X-Git-Url: https://git.saurik.com/bison.git/blobdiff_plain/5f7e08323a452b0b55ad94b8c866970a2929d9ad..7742ddebf1f73cc15138d41c11a39f32b43c3ef5:/src/reader.c diff --git a/src/reader.c b/src/reader.c index 6206adca..9bec971a 100644 --- a/src/reader.c +++ b/src/reader.c @@ -21,7 +21,6 @@ #include "system.h" -#include "obstack.h" #include "quotearg.h" #include "quote.h" #include "getargs.h" @@ -36,13 +35,14 @@ #include "conflicts.h" #include "muscle_tab.h" -/* Number of slots allocated (but not necessarily used yet) in `rline' */ -static int rline_allocated; - typedef struct symbol_list { struct symbol_list *next; bucket *sym; + int line; + /* The action is attached to the LHS of a rule. */ + const char *action; + int action_line; bucket *ruleprec; } symbol_list; @@ -66,6 +66,19 @@ static int lastprec; static bucket *errtoken; static bucket *undeftoken; + + +static symbol_list * +symbol_list_new (bucket *sym) +{ + symbol_list *res = XMALLOC (symbol_list, 1); + res->next = NULL; + res->sym = sym; + res->line = lineno; + res->ruleprec = NULL; + return res; +} + /*===================\ @@ -123,7 +136,7 @@ read_signed_integer (FILE *stream) `--------------------------------------------------------------*/ static char * -get_type_name (int n, symbol_list * rule) +get_type_name (int n, symbol_list *rule) { int i; symbol_list *rp; @@ -219,25 +232,22 @@ 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))) { @@ -252,9 +262,7 @@ 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; @@ -264,26 +272,20 @@ 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); + obstack_1grow (oout, c); if (cplus_comment) ended = 1; else @@ -293,27 +295,13 @@ 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 | @@ -426,14 +414,12 @@ copy_definition (void) /* -1 while reading a character if prev char was %. */ int after_percent; -#if 0 if (!no_lines_flag) { obstack_fgrow2 (&attrs_obstack, muscle_find ("linef"), lineno, quotearg_style (c_quoting_style, muscle_find("filename"))); } -#endif after_percent = 0; @@ -568,7 +554,8 @@ parse_token_decl (symbol_class what_is, symbol_class what_is_not) else { complain (_("`%s' is invalid in %s"), - token_buffer, (what_is == token_sym) ? "%token" : "%nterm"); + token_buffer, + (what_is == token_sym) ? "%token" : "%nterm"); skip_to_char ('%'); } } @@ -740,43 +727,23 @@ parse_union_decl (void) { int c; 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'); - obstack_init (&union_obstack); obstack_sgrow (&union_obstack, "union"); - if (defines_flag) - obstack_sgrow (&defines_obstack, prologue); - c = getc (finput); - - while (c != EOF) + while (!done) { + c = xgetc (finput); + /* 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) { @@ -785,7 +752,7 @@ typedef union"; break; case '/': - copy_comment2 (finput, &defines_obstack, &union_obstack); + copy_comment (finput, &union_obstack); break; case '{': @@ -793,26 +760,22 @@ typedef union"; break; case '}': + /* FIXME: Errr. How could this happen???. --akim */ if (count == 0) complain (_("unmatched %s"), "`}'"); count--; - if (count <= 0) - { - if (defines_flag) - obstack_sgrow (&defines_obstack, epilogue); - /* JF don't choke on trailing semi */ - c = skip_white_space (); - if (c != ';') - ungetc (c, finput); - obstack_1grow (&union_obstack, 0); - muscle_insert ("stype", obstack_finish (&union_obstack)); - return; - } + if (!count) + done = TRUE; + break; } - - c = getc (finput); } + /* 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)); } @@ -952,14 +915,55 @@ parse_muscle_decl (void) } + +/*---------------------------------. +| Parse a double quoted parameter. | +`---------------------------------*/ + +static const char * +parse_dquoted_param (const char *from) +{ + 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; +} + /*----------------------------------. | Parse