUP ();
+ result->type = CPP_GREATER;
}
break;
case '%':
- lex_percent (buffer, result);
- if (result->type == CPP_HASH)
- goto do_hash;
- break;
-
- case '.':
- lex_dot (pfile, result);
- break;
-
- case '+':
- result->type = CPP_PLUS;
- c = get_effective_char (buffer);
+ c = get_effective_char (pfile);
if (c == '=')
- ACCEPT_CHAR (CPP_PLUS_EQ);
- else if (c == '+')
- ACCEPT_CHAR (CPP_PLUS_PLUS);
- break;
+ result->type = CPP_MOD_EQ;
+ else if (CPP_OPTION (pfile, digraphs) && c == ':')
+ {
+ result->flags |= DIGRAPH;
+ result->type = CPP_HASH;
+ if (get_effective_char (pfile) == '%')
+ {
+ const unsigned char *pos = buffer->cur;
- case '-':
- result->type = CPP_MINUS;
- c = get_effective_char (buffer);
- if (c == '>')
+ if (get_effective_char (pfile) == ':')
+ result->type = CPP_PASTE;
+ else
+ buffer->cur = pos - 1;
+ }
+ else
+ BACKUP ();
+ }
+ else if (CPP_OPTION (pfile, digraphs) && c == '>')
{
- ACCEPT_CHAR (CPP_DEREF);
- if (CPP_OPTION (pfile, cplusplus)
- && get_effective_char (buffer) == '*')
- ACCEPT_CHAR (CPP_DEREF_STAR);
+ result->flags |= DIGRAPH;
+ result->type = CPP_CLOSE_BRACE;
+ }
+ else
+ {
+ BACKUP ();
+ result->type = CPP_MOD;
}
- else if (c == '=')
- ACCEPT_CHAR (CPP_MINUS_EQ);
- else if (c == '-')
- ACCEPT_CHAR (CPP_MINUS_MINUS);
- break;
-
- case '*':
- result->type = CPP_MULT;
- if (get_effective_char (buffer) == '=')
- ACCEPT_CHAR (CPP_MULT_EQ);
- break;
-
- case '=':
- result->type = CPP_EQ;
- if (get_effective_char (buffer) == '=')
- ACCEPT_CHAR (CPP_EQ_EQ);
break;
- case '!':
- result->type = CPP_NOT;
- if (get_effective_char (buffer) == '=')
- ACCEPT_CHAR (CPP_NOT_EQ);
- break;
+ case '.':
+ result->type = CPP_DOT;
+ c = get_effective_char (pfile);
+ if (c == '.')
+ {
+ const unsigned char *pos = buffer->cur;
- case '&':
- result->type = CPP_AND;
- c = get_effective_char (buffer);
- if (c == '=')
- ACCEPT_CHAR (CPP_AND_EQ);
- else if (c == '&')
- ACCEPT_CHAR (CPP_AND_AND);
- break;
-
- case '#':
- c = buffer->extra_char; /* Can be set by error condition below. */
- if (c != EOF)
+ if (get_effective_char (pfile) == '.')
+ result->type = CPP_ELLIPSIS;
+ else
+ buffer->cur = pos - 1;
+ }
+ /* All known character sets have 0...9 contiguous. */
+ else if (ISDIGIT (c))
{
- buffer->read_ahead = c;
- buffer->extra_char = EOF;
+ result->type = CPP_NUMBER;
+ parse_number (pfile, &result->val.str, c, 1);
}
+ else if (c == '*' && CPP_OPTION (pfile, cplusplus))
+ result->type = CPP_DOT_STAR;
else
- c = get_effective_char (buffer);
+ BACKUP ();
+ break;
- if (c == '#')
+ case '+':
+ c = get_effective_char (pfile);
+ if (c == '+')
+ result->type = CPP_PLUS_PLUS;
+ else if (c == '=')
+ result->type = CPP_PLUS_EQ;
+ else
{
- ACCEPT_CHAR (CPP_PASTE);
- break;
+ BACKUP ();
+ result->type = CPP_PLUS;
}
+ break;
- result->type = CPP_HASH;
- do_hash:
- if (bol)
+ case '-':
+ c = get_effective_char (pfile);
+ if (c == '>')
{
- if (pfile->state.parsing_args)
- {
- /* 6.10.3 paragraph 11: If there are sequences of
- preprocessing tokens within the list of arguments that
- would otherwise act as preprocessing directives, the
- behavior is undefined.
