X-Git-Url: https://git.saurik.com/bison.git/blobdiff_plain/a870c5670ed51b4069baf8268d6fc6efa38fc95a..2666f928214f2accec70e46ced834828706777ae:/src/reader.c diff --git a/src/reader.c b/src/reader.c index 975a2028..509aa3ce 100644 --- a/src/reader.c +++ b/src/reader.c @@ -41,6 +41,9 @@ 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; @@ -230,25 +233,22 @@ 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))) { @@ -263,9 +263,7 @@ 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; @@ -275,26 +273,20 @@ 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 @@ -304,27 +296,13 @@ 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 | @@ -437,14 +415,12 @@ 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; @@ -752,43 +728,23 @@ 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) { @@ -797,7 +753,7 @@ typedef union"; break; case '/': - copy_comment2 (finput, &defines_obstack, &union_obstack); + copy_comment (finput, &union_obstack); break; case '{': @@ -805,26 +761,31 @@ 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)); + + if (defines_flag) + obstack_fgrow1 (&defines_obstack, "\ +#ifndef YYSTYPE\n\ +typedef %s +yystype;\n\ +# define YYSTYPE yystype\n\ +#endif\n", + muscle_find ("stype")); } @@ -987,13 +948,8 @@ parse_dquoted_param (const char *from) return NULL; } - for (;;) - { - if (literalchar (NULL, &c, '\"')) - obstack_1grow (¶m_obstack, c); - else - break; - } + while ((c = literalchar ()) != '"') + obstack_1grow (¶m_obstack, c); obstack_1grow (¶m_obstack, '\0'); param = obstack_finish (¶m_obstack); @@ -1138,16 +1094,6 @@ copy_action (symbol_list *rule, int stack_offset) 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); @@ -1205,7 +1151,9 @@ 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; } /*-------------------------------------------------------------------. @@ -1329,332 +1277,246 @@ 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++; + 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 = symbol_list_new (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) - { - /* 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; + 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 */ @@ -1690,6 +1552,13 @@ 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); @@ -1904,25 +1773,13 @@ symbols_output (void) obstack_free (&tokendefs, NULL); } -#if 0 - if (!no_parser_flag) - output_token_defines (&table_obstack); -#endif - if (defines_flag) { output_token_defines (&defines_obstack); if (!pure_parser) - { - if (spec_name_prefix) - obstack_fgrow1 (&defines_obstack, "\nextern YYSTYPE %slval;\n", - spec_name_prefix); - else - obstack_sgrow (&defines_obstack, - "\nextern YYSTYPE yylval;\n"); - } - + obstack_fgrow1 (&defines_obstack, "\nextern YYSTYPE %slval;\n", + spec_name_prefix); if (semantic_parser) { int i; @@ -1971,6 +1828,8 @@ packgram (void) 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) @@ -2073,7 +1932,6 @@ reader (void) /* Assign the symbols their symbol numbers. Write #defines for the token symbols into FDEFINES if requested. */ packsymbols (); - symbols_output (); /* Convert the grammar into the format described in gram.h. */ packgram (); /* Output the headers. */