X-Git-Url: https://git.saurik.com/bison.git/blobdiff_plain/2f1afb737f4b8e0a4ea3f5d02e25858cc709e466..3e6656f9ab66ae7e629f8e5d5b41810bf070f09d:/src/reader.c diff --git a/src/reader.c b/src/reader.c index 63d993f9..64601e34 100644 --- a/src/reader.c +++ b/src/reader.c @@ -22,12 +22,10 @@ #include "system.h" #include "quotearg.h" -#include "quote.h" #include "getargs.h" #include "files.h" #include "symtab.h" -#include "options.h" -#include "lex.h" +#include "symlist.h" #include "gram.h" #include "complain.h" #include "output.h" @@ -35,1447 +33,363 @@ #include "conflicts.h" #include "muscle_tab.h" -typedef struct symbol_list -{ - struct symbol_list *next; - symbol_t *sym; - int line; - - /* The action is attached to the LHS of a rule. */ - const char *action; - int action_line; - - symbol_t *ruleprec; -} symbol_list; - -int lineno; -static symbol_list *grammar = NULL; +static symbol_list_t *grammar = NULL; static int start_flag = 0; - -/* Nonzero if components of semantic values are used, implying - they must be unions. */ -static int value_components_used; +merger_list *merge_functions; /* Nonzero if %union has been seen. */ -static int typed = 0; - -/* Incremented for each %left, %right or %nonassoc seen */ -static int lastprec = 0; - -static symbol_list * -symbol_list_new (symbol_t *sym) -{ - symbol_list *res = XMALLOC (symbol_list, 1); - res->next = NULL; - res->sym = sym; - res->line = lineno; - res->action = NULL; - res->action_line = 0; - res->ruleprec = NULL; - return res; -} - -/*===================\ -| Low level lexing. | -\===================*/ +int typed = 0; + +/*-----------------------. +| Set the start symbol. | +`-----------------------*/ -static void -skip_to_char (int target) +void +grammar_start_symbol_set (symbol_t *s, location_t l) { - int c; - if (target == '\n') - complain (_(" Skipping to next \\n")); + if (start_flag) + complain_at (l, _("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 = 1; + startsymbol = s; + startsymbol_location = l; } - - ungetc (c, stream); - - return sign * n; } - -/*--------------------------------------------------------------. -| Get the data type (alternative in the union) of the value for | -| symbol N in rule RULE. | -`--------------------------------------------------------------*/ - -static char * -get_type_name (int n, symbol_list *rule) -{ - int i; - symbol_list *rp; - if (n < 0) - { - complain (_("invalid $ value")); - return NULL; - } - - rp = rule; - i = 0; - while (i < n) - { - rp = rp->next; - if (rp == NULL || rp->sym == NULL) - { - complain (_("invalid $ value")); - return NULL; - } - ++i; - } - - return rp->sym->type_name; -} - -/*------------------------------------------------------------------. -| Copy the character C to OOUT, and insert quadigraphs when needed. | -`------------------------------------------------------------------*/ +/*----------------------------------------------------------------. +| There are two prologues: one before %union, one after. Augment | +| the current one. | +`----------------------------------------------------------------*/ -static inline void -copy_character (struct obstack *oout, int c) +void +prologue_augment (const char *prologue, location_t location) { - switch (c) - { - case '[': - obstack_sgrow (oout, "@<:@"); - break; - - case ']': - obstack_sgrow (oout, "@:>@"); - break; - - default: - obstack_1grow (oout, c); - } + struct obstack *oout = + !typed ? &pre_prologue_obstack : &post_prologue_obstack; + + obstack_fgrow1 (oout, "]b4_syncline([[%d]], [[", + location.first_line); + MUSCLE_OBSTACK_SGROW (oout, quotearg_style (c_quoting_style, location.file)); + obstack_sgrow (oout, "]])[\n"); + obstack_sgrow (oout, prologue); } -/*------------------------------------------------------------. -| 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; - } - copy_character (oout, c); - if (c == '\\') - { - c = getc (fin); - if (c == EOF) - fatal (_("unterminated string at end of file")); - copy_character (oout, c); +/*----------------------. +| Handle the epilogue. | +`----------------------*/ - if (c == '\n') - ++lineno; - } - - c = getc (fin); - } - - if (store) - obstack_1grow (oout, c); -} - -/* FIXME. */ - -static inline void -copy_string (FILE *fin, struct obstack *oout, int match) +void +epilogue_set (const char *epilogue, location_t location) { - copy_string2 (fin, oout, match, 1); + obstack_fgrow1 (&muscle_obstack, "]b4_syncline([[%d]], [[", + location.first_line); + MUSCLE_OBSTACK_SGROW (&muscle_obstack, + quotearg_style (c_quoting_style, location.file)); + obstack_sgrow (&muscle_obstack, "]])[\n"); + obstack_sgrow (&muscle_obstack, epilogue); + obstack_1grow (&muscle_obstack, 0); + muscle_insert ("epilogue", obstack_finish (&muscle_obstack)); } -/* FIXME. */ - -static inline void -copy_identifier (FILE *fin, struct obstack *oout) -{ - int c; - - while (isalnum (c = getc (fin)) || c == '_') - obstack_1grow (oout, c); - - ungetc (c, fin); -} + -/*------------------------------------------------------------------. -| Dump the wannabee comment from IN to OOUT. In fact we just saw a | -| `/', which might or might not be a comment. In any case, copy | -| what we saw. | -`------------------------------------------------------------------*/ +/*-------------------------------------------------------------------. +| Return the merger index for a merging function named NAME, whose | +| arguments have type TYPE. Records the function, if new, in | +| merger_list. | +`-------------------------------------------------------------------*/ -static inline void -copy_comment (FILE *fin, struct obstack *oout) +static int +get_merge_function (const char* name, const char* type, + location_t loc) { - int cplus_comment; - int ended; - int c; + merger_list *syms; + merger_list head; + int n; - /* We read a `/', output it. */ - obstack_1grow (oout, '/'); + if (! glr_parser) + return 0; - switch ((c = getc (fin))) - { - case '/': - cplus_comment = 1; - break; - case '*': - cplus_comment = 0; - break; - default: - ungetc (c, fin); - return; - } + if (type == NULL) + type = ""; - obstack_1grow (oout, c); - c = getc (fin); - - ended = 0; - while (!ended) - { - if (!cplus_comment && c == '*') - { - while (c == '*') - { - obstack_1grow (oout, c); - c = getc (fin); - } - - if (c == '/') - { - obstack_1grow (oout, c); - ended = 1; - } - } - else if (c == '\n') - { - ++lineno; - obstack_1grow (oout, c); - if (cplus_comment) - ended = 1; - else - c = getc (fin); - } - else if (c == EOF) - fatal (_("unterminated comment")); - else - { - copy_character (oout, c); - c = getc (fin); - } - } + head.next = merge_functions; + for (syms = &head, n = 1; syms->next != NULL; syms = syms->next, n += 1) + if (strcmp (name, syms->next->name) == 0) + break; + if (syms->next == NULL) + { + syms->next = XMALLOC (merger_list, 1); + syms->next->name = xstrdup (name); + syms->next->type = xstrdup (type); + syms->next->next = NULL; + merge_functions = head.next; + } + else if (strcmp (type, syms->next->type) != 0) + warn_at (loc, _("result type clash on merge function %s: <%s> != <%s>"), + name, type, syms->next->type); + return n; } +/*--------------------------------------. +| Free all merge-function definitions. | +`--------------------------------------*/ -/*-------------------------------------------------------------------. -| FIN is pointing to a location (i.e., a `@'). Output to OOUT a | -| reference to this location. RULE_LENGTH is the number of values in | -| the current rule so far, which says where to find `$0' with | -| respect to the top of the stack. | -`-------------------------------------------------------------------*/ - -static inline void -copy_at (FILE *fin, struct obstack *oout, int rule_length) +void +free_merger_functions (void) { - int c = getc (fin); - locations_flag = 1; - - if (c == '$') - { - obstack_sgrow (oout, "]b4_lhs_location["); - } - else if (isdigit (c) || c == '-') + merger_list *L0; + if (! glr_parser) + return; + L0 = merge_functions; + while (L0 != NULL) { - int n; - - ungetc (c, fin); - n = read_signed_integer (fin); - if (n > rule_length) - complain (_("invalid value: %s%d"), "@", n); - else - obstack_fgrow2 (oout, "]b4_rhs_location([%d], [%d])[", - rule_length, n); - } - else - { - char buf[] = "@c"; - buf[1] = c; - complain (_("%s is invalid"), quote (buf)); + merger_list *L1 = L0->next; + free (L0); + L0 = L1; } } - -/*------------------------------------------------------------------. -| FIN is pointing to a wannabee semantic value (i.e., a `$'). | -| | -| Possible inputs: $[]($|integer) | -| | -| Output to OOUT a reference to this semantic value. RULE_LENGTH is | -| the number of values in the current rule so far, which says where | -| to find `$0' with respect to the top of the stack. | -`------------------------------------------------------------------*/ - -static inline void -copy_dollar (FILE *fin, struct obstack *oout, - symbol_list *rule, int rule_length) -{ - int c = getc (fin); - const char *type_name = NULL; - - /* Get the type name if explicit. */ - if (c == '<') - { - read_type_name (fin); - type_name = token_buffer; - value_components_used = 1; - c = getc (fin); - } - - if (c == '$') - { - if (!type_name) - type_name = get_type_name (0, rule); - if (!type_name && typed) - complain (_("$$ of `%s' has no declared type"), - rule->sym->tag); - if (!type_name) - type_name = ""; - obstack_fgrow1 (oout, - "]b4_lhs_value([%s])[", type_name); - } - else if (isdigit (c) || c == '-') - { - int n; - ungetc (c, fin); - n = read_signed_integer (fin); - - if (n > rule_length) - complain (_("invalid value: %s%d"), "$", n); - else - { - if (!type_name && n > 0) - type_name = get_type_name (n, rule); - if (!type_name && typed) - complain (_("$%d of `%s' has no declared type"), - n, rule->sym->tag); - if (!type_name) - type_name = ""; - obstack_fgrow3 (oout, "]b4_rhs_value([%d], [%d], [%s])[", - rule_length, n, type_name); - } - } - else - { - char buf[] = "$c"; - buf[1] = c; - complain (_("%s is invalid"), quote (buf)); - } -} /*-------------------------------------------------------------------. -| Copy the contents of a `%{ ... %}' into the definitions file. The | -| `%{' has already been read. Return after reading the `%}'. | +| Parse the input grammar into a one symbol_list_t structure. Each | +| rule is represented by a sequence of symbols: the left hand side | +| followed by the contents of the right hand side, followed by a | +| null pointer instead of a symbol to terminate the rule. The next | +| symbol is the lhs of the following rule. | +| | +| All actions are copied out, labelled by the rule number they apply | +| to. | +| | +| Bison used to allow some %directives in the rules sections, but | +| this is no longer consider appropriate: (i) the documented grammar | +| doesn't claim it, (ii), it would promote bad style, (iii), error | +| recovery for %directives consists in skipping the junk until a `%' | +| is seen and helrp synchronizing. This scheme is definitely wrong | +| in the rules section. | `-------------------------------------------------------------------*/ -static void -copy_definition (struct obstack *oout) -{ - int c; - /* -1 while reading a character if prev char was %. */ - int after_percent; +/* The (currently) last symbol of GRAMMAR. */ +symbol_list_t *grammar_end = NULL; - if (!no_lines_flag) - { - obstack_fgrow2 (oout, muscle_find ("linef"), - lineno, quotearg_style (c_quoting_style, - muscle_find ("filename"))); - } - - after_percent = 0; - - c = getc (finput); - - for (;;) - { - switch (c) - { - case '\n': - obstack_1grow (oout, c); - ++lineno; - break; - - case '%': - after_percent = -1; - break; - - case '\'': - case '"': - copy_string (finput, oout, c); - break; - - case '/': - copy_comment (finput, oout); - break; - - case EOF: - fatal ("%s", _("unterminated `%{' definition")); - - default: - copy_character (oout, c); - } - - c = getc (finput); - - if (after_percent) - { - if (c == '}') - return; - obstack_1grow (oout, '%'); - } - after_percent = 0; - } -} - - -/*-------------------------------------------------------------------. -| Parse what comes after %token or %nterm. For %token, WHAT_IS is | -| token_sym and WHAT_IS_NOT is nterm_sym. For %nterm, the arguments | -| are reversed. | -`-------------------------------------------------------------------*/ - -static void -parse_token_decl (symbol_class what_is, symbol_class what_is_not) +/* Append S to the GRAMMAR. */ +void +grammar_symbol_append (symbol_t *symbol, location_t location) { - token_t token = tok_undef; - char *typename = NULL; - - /* The symbol being defined. */ - symbol_t *symbol = NULL; - - /* 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); + symbol_list_t *p = symbol_list_new (symbol, location); - token = lex (); - if (token == tok_comma) - { - symbol = NULL; - continue; - } - if (token == tok_typename) - { - typename = xstrdup (token_buffer); - value_components_used = 1; - symbol = NULL; - } - else if (token == tok_identifier && *symval->tag == '\"' && symbol) - { - symbol_make_alias (symbol, symval, typename); - symbol = NULL; - } - else if (token == tok_identifier) - { - int oldclass = symval->class; - symbol = symval; - - if (symbol->class == what_is_not) - complain (_("symbol %s redefined"), symbol->tag); - symbol->class = what_is; - if (what_is == nterm_sym && oldclass != nterm_sym) - symbol->number = nvars++; - if (what_is == token_sym && symbol->number == NUMBER_UNDEFINED) - symbol->number = ntokens++; - - if (typename) - { - if (symbol->type_name == NULL) - symbol->type_name = typename; - else if (strcmp (typename, symbol->type_name) != 0) - complain (_("type redeclaration for %s"), symbol->tag); - } - } - else if (symbol && token == tok_number) - { - symbol->user_token_number = numval; - /* User defined EOF token? */ - if (numval == 0) - { - eoftoken = symbol; - eoftoken->number = 0; - /* It is always mapped to 0, so it was already counted in - NTOKENS. */ - --ntokens; - } - } - else - { - complain (_("`%s' is invalid in %s"), - token_buffer, - (what_is == token_sym) ? "%token" : "%nterm"); - skip_to_char ('%'); - } - } + if (grammar_end) + grammar_end->next = p; + else + grammar = p; + grammar_end = p; } +/* 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). */ +symbol_list_t *current_rule = NULL; +symbol_list_t *previous_rule_end = NULL; -/*------------------------------. -| Parse what comes after %start | -`------------------------------*/ - -static void -parse_start_decl (void) -{ - if (start_flag) - complain (_("multiple %s declarations"), "%start"); - if (lex () != tok_identifier) - complain (_("invalid %s declaration"), "%start"); - else - { - start_flag = 1; - startsymbol = symval; - } -} -/*-----------------------------------------------------------. -| read in a %type declaration and record its