X-Git-Url: https://git.saurik.com/bison.git/blobdiff_plain/428046f8d4365e2cd60f8a5f645c2f3e23e3f7af..e91ff6ce3d2653b41265591dafd4bdb32ae114c2:/src/reader.c diff --git a/src/reader.c b/src/reader.c index c06f4399..6a99cfff 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. @@ -21,7 +21,6 @@ #include "system.h" -#include "obstack.h" #include "quotearg.h" #include "quote.h" #include "getargs.h" @@ -39,47 +38,215 @@ 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; - bucket *ruleprec; -} -symbol_list; + + /* The guard is attached to the LHS of a rule. */ + const char *guard; + int guard_line; + symbol_t *ruleprec; +} symbol_list; int lineno; -char **tags; -short *user_toknums; -static symbol_list *grammar; -static int start_flag; -static bucket *startval; +static symbol_list *grammar = NULL; +static int start_flag = 0; +static symbol_t *startval = NULL; /* Nonzero if components of semantic values are used, implying they must be unions. */ static int value_components_used; /* Nonzero if %union has been seen. */ -static int typed; +static int typed = 0; /* Incremented for each %left, %right or %nonassoc seen */ -static int lastprec; - -static bucket *errtoken; -static bucket *undeftoken; +static int lastprec = 0; +symbol_t *errtoken = NULL; +symbol_t *undeftoken = NULL; +symbol_t *eoftoken = NULL; +symbol_t *axiom = NULL; static symbol_list * -symbol_list_new (bucket *sym) +symbol_list_new (symbol_t *sym) { symbol_list *res = XMALLOC (symbol_list, 1); res->next = NULL; res->sym = sym; res->line = lineno; + res->action = NULL; + res->action_line = 0; + res->guard = NULL; + res->guard_line = 0; res->ruleprec = NULL; return res; } +/*------------------------. +| Operations on symbols. | +`------------------------*/ + + +/*-----------------------------------------------------------. +| If THIS is not defined, report an error, and consider it a | +| nonterminal. | +`-----------------------------------------------------------*/ + +static bool +symbol_check_defined (symbol_t *this) +{ + if (this->class == unknown_sym) + { + complain + (_("symbol %s is used, but is not defined as a token and has no rules"), + this->tag); + this->class = nterm_sym; + this->number = nvars++; + } + + return TRUE; +} + + +/*-------------------------------------------------------------------. +| Assign a symbol number, and write the definition of the token name | +| into FDEFINES. Put in SYMBOLS. | +`-------------------------------------------------------------------*/ + +static bool +symbol_make_alias (symbol_t *symbol, char *typename) +{ + if (symval->alias) + warn (_("symbol `%s' used more than once as a literal string"), + symval->tag); + else if (symbol->alias) + warn (_("symbol `%s' given more than one literal string"), + symbol->tag); + else + { + symval->class = token_sym; + symval->type_name = typename; + symval->user_token_number = symbol->user_token_number; + symbol->user_token_number = SALIAS; + symval->alias = symbol; + symbol->alias = symval; + /* symbol and symval combined are only one symbol */ + nsyms--; + ntokens--; + assert (ntokens == symbol->number || ntokens == symval->number); + symbol->number = symval->number = + (symval->number < symbol->number) ? symval->number : symbol->number; + } + + return TRUE; +} + +/*---------------------------------------------------------. +| Check that THIS, and its alias, have same precedence and | +| associativity. | +`---------------------------------------------------------*/ + +static bool +symbol_check_alias_consistence (symbol_t *this) +{ + /* Check only those who _are_ the aliases. */ + if (this->alias && this->user_token_number == SALIAS) + { + if (this->prec != this->alias->prec) + { + if (this->prec != 0 && this->alias->prec != 0) + complain (_("conflicting precedences for %s and %s"), + this->tag, this->alias->tag); + if (this->prec != 0) + this->alias->prec = this->prec; + else + this->prec = this->alias->prec; + } + + if (this->assoc != this->alias->assoc) + { + if (this->assoc != 0 && this->alias->assoc != 0) + complain (_("conflicting assoc values for %s and %s"), + this->tag, this->alias->tag); + if (this->assoc != 0) + this->alias->assoc = this->assoc; + else + this->assoc = this->alias->assoc; + } + } + return TRUE; +} + + +/*-------------------------------------------------------------------. +| Assign a symbol number, and write the definition of the token name | +| into FDEFINES. Put in SYMBOLS. | +`-------------------------------------------------------------------*/ + +static bool +symbol_pack (symbol_t *this) +{ + if (this->class == nterm_sym) + { + this->number += ntokens; + } + else if (this->alias) + { + /* This symbol and its alias are a single token defn. + Allocate a tokno, and assign to both check agreement of + prec and assoc fields and make both the same */ + if (this->number == NUMBER_UNDEFINED) + { + if (this == eoftoken || this->alias == eoftoken) + this->number = this->alias->number = 0; + else + { + assert (this->alias->number != NUMBER_UNDEFINED); + this->number = this->alias->number; + } + } + /* Do not do processing below for SALIASs. */ + if (this->user_token_number == SALIAS) + return TRUE; + } + else /* this->class == token_sym */ + { + assert (this->number != NUMBER_UNDEFINED); + } + + symbols[this->number] = this; + return TRUE; +} + + + + +/*--------------------------------------------------. +| Put THIS in TOKEN_TRANSLATIONS if it is a token. | +`--------------------------------------------------*/ + +static bool +symbol_translation (symbol_t *this) +{ + /* Non-terminal? */ + if (this->class == token_sym + && this->user_token_number != SALIAS) + { + /* A token which translation has already been set? */ + if (token_translations[this->user_token_number] != undeftoken->number) + complain (_("tokens %s and %s both assigned number %d"), + symbols[token_translations[this->user_token_number]]->tag, + this->tag, this->user_token_number); + + token_translations[this->user_token_number] = this->number; + } + + return TRUE; +} /*===================\ @@ -159,7 +326,7 @@ get_type_name (int n, symbol_list *rule) complain (_("invalid $ value")); return NULL; } - i++; + ++i; } return rp->sym->type_name; @@ -202,7 +369,7 @@ copy_string2 (FILE *fin, struct obstack *oout, int match, int store) obstack_1grow (oout, c); if (c == '\n') - lineno++; + ++lineno; } c = getc (fin); @@ -233,25 +400,22 @@ copy_identifier (FILE *fin, struct obstack *oout) ungetc (c, fin); } -/*-----------------------------------------------------------------. -| Dump the wannabee comment from IN to OUT1 and OUT2 (which can be | -| NULL). In fact we just saw a `/', which might or might not be a | -| comment. In any case, copy what we saw. | -| | -| OUT2 might be NULL. | -`-----------------------------------------------------------------*/ + +/*------------------------------------------------------------------. +| Dump the wannabee comment from IN to OOUT. In fact we just saw a | +| `/', which might or might not be a comment. In any case, copy | +| what we saw. | +`------------------------------------------------------------------*/ static inline void -copy_comment2 (FILE *fin, struct obstack *oout1, struct obstack *oout2) +copy_comment (FILE *fin, struct obstack *oout) { int cplus_comment; int ended; int c; /* We read a `/', output it. */ - obstack_1grow (oout1, '/'); - if (oout2) - obstack_1grow (oout2, '/'); + obstack_1grow (oout, '/'); switch ((c = getc (fin))) { @@ -266,9 +430,7 @@ copy_comment2 (FILE *fin, struct obstack *oout1, struct obstack *oout2) return; } - obstack_1grow (oout1, c); - if (oout2) - obstack_1grow (oout2, c); + obstack_1grow (oout, c); c = getc (fin); ended = 0; @@ -278,26 +440,20 @@ copy_comment2 (FILE *fin, struct obstack *oout1, struct obstack *oout2) { while (c == '*') { - obstack_1grow (oout1, c); - if (oout2) - obstack_1grow (oout2, c); + obstack_1grow (oout, c); c = getc (fin); } if (c == '/') { - obstack_1grow (oout1, c); - if (oout2) - obstack_1grow (oout2, c); + obstack_1grow (oout, c); ended = 1; } } else if (c == '\n') { - lineno++; - obstack_1grow (oout1, c); - if (oout2) - obstack_1grow (oout2, c); + ++lineno; + obstack_1grow (oout, c); if (cplus_comment) ended = 1; else @@ -307,27 +463,13 @@ copy_comment2 (FILE *fin, struct obstack *oout1, struct obstack *oout2) fatal (_("unterminated comment")); else { - obstack_1grow (oout1, c); - if (oout2) - obstack_1grow (oout2, c); + obstack_1grow (oout, c); c = getc (fin); } } } -/*-------------------------------------------------------------------. -| Dump the comment (actually the current string starting with a `/') | -| from FIN to OOUT. | -`-------------------------------------------------------------------*/ - -static inline void -copy_comment (FILE *fin, struct obstack *oout) -{ - copy_comment2 (fin, oout, NULL); -} - - /*-----------------------------------------------------------------. | FIN is pointing to a location (i.e., a `@'). Output to OOUT a | | reference to this location. STACK_OFFSET is the number of values | @@ -352,9 +494,16 @@ copy_at (FILE *fin, struct obstack *oout, int stack_offset) ungetc (c, fin); n = read_signed_integer (fin); - - obstack_fgrow1 (oout, "yylsp[%d]", n - stack_offset); - locations_flag = 1; + if (n > stack_offset) + 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; + } } else { @@ -409,16 +558,24 @@ copy_dollar (FILE *fin, struct obstack *oout, ungetc (c, fin); n = read_signed_integer (fin); - if (!type_name && n > 0) - type_name = get_type_name (n, rule); - - obstack_fgrow1 (oout, "yyvsp[%d]", n - stack_offset); - - if (type_name) - obstack_fgrow1 (oout, ".%s", type_name); - if (!type_name && typed) - complain (_("$%d of `%s' has no declared type"), - n, rule->sym->tag); + if (n > stack_offset) + complain (_("invalid value: %s%d"), "$", 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); + } } else { @@ -444,7 +601,7 @@ copy_definition (void) { obstack_fgrow2 (&attrs_obstack, muscle_find ("linef"), lineno, quotearg_style (c_quoting_style, - muscle_find("filename"))); + muscle_find ("filename"))); } after_percent = 0; @@ -457,7 +614,7 @@ copy_definition (void) { case '\n': obstack_1grow (&attrs_obstack, c); - lineno++; + ++lineno; break; case '%': @@ -506,7 +663,7 @@ 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. */ @@ -535,23 +692,7 @@ 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) @@ -563,7 +704,9 @@ 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) { @@ -576,6 +719,15 @@ parse_token_decl (symbol_class what_is, symbol_class what_is_not) else if (symbol && token == tok_number) { symbol->user_token_number = numval; + /* User defined EOF token? */ + if (numval == 0) + { + eoftoken = symbol; + eoftoken->number = 0; + /* It is always mapped to 0, so it was already counted in + NTOKENS. */ + --ntokens; + } } else { @@ -674,7 +826,8 @@ 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 +857,11 @@ 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 +878,9 @@ 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; @@ -755,30 +912,14 @@ parse_union_decl (void) int count = 0; bool done = FALSE; struct obstack union_obstack; - const char *prologue = "\ -#ifndef YYSTYPE\n\ -typedef union"; - const char *epilogue = "\ - yystype;\n\ -# define YYSTYPE yystype\n\ -#endif\n"; - if (typed) complain (_("multiple %s declarations"), "%union"); typed = 1; - /* FIXME: I'm worried: are you sure attrs_obstack is properly - filled? */ - /* I don't see any reasons to keep this line, because we should - create a special skeleton for this option. */ - if (no_lines_flag) - obstack_1grow (&attrs_obstack, '\n'); - + MUSCLE_INSERT_INT ("stype_line", lineno); obstack_init (&union_obstack); obstack_sgrow (&union_obstack, "union"); - if (defines_flag) - obstack_sgrow (&defines_obstack, prologue); while (!done) { @@ -786,24 +927,20 @@ typedef union"; /* If C contains '/', it is output by copy_comment (). */ if (c != '/') - { - obstack_1grow (&union_obstack, c); - if (defines_flag) - obstack_1grow (&defines_obstack, c); - } + obstack_1grow (&union_obstack, c); switch (c) { case '\n': - lineno++; + ++lineno; break; case '/': - copy_comment2 (finput, &defines_obstack, &union_obstack); + copy_comment (finput, &union_obstack); break; case '{': - count++; + ++count; break; case '}': @@ -817,8 +954,6 @@ typedef union"; } } - if (defines_flag) - obstack_sgrow (&defines_obstack, epilogue); /* JF don't choke on trailing semi */ c = skip_white_space (); if (c != ';') @@ -870,7 +1005,7 @@ static void parse_thong_decl (void) { token_t token; - struct bucket *symbol; + symbol_t *symbol; char *typename = 0; int usrtoknum = SUNDEF; @@ -923,12 +1058,13 @@ parse_thong_decl (void) nsyms--; } + static void parse_muscle_decl (void) { int ch = ungetc (skip_white_space (), finput); - char* muscle_key; - char* muscle_value; + char *muscle_key; + char *muscle_value; /* Read key. */ if (!isalpha (ch) && ch != '_') @@ -1095,7 +1231,7 @@ read_declarations (void) case tok_stropt: case tok_intopt: case tok_obsolete: - abort (); + assert (0); break; case tok_illegal: @@ -1121,178 +1257,89 @@ read_declarations (void) | matching `}' into the actions file. STACK_OFFSET is the number of | | values in the current rule so far, which says where to find `$0' | | with respect to the top of the stack. | +| | +| This routine is used both for actions and guards. Only | +| ACTION_OBSTACK is used, but this is fine, since we use only | +| pointers to relevant portions inside this obstack. | `-------------------------------------------------------------------*/ static void -copy_action (symbol_list *rule, int stack_offset) +parse_braces (symbol_list *rule, int stack_offset) { int c; int count; - /* offset is always 0 if parser has already popped the stack pointer */ - if (semantic_parser) - stack_offset = 0; - count = 1; - c = getc (finput); - while (count > 0) { - while (c != '}') - { - switch (c) - { - case '\n': - obstack_1grow (&action_obstack, c); - lineno++; - break; - - case '{': - obstack_1grow (&action_obstack, c); - count++; - break; - - case '\'': - case '"': - copy_string (finput, &action_obstack, c); - break; - - case '/': - copy_comment (finput, &action_obstack); - break; - - case '$': - 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 '}' */ + while ((c = getc (finput)) != '}') + switch (c) + { + case '\n': + obstack_1grow (&action_obstack, c); + ++lineno; + break; + + case '{': + obstack_1grow (&action_obstack, c); + ++count; + break; + + case '\'': + case '"': + copy_string (finput, &action_obstack, c); + break; + + case '/': + copy_comment (finput, &action_obstack); + break; + + case '$': + copy_dollar (finput, &action_obstack, + rule, stack_offset); + break; + + case '@': + copy_at (finput, &action_obstack, + stack_offset); + break; + + case EOF: + fatal (_("unmatched %s"), "`{'"); + + default: + obstack_1grow (&action_obstack, c); + } + /* Above loop exits when C is '}'. */ if (--count) - { - obstack_1grow (&action_obstack, c); - c = getc (finput); - } + obstack_1grow (&action_obstack, c); } obstack_1grow (&action_obstack, '\0'); - rule->action = obstack_finish (&action_obstack); - rule->action_line = lineno; } - -/*-------------------------------------------------------------------. -| 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 `$0' with | -| respect to the top of the stack, for the simple parser in which | -| the stack is not popped until after the guard is run. | -`-------------------------------------------------------------------*/ + static void -copy_guard (symbol_list *rule, int stack_offset) +parse_action (symbol_list *rule, int stack_offset) { - int c; - int count; - int brace_flag = 0; - - /* offset is always 0 if parser has already popped the stack pointer */ - if (semantic_parser) - stack_offset = 0; - - obstack_fgrow1 (&guard_obstack, "\ncase %d:\n", nrules); - if (!no_lines_flag) - obstack_fgrow2 (&guard_obstack, muscle_find ("linef"), - lineno, quotearg_style (c_quoting_style, - muscle_find ("filename"))); - obstack_1grow (&guard_obstack, '{'); - - count = 0; - c = getc (finput); - - while (brace_flag ? (count > 0) : (c != ';')) - { - switch (c) - { - case '\n': - obstack_1grow (&guard_obstack, c); - lineno++; - break; - - case '{': - obstack_1grow (&guard_obstack, c); - brace_flag = 1; - count++; - break; - - case '}': - obstack_1grow (&guard_obstack, c); - if (count > 0) - count--; - else - { - complain (_("unmatched %s"), "`}'"); - c = getc (finput); /* skip it */ - } - break; - - case '\'': - case '"': - copy_string (finput, &guard_obstack, c); - break; - - case '/': - copy_comment (finput, &guard_obstack); - break; - - case '$': - 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); - } + rule->action_line = lineno; + parse_braces (rule, stack_offset); + rule->action = obstack_finish (&action_obstack); +} - 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 +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); } + /*-------------------------------------------------------------------. @@ -1300,20 +1347,20 @@ copy_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; } @@ -1339,10 +1386,9 @@ 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. */ @@ -1359,7 +1405,7 @@ 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) { @@ -1386,8 +1432,8 @@ readgram (void) } /* start a new rule and record its lhs. */ - nrules++; - nitems++; + ++nrules; + ++nritems; p = symbol_list_new (lhs); @@ -1405,8 +1451,8 @@ 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); @@ -1430,7 +1476,7 @@ readgram (void) If one does, exit this rule now. */ if (t == tok_identifier) { - bucket *ssave; + symbol_t *ssave; token_t t1; ssave = symval; @@ -1438,7 +1484,10 @@ 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; @@ -1456,16 +1505,21 @@ 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 @@ -1478,7 +1532,7 @@ readgram (void) /* Insert the dummy generated by that rule into this rule. */ - nitems++; + ++nritems; p = symbol_list_new (sdummy); p1->next = p; p1 = p; @@ -1488,18 +1542,18 @@ readgram (void) if (t == tok_identifier) { - nitems++; + ++nritems; p = symbol_list_new (symval); p1->next = p; p1 = p; } else /* handle an action. */ { - copy_action (crule, rulelength); + 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 */ @@ -1514,22 +1568,24 @@ readgram (void) crule->ruleprec = symval; t = lex (); } + if (t == tok_guard) { if (!semantic_parser) complain (_("%%guard present but %%semantic_parser not specified")); - copy_guard (crule, rulelength); + parse_guard (crule, rulelength); t = lex (); } - else if (t == tok_left_curly) + + 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); + parse_action (crule, rulelength); action_flag = 1; - xactions++; /* -wjh */ + ++xactions; /* -wjh */ t = lex (); } /* If $$ is being set in default way, report if any type @@ -1547,6 +1603,8 @@ readgram (void) /* Warn if there is no default for $$ 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 (); } @@ -1556,28 +1614,34 @@ 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 @@ -1586,7 +1650,7 @@ readgram (void) static void read_additionnal_code (void) { - char c; + int c; struct obstack el_obstack; obstack_init (&el_obstack); @@ -1595,7 +1659,7 @@ read_additionnal_code (void) { obstack_fgrow2 (&el_obstack, muscle_find ("linef"), lineno, quotearg_style (c_quoting_style, - muscle_find("filename"))); + muscle_find ("filename"))); } while ((c = getc (finput)) != EOF) @@ -1606,244 +1670,87 @@ read_additionnal_code (void) } -/*--------------------------------------------------------------. -| For named tokens, but not literal ones, define the name. The | -| value is the user token number. | -`--------------------------------------------------------------*/ +/*------------------------------------------------------------------. +| Set TOKEN_TRANSLATIONS. Check that no two symbols share the same | +| number. | +`------------------------------------------------------------------*/ static void -output_token_defines (struct obstack *oout) +token_translations_init (void) { - bucket *bp; - char *cp, *symbol; - char c; + int num_256_available_p = TRUE; + int i; - for (bp = firstsymbol; bp; bp = bp->next) + /* Find the highest user token number, and whether 256, the POSIX + preferred user token number for the error token, is used. */ + max_user_token_number = 0; + for (i = 0; i < ntokens; ++i) { - symbol = bp->tag; /* get symbol */ - - if (bp->value >= ntokens) - continue; - if (bp->user_token_number == SALIAS) - continue; - if ('\'' == *symbol) - continue; /* skip literal character */ - if (bp == errtoken) - continue; /* skip error token */ - if ('\"' == *symbol) + symbol_t *this = symbols[i]; + if (this->user_token_number != SUNDEF) { - /* use literal string only if given a symbol with an alias */ - if (bp->alias) - symbol = bp->alias->tag; - else - continue; + if (this->user_token_number > max_user_token_number) + max_user_token_number = this->user_token_number; + if (this->user_token_number == 256) + num_256_available_p = FALSE; } - - /* Don't #define nonliteral tokens whose names contain periods. */ - cp = symbol; - while ((c = *cp++) && c != '.'); - if (c != '\0') - continue; - - obstack_fgrow2 (oout, "# define\t%s\t%d\n", - symbol, bp->user_token_number); - if (semantic_parser) - /* FIXME: This is certainly dead wrong, and should be just as - above. --akim. */ - obstack_fgrow2 (oout, "# define\tT%s\t%d\n", symbol, bp->value); } -} + /* If 256 is not used, assign it to error, to follow POSIX. */ + if (num_256_available_p && errtoken->user_token_number == SUNDEF) + errtoken->user_token_number = 256; -/*------------------------------------------------------------------. -| Set TOKEN_TRANSLATIONS. Check that no two symbols share the same | -| number. | -`------------------------------------------------------------------*/ + /* Set the missing user numbers. */ + if (max_user_token_number < 256) + max_user_token_number = 256; -static void -token_translations_init (void) -{ - bucket *bp = NULL; - int i; + for (i = 0; i < ntokens; ++i) + { + symbol_t *this = symbols[i]; + if (this->user_token_number == SUNDEF) + this->user_token_number = ++max_user_token_number; + if (this->user_token_number > max_user_token_number) + max_user_token_number = this->user_token_number; + } - token_translations = XCALLOC (short, max_user_token_number + 1); + token_translations = XCALLOC (token_number_t, max_user_token_number + 1); /* Initialize all entries for literal tokens to 2, the internal token number for $undefined., which represents all invalid inputs. */ - for (i = 0; i <= max_user_token_number; i++) - token_translations[i] = 2; - - for (bp = firstsymbol; bp; bp = bp->next) - { - /* Non-terminal? */ - if (bp->value >= ntokens) - continue; - /* A token string alias? */ - if (bp->user_token_number == SALIAS) - continue; - - assert (bp->user_token_number != SUNDEF); - - /* A token which translation has already been set? */ - if (token_translations[bp->user_token_number] != 2) - complain (_("tokens %s and %s both assigned number %d"), - tags[token_translations[bp->user_token_number]], - bp->tag, bp->user_token_number); - token_translations[bp->user_token_number] = bp->value; - } + for (i = 0; i < max_user_token_number + 1; i++) + token_translations[i] = undeftoken->number; + symbols_do (symbol_translation, NULL); } -/*------------------------------------------------------------------. -| Assign symbol numbers, and write definition of token names into | -| FDEFINES. Set up vectors TAGS and SPREC of names and precedences | -| of symbols. | -`------------------------------------------------------------------*/ +/*----------------------------------------------------------------. +| Assign symbol numbers, and write definition of token names into | +| FDEFINES. Set up vectors SYMBOL_TABLE, TAGS of symbols. | +`----------------------------------------------------------------*/ static void packsymbols (void) { - bucket *bp = NULL; - int tokno = 1; - int last_user_token_number; - static char DOLLAR[] = "$"; - - tags = XCALLOC (char *, nsyms + 1); - user_toknums = XCALLOC (short, nsyms + 1); - - 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; + symbols = XCALLOC (symbol_t *, nsyms); - for (bp = firstsymbol; bp; bp = bp->next) - { - if (bp->class == nterm_sym) - { - bp->value += ntokens; - } - else if (bp->alias) - { - /* this symbol and its alias are a single token defn. - allocate a tokno, and assign to both check agreement of - ->prec and ->assoc fields and make both the same */ - if (bp->value == 0) - bp->value = bp->alias->value = tokno++; - - if (bp->prec != bp->alias->prec) - { - if (bp->prec != 0 && bp->alias->prec != 0 - && bp->user_token_number == SALIAS) - complain (_("conflicting precedences for %s and %s"), - bp->tag, bp->alias->tag); - if (bp->prec != 0) - bp->alias->prec = bp->prec; - else - bp->prec = bp->alias->prec; - } - - if (bp->assoc != bp->alias->assoc) - { - if (bp->assoc != 0 && bp->alias->assoc != 0 - && bp->user_token_number == SALIAS) - complain (_("conflicting assoc values for %s and %s"), - bp->tag, bp->alias->tag); - if (bp->assoc != 0) - bp->alias->assoc = bp->assoc; - else - bp->assoc = bp->alias->assoc; - } - - if (bp->user_token_number == SALIAS) - continue; /* do not do processing below for SALIASs */ - - } - else /* bp->class == token_sym */ - { - bp->value = tokno++; - } - - if (bp->class == token_sym) - { - if (bp->user_token_number == SUNDEF) - bp->user_token_number = ++last_user_token_number; - if (bp->user_token_number > max_user_token_number) - max_user_token_number = bp->user_token_number; - } - - tags[bp->value] = bp->tag; - user_toknums[bp->value] = bp->user_token_number; - sprec[bp->value] = bp->prec; - sassoc[bp->value] = bp->assoc; - } + symbols_do (symbol_check_alias_consistence, NULL); + symbols_do (symbol_pack, NULL); token_translations_init (); - error_token_number = errtoken->value; - 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; -} - - -/*-----------------------------------. -| Output definition of token names. | -`-----------------------------------*/ - -static void -symbols_output (void) -{ - { - struct obstack tokendefs; - obstack_init (&tokendefs); - output_token_defines (&tokendefs); - obstack_1grow (&tokendefs, 0); - muscle_insert ("tokendef", xstrdup (obstack_finish (&tokendefs))); - obstack_free (&tokendefs, NULL); - } - - if (defines_flag) - { - output_token_defines (&defines_obstack); - - if (!pure_parser) - obstack_fgrow1 (&defines_obstack, "\nextern YYSTYPE %slval;\n", - spec_name_prefix); - if (semantic_parser) - { - int i; - - for (i = ntokens; i < nsyms; i++) - { - /* don't make these for dummy nonterminals made by gensym. */ - if (*tags[i] != '@') - obstack_fgrow2 (&defines_obstack, - "# define\tNT%s\t%d\n", tags[i], i); - } -#if 0 - /* `fdefines' is now a temporary file, so we need to copy its - contents in `done', so we can't close it here. */ - fclose (fdefines); - fdefines = NULL; -#endif - } - } + start_symbol = startval->number; } /*---------------------------------------------------------------. | Convert the rules into the representation using RRHS, RLHS and | -| RITEMS. | +| RITEM. | `---------------------------------------------------------------*/ static void @@ -1853,8 +1760,8 @@ packgram (void) int ruleno; symbol_list *p; - 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; @@ -1862,25 +1769,28 @@ 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; + 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; + rules[ruleno].guard = p->guard; + rules[ruleno].guard_line = p->guard_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; } @@ -1889,19 +1799,18 @@ 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; + assert (itemno == nritems); if (trace_flag) ritem_print (stderr); @@ -1918,22 +1827,6 @@ packgram (void) void reader (void) { - start_flag = 0; - startval = NULL; /* start symbol not specified yet. */ - - nsyms = 1; - nvars = 0; - nrules = 0; - nitems = 0; - - typed = 0; - lastprec = 0; - - semantic_parser = 0; - pure_parser = 0; - - grammar = NULL; - lex_init (); lineno = 1; @@ -1941,38 +1834,71 @@ 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); + + finput = xfopen (infile, "r"); /* Read the declaration section. Copy %{ ... %} groups to TABLE_OBSTACK and FDEFINES file. Also notice any %token, %left, etc. found there. */ read_declarations (); + + /* If the user did not define her EOFTOKEN, do it now. */ + if (!eoftoken) + { + eoftoken = getsym ("$"); + eoftoken->class = token_sym; + eoftoken->number = 0; + /* Value specified by POSIX. */ + eoftoken->user_token_number = 0; + } + /* Read in the grammar, build grammar in list form. Write out guards and actions. */ readgram (); /* Some C code is given at the end of the grammar file. */ read_additionnal_code (); - /* Now we know whether we need the line-number stack. If we do, - write its type into the .tab.h file. - This is no longer need with header skeleton. */ + lex_free (); + xfclose (finput); /* Assign the symbols their symbol numbers. Write #defines for the token symbols into FDEFINES if requested. */ packsymbols (); + /* Convert the grammar into the format described in gram.h. */ packgram (); - /* Output the headers. */ - symbols_output (); + + /* 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 (); }