X-Git-Url: https://git.saurik.com/bison.git/blobdiff_plain/007a50a493714fdda37f9600e068cf705c8ec3f1..b87f8b2159820e022e3b726b8369f74d64712f8f:/src/reader.c?ds=sidebyside diff --git a/src/reader.c b/src/reader.c index f76f5c84..75957010 100644 --- a/src/reader.c +++ b/src/reader.c @@ -45,9 +45,6 @@ typedef struct symbol_list const char *action; int action_line; - /* The guard is attached to the LHS of a rule. */ - const char *guard; - int guard_line; symbol_t *ruleprec; } symbol_list; @@ -80,8 +77,6 @@ symbol_list_new (symbol_t *sym) res->line = lineno; res->action = NULL; res->action_line = 0; - res->guard = NULL; - res->guard_line = 0; res->ruleprec = NULL; return res; } @@ -131,7 +126,7 @@ symbol_make_alias (symbol_t *symbol, char *typename) symval->class = token_sym; symval->type_name = typename; symval->user_token_number = symbol->user_token_number; - symbol->user_token_number = SALIAS; + symbol->user_token_number = USER_NUMBER_ALIAS; symval->alias = symbol; symbol->alias = symval; /* symbol and symval combined are only one symbol */ @@ -154,7 +149,7 @@ static bool symbol_check_alias_consistence (symbol_t *this) { /* Check only those who _are_ the aliases. */ - if (this->alias && this->user_token_number == SALIAS) + if (this->alias && this->user_token_number == USER_NUMBER_ALIAS) { if (this->prec != this->alias->prec) { @@ -199,23 +194,23 @@ symbol_pack (symbol_t *this) /* This symbol and its alias are a single token defn. Allocate a tokno, and assign to both check agreement of prec and assoc fields and make both the same */ - if (this->number == -1) + if (this->number == NUMBER_UNDEFINED) { if (this == eoftoken || this->alias == eoftoken) this->number = this->alias->number = 0; else { - assert (this->alias->number != -1); + assert (this->alias->number != NUMBER_UNDEFINED); this->number = this->alias->number; } } - /* Do not do processing below for SALIASs. */ - if (this->user_token_number == SALIAS) + /* Do not do processing below for USER_NUMBER_ALIASs. */ + if (this->user_token_number == USER_NUMBER_ALIAS) return TRUE; } else /* this->class == token_sym */ { - assert (this->number != -1); + assert (this->number != NUMBER_UNDEFINED); } symbols[this->number] = this; @@ -234,7 +229,7 @@ symbol_translation (symbol_t *this) { /* Non-terminal? */ if (this->class == token_sym - && this->user_token_number != SALIAS) + && this->user_token_number != USER_NUMBER_ALIAS) { /* A token which translation has already been set? */ if (token_translations[this->user_token_number] != undeftoken->number) @@ -332,6 +327,28 @@ get_type_name (int n, symbol_list *rule) return rp->sym->type_name; } +/*------------------------------------------------------------------. +| Copy the character C to OOUT, and insert quadigraphs when needed. | +`------------------------------------------------------------------*/ + +static inline void +copy_character (struct obstack *oout, int c) +{ + switch (c) + { + case '[': + obstack_sgrow (oout, "@<:@"); + break; + + case ']': + obstack_sgrow (oout, "@:>@"); + break; + + default: + obstack_1grow (oout, c); + } +} + /*------------------------------------------------------------. | Dump the string from FIN to OOUT if non null. MATCH is the | | delimiter of the string (either ' or "). | @@ -359,14 +376,14 @@ copy_string2 (FILE *fin, struct obstack *oout, int match, int store) continue; } - obstack_1grow (oout, c); + copy_character (oout, c); if (c == '\\') { c = getc (fin); if (c == EOF) fatal (_("unterminated string at end of file")); - obstack_1grow (oout, c); + copy_character (oout, c); if (c == '\n') ++lineno; @@ -463,30 +480,29 @@ copy_comment (FILE *fin, struct obstack *oout) fatal (_("unterminated comment")); else { - obstack_1grow (oout, c); + copy_character (oout, c); c = getc (fin); } } } -/*-----------------------------------------------------------------. -| FIN is pointing to a location (i.e., a `@'). Output to OOUT a | -| reference to this location. STACK_OFFSET is the number of values | -| in the current rule so far, which says where to find `$0' with | -| respect to the top of the stack. | -`-----------------------------------------------------------------*/ +/*-------------------------------------------------------------------. +| FIN is pointing to a location (i.e., a `@'). Output to OOUT a | +| reference to this location. RULE_LENGTH is the number of values in | +| the current rule so far, which says where to find `$0' with | +| respect to the top of the stack. | +`-------------------------------------------------------------------*/ static inline void -copy_at (FILE *fin, struct obstack *oout, int stack_offset) +copy_at (FILE *fin, struct obstack *oout, int rule_length) { - int c; + int c = getc (fin); + locations_flag = 1; - c = getc (fin); if (c == '$') { - obstack_sgrow (oout, "yyloc"); - locations_flag = 1; + obstack_sgrow (oout, "]b4_lhs_location["); } else if (isdigit (c) || c == '-') { @@ -494,16 +510,11 @@ copy_at (FILE *fin, struct obstack *oout, int stack_offset) ungetc (c, fin); n = read_signed_integer (fin); - if (n > stack_offset) + if (n > rule_length) complain (_("invalid value: %s%d"), "@", n); else - { - /* Offset is always 0 if parser has already popped the stack - pointer. */ - obstack_fgrow1 (oout, "yylsp[%d]", - n - (semantic_parser ? 