/* Input parser for bison
- Copyright 1984, 1986, 1989, 1992, 1998, 2000, 2001
+ Copyright (C) 1984, 1986, 1989, 1992, 1998, 2000, 2001, 2002
Free Software Foundation, Inc.
This file is part of Bison, the GNU Compiler Compiler.
#include "system.h"
-#include "obstack.h"
#include "quotearg.h"
#include "quote.h"
#include "getargs.h"
#include "conflicts.h"
#include "muscle_tab.h"
-/* Number of slots allocated (but not necessarily used yet) in `rline' */
-static int rline_allocated;
-
typedef struct symbol_list
{
struct symbol_list *next;
- bucket *sym;
- bucket *ruleprec;
-}
-symbol_list;
+ symbol_t *sym;
+ int line;
+
+ /* The action is attached to the LHS of a rule. */
+ const char *action;
+ int action_line;
+
+ symbol_t *ruleprec;
+} symbol_list;
int lineno;
-char **tags;
-short *user_toknums;
-static symbol_list *grammar;
-static int start_flag;
-static bucket *startval;
+static symbol_list *grammar = NULL;
+static int start_flag = 0;
+static symbol_t *startval = NULL;
/* Nonzero if components of semantic values are used, implying
they must be unions. */
static int value_components_used;
/* Nonzero if %union has been seen. */
-static int typed;
+static int typed = 0;
/* Incremented for each %left, %right or %nonassoc seen */
-static int lastprec;
+static int lastprec = 0;
+
+symbol_t *errtoken = NULL;
+symbol_t *undeftoken = NULL;
+symbol_t *eoftoken = NULL;
+symbol_t *axiom = NULL;
+
+static symbol_list *
+symbol_list_new (symbol_t *sym)
+{
+ symbol_list *res = XMALLOC (symbol_list, 1);
+ res->next = NULL;
+ res->sym = sym;
+ res->line = lineno;
+ res->action = NULL;
+ res->action_line = 0;
+ res->ruleprec = NULL;
+ return res;
+}
+
+/*------------------------.
+| Operations on symbols. |
+`------------------------*/
+
+
+/*-----------------------------------------------------------.
+| If THIS is not defined, report an error, and consider it a |
+| nonterminal. |
+`-----------------------------------------------------------*/
+
+static bool
+symbol_check_defined (symbol_t *this)
+{
+ if (this->class == unknown_sym)
+ {
+ complain
+ (_("symbol %s is used, but is not defined as a token and has no rules"),
+ this->tag);
+ this->class = nterm_sym;
+ this->number = nvars++;
+ }
+
+ return TRUE;
+}
+
+
+/*-------------------------------------------------------------------.
+| Assign a symbol number, and write the definition of the token name |
+| into FDEFINES. Put in SYMBOLS. |
+`-------------------------------------------------------------------*/
+
+static bool
+symbol_make_alias (symbol_t *symbol, char *typename)
+{
+ if (symval->alias)
+ warn (_("symbol `%s' used more than once as a literal string"),
+ symval->tag);
+ else if (symbol->alias)
+ warn (_("symbol `%s' given more than one literal string"),
+ symbol->tag);
+ else
+ {
+ symval->class = token_sym;
+ symval->type_name = typename;
+ symval->user_token_number = symbol->user_token_number;
+ symbol->user_token_number = USER_NUMBER_ALIAS;
+ symval->alias = symbol;
+ symbol->alias = symval;
+ /* symbol and symval combined are only one symbol */
+ nsyms--;
+ ntokens--;
+ assert (ntokens == symbol->number || ntokens == symval->number);
+ symbol->number = symval->number =
+ (symval->number < symbol->number) ? symval->number : symbol->number;
+ }
+
+ return TRUE;
+}
+
+/*---------------------------------------------------------.
+| Check that THIS, and its alias, have same precedence and |
+| associativity. |
+`---------------------------------------------------------*/
+
+static bool
+symbol_check_alias_consistence (symbol_t *this)
+{
+ /* Check only those who _are_ the aliases. */
+ if (this->alias && this->user_token_number == USER_NUMBER_ALIAS)
+ {
+ if (this->prec != this->alias->prec)
+ {
+ if (this->prec != 0 && this->alias->prec != 0)
+ complain (_("conflicting precedences for %s and %s"),
+ this->tag, this->alias->tag);
+ if (this->prec != 0)
+ this->alias->prec = this->prec;
+ else
+ this->prec = this->alias->prec;
+ }
+
+ if (this->assoc != this->alias->assoc)
+ {
+ if (this->assoc != 0 && this->alias->assoc != 0)
+ complain (_("conflicting assoc values for %s and %s"),
+ this->tag, this->alias->tag);
+ if (this->assoc != 0)
+ this->alias->assoc = this->assoc;
+ else
+ this->assoc = this->alias->assoc;
+ }
+ }
+ return TRUE;
+}
+
+
+/*-------------------------------------------------------------------.
+| Assign a symbol number, and write the definition of the token name |
+| into FDEFINES. Put in SYMBOLS. |
+`-------------------------------------------------------------------*/
+
+static bool
+symbol_pack (symbol_t *this)
+{
+ if (this->class == nterm_sym)
+ {
+ this->number += ntokens;
+ }
+ else if (this->alias)
+ {
+ /* This symbol and its alias are a single token defn.
+ Allocate a tokno, and assign to both check agreement of
+ prec and assoc fields and make both the same */
+ if (this->number == NUMBER_UNDEFINED)
+ {
+ if (this == eoftoken || this->alias == eoftoken)
+ this->number = this->alias->number = 0;
+ else
+ {
+ assert (this->alias->number != NUMBER_UNDEFINED);
+ this->number = this->alias->number;
+ }
+ }
+ /* Do not do processing below for USER_NUMBER_ALIASs. */
+ if (this->user_token_number == USER_NUMBER_ALIAS)
+ return TRUE;
+ }
+ else /* this->class == token_sym */
+ {
+ assert (this->number != NUMBER_UNDEFINED);
+ }
-static bucket *errtoken;
-static bucket *undeftoken;
+ symbols[this->number] = this;
+ return TRUE;
+}
+
+
+
+
+/*--------------------------------------------------.
+| Put THIS in TOKEN_TRANSLATIONS if it is a token. |
+`--------------------------------------------------*/
+
+static bool
+symbol_translation (symbol_t *this)
+{
+ /* Non-terminal? */
+ if (this->class == token_sym
+ && this->user_token_number != USER_NUMBER_ALIAS)
+ {
+ /* A token which translation has already been set? */
+ if (token_translations[this->user_token_number] != undeftoken->number)
+ complain (_("tokens %s and %s both assigned number %d"),
+ symbols[token_translations[this->user_token_number]]->tag,
+ this->tag, this->user_token_number);
+
+ token_translations[this->user_token_number] = this->number;
+ }
+
+ return TRUE;
+}
\f
/*===================\
`--------------------------------------------------------------*/
static char *
-get_type_name (int n, symbol_list * rule)
+get_type_name (int n, symbol_list *rule)
{
int i;
symbol_list *rp;
complain (_("invalid $ value"));
return NULL;
}
- i++;
+ ++i;
}
return rp->sym->type_name;
}
\f
+/*------------------------------------------------------------------.
+| Copy the character C to OOUT, and insert quadigraphs when needed. |
+`------------------------------------------------------------------*/
+
+static inline void
+copy_character (struct obstack *oout, int c)
+{
+ switch (c)
+ {
+ case '[':
+ obstack_sgrow (oout, "@<:@");
+ break;
+
+ case ']':
+ obstack_sgrow (oout, "@:>@");
+ break;
+
+ default:
+ obstack_1grow (oout, c);
+ }
+}
+
/*------------------------------------------------------------.
| Dump the string from FIN to OOUT if non null. MATCH is the |
| delimiter of the string (either ' or "). |
continue;
}
- obstack_1grow (oout, c);
+ copy_character (oout, c);
if (c == '\\')
{
c = getc (fin);
if (c == EOF)
fatal (_("unterminated string at end of file"));
- obstack_1grow (oout, c);
+ copy_character (oout, c);
if (c == '\n')
- lineno++;
+ ++lineno;
}
c = getc (fin);
ungetc (c, fin);
}
-/*-----------------------------------------------------------------.