-
- This implementation will report a hard error, terminate
- the macro invocation, and proceed to process the
- directive. */
- cpp_error (pfile,
- "directives may not be used inside a macro argument");
-
- /* Put a '#' in lookahead, return CPP_EOF for parse_arg. */
- buffer->extra_char = buffer->read_ahead;
- buffer->read_ahead = '#';
- pfile->state.next_bol = 1;
- result->type = CPP_EOF;
-
- /* Get whitespace right - newline_in_args sets it. */
- if (pfile->lexer_pos.col == 1)
- result->flags &= ~PREV_WHITE;
- }
- else
+ result->type = CPP_DEREF;
+ if (CPP_OPTION (pfile, cplusplus))
{
- /* This is the hash introducing a directive. */
- if (_cpp_handle_directive (pfile, result->flags & PREV_WHITE))
- goto done_directive; /* bol still 1. */
- /* This is in fact an assembler #. */
+ if (get_effective_char (pfile) == '*')
+ result->type = CPP_DEREF_STAR;
+ else
+ BACKUP ();
}
}
+ else if (c == '-')
+ result->type = CPP_MINUS_MINUS;
+ else if (c == '=')
+ result->type = CPP_MINUS_EQ;
+ else
+ {
+ BACKUP ();
+ result->type = CPP_MINUS;
+ }
break;
- case '|':
- result->type = CPP_OR;
- c = get_effective_char (buffer);
- if (c == '=')
- ACCEPT_CHAR (CPP_OR_EQ);
- else if (c == '|')
- ACCEPT_CHAR (CPP_OR_OR);
+ case '&':
+ c = get_effective_char (pfile);
+ if (c == '&')
+ result->type = CPP_AND_AND;
+ else if (c == '=')
+ result->type = CPP_AND_EQ;
+ else
+ {
+ BACKUP ();
+ result->type = CPP_AND;
+ }
break;
-
- case '^':
- result->type = CPP_XOR;
- if (get_effective_char (buffer) == '=')
- ACCEPT_CHAR (CPP_XOR_EQ);
+
+ case '|':
+ c = get_effective_char (pfile);
+ if (c == '|')
+ result->type = CPP_OR_OR;
+ else if (c == '=')
+ result->type = CPP_OR_EQ;
+ else
+ {
+ BACKUP ();
+ result->type = CPP_OR;
+ }
break;
case ':':
- result->type = CPP_COLON;
- c = get_effective_char (buffer);
+ c = get_effective_char (pfile);
if (c == ':' && CPP_OPTION (pfile, cplusplus))
- ACCEPT_CHAR (CPP_SCOPE);
+ result->type = CPP_SCOPE;
else if (c == '>' && CPP_OPTION (pfile, digraphs))
{
result->flags |= DIGRAPH;
- ACCEPT_CHAR (CPP_CLOSE_SQUARE);
+ result->type = CPP_CLOSE_SQUARE;
+ }
+ else
+ {
+ BACKUP ();
+ result->type = CPP_COLON;
}
break;
+ case '*': IF_NEXT_IS ('=', CPP_MULT_EQ, CPP_MULT); break;
+ case '=': IF_NEXT_IS ('=', CPP_EQ_EQ, CPP_EQ); break;
+ case '!': IF_NEXT_IS ('=', CPP_NOT_EQ, CPP_NOT); break;
+ case '^': IF_NEXT_IS ('=', CPP_XOR_EQ, CPP_XOR); break;
+ case '#': IF_NEXT_IS ('#', CPP_PASTE, CPP_HASH); break;
+
case '~': result->type = CPP_COMPL; break;
case ',': result->type = CPP_COMMA; break;
case '(': result->type = CPP_OPEN_PAREN; break;
case '}': result->type = CPP_CLOSE_BRACE; break;
case ';': result->type = CPP_SEMICOLON; break;
- case '@':
- if (CPP_OPTION (pfile, objc))
- {
- /* In Objective C, '@' may begin keywords or strings, like
- @keyword or @"string". It would be nice to call
- get_effective_char here and test the result. However, we
- would then need to pass 2 characters to parse_identifier,
- making it ugly and slowing down its main loop. Instead,
- we assume we have an identifier, and recover if not. */
- result->type = CPP_NAME;
- result->val.node = parse_identifier (pfile, c);
- if (result->val.node->length != 1)
- break;
+ /* @ is a punctuator in Objective C. */
+ case '@': result->type = CPP_ATSIGN; break;
- /* OK, so it wasn't an identifier. Maybe a string? */
- if (buffer->read_ahead == '"')
- {
- c = '"';
- ACCEPT_CHAR (CPP_OSTRING);
- goto make_string;
- }
- }
- goto random_char;
+ case '$':
+ if (CPP_OPTION (pfile, dollars_in_ident))
+ goto start_ident;
+ /* Fall through... */
random_char:
default:
break;
}
- if (pfile->skipping)
- goto skip;
-
- /* If not in a directive, this token invalidates controlling macros. */
- if (!pfile->state.in_directive)
- pfile->mi_state = MI_FAILED;
+ return result;
}
-/* An upper bound on the number of bytes needed to spell a token,
+/* An upper bound on the number of bytes needed to spell TOKEN,
including preceding whitespace. */
unsigned int
cpp_token_len (token)
switch (TOKEN_SPELL (token))
{
- default: len = 0; break;
- case SPELL_STRING: len = token->val.str.len; break;
- case SPELL_IDENT: len = token->val.node->length; break;
+ default: len = 0; break;
+ case SPELL_NUMBER:
+ case SPELL_STRING: len = token->val.str.len; break;
+ case SPELL_IDENT: len = NODE_LEN (token->val.node); break;
}
- /* 1 for whitespace, 4 for comment delimeters. */
+ /* 1 for whitespace, 4 for comment delimiters. */
return len + 5;
}
unsigned char c;
if (token->flags & DIGRAPH)
- spelling = digraph_spellings[token->type - CPP_FIRST_DIGRAPH];
+ spelling
+ = digraph_spellings[(int) token->type - (int) CPP_FIRST_DIGRAPH];