0 : stack_offset)); - locations_flag = 1; - } + obstack_fgrow2 (oout, "]b4_rhs_location([%d], [%d])[", + rule_length, n); } else { @@ -514,19 +525,19 @@ copy_at (FILE *fin, struct obstack *oout, int stack_offset) } -/*-------------------------------------------------------------------. -| FIN is pointing to a wannabee semantic value (i.e., a `$'). | -| | -| Possible inputs: $[]($|integer) | -| | -| Output to OOUT a reference to this semantic value. STACK_OFFSET is | -| the number of values in the current rule so far, which says where | -| to find `$0' with respect to the top of the stack. | -`-------------------------------------------------------------------*/ +/*------------------------------------------------------------------. +| FIN is pointing to a wannabee semantic value (i.e., a `$'). | +| | +| Possible inputs: $[]($|integer) | +| | +| Output to OOUT a reference to this semantic value. RULE_LENGTH is | +| the number of values in the current rule so far, which says where | +| to find `$0' with respect to the top of the stack. | +`------------------------------------------------------------------*/ static inline void copy_dollar (FILE *fin, struct obstack *oout, - symbol_list *rule, int stack_offset) + symbol_list *rule, int rule_length) { int c = getc (fin); const char *type_name = NULL; @@ -542,15 +553,15 @@ copy_dollar (FILE *fin, struct obstack *oout, if (c == '$') { - obstack_sgrow (oout, "yyval"); - if (!type_name) type_name = get_type_name (0, rule); - if (type_name) - obstack_fgrow1 (oout, ".%s", type_name); if (!type_name && typed) complain (_("$$ of `%s' has no declared type"), rule->sym->tag); + if (!type_name) + type_name = ""; + obstack_fgrow1 (oout, + "]b4_lhs_value([%s])[", type_name); } else if (isdigit (c) || c == '-') { @@ -558,23 +569,19 @@ copy_dollar (FILE *fin, struct obstack *oout, ungetc (c, fin); n = read_signed_integer (fin); - if (n > stack_offset) + if (n > rule_length) complain (_("invalid value: %s%d"), "$", n); else { if (!type_name && n > 0) type_name = get_type_name (n, rule); - - /* Offset is always 0 if parser has already popped the stack - pointer. */ - obstack_fgrow1 (oout, "yyvsp[%d]", - n - (semantic_parser ? 0 : stack_offset)); - - if (type_name) - obstack_fgrow1 (oout, ".%s", type_name); if (!type_name && typed) complain (_("$%d of `%s' has no declared type"), n, rule->sym->tag); + if (!type_name) + type_name = ""; + obstack_fgrow3 (oout, "]b4_rhs_value([%d], [%d], [%s])[", + rule_length, n, type_name); } } else @@ -591,7 +598,7 @@ copy_dollar (FILE *fin, struct obstack *oout, `-------------------------------------------------------------------*/ static void -copy_definition (void) +copy_definition (struct obstack *oout) { int c; /* -1 while reading a character if prev char was %. */ @@ -599,7 +606,7 @@ copy_definition (void) if (!no_lines_flag) { - obstack_fgrow2 (&attrs_obstack, muscle_find ("linef"), + obstack_fgrow2 (oout, muscle_find ("linef"), lineno, quotearg_style (c_quoting_style, muscle_find ("filename"))); } @@ -613,7 +620,7 @@ copy_definition (void) switch (c) { case '\n': - obstack_1grow (&attrs_obstack, c); + obstack_1grow (oout, c); ++lineno; break; @@ -623,18 +630,18 @@ copy_definition (void) case '\'': case '"': - copy_string (finput, &attrs_obstack, c); + copy_string (finput, oout, c); break; case '/': - copy_comment (finput, &attrs_obstack); + copy_comment (finput, oout); break; case EOF: fatal ("%s", _("unterminated `%{' definition")); default: - obstack_1grow (&attrs_obstack, c); + copy_character (oout, c); } c = getc (finput); @@ -643,7 +650,7 @@ copy_definition (void) { if (c == '}') return; - obstack_1grow (&attrs_obstack, '%'); + obstack_1grow (oout, '%'); } after_percent = 0; } @@ -705,7 +712,7 @@ parse_token_decl (symbol_class what_is, symbol_class what_is_not) symbol->class = what_is; if (what_is == nterm_sym && oldclass != nterm_sym) symbol->number = nvars++; - if (what_is == token_sym && symbol->number == -1) + if (what_is == token_sym && symbol->number == NUMBER_UNDEFINED) symbol->number = ntokens++; if (typename) @@ -857,7 +864,7 @@ parse_assoc_decl (associativity assoc) symval->assoc = assoc; if (symval->class == nterm_sym) complain (_("symbol %s redefined"), symval->tag); - if (symval->number == -1) + if (symval->number == NUMBER_UNDEFINED) { symval->number = ntokens++; symval->class = token_sym; @@ -991,10 +998,10 @@ parse_expect_decl (void) | | | Two symbols are entered in the table, one for the token symbol and | | one for the literal. Both are given the , if any, from the | -| declaration. The ->user_token_number of the first is SALIAS and | -| the ->user_token_number of the second is set to the number, if | -| any, from the declaration. The two symbols are linked via | -| pointers in their ->alias fields. | +| declaration. The ->user_token_number of the first is | +| USER_NUMBER_ALIAS and the ->user_token_number of the second is set | +| to the number, if any, from the declaration. The two symbols are | +| linked via pointers in their ->alias fields. | | | | During OUTPUT_DEFINES_TABLE, the symbol is reported thereafter, | | only the literal string is retained it is the literal string that | @@ -1007,7 +1014,7 @@ parse_thong_decl (void) token_t token; symbol_t *symbol; char *typename = 0; - int usrtoknum = SUNDEF; + int usrtoknum = USER_NUMBER_UNDEFINED; token = lex (); /* fetch typename or first token */ if (token == tok_typename) @@ -1028,7 +1035,7 @@ parse_thong_decl (void) } symval->class = token_sym; symval->type_name = typename; - symval->user_token_number = SALIAS; + symval->user_token_number = USER_NUMBER_ALIAS; symbol = symval; token = lex (); /* get number or literal string */ @@ -1154,7 +1161,7 @@ parse_skel_decl (void) /*----------------------------------------------------------------. | Read from finput until `%%' is seen. Discard the `%%'. Handle | | any `%' declarations, and copy the contents of any `%{ ... %}' | -| groups to ATTRS_OBSTACK. | +| groups to PRE_PROLOGUE_OBSTACK or POST_PROLOGUE_OBSTACK. | `----------------------------------------------------------------*/ static void @@ -1174,7 +1181,10 @@ read_declarations (void) return; case tok_percent_left_curly: - copy_definition (); + if (!typed) + copy_definition (&pre_prologue_obstack); + else + copy_definition (&post_prologue_obstack); break; case tok_token: @@ -1252,26 +1262,26 @@ read_declarations (void) } } -/*-------------------------------------------------------------------. -| Assuming that a `{' has just been seen, copy everything up to the | -| matching `}' into the actions file. STACK_OFFSET is the number of | -| values in the current rule so far, which says where to find `$0' | -| with respect to the top of the stack. | -| | -| This routine is used both for actions and guards. Only | -| ACTION_OBSTACK is used, but this is fine, since we use only | -| pointers to relevant portions inside this obstack. | -`-------------------------------------------------------------------*/ +/*------------------------------------------------------------------. +| Assuming that a `{' has just been seen, copy everything up to the | +| matching `}' into ACTION_OBSTACK. | +| | +| RULE_LENGTH is the number of values in the current rule so far, | +| which says where to find `$0' with respect to the top of the | +| stack. It is not the same as the rule->length in the case of mid | +| rule actions. | +| | +| This routine is used for actions. | +`------------------------------------------------------------------*/ static void -parse_braces (symbol_list *rule, int stack_offset) +parse_action (symbol_list *rule, int rule_length) { - int c; - int count; - - count = 1; + int count = 1; + rule->action_line = lineno; while (count > 0) { + int c; while ((c = getc (finput)) != '}') switch (c) { @@ -1295,13 +1305,11 @@ parse_braces (symbol_list *rule, int stack_offset) break; case '$': - copy_dollar (finput, &action_obstack, - rule, stack_offset); + copy_dollar (finput, &action_obstack, rule, rule_length); break; case '@': - copy_at (finput, &action_obstack, - stack_offset); + copy_at (finput, &action_obstack, rule_length); break; case EOF: @@ -1317,29 +1325,9 @@ parse_braces (symbol_list *rule, int stack_offset) } obstack_1grow (&action_obstack, '\0'); -} - - -static void -parse_action (symbol_list *rule, int stack_offset) -{ - rule->action_line = lineno; - parse_braces (rule, stack_offset); rule->action = obstack_finish (&action_obstack); } - -static void -parse_guard (symbol_list *rule, int stack_offset) -{ - token_t t = lex (); - if (t != tok_left_curly) - complain (_("invalid %s declaration"), "%guard"); - rule->guard_line = lineno; - parse_braces (rule, stack_offset); - rule->guard = obstack_finish (&action_obstack); -} - /*-------------------------------------------------------------------. @@ -1371,8 +1359,8 @@ gensym (void) | null pointer instead of a symbol to terminate the rule. The next | | symbol is the lhs of the following rule. | | | -| All guards and actions are copied out to the appropriate files, | -| labelled by the rule number they apply to. | +| All actions are copied out, labelled by the rule