what comes after %skeleton. | `----------------------------------*/ -void +static void parse_skel_decl (void) { - /* Complete with parse_dquoted_param () on the CVS branch 1.29. */ + skeleton = parse_dquoted_param ("%skeleton"); } /*----------------------------------------------------------------. @@ -1042,10 +1046,10 @@ read_declarations (void) case tok_stropt: case tok_intopt: case tok_obsolete: - case tok_illegal: abort (); break; + case tok_illegal: default: complain (_("unrecognized: %s"), token_buffer); skip_to_char ('%'); @@ -1075,22 +1079,11 @@ copy_action (symbol_list *rule, int stack_offset) { int c; int count; - char buf[4096]; /* offset is always 0 if parser has already popped the stack pointer */ if (semantic_parser) stack_offset = 0; - obstack_fgrow1 (&action_obstack, "\ncase %d:\n", nrules); - - if (!no_lines_flag) - { - obstack_fgrow2 (&action_obstack, muscle_find ("linef"), - lineno, quotearg_style (c_quoting_style, - muscle_find ("filename"))); - } - obstack_1grow (&action_obstack, '{'); - count = 1; c = getc (finput); @@ -1148,7 +1141,9 @@ copy_action (symbol_list *rule, int stack_offset) } } - obstack_sgrow (&action_obstack, ";\n break;}"); + obstack_1grow (&action_obstack, '\0'); + rule->action = obstack_finish (&action_obstack); + rule->action_line = lineno; } /*-------------------------------------------------------------------. @@ -1251,20 +1246,6 @@ copy_guard (symbol_list *rule, int stack_offset) } -static void -record_rule_line (void) -{ - /* Record each rule's source line number in rline table. */ - - if (nrules >= rline_allocated) - { - rline_allocated = nrules * 2; - rline = XREALLOC (rline, short, rline_allocated); - } - rline[nrules] = lineno; -} - - /*-------------------------------------------------------------------. | Generate a dummy symbol, a nonterminal, whose name cannot conflict | | with the user's names. | @@ -1286,337 +1267,246 @@ gensym (void) sym->value = nvars++; return sym; } - -#if 0 -/*------------------------------------------------------------------. -| read in a %type declaration and record its information for | -| get_type_name to access. This is unused. It is only called from | -| the #if 0 part of readgram | -`------------------------------------------------------------------*/ - -static int -get_type (void) -{ - int k; - token_t token; - char *name; - - token = lex (); - - if (token != tok_typename) - { - complain (_("invalid %s declaration"), "%type"); - return t; - } - - name = xstrdup (token_buffer); - - for (;;) - { - token = lex (); - - switch (token) - { - case tok_semicolon: - return lex (); - - case tok_comma: - break; - - case tok_identifier: - if (symval->type_name == NULL) - symval->type_name = name; - else if (strcmp (name, symval->type_name) != 0) - complain (_("type redeclaration for %s"), symval->tag); - - break; - - default: - return token; - } - } -} - -#endif -/*------------------------------------------------------------------. -| Parse the input grammar into a one symbol_list structure. Each | -| rule is represented by a sequence of symbols: the left hand side | -| followed by the contents of the right hand side, followed by a | -| null pointer instead of a symbol to terminate the rule. The next | -| symbol is the lhs of the following rule. | -| | -| All guards and actions are copied out to the appropriate files, | -| labelled by the rule number they apply to. | -`------------------------------------------------------------------*/ +/*-------------------------------------------------------------------. +| Parse the input grammar into a one symbol_list structure. Each | +| rule is represented by a sequence of symbols: the left hand side | +| followed by the contents of the