else if (token->flags & NAMED_OP)
goto spell_ident;
else
}
break;
+ case SPELL_CHAR:
+ *buffer++ = token->val.c;
+ break;
+
+ spell_ident:
case SPELL_IDENT:
- spell_ident:
- memcpy (buffer, token->val.node->name, token->val.node->length);
- buffer += token->val.node->length;
+ memcpy (buffer, NODE_NAME (token->val.node), NODE_LEN (token->val.node));
+ buffer += NODE_LEN (token->val.node);
+ break;
+
+ case SPELL_NUMBER:
+ memcpy (buffer, token->val.str.text, token->val.str.len);
+ buffer += token->val.str.len;
break;
case SPELL_STRING:
{
case CPP_STRING: left = '"'; right = '"'; tag = '\0'; break;
case CPP_WSTRING: left = '"'; right = '"'; tag = 'L'; break;
- case CPP_OSTRING: left = '"'; right = '"'; tag = '@'; break;
case CPP_CHAR: left = '\''; right = '\''; tag = '\0'; break;
case CPP_WCHAR: left = '\''; right = '\''; tag = 'L'; break;
case CPP_HEADER_NAME: left = '<'; right = '>'; tag = '\0'; break;
- default: left = '\0'; right = '\0'; tag = '\0'; break;
+ default:
+ cpp_ice (pfile, "unknown string token %s\n", TOKEN_NAME (token));
+ return buffer;
}
if (tag) *buffer++ = tag;
- if (left) *buffer++ = left;
+ *buffer++ = left;
memcpy (buffer, token->val.str.text, token->val.str.len);
buffer += token->val.str.len;
- if (right) *buffer++ = right;
+ *buffer++ = right;
}
break;
- case SPELL_CHAR:
- *buffer++ = token->val.c;
- break;
-
case SPELL_NONE:
- cpp_ice (pfile, "Unspellable token %s", TOKEN_NAME (token));
+ cpp_ice (pfile, "unspellable token %s", TOKEN_NAME (token));
break;
}
return buffer;
}
-/* Returns a token as a null-terminated string. The string is
- temporary, and automatically freed later. Useful for diagnostics. */
+/* Returns TOKEN spelt as a null-terminated string. The string is
+ freed when the reader is destroyed. Useful for diagnostics. */
unsigned char *
cpp_token_as_text (pfile, token)
cpp_reader *pfile;
const cpp_token *token;
{
unsigned int len = cpp_token_len (token);
- unsigned char *start = _cpp_pool_alloc (&pfile->ident_pool, len), *end;
+ unsigned char *start = _cpp_unaligned_alloc (pfile, len), *end;
end = cpp_spell_token (pfile, token, start);
end[0] = '\0';
return start;
}
-/* Used by C front ends. Should really move to using cpp_token_as_text. */
+/* Used by C front ends, which really should move to using
+ cpp_token_as_text. */
const char *
cpp_type2name (type)
enum cpp_ttype type;
return (const char *) token_spellings[type].name;
}
-/* Writes the spelling of token to FP. Separate from cpp_spell_token
- for efficiency - to avoid double-buffering. Also, outputs a space
- if PREV_WHITE is flagged. */
+/* Writes the spelling of token to FP, without any preceding space.
+ Separated from cpp_spell_token for efficiency - to avoid stdio
+ double-buffering. */
void
cpp_output_token (token, fp)
const cpp_token *token;
FILE *fp;
{
- if (token->flags & PREV_WHITE)
- putc (' ', fp);
-
switch (TOKEN_SPELL (token))
{
case SPELL_OPERATOR:
{
const unsigned char *spelling;
+ int c;
if (token->flags & DIGRAPH)
- spelling = digraph_spellings[token->type - CPP_FIRST_DIGRAPH];
+ spelling
+ = digraph_spellings[(int) token->type - (int) CPP_FIRST_DIGRAPH];
else if (token->flags & NAMED_OP)
goto spell_ident;
else
spelling = TOKEN_NAME (token);
- ufputs (spelling, fp);
+ c = *spelling;
+ do
+ putc (c, fp);
+ while ((c = *++spelling) != '\0');
}
break;
+ case SPELL_CHAR:
+ putc (token->val.c, fp);
+ break;
+
spell_ident:
case SPELL_IDENT:
- ufputs (token->val.node->name, fp);
+ fwrite (NODE_NAME (token->val.node), 1, NODE_LEN (token->val.node), fp);
break;
+ case SPELL_NUMBER:
+ fwrite (token->val.str.text, 1, token->val.str.len, fp);
+ break;
+
case SPELL_STRING:
{
int left, right, tag;
{
case CPP_STRING: left = '"'; right = '"'; tag = '\0'; break;
case CPP_WSTRING: left = '"'; right = '"'; tag = 'L'; break;
- case CPP_OSTRING: left = '"'; right = '"'; tag = '@'; break;
case CPP_CHAR: left = '\''; right = '\''; tag = '\0'; break;
case CPP_WCHAR: left = '\''; right = '\''; tag = 'L'; break;
case CPP_HEADER_NAME: left = '<'; right = '>'; tag = '\0'; break;
- default: left = '\0'; right = '\0'; tag = '\0'; break;
+ default:
+ fprintf (stderr, "impossible STRING token %s\n", TOKEN_NAME (token));
+ return;
}
if (tag) putc (tag, fp);
- if (left) putc (left, fp);
+ putc (left, fp);
fwrite (token->val.str.text, 1, token->val.str.len, fp);
- if (right) putc (right, fp);
+ putc (right, fp);
}
break;
- case SPELL_CHAR:
- putc (token->val.c, fp);
- break;
-
case SPELL_NONE:
/* An error, most probably. */
break;
return (a->type != CPP_MACRO_ARG || a->val.arg_no == b->val.arg_no);
case SPELL_IDENT:
return a->val.node == b->val.node;
+ case SPELL_NUMBER:
case SPELL_STRING:
return (a->val.str.len == b->val.str.len
&& !memcmp (a->val.str.text, b->val.str.text,
return 0;
}
-#if 0
-/* Compare two token lists. */
-int
-_cpp_equiv_toklists (a, b)
- const struct toklist *a, *b;
-{
- unsigned int i, count;
-
- count = a->limit - a->first;
- if (count != (b->limit - b->first))
- return 0;
-
- for (i = 0; i < count; i++)
- if (! _cpp_equiv_tokens (&a->first[i], &b->first[i]))
- return 0;
-
- return 1;
-}
-#endif
-
-/* Determine whether two tokens can be pasted together, and if so,
- what the resulting token is. Returns CPP_EOF if the tokens cannot
- be pasted, or the appropriate type for the merged token if they
- can. */
-enum cpp_ttype
-cpp_can_paste (pfile, token1, token2, digraph)
- cpp_reader * pfile;
- const cpp_token *token1, *token2;
- int* digraph;
-{
- enum cpp_ttype a = token1->type, b = token2->type;
- int cxx = CPP_OPTION (pfile, cplusplus);
-
- /* Treat named operators as if they were ordinary NAMEs. */
- if (token1->flags & NAMED_OP)
- a = CPP_NAME;
- if (token2->flags & NAMED_OP)
- b = CPP_NAME;
-
- if (a <= CPP_LAST_EQ && b == CPP_EQ)
- return a + (CPP_EQ_EQ - CPP_EQ);
-
- switch (a)
- {
- case CPP_GREATER:
- if (b == a) return CPP_RSHIFT;
- if (b == CPP_QUERY && cxx) return CPP_MAX;
- if (b == CPP_GREATER_EQ) return CPP_RSHIFT_EQ;
- break;
- case CPP_LESS:
- if (b == a) return CPP_LSHIFT;
- if (b == CPP_QUERY && cxx) return CPP_MIN;
- if (b == CPP_LESS_EQ) return CPP_LSHIFT_EQ;
- if (CPP_OPTION (pfile, digraphs))
- {
- if (b == CPP_COLON)
- {*digraph = 1; return CPP_OPEN_SQUARE;} /* <: digraph */
- if (b == CPP_MOD)
- {*digraph = 1; return CPP_OPEN_BRACE;} /* <% digraph */
- }
- break;
-
- case CPP_PLUS: if (b == a) return CPP_PLUS_PLUS; break;
- case CPP_AND: if (b == a) return CPP_AND_AND; break;
- case CPP_OR: if (b == a) return CPP_OR_OR; break;
-
- case CPP_MINUS:
- if (b == a) return CPP_MINUS_MINUS;
- if (b == CPP_GREATER) return CPP_DEREF;
- break;
- case CPP_COLON:
- if (b == a && cxx) return CPP_SCOPE;
- if (b == CPP_GREATER && CPP_OPTION (pfile, digraphs))
- {*digraph = 1; return CPP_CLOSE_SQUARE;} /* :> digraph */
- break;
-
- case CPP_MOD:
- if (CPP_OPTION (pfile, digraphs))
- {
- if (b == CPP_GREATER)
- {*digraph = 1; return CPP_CLOSE_BRACE;} /* %> digraph */
- if (b == CPP_COLON)
- {*digraph = 1; return CPP_HASH;} /* %: digraph */
- }
- break;
- case CPP_DEREF:
- if (b == CPP_MULT && cxx) return CPP_DEREF_STAR;
- break;
- case CPP_DOT:
- if (b == CPP_MULT && cxx) return CPP_DOT_STAR;
- if (b == CPP_NUMBER) return CPP_NUMBER;
- break;
-
- case CPP_HASH:
- if (b == a && (token1->flags & DIGRAPH) == (token2->flags & DIGRAPH))
- /* %:%: digraph */
- {*digraph = (token1->flags & DIGRAPH); return CPP_PASTE;}
- break;
-
- case CPP_NAME:
- if (b == CPP_NAME) return CPP_NAME;
- if (b == CPP_NUMBER
- && name_p (pfile, &token2->val.str)) return CPP_NAME;
- if (b == CPP_CHAR
- && token1->val.node == pfile->spec_nodes.n_L) return CPP_WCHAR;
- if (b == CPP_STRING
- && token1->val.node == pfile->spec_nodes.n_L) return CPP_WSTRING;
- break;
-
- case CPP_NUMBER:
- if (b == CPP_NUMBER) return CPP_NUMBER;
- if (b == CPP_NAME) return CPP_NUMBER;
- if (b == CPP_DOT) return CPP_NUMBER;
- /* Numbers cannot have length zero, so this is safe. */
- if ((b == CPP_PLUS || b == CPP_MINUS)
- && VALID_SIGN ('+', token1->val.str.text[token1->val.str.len - 1]))
- return CPP_NUMBER;
- break;
-
- case CPP_OTHER:
- if (CPP_OPTION (pfile, objc) && token1->val.c == '@')
- {
- if (b == CPP_NAME) return CPP_NAME;
- if (b == CPP_STRING) return CPP_OSTRING;
- }
-
- default:
- break;
- }
-
- return CPP_EOF;
-}
-
/* Returns nonzero if a space should be inserted to avoid an
accidental token paste for output. For simplicity, it is
conservative, and occasionally advises a space where one is not
needed, e.g. "." and ".2". */
-
int
cpp_avoid_paste (pfile, token1, token2)
cpp_reader *pfile;
c = EOF;
if (token2->flags & DIGRAPH)
- c = digraph_spellings[b - CPP_FIRST_DIGRAPH][0];
+ c = digraph_spellings[(int) b - (int) CPP_FIRST_DIGRAPH][0];
else if (token_spellings[b].category == SPELL_OPERATOR)
c = token_spellings[b].name[0];
/* Quickly get everything that can paste with an '='. */
- if (a <= CPP_LAST_EQ && c == '=')
+ if ((int) a <= (int) CPP_LAST_EQ && c == '=')
return 1;
switch (a)
case CPP_OR: return c == '|';
case CPP_COLON: return c == ':' || c == '>';
case CPP_DEREF: return c == '*';
- case CPP_DOT: return c == '.' || c == '%';
+ case CPP_DOT: return c == '.' || c == '%' || b == CPP_NUMBER;
case CPP_HASH: return c == '#' || c == '%'; /* Digraph form. */
case CPP_NAME: return ((b == CPP_NUMBER
&& name_p (pfile, &token2->val.str))
}
