X-Git-Url: https://git.saurik.com/bison.git/blobdiff_plain/6b7e85b9940d4313ab2f1663a449fdcb8b61df4b..595284e9d0277517bac8f3fc7bcdd529d97aa6f9:/src/reader.c diff --git a/src/reader.c b/src/reader.c index bbdb40ae..f658e64e 100644 --- a/src/reader.c +++ b/src/reader.c @@ -1,2021 +1,757 @@ -/* Input parser for bison - Copyright 1984, 1986, 1989, 1992, 1998, 2000, 2001 - 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 "obstack.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" - -/* 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; +#include "symlist.h" +#include "symtab.h" +#include "scan-gram.h" +#include "scan-code.h" -int lineno; -char **tags; -short *user_toknums; -static symbol_list *grammar; -static int start_flag; -static bucket *startval; +static void prepare_percent_define_front_end_variables (void); +static void check_and_convert_grammar (void); -/* Nonzero if components of semantic values are used, implying - they must be unions. */ -static int value_components_used; +static symbol_list *grammar = NULL; +static bool start_flag = false; +merger_list *merge_functions; -/* Nonzero if %union has been seen. */ -static int typed; +/* Was %union seen? */ +bool union_seen = false; -/* Incremented for each %left, %right or %nonassoc seen */ -static int lastprec; +/* Was a tag seen? */ +bool tag_seen = false; -static bucket *errtoken; -static bucket *undeftoken; +/* 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; -} -/*------------------------------------------------------------. -| 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) -{ - int c; - - if (store) - obstack_1grow (oout, match); - - c = getc (fin); - - while (c != match) - { - if (c == EOF) - fatal (_("unterminated string at end of file")); - if (c == '\n') - { - complain (_("unterminated string")); - ungetc (c, fin); - c = match; /* invent terminator */ - continue; - } - - obstack_1grow (oout, c); - - if (c == '\\') - { - c = getc (fin); - if (c == EOF) - fatal (_("unterminated string at end of file")); - obstack_1grow (oout, c); - - if (c == '\n') - lineno++; - } - - c = getc (fin); - } +/*------------------------------------------------------------------------. +| Return the merger index for a merging function named NAME. Records the | +| function, if new, in MERGER_LIST. | +`------------------------------------------------------------------------*/ - if (store) - obstack_1grow (oout, c); -} - -/* FIXME. */ - -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 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. | -`-----------------------------------------------------------------*/ - -static inline void -copy_comment2 (FILE *fin, struct obstack *oout1, struct obstack *oout2) +static int +get_merge_function (uniqstr name) { - int cplus_comment; - int ended; - int c; + merger_list *syms; + merger_list head; + int n; - /* We read a `/', output it. */ - obstack_1grow (oout1, '/'); - if (oout2) - obstack_1grow (oout2, '/'); + if (! glr_parser) + return 0; - switch ((c = getc (fin))) - { - case '/': - cplus_comment = 1; + head.next = merge_functions; + for (syms = &head, n = 1; syms->next; syms = syms->next, n += 1) + if (UNIQSTR_EQ (name, syms->next->name)) break; - case '*': - cplus_comment = 0; - break; - default: - ungetc (c, fin); - return; - } - - obstack_1grow (oout1, c); - if (oout2) - obstack_1grow (oout2, c); - c = getc (fin); - - ended = 0; - while (!ended) + if (syms->next == NULL) { - if (!cplus_comment && c == '*') - { - while (c == '*') - { - obstack_1grow (oout1, c); - if (oout2) - obstack_1grow (oout2, c); - c = getc (fin); - } - - if (c == '/') - { - obstack_1grow (oout1, c); - if (oout2) - obstack_1grow (oout2, c); - ended = 1; - } - } - else if (c == '\n') - { - lineno++; - obstack_1grow (oout1, c); - if (oout2) - obstack_1grow (oout2, c); - if (cplus_comment) - ended = 1; - else - c = getc (fin); - } - else if (c == EOF) - fatal (_("unterminated comment")); - else - { - obstack_1grow (oout1, c); - if (oout2) - obstack_1grow (oout2, c); - 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; } + return n; } +/*-------------------------------------------------------------------------. +| 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. | +`-------------------------------------------------------------------------*/ -/*-------------------------------------------------------------------. -| Dump the comment (actually the current string starting with a `/') | -| from FIN to OOUT. | -`-------------------------------------------------------------------*/ - -static inline void -copy_comment (FILE *fin, struct obstack *oout) +static void +record_merge_function_type (int merger, uniqstr type, location declaration_loc) { - copy_comment2 (fin, oout, NULL); -} - + int merger_find; + merger_list *merge_function; -/*-----------------------------------------------------------------. -| 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 `$0' with | -| respect to the top of the