-| Dump the wannabee comment from IN to OUT1 and OUT2 (which can be |
-| NULL). In fact we just saw a `/', which might or might not be a |
-| comment. In any case, copy what we saw. |
-| |
-| OUT2 might be NULL. |
-`-----------------------------------------------------------------*/
+
+/*------------------------------------------------------------------.
+| Dump the wannabee comment from IN to OOUT. In fact we just saw a |
+| `/', which might or might not be a comment. In any case, copy |
+| what we saw. |
+`------------------------------------------------------------------*/
static inline void
-copy_comment2 (FILE *fin, struct obstack *oout1, struct obstack *oout2)
+copy_comment (FILE *fin, struct obstack *oout)
{
int cplus_comment;
int ended;
int c;
/* We read a `/', output it. */
- obstack_1grow (oout1, '/');
- if (oout2)
- obstack_1grow (oout2, '/');
+ obstack_1grow (oout, '/');
switch ((c = getc (fin)))
{
return;
}
- obstack_1grow (oout1, c);
- if (oout2)
- obstack_1grow (oout2, c);
+ obstack_1grow (oout, c);
c = getc (fin);
ended = 0;
{
while (c == '*')
{
- obstack_1grow (oout1, c);
- if (oout2)
- obstack_1grow (oout2, c);
+ obstack_1grow (oout, c);
c = getc (fin);
}
if (c == '/')
{
- obstack_1grow (oout1, c);
- if (oout2)
- obstack_1grow (oout2, c);
+ obstack_1grow (oout, c);
ended = 1;
}
}
else if (c == '\n')
{
- lineno++;
- obstack_1grow (oout1, c);
- if (oout2)
- obstack_1grow (oout2, c);
+ ++lineno;
+ obstack_1grow (oout, c);
if (cplus_comment)
ended = 1;
else
fatal (_("unterminated comment"));
else
{
- obstack_1grow (oout1, c);
- if (oout2)
- obstack_1grow (oout2, c);
+ copy_character (oout, c);
c = getc (fin);
}
}
/*-------------------------------------------------------------------.
-| Dump the comment (actually the current string starting with a `/') |
-| from FIN to OOUT. |
+| FIN is pointing to a location (i.e., a `@'). Output to OOUT a |
+| reference to this location. RULE_LENGTH is the number of values in |
+| the current rule so far, which says where to find `$0' with |
+| respect to the top of the stack. |
`-------------------------------------------------------------------*/
static inline void
-copy_comment (FILE *fin, struct obstack *oout)
-{
- copy_comment2 (fin, oout, NULL);
-}
-
-
-/*-----------------------------------------------------------------.
-| FIN is pointing to a location (i.e., a `@'). Output to OOUT a |
-| reference to this location. STACK_OFFSET is the number of values |
-| in the current rule so far, which says where to find `$0' with |
-| respect to the top of the stack. |
-`-----------------------------------------------------------------*/
-
-static inline void
-copy_at (FILE *fin, struct obstack *oout, int stack_offset)
+copy_at (FILE *fin, struct obstack *oout, int rule_length)
{
- int c;
+ int c = getc (fin);
+ locations_flag = 1;
- c = getc (fin);
if (c == '$')
{
- obstack_sgrow (oout, "yyloc");
- locations_flag = 1;
+ obstack_sgrow (oout, "]b4_lhs_location[");
}
else if (isdigit (c) || c == '-')
{
ungetc (c, fin);
n = read_signed_integer (fin);
-
- obstack_fgrow1 (oout, "yylsp[%d]", n - stack_offset);
- locations_flag = 1;
+ if (n > rule_length)
+ complain (_("invalid value: %s%d"), "@", n);
+ else
+ obstack_fgrow2 (oout, "]b4_rhs_location([%d], [%d])[",
+ rule_length, n);
}
else
{
}
-/*-------------------------------------------------------------------.
-| FIN is pointing to a wannabee semantic value (i.e., a `$'). |
-| |
-| Possible inputs: $[<TYPENAME>]($|integer) |
-| |
-| Output to OOUT a reference to this semantic value. STACK_OFFSET is |
-| the number of values in the current rule so far, which says where |
-| to find `$0' with respect to the top of the stack. |
-`-------------------------------------------------------------------*/
+/*------------------------------------------------------------------.
+| FIN is pointing to a wannabee semantic value (i.e., a `$'). |
+| |
+| Possible inputs: $[<TYPENAME>]($|integer) |
+| |
+| Output to OOUT a reference to this semantic value. RULE_LENGTH is |
+| the number of values in the current rule so far, which says where |
+| to find `$0' with respect to the top of the stack. |
+`------------------------------------------------------------------*/
static inline void
copy_dollar (FILE *fin, struct obstack *oout,
- symbol_list *rule, int stack_offset)
+ symbol_list *rule, int rule_length)
{
int c = getc (fin);
const char *type_name = NULL;
if (c == '$')
{
- obstack_sgrow (oout, "yyval");
-
if (!type_name)
type_name = get_type_name (0, rule);
- if (type_name)
- obstack_fgrow1 (oout, ".%s", type_name);
if (!type_name && typed)
complain (_("$$ of `%s' has no declared type"),
rule->sym->tag);
+ if (!type_name)
+ type_name = "";
+ obstack_fgrow1 (oout,
+ "]b4_lhs_value([%s])[", type_name);
}
else if (isdigit (c) || c == '-')
{
ungetc (c, fin);
n = read_signed_integer (fin);
- if (!type_name && n > 0)
- type_name = get_type_name (n, rule);
-
- obstack_fgrow1 (oout, "yyvsp[%d]", n - stack_offset);
-
- if (type_name)
- obstack_fgrow1 (oout, ".%s", type_name);
- if (!type_name && typed)
- complain (_("$%d of `%s' has no declared type"),
- n, rule->sym->tag);
+ if (n > rule_length)
+ complain (_("invalid value: %s%d"), "$", n);
+ else
+ {
+ if (!type_name && n > 0)
+ type_name = get_type_name (n, rule);
+ if (!type_name && typed)
+ complain (_("$%d of `%s' has no declared type"),
+ n, rule->sym->tag);
+ if (!type_name)
+ type_name = "";
+ obstack_fgrow3 (oout, "]b4_rhs_value([%d], [%d], [%s])[",
+ rule_length, n, type_name);
+ }
}
else
{
`-------------------------------------------------------------------*/
static void
-copy_definition (void)
+copy_definition (struct obstack *oout)
{
int c;
/* -1 while reading a character if prev char was %. */
int after_percent;
-#if 0
if (!no_lines_flag)
{
- obstack_fgrow2 (&attrs_obstack, muscle_find ("linef"),
+ obstack_fgrow2 (oout, muscle_find ("linef"),
lineno, quotearg_style (c_quoting_style,
- muscle_find("filename")));
+ muscle_find ("filename")));
}
-#endif
after_percent = 0;
switch (c)
{
case '\n':
- obstack_1grow (&attrs_obstack, c);
- lineno++;
+ obstack_1grow (oout, c);
+ ++lineno;
break;
case '%':
case '\'':
case '"':
- copy_string (finput, &attrs_obstack, c);
+ copy_string (finput, oout, c);
break;
case '/':
- copy_comment (finput, &attrs_obstack);
+ copy_comment (finput, oout);
break;
case EOF:
fatal ("%s", _("unterminated `%{' definition"));
default:
- obstack_1grow (&attrs_obstack, c);
+ copy_character (oout, c);
}
c = getc (finput);
{
if (c == '}')
return;
- obstack_1grow (&attrs_obstack, '%');
+ obstack_1grow (oout, '%');
}
after_percent = 0;
}
char *typename = NULL;
/* The symbol being defined. */
- struct bucket *symbol = NULL;
+ symbol_t *symbol = NULL;
/* After `%token' and `%nterm', any number of symbols maybe be
defined. */
}
else if (token == tok_identifier && *symval->tag == '\"' && symbol)
{
- if (symval->alias)
- warn (_("symbol `%s' used more than once as a literal string"),
- symval->tag);
- else if (symbol->alias)