/* Output all the remaining tokens on the current line, and a newline
- character, to FP. Leading whitespace is removed. */
+ character, to FP. Leading whitespace is removed. If there are
+ macros, special token padding is not performed. */
void
cpp_output_line (pfile, fp)
cpp_reader *pfile;
FILE *fp;
{
- cpp_token token;
+ const cpp_token *token;
- cpp_get_token (pfile, &token);
- token.flags &= ~PREV_WHITE;
- while (token.type != CPP_EOF)
+ token = cpp_get_token (pfile);
+ while (token->type != CPP_EOF)
{
- cpp_output_token (&token, fp);
- cpp_get_token (pfile, &token);
+ cpp_output_token (token, fp);
+ token = cpp_get_token (pfile);
+ if (token->flags & PREV_WHITE)
+ putc (' ', fp);
}
putc ('\n', fp);
}
-/* Memory pools. */
-
-struct dummy
+/* Returns the value of a hexadecimal digit. */
+static unsigned int
+hex_digit_value (c)
+ unsigned int c;
{
- char c;
- union
- {
- double d;
- int *p;
- } u;
-};
+ if (hex_p (c))
+ return hex_value (c);
+ else
+ abort ();
+}
-#define DEFAULT_ALIGNMENT (offsetof (struct dummy, u))
+/* Parse a '\uNNNN' or '\UNNNNNNNN' sequence. Returns 1 to indicate
+ failure if cpplib is not parsing C++ or C99. Such failure is
+ silent, and no variables are updated. Otherwise returns 0, and
+ warns if -Wtraditional.
+
+ [lex.charset]: The character designated by the universal character
+ name \UNNNNNNNN is that character whose character short name in
+ ISO/IEC 10646 is NNNNNNNN; the character designated by the
+ universal character name \uNNNN is that character whose character
+ short name in ISO/IEC 10646 is 0000NNNN. If the hexadecimal value
+ for a universal character name is less than 0x20 or in the range
+ 0x7F-0x9F (inclusive), or if the universal character name
+ designates a character in the basic source character set, then the
+ program is ill-formed.
+
+ We assume that wchar_t is Unicode, so we don't need to do any
+ mapping. Is this ever wrong?
+
+ PC points to the 'u' or 'U', PSTR is points to the byte after PC,
+ LIMIT is the end of the string or charconst. PSTR is updated to
+ point after the UCS on return, and the UCS is written into PC. */
static int
-chunk_suitable (pool, chunk, size)
- cpp_pool *pool;
- cpp_chunk *chunk;
- unsigned int size;
+maybe_read_ucs (pfile, pstr, limit, pc)
+ cpp_reader *pfile;
+ const unsigned char **pstr;
+ const unsigned char *limit;
+ unsigned int *pc;
{
- /* Being at least twice SIZE means we can use memcpy in
- _cpp_next_chunk rather than memmove. Besides, it's a good idea
- anyway. */
- return (chunk && pool->locked != chunk
- && (unsigned int) (chunk->limit - chunk->base) >= size * 2);
+ const unsigned char *p = *pstr;
+ unsigned int code = 0;
+ unsigned int c = *pc, length;
+
+ /* Only attempt to interpret a UCS for C++ and C99. */
+ if (! (CPP_OPTION (pfile, cplusplus) || CPP_OPTION (pfile, c99)))
+ return 1;
+
+ if (CPP_WTRADITIONAL (pfile))
+ cpp_warning (pfile, "the meaning of '\\%c' varies with -traditional", c);
+
+ length = (c == 'u' ? 4: 8);
+
+ if ((size_t) (limit - p) < length)
+ {
+ cpp_error (pfile, "incomplete universal-character-name");
+ /* Skip to the end to avoid more diagnostics. */
+ p = limit;
+ }
+ else
+ {
+ for (; length; length--, p++)
+ {
+ c = *p;
+ if (ISXDIGIT (c))
+ code = (code << 4) + hex_digit_value (c);
+ else
+ {
+ cpp_error (pfile,
+ "non-hex digit '%c' in universal-character-name", c);
+ /* We shouldn't skip in case there are multibyte chars. */
+ break;
+ }
+ }
+ }
+
+#ifdef TARGET_EBCDIC
+ cpp_error (pfile, "universal-character-name on EBCDIC target");
+ code = 0x3f; /* EBCDIC invalid character */
+#else
+ /* True extended characters are OK. */
+ if (code >= 0xa0
+ && !(code & 0x80000000)
+ && !(code >= 0xD800 && code <= 0xDFFF))
+ ;
+ /* The standard permits $, @ and ` to be specified as UCNs. We use
+ hex escapes so that this also works with EBCDIC hosts. */
+ else if (code == 0x24 || code == 0x40 || code == 0x60)
+ ;
+ /* Don't give another error if one occurred above. */
+ else if (length == 0)
+ cpp_error (pfile, "universal-character-name out of range");
+#endif
+
+ *pstr = p;
+ *pc = code;
+ return 0;
}
-/* Returns the end of the new pool. PTR points to a char in the old
- pool, and is updated to point to the same char in the new pool. */
-unsigned char *
-_cpp_next_chunk (pool, len, ptr)
- cpp_pool *pool;
- unsigned int len;
- unsigned char **ptr;
+/* Interpret an escape sequence, and return its value. PSTR points to
+ the input pointer, which is just after the backslash. LIMIT is how
+ much text we have. MASK is a bitmask for the precision for the
+ destination type (char or wchar_t). TRADITIONAL, if true, does not
+ interpret escapes that did not exist in traditional C.