stack. | -`-----------------------------------------------------------------*/ + if (merger <= 0) + return; -static inline void -copy_at (FILE *fin, struct obstack *oout, int stack_offset) -{ - int c; + if (type == NULL) + type = uniqstr_new (""); - c = getc (fin); - if (c == '$') + 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)) { - obstack_sgrow (oout, "yyloc"); - locations_flag = 1; + 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")); } - else if (isdigit (c) || c == '-') - { - int n; + merge_function->type = uniqstr_new (type); + merge_function->type_declaration_location = declaration_loc; +} - ungetc (c, fin); - n = read_signed_integer (fin); +/*--------------------------------------. +| Free all merge-function definitions. | +`--------------------------------------*/ - obstack_fgrow1 (oout, "yylsp[%d]", n - stack_offset); - locations_flag = 1; - } - else +void +free_merger_functions (void) +{ + 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) | +| 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. | | | -| 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 `$0' with respect to the top of the stack. | +| All actions are copied out, labelled by the rule number they apply | +| to. | `-------------------------------------------------------------------*/ -static inline void -copy_dollar (FILE *fin, struct obstack *oout, - symbol_list *rule, int stack_offset) -{ - int c = getc (fin); - const char *type_name = NULL; +/* The (currently) last symbol of GRAMMAR. */ +static symbol_list *grammar_end = NULL; - /* Get the type name if explicit. */ - if (c == '<') - { - read_type_name (fin); - type_name = token_buffer; - value_components_used = 1; - c = getc (fin); - } +/* Append SYM to the grammar. */ +static symbol_list * +grammar_symbol_append (symbol *sym, location loc) +{ + symbol_list *p = symbol_list_sym_new (sym, loc); - if (c == '$') - { - 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); - } - else if (isdigit (c) || c == '-') - { - int n; - ungetc (c, fin); - n = read_signed_integer (fin); + if (grammar_end) + grammar_end->next = p; + else + grammar = p; - if (!type_name && n > 0) - type_name = get_type_name (n, rule); + grammar_end = p; - obstack_fgrow1 (oout, "yyvsp[%d]", n - stack_offset); + /* A null SYM stands for an end of rule; it is not an actual + part of it. */ + if (sym) + ++nritems; - 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); - } - else - { - char buf[] = "$c"; - buf[1] = c; - complain (_("%s is invalid"), quote (buf)); - } + return p; } - -/*-------------------------------------------------------------------. -| Copy the contents of a `%{ ... %}' into the definitions file. The | -| `%{' has already been read. Return after reading the `%}'. | -`-------------------------------------------------------------------*/ static void -copy_definition (void) +assign_named_ref (symbol_list *p, named_ref *name) { - int c; - /* -1 while reading a character if prev char was %. */ - int after_percent; + symbol *sym = p->content.sym; -#if 0 - if (!no_lines_flag) + if (name->id == sym->tag) { - obstack_fgrow2 (&attrs_obstack, muscle_find ("linef"), - lineno, quotearg_style (c_quoting_style, - muscle_find("filename"))); - } -#endif - - after_percent = 0; - - c = getc (finput); - - for (;;) - { - switch (c) - { - case '\n': - obstack_1grow (&attrs_obstack, c); - lineno++; - break; - - case '%': - after_percent = -1; - break; - - case '\'': - case '"': - copy_string (finput, &attrs_obstack, c); - break; - - case '/': - copy_comment (finput, &attrs_obstack); - break; - - case EOF: - fatal ("%s", _("unterminated `%{' definition")); - - default: - obstack_1grow (&attrs_obstack, c); - } - - c = getc (finput); - - if (after_percent) - { - if (c == '}') - return; - obstack_1grow (&attrs_obstack, '%'); - } - after_percent = 0; + complain_at (name->loc, Wother, + _("duplicated symbol name for %s ignored"), + quote (sym->tag)); + named_ref_free (name); } + else + p->named_ref = name; } -/*-------------------------------------------------------------------. -| 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. | -`-------------------------------------------------------------------*/ +/* 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; -static void -parse_token_decl (symbol_class what_is, symbol_class what_is_not) -{ - token_t token = tok_undef; - char *typename = NULL; - /* The symbol being defined. */ - struct bucket *symbol = NULL; +/*----------------------------------------------. +| Create a new rule for LHS in to the GRAMMAR. | +`----------------------------------------------*/ - /* 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) - { - 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 = 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->value = nvars++; - - 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; - } - else - { - complain (_("`%s' is invalid in %s"), - token_buffer, (what_is == token_sym) ? "%token" : "%nterm"); - skip_to_char ('%'); - } - } +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; + p = grammar_symbol_append (lhs, loc); + if (lhs_name) + assign_named_ref (p, named_ref_copy (lhs_name)); -/*------------------------------. -| Parse what comes after %start | -`------------------------------*/ + current_rule = grammar_end; -static void -parse_start_decl (void) -{ - if (start_flag) - complain (_("multiple %s declarations"), "%start"); - if (lex () != tok_identifier) - complain (_("invalid %s declaration"), "%start"); - else + /* Mark the rule's lhs as a nonterminal if not already so. */ + if (lhs->class == unknown_sym) { - start_flag = 1; - startval = symval; + lhs->class = nterm_sym; + lhs->number = nvars; + ++nvars; } + else if (lhs->class == token_sym) + complain_at (loc, complaint, _("rule given for %s, which is a token"), + lhs->tag); } -/*-----------------------------------------------------------. -| read in a %type declaration and record its information for | -| get_type_name to access | -`-----------------------------------------------------------*/ -static void -parse_type_decl (void) -{ - char *name; +/*----------------------------------------------------------------------. +| 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; }". | +`----------------------------------------------------------------------*/ - if (lex () != tok_typename) - { - complain ("%s", _("%type declaration has no ")); - skip_to_char ('%'); - return; - } - - name = xstrdup (token_buffer); - - for (;;) +static bool +symbol_should_be_used (symbol_list const *s, bool *midrule_warning) +{ + 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)) { - 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_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 ('%'); - } + *midrule_warning = true; + return true; } + return false; } - - /*----------------------------------------------------------------. -| Read in a %left, %right or %nonassoc declaration and record its | -| information. | +| Check that the rule R is properly defined. For instance, there | +| should be no type clash on the default action. | `----------------------------------------------------------------*/ static void -parse_assoc_decl (associativity assoc) -{ - char *name = NULL; - int prev = 0; - - lastprec++; /* Assign a new precedence level, never 0. */ - - 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); - - 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); - 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; +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 $$ but we need one. */ + else + complain_at (r->location, Wother, + _("empty rule for typed nonterminal, and no action")); } -} - + /* Check that symbol values that should be used are in fact used. */ + { + symbol_list const *l = r; + int n = 0; + for (; l && l->content.sym; l = l->next, ++n) + { + bool midrule_warning = false; + if (!l->action_props.is_value_used + && symbol_should_be_used (l, &midrule_warning) + /* The default action, $$ = $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: $$")); + } + } + } -/*--------------------------------------------------------------. -| 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; - 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'); - - obstack_init (&union_obstack); - obstack_sgrow (&union_obstack, "union"); - if (defines_flag) - obstack_sgrow (&defines_obstack, "typedef union"); - - c = getc (finput); - - while (c != EOF) - { - /* If C contains '/', it is output by copy_comment (). */ - if (c != '/') - { - obstack_1grow (&union_obstack, c); - if (defines_flag) - obstack_1grow (&defines_obstack, c); - } - - switch (c) - { - case '\n': - lineno++; - break; - - case '/': - copy_comment2 (finput, &defines_obstack, &union_obstack); - break; - - case '{': - count++; - break; - - case '}': - 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; - } - } - - c = getc (finput); - } - + /* 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); } -/*-------------------------------------------------------. -| Parse the declaration %expect N which says to expect N | -| shift-reduce conflicts. | -`-------------------------------------------------------*/ +/*-------------------------------------. +| End the currently being grown rule. | +`-------------------------------------*/ -static void -parse_expect_decl (void) +void +grammar_current_rule_end (location loc) { - 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); + /* 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; } /*-------------------------------------------------------------------. -| 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 | +| 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 -parse_thong_decl (void) -{ - token_t token; - struct bucket *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 */ +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; - 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; + /* End the dummy's rule. */ + midrule->next = symbol_list_sym_new (NULL, dummy_location); + midrule->next->next = current_rule; - symval->alias = symbol; - symbol->alias = symval; + previous_rule_end = midrule->next; - /* symbol and symval combined are only one symbol. */ - nsyms--; + /* 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); } -static void -parse_muscle_decl (void) -{ - int ch = ungetc (skip_white_space (), finput); - char* muscle_key; - char* muscle_value; - - /* 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); - - /* Read value. */ - ch = skip_white_space (); - if (ch != '"') - { - ungetc (ch, finput); - if (ch != EOF) - { - complain (_("invalid %s declaration"), "%define"); - skip_to_char ('%'); - return; - } - else - fatal (_("Premature EOF after %s"), "\""); - } - copy_string2 (finput, &muscle_obstack, '"', 0); - obstack_1grow (&muscle_obstack, 0); - muscle_value = obstack_finish (&muscle_obstack); +/* Set the precedence symbol of the current rule to PRECSYM. */ - /* Store the (key, value) pair in the environment. */ - muscle_insert (muscle_key, muscle_value); -} - - -/*----------------------------------. -| Parse what comes after %skeleton. | -`----------------------------------*/ +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. */ void -parse_skel_decl (void) +grammar_current_rule_dprec_set (int dprec, location loc) { - /* Complete with parse_dquoted_param () on the CVS branch 1.29. */ + 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; } -/*----------------------------------------------------------------. -| Read from finput until `%%' is seen. Discard the `%%'. Handle | -| any `%' declarations, and copy the contents of any `%{ ... %}' | -| groups to ATTRS_OBSTACK. | -`----------------------------------------------------------------*/ - -static void -read_declarations (void) -{ - int c; - int tok; - - for (;;) - { - c = skip_white_space (); - - if (c == '%') - { - tok = parse_percent_token (); - - switch (tok) - { - case tok_two_percents: - return; - - case tok_percent_left_curly: - copy_definition (); - 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; - - default: - complain (_("unrecognized: %s"), token_buffer); - skip_to_char ('%'); - } - } - else if (c == EOF) - fatal (_("no input grammar")); - else - { - char buf[] = "c"; - buf[0] = c; - complain (_("unknown character: %s"), quote (buf)); - skip_to_char ('%'); - } - } -} - -/*-------------------------------------------------------------------. -| 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. | -`-------------------------------------------------------------------*/ - -static void -copy_action (symbol_list *rule, int stack_offset) -{ - 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, '{'); +/* Attach a merge function NAME with argument type TYPE to current + rule. */ - count = 1; - c = getc (finput); - - 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 '}' */ - - if (--count) - { - obstack_1grow (&action_obstack, c); - c = getc (finput); - } - } - - obstack_sgrow (&action_obstack, ";\n break;}"); -} - -/*-------------------------------------------------------------------. -| 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); -} - - -static void -record_rule_line (void) +void +grammar_current_rule_merge_set (uniqstr name, location loc) { - /* 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; + 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. */ -/*-------------------------------------------------------------------. -| Generate a dummy symbol, a nonterminal, whose name cannot conflict | -| with the user's names. | -`-------------------------------------------------------------------*/ - -static bucket * -gensym (void) +void +grammar_current_rule_symbol_append (symbol *sym, location loc, + named_ref *name) { - /* Incremented for each generated symbol */ - static int gensym_count = 0; - static char buf[256]; - - bucket *sym; - - sprintf (buf, "@%d", ++gensym_count); - token_buffer = buf; - sym = getsym (token_buffer); - sym->class = nterm_sym; - sym->value = nvars++; - return sym; + 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 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 | -`------------------------------------------------------------------*/ +/* Attach an ACTION to the current rule. */ -static int -get_type (void) +void +grammar_current_rule_action_append (const char *action, location loc, + named_ref *name, bool is_predicate) { - 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; - } - } + 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); } -#endif -/*------------------------------------------------------------------. -| 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. | -`------------------------------------------------------------------*/ +/*---------------------------------------------------------------. +| Convert the rules into the representation using RRHS, RLHS and | +| RITEM. | +`---------------------------------------------------------------*/ static void -readgram (void) +packgram (void) { - token_t t; - bucket *lhs = NULL; - symbol_list *p; - symbol_list *p1; - bucket *bp; + unsigned int itemno = 0; + rule_number ruleno = 0; + symbol_list *p = grammar; - /* Points to first symbol_list of current rule. its symbol is the - lhs of the rule. */ - symbol_list *crule; - /* Points to the symbol_list preceding crule. */ - symbol_list *crule1; + ritem = xnmalloc (nritems + 1, sizeof *ritem); - p1 = NULL; + /* This sentinel is used by build_relations in gram.c. */ + *ritem++ = 0; - t = lex (); + rules = xnmalloc (nrules, sizeof *rules); - while (t != tok_two_percents && t != tok_eof) + while (p) { - 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 (); - - p = XCALLOC (symbol_list, 1); - p->sym = 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->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. */ - - 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) - 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 $$ but we need one. */ - else if (!xactions && !first_rhs && lhs->type_name != 0) - complain (_("empty rule for typed nonterminal, and no action")); - if (t == tok_semicolon) - t = lex (); - } -#if 0 - /* these things can appear as alternatives to rules. */ -/* NO, they cannot. - a) none of the documentation allows them - b) most of them scan forward until finding a next % - thus they may swallow lots of intervening rules -*/ - else if (t == tok_token) - { - parse_token_decl (token_sym, nterm_sym); - t = lex (); - } - else if (t == tok_nterm) - { - parse_token_decl (nterm_sym, token_sym); - t = lex (); - } - else if (t == tok_type) - { - t = get_type (); - } - else if (t == tok_union) - { - parse_union_decl (); - t = lex (); - } - else if (t == tok_expect) - { - parse_expect_decl (); - t = lex (); - } - else if (t == tok_start) - { - parse_start_decl (); - t = lex (); - } -#endif + int rule_length = 0; + symbol *ruleprec = p->ruleprec; + record_merge_function_type (p->merger, p->content.sym->type_name, + p->merger_declaration_location); + rules[ruleno].user_number = ruleno; + rules[ruleno].number = ruleno; + rules[ruleno].lhs = p->content.sym; + rules[ruleno].rhs = ritem + itemno; + rules[ruleno].prec = NULL; + rules[ruleno].dprec = p->dprec; + rules[ruleno].merger = p->merger; + rules[ruleno].precsym = NULL; + rules[ruleno].location = p->location; + rules[ruleno].useful = true; + rules[ruleno].action = p->action_props.code; + rules[ruleno].action_location = p->action_props.location; + rules[ruleno].is_predicate = p->action_props.is_predicate; + + /* If the midrule's $$ 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; + } - 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. */ - - 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; -} - -/* 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) -{ - char c; - struct obstack el_obstack; + /* If this rule has a %prec, + the specified symbol's precedence replaces the default. */ + if (ruleprec) + { + 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); - obstack_init (&el_obstack); + if (p) + p = p->next; + } - while ((c = getc (finput)) != EOF) - obstack_1grow (&el_obstack, c); + aver (itemno == nritems); - obstack_1grow (&el_obstack, 0); - muscle_insert ("epilogue", obstack_finish (&el_obstack)); + if (trace_flag & trace_sets) + ritem_print (stderr); } - -/*--------------------------------------------------------------. -| For named tokens, but not literal ones, define the name. The | -| value is the user token number. | -`--------------------------------------------------------------*/ +/*------------------------------------------------------------------. +| 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. | +`------------------------------------------------------------------*/ -static void -output_token_defines (struct obstack *oout) +void +reader (void) { - bucket *bp; - char *cp, *symbol; - char c; + /* Initialize the symbol table. */ + symbols_new (); - for (bp = firstsymbol; bp; bp = bp->next) - { - 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) - { - /* use literal string only if given a symbol with an alias */ - if (bp->alias) - symbol = bp->alias->tag; - else - continue; - } - - /* 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); - } -} + /* Construct the accept symbol. */ + accept = symbol_get ("$accept", empty_location); + accept->class = nterm_sym; + accept->number = nvars++; + /* Construct the error token */ + errtoken = symbol_get ("error", empty_location); + errtoken->class = token_sym; + errtoken->number = ntokens++; -/*------------------------------------------------------------------. -| Set TOKEN_TRANSLATIONS. Check that no two symbols share the same | -| number. | -`------------------------------------------------------------------*/ + /* Construct a token that represents all undefined literal tokens. + It is always token number 2. */ + undeftoken = symbol_get ("$undefined", empty_location); + undeftoken->class = token_sym; + undeftoken->number = ntokens++; -static