right hand side, followed by a | +| null pointer instead of a symbol to terminate the rule. The next | +| symbol is the lhs of the following rule. | +| | +| All guards and actions are copied out to the appropriate files, | +| labelled by the rule number they apply to. | +| | +| Bison used to allow some %directives in the rules sections, but | +| this is no longer consider appropriate: (i) the documented grammar | +| doesn't claim it, (ii), it would promote bad style, (iii), error | +| recovery for %directives consists in skipping the junk until a `%' | +| is seen and helrp synchronizing. This scheme is definitely wrong | +| in the rules section. | +`-------------------------------------------------------------------*/ static void readgram (void) { token_t t; bucket *lhs = NULL; - symbol_list *p; - symbol_list *p1; + 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. */ - symbol_list *crule; + symbol_list *crule = NULL; /* Points to the symbol_list preceding crule. */ - symbol_list *crule1; - - p1 = NULL; + 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 */ - bucket *first_rhs = 0; - - if (t == tok_identifier) - { - lhs = symval; - - if (!start_flag) - { - startval = lhs; - start_flag = 1; - } - - t = lex (); - if (t != tok_colon) - { - complain (_("ill-formed rule: initial symbol not followed by colon")); - unlex (t); - } - } - - if (nrules == 0 && t == tok_bar) - { - complain (_("grammar starts with vertical bar")); - lhs = symval; /* BOGUS: use a random symval */ - } - /* start a new rule and record its lhs. */ - - nrules++; - nitems++; - - record_rule_line (); + if (t == tok_identifier || t == tok_bar) + { + int action_flag = 0; + /* Number of symbols in rhs of this rule so far */ + int rulelength = 0; + int xactions = 0; /* JF for error checking */ + bucket *first_rhs = 0; + + if (t == tok_identifier) + { + lhs = symval; + + if (!start_flag) + { + startval = lhs; + start_flag = 1; + } - p = XCALLOC (symbol_list, 1); - p->sym = lhs; + t = lex (); + if (t != tok_colon) + { + complain (_("ill-formed rule: initial symbol not followed by colon")); + unlex (t); + } + } + + if (nrules == 0 && t == tok_bar) + { + complain (_("grammar starts with vertical bar")); + lhs = symval; /* BOGUS: use a random symval */ + } + /* start a new rule and record its lhs. */ + + nrules++; + nitems++; + + p = symbol_list_new (lhs); + + crule1 = p1; + if (p1) + p1->next = p; + else + grammar = p; - crule1 = p1; - if (p1) - p1->next = p; - else - grammar = p; + p1 = p; + crule = p; - p1 = p; - crule = p; + /* mark the rule's lhs as a nonterminal if not already so. */ - /* mark the rule's lhs as a nonterminal if not already so. */ + if (lhs->class == unknown_sym) + { + lhs->class = nterm_sym; + lhs->value = nvars; + nvars++; + } + else if (lhs->class == token_sym) + complain (_("rule given for %s, which is a token"), lhs->tag); - if (lhs->class == unknown_sym) - { - lhs->class = nterm_sym; - lhs->value = nvars; - nvars++; - } - else if (lhs->class == token_sym) - complain (_("rule given for %s, which is a token"), lhs->tag); - - /* read the rhs of the rule. */ + /* read the rhs of the rule. */ - for (;;) - { - t = lex (); - if (t == tok_prec) - { - t = lex (); - crule->ruleprec = symval; - t = lex (); - } - - if (!