+
+ Handles all relevant diagnostics. */
+unsigned int
+cpp_parse_escape (pfile, pstr, limit, mask, traditional)
+ cpp_reader *pfile;
+ const unsigned char **pstr;
+ const unsigned char *limit;
+ unsigned HOST_WIDE_INT mask;
+ int traditional;
{
- cpp_chunk *chunk = pool->cur->next;
+ int unknown = 0;
+ const unsigned char *str = *pstr;
+ unsigned int c = *str++;
- /* LEN is the minimum size we want in the new pool. */
- len += POOL_ROOM (pool);
- if (! chunk_suitable (pool, chunk, len))
+ switch (c)
{
- chunk = new_chunk (POOL_SIZE (pool) * 2 + len);
+ case '\\': case '\'': case '"': case '?': break;
+ case 'b': c = TARGET_BS; break;
+ case 'f': c = TARGET_FF; break;
+ case 'n': c = TARGET_NEWLINE; break;
+ case 'r': c = TARGET_CR; break;
+ case 't': c = TARGET_TAB; break;
+ case 'v': c = TARGET_VT; break;
+
+ case '(': case '{': case '[': case '%':
+ /* '\(', etc, are used at beginning of line to avoid confusing Emacs.
+ '\%' is used to prevent SCCS from getting confused. */
+ unknown = CPP_PEDANTIC (pfile);
+ break;
+
+ case 'a':
+ if (CPP_WTRADITIONAL (pfile))
+ cpp_warning (pfile, "the meaning of '\\a' varies with -traditional");
+ if (!traditional)
+ c = TARGET_BELL;
+ break;
+
+ case 'e': case 'E':
+ if (CPP_PEDANTIC (pfile))
+ cpp_pedwarn (pfile, "non-ISO-standard escape sequence, '\\%c'", c);
+ c = TARGET_ESC;
+ break;
+
+ case 'u': case 'U':
+ unknown = maybe_read_ucs (pfile, &str, limit, &c);
+ break;
+
+ case 'x':
+ if (CPP_WTRADITIONAL (pfile))
+ cpp_warning (pfile, "the meaning of '\\x' varies with -traditional");
+
+ if (!traditional)
+ {
+ unsigned int i = 0, overflow = 0;
+ int digits_found = 0;
+
+ while (str < limit)
+ {
+ c = *str;
+ if (! ISXDIGIT (c))
+ break;
+ str++;
+ overflow |= i ^ (i << 4 >> 4);
+ i = (i << 4) + hex_digit_value (c);
+ digits_found = 1;
+ }
+
+ if (!digits_found)
+ cpp_error (pfile, "\\x used with no following hex digits");
+
+ if (overflow | (i != (i & mask)))
+ {
+ cpp_pedwarn (pfile, "hex escape sequence out of range");
+ i &= mask;
+ }
+ c = i;
+ }
+ break;
- chunk->next = pool->cur->next;
- pool->cur->next = chunk;
+ case '0': case '1': case '2': case '3':
+ case '4': case '5': case '6': case '7':
+ {
+ unsigned int i = c - '0';
+ int count = 0;
+
+ while (str < limit && ++count < 3)
+ {
+ c = *str;
+ if (c < '0' || c > '7')
+ break;
+ str++;
+ i = (i << 3) + c - '0';
+ }
+
+ if (i != (i & mask))
+ {
+ cpp_pedwarn (pfile, "octal escape sequence out of range");
+ i &= mask;
+ }
+ c = i;
+ }
+ break;
+
+ default:
+ unknown = 1;
+ break;
}
- /* Update the pointer before changing chunk's front. */
- if (ptr)
- *ptr += chunk->base - POOL_FRONT (pool);
+ if (unknown)
+ {
+ if (ISGRAPH (c))
+ cpp_pedwarn (pfile, "unknown escape sequence '\\%c'", c);
+ else
+ cpp_pedwarn (pfile, "unknown escape sequence: '\\%03o'", c);
+ }
- memcpy (chunk->base, POOL_FRONT (pool), POOL_ROOM (pool));
- chunk->front = chunk->base;
+ if (c > mask)
+ cpp_pedwarn (pfile, "escape sequence out of range for character");
- pool->cur = chunk;
- return POOL_LIMIT (pool);
+ *pstr = str;
+ return c;
}
-static cpp_chunk *
-new_chunk (size)
- unsigned int size;
+#ifndef MAX_CHAR_TYPE_SIZE
+#define MAX_CHAR_TYPE_SIZE CHAR_TYPE_SIZE
+#endif
+
+#ifndef MAX_WCHAR_TYPE_SIZE
+#define MAX_WCHAR_TYPE_SIZE WCHAR_TYPE_SIZE
+#endif
+
+/* Interpret a (possibly wide) character constant in TOKEN.