information for | -| get_type_name to access | -`-----------------------------------------------------------*/ +/*----------------------------------------------. +| Create a new rule for LHS in to the GRAMMAR. | +`----------------------------------------------*/ -static void -parse_type_decl (void) +void +grammar_rule_begin (symbol_t *lhs, location_t location) { - char *name; - - if (lex () != tok_typename) + if (!start_flag) { - complain ("%s", _("%type declaration has no ")); - skip_to_char ('%'); - return; + startsymbol = lhs; + startsymbol_location = location; + start_flag = 1; } - name = xstrdup (token_buffer); - - for (;;) - { - token_t t; - int tmp_char = ungetc (skip_white_space (), finput); - - if (tmp_char == '%') - return; - if (tmp_char == EOF) - fatal (_("Premature EOF after %s"), token_buffer); + /* Start a new rule and record its lhs. */ + ++nrules; + ++nritems; - 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 ('%'); - } - } -} + previous_rule_end = grammar_end; + grammar_symbol_append (lhs, location); + current_rule = grammar_end; + /* Mark the rule's lhs as a nonterminal if not already so. */ - -/*----------------------------------------------------------------. -| Read in a %left, %right or %nonassoc declaration and record its | -| information. | -`----------------------------------------------------------------*/ - -static void -parse_assoc_decl (associativity assoc) -{ - char *name = NULL; - int prev = 0; - - /* Assign a new precedence level, never 0. */ - ++lastprec; - - for (;;) + if (lhs->class == unknown_sym) { - token_t t; - int tmp_char = ungetc (skip_white_space (), finput); - - if (tmp_char == '%') - return; - if (tmp_char == EOF) - fatal (_("Premature EOF after %s"), token_buffer); - - t = lex (); - - switch (t) - { - case tok_typename: - name = xstrdup (token_buffer); - break; - - case tok_comma: - break; - - case tok_identifier: - if (symval->prec != 0) - complain (_("redefining precedence of %s"), symval->tag); - symval->prec = lastprec; - symval->assoc = assoc; - if (symval->class == nterm_sym) - complain (_("symbol %s redefined"), symval->tag); - if (symval->number == NUMBER_UNDEFINED) - { - symval->number = ntokens++; - symval->class = token_sym; - } - if (name) - { /* record the type, if one is specified */ - if (symval->type_name == NULL) - symval->type_name = name; - else if (strcmp (name, symval->type_name) != 0) - complain (_("type redeclaration for %s"), symval->tag); - } - break; - - case tok_number: - if (prev == tok_identifier) - { - symval->user_token_number = numval; - } - else - { - complain - (_("invalid text (%s) - number should be after identifier"), - token_buffer); - skip_to_char ('%'); - } - break; - - case tok_semicolon: - return; - - default: - complain (_("unexpected item: %s"), token_buffer); - skip_to_char ('%'); - } - - prev = t; + lhs->class = nterm_sym; + lhs->number = nvars; + ++nvars; } + else if (lhs->class == token_sym) + complain_at (location, _("rule given for %s, which is a token"), lhs->tag); } - - -/*--------------------------------------------------------------. -| 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. | -`--------------------------------------------------------------*/ +/* Check that the last rule (CURRENT_RULE) is properly defined. For + instance, there should be no type clash on the default action. */ static void -parse_union_decl (void) +grammar_current_rule_check (void) { - int c; - int count = 0; - bool done = FALSE; - struct obstack union_obstack; - if (typed) - complain (_("multiple %s declarations"), "%union"); - - typed = 1; - - MUSCLE_INSERT_INT ("stype_line", lineno); - obstack_init (&union_obstack); - obstack_sgrow (&union_obstack, "union"); - - while (!done) - { - c = xgetc (finput); - - /* If C contains '/', it is output by copy_comment (). */ - if (c != '/') - obstack_1grow (&union_obstack, c); - - switch (c) - { - case '\n': - ++lineno; - break; - - case '/': - copy_comment (finput, &union_obstack); - break; - - case '{': - ++count; - break; - - case '}': - /* FIXME: Errr. How could this happen???. --akim */ - if (count == 0) - complain (_("unmatched %s"), "`}'"); - count--; - if (!count) - done = TRUE; - break; - } - } - - /* JF don't choke on trailing semi */ - c = skip_white_space (); - if (c != ';') - ungetc (c, finput); - obstack_1grow (&union_obstack, 0); - muscle_insert ("stype", obstack_finish (&union_obstack)); + symbol_t *lhs = current_rule->sym; + char const *lhs_type = lhs->type_name; + symbol_t *first_rhs = current_rule->next->sym; + + /* If there is an action, then there is nothing we can do: the user + is allowed to shoot herself in the foot. */ + if (current_rule->action) + return; + + /* Don't worry about the default action if $$ is untyped, since $$'s + value can't be used. */ + if (! lhs_type) + return; + + /* If $$ is being set in default way, report if any type mismatch. */ + if (first_rhs) + { + const char *rhs_type = first_rhs->type_name ? first_rhs->type_name : ""; + if (strcmp (lhs_type, rhs_type)) + complain_at (current_rule->location, + _("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 (current_rule->location, + _("empty rule for typed nonterminal, and no action")); } -/*-------------------------------------------------------. -| 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_rule_end (location_t location) { - 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 = location; + grammar_current_rule_check (); } /*-------------------------------------------------------------------. -| 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 | -| 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 | -| 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; - symbol_t *symbol; - char *typename = 0; - int usrtoknum = USER_NUMBER_UNDEFINED; - - 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 = USER_NUMBER_ALIAS; - 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 */ - - if (token != tok_identifier || *symval->tag != '\"') - { - complain (_("expected string constant instead of %s"), token_buffer); - skip_to_char ('%'); - return; - } - symval->class = token_sym; - symval->type_name = typename; - symval->user_token_number = usrtoknum; - - symval->alias = symbol; - symbol->alias = symval; - - /* symbol and symval combined are only one symbol. */ - nsyms--; -} - - -static void -parse_muscle_decl (void) +void +grammar_midrule_action (void) { - int ch = ungetc (skip_white_space (), finput); - char *muscle_key; - char *muscle_value; + /* 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_t dummy_location = current_rule->action_location; + symbol_t *dummy = dummy_symbol_get (dummy_location); + symbol_list_t *midrule = symbol_list_new (dummy, dummy_location); + + /* 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; + midrule->action = current_rule->action; + midrule->action_location = dummy_location; + current_rule->action = NULL; + + if (previous_rule_end) + previous_rule_end->next = midrule; + else + grammar = midrule; - /* 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); + /* End the dummy's rule. */ + previous_rule_end = symbol_list_new (NULL, dummy_location); + previous_rule_end->next = current_rule; - /* 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); + midrule->next = previous_rule_end; - /* Store the (key, value) pair in the environment. */ - muscle_insert (muscle_key, muscle_value); + /* Insert the dummy nonterminal replacing the midrule action into + the current rule. */ + grammar_current_rule_symbol_append (dummy, dummy_location); } +/* Set the precedence symbol of the current rule to PRECSYM. */ - -/*---------------------------------. -| Parse a double quoted parameter. | -`---------------------------------*/ - -static const char * -parse_dquoted_param (const char *from) +void +grammar_current_rule_prec_set (symbol_t *precsym, location_t location) { - struct obstack param_obstack; - const char *param = NULL; - int c; - - obstack_init (¶m_obstack); - c = skip_white_space (); - - if (c != '"') - { - complain (_("invalid %s declaration"), from); - ungetc (c, finput); - skip_to_char ('%'); - return NULL; - } - - while ((c = literalchar ()) != '"') - obstack_1grow (¶m_obstack, c); - - obstack_1grow (¶m_obstack, '\0'); - param = obstack_finish (¶m_obstack); - - if (c != '"' || strlen (param) == 0) - { - complain (_("invalid %s declaration"), from); - if (c != '"') - ungetc (c, finput); - skip_to_char ('%'); - return NULL; - } - - return param; + if (current_rule->ruleprec) + complain_at (location, _("only one %s allowed per rule"), "%prec"); + current_rule->ruleprec = precsym; } -/*----------------------------------. -| Parse what comes after %skeleton. | -`----------------------------------*/ +/* Attach dynamic precedence DPREC to the current rule. */ -static void -parse_skel_decl (void) +void +grammar_current_rule_dprec_set (int dprec, location_t location) { - skeleton = parse_dquoted_param ("%skeleton"); + if (! glr_parser) + warn_at (location, _("%s affects only GLR parsers"), "%dprec"); + if (dprec <= 0) + complain_at (location, + _("%s must be followed by positive number"), "%dprec"); + else if (current_rule->dprec != 0) + complain_at (location, _("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 PRE_PROLOGUE_OBSTACK or POST_PROLOGUE_OBSTACK. | -`----------------------------------------------------------------*/ - -static void -read_declarations (void) -{ - for (;;) - { - int c = skip_white_space (); +/* Attach a merge function NAME with argument type TYPE to current + rule. */ - if (c == '%') - { - token_t tok = parse_percent_token (); - - switch (tok) - { - case tok_two_percents: - return; - - case tok_percent_left_curly: - if (!typed) - copy_definition (&pre_prologue_obstack); - else - copy_definition (&post_prologue_obstack); - break; - - case tok_token: - parse_token_decl (token_sym, nterm_sym); - break; - - case