- warn (_("symbol `%s' given more than one literal string"),
- symbol->tag);
- else
- {
- symval->class = token_sym;
- symval->type_name = typename;
- symval->user_token_number = symbol->user_token_number;
- symbol->user_token_number = SALIAS;
- symval->alias = symbol;
- symbol->alias = symval;
- /* symbol and symval combined are only one symbol */
- nsyms--;
- }
+ symbol_make_alias (symbol, typename);
symbol = NULL;
}
else if (token == tok_identifier)
complain (_("symbol %s redefined"), symbol->tag);
symbol->class = what_is;
if (what_is == nterm_sym && oldclass != nterm_sym)
- symbol->value = nvars++;
+ symbol->number = nvars++;
+ if (what_is == token_sym && symbol->number == NUMBER_UNDEFINED)
+ symbol->number = ntokens++;
if (typename)
{
else if (symbol && token == tok_number)
{
symbol->user_token_number = numval;
+ /* User defined EOF token? */
+ if (numval == 0)
+ {
+ eoftoken = symbol;
+ eoftoken->number = 0;
+ /* It is always mapped to 0, so it was already counted in
+ NTOKENS. */
+ --ntokens;
+ }
}
else
{
complain (_("`%s' is invalid in %s"),
- token_buffer, (what_is == token_sym) ? "%token" : "%nterm");
+ token_buffer,
+ (what_is == token_sym) ? "%token" : "%nterm");
skip_to_char ('%');
}
}
char *name = NULL;
int prev = 0;
- lastprec++; /* Assign a new precedence level, never 0. */
+ /* Assign a new precedence level, never 0. */
+ ++lastprec;
for (;;)
{
symval->assoc = assoc;
if (symval->class == nterm_sym)
complain (_("symbol %s redefined"), symval->tag);
- symval->class = token_sym;
+ if (symval->number == NUMBER_UNDEFINED)
+ {
+ symval->number = ntokens++;
+ symval->class = token_sym;
+ }
if (name)
{ /* record the type, if one is specified */
if (symval->type_name == NULL)
}
else
{
- complain (_
- ("invalid text (%s) - number should be after identifier"),
-token_buffer);
+ complain
+ (_("invalid text (%s) - number should be after identifier"),
+ token_buffer);
skip_to_char ('%');
}
break;
{
int c;
int count = 0;
+ bool done = FALSE;
struct obstack union_obstack;
-
if (typed)
complain (_("multiple %s declarations"), "%union");
typed = 1;
- /* FIXME: I'm worried: are you sure attrs_obstack is properly
- filled? */
- if (no_lines_flag)
- obstack_1grow (&attrs_obstack, '\n');
-
+ MUSCLE_INSERT_INT ("stype_line", lineno);
obstack_init (&union_obstack);
obstack_sgrow (&union_obstack, "union");
- if (defines_flag)
- obstack_sgrow (&defines_obstack, "typedef union");
- c = getc (finput);
-
- while (c != EOF)
+ while (!done)
{
+ c = xgetc (finput);
+
/* If C contains '/', it is output by copy_comment (). */
if (c != '/')
- {
- obstack_1grow (&union_obstack, c);
- if (defines_flag)
- obstack_1grow (&defines_obstack, c);
- }
+ obstack_1grow (&union_obstack, c);
switch (c)
{
case '\n':
- lineno++;
+ ++lineno;
break;
case '/':
- copy_comment2 (finput, &defines_obstack, &union_obstack);
+ copy_comment (finput, &union_obstack);
break;
case '{':
- count++;
+ ++count;
break;
case '}':
+ /* FIXME: Errr. How could this happen???. --akim */
if (count == 0)
complain (_("unmatched %s"), "`}'");
count--;
- if (count <= 0)
- {
- if (defines_flag)
- obstack_sgrow (&defines_obstack, " YYSTYPE;\n");
- /* JF don't choke on trailing semi */
- c = skip_white_space ();
- if (c != ';')
- ungetc (c, finput);
- obstack_1grow (&union_obstack, 0);
- muscle_insert ("stype", obstack_finish (&union_obstack));
- return;
- }
+ if (!count)
+ done = TRUE;
+ break;
}
-
- c = getc (finput);
}
+ /* JF don't choke on trailing semi */
+ c = skip_white_space ();
+ if (c != ';')
+ ungetc (c, finput);
+ obstack_1grow (&union_obstack, 0);
+ muscle_insert ("stype", obstack_finish (&union_obstack));
}
| |
| Two symbols are entered in the table, one for the token symbol and |
| one for the literal. Both are given the <type>, if any, from the |
-| declaration. The ->user_token_number of the first is SALIAS and |
-| the ->user_token_number of the second is set to the number, if |
-| any, from the declaration. The two symbols are linked via |
-| pointers in their ->alias fields. |
+| declaration. The ->user_token_number of the first is |
+| USER_NUMBER_ALIAS and the ->user_token_number of the second is set |
+| to the number, if any, from the declaration. The two symbols are |
+| linked via pointers in their ->alias fields. |
| |
| During OUTPUT_DEFINES_TABLE, the symbol is reported thereafter, |
| only the literal string is retained it is the literal string that |
parse_thong_decl (void)
{
token_t token;
- struct bucket *symbol;
+ symbol_t *symbol;
char *typename = 0;
- int usrtoknum = SUNDEF;
+ int usrtoknum = USER_NUMBER_UNDEFINED;
token = lex (); /* fetch typename or first token */
if (token == tok_typename)
}
symval->class = token_sym;
symval->type_name = typename;
- symval->user_token_number = SALIAS;
+ symval->user_token_number = USER_NUMBER_ALIAS;
symbol = symval;
token = lex (); /* get number or literal string */
nsyms--;
}
+
static void
parse_muscle_decl (void)
{
int ch = ungetc (skip_white_space (), finput);
- char* muscle_key;
- char* muscle_value;
+ char *muscle_key;
+ char *muscle_value;
/* Read key. */
if (!isalpha (ch) && ch != '_')
}
+
+/*---------------------------------.
+| Parse a double quoted parameter. |
+`---------------------------------*/
+
+static const char *
+parse_dquoted_param (const char *from)
+{
+ struct obstack param_obstack;
+ const char *param = NULL;
+ int c;
+
+ obstack_init (¶m_obstack);
+ c = skip_white_space ();
+
+ if (c != '"')
+ {
+ complain (_("invalid %s declaration"), from);
+ ungetc (c, finput);
+ skip_to_char ('%');
+ return NULL;
+ }
+
+ while ((c = literalchar ()) != '"')
+ obstack_1grow (¶m_obstack, c);
+
+ obstack_1grow (¶m_obstack, '\0');
+ param = obstack_finish (¶m_obstack);
+
+ if (c != '"' || strlen (param) == 0)
+ {
+ complain (_("invalid %s declaration"), from);
+ if (c != '"')
+ ungetc (c, finput);
+ skip_to_char ('%');
+ return NULL;
+ }
+
+ return param;
+}
+
/*----------------------------------.
| Parse what comes after %skeleton. |
`----------------------------------*/
-void
+static void
parse_skel_decl (void)
{
- /* Complete with parse_dquoted_param () on the CVS branch 1.29. */
+ skeleton = parse_dquoted_param ("%skeleton");
}
/*----------------------------------------------------------------.
| Read from finput until `%%' is seen. Discard the `%%'. Handle |
| any `%' declarations, and copy the contents of any `%{ ... %}' |
-| groups to ATTRS_OBSTACK. |
+| groups to PRE_PROLOGUE_OBSTACK or POST_PROLOGUE_OBSTACK. |
`----------------------------------------------------------------*/
static void
return;
case tok_percent_left_curly:
- copy_definition ();
+ if (!typed)
+ copy_definition (&pre_prologue_obstack);
+ else
+ copy_definition (&post_prologue_obstack);
break;
case tok_token:
case tok_stropt:
case tok_intopt:
case tok_obsolete:
- case tok_illegal:
- abort ();
+ assert (0);
break;
+ case tok_illegal:
default:
complain (_("unrecognized: %s"), token_buffer);
skip_to_char ('%');
}
}
\f
-/*-------------------------------------------------------------------.
