X-Git-Url: https://git.saurik.com/bison.git/blobdiff_plain/75142d452b7f6bc3b6be3d8a08c0f602fd157ec5..b87f8b2159820e022e3b726b8369f74d64712f8f:/src/reader.c diff --git a/src/reader.c b/src/reader.c index 2f583f1e..75957010 100644 --- a/src/reader.c +++ b/src/reader.c @@ -1,5 +1,5 @@ /* Input parser for bison - Copyright 1984, 1986, 1989, 1992, 1998, 2000, 2001 + Copyright (C) 1984, 1986, 1989, 1992, 1998, 2000, 2001, 2002 Free Software Foundation, Inc. This file is part of Bison, the GNU Compiler Compiler. @@ -38,42 +38,38 @@ typedef struct symbol_list { struct symbol_list *next; - bucket *sym; + symbol_t *sym; int line; /* The action is attached to the LHS of a rule. */ const char *action; int action_line; - /* The guard is attached to the LHS of a rule. */ - const char *guard; - int guard_line; - bucket *ruleprec; + symbol_t *ruleprec; } symbol_list; int lineno; -char **tags; -short *user_toknums; -static symbol_list *grammar; -static int start_flag; -static bucket *startval; +static symbol_list *grammar = NULL; +static int start_flag = 0; +static symbol_t *startval = NULL; /* Nonzero if components of semantic values are used, implying they must be unions. */ static int value_components_used; /* Nonzero if %union has been seen. */ -static int typed; +static int typed = 0; /* Incremented for each %left, %right or %nonassoc seen */ -static int lastprec; +static int lastprec = 0; -static bucket *errtoken = NULL; -static bucket *undeftoken = NULL; -static bucket *eoftoken = NULL; +symbol_t *errtoken = NULL; +symbol_t *undeftoken = NULL; +symbol_t *eoftoken = NULL; +symbol_t *axiom = NULL; static symbol_list * -symbol_list_new (bucket *sym) +symbol_list_new (symbol_t *sym) { symbol_list *res = XMALLOC (symbol_list, 1); res->next = NULL; @@ -81,12 +77,171 @@ symbol_list_new (bucket *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 = USER_NUMBER_ALIAS; + symval->alias = symbol; + symbol->alias = symval; + /* symbol and symval combined are only one symbol */ + nsyms--; + ntokens--; + assert (ntokens == symbol->number || ntokens == symval->number); + symbol->number = symval->number = + (symval->number < symbol->number) ? symval->number : symbol->number; + } + + return TRUE; +} + +/*---------------------------------------------------------. +| Check that THIS, and its alias, have same precedence and | +| associativity. | +`---------------------------------------------------------*/ + +static bool +symbol_check_alias_consistence (symbol_t *this) +{ + /* Check only those who _are_ the aliases. */ + if (this->alias && this->user_token_number == USER_NUMBER_ALIAS) + { + if (this->prec != this->alias->prec) + { + if (this->prec != 0 && this->alias->prec != 0) + complain (_("conflicting precedences for %s and %s"), + this->tag, this->alias->tag); + if (this->prec != 0) + this->alias->prec = this->prec; + else + this->prec = this->alias->prec; + } + + if (this->assoc != this->alias->assoc) + { + if (this->assoc != 0 && this->alias->assoc != 0) + complain (_("conflicting assoc values for %s and %s"), + this->tag, this->alias->tag); + if (this->assoc != 0) + this->alias->assoc = this->assoc; + else + this->assoc = this->alias->assoc; + } + } + return TRUE; +} + + +/*-------------------------------------------------------------------. +| Assign a symbol number, and write the definition of the token name | +| into FDEFINES. Put in SYMBOLS. | +`-------------------------------------------------------------------*/ + +static bool +symbol_pack (symbol_t *this) +{ + if (this->class == nterm_sym) + { + this->number += ntokens; + } + else if (this->alias) + { + /* This symbol and its alias are a single token defn. + Allocate a tokno, and assign to both check agreement of + prec and assoc fields and make both the same */ + if (this->number == NUMBER_UNDEFINED) + { + if (this == eoftoken || this->alias == eoftoken) + this->number = this->alias->number = 0; + else + { + assert (this->alias->number != NUMBER_UNDEFINED); + this->number = this->alias->number; + } + } + /* Do not do processing below for USER_NUMBER_ALIASs. */ + if (this->user_token_number == USER_NUMBER_ALIAS) + return TRUE; + } + else /* this->class == token_sym */ + { + assert (this->number != NUMBER_UNDEFINED); + } + + symbols[this->number] = this; + return TRUE; +} + + + + +/*--------------------------------------------------. +| Put THIS in TOKEN_TRANSLATIONS if it is a token. | +`--------------------------------------------------*/ + +static bool +symbol_translation (symbol_t *this) +{ + /* Non-terminal? */ + if (this->class == token_sym + && this->user_token_number != USER_NUMBER_ALIAS) + { + /* A token which translation has already been set? */ + if (token_translations[this->user_token_number] != undeftoken->number) + complain (_("tokens %s and %s both assigned number %d"), + symbols[token_translations[this->user_token_number]]->tag, + this->tag, this->user_token_number); + + token_translations[this->user_token_number] = this->number; + } + + return TRUE; +} /*===================\ @@ -166,12 +321,34 @@ get_type_name (int n, symbol_list *rule) complain (_("invalid $ value")); return NULL; } - i++; + ++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; + + 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 "). | @@ -199,17 +376,17 @@ copy_string2 (FILE *fin, struct obstack *oout, int match, int store) continue; } - obstack_1grow (oout, c); + copy_character (oout, c); if (c == '\\') { c = getc (fin); if (c == EOF) fatal (_("unterminated string at end of file")); - obstack_1grow (oout, c); + copy_character (oout, c); if (c == '\n') - lineno++; + ++lineno; } c = getc (fin); @@ -292,7 +469,7 @@ copy_comment (FILE *fin, struct obstack *oout) } else if (c == '\n') { - lineno++; + ++lineno; obstack_1grow (oout, c); if (cplus_comment) ended = 1; @@ -303,30 +480,29 @@ copy_comment (FILE *fin, struct obstack *oout) fatal (_("unterminated comment")); else { - obstack_1grow (oout, c); + copy_character (oout, c); c = getc (fin); } } } -/*-----------------------------------------------------------------. -| 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. | -`-----------------------------------------------------------------*/ +/*-------------------------------------------------------------------. +| 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 `$0' with | +| respect to the top of the stack. | +`-------------------------------------------------------------------*/ static inline void -copy_at (FILE *fin, struct obstack *oout, int stack_offset) +copy_at (FILE *fin, struct obstack *oout, int rule_length) { - int c; + int c = getc (fin); + locations_flag = 1; - c = getc (fin); if (c == '$') { - obstack_sgrow (oout, "yyloc"); - locations_flag = 1; + obstack_sgrow (oout, "]b4_lhs_location["); } else if (isdigit (c) || c == '-') { @@ -334,16 +510,11 @@ copy_at (FILE *fin, struct obstack *oout, int stack_offset) ungetc (c, fin); n = read_signed_integer (fin); - if (n > stack_offset) + if (n > rule_length) complain (_("invalid value: %s%d"), "@", n); else - { - /* Offset is always 0 if parser has already popped the stack - pointer. */ - obstack_fgrow1 (oout, "yylsp[%d]", - n - (semantic_parser ? 0 : stack_offset)); - locations_flag = 1; - } + obstack_fgrow2 (oout, "]b4_rhs_location([%d], [%d])[", + rule_length, n); } else { @@ -354,19 +525,19 @@ copy_at (FILE *fin, struct obstack *oout, int stack_offset) } -/*-------------------------------------------------------------------. -| FIN is pointing to a wannabee semantic value (i.e., a `$'). | -| | -| Possible inputs: $[]($|integer) | -| | -| Output to OOUT a reference to this semantic value. STACK_OFFSET is | -| the number of values in the current rule so far, which says where | -| to find `$0' with respect to the top of the stack. | -`-------------------------------------------------------------------*/ +/*------------------------------------------------------------------. +| FIN