tok_nterm: - parse_token_decl (nterm_sym, token_sym); - break; - - case tok_type: - parse_type_decl (); - break; - - case tok_start: - parse_start_decl (); - break; - - case tok_union: - parse_union_decl (); - break; - - case tok_expect: - parse_expect_decl (); - break; - - case tok_thong: - parse_thong_decl (); - break; - - case tok_left: - parse_assoc_decl (left_assoc); - break; - - case tok_right: - parse_assoc_decl (right_assoc); - break; - - case tok_nonassoc: - parse_assoc_decl (non_assoc); - break; - - case tok_define: - parse_muscle_decl (); - break; - - case tok_skel: - parse_skel_decl (); - break; - - case tok_noop: - break; - - case tok_stropt: - case tok_intopt: - case tok_obsolete: - assert (0); - break; - - case tok_illegal: - default: - complain (_("unrecognized: %s"), token_buffer); - skip_to_char ('%'); - } - } - else if (c == EOF) - fatal (_("no input grammar")); - 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 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_action (symbol_list *rule, int rule_length) +void +grammar_current_rule_merge_set (const char* name, location_t location) { - int count = 1; - rule->action_line = lineno; - while (count > 0) - { - int c; - while ((c = getc (finput)) != '}') - switch (c) - { - case '\n': - copy_character (&action_obstack, c); - ++lineno; - break; - - case '{': - copy_character (&action_obstack, c); - ++count; - break; - - case '\'': - case '"': - copy_string (finput, &action_obstack, c); - break; - - case '/': - copy_comment (finput, &action_obstack); - break; - - case '$': - copy_dollar (finput, &action_obstack, rule, rule_length); - break; - - case '@': - copy_at (finput, &action_obstack, rule_length); - break; - - case EOF: - fatal (_("unmatched %s"), "`{'"); - - default: - copy_character (&action_obstack, c); - } - - /* Above loop exits when C is '}'. */ - if (--count) - copy_character (&action_obstack, c); - } - - obstack_1grow (&action_obstack, '\0'); - rule->action = obstack_finish (&action_obstack); + if (! glr_parser) + warn_at (location, _("%s affects only GLR parsers"), "%merge"); + if (current_rule->merger != 0) + complain_at (location, _("only one %s allowed per rule"), "%merge"); + current_rule->merger = + get_merge_function (name, current_rule->sym->type_name, location); } - - -/*-------------------------------------------------------------------. -| Generate a dummy symbol, a nonterminal, whose name cannot conflict | -| with the user's names. | -`-------------------------------------------------------------------*/ +/* Attach a SYMBOL to the current rule. If needed, move the previous + action as a mid-rule action. */ -static symbol_t * -gensym (void) +void +grammar_current_rule_symbol_append (symbol_t *symbol, location_t location) { - /* Incremented for each generated symbol */ - static int gensym_count = 0; - static char buf[256]; - - symbol_t *sym; - - sprintf (buf, "@%d", ++gensym_count); - token_buffer = buf; - sym = getsym (token_buffer); - sym->class = nterm_sym; - sym->number = nvars++; - return sym; + if (current_rule->action) + grammar_midrule_action (); + ++nritems; + grammar_symbol_append (symbol, location); } - -/*-------------------------------------------------------------------. -| Parse the input grammar into a one symbol_list structure. Each | -| rule is represented by a sequence of symbols: the left hand side | -| followed by the contents of the right hand side, followed by a | -| null pointer instead of a symbol to terminate the rule. The next | -| symbol is the lhs of the following rule. | -| | -| All actions are copied out, labelled by the rule number they apply | -| to. | -| | -| Bison used to allow some %directives in the rules sections, but | -| this is no longer consider appropriate: (i) the documented grammar | -| doesn't claim it, (ii), it would promote bad style, (iii), error | -| recovery for %directives consists in skipping the junk until a `%' | -| is seen and helrp synchronizing. This scheme is definitely wrong | -| in the rules section. | -`-------------------------------------------------------------------*/ - -static void -readgram (void) -{ - token_t t; - symbol_t *lhs = NULL; - symbol_list *p = NULL; - symbol_list *p1 = NULL; - - /* Points to first symbol_list of current rule. its symbol is the - lhs of the rule. */ - symbol_list *crule = NULL; - /* Points to the symbol_list preceding crule. */ - symbol_list *crule1 = NULL; - - t = lex (); - - while (t != tok_two_percents && t != tok_eof) - if (t == tok_identifier || t == tok_bar) - { - int action_flag = 0; - /* Number of symbols in rhs of this rule so far */ - int rulelength = 0; - int xactions = 0; /* JF for error checking */ - symbol_t *first_rhs = 0; - - if (t == tok_identifier) - { - lhs = symval; - - if (!start_flag) - { - startsymbol = lhs; - start_flag = 1; - } - - t = lex (); - if (t != tok_colon) - { - complain (_("ill-formed rule: initial symbol not followed