-| Assuming that a `{' has just been seen, copy everything up to the |
-| matching `}' into the actions file. STACK_OFFSET is the number of |
-| values in the current rule so far, which says where to find `$0' |
-| with respect to the top of the stack. |
-`-------------------------------------------------------------------*/
+/*------------------------------------------------------------------.
+| Assuming that a `{' has just been seen, copy everything up to the |
+| matching `}' into ACTION_OBSTACK. |
+| |
+| RULE_LENGTH is the number of values in the current rule so far, |
+| which says where to find `$0' with respect to the top of the |
+| stack. It is not the same as the rule->length in the case of mid |
+| rule actions. |
+| |
+| This routine is used for actions. |
+`------------------------------------------------------------------*/
static void
-copy_action (symbol_list *rule, int stack_offset)
+parse_action (symbol_list *rule, int rule_length)
{
- int c;
- int count;
- char buf[4096];
-
- /* offset is always 0 if parser has already popped the stack pointer */
- if (semantic_parser)
- stack_offset = 0;
-
- obstack_fgrow1 (&action_obstack, "\ncase %d:\n", nrules);
-
- if (!no_lines_flag)
- {
- obstack_fgrow2 (&action_obstack, muscle_find ("linef"),
- lineno, quotearg_style (c_quoting_style,
- muscle_find ("filename")));
- }
- obstack_1grow (&action_obstack, '{');
-
- count = 1;
- c = getc (finput);
-
+ int count = 1;
+ rule->action_line = lineno;
while (count > 0)
{
- while (c != '}')
- {
- switch (c)
- {
- case '\n':
- obstack_1grow (&action_obstack, c);
- lineno++;
- break;
-
- case '{':
- obstack_1grow (&action_obstack, c);
- count++;
- break;
-
- case '\'':
- case '"':
- copy_string (finput, &action_obstack, c);
- break;
-
- case '/':
- copy_comment (finput, &action_obstack);
- break;
-
- case '$':
- copy_dollar (finput, &action_obstack,
- rule, stack_offset);
- break;
-
- case '@':
- copy_at (finput, &action_obstack,
- stack_offset);
- break;
-
- case EOF:
- fatal (_("unmatched %s"), "`{'");
-
- default:
- obstack_1grow (&action_obstack, c);
- }
-
- c = getc (finput);
- }
-
- /* above loop exits when c is '}' */
-
+ int c;
+ while ((c = getc (finput)) != '}')
+ switch (c)
+ {
+ case '\n':
+ copy_character (&action_obstack, c);
+ ++lineno;
+ break;
+
+ case '{':
+ copy_character (&action_obstack, c);
+ ++count;
+ break;
+
+ case '\'':
+ case '"':
+ copy_string (finput, &action_obstack, c);
+ break;
+
+ case '/':
+ copy_comment (finput, &action_obstack);
+ break;
+
+ case '$':
+ copy_dollar (finput, &action_obstack, rule, rule_length);
+ break;
+
+ case '@':
+ copy_at (finput, &action_obstack, rule_length);
+ break;
+
+ case EOF:
+ fatal (_("unmatched %s"), "`{'");
+
+ default:
+ copy_character (&action_obstack, c);
+ }
+
+ /* Above loop exits when C is '}'. */
if (--count)
- {
- obstack_1grow (&action_obstack, c);
- c = getc (finput);
- }
+ copy_character (&action_obstack, c);
}
- obstack_sgrow (&action_obstack, ";\n break;}");
+ obstack_1grow (&action_obstack, '\0');
+ rule->action = obstack_finish (&action_obstack);
}
-\f
-/*-------------------------------------------------------------------.
-| After `%guard' is seen in the input file, copy the actual guard |
-| into the guards file. If the guard is followed by an action, copy |
-| that into the actions file. STACK_OFFSET is the number of values |
-| in the current rule so far, which says where to find `$0' with |
-| respect to the top of the stack, for the simple parser in which |
-| the stack is not popped until after the guard is run. |
-`-------------------------------------------------------------------*/
-
-static void
-copy_guard (symbol_list *rule, int stack_offset)
-{
- int c;
- int count;
- int brace_flag = 0;
-
- /* offset is always 0 if parser has already popped the stack pointer */
- if (semantic_parser)
- stack_offset = 0;
-
- obstack_fgrow1 (&guard_obstack, "\ncase %d:\n", nrules);
- if (!no_lines_flag)
- obstack_fgrow2 (&guard_obstack, muscle_find ("linef"),
- lineno, quotearg_style (c_quoting_style,
- muscle_find ("filename")));
- obstack_1grow (&guard_obstack, '{');
-
- count = 0;
- c = getc (finput);
-
- while (brace_flag ? (count > 0) : (c != ';'))
- {
- switch (c)
- {
- case '\n':
- obstack_1grow (&guard_obstack, c);
- lineno++;
- break;
-
- case '{':
- obstack_1grow (&guard_obstack, c);
- brace_flag = 1;
- count++;
- break;
-
- case '}':
- obstack_1grow (&guard_obstack, c);
- if (count > 0)
- count--;
- else
- {
- complain (_("unmatched %s"), "`}'");
- c = getc (finput); /* skip it */
- }
- break;
- case '\'':
- case '"':
- copy_string (finput, &guard_obstack, c);
- break;
-
- case '/':
- copy_comment (finput, &guard_obstack);
- break;
-
- case '$':
- copy_dollar (finput, &guard_obstack, rule, stack_offset);
- break;
-
- case '@':
- copy_at (finput, &guard_obstack, stack_offset);
- break;
-
- case EOF:
- fatal ("%s", _("unterminated %guard clause"));
-
- default:
- obstack_1grow (&guard_obstack, c);
- }
-
- if (c != '}' || count != 0)
- c = getc (finput);
- }
-
- c = skip_white_space ();
-
- obstack_sgrow (&guard_obstack, ";\n break;}");
- if (c == '{')
- copy_action (rule, stack_offset);
- else if (c == '=')
- {
- c = getc (finput); /* why not skip_white_space -wjh */
- if (c == '{')
- copy_action (rule, stack_offset);
- }
- else
- ungetc (c, finput);
-}
\f
-static void
-record_rule_line (void)
-{
- /* Record each rule's source line number in rline table. */
-
- if (nrules >= rline_allocated)
- {
- rline_allocated = nrules * 2;
- rline = XREALLOC (rline, short, rline_allocated);
- }
- rline[nrules] = lineno;
-}
-
-
/*-------------------------------------------------------------------.
| Generate a dummy symbol, a nonterminal, whose name cannot conflict |
| with the user's names. |
`-------------------------------------------------------------------*/
-static bucket *
+static symbol_t *
gensym (void)
{
/* Incremented for each generated symbol */
static int gensym_count = 0;
static char buf[256];
- bucket *sym;
+ symbol_t *sym;
sprintf (buf, "@%d", ++gensym_count);
token_buffer = buf;
sym = getsym (token_buffer);
sym->class = nterm_sym;
- sym->value = nvars++;
+ sym->number = nvars++;
return sym;
}
-
-#if 0
-/*------------------------------------------------------------------.
