X-Git-Url: https://git.saurik.com/bison.git/blobdiff_plain/82b6cb3fb20b5036b201a07c4ea1b2a59b904d01..11b192127cac086338bfaa43f3bab3acda8e3a20:/src/reader.c diff --git a/src/reader.c b/src/reader.c index 0e1fc2bd..f658e64e 100644 --- a/src/reader.c +++ b/src/reader.c @@ -1,1766 +1,502 @@ -/* Input parser for bison - Copyright (C) 1984, 1986, 1989, 1992, 1998, 2000, 2001, 2002 - Free Software Foundation, Inc. +/* Input parser for Bison + + Copyright (C) 1984, 1986, 1989, 1992, 1998, 2000-2003, 2005-2007, + 2009-2012 Free Software Foundation, Inc. This file is part of Bison, the GNU Compiler Compiler. - Bison is free software; you can redistribute it and/or modify + This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by - the Free Software Foundation; either version 2, or (at your option) - any later version. + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. - Bison is distributed in the hope that it will be useful, + This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License - along with Bison; see the file COPYING. If not, write to - the Free Software Foundation, Inc., 59 Temple Place - Suite 330, - Boston, MA 02111-1307, USA. */ - + along with this program. If not, see . */ +#include #include "system.h" -#include "quotearg.h" -#include "quote.h" -#include "getargs.h" + +#include + +#include "complain.h" +#include "conflicts.h" #include "files.h" -#include "symtab.h" -#include "options.h" -#include "lex.h" +#include "getargs.h" #include "gram.h" -#include "complain.h" -#include "output.h" +#include "muscle-tab.h" #include "reader.h" -#include "conflicts.h" -#include "muscle_tab.h" - -typedef struct symbol_list -{ - struct symbol_list *next; - symbol_t *sym; - int line; - - /* The action is attached to the LHS of a rule. */ - const char *action; - int action_line; +#include "symlist.h" +#include "symtab.h" +#include "scan-gram.h" +#include "scan-code.h" - /* The guard is attached to the LHS of a rule. */ - const char *guard; - int guard_line; - symbol_t *ruleprec; -} symbol_list; +static void prepare_percent_define_front_end_variables (void); +static void check_and_convert_grammar (void); -int lineno; 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 = 0; - -/* Incremented for each %left, %right or %nonassoc seen */ -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->guard = NULL; - res->guard_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 = SALIAS; - 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 == SALIAS) - { - 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 SALIASs. */ - if (this->user_token_number == SALIAS) - return TRUE; - } - else /* this->class == token_sym */ - { - assert (this->number != NUMBER_UNDEFINED); - } - - symbols[this->number] = this; - return TRUE; -} - +static bool start_flag = false; +merger_list *merge_functions; +/* Was %union seen? */ +bool union_seen = false; +/* Was a tag seen? */ +bool tag_seen = false; -/*--------------------------------------------------. -| 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 != SALIAS) - { - /* 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; -} +/* Should rules have a default precedence? */ +bool default_prec = true; +/*-----------------------. +| Set the start symbol. | +`-----------------------*/ -/*===================\ -| Low level lexing. | -\===================*/ - -static void -skip_to_char (int target) +void +grammar_start_symbol_set (symbol *sym, location loc) { - int c; - if (target == '\n') - complain (_(" Skipping to next \\n")); + if (start_flag) + complain_at (loc, complaint, _("multiple %s declarations"), "%start"); else - complain (_(" Skipping to next %c"), target); - - do - c = skip_white_space (); - while (c != target && c != EOF); - if (c != EOF) - ungetc (c, finput); -} - - -/*---------------------------------------------------------. -| Read a signed integer from STREAM and return its value. | -`---------------------------------------------------------*/ - -static inline int -read_signed_integer (FILE *stream) -{ - int c = getc (stream); - int sign = 1; - int n = 0; - - if (c == '-') - { - c = getc (stream); - sign = -1; - } - - while (isdigit (c)) { - n = 10 * n + (c - '0'); - c = getc (stream); + start_flag = true; + startsymbol = sym; + startsymbol_location = loc; } - - ungetc (c, stream); - - return sign * n; } - -/*--------------------------------------------------------------. -| Get the data type (alternative in the union) of the value for | -| symbol N in rule RULE. | -`--------------------------------------------------------------*/ - -static char * -get_type_name (int n, symbol_list *rule) -{ - int i; - symbol_list *rp; - - if (n < 0) - { - complain (_("invalid $ value")); - return NULL; - } - - rp = rule; - i = 0; - - while (i < n) - { - rp = rp->next; - if (rp == NULL || rp->sym == NULL) - { - complain (_("invalid $ value")); - return NULL; - } - ++i; - } - return rp->sym->type_name; -} -/*------------------------------------------------------------------. -| 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; +/*------------------------------------------------------------------------. +| Return the merger index for a merging function named NAME. Records the | +| function, if new, in MERGER_LIST. | +`------------------------------------------------------------------------*/ - 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 "). | -`------------------------------------------------------------*/ - -static inline void -copy_string2 (FILE *fin, struct obstack *oout, int match, int store) +static int +get_merge_function (uniqstr name) { - int c; - - if (store) - obstack_1grow (oout, match); + merger_list *syms; + merger_list head; + int n; - c = getc (fin); + if (! glr_parser) + return 0; - while (c != match) + head.next = merge_functions; + for (syms = &head, n = 1; syms->next; syms = syms->next, n += 1) + if (UNIQSTR_EQ (name, syms->next->name)) + break; + if (syms->next == NULL) { - if (c == EOF) - fatal (_("unterminated string at end of file")); - if (c == '\n') - { - complain (_("unterminated string")); - ungetc (c, fin); - c = match; /* invent terminator */ - continue; - } - - copy_character (oout, c); - - if (c == '\\') - { - c = getc (fin); - if (c == EOF) - fatal (_("unterminated string at end of file")); - copy_character (oout, c); - - if (c == '\n') - ++lineno; - } - - c = getc (fin); + syms->next = xmalloc (sizeof syms->next[0]); + syms->next->name = uniqstr_new (name); + /* After all symbol type declarations have been parsed, packgram invokes + record_merge_function_type to set the type. */ + syms->next->type = NULL; + syms->next->next = NULL; + merge_functions = head.next; } - - if (store) - obstack_1grow (oout, c); + return n; } -/* FIXME. */ +/*-------------------------------------------------------------------------. +| For the existing merging function with index MERGER, record the result | +| type as TYPE as required by the lhs of the rule whose %merge declaration | +| is at DECLARATION_LOC. | +`-------------------------------------------------------------------------*/ -static inline void -copy_string (FILE *fin, struct obstack *oout, int match) -{ - copy_string2 (fin, oout, match, 1); -} - -/* FIXME. */ - -static inline void -copy_identifier (FILE *fin, struct obstack *oout) -{ - int c; - - while (isalnum (c = getc (fin)) || c == '_') - obstack_1grow (oout, c); - - ungetc (c, fin); -} - - -/*------------------------------------------------------------------. -| 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_comment (FILE *fin, struct obstack *oout) +static void +record_merge_function_type (int merger, uniqstr type, location declaration_loc) { - int cplus_comment; - int ended; - int c; + int merger_find; + merger_list *merge_function; - /* We read a `/', output it. */ - obstack_1grow (oout, '/'); + if (merger <= 0) + return; - switch ((c = getc (fin))) - { - case '/': - cplus_comment = 1; - break; - case '*': - cplus_comment = 0; - break; - default: - ungetc (c, fin); - return; - } + if (type == NULL) + type = uniqstr_new (""); - obstack_1grow (oout, c); - c = getc (fin); - - ended = 0; - while (!ended) + merger_find = 1; + for (merge_function = merge_functions; + merge_function != NULL && merger_find != merger; + merge_function = merge_function->next) + merger_find += 1; + aver (merge_function != NULL && merger_find == merger); + if (merge_function->type != NULL && !UNIQSTR_EQ (merge_function->type, type)) { - if (!cplus_comment && c == '*') - { - while (c == '*') - { - obstack_1grow (oout, c); - c = getc (fin); - } - - if (c == '/') - { - obstack_1grow (oout, c); - ended = 1; - } - } - else if (c == '\n') - { - ++lineno; - obstack_1grow (oout, c); - if (cplus_comment) - ended = 1; - else - c = getc (fin); - } - else if (c == EOF) - fatal (_("unterminated comment")); - else - { - copy_character (oout, c); - c = getc (fin); - } + unsigned indent = 0; + complain_at_indent (declaration_loc, complaint, &indent, + _("result type clash on merge function %s: " + "<%s> != <%s>"), + quote (merge_function->name), type, + merge_function->type); + indent += SUB_INDENT; + complain_at_indent (merge_function->type_declaration_location, complaint, + &indent, + _("previous declaration")); } + merge_function->type = uniqstr_new (type); + merge_function->type_declaration_location = declaration_loc; } +/*--------------------------------------. +| Free all merge-function definitions. | +`--------------------------------------*/ -/*-------------------------------------------------------------------. -| 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 `$0' with | -| respect to the top of the stack. | -`-------------------------------------------------------------------*/ - -static inline void -copy_at (FILE *fin, struct obstack *oout, int rule_length) +void +free_merger_functions (void) { - int c = getc (fin); - locations_flag = 1; - - if (c == '$') - { - obstack_sgrow (oout, "]b4_lhs_location["); - } - else if (isdigit (c) || c == '-') - { - int n; - - ungetc (c, fin); - n = read_signed_integer (fin); - if (n > rule_length) - complain (_("invalid value: %s%d"), "@", n); - else - obstack_fgrow2 (oout, "]b4_rhs_location([%d], [%d])[", - rule_length, n); - } - else + merger_list *L0 = merge_functions; + while (L0) { - char buf[] = "@c"; - buf[1] = c; - complain (_("%s is invalid"), quote (buf)); + merger_list *L1 = L0->next; + free (L0); + L0 = L1; } } - -/*------------------------------------------------------------------. -| FIN is pointing to a wannabee semantic value (i.e., a `$'). | -| | -| Possible inputs: $[]($|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 `$0' with respect to the top of the stack. | -`------------------------------------------------------------------*/ - -static inline void -copy_dollar (FILE *fin, struct obstack *oout, - symbol_list *rule, int rule_length) -{ - int c = getc (fin); - const char *type_name = NULL; - - /* Get the type name if explicit. */ - if (c == '<') - { - read_type_name (fin); - type_name = token_buffer; - value_components_used = 1; - c = getc (fin); - } - - if (c == '$') - { - if (!type_name) - type_name = get_type_name (0, rule); - 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 == '-') - { - int n; - ungetc (c, fin); - n = read_signed_integer (fin); - - 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 - { - char buf[] = "$c"; - buf[1] = c; - complain (_("%s is invalid"), quote (buf)); - } -} /*-------------------------------------------------------------------. -| Copy the contents of a `%{ ... %}' into the definitions file. The | -| `%{' has already been read. Return after reading the `%}'. | +| 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. | `-------------------------------------------------------------------*/ -static void -copy_definition (struct obstack *oout) -{ - int c; - /* -1 while reading a character if prev char was %. */ - int after_percent; - - if (!no_lines_flag) - { - obstack_fgrow2 (oout, muscle_find ("linef"), - lineno, quotearg_style (c_quoting_style, - muscle_find ("filename"))); - } +/* The (currently) last symbol of GRAMMAR. */ +static symbol_list *grammar_end = NULL; - after_percent = 0; - - c = getc (finput); - - for (;;) - { - switch (c) - { - case '\n': - obstack_1grow (oout, c); - ++lineno; - break; - - case '%': - after_percent = -1; - break; - - case '\'': - case '"': - copy_string (finput, oout, c); - break; - - case '/': - copy_comment (finput, oout); - break; - - case EOF: - fatal ("%s", _("unterminated `%{' definition")); - - default: - copy_character (oout, c); - } - - c = getc (finput); - - if (after_percent) - { - if (c == '}') - return; - obstack_1grow (oout, '%'); - } - after_percent = 0; - } -} - - -/*-------------------------------------------------------------------. -| Parse what comes after %token or %nterm. For %token, WHAT_IS is | -| token_sym and WHAT_IS_NOT is nterm_sym. For %nterm, the arguments | -| are reversed. | -`-------------------------------------------------------------------*/ - -static void -parse_token_decl (symbol_class what_is, symbol_class what_is_not) +/* Append SYM to the grammar. */ +static symbol_list * +grammar_symbol_append (symbol *sym, location loc) { - token_t token = tok_undef; - char *typename = NULL; - - /* The symbol being defined. */ - symbol_t *symbol = NULL; + symbol_list *p = symbol_list_sym_new (sym, loc); - /* After `%token' and `%nterm', any number of symbols maybe be - defined. */ - for (;;) - { - int tmp_char = ungetc (skip_white_space (), finput); - - /* `%' (for instance from `%token', or from `%%' etc.) is the - only valid means to end this declaration. */ - if (tmp_char == '%') - return; - if (tmp_char == EOF) - fatal (_("Premature EOF after %s"), token_buffer); - - token = lex (); - if (token == tok_comma) - { - symbol = NULL; - continue; - } - if (token == tok_typename) - { - typename = xstrdup (token_buffer); - value_components_used = 1; - symbol = NULL; - } - else if (token == tok_identifier && *symval->tag == '\"' && symbol) - { - symbol_make_alias (symbol, typename); - symbol = NULL; - } - else if (token == tok_identifier) - { - int oldclass = symval->class; - symbol = symval; - - if (symbol->class == what_is_not) - complain (_("symbol %s redefined"), symbol->tag); - symbol->class = what_is; - if (what_is == nterm_sym && oldclass != nterm_sym) - symbol->number = nvars++; - if (what_is == token_sym && symbol->number == NUMBER_UNDEFINED) - symbol->number = ntokens++; - - if (typename) - { - if (symbol->type_name == NULL) - symbol->type_name = typename; - else if (strcmp (typename, symbol->type_name) != 0) - complain (_("type redeclaration for %s"), symbol->tag); - } - } - 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"); - skip_to_char ('%'); - } - } - -} - - -/*------------------------------. -| Parse what comes after %start | -`------------------------------*/ - -static void -parse_start_decl (void) -{ - if (start_flag) - complain (_("multiple %s declarations"), "%start"); - if (lex () != tok_identifier) - complain (_("invalid %s declaration"), "%start"); + if (grammar_end) + grammar_end->next = p; else - { - start_flag = 1; - startval = symval; - } -} - -/*-----------------------------------------------------------. -| read in a %type declaration and record its information for | -| get_type_name to access | -`-----------------------------------------------------------*/ - -static void -parse_type_decl (void) -{ - char *name; - - if (lex () != tok_typename) - { - complain ("%s", _("%type declaration has no ")); - skip_to_char ('%'); - return; - } - - name = xstrdup (token_buffer); - - for (;;) - { - token_t t; - int tmp_char = ungetc (skip_white_space (), finput); - - if (tmp_char == '%') - return; - if (tmp_char == EOF) - fatal (_("Premature EOF after %s"), token_buffer); + grammar = p; - t = lex (); + grammar_end = p; - switch (t) - { + /* A null SYM stands for an end of rule; it is not an actual + part of it. */ + if (sym) + ++nritems; - case tok_comma: - case tok_semicolon: - 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: - complain (_("invalid %%type declaration due to item: %s"), - token_buffer); - skip_to_char ('%'); - } - } + return p; } - - -/*----------------------------------------------------------------. -| Read in a %left, %right or %nonassoc declaration and record its | -| information. | -`----------------------------------------------------------------*/ - static void -parse_assoc_decl (associativity assoc) +assign_named_ref (symbol_list *p, named_ref *name) { - char *name = NULL; - int prev = 0; - - /* Assign a new precedence level, never 0. */ - ++lastprec; + symbol *sym = p->content.sym; - for (;;) + if (name->id == sym->tag) { - token_t t; - int tmp_char = ungetc (skip_white_space (), finput); - - if (tmp_char == '%') - return; - if (tmp_char == EOF) - fatal (_("Premature EOF after %s"), token_buffer); - - t = lex (); - - switch (t) - { - case tok_typename: - name = xstrdup (token_buffer); - break; - - case tok_comma: - break; - - case tok_identifier: - if (symval->prec != 0) - complain (_("redefining precedence of %s"), symval->tag); - symval->prec = lastprec; - symval->assoc = assoc; - if (symval->class == nterm_sym) - complain (_("symbol %s redefined"), symval->tag); - 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) - symval->type_name = name; - else if (strcmp (name, symval->type_name) != 0) - complain (_("type redeclaration for %s"), symval->tag); - } - break; - - case tok_number: - if (prev == tok_identifier) - { - symval->user_token_number = numval; - } - else - { - complain - (_("invalid text (%s) - number should be after identifier"), - token_buffer); - skip_to_char ('%'); - } - break; - - case tok_semicolon: - return; - - default: - complain (_("unexpected item: %s"), token_buffer); - skip_to_char ('%'); - } - - prev = t; + complain_at (name->loc, Wother, + _("duplicated symbol name for %s ignored"), + quote (sym->tag)); + named_ref_free (name); } -} - - - -/*--------------------------------------------------------------. -| Copy the union declaration into the stype muscle | -| (and fdefines), where it is made into the definition of | -| YYSTYPE, the type of elements of the parser value stack. | -`--------------------------------------------------------------*/ - -static void -parse_union_decl (void) -{ - int c; - int count = 0; - bool done = FALSE; - struct obstack union_obstack; - if (typed) - complain (_("multiple %s declarations"), "%union"); - - typed = 1; - - MUSCLE_INSERT_INT ("stype_line", lineno); - obstack_init (&union_obstack); - obstack_sgrow (&union_obstack, "union"); - - while (!done) - { - c = xgetc (finput); - - /* If C contains '/', it is output by copy_comment (). */ - if (c != '/') - obstack_1grow (&union_obstack, c); - - switch (c) - { - case '\n': - ++lineno; - break; - - case '/': - copy_comment (finput, &union_obstack); - break; - - case '{': - ++count; - break; - - case '}': - /* FIXME: Errr. How could this happen???. --akim */ - if (count == 0) - complain (_("unmatched %s"), "`}'"); - count--; - if (!count) - done = TRUE; - break; - } - } - - /* 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)); -} - - -/*-------------------------------------------------------. -| Parse the declaration %expect N which says to expect N | -| shift-reduce conflicts. | -`-------------------------------------------------------*/ - -static void -parse_expect_decl (void) -{ - int c = skip_white_space (); - ungetc (c, finput); - - if (!isdigit (c)) - complain (_("argument of %%expect is not an integer")); else - expected_conflicts = read_signed_integer (finput); + p->named_ref = name; } -/*-------------------------------------------------------------------. -| Parse what comes after %thong. the full syntax is | -| | -| %thong token number literal | -| | -| the or number may be omitted. The number specifies the | -| user_token_number. | -| | -| Two symbols are entered in the table, one for the token symbol and | -| one for the literal. Both are given the , 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. | -| | -| During OUTPUT_DEFINES_TABLE, the symbol is reported thereafter, | -| only the literal string is retained it is the literal string that | -| is output to yytname | -`-------------------------------------------------------------------*/ - -static void -parse_thong_decl (void) -{ - token_t token; - symbol_t *symbol; - char *typename = 0; - int usrtoknum = SUNDEF; - - token = lex (); /* fetch typename or first token */ - if (token == tok_typename) - { - typename = xstrdup (token_buffer); - value_components_used = 1; - token = lex (); /* fetch first token */ - } - - /* process first token */ - - if (token != tok_identifier) - { - complain (_("unrecognized item %s, expected an identifier"), - token_buffer); - skip_to_char ('%'); - return; - } - symval->class = token_sym; - symval->type_name = typename; - symval->user_token_number = SALIAS; - symbol = symval; - - token = lex (); /* get number or literal string */ - - if (token == tok_number) - { - usrtoknum = numval; - token = lex (); /* okay, did number, now get literal */ - } - - /* process literal string token */ +/* The rule currently being defined, and the previous rule. + CURRENT_RULE points to the first LHS of the current rule, while + PREVIOUS_RULE_END points to the *end* of the previous rule (NULL). */ +static symbol_list *current_rule = NULL; +static symbol_list *previous_rule_end = NULL; - if (token != tok_identifier || *symval->tag != '\"') - { - complain (_("expected string constant instead of %s"), token_buffer); - skip_to_char ('%'); - return; - } - symval->class = token_sym; - symval->type_name = typename; - symval->user_token_number = usrtoknum; - symval->alias = symbol; - symbol->alias = symval; +/*----------------------------------------------. +| Create a new rule for LHS in to the GRAMMAR. | +`----------------------------------------------*/ - /* symbol and symval combined are only one symbol. */ - nsyms--; -} +void +grammar_current_rule_begin (symbol *lhs, location loc, + named_ref *lhs_name) +{ + symbol_list* p; + /* Start a new rule and record its lhs. */ + ++nrules; + previous_rule_end = grammar_end; -static void -parse_muscle_decl (void) -{ - int ch = ungetc (skip_white_space (), finput); - char *muscle_key; - char *muscle_value; + p = grammar_symbol_append (lhs, loc); + if (lhs_name) + assign_named_ref (p, named_ref_copy (lhs_name)); - /* Read key. */ - if (!isalpha (ch) && ch != '_') - { - complain (_("invalid %s declaration"), "%define"); - skip_to_char ('%'); - return; - } - copy_identifier (finput, &muscle_obstack); - obstack_1grow (&muscle_obstack, 0); - muscle_key = obstack_finish (&muscle_obstack); + current_rule = grammar_end; - /* Read value. */ - ch = skip_white_space (); - if (ch != '"') + /* Mark the rule's lhs as a nonterminal if not already so. */ + if (lhs->class == unknown_sym) { - ungetc (ch, finput); - if (ch != EOF) - { - complain (_("invalid %s declaration"), "%define"); - skip_to_char ('%'); - return; - } - else - fatal (_("Premature EOF after %s"), "\""); + lhs->class = nterm_sym; + lhs->number = nvars; + ++nvars; } - copy_string2 (finput, &muscle_obstack, '"', 0); - obstack_1grow (&muscle_obstack, 0); - muscle_value = obstack_finish (&muscle_obstack); - - /* Store the (key, value) pair in the environment. */ - muscle_insert (muscle_key, muscle_value); + else if (lhs->class == token_sym) + complain_at (loc, complaint, _("rule given for %s, which is a token"), + lhs->tag); } +/*----------------------------------------------------------------------. +| A symbol should be used if either: | +| 1. It has a destructor. | +| 2. The symbol is a mid-rule symbol (i.e., the generated LHS | +| replacing a mid-rule action) that was assigned to or used, as in | +| "exp: { $$ = 1; } { $$ = $1; }". | +`----------------------------------------------------------------------*/ -/*---------------------------------. -| Parse a double quoted parameter. | -`---------------------------------*/ - -static const char * -parse_dquoted_param (const char *from) +static bool +symbol_should_be_used (symbol_list const *s, bool *midrule_warning) { - 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) + if (symbol_code_props_get (s->content.sym, destructor)->code) + return true; + if ((s->midrule && s->midrule->action_props.is_value_used) + || (s->midrule_parent_rule + && symbol_list_n_get (s->midrule_parent_rule, + s->midrule_parent_rhs_index) + ->action_props.is_value_used)) { - complain (_("invalid %s declaration"), from); - if (c != '"') - ungetc (c, finput); - skip_to_char ('%'); - return NULL; + *midrule_warning = true; + return true; } - - return