+ WARN_MULTI warns about multi-character charconsts, if not
+ TRADITIONAL. TRADITIONAL also indicates not to interpret escapes
+ that did not exist in traditional C. PCHARS_SEEN points to a
+ variable that is filled in with the number of characters seen. */
+HOST_WIDE_INT
+cpp_interpret_charconst (pfile, token, warn_multi, traditional, pchars_seen)
+ cpp_reader *pfile;
+ const cpp_token *token;
+ int warn_multi;
+ int traditional;
+ unsigned int *pchars_seen;
+{
+ const unsigned char *str = token->val.str.text;
+ const unsigned char *limit = str + token->val.str.len;
+ unsigned int chars_seen = 0;
+ unsigned int width, max_chars, c;
+ unsigned HOST_WIDE_INT mask;
+ HOST_WIDE_INT result = 0;
+
+#ifdef MULTIBYTE_CHARS
+ (void) local_mbtowc (NULL, NULL, 0);
+#endif
+
+ /* Width in bits. */
+ if (token->type == CPP_CHAR)
+ width = MAX_CHAR_TYPE_SIZE;
+ else
+ width = MAX_WCHAR_TYPE_SIZE;
+
+ if (width < HOST_BITS_PER_WIDE_INT)
+ mask = ((unsigned HOST_WIDE_INT) 1 << width) - 1;
+ else
+ mask = ~0;
+ max_chars = HOST_BITS_PER_WIDE_INT / width;
+
+ while (str < limit)
+ {
+#ifdef MULTIBYTE_CHARS
+ wchar_t wc;
+ int char_len;
+
+ char_len = local_mbtowc (&wc, str, limit - str);
+ if (char_len == -1)
+ {
+ cpp_warning (pfile, "ignoring invalid multibyte character");
+ c = *str++;
+ }
+ else
+ {
+ str += char_len;
+ c = wc;
+ }
+#else
+ c = *str++;
+#endif
+
+ if (c == '\\')
+ c = cpp_parse_escape (pfile, &str, limit, mask, traditional);
+
+#ifdef MAP_CHARACTER
+ if (ISPRINT (c))
+ c = MAP_CHARACTER (c);
+#endif
+
+ /* Merge character into result; ignore excess chars. */
+ if (++chars_seen <= max_chars)
+ {
+ if (width < HOST_BITS_PER_WIDE_INT)
+ result = (result << width) | (c & mask);
+ else
+ result = c;
+ }
+ }
+
+ if (chars_seen == 0)
+ cpp_error (pfile, "empty character constant");
+ else if (chars_seen > max_chars)
+ {
+ chars_seen = max_chars;
+ cpp_warning (pfile, "character constant too long");
+ }
+ else if (chars_seen > 1 && !traditional && warn_multi)
+ cpp_warning (pfile, "multi-character character constant");
+
+ /* If char type is signed, sign-extend the constant. The
+ __CHAR_UNSIGNED__ macro is set by the driver if appropriate. */
+ if (token->type == CPP_CHAR && chars_seen)
+ {
+ unsigned int nbits = chars_seen * width;
+
+ mask = (unsigned HOST_WIDE_INT) ~0 >> (HOST_BITS_PER_WIDE_INT - nbits);
+ if (pfile->spec_nodes.n__CHAR_UNSIGNED__->type == NT_MACRO
+ || ((result >> (nbits - 1)) & 1) == 0)
+ result &= mask;
+ else
+ result |= ~mask;
+ }
+
+ *pchars_seen = chars_seen;
+ return result;
+}
+
+/* Memory buffers. Changing these three constants can have a dramatic
+ effect on performance. The values here are reasonable defaults,
+ but might be tuned. If you adjust them, be sure to test across a
+ range of uses of cpplib, including heavy nested function-like macro
+ expansion. Also check the change in peak memory usage (NJAMD is a
+ good tool for this). */
+#define MIN_BUFF_SIZE 8000
+#define BUFF_SIZE_UPPER_BOUND(MIN_SIZE) (MIN_BUFF_SIZE + (MIN_SIZE) * 3 / 2)
+#define EXTENDED_BUFF_SIZE(BUFF, MIN_EXTRA) \
+ (MIN_EXTRA + ((BUFF)->limit - (BUFF)->cur) * 2)
+
+#if MIN_BUFF_SIZE > BUFF_SIZE_UPPER_BOUND (0)
+ #error BUFF_SIZE_UPPER_BOUND must be at least as large as MIN_BUFF_SIZE!
+#endif
+
+struct dummy
{
+ char c;
+ union
+ {
+ double d;
+ int *p;
+ } u;
+};
+
+#define DEFAULT_ALIGNMENT (offsetof (struct dummy, u))
+#define CPP_ALIGN(size, align) (((size) + ((align) - 1)) & ~((align) - 1))
+
+/* Create a new allocation buffer. Place the control block at the end
+ of the buffer, so that buffer overflows will cause immediate chaos. */
+static _cpp_buff *
+new_buff (len)
+ size_t len;
+{
+ _cpp_buff *result;
unsigned char *base;
- cpp_chunk *result;
- size = ALIGN (size, DEFAULT_ALIGNMENT);
- base = (unsigned char *) xmalloc (size + sizeof (cpp_chunk));
- /* Put the chunk descriptor at the end. Then chunk overruns will
- cause obvious chaos. */
- result = (cpp_chunk *) (base + size);
- result->base = base;
- result->front = base;
- result->limit = base + size;
- result->next = 0;
+ if (len < MIN_BUFF_SIZE)
+ len = MIN_BUFF_SIZE;
+ len = CPP_ALIGN (len, DEFAULT_ALIGNMENT);
+ base = xmalloc (len + sizeof (_cpp_buff));
+ result = (_cpp_buff *) (base + len);
+ result->base = base;
+ result->cur = base;
+ result->limit = base + len;
+ result->next = NULL;
return result;
}
+/* Place a chain of unwanted allocation buffers on the free list. */
void
-_cpp_init_pool (pool, size, align, temp)
- cpp_pool *pool;
- unsigned int size, align, temp;
+_cpp_release_buff (pfile, buff)
+ cpp_reader *pfile;
+ _cpp_buff *buff;