-| read in a %type declaration and record its information for |
-| get_type_name to access. This is unused. It is only called from |
-| the #if 0 part of readgram |
-`------------------------------------------------------------------*/
-
-static int
-get_type (void)
-{
- int k;
- token_t token;
- char *name;
-
- token = lex ();
-
- if (token != tok_typename)
- {
- complain (_("invalid %s declaration"), "%type");
- return t;
- }
-
- name = xstrdup (token_buffer);
-
- for (;;)
- {
- token = lex ();
-
- switch (token)
- {
- case tok_semicolon:
- return lex ();
-
- case tok_comma:
- break;
-
- case tok_identifier:
- if (symval->type_name == NULL)
- symval->type_name = name;
- else if (strcmp (name, symval->type_name) != 0)
- complain (_("type redeclaration for %s"), symval->tag);
-
- break;
-
- default:
- return token;
- }
- }
-}
-
-#endif
\f
-/*------------------------------------------------------------------.
-| Parse the input grammar into a one symbol_list structure. Each |
-| rule is represented by a sequence of symbols: the left hand side |
-| followed by the contents of the right hand side, followed by a |
-| null pointer instead of a symbol to terminate the rule. The next |
-| symbol is the lhs of the following rule. |
-| |
-| All guards and actions are copied out to the appropriate files, |
-| labelled by the rule number they apply to. |
-`------------------------------------------------------------------*/
+/*-------------------------------------------------------------------.
+| Parse the input grammar into a one symbol_list structure. Each |
+| rule is represented by a sequence of symbols: the left hand side |
+| followed by the contents of the right hand side, followed by a |
+| null pointer instead of a symbol to terminate the rule. The next |
+| symbol is the lhs of the following rule. |
+| |
+| All actions are copied out, labelled by the rule number they apply |
+| to. |
+| |
+| Bison used to allow some %directives in the rules sections, but |
+| this is no longer consider appropriate: (i) the documented grammar |
+| doesn't claim it, (ii), it would promote bad style, (iii), error |
+| recovery for %directives consists in skipping the junk until a `%' |
+| is seen and helrp synchronizing. This scheme is definitely wrong |
+| in the rules section. |
+`-------------------------------------------------------------------*/
static void
readgram (void)
{
token_t t;
- bucket *lhs = NULL;
- symbol_list *p;
- symbol_list *p1;
- bucket *bp;
+ symbol_t *lhs = NULL;
+ symbol_list *p = NULL;
+ symbol_list *p1 = NULL;
/* Points to first symbol_list of current rule. its symbol is the
lhs of the rule. */
- symbol_list *crule;
+ symbol_list *crule = NULL;
/* Points to the symbol_list preceding crule. */
- symbol_list *crule1;
-
- p1 = NULL;
+ symbol_list *crule1 = NULL;
t = lex ();
while (t != tok_two_percents && t != tok_eof)
- {
- if (t == tok_identifier || t == tok_bar)
- {
- int action_flag = 0;
- /* Number of symbols in rhs of this rule so far */
- int rulelength = 0;
- int xactions = 0; /* JF for error checking */
- bucket *first_rhs = 0;
-
- if (t == tok_identifier)
- {
- lhs = symval;
-
- if (!start_flag)
- {
- startval = lhs;
- start_flag = 1;
- }
-
- t = lex ();
- if (t != tok_colon)
- {
- complain (_("ill-formed rule: initial symbol not followed by colon"));
- unlex (t);
- }
- }
-
- if (nrules == 0 && t == tok_bar)
- {
- complain (_("grammar starts with vertical bar"));
- lhs = symval; /* BOGUS: use a random symval */
- }
- /* start a new rule and record its lhs. */
-
- nrules++;
- nitems++;
-
- record_rule_line ();
+ if (t == tok_identifier || t == tok_bar)
+ {
+ int action_flag = 0;
+ /* Number of symbols in rhs of this rule so far */
+ int rulelength = 0;
+ int xactions = 0; /* JF for error checking */
+ symbol_t *first_rhs = 0;
+
+ if (t == tok_identifier)
+ {
+ lhs = symval;
+
+ if (!start_flag)
+ {
+ startval = lhs;
+ start_flag = 1;
+ }
- p = XCALLOC (symbol_list, 1);
- p->sym = lhs;
+ t = lex ();
+ if (t != tok_colon)
+ {
+ complain (_("ill-formed rule: initial symbol not followed by colon"));
+ unlex (t);
+ }
+ }
+
+ if (nrules == 0 && t == tok_bar)
+ {
+ complain (_("grammar starts with vertical bar"));
+ lhs = symval; /* BOGUS: use a random symval */
+ }
+ /* start a new rule and record its lhs. */
+
+ ++nrules;
+ ++nritems;
+
+ p = symbol_list_new (lhs);
+
+ crule1 = p1;
+ if (p1)
+ p1->next = p;
+ else
+ grammar = p;
- crule1 = p1;
- if (p1)
- p1->next = p;
- else
- grammar = p;
+ p1 = p;
+ crule = p;
- p1 = p;
- crule = p;
+ /* mark the rule's lhs as a nonterminal if not already so. */
- /* mark the rule's lhs as a nonterminal if not already so. */
+ if (lhs->class == unknown_sym)
+ {
+ lhs->class = nterm_sym;
+ lhs->number = nvars;
+ ++nvars;
+ }
+ else if (lhs->class == token_sym)
+ complain (_("rule given for %s, which is a token"), lhs->tag);
- if (lhs->class == unknown_sym)
- {
- lhs->class = nterm_sym;
- lhs->value = nvars;
- nvars++;
- }
- else if (lhs->class == token_sym)
- complain (_("rule given for %s, which is a token"), lhs->tag);
+ /* read the rhs of the rule. */
- /* read the rhs of the rule. */
+ for (;;)
+ {
+ t = lex ();
+ if (t == tok_prec)
+ {
+ t = lex ();
+ crule->ruleprec = symval;
+ t = lex ();
+ }
+
+ if (!(t == tok_identifier || t == tok_left_curly))
+ break;
- for (;;)
- {
- t = lex ();
- if (t == tok_prec)
- {
- t = lex ();
- crule->ruleprec = symval;
- t = lex ();
- }
-
- if (!(t == tok_identifier || t == tok_left_curly))
- break;
-
- /* If next token is an identifier, see if a colon follows it.
- If one does, exit this rule now. */
- if (t == tok_identifier)
- {
- bucket *ssave;
- token_t t1;
-
- ssave = symval;
- t1 = lex ();
- unlex (t1);
- symval = ssave;
- if (t1 == tok_colon)
+ /* If next token is an identifier, see if a colon follows it.