is pointing to a wannabee semantic value (i.e., a `$'). | +| | +| Possible inputs: $[]($|integer) | +| | +| Output to OOUT a reference to this semantic value. RULE_LENGTH is | +| the number of values in the current rule so far, which says where | +| to find `$0' with respect to the top of the stack. | +`------------------------------------------------------------------*/ static inline void copy_dollar (FILE *fin, struct obstack *oout, - symbol_list *rule, int stack_offset) + symbol_list *rule, int rule_length) { int c = getc (fin); const char *type_name = NULL; @@ -382,15 +553,15 @@ copy_dollar (FILE *fin, struct obstack *oout, if (c == '$') { - obstack_sgrow (oout, "yyval"); - if (!type_name) type_name = get_type_name (0, rule); - if (type_name) - obstack_fgrow1 (oout, ".%s", type_name); if (!type_name && typed) complain (_("$$ of `%s' has no declared type"), rule->sym->tag); + if (!type_name) + type_name = ""; + obstack_fgrow1 (oout, + "]b4_lhs_value([%s])[", type_name); } else if (isdigit (c) || c == '-') { @@ -398,23 +569,19 @@ copy_dollar (FILE *fin, struct obstack *oout, ungetc (c, fin); n = read_signed_integer (fin); - if (n > stack_offset) + if (n > rule_length) complain (_("invalid value: %s%d"), "$", n); else { if (!type_name && n > 0) type_name = get_type_name (n, rule); - - /* Offset is always 0 if parser has already popped the stack - pointer. */ - obstack_fgrow1 (oout, "yyvsp[%d]", - n - (semantic_parser ? 0 : stack_offset)); - - if (type_name) - obstack_fgrow1 (oout, ".%s", type_name); if (!type_name && typed) complain (_("$%d of `%s' has no declared type"), n, rule->sym->tag); + if (!type_name) + type_name = ""; + obstack_fgrow3 (oout, "]b4_rhs_value([%d], [%d], [%s])[", + rule_length, n, type_name); } } else @@ -431,7 +598,7 @@ copy_dollar (FILE *fin, struct obstack *oout, `-------------------------------------------------------------------*/ static void -copy_definition (void) +copy_definition (struct obstack *oout) { int c; /* -1 while reading a character if prev char was %. */ @@ -439,7 +606,7 @@ copy_definition (void) if (!no_lines_flag) { - obstack_fgrow2 (&attrs_obstack, muscle_find ("linef"), + obstack_fgrow2 (oout, muscle_find ("linef"), lineno, quotearg_style (c_quoting_style, muscle_find ("filename"))); } @@ -453,8 +620,8 @@ copy_definition (void) switch (c) { case '\n': - obstack_1grow (&attrs_obstack, c); - lineno++; + obstack_1grow (oout, c); + ++lineno; break; case '%': @@ -463,18 +630,18 @@ copy_definition (void) case '\'': case '"': - copy_string (finput, &attrs_obstack, c); + copy_string (finput, oout, c); break; case '/': - copy_comment (finput, &attrs_obstack); + copy_comment (finput, oout); break; case EOF: fatal ("%s", _("unterminated `%{' definition")); default: - obstack_1grow (&attrs_obstack, c); + copy_character (oout, c); } c = getc (finput); @@ -483,7 +650,7 @@ copy_definition (void) { if (c == '}') return; - obstack_1grow (&attrs_obstack, '%'); + obstack_1grow (oout, '%'); } after_percent = 0; } @@ -503,7 +670,7 @@ parse_token_decl (symbol_class what_is, symbol_class what_is_not) char *typename = NULL; /* The symbol being defined. */ - struct bucket *symbol = NULL; + symbol_t *symbol = NULL; /* After `%token' and `%nterm', any number of symbols maybe be defined. */ @@ -532,23 +699,7 @@ parse_token_decl (symbol_class what_is, symbol_class what_is_not) } else if (token == tok_identifier && *symval->tag == '\"' && symbol) { - if (symval->alias) - warn (_("symbol `%s' used more than once as a literal string"), - symval->tag); - else if (symbol->alias) - warn (_("symbol `%s' given more than one literal string"), - symbol->tag); - else - { - symval->class = token_sym; - symval->type_name = typename; - symval->user_token_number = symbol->user_token_number; - symbol->user_token_number = SALIAS; - symval->alias = symbol; - symbol->alias = symval; - /* symbol and symval combined are only one symbol */ - nsyms--; - } + symbol_make_alias (symbol, typename); symbol = NULL; } else if (token == tok_identifier) @@ -560,7 +711,9 @@ parse_token_decl (symbol_class what_is, 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++; + symbol->number = nvars++; + if (what_is == token_sym && symbol->number == NUMBER_UNDEFINED) + symbol->number = ntokens++; if (typename) { @@ -575,7 +728,13 @@ parse_token_decl (symbol_class what_is, symbol_class what_is_not) symbol->user_token_number = numval; /* User defined EOF token? */ if (numval == 0) - eoftoken = symbol; + { + eoftoken = symbol; + eoftoken->number = 0; + /* It is always mapped to 0, so it was already counted in + NTOKENS. */ + --ntokens; + } } else { @@ -674,7 +833,8 @@ parse_assoc_decl (associativity assoc) char *name = NULL; int prev = 0; - lastprec++; /* Assign a new precedence level, never 0. */ + /* Assign a new precedence level, never 0. */ + ++lastprec; for (;;) { @@ -704,7 +864,11 @@ parse_assoc_decl (associativity assoc) symval->assoc = assoc; if (symval->class == nterm_sym) complain (_("symbol %s redefined"), symval->tag); - symval->class = token_sym; + if (symval->number == NUMBER_UNDEFINED) + { + symval->number = ntokens++; + symval->class = token_sym; + } if (name) { /* record the type, if one is specified */ if (symval->type_name == NULL) @@ -721,9 +885,9 @@ parse_assoc_decl (associativity assoc) } else { - complain (_ - ("invalid text (%s) - number should be after identifier"), -token_buffer); + complain + (_("invalid text (%s) - number should be after identifier"), + token_buffer); skip_to_char ('%'); } break; @@ -760,6 +924,7 @@ parse_union_decl (void) typed = 1; + MUSCLE_INSERT_INT ("stype_line", lineno); obstack_init (&union_obstack); obstack_sgrow (&union_obstack, "union"); @@ -774,7 +939,7 @@ parse_union_decl (void) switch (c) { case '\n': - lineno++; + ++lineno; break; case '/': @@ -782,7 +947,7 @@ parse_union_decl (void) break; case '{': - count++; + ++count; break; case '}': @@ -833,10 +998,10 @@ parse_expect_decl (void) | | | 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. | +| declaration. The ->user_token_number of the first is | +| USER_NUMBER_ALIAS and the ->user_token_number of the second is set | +| to the number, if any, from the declaration. The two symbols are | +| linked via pointers in their ->alias fields. | | | | During OUTPUT_DEFINES_TABLE, the symbol is reported thereafter, | | only the literal string is retained it is the literal string that | @@ -847,9 +1012,9 @@ static void parse_thong_decl (void) { token_t token; - struct bucket *symbol; + symbol_t *symbol; char *typename = 0; - int usrtoknum = SUNDEF; + int usrtoknum = USER_NUMBER_UNDEFINED; token = lex (); /* fetch typename or first token */ if (token == tok_typename) @@ -870,7 +1035,7 @@ parse_thong_decl (void) } symval->class = token_sym; symval->type_name = typename; - symval->user_token_number = SALIAS; + symval->user_token_number = USER_NUMBER_ALIAS; symbol = symval; token = lex (); /* get number or literal string */ @@ -996,7 +1161,7 @@ parse_skel_decl (void) /*----------------------------------------------------------------. | Read from finput until `%%' is seen. Discard the `%%'. Handle | | any `%' declarations, and copy the contents of any `%{ ... %}' | -| groups to ATTRS_OBSTACK. | +| groups to PRE_PROLOGUE_OBSTACK or POST_PROLOGUE_OBSTACK. | `----------------------------------------------------------------*/ static void @@ -1016,7 +1181,10 @@ read_declarations (void) return; case tok_percent_left_curly: - copy_definition (); + if (!typed) + copy_definition (&pre_prologue_obstack); + else + copy_definition (&post_prologue_obstack); break; case tok_token: @@ -1073,7 +1241,7 @@ read_declarations (void) case tok_stropt: case tok_intopt: case tok_obsolete: - abort (); + assert (0); break; case tok_illegal: @@ -1094,37 +1262,37 @@ read_declarations (void) } } -/*-------------------------------------------------------------------. -| 