void -token_translations_init (void) -{ - bucket *bp = NULL; - int i; + gram_in = xfopen (grammar_file, "r"); - token_translations = XCALLOC (short, max_user_token_number + 1); + gram__flex_debug = trace_flag & trace_scan; + gram_debug = trace_flag & trace_parse; + gram_scanner_initialize (); + gram_parse (); + prepare_percent_define_front_end_variables (); - /* Initialize all entries for literal tokens to 2, the internal - token number for $undefined., which represents all invalid - inputs. */ - for (i = 0; i <= max_user_token_number; i++) - token_translations[i] = 2; + if (! complaint_issued) + check_and_convert_grammar (); - for (bp = firstsymbol; bp; bp = bp->next) - { - /* Non-terminal? */ - if (bp->value >= ntokens) - continue; - /* A token string alias? */ - if (bp->user_token_number == SALIAS) - continue; - - assert (bp->user_token_number != SUNDEF); - - /* A token which translation has already been set? */ - if (token_translations[bp->user_token_number] != 2) - complain (_("tokens %s and %s both assigned number %d"), - tags[token_translations[bp->user_token_number]], - bp->tag, bp->user_token_number); - token_translations[bp->user_token_number] = bp->value; - } + xfclose (gram_in); } - -/*------------------------------------------------------------------. -| Assign symbol numbers, and write definition of token names into | -| FDEFINES. Set up vectors TAGS and SPREC of names and precedences | -| of symbols. | -`------------------------------------------------------------------*/ - static void -packsymbols (void) +prepare_percent_define_front_end_variables (void) { - bucket *bp = NULL; - int tokno = 1; - int last_user_token_number; - static char DOLLAR[] = "$"; - - tags = XCALLOC (char *, nsyms + 1); - user_toknums = XCALLOC (short, nsyms + 1); - - 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; - } - - token_translations_init (); - - error_token_number = errtoken->value; - + /* Set %define front-end variable defaults. */ + muscle_percent_define_default ("lr.keep-unreachable-states", "false"); { - 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); + 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); } -#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 - } - } + /* 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); + } } -/*---------------------------------------------------------------. -| Convert the rules into the representation using RRHS, RLHS and | -| RITEMS. | -`---------------------------------------------------------------*/ +/*-------------------------------------------------------------. +| Check the grammar that has just been read, and convert it to | +| internal form. | +`-------------------------------------------------------------*/ static void -packgram (void) +check_and_convert_grammar (void) { - 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; - - itemno = 0; - ruleno = 1; + /* Grammar has been read. Do some checking. */ + if (nrules == 0) + complain (fatal, _("no rules in the input grammar")); - p = grammar; - while (p) + /* If the user did not define her ENDTOKEN, do it now. */ + if (!endtoken) { - rlhs[ruleno] = p->sym->value; - rrhs[ruleno] = itemno; - ruleprec = p->ruleprec; - - p = p->next; - while (p && p->sym) - { - ritem[itemno++] = p->sym->value; - /* 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; - } - if (p) - p = p->next; - } - - /* If this rule has a %prec, - the specified symbol's precedence replaces the default. */ - if (ruleprec) - { - rprec[ruleno] = ruleprec->prec; - rassoc[ruleno] = ruleprec->assoc; - rprecsym[ruleno] = ruleprec->value; - } - - ritem[itemno++] = -ruleno; - ruleno++; - - if (p) - p = p->next; + endtoken = symbol_get ("$end", empty_location); + endtoken->class = token_sym; + endtoken->number = 0; + /* Value specified by POSIX. */ + endtoken->user_token_number = 0; } - ritem[itemno] = 0; -} - -/*-------------------------------------------------------------------. -| 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. | -`-------------------------------------------------------------------*/ - -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; - - /* Initialize the muscle obstack. */ - obstack_init (&muscle_obstack); - - /* Initialize the symbol table. */ - tabinit (); - - /* Construct the error token */ - errtoken = getsym ("error"); - errtoken->class = token_sym; - errtoken->user_token_number = 256; /* Value specified by POSIX. */ + /* 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): + + 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; + } - /* 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; - - /* Read the declaration section. Copy %{ ... %} groups to - TABLE_OBSTACK and FDEFINES file. Also notice any %token, %left, - etc. found there. */ - read_declarations (); - /* 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 (); - - /* 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. */ + 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. */ + symbol_list_free (grammar); }