+ If one does, exit this rule now. */
+ if (t == tok_identifier)
+ {
+ symbol_t *ssave;
+ token_t t1;
+
+ ssave = symval;
+ t1 = lex ();
+ unlex (t1);
+ symval = ssave;
+ if (t1 == tok_colon)
+ {
+ warn (_("previous rule lacks an ending `;'"));
break;
-
- if (!first_rhs) /* JF */
- first_rhs = symval;
- /* Not followed by colon =>
- process as part of this rule's rhs. */
- }
-
- /* If we just passed an action, that action was in the middle
- of a rule, so make a dummy rule to reduce it to a
- non-terminal. */
- if (action_flag)
- {
- bucket *sdummy;
-
- /* Since the action was written out with this rule's
- number, we must give the new rule this number by
- inserting the new rule before it. */
-
- /* Make a dummy nonterminal, a gensym. */
- sdummy = gensym ();
-
- /* Make a new rule, whose body is empty,
- before the current one, so that the action
- just read can belong to it. */
- nrules++;
- nitems++;
- record_rule_line ();
- p = XCALLOC (symbol_list, 1);
- if (crule1)
- crule1->next = p;
- else
- grammar = p;
- p->sym = sdummy;
- crule1 = XCALLOC (symbol_list, 1);
- p->next = crule1;
- crule1->next = crule;
-
- /* Insert the dummy generated by that rule into this
- rule. */
- nitems++;
- p = XCALLOC (symbol_list, 1);
- p->sym = sdummy;
- p1->next = p;
- p1 = p;
-
- action_flag = 0;
- }
-
- if (t == tok_identifier)
- {
- nitems++;
- p = XCALLOC (symbol_list, 1);
- p->sym = symval;
- p1->next = p;
- p1 = p;
- }
- else /* handle an action. */
- {
- copy_action (crule, rulelength);
- action_flag = 1;
- xactions++; /* JF */
- }
- rulelength++;
- } /* end of read rhs of rule */
-
- /* Put an empty link in the list to mark the end of this rule */
- p = XCALLOC (symbol_list, 1);
- p1->next = p;
- p1 = p;
-
- if (t == tok_prec)
- {
- complain (_("two @prec's in a row"));
- t = lex ();
- crule->ruleprec = symval;
- t = lex ();
- }
- if (t == tok_guard)
- {
- if (!semantic_parser)
- complain (_("%%guard present but %%semantic_parser not specified"));
-
- copy_guard (crule, rulelength);
- t = lex ();
- }
- else if (t == tok_left_curly)
- {
- /* This case never occurs -wjh */
- if (action_flag)
- complain (_("two actions at end of one rule"));
- copy_action (crule, rulelength);
- action_flag = 1;
- xactions++; /* -wjh */
- t = lex ();
- }
- /* If $$ is being set in default way, report if any type
- mismatch. */
- else if (!xactions
- && first_rhs && lhs->type_name != first_rhs->type_name)
- {
- if (lhs->type_name == 0
- || first_rhs->type_name == 0
- || strcmp (lhs->type_name, first_rhs->type_name))
- complain (_("type clash (`%s' `%s') on default action"),
- lhs->type_name ? lhs->type_name : "",
- first_rhs->type_name ? first_rhs->type_name : "");
- }
- /* Warn if there is no default for $$ but we need one. */
- else if (!xactions && !first_rhs && lhs->type_name != 0)
- complain (_("empty rule for typed nonterminal, and no action"));
- if (t == tok_semicolon)
+ }
+
+ if (!first_rhs) /* JF */
+ first_rhs = symval;
+ /* Not followed by colon =>
+ process as part of this rule's rhs. */
+ }
+
+ /* If we just passed an action, that action was in the middle
+ of a rule, so make a dummy rule to reduce it to a
+ non-terminal. */
+ if (action_flag)
+ {
+ /* Since the action was written out with this rule's
+ number, we must give the new rule this number by
+ inserting the new rule before it. */
+
+ /* Make a dummy nonterminal, a gensym. */
+ symbol_t *sdummy = gensym ();
+
+ /* Make a new rule, whose body is empty, before the
+ current one, so that the action just read can
+ belong to it. */
+ ++nrules;
+ ++nritems;
+ p = symbol_list_new (sdummy);
+ /* Attach its lineno to that of the host rule. */
+ p->line = crule->line;
+ /* Move the action from the host rule to this one. */
+ p->action = crule->action;
+ p->action_line = crule->action_line;
+ crule->action = NULL;
+
+ if (crule1)
+ crule1->next = p;
+ else
+ grammar = p;
+ /* End of the rule. */
+ crule1 = symbol_list_new (NULL);
+ crule1->next = crule;
+
+ p->next = crule1;
+
+ /* Insert the dummy generated by that rule into this
+ rule. */
+ ++nritems;
+ p = symbol_list_new (sdummy);
+ p1->next = p;
+ p1 = p;
+
+ action_flag = 0;
+ }
+
+ if (t == tok_identifier)
+ {
+ ++nritems;
+ p = symbol_list_new (symval);
+ p1->next = p;
+ p1 = p;
+ }
+ else /* handle an action. */
+ {
+ parse_action (crule, rulelength);
+ action_flag = 1;
+ ++xactions; /* JF */
+ }
+ ++rulelength;
+ } /* end of read rhs of rule */
+
+ /* Put an empty link in the list to mark the end of this rule */
+ p = symbol_list_new (NULL);
+ p1->next = p;
+ p1 = p;
+
+ if (t == tok_prec)
+ {
+ complain (_("two @prec's in a row"));
t = lex ();
- }
-#if 0
- /* these things can appear as alternatives to rules. */
-/* NO, they cannot.
- a) none of the documentation allows them
- b) most of them scan forward until finding a next %
- thus they may swallow lots of intervening rules
-*/
- else if (t == tok_token)
- {
- parse_token_decl (token_sym, nterm_sym);
- t = lex ();
- }
- else if (t == tok_nterm)
- {
- parse_token_decl (nterm_sym, token_sym);
- t = lex ();
- }
- else if (t == tok_type)
- {
- t = get_type ();
- }
- else if (t == tok_union)
- {
- parse_union_decl ();
- t = lex ();
- }
- else if (t == tok_expect)
- {
- parse_expect_decl ();
- t = lex ();
- }
- else if (t == tok_start)
- {
- parse_start_decl ();
- t = lex ();
- }
-#endif
-
- else
- {
- complain (_("invalid input: %s"), quote (token_buffer));
+ crule->ruleprec = symval;
+ t = lex ();
+ }
+
+ if (t == tok_left_curly)
+ {
+ /* This case never occurs -wjh */
+ if (action_flag)
+ complain (_("two actions at end of one rule"));
+ parse_action (crule, rulelength);
+ action_flag = 1;
+ ++xactions; /* -wjh */
+ t = lex ();
+ }
+ /* If $$ is being set in default way, report if any type
+ mismatch. */
+ else if (!xactions
+ && first_rhs && lhs->type_name != first_rhs->type_name)
+ {
+ if (lhs->type_name == 0
+ || first_rhs->type_name == 0
+ || strcmp (lhs->type_name, first_rhs->type_name))
+ complain (_("type clash (`%s' `%s') on default action"),
+ lhs->type_name ? lhs->type_name : "",
+ first_rhs->type_name ? first_rhs->type_name : "");
+ }
+ /* Warn if there is no default for $$ but we need one. */
+ else if (!xactions && !first_rhs && lhs->type_name != 0)
+ complain (_("empty rule for typed nonterminal, and no action"));
+ if (t == tok_two_percents || t == tok_eof)
+ warn (_("previous rule lacks an ending `;'"));
+ if (t == tok_semicolon)
t = lex ();
- }
- }
+ }
+ else
+ {
+ complain (_("invalid input: %s"), quote (token_buffer));
+ t = lex ();
+ }
/* grammar has been read. Do some checking */
- if (nsyms > MAXSHORT)
- fatal (_("too many symbols (tokens plus nonterminals); maximum %d"),
- MAXSHORT);
if (nrules == 0)
fatal (_("no rules in the input grammar"));
/* Report any undefined symbols and consider them nonterminals. */
+ symbols_do (symbol_check_defined, NULL);
+
+ /* Insert the initial rule, which line is that of the first rule
+ (not that of the start symbol):
+
+ axiom: %start EOF. */
+ p = symbol_list_new (axiom);
+ p->line = grammar->line;
+ p->next = symbol_list_new (startval);
+ p->next->next = symbol_list_new (eoftoken);
+ p->next->next->next = symbol_list_new (NULL);
+ p->next->next->next->next = grammar;
+ nrules += 1;
+ nritems += 3;
+ grammar = p;
+ startval = axiom;
+
+ if (nsyms > SHRT_MAX)
+ fatal (_("too many symbols (tokens plus nonterminals); maximum %d"),
+ SHRT_MAX);
- for (bp = firstsymbol; bp; bp = bp->next)
- if (bp->class == unknown_sym)
- {
- complain (_
- ("symbol %s is used, but is not defined as a token and has no rules"),
- bp->tag);
- bp->class = nterm_sym;
- bp->value = nvars++;
- }
-
- ntokens = nsyms - nvars;
+ assert (nsyms == ntokens + nvars);
}
/* At the end of the grammar file, some C source code must
static void
read_additionnal_code (void)
{
- char c;
+ int c;
struct obstack el_obstack;
obstack_init (&el_obstack);
+ if (!no_lines_flag)
+ {
+ obstack_fgrow2 (&el_obstack, muscle_find ("linef"),
+ lineno, quotearg_style (c_quoting_style,
+ muscle_find ("filename")));
+ }
+
while ((c = getc (finput)) != EOF)
- obstack_1grow (&el_obstack, c);
+ copy_character (&el_obstack, c);
obstack_1grow (&el_obstack, 0);
muscle_insert ("epilogue", obstack_finish (&el_obstack));
}
\f
-/*--------------------------------------------------------------.