(t == tok_identifier || t == tok_left_curly)) - break; - - /* If next token is an identifier, see if a colon follows it. - If one does, exit this rule now. */ - if (t == tok_identifier) - { - bucket *ssave; - token_t t1; - - ssave = symval; - t1 = lex (); - unlex (t1); - symval = ssave; - if (t1 == tok_colon) - break; - - if (!first_rhs) /* JF */ - first_rhs = symval; - /* Not followed by colon => - process as part of this rule's rhs. */ - } - - /* If we just passed an action, that action was in the middle - of a rule, so make a dummy rule to reduce it to a - non-terminal. */ - if (action_flag) - { - bucket *sdummy; - - /* Since the action was written out with this rule's - number, we must give the new rule this number by - inserting the new rule before it. */ - - /* Make a dummy nonterminal, a gensym. */ - sdummy = gensym (); - - /* Make a new rule, whose body is empty, - before the current one, so that the action - just read can belong to it. */ - nrules++; - nitems++; - record_rule_line (); - p = XCALLOC (symbol_list, 1); - if (crule1) - crule1->next = p; - else - grammar = p; - p->sym = sdummy; - crule1 = XCALLOC (symbol_list, 1); - p->next = crule1; - crule1->next = crule; - - /* Insert the dummy generated by that rule into this - rule. */ - nitems++; - p = XCALLOC (symbol_list, 1); - p->sym = sdummy; - p1->next = p; - p1 = p; - - action_flag = 0; - } - - if (t == tok_identifier) - { - nitems++; - p = XCALLOC (symbol_list, 1); - p->sym = symval; - p1->next = p; - p1 = p; - } - else /* handle an action. */ - { - copy_action (crule, rulelength); - action_flag = 1; - xactions++; /* JF */ - } - rulelength++; - } /* end of read rhs of rule */ - - /* Put an empty link in the list to mark the end of this rule */ - p = XCALLOC (symbol_list, 1); - p1->next = p; - p1 = p; + for (;;) + { + t = lex (); + if (t == tok_prec) + { + t = lex (); + crule->ruleprec = symval; + t = lex (); + } + + if (!(t == tok_identifier || t == tok_left_curly)) + break; - if (t == tok_prec) - { - complain (_("two @prec's in a row")); - t = lex (); - crule->ruleprec = symval; - t = lex (); - } - if (t == tok_guard) - { - if (!semantic_parser) - complain (_("%%guard present but %%semantic_parser not specified")); + /* If next token is an identifier, see if a colon follows it. + If one does, exit this rule now. */ + if (t == tok_identifier) + { + bucket *ssave; + token_t t1; + + ssave = symval; + t1 = lex (); + unlex (t1); + symval = ssave; + if (t1 == tok_colon) + break; + + if (!first_rhs) /* JF */ + first_rhs = symval; + /* Not followed by colon => + process as part of this rule's rhs. */ + } + + /* If we just passed an action, that action was in the middle + of a rule, so make a dummy rule to reduce it to a + non-terminal. */ + if (action_flag) + { + /* 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. */ + bucket *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++; + p = symbol_list_new (sdummy); + /* Attach its lineno to that of the host rule. */ + p->line = crule->line; + 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. */ + nitems++; + p = symbol_list_new (sdummy); + p1->next = p; + p1 = p; + + action_flag = 0; + } + + if (t == tok_identifier) + { + nitems++; + p = symbol_list_new (symval); + p1->next = p; + p1 = p; + } + else /* handle an action. */ + { + copy_action (crule, rulelength); + action_flag = 1; + xactions++; /* JF */ + } + rulelength++; + } /* end of read rhs of rule */ + + /* Put an empty link in the list to mark the end of this rule */ + p = symbol_list_new (NULL); + p1->next = p; + p1 = p; + + if (t == tok_prec) + { + complain (_("two @prec's in a row")); + t = lex (); + crule->ruleprec = symval; + t = lex (); + } + if (t == tok_guard) + { + if (!semantic_parser) + complain (_("%%guard present but %%semantic_parser not specified")); - copy_guard (crule, rulelength); - t = lex (); - } - else if (t == tok_left_curly) - { - /* This case never occurs -wjh */ - if (action_flag) - complain (_("two actions at end of one rule")); - copy_action (crule, rulelength); - action_flag = 1; - xactions++; /* -wjh */ - t = lex (); - } - /* If $$ is being set in default way, report if any type - mismatch. */ - else if (!xactions - && first_rhs && lhs->type_name != first_rhs->type_name) - { - if (lhs->type_name == 0 - || first_rhs->type_name == 0 - || strcmp (lhs->type_name, first_rhs->type_name)) - complain (_("type clash (`%s' `%s') on default action"), - lhs->type_name ? lhs->type_name : "", - first_rhs->type_name ? first_rhs->type_name : ""); - } - /* Warn if there is no default for $$ but we need one. */ - else if (!xactions && !first_rhs && lhs->type_name != 0) - complain (_("empty rule for typed nonterminal, and no action")); - if (t == tok_semicolon) + copy_guard (crule, rulelength); t = lex (); - } -#if 0 - /* these things can appear as alternatives to rules. */ -/* NO, they cannot. - a) none of the documentation allows them - b) most of them scan forward until finding a next % - thus they may swallow lots of intervening rules -*/ - else if (t == tok_token) - { - parse_token_decl (token_sym, nterm_sym); - t = lex (); - } - else if (t == tok_nterm) - { - parse_token_decl (nterm_sym, token_sym); - t = lex (); - } - else if (t == tok_type) - { - t = get_type (); - } - else if (t == tok_union) - { - parse_union_decl (); - t = lex (); - } - else if (t == tok_expect) - { - parse_expect_decl (); - t = lex (); - } - else if (t == tok_start) - { - parse_start_decl (); + } + else if (t == tok_left_curly) + { + /* This case never occurs -wjh */ + if (action_flag) + complain (_("two actions at end of one rule")); + copy_action (crule, rulelength); + action_flag = 1; + xactions++; /* -wjh */ + t = lex (); + } + /* If $$ is being set in default way, report if any type + mismatch. */ + else if (!xactions + && first_rhs && lhs->type_name != first_rhs->type_name) + { + if (lhs->type_name == 0 + || first_rhs->type_name == 0 + || strcmp (lhs->type_name, first_rhs->type_name)) + complain (_("type clash (`%s' `%s') on default action"), + lhs->type_name ? lhs->type_name : "", + first_rhs->type_name ? first_rhs->type_name : ""); + } + /* Warn if there is no default for $$ but we need one. */ + else if (!xactions && !first_rhs && lhs->type_name != 0) + complain (_("empty rule for typed nonterminal, and no action")); + if (t == tok_semicolon) t = lex (); - } -#endif + } + else + { + complain (_("invalid input: %s"), quote (token_buffer)); + t = lex (); + } - else - { - complain (_("invalid input: %s"), quote (token_buffer)); - t = lex (); - } - } /* grammar has been read. Do some checking */ @@ -1652,6 +1542,13 @@ read_additionnal_code (void) 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) obstack_1grow (&el_obstack, c); @@ -1660,55 +1557,6 @@ read_additionnal_code (void) } -/*--------------------------------------------------------------. -| For named tokens, but not literal ones, define the name. The | -| value is the user token number. | -`--------------------------------------------------------------*/ - -static void -output_token_defines (struct obstack *oout) -{ - bucket *bp; - char *cp, *symbol; - char c; - - for (bp = firstsymbol; bp; bp = bp->next) - { - symbol = bp->tag; /* get symbol */ - - if (bp->value >= ntokens) - continue; - if (bp->user_token_number == SALIAS) - continue; - if ('\'' == *symbol) - continue; /* skip literal character */ - if (bp == errtoken) - continue; /* skip error token */ - if ('\"' == *symbol) - { - /* use literal string only if given a symbol with an alias */ - if (bp->alias) - symbol = bp->alias->tag; - else - continue; - } - - /* Don't #define nonliteral tokens whose names contain periods. */ - cp = symbol; - while ((c = *cp++) && c != '.'); - if (c != '\0') - continue; - - obstack_fgrow2 (oout, "# define\t%s\t%d\n", - symbol, bp->user_token_number); - if (semantic_parser) - /* FIXME: This is certainly dead wrong, and should be just as - above. --akim. */ - obstack_fgrow2 (oout, "# define\tT%s\t%d\n", symbol, bp->value); - } -} - - /*------------------------------------------------------------------. | Set