param; -} - -/*----------------------------------. -| Parse what comes after %skeleton. | -`----------------------------------*/ - -static void -parse_skel_decl (void) -{ - skeleton = parse_dquoted_param ("%skeleton"); + return false; } /*----------------------------------------------------------------. -| Read from finput until `%%' is seen. Discard the `%%'. Handle | -| any `%' declarations, and copy the contents of any `%{ ... %}' | -| groups to PRE_PROLOGUE_OBSTACK or POST_PROLOGUE_OBSTACK. | +| Check that the rule R is properly defined. For instance, there | +| should be no type clash on the default action. | `----------------------------------------------------------------*/ static void -read_declarations (void) -{ - for (;;) - { - int c = skip_white_space (); - - if (c == '%') - { - token_t tok = parse_percent_token (); - - switch (tok) - { - case tok_two_percents: - return; - - case tok_percent_left_curly: - if (!typed) - copy_definition (&pre_prologue_obstack); - else - copy_definition (&post_prologue_obstack); - break; - - case tok_token: - parse_token_decl (token_sym, nterm_sym); - break; - - case tok_nterm: - parse_token_decl (nterm_sym, token_sym); - break; - - case tok_type: - parse_type_decl (); - break; - - case tok_start: - parse_start_decl (); - break; - - case tok_union: - parse_union_decl (); - break; - - case tok_expect: - parse_expect_decl (); - break; - - case tok_thong: - parse_thong_decl (); - break; - - case tok_left: - parse_assoc_decl (left_assoc); - break; - - case tok_right: - parse_assoc_decl (right_assoc); - break; - - case tok_nonassoc: - parse_assoc_decl (non_assoc); - break; - - case tok_define: - parse_muscle_decl (); - break; - - case tok_skel: - parse_skel_decl (); - break; - - case tok_noop: - break; - - case tok_stropt: - case tok_intopt: - case tok_obsolete: - assert (0); - break; - - case tok_illegal: - default: - complain (_("unrecognized: %s"), token_buffer); - skip_to_char ('%'); - } - } - else if (c == EOF) - fatal (_("no input grammar")); +grammar_rule_check (const symbol_list *r) +{ + /* Type check. + + If there is an action, then there is nothing we can do: the user + is allowed to shoot herself in the foot. + + Don't worry about the default action if $$ is untyped, since $$'s + value can't be used. */ + if (!r->action_props.code && r->content.sym->type_name) + { + symbol *first_rhs = r->next->content.sym; + /* If $$ is being set in default way, report if any type mismatch. */ + if (first_rhs) + { + char const *lhs_type = r->content.sym->type_name; + const char *rhs_type = + first_rhs->type_name ? first_rhs->type_name : ""; + if (!UNIQSTR_EQ (lhs_type, rhs_type)) + complain_at (r->location, Wother, + _("type clash on default action: <%s> != <%s>"), + lhs_type, rhs_type); + } + /* Warn if there is no default for $$ but we need one. */ else - { - char buf[] = "c"; - buf[0] = c; - complain (_("unknown character: %s"), quote (buf)); - skip_to_char ('%'); - } + complain_at (r->location, Wother, + _("empty rule for typed nonterminal, and no action")); } -} - -/*------------------------------------------------------------------. -| Assuming that a `{' has just been seen, copy everything up to the | -| matching `}' into the actions file. RULE_LENGTH is the number of | -| values in the current rule so far, which says where to find `$0' | -| with respect to the top of the stack. | -| | -| This routine is used both for actions and guards. Only | -| ACTION_OBSTACK is used, but this is fine, since we use only | -| pointers to relevant portions inside this obstack. | -`------------------------------------------------------------------*/ -static void -parse_braces (symbol_list *rule, int rule_length) -{ - int c; - int count; - - count = 1; - while (count > 0) - { - while ((c = getc (finput)) != '}') - switch (c) - { - case '\n': - obstack_1grow (&action_obstack, c); - ++lineno; - break; - - case '{': - obstack_1grow (&action_obstack, c); - ++count; - break; - - case '\'': - case '"': - copy_string (finput, &action_obstack, c); - break; - - case '/': - copy_comment (finput, &action_obstack); - break; - - case '$': - copy_dollar (finput, &action_obstack, rule, rule_length); - break; - - case '@': - copy_at (finput, &action_obstack, rule_length); - break; - - case EOF: - fatal (_("unmatched %s"), "`{'"); - - default: - obstack_1grow (&action_obstack, c); - } - - /* Above loop exits when C is '}'. */ - if (--count) - obstack_1grow (&action_obstack, c); - } + /* Check that symbol values that should be used are in fact used. */ + { + symbol_list const *l = r; + int n = 0; + for (; l && l->content.sym; l = l->next, ++n) + { + bool midrule_warning = false; + if (!l->action_props.is_value_used + && symbol_should_be_used (l, &midrule_warning) + /* The default action, $$ = $1, `uses' both. */ + && (r->action_props.code || (n != 0 && n != 1))) + { + warnings warn_flag = midrule_warning ? Wmidrule_values : Wother; + if (n) + complain_at (r->location, warn_flag, _("unused value: $%d"), n); + else + complain_at (r->location, warn_flag, _("unset value: $$")); + } + } + } - obstack_1grow (&action_obstack, '\0'); + /* See comments in grammar_current_rule_prec_set for how POSIX + mandates this complaint. It's only for identifiers, so skip + it for char literals and strings, which are always tokens. */ + if (r->ruleprec + && r->ruleprec->tag[0] != '\'' && r->ruleprec->tag[0] != '"' + && r->ruleprec->status != declared && !r->ruleprec->prec) + complain_at (r->location, Wother, + _("token for %%prec is not defined: %s"), r->ruleprec->tag); } -static void -parse_action (symbol_list *rule, int rule_length) -{ - rule->action_line = lineno; - parse_braces (rule, rule_length); - rule->action = obstack_finish (&action_obstack); -} +/*-------------------------------------. +| End the currently being grown rule. | +`-------------------------------------*/ - -static void -parse_guard (symbol_list *rule, int rule_length) +void +grammar_current_rule_end (location loc) { - token_t t = lex (); - if (t != tok_left_curly) - complain (_("invalid %s declaration"), "%guard"); - rule->guard_line = lineno; - parse_braces (rule, rule_length); - rule->guard = obstack_finish (&action_obstack); + /* Put an empty link in the list to mark the end of this rule */ + grammar_symbol_append (NULL, grammar_end->location); + current_rule->location = loc; } - - -/*-------------------------------------------------------------------. -| Generate a dummy symbol, a nonterminal, whose name cannot conflict | -| with the user's names. | -`-------------------------------------------------------------------*/ -static symbol_t * -gensym (void) -{ - /* Incremented for each generated symbol */ - static int gensym_count = 0; - static char buf[256]; - - symbol_t *sym; - - sprintf (buf, "@%d", ++gensym_count); - token_buffer = buf; - sym = getsym (token_buffer); - sym->class = nterm_sym; - sym->number = nvars++; - return sym; -} - /*-------------------------------------------------------------------. -| 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. | -| | -| 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. | +| The previous action turns out the be a mid-rule action. Attach it | +| to the current rule, i.e., create a dummy symbol, attach it this | +| mid-rule action, and append this dummy nonterminal to the current | +| rule. | `-------------------------------------------------------------------*/ -static void -readgram (void) -{ - token_t t; - 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 = NULL; - /* Points to the symbol_list preceding crule. */ - 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 */ - symbol_t *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; - ++nritems; - - p = symbol_list_new (lhs); - - crule1 = p1; - if (p1) - p1->next = p; - else - grammar = p; - - p1 = p; - crule = p; - - /* 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); - - /* 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; - - /* 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) - { - /* 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 (); - crule->ruleprec = symval; - t = lex (); - } - - if (t == tok_guard) - { - if (!semantic_parser) - complain (_("%%guard present but %%semantic_parser not specified")); - - parse_guard (crule, rulelength); - 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 (nrules == 0) - fatal (_("no rules in the input grammar")); +void +grammar_midrule_action (void) +{ + /* 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, whose location is that of the midrule + action. Create the MIDRULE. */ + location dummy_location = current_rule->action_props.location; + symbol *dummy = dummy_symbol_get (dummy_location); + symbol_list *midrule = symbol_list_sym_new (dummy, dummy_location); + + /* Remember named_ref of previous action. */ + named_ref *action_name = current_rule->action_props.named_ref; + + /* Make a new rule, whose body is empty, before the current one, so + that the action just read can belong to it. */ + ++nrules; + ++nritems; + /* Attach its location and actions to that of the DUMMY. */ + midrule->location = dummy_location; + code_props_rule_action_init (&midrule->action_props, + current_rule->action_props.code, + current_rule->action_props.location, + midrule, 0, + current_rule->action_props.is_predicate); + code_props_none_init (¤t_rule->action_props); + + if (previous_rule_end) + previous_rule_end->next = midrule; + else + grammar = midrule; - /* Report any undefined symbols and consider them nonterminals. */ - symbols_do (symbol_check_defined, NULL); + /* End the dummy's rule. */ + midrule->next = symbol_list_sym_new (NULL, dummy_location); + midrule->next->next = current_rule; - /* Insert the initial rule, which line is that of the first rule - (not that of the start symbol): + previous_rule_end = midrule->next; - 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); - - assert (nsyms == ntokens + nvars); + /* Insert the dummy nonterminal replacing the midrule action into + the current rule. Bind it to its dedicated rule. */ + grammar_current_rule_symbol_append (dummy, dummy_location, + action_name); + grammar_end->midrule = midrule; + midrule->midrule_parent_rule = current_rule; + midrule->midrule_parent_rhs_index = symbol_list_length (current_rule->next); } -/* At the end of the grammar file, some C source code must - be stored. It is going to be associated to the epilogue - directive. */ -static void -read_additionnal_code (void) -{ - 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"))); - } +/* Set the precedence symbol of the current rule to PRECSYM. */ - while ((c = getc (finput)) != EOF) - copy_character (&el_obstack, c); +void +grammar_current_rule_prec_set (symbol *precsym, location loc) +{ + /* POSIX says that any identifier is a nonterminal if it does not + appear on the LHS of a grammar rule and is not defined by %token + or by one of the directives that assigns precedence to a token. We + ignore this here because the only kind of identifier that POSIX + allows to follow a %prec is a token and because assuming it's a + token now can produce more logical error messages. Nevertheless, + grammar_rule_check does obey what we believe is the real intent of + POSIX here: that an error be reported for any identifier that + appears after %prec but that is not defined separately as a + token. */ + symbol_class_set (precsym, token_sym, loc, false); + if (current_rule->ruleprec) + complain_at (loc, complaint, _("only one %s allowed per rule"), "%prec"); + current_rule->ruleprec = precsym; +} + +/* Attach dynamic precedence DPREC to the current rule. */ - obstack_1grow (&el_obstack, 0); - muscle_insert ("epilogue", obstack_finish (&el_obstack)); +void +grammar_current_rule_dprec_set (int dprec, location loc) +{ + if (! glr_parser) + complain_at (loc, Wother, _("%s affects only GLR parsers"), + "%dprec"); + if (dprec <= 0) + complain_at (loc, complaint, _("%s must be followed by positive number"), + "%dprec"); + else if (current_rule->dprec != 0) + complain_at (loc, complaint, _("only one %s allowed per rule"), "%dprec"); + current_rule->dprec = dprec; } - -/*------------------------------------------------------------------. -| Set TOKEN_TRANSLATIONS. Check that no two symbols share the same | -| number. | -`------------------------------------------------------------------*/ +/* Attach a merge function NAME with argument type TYPE to current + rule. */ -static void -token_translations_init (void) +void +grammar_current_rule_merge_set (uniqstr name, location loc) { - int num_256_available_p = TRUE; - int i; - - /* 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_t *this = symbols[i]; - if (this->user_token_number != SUNDEF) - { - 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; - } - } - - /* If 256 is not used, assign it to error, to follow POSIX. */ - if (num_256_available_p && errtoken->user_token_number == SUNDEF) - errtoken->user_token_number = 256; - - /* Set the missing user numbers. */ - if (max_user_token_number < 256) - max_user_token_number = 256; - - for (i = 0; i < ntokens; ++i) - { - symbol_t *this = symbols[i]; - if (this->user_token_number == SUNDEF) - 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 (token_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 + 1; i++) - token_translations[i] = undeftoken->number; - symbols_do (symbol_translation, NULL); + if (! glr_parser) + complain_at (loc, Wother, _("%s affects only GLR parsers"), + "%merge"); + if (current_rule->merger != 0) + complain_at (loc, complaint, _("only one %s allowed per rule"), "%merge"); + current_rule->merger = get_merge_function (name); + current_rule->merger_declaration_location = loc; } +/* Attach SYM to the current rule. If needed, move the previous + action as a mid-rule action. */ -/*----------------------------------------------------------------. -| Assign symbol numbers, and write definition of token names into | -| FDEFINES. Set up vectors SYMBOL_TABLE, TAGS of symbols. | -`----------------------------------------------------------------*/ - -static void -packsymbols (void) +void +grammar_current_rule_symbol_append (symbol *sym, location loc, + named_ref *name) { - symbols = XCALLOC (symbol_t *, nsyms); - - symbols_do (symbol_check_alias_consistence, NULL); - symbols_do (symbol_pack, NULL); - - token_translations_init (); + symbol_list *p; + if (current_rule->action_props.code) + grammar_midrule_action (); + p = grammar_symbol_append (sym, loc); + if (name) + assign_named_ref(p, name); + if (sym->status == undeclared || sym->status == used) + sym->status = needed; +} - 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); +/* Attach an ACTION to the current rule. */ - start_symbol = startval->number; +void +grammar_current_rule_action_append (const char *action, location loc, + named_ref *name, bool is_predicate) +{ + if (current_rule->action_props.code) + grammar_midrule_action (); + /* After all symbol declarations have been parsed, packgram invokes + code_props_translate_code. */ + code_props_rule_action_init (¤t_rule->action_props, action, loc, + current_rule, name, is_predicate); } - + /*---------------------------------------------------------------. | Convert the rules into the representation using RRHS, RLHS and | | RITEM. | @@ -1769,150 +505,253 @@ packsymbols (void) static void packgram (void) { - int itemno; - int ruleno; - symbol_list *p; + unsigned int itemno = 0; + rule_number ruleno = 0; + symbol_list *p = grammar; + + ritem = xnmalloc (nritems + 1, sizeof *ritem); - ritem = XCALLOC (item_number_t, nritems + 1); - rules = XCALLOC (rule_t, nrules) - 1; + /* This sentinel is used by build_relations in gram.c. */ + *ritem++ = 0; - itemno = 0; - ruleno = 1; + rules = xnmalloc (nrules, sizeof *rules); - p = grammar; while (p) { - symbol_t *ruleprec = p->ruleprec; + int rule_length = 0; + symbol *ruleprec = p->ruleprec; + record_merge_function_type (p->merger, p->content.sym->type_name, + p->merger_declaration_location); rules[ruleno].user_number = ruleno; rules[ruleno].number = ruleno; - rules[ruleno].lhs = p->sym; + rules[ruleno].lhs = p->content.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; - rules[ruleno].guard = p->guard; - rules[ruleno].guard_line = p->guard_line; - - p = p->next; - while (p && p->sym) - { - /* item_number_t = token_number_t. - But the former needs to contain more: negative rule numbers. */ - ritem[itemno++] = token_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) - rules[ruleno].prec = p->sym; - if (p) - p = p->next; - } + rules[ruleno].prec = NULL; + rules[ruleno].dprec = p->dprec; + rules[ruleno].merger = p->merger; + rules[ruleno].precsym = NULL; + rules[ruleno].location = p->location; + rules[ruleno].useful = true; + rules[ruleno].action = p->action_props.code; + rules[ruleno].action_location = p->action_props.location; + rules[ruleno].is_predicate = p->action_props.is_predicate; + + /* If the midrule's $$ is set or its $n is used, remove the `$' from the + symbol name so that it's a user-defined symbol so that the default + %destructor and %printer apply. */ + if (p->midrule_parent_rule + && (p->action_props.is_value_used + || symbol_list_n_get (p->midrule_parent_rule, + p->midrule_parent_rhs_index) + ->action_props.is_value_used)) + p->content.sym->tag += 1; + + /* Don't check the generated rule 0. It has no action, so some rhs + symbols may appear unused, but the parsing algorithm ensures that + %destructor's are invoked appropriately. */ + if (p != grammar) + grammar_rule_check (p); + + for (p = p->next; p && p->content.sym; p = p->next) + { + ++rule_length; + + /* Don't allow rule_length == INT_MAX, since that might + cause confusion with strtol if INT_MAX == LONG_MAX. */ + if (rule_length == INT_MAX) + complain_at (rules[ruleno].location, fatal, _("rule is too long")); + + /* item_number = symbol_number. + But the former needs to contain more: negative rule numbers. */ + ritem[itemno++] = + symbol_number_as_item_number (p->content.sym->number); + /* A rule gets by default the precedence and associativity + of its last token. */ + if (p->content.sym->class == token_sym && default_prec) + rules[ruleno].prec = p->content.sym; + } /* If this rule has a %prec, the specified symbol's precedence replaces the default. */ if (ruleprec) - { - rules[ruleno].precsym = ruleprec; - rules[ruleno].prec = ruleprec; - } - ritem[itemno++] = -ruleno; + { + rules[ruleno].precsym = ruleprec; + rules[ruleno].prec = ruleprec; + } + /* An item ends by the rule number (negated). */ + ritem[itemno++] = rule_number_as_item_number (ruleno); + aver (itemno < ITEM_NUMBER_MAX); ++ruleno; + aver (ruleno < RULE_NUMBER_MAX); if (p) - p = p->next; + p = p->next; } - ritem[itemno] = 0; - assert (itemno == nritems); + aver (itemno == nritems); - if (trace_flag) + if (trace_flag & trace_sets) ritem_print (stderr); } -/*-------------------------------------------------------------------. -| 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) { - lex_init (); - lineno = 1; - - /* Initialize the muscle obstack. */ - obstack_init (&muscle_obstack); - /* Initialize the symbol table. */ symbols_new (); - /* Construct the axiom symbol. */ - axiom = getsym ("$axiom"); - axiom->class = nterm_sym; - axiom->number = nvars++; + /* Construct the accept symbol. */ + accept = symbol_get ("$accept", empty_location); + accept->class = nterm_sym; + accept->number = nvars++; /* Construct the error token */ - errtoken = getsym ("error"); + errtoken = symbol_get ("error", empty_location); errtoken->class = token_sym; errtoken->number = ntokens++; /* Construct a token that represents all undefined literal tokens. It is always token number 2. */ - undeftoken = getsym ("$undefined."); + undeftoken = symbol_get ("$undefined", empty_location); undeftoken->class = token_sym; undeftoken->number = ntokens++; - /* Initialize the obstacks. */ - obstack_init (&action_obstack); - obstack_init (&output_obstack); - obstack_init (&pre_prologue_obstack); - obstack_init (&post_prologue_obstack); + gram_in = xfopen (grammar_file, "r"); + + gram__flex_debug = trace_flag & trace_scan; + gram_debug = trace_flag & trace_parse; + gram_scanner_initialize (); + gram_parse (); + prepare_percent_define_front_end_variables (); - finput = xfopen (infile, "r"); + if (! complaint_issued) + check_and_convert_grammar (); - /* Read the declaration section. Copy %{ ... %} groups to - TABLE_OBSTACK and FDEFINES file. Also notice any %token, %left, - etc. found there. */ - read_declarations (); + xfclose (gram_in); +} + +static void +prepare_percent_define_front_end_variables (void) +{ + /* Set %define front-end variable defaults. */ + muscle_percent_define_default ("lr.keep-unreachable-states", "false"); + { + char *lr_type; + /* IELR would be a better default, but LALR is historically the + default. */ + muscle_percent_define_default ("lr.type", "lalr"); + lr_type = muscle_percent_define_get ("lr.type"); + if (STRNEQ (lr_type, "canonical-lr")) + muscle_percent_define_default ("lr.default-reductions", "most"); + else + muscle_percent_define_default ("lr.default-reductions", "accepting"); + free (lr_type); + } + + /* Check %define front-end variables. */ + { + static char const * const values[] = { + "lr.type", "lalr", "ielr", "canonical-lr", NULL, + "lr.default-reductions", "most", "consistent", "accepting", NULL, + NULL + }; + muscle_percent_define_check_values (values); + } +} - /* If the user did not define her EOFTOKEN, do it now. */ - if (!eoftoken) + +/*-------------------------------------------------------------. +| Check the grammar that has just been read, and convert it to | +| internal form. | +`-------------------------------------------------------------*/ + +static void +check_and_convert_grammar (void) +{ + /* Grammar has been read. Do some checking. */ + if (nrules == 0) + complain (fatal, _("no rules in the input grammar")); + + /* If the user did not define her ENDTOKEN, do it now. */ + if (!endtoken) { - eoftoken = getsym ("$"); - eoftoken->class = token_sym; - eoftoken->number = 0; + endtoken = symbol_get ("$end", empty_location); + endtoken->class = token_sym; + endtoken->number = 0; /* Value specified by POSIX. */ - eoftoken->user_token_number = 0; + endtoken->user_token_number = 0; } - /* Read in the grammar, build grammar in list form. Write out - guards and actions. */ - readgram (); - /* Some C code is given at the end of the grammar file. */ - read_additionnal_code (); + /* Report any undefined symbols and consider them nonterminals. */ + symbols_check_defined (); + + /* Find the start symbol if no %start. */ + if (!start_flag) + { + symbol_list *node; + for (node = grammar; + node != NULL && symbol_is_dummy (node->content.sym); + node = node->next) + { + for (node = node->next; + node != NULL && node->content.sym != NULL; + node = node->next) + ; + } + aver (node != NULL); + grammar_start_symbol_set (node->content.sym, + node->content.sym->location); + } + + /* Insert the initial rule, whose line is that of the first rule + (not that of the start symbol): - lex_free (); - xfclose (finput); + accept: %start EOF. */ + { + symbol_list *p = symbol_list_sym_new (accept, empty_location); + p->location = grammar->location; + p->next = symbol_list_sym_new (startsymbol, empty_location); + p->next->next = symbol_list_sym_new (endtoken, empty_location); + p->next->next->next = symbol_list_sym_new (NULL, empty_location); + p->next->next->next->next = grammar; + nrules += 1; + nritems += 3; + grammar = p; + } + + aver (nsyms <= SYMBOL_NUMBER_MAXIMUM && nsyms == ntokens + nvars); /* Assign the symbols their symbol numbers. Write #defines for the token symbols into FDEFINES if requested. */ - packsymbols (); + symbols_pack (); + + /* Scan rule actions after invoking symbol_check_alias_consistency (in + symbols_pack above) so that token types are set correctly before the rule + action type checking. + + Before invoking grammar_rule_check (in packgram below) on any rule, make + sure all actions have already been scanned in order to set `used' flags. + Otherwise, checking that a midrule's $$ should be set will not always work + properly because the check must forward-reference the midrule's parent + rule. For the same reason, all the `used' flags must be set before + checking whether to remove `$' from any midrule symbol name (also in + packgram). */ + { + symbol_list *sym; + for (sym = grammar; sym; sym = sym->next) + code_props_translate_code (&sym->action_props); + } /* 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); -} - -void -grammar_free (void) -{ - XFREE (ritem); - free (rules + 1); - /* Free the symbol table data structure. */ - symbols_free (); + symbol_list_free (grammar); }