{
- if (align == 0)
- align = DEFAULT_ALIGNMENT;
- if (align & (align - 1))
- abort ();
- pool->align = align;
- pool->cur = new_chunk (size);
- pool->locked = 0;
- pool->locks = 0;
- if (temp)
- pool->cur->next = pool->cur;
+ _cpp_buff *end = buff;
+
+ while (end->next)
+ end = end->next;
+ end->next = pfile->free_buffs;
+ pfile->free_buffs = buff;
}
-void
-_cpp_lock_pool (pool)
- cpp_pool *pool;
+/* Return a free buffer of size at least MIN_SIZE. */
+_cpp_buff *
+_cpp_get_buff (pfile, min_size)
+ cpp_reader *pfile;
+ size_t min_size;
+{
+ _cpp_buff *result, **p;
+
+ for (p = &pfile->free_buffs;; p = &(*p)->next)
+ {
+ size_t size;
+
+ if (*p == NULL)
+ return new_buff (min_size);
+ result = *p;
+ size = result->limit - result->base;
+ /* Return a buffer that's big enough, but don't waste one that's
+ way too big. */
+ if (size >= min_size && size <= BUFF_SIZE_UPPER_BOUND (min_size))
+ break;
+ }
+
+ *p = result->next;
+ result->next = NULL;
+ result->cur = result->base;
+ return result;
+}
+
+/* Creates a new buffer with enough space to hold the uncommitted
+ remaining bytes of BUFF, and at least MIN_EXTRA more bytes. Copies
+ the excess bytes to the new buffer. Chains the new buffer after
+ BUFF, and returns the new buffer. */
+_cpp_buff *
+_cpp_append_extend_buff (pfile, buff, min_extra)
+ cpp_reader *pfile;
+ _cpp_buff *buff;
+ size_t min_extra;
{
- if (pool->locks++ == 0)
- pool->locked = pool->cur;
+ size_t size = EXTENDED_BUFF_SIZE (buff, min_extra);
+ _cpp_buff *new_buff = _cpp_get_buff (pfile, size);
+
+ buff->next = new_buff;
+ memcpy (new_buff->base, buff->cur, BUFF_ROOM (buff));
+ return new_buff;
}
+/* Creates a new buffer with enough space to hold the uncommitted
+ remaining bytes of the buffer pointed to by BUFF, and at least
+ MIN_EXTRA more bytes. Copies the excess bytes to the new buffer.
+ Chains the new buffer before the buffer pointed to by BUFF, and
+ updates the pointer to point to the new buffer. */
void
-_cpp_unlock_pool (pool)
- cpp_pool *pool;
+_cpp_extend_buff (pfile, pbuff, min_extra)
+ cpp_reader *pfile;
+ _cpp_buff **pbuff;
+ size_t min_extra;
{
- if (--pool->locks == 0)
- pool->locked = 0;
+ _cpp_buff *new_buff, *old_buff = *pbuff;
+ size_t size = EXTENDED_BUFF_SIZE (old_buff, min_extra);
+
+ new_buff = _cpp_get_buff (pfile, size);
+ memcpy (new_buff->base, old_buff->cur, BUFF_ROOM (old_buff));
+ new_buff->next = old_buff;
+ *pbuff = new_buff;
}
+/* Free a chain of buffers starting at BUFF. */
void
-_cpp_free_pool (pool)
- cpp_pool *pool;
+_cpp_free_buff (buff)
+ _cpp_buff *buff;
{
- cpp_chunk *chunk = pool->cur, *next;
+ _cpp_buff *next;
- do
+ for (; buff; buff = next)
{
- next = chunk->next;
- free (chunk->base);
- chunk = next;
+ next = buff->next;
+ free (buff->base);
}
- while (chunk && chunk != pool->cur);
}
-/* Reserve LEN bytes from a memory pool. */
+/* Allocate permanent, unaligned storage of length LEN. */
unsigned char *
-_cpp_pool_reserve (pool, len)
- cpp_pool *pool;
- unsigned int len;
+_cpp_unaligned_alloc (pfile, len)
+ cpp_reader *pfile;
+ size_t len;
{
- len = ALIGN (len, pool->align);
- if (len > (unsigned int) POOL_ROOM (pool))
- _cpp_next_chunk (pool, len, 0);
+ _cpp_buff *buff = pfile->u_buff;
+ unsigned char *result = buff->cur;
- return POOL_FRONT (pool);
+ if (len > (size_t) (buff->limit - result))
+ {
+ buff = _cpp_get_buff (pfile, len);
+ buff->next = pfile->u_buff;
+ pfile->u_buff = buff;
+ result = buff->cur;
+ }
+
+ buff->cur = result + len;
+ return result;
}
-/* Allocate LEN bytes from a memory pool. */
+/* Allocate permanent, unaligned storage of length LEN from a_buff.
+ That buffer is used for growing allocations when saving macro
+ replacement lists in a #define, and when parsing an answer to an
+ assertion in #assert, #unassert or #if (and therefore possibly
+ whilst expanding macros). It therefore must not be used by any
+ code that they might call: specifically the lexer and the guts of
+ the macro expander.
+
+ All existing other uses clearly fit this restriction: storing
+ registered pragmas during initialization. */
unsigned char *
-_cpp_pool_alloc (pool, len)
- cpp_pool *pool;
- unsigned int len;
+_cpp_aligned_alloc (pfile, len)
+ cpp_reader *pfile;
+ size_t len;
{
- unsigned char *result = _cpp_pool_reserve (pool, len);
+ _cpp_buff *buff = pfile->a_buff;
+ unsigned char *result = buff->cur;
+
+ if (len > (size_t) (buff->limit - result))
+ {
+ buff = _cpp_get_buff (pfile, len);
+ buff->next = pfile->a_buff;
+ pfile->a_buff = buff;
+ result = buff->cur;
+ }
- POOL_COMMIT (pool, len);
+ buff->cur = result + len;
return result;
}