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. | -| | -| 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. | -`-------------------------------------------------------------------*/ +/*------------------------------------------------------------------. +| Assuming that a `{' has just been seen, copy everything up to the | +| matching `}' into ACTION_OBSTACK. | +| | +| RULE_LENGTH is the number of values in the current rule so far, | +| which says where to find `$0' with respect to the top of the | +| stack. It is not the same as the rule->length in the case of mid | +| rule actions. | +| | +| This routine is used for actions. | +`------------------------------------------------------------------*/ static void -parse_braces (symbol_list *rule, int stack_offset) +parse_action (symbol_list *rule, int rule_length) { - int c; - int count; - - count = 1; + int count = 1; + rule->action_line = lineno; while (count > 0) { + int c; while ((c = getc (finput)) != '}') switch (c) { case '\n': obstack_1grow (&action_obstack, c); - lineno++; + ++lineno; break; case '{': obstack_1grow (&action_obstack, c); - count++; + ++count; break; case '\'': @@ -1137,13 +1305,11 @@ parse_braces (symbol_list *rule, int stack_offset) break; case '$': - copy_dollar (finput, &action_obstack, - rule, stack_offset); + copy_dollar (finput, &action_obstack, rule, rule_length); break; case '@': - copy_at (finput, &action_obstack, - stack_offset); + copy_at (finput, &action_obstack, rule_length); break; case EOF: @@ -1155,36 +1321,13 @@ parse_braces (symbol_list *rule, int stack_offset) /* Above loop exits when C is '}'. */ if (--count) - { - obstack_1grow (&action_obstack, c); - c = getc (finput); - } + obstack_1grow (&action_obstack, c); } obstack_1grow (&action_obstack, '\0'); -} - - -static void -parse_action (symbol_list *rule, int stack_offset) -{ - rule->action_line = lineno; - parse_braces (rule, stack_offset); rule->action = obstack_finish (&action_obstack); } - -static void -parse_guard (symbol_list *rule, int stack_offset) -{ - token_t t = lex (); - if (t != tok_left_curly) - complain (_("invalid %s declaration"), "%guard"); - rule->guard_line = lineno; - parse_braces (rule, stack_offset); - rule->guard = obstack_finish (&action_obstack); -} - /*-------------------------------------------------------------------. @@ -1192,20 +1335,20 @@ parse_guard (symbol_list *rule, int stack_offset) | with the user's names. | `-------------------------------------------------------------------*/ -static bucket * +static symbol_t * gensym (void) { /* Incremented for each generated symbol */ static int gensym_count = 0; static char buf[256]; - bucket *sym; + symbol_t *sym; sprintf (buf, "@%d", ++gensym_count); token_buffer = buf; sym = getsym (token_buffer); sym->class = nterm_sym; - sym->value = nvars++; + sym->number = nvars++; return sym; } @@ -1216,8 +1359,8 @@ gensym (void) | 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. | +| All actions are copied out, labelled by the rule number they apply | +| to. | | | | Bison used to allow some %directives in the rules sections, but | | this is no longer consider appropriate: (i) the documented grammar | @@ -1231,10 +1374,9 @@ static void readgram (void) { token_t t; - bucket *lhs = NULL; + symbol_t *lhs = NULL; symbol_list *p = NULL; symbol_list *p1 = NULL; - bucket *bp; /* Points to first symbol_list of current rule. its symbol is the lhs of the rule. */ @@ -1251,7 +1393,7 @@ readgram (void) /* Number of symbols in rhs of this rule so far */ int rulelength = 0; int xactions = 0; /* JF for error checking */ - bucket *first_rhs = 0; + symbol_t *first_rhs = 0; if (t == tok_identifier) { @@ -1278,8 +1420,8 @@ readgram (void) } /* start a new rule and record its lhs. */ - nrules++; - nitems++; + ++nrules; + ++nritems; p = symbol_list_new (lhs); @@ -1297,8 +1439,8 @@ readgram (void) if (lhs->class == unknown_sym) { lhs->class = nterm_sym; - lhs->value = nvars; - nvars++; + lhs->number = nvars; + ++nvars; } else if (lhs->class == token_sym) complain (_("rule given for %s, which is a token"), lhs->tag); @@ -1322,7 +1464,7 @@ readgram (void) If one does, exit this rule now. */ if (t == tok_identifier) { - bucket *ssave; + symbol_t *ssave; token_t t1; ssave = symval; @@ -1330,7 +1472,10 @@ readgram (void) unlex (t1); symval = ssave; if (t1 == tok_colon) - break; + { + warn (_("previous rule lacks an ending `;'")); + break; + } if (!first_rhs) /* JF */ first_rhs = symval; @@ -1348,16 +1493,21 @@ readgram (void) inserting the new rule before it. */ /* Make a dummy nonterminal, a gensym. */ - bucket *sdummy = 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++; - nitems++; + ++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 @@ -1370,7 +1520,7 @@ readgram (void) /* Insert the dummy generated by that rule into this rule. */ - nitems++; + ++nritems; p = symbol_list_new (sdummy); p1->next = p; p1 = p; @@ -1380,7 +1530,7 @@ readgram (void) if (t == tok_identifier) { - nitems++; + ++nritems; p = symbol_list_new (symval); p1->next = p; p1 = p; @@ -1389,9 +1539,9 @@ readgram (void) { parse_action (crule, rulelength); action_flag = 1; - xactions++; /* JF */ + ++xactions; /* JF */ } - rulelength++; + ++rulelength; } /* end of read rhs of rule */ /* Put an empty link in the list to mark the end of this rule */ @@ -1407,15 +1557,6 @@ readgram (void) 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 */ @@ -1423,7 +1564,7 @@ readgram (void) complain (_("two actions at end of one rule")); parse_action (crule, rulelength); action_flag = 1; - xactions++; /* -wjh */ + ++xactions; /* -wjh */ t = lex (); } /* If $$ is being set in default way, report if any type @@ -1441,6 +1582,8 @@ readgram (void) /* 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 (); } @@ -1450,28 +1593,34 @@ readgram (void) t = lex (); } - /* grammar has been read. Do some checking */ - if (nsyms > MAXSHORT) - fatal (_("too many symbols (tokens plus nonterminals); maximum %d"), - MAXSHORT); if (nrules == 0) fatal (_("no rules in the input grammar")); /* Report any undefined symbols and consider them nonterminals. */ + symbols_do (symbol_check_defined, NULL); + + /* Insert the initial rule, which line is that of the first rule + (not that of the start symbol): + + axiom: %start EOF. */ + p = symbol_list_new (axiom); + p->line = grammar->line; + p->next = symbol_list_new (startval); + p->next->next = symbol_list_new (eoftoken); + p->next->next->next = symbol_list_new (NULL); + p->next->next->next->next = grammar; + nrules += 1; + nritems += 3; + grammar = p; + startval = axiom; + + if (nsyms > SHRT_MAX) + fatal (_("too many symbols (tokens plus nonterminals); maximum %d"), + SHRT_MAX); - for (bp = firstsymbol; bp; bp = bp->next) - if (bp->class == unknown_sym) - { - complain (_ - ("symbol %s is used, but is not defined as a token and has no rules"), - bp->tag); - bp->class = nterm_sym; - bp->value = nvars++; - } - - ntokens = nsyms - nvars; + assert (nsyms == ntokens + nvars); } /* At the end of the grammar file, some C source code must @@ -1480,7 +1629,7 @@ readgram (void) static void read_additionnal_code (void) { - char c; + int c; struct obstack el_obstack; obstack_init (&el_obstack); @@ -1493,7 +1642,7 @@ read_additionnal_code (void) } while ((c = getc (finput)) != EOF) - obstack_1grow (&el_obstack, c); + copy_character (&el_obstack, c); obstack_1grow (&el_obstack, 0); muscle_insert ("epilogue", obstack_finish (&el_obstack)); @@ -1508,198 +1657,80 @@ read_additionnal_code (void) static void token_translations_init (void) { - bucket *bp = NULL; + int num_256_available_p = TRUE; int i; - token_translations = XCALLOC (short, max_user_token_number + 1); + /* 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 != USER_NUMBER_UNDEFINED) + { + 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; + } + } - /* 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 256 is not used, assign it to error, to follow POSIX. */ + if (num_256_available_p + && errtoken->user_token_number == USER_NUMBER_UNDEFINED) + errtoken->user_token_number = 256; + + /* Set the missing user numbers. */ + if (max_user_token_number < 256) + max_user_token_number = 256; - for (bp = firstsymbol; bp; bp = bp->next) + for (i = 0; i < ntokens; ++i) { - /* Non-terminal? */ - if (bp->value >= ntokens) - continue; - /* A token string alias? */ - if (bp->user_token_number == SALIAS) - continue; + symbol_t *this = symbols[i]; + if (this->user_token_number == USER_NUMBER_UNDEFINED) + this->user_token_number = ++max_user_token_number; + if (this->user_token_number > max_user_token_number) + max_user_token_number = this->user_token_number; + } - assert (bp->user_token_number != SUNDEF); + token_translations = XCALLOC (token_number_t, max_user_token_number + 1); - /* 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; - } + /* 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); } -/*------------------------------------------------------------------. -| Assign symbol numbers, and write definition of token names into | -| FDEFINES. Set up vectors TAGS and SPREC of names and precedences | -| of symbols. | -`------------------------------------------------------------------*/ +/*----------------------------------------------------------------. +| Assign symbol numbers, and write definition of token names into | +| FDEFINES. Set up vectors SYMBOL_TABLE, TAGS of symbols. | +`----------------------------------------------------------------*/ static void packsymbols (void) { - bucket *bp = NULL; - int tokno = 1; - int last_user_token_number; - - tags = XCALLOC (char *, nsyms + 1); - user_toknums = XCALLOC (short, nsyms + 1); - - sprec = XCALLOC (short, nsyms); - sassoc = XCALLOC (short, nsyms); - - 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 == -1) - { - if (bp == eoftoken || bp->alias == eoftoken) - bp->value = bp->alias->value = 0; - else - { - bp->value = bp->alias->value = tokno++; - } - } + symbols = XCALLOC (symbol_t *, nsyms); - 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; - } - - /* Do not do processing below for SALIASs. */ - if (bp->user_token_number == SALIAS) - continue; - - } - else /* bp->class == token_sym */ - { - if (bp == eoftoken) - bp->value = 0; - else - bp->value = tokno++; - } - - if (bp->class == token_sym) - { - if (bp->user_token_number == SUNDEF) - bp->user_token_number = ++last_user_token_number; - if (bp->user_token_number > max_user_token_number) - max_user_token_number = bp->user_token_number; - } - - tags[bp->value] = bp->tag; - user_toknums[bp->value] = bp->user_token_number; - sprec[bp->value] = bp->prec; - sassoc[bp->value] = bp->assoc; - } + symbols_do (symbol_check_alias_consistence, NULL); + symbols_do (symbol_pack, NULL); token_translations_init (); - error_token_number = errtoken->value; - 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; -} - - -/*---------------------------------------------------------------. -| Save the definition of token names in the `TOKENDEFS' muscle. | -`---------------------------------------------------------------*/ - -static void -symbols_save (void) -{ - struct obstack tokendefs; - bucket *bp; - obstack_init (&tokendefs); - - for (bp = firstsymbol; bp; bp = bp->next) - { - char *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. */ - if (strchr (symbol, '.')) - continue; - - obstack_fgrow2 (&tokendefs, "# define %s\t%d\n", - symbol, bp->user_token_number); - if (semantic_parser) - /* FIXME: This is probably wrong, and should be just as - above. --akim. */ - obstack_fgrow2 (&tokendefs, "# define T%s\t%d\n", symbol, bp->value); - } - - obstack_1grow (&tokendefs, 0); - muscle_insert ("tokendef", xstrdup (obstack_finish (&tokendefs))); - obstack_free (&tokendefs, NULL); + start_symbol = startval->number; } /*---------------------------------------------------------------. | Convert the rules into the representation using RRHS, RLHS and | -| RITEMS. | +| RITEM. | `---------------------------------------------------------------*/ static void @@ -1709,12 +1740,8 @@ packgram (void) int ruleno; symbol_list *p; - /* We use short to index items. */ - if (nitems >= MAXSHORT) - fatal (_("too many items (max %d)"), MAXSHORT); - - ritem = XCALLOC (short, nitems + 1); - rule_table = XCALLOC (rule_t, nrules) - 1; + ritem = XCALLOC (item_number_t, nritems + 1); + rules = XCALLOC (rule_t, nrules) - 1; itemno = 0; ruleno = 1; @@ -1722,27 +1749,26 @@ packgram (void) p = grammar; while (p) { - bucket *ruleprec = p->ruleprec; - rule_table[ruleno].lhs = p->sym->value; - rule_table[ruleno].rhs = itemno; - rule_table[ruleno].line = p->line; - rule_table[ruleno].useful = TRUE; - rule_table[ruleno].action = p->action; - rule_table[ruleno].action_line = p->action_line; - rule_table[ruleno].guard = p->guard; - rule_table[ruleno].guard_line = p->guard_line; + symbol_t *ruleprec = p->ruleprec; + rules[ruleno].user_number = ruleno; + rules[ruleno].number = ruleno; + rules[ruleno].lhs = p->sym; + rules[ruleno].rhs = ritem + itemno; + rules[ruleno].line = p->line; + rules[ruleno].useful = TRUE; + rules[ruleno].action = p->action; + rules[ruleno].action_line = p->action_line; p = p->next; while (p && p->sym) { - ritem[itemno++] = p->sym->value; + /* item_number_t = 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) - { - rule_table[ruleno].prec = p->sym->prec; - rule_table[ruleno].assoc = p->sym->assoc; - } + rules[ruleno].prec = p->sym; if (p) p = p->next; } @@ -1751,53 +1777,33 @@ packgram (void) the specified symbol's precedence replaces the default. */ if (ruleprec) { - rule_table[ruleno].prec = ruleprec->prec; - rule_table[ruleno].assoc = ruleprec->assoc; - rule_table[ruleno].precsym = ruleprec->value; + rules[ruleno].precsym = ruleprec; + rules[ruleno].prec = ruleprec; } - ritem[itemno++] = -ruleno; - ruleno++; + ++ruleno; if (p) p = p->next; } ritem[itemno] = 0; - nritems = itemno; - assert (nritems == nitems); + assert (itemno == nritems); if (trace_flag) 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) { - start_flag = 0; - startval = NULL; /* start symbol not specified yet. */ - - nsyms = 0; - nvars = 0; - nrules = 0; - nitems = 0; - - typed = 0; - lastprec = 0; - - semantic_parser = 0; - pure_parser = 0; - - grammar = NULL; - lex_init (); lineno = 1; @@ -1805,23 +1811,29 @@ reader (void) obstack_init (&muscle_obstack); /* Initialize the symbol table. */ - tabinit (); + symbols_new (); + + /* Construct the axiom symbol. */ + axiom = getsym ("$axiom"); + axiom->class = nterm_sym; + axiom->number = nvars++; /* Construct the error token */ errtoken = getsym ("error"); errtoken->class = token_sym; - errtoken->user_token_number = 256; /* Value specified by POSIX. */ + errtoken->number = ntokens++; /* Construct a token that represents all undefined literal tokens. It is always token number 2. */ undeftoken = getsym ("$undefined."); undeftoken->class = token_sym; - undeftoken->user_token_number = 2; + undeftoken->number = ntokens++; /* Initialize the obstacks. */ obstack_init (&action_obstack); - obstack_init (&attrs_obstack); obstack_init (&output_obstack); + obstack_init (&pre_prologue_obstack); + obstack_init (&post_prologue_obstack); finput = xfopen (infile, "r"); @@ -1835,12 +1847,13 @@ reader (void) { eoftoken = getsym ("$"); eoftoken->class = token_sym; + eoftoken->number = 0; /* Value specified by POSIX. */ eoftoken->user_token_number = 0; } /* Read in the grammar, build grammar in list form. Write out - guards and actions. */ + actions. */ readgram (); /* Some C code is given at the end of the grammar file. */ read_additionnal_code (); @@ -1851,9 +1864,19 @@ reader (void) /* Assign the symbols their symbol numbers. Write #defines for the token symbols into FDEFINES if requested. */ packsymbols (); - /* Save them. */ - symbols_save (); /* 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 (); }