TOKEN_TRANSLATIONS. Check that no two symbols share the same | | number. | @@ -1850,60 +1698,57 @@ packsymbols (void) } -/*-----------------------------------. -| Output definition of token names. | -`-----------------------------------*/ +/*---------------------------------------------------------------. +| Save the definition of token names in the `TOKENDEFS' muscle. | +`---------------------------------------------------------------*/ static void -symbols_output (void) +symbols_save (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 0 - if (!no_parser_flag) - output_token_defines (&table_obstack); -#endif - - if (defines_flag) + struct obstack tokendefs; + bucket *bp; + char *cp, *symbol; + char c; + obstack_init (&tokendefs); + + for (bp = firstsymbol; bp; bp = bp->next) { - output_token_defines (&defines_obstack); + symbol = bp->tag; /* get symbol */ - if (!pure_parser) + 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) { - if (spec_name_prefix) - obstack_fgrow1 (&defines_obstack, "\nextern YYSTYPE %slval;\n", - spec_name_prefix); + /* use literal string only if given a symbol with an alias */ + if (bp->alias) + symbol = bp->alias->tag; else - obstack_sgrow (&defines_obstack, - "\nextern YYSTYPE yylval;\n"); + continue; } - if (semantic_parser) - { - int i; + /* Don't #define nonliteral tokens whose names contain periods. */ + cp = symbol; + while ((c = *cp++) && c != '.'); + if (c != '\0') + continue; - 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 - } + 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); } @@ -1919,14 +1764,8 @@ packgram (void) int ruleno; symbol_list *p; - bucket *ruleprec; - ritem = XCALLOC (short, nitems + 1); - rlhs = XCALLOC (short, nrules) - 1; - rrhs = XCALLOC (short, nrules) - 1; - rprec = XCALLOC (short, nrules) - 1; - rprecsym = XCALLOC (short, nrules) - 1; - rassoc = XCALLOC (short, nrules) - 1; + rule_table = XCALLOC (rule_t, nrules) - 1; itemno = 0; ruleno = 1; @@ -1934,9 +1773,13 @@ packgram (void) p = grammar; while (p) { - rlhs[ruleno] = p->sym->value; - rrhs[ruleno] = itemno; - ruleprec = p->ruleprec; + 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; p = p->next; while (p && p->sym) @@ -1946,8 +1789,8 @@ packgram (void) of the last token in it. */ if (p->sym->class == token_sym) { - rprec[ruleno] = p->sym->prec; - rassoc[ruleno] = p->sym->assoc; + rule_table[ruleno].prec = p->sym->prec; + rule_table[ruleno].assoc = p->sym->assoc; } if (p) p = p->next; @@ -1957,9 +1800,9 @@ packgram (void) the specified symbol's precedence replaces the default. */ if (ruleprec) { - rprec[ruleno] = ruleprec->prec; - rassoc[ruleno] = ruleprec->assoc; - rprecsym[ruleno] = ruleprec->value; + rule_table[ruleno].prec = ruleprec->prec; + rule_table[ruleno].assoc = ruleprec->assoc; + rule_table[ruleno].precsym = ruleprec->value; } ritem[itemno++] = -ruleno; @@ -1970,6 +1813,9 @@ packgram (void) } ritem[itemno] = 0; + + if (trace_flag) + ritem_print (stderr); } /*-------------------------------------------------------------------. @@ -1990,8 +1836,6 @@ reader (void) nvars = 0; nrules = 0; nitems = 0; - rline_allocated = 10; - rline = XCALLOC (short, rline_allocated); typed = 0; lastprec = 0; @@ -2021,6 +1865,14 @@ reader (void) undeftoken->class = token_sym; undeftoken->user_token_number = 2; + /* Initialize the obstacks. */ + obstack_init (&action_obstack); + obstack_init (&attrs_obstack); + obstack_init (&guard_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. */ @@ -2031,14 +1883,15 @@ reader (void) /* 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 (); - symbols_output (); + /* Save them. */ + symbols_save (); + /* Convert the grammar into the format described in gram.h. */ packgram (); }