-| For named tokens, but not literal ones, define the name. The |
-| value is the user token number. |
-`--------------------------------------------------------------*/
+/*------------------------------------------------------------------.
+| Set TOKEN_TRANSLATIONS. Check that no two symbols share the same |
+| number. |
+`------------------------------------------------------------------*/
static void
-output_token_defines (struct obstack *oout)
+token_translations_init (void)
{
- bucket *bp;
- char *cp, *symbol;
- char c;
+ int num_256_available_p = TRUE;
+ int i;
- for (bp = firstsymbol; bp; bp = bp->next)
+ /* Find the highest user token number, and whether 256, the POSIX
+ preferred user token number for the error token, is used. */
+ max_user_token_number = 0;
+ for (i = 0; i < ntokens; ++i)
{
- symbol = bp->tag; /* get symbol */
-
- if (bp->value >= ntokens)
- continue;
- if (bp->user_token_number == SALIAS)
- continue;
- if ('\'' == *symbol)
- continue; /* skip literal character */
- if (bp == errtoken)
- continue; /* skip error token */
- if ('\"' == *symbol)
+ symbol_t *this = symbols[i];
+ if (this->user_token_number != USER_NUMBER_UNDEFINED)
{
- /* use literal string only if given a symbol with an alias */
- if (bp->alias)
- symbol = bp->alias->tag;
- else
- continue;
+ if (this->user_token_number > max_user_token_number)
+ max_user_token_number = this->user_token_number;
+ if (this->user_token_number == 256)
+ num_256_available_p = FALSE;
}
-
- /* Don't #define nonliteral tokens whose names contain periods. */
- cp = symbol;
- while ((c = *cp++) && c != '.');
- if (c != '\0')
- continue;
-
- obstack_fgrow2 (oout, "# define\t%s\t%d\n",
- symbol, bp->user_token_number);
- if (semantic_parser)
- /* FIXME: This is certainly dead wrong, and should be just as
- above. --akim. */
- obstack_fgrow2 (oout, "# define\tT%s\t%d\n", symbol, bp->value);
}
-}
+ /* If 256 is not used, assign it to error, to follow POSIX. */
+ if (num_256_available_p
+ && errtoken->user_token_number == USER_NUMBER_UNDEFINED)
+ errtoken->user_token_number = 256;
-/*------------------------------------------------------------------.
-| Set TOKEN_TRANSLATIONS. Check that no two symbols share the same |
-| number. |
-`------------------------------------------------------------------*/
+ /* Set the missing user numbers. */
+ if (max_user_token_number < 256)
+ max_user_token_number = 256;
-static void
-token_translations_init (void)
-{
- bucket *bp = NULL;
- int i;
+ for (i = 0; i < ntokens; ++i)
+ {
+ symbol_t *this = symbols[i];
+ if (this->user_token_number == USER_NUMBER_UNDEFINED)
+ this->user_token_number = ++max_user_token_number;
+ if (this->user_token_number > max_user_token_number)
+ max_user_token_number = this->user_token_number;
+ }
- token_translations = XCALLOC (short, max_user_token_number + 1);
+ token_translations = XCALLOC (symbol_number_t, max_user_token_number + 1);
/* Initialize all entries for literal tokens to 2, the internal
token number for $undefined., which represents all invalid
inputs. */
- for (i = 0; i <= max_user_token_number; i++)
- token_translations[i] = 2;
-
- for (bp = firstsymbol; bp; bp = bp->next)
- {
- /* Non-terminal? */
- if (bp->value >= ntokens)
- continue;
- /* A token string alias? */
- if (bp->user_token_number == SALIAS)
- continue;
-
- assert (bp->user_token_number != SUNDEF);
-
- /* A token which translation has already been set? */
- if (token_translations[bp->user_token_number] != 2)
- complain (_("tokens %s and %s both assigned number %d"),
- tags[token_translations[bp->user_token_number]],
- bp->tag, bp->user_token_number);
- token_translations[bp->user_token_number] = bp->value;
- }
+ for (i = 0; i < max_user_token_number + 1; i++)
+ token_translations[i] = undeftoken->number;
+ symbols_do (symbol_translation, NULL);
}
-/*------------------------------------------------------------------.
-| Assign symbol numbers, and write definition of token names into |
-| FDEFINES. Set up vectors TAGS and SPREC of names and precedences |
-| of symbols. |
-`------------------------------------------------------------------*/
+/*----------------------------------------------------------------.
+| Assign symbol numbers, and write definition of token names into |
+| FDEFINES. Set up vectors SYMBOL_TABLE, TAGS of symbols. |
+`----------------------------------------------------------------*/
static void
packsymbols (void)
{
- bucket *bp = NULL;
- int tokno = 1;
- int last_user_token_number;
- static char DOLLAR[] = "$";
-
- tags = XCALLOC (char *, nsyms + 1);
- user_toknums = XCALLOC (short, nsyms + 1);
+ symbols = XCALLOC (symbol_t *, nsyms);
- sprec = XCALLOC (short, nsyms);
- sassoc = XCALLOC (short, nsyms);
-
- /* The EOF token. */
- tags[0] = DOLLAR;
- user_toknums[0] = 0;
-
- max_user_token_number = 256;
- last_user_token_number = 256;
-
- for (bp = firstsymbol; bp; bp = bp->next)
- {
- if (bp->class == nterm_sym)
- {
- bp->value += ntokens;
- }
- else if (bp->alias)
- {
- /* this symbol and its alias are a single token defn.
- allocate a tokno, and assign to both check agreement of
- ->prec and ->assoc fields and make both the same */
- if (bp->value == 0)
- bp->value = bp->alias->value = tokno++;
-
- if (bp->prec != bp->alias->prec)
- {
- if (bp->prec != 0 && bp->alias->prec != 0
- && bp->user_token_number == SALIAS)
- complain (_("conflicting precedences for %s and %s"),
- bp->tag, bp->alias->tag);
- if (bp->prec != 0)
- bp->alias->prec = bp->prec;
- else
- bp->prec = bp->alias->prec;
- }
-
- if (bp->assoc != bp->alias->assoc)
- {
- if (bp->assoc != 0 && bp->alias->assoc != 0
- && bp->user_token_number == SALIAS)
- complain (_("conflicting assoc values for %s and %s"),
- bp->tag, bp->alias->tag);
- if (bp->assoc != 0)
- bp->alias->assoc = bp->assoc;
- else
- bp->assoc = bp->alias->assoc;
- }
-
- if (bp->user_token_number == SALIAS)
- continue; /* do not do processing below for SALIASs */
-
- }
- else /* bp->class == token_sym */
- {
- bp->value = tokno++;
- }
-
- if (bp->class == token_sym)
- {
- if (bp->user_token_number == SUNDEF)
- bp->user_token_number = ++last_user_token_number;
- if (bp->user_token_number > max_user_token_number)
- max_user_token_number = bp->user_token_number;
- }
-
- tags[bp->value] = bp->tag;
- user_toknums[bp->value] = bp->user_token_number;
- sprec[bp->value] = bp->prec;
- sassoc[bp->value] = bp->assoc;
- }
+ symbols_do (symbol_check_alias_consistence, NULL);
+ symbols_do (symbol_pack, NULL);
token_translations_init ();
- error_token_number = errtoken->value;
-
- {
- struct obstack tokendefs;
- obstack_init (&tokendefs);
- output_token_defines (&tokendefs);
- obstack_1grow (&tokendefs, 0);
- muscle_insert ("tokendef", xstrdup (obstack_finish (&tokendefs)));
- obstack_free (&tokendefs, NULL);
- }
-
-#if 0
- if (!no_parser_flag)
- output_token_defines (&table_obstack);
-#endif
-
if (startval->class == unknown_sym)
fatal (_("the start symbol %s is undefined"), startval->tag);
else if (startval->class == token_sym)
fatal (_("the start symbol %s is a token"), startval->tag);
- start_symbol = startval->value;
-
- if (defines_flag)
- {
- output_token_defines (&defines_obstack);
-
- if (!pure_parser)
- {
- if (spec_name_prefix)
- obstack_fgrow1 (&defines_obstack, "\nextern YYSTYPE %slval;\n",
- spec_name_prefix);
- else
- obstack_sgrow (&defines_obstack,
- "\nextern YYSTYPE yylval;\n");
- }
-
- if (semantic_parser)
- {
- int i;
-
- for (i = ntokens; i < nsyms; i++)
- {
- /* don't make these for dummy nonterminals made by gensym. */
- if (*tags[i] != '@')
- obstack_fgrow2 (&defines_obstack,
- "# define\tNT%s\t%d\n", tags[i], i);
- }
-#if 0
- /* `fdefines' is now a temporary file, so we need to copy its
- contents in `done', so we can't close it here. */
- fclose (fdefines);
- fdefines = NULL;
-#endif
- }
- }
+ start_symbol = startval->number;
}
/*---------------------------------------------------------------.
| Convert the rules into the representation using RRHS, RLHS and |
-| RITEMS. |
+| RITEM. |
`---------------------------------------------------------------*/
static void
packgram (void)
{
- int itemno;
+ unsigned int itemno;
int ruleno;
symbol_list *p;
- bucket *ruleprec;
-
- ritem = XCALLOC (short, nitems + 1);
- rlhs = XCALLOC (short, nrules) - 1;
- rrhs = XCALLOC (short, nrules) - 1;
- rprec = XCALLOC (short, nrules) - 1;
- rprecsym = XCALLOC (short, nrules) - 1;
- rassoc = XCALLOC (short, nrules) - 1;
+ ritem = XCALLOC (item_number_t, nritems);
+ rules = XCALLOC (rule_t, nrules) - 1;
itemno = 0;
ruleno = 1;
p = grammar;
while (p)
{
- rlhs[ruleno] = p->sym->value;
- rrhs[ruleno] = itemno;
- ruleprec = p->ruleprec;
+ symbol_t *ruleprec = p->ruleprec;
+ rules[ruleno].user_number = ruleno;
+ rules[ruleno].number = ruleno;
+ rules[ruleno].lhs = p->sym;
+ rules[ruleno].rhs = ritem + itemno;
+ rules[ruleno].line = p->line;
+ rules[ruleno].useful = TRUE;
+ rules[ruleno].action = p->action;
+ rules[ruleno].action_line = p->action_line;
p = p->next;
while (p && p->sym)
{
- ritem[itemno++] = p->sym->value;
+ /* item_number_t = symbol_number_t.
+ But the former needs to contain more: negative rule numbers. */
+ ritem[itemno++] = symbol_number_as_item_number (p->sym->number);
/* A rule gets by default the precedence and associativity
of the last token in it. */
if (p->sym->class == token_sym)
- {
- rprec[ruleno] = p->sym->prec;
- rassoc[ruleno] = p->sym->assoc;
- }
+ rules[ruleno].prec = p->sym;
if (p)
p = p->next;
}
the specified symbol's precedence replaces the default. */
if (ruleprec)
{
- rprec[ruleno] = ruleprec->prec;
- rassoc[ruleno] = ruleprec->assoc;
- rprecsym[ruleno] = ruleprec->value;
+ rules[ruleno].precsym = ruleprec;
+ rules[ruleno].prec = ruleprec;
}
-
ritem[itemno++] = -ruleno;
- ruleno++;
+ ++ruleno;
if (p)
p = p->next;
}
- ritem[itemno] = 0;
+ assert (itemno == nritems);
+
+ if (trace_flag)
+ ritem_print (stderr);
}
\f
-/*-------------------------------------------------------------------.
-| Read in the grammar specification and record it in the format |
-| described in gram.h. All guards are copied into the GUARD_OBSTACK |
-| and all actions into ACTION_OBSTACK, in each case forming the body |
-| of a C function (YYGUARD or YYACTION) which contains a switch |
-| statement to decide which guard or action to execute. |
-`-------------------------------------------------------------------*/
+/*------------------------------------------------------------------.
+| Read in the grammar specification and record it in the format |
+| described in gram.h. All actions are copied into ACTION_OBSTACK, |
+| in each case forming the body of a C function (YYACTION) which |
+| contains a switch statement to decide which action to execute. |
+`------------------------------------------------------------------*/
void
reader (void)
{
- start_flag = 0;
- startval = NULL; /* start symbol not specified yet. */
-
- nsyms = 1;
- nvars = 0;
- nrules = 0;
- nitems = 0;
- rline_allocated = 10;
- rline = XCALLOC (short, rline_allocated);
-
- typed = 0;
- lastprec = 0;
-
- semantic_parser = 0;
- pure_parser = 0;
-
- grammar = NULL;
-
lex_init ();
lineno = 1;
obstack_init (&muscle_obstack);
/* Initialize the symbol table. */
- tabinit ();
+ symbols_new ();
+
+ /* Construct the axiom symbol. */
+ axiom = getsym ("$axiom");
+ axiom->class = nterm_sym;
+ axiom->number = nvars++;
/* Construct the error token */
errtoken = getsym ("error");
errtoken->class = token_sym;
- errtoken->user_token_number = 256; /* Value specified by POSIX. */
+ errtoken->number = ntokens++;
/* Construct a token that represents all undefined literal tokens.
It is always token number 2. */
undeftoken = getsym ("$undefined.");
undeftoken->class = token_sym;
- undeftoken->user_token_number = 2;
+ undeftoken->number = ntokens++;
+
+ /* Initialize the obstacks. */
+ obstack_init (&action_obstack);
+ obstack_init (&output_obstack);
+ obstack_init (&pre_prologue_obstack);
+ obstack_init (&post_prologue_obstack);
+
+ finput = xfopen (infile, "r");
/* Read the declaration section. Copy %{ ... %} groups to
TABLE_OBSTACK and FDEFINES file. Also notice any %token, %left,
etc. found there. */
read_declarations ();
+
+ /* If the user did not define her EOFTOKEN, do it now. */
+ if (!eoftoken)
+ {
+ eoftoken = getsym ("$");
+ eoftoken->class = token_sym;
+ eoftoken->number = 0;
+ /* Value specified by POSIX. */
+ eoftoken->user_token_number = 0;
+ }
+
/* Read in the grammar, build grammar in list form. Write out
- guards and actions. */
+ actions. */
readgram ();
/* Some C code is given at the end of the grammar file. */
read_additionnal_code ();
- /* Now we know whether we need the line-number stack. If we do,
- write its type into the .tab.h file.
- This is no longer need with header skeleton. */
+ lex_free ();
+ xfclose (finput);
/* Assign the symbols their symbol numbers. Write #defines for the
token symbols into FDEFINES if requested. */
packsymbols ();
+
/* Convert the grammar into the format described in gram.h. */
packgram ();
+
+ /* The grammar as a symbol_list is no longer needed. */
+ LIST_FREE (symbol_list, grammar);
}
projects / bison.git / blobdiff