by colon")); - unlex (t); - } - } - - if (nrules == 0 && t == tok_bar) - { - complain (_("grammar starts with vertical bar")); - lhs = symval; /* BOGUS: use a random symval */ - } - /* start a new rule and record its lhs. */ - - ++nrules; - ++nritems; - - p = symbol_list_new (lhs); - - crule1 = p1; - if (p1) - p1->next = p; - else - grammar = p; - - p1 = p; - crule = p; - - /* mark the rule's lhs as a nonterminal if not already so. */ - - if (lhs->class == unknown_sym) - { - lhs->class = nterm_sym; - lhs->number = nvars; - ++nvars; - } - else if (lhs->class == token_sym) - complain (_("rule given for %s, which is a token"), lhs->tag); - - /* read the rhs of the rule. */ - - for (;;) - { - t = lex (); - if (t == tok_prec) - { - t = lex (); - crule->ruleprec = symval; - t = lex (); - } - - if (!(t == tok_identifier || t == tok_left_curly)) - break; - - /* If next token is an identifier, see if a colon follows it. - If one does, exit this rule now. */ - if (t == tok_identifier) - { - symbol_t *ssave; - token_t t1; - - ssave = symval; - t1 = lex (); - unlex (t1); - symval = ssave; - if (t1 == tok_colon) - { - warn (_("previous rule lacks an ending `;'")); - break; - } - - if (!first_rhs) /* JF */ - first_rhs = symval; - /* Not followed by colon => - process as part of this rule's rhs. */ - } - - /* If we just passed an action, that action was in the middle - of a rule, so make a dummy rule to reduce it to a - non-terminal. */ - if (action_flag) - { - /* Since the action was written out with this rule's - number, we must give the new rule this number by - inserting the new rule before it. */ - - /* Make a dummy nonterminal, a gensym. */ - symbol_t *sdummy = gensym (); - - /* Make a new rule, whose body is empty, before the - current one, so that the action just read can - belong to it. */ - ++nrules; - ++nritems; - p = symbol_list_new (sdummy); - /* Attach its lineno to that of the host rule. */ - p->line = crule->line; - /* Move the action from the host rule to this one. */ - p->action = crule->action; - p->action_line = crule->action_line; - crule->action = NULL; - - if (crule1) - crule1->next = p; - else - grammar = p; - /* End of the rule. */ - crule1 = symbol_list_new (NULL); - crule1->next = crule; - - p->next = crule1; - - /* Insert the dummy generated by that rule into this - rule. */ - ++nritems; - p = symbol_list_new (sdummy); - p1->next = p; - p1 = p; - - action_flag = 0; - } - - if (t == tok_identifier) - { - ++nritems; - p = symbol_list_new (symval); - p1->next = p; - p1 = p; - } - else /* handle an action. */ - { - parse_action (crule, rulelength); - action_flag = 1; - ++xactions; /* JF */ - } - ++rulelength; - } /* end of read rhs of rule */ - - /* Put an empty link in the list to mark the end of this rule */ - p = symbol_list_new (NULL); - p1->next = p; - p1 = p; - - if (t == tok_prec) - { - complain (_("two @prec's in a row")); - t = lex (); - crule->ruleprec = symval; - t = lex (); - } - - if (t == tok_left_curly) - { - /* This case never occurs -wjh */ - if (action_flag) - complain (_("two actions at end of one rule")); - parse_action (crule, rulelength); - action_flag = 1; - ++xactions; /* -wjh */ - t = lex (); - } - /* If $$ is being set in default way, report if any type - mismatch. */ - else if (!xactions - && first_rhs && lhs->type_name != first_rhs->type_name) - { - if (lhs->type_name == 0 - || first_rhs->type_name == 0 - || strcmp (lhs->type_name, first_rhs->type_name)) - complain (_("type clash (`%s' `%s') on default action"), - lhs->type_name ? lhs->type_name : "", - first_rhs->type_name ? first_rhs->type_name : ""); - } - /* Warn if there is no default for $$ but we need one. */ - else if (!xactions && !first_rhs && lhs->type_name != 0) - complain (_("empty rule for typed nonterminal, and no action")); - if (t == tok_two_percents || t == tok_eof) - warn (_("previous rule lacks an ending `;'")); - if (t == tok_semicolon) - t = lex (); - } - else - { - complain (_("invalid input: %s"), quote (token_buffer)); - t = lex (); - } - - /* grammar has been read. Do some checking */ - if (nrules == 0) - fatal (_("no rules in the input grammar")); - - /* Report any undefined symbols and consider them nonterminals. */ - symbols_check_defined (); - - /* 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 (startsymbol); - 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; - - if (nsyms > SHRT_MAX) - fatal (_("too many symbols (tokens plus nonterminals); maximum %d"), - SHRT_MAX); - - assert (nsyms == ntokens + nvars); -} +/* Attach an ACTION to the current rule. If needed, move the previous + action as a mid-rule action. */ -/* 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) +void +grammar_current_rule_action_append (const char *action, location_t location) { - int c; - struct obstack el_obstack; - - obstack_init (&el_obstack); - - if (!no_lines_flag) - { - obstack_fgrow2 (&el_obstack, muscle_find ("linef"), - lineno, quotearg_style (c_quoting_style, - muscle_find ("filename"))); - } - - while ((c = getc (finput)) != EOF) - copy_character (&el_obstack, c); - - obstack_1grow (&el_obstack, 0); - muscle_insert ("epilogue", obstack_finish (&el_obstack)); + if (current_rule->action) + grammar_midrule_action (); + current_rule->action = action; + current_rule->action_location = location; } @@ -1487,17 +401,13 @@ read_additionnal_code (void) static void packgram (void) { - unsigned int itemno; - int ruleno; - symbol_list *p; + unsigned int itemno = 0; + rule_number_t ruleno = 0; + symbol_list_t *p = grammar; ritem = XCALLOC (item_number_t, nritems); - rules = XCALLOC (rule_t, nrules) - 1; - - itemno = 0; - ruleno = 1; + rules = XCALLOC (rule_t, nrules); - p = grammar; while (p) { symbol_t *ruleprec = p->ruleprec; @@ -1505,10 +415,12 @@ packgram (void) 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].location = p->location; + rules[ruleno].useful = true; rules[ruleno].action = p->action; - rules[ruleno].action_line = p->action_line; + rules[ruleno].action_location = p->action_location; + rules[ruleno].dprec = p->dprec; + rules[ruleno].merger = p->merger; p = p->next; while (p && p->sym) @@ -1531,7 +443,7 @@ packgram (void) rules[ruleno].precsym = ruleprec; rules[ruleno].prec = ruleprec; } - ritem[itemno++] = -ruleno; + ritem[itemno++] = rule_number_as_item_number (ruleno); ++ruleno; if (p) @@ -1540,7 +452,7 @@ packgram (void) assert (itemno == nritems); - if (trace_flag) + if (trace_flag & trace_sets) ritem_print (stderr); } @@ -1554,61 +466,83 @@ packgram (void) void reader (void) { - lex_init (); - lineno = 1; - - /* Initialize the muscle obstack. */ - obstack_init (&muscle_obstack); + gram_control_t gram_control; /* Initialize the symbol table. */ symbols_new (); - /* Construct the axiom symbol. */ - axiom = getsym ("$axiom"); - axiom->class = nterm_sym; - axiom->number = nvars++; + /* Construct the accept symbol. */ + accept = symbol_get ("$accept", empty_location); + accept->class = nterm_sym; + accept->number = nvars++; /* Construct the error token */ - errtoken = getsym ("error"); + errtoken = symbol_get ("error", empty_location); errtoken->class = token_sym; errtoken->number = ntokens++; /* Construct a token that represents all undefined literal tokens. It is always token number 2. */ - undeftoken = getsym ("$undefined."); + undeftoken = symbol_get ("$undefined", empty_location); undeftoken->class = token_sym; undeftoken->number = ntokens++; /* Initialize the obstacks. */ - obstack_init (&action_obstack); - obstack_init (&output_obstack); obstack_init (&pre_prologue_obstack); obstack_init (&post_prologue_obstack); finput = xfopen (infile, "r"); + gram_in = finput; + + gram__flex_debug = trace_flag & trace_scan; + gram_debug = trace_flag & trace_parse; + scanner_initialize (); + gram_parse (&gram_control); - /* Read the declaration section. Copy %{ ... %} groups to - TABLE_OBSTACK and FDEFINES file. Also notice any %token, %left, - etc. found there. */ - read_declarations (); + /* If something went wrong during the parsing, don't try to + continue. */ + if (complaint_issued) + return; - /* If the user did not define her EOFTOKEN, do it now. */ - if (!eoftoken) + /* Grammar has been read. Do some checking */ + if (nrules == 0) + fatal (_("no rules in the input grammar")); + + /* Report any undefined symbols and consider them nonterminals. */ + symbols_check_defined (); + + /* If the user did not define her ENDTOKEN, do it now. */ + if (!endtoken) { - eoftoken = getsym ("$"); - eoftoken->class = token_sym; - eoftoken->number = 0; + endtoken = symbol_get ("$end", empty_location); + endtoken->class = token_sym; + endtoken->number = 0; /* Value specified by POSIX. */ - eoftoken->user_token_number = 0; + endtoken->user_token_number = 0; } - /* Read in the grammar, build grammar in list form. Write out - actions. */ - readgram (); - /* Some C code is given at the end of the grammar file. */ - read_additionnal_code (); + /* Insert the initial rule, which line is that of the first rule + (not that of the start symbol): + + accept: %start EOF. */ + { + symbol_list_t *p = symbol_list_new (accept, empty_location); + p->location = grammar->location; + p->next = symbol_list_new (startsymbol, empty_location); + p->next->next = symbol_list_new (endtoken, empty_location); + p->next->next->next = symbol_list_new (NULL, empty_location); + p->next->next->next->next = grammar; + nrules += 1; + nritems += 3; + grammar = p; + } + + if (SYMBOL_NUMBER_MAX < nsyms) + fatal (_("too many symbols (tokens plus nonterminals); maximum %d"), + SYMBOL_NUMBER_MAX); + + assert (nsyms == ntokens + nvars); - lex_free (); xfclose (finput); /* Assign the symbols their symbol numbers. Write #defines for the @@ -1618,6 +552,6 @@ reader (void) /* 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); + /* The grammar as a symbol_list_t is no longer needed. */ + LIST_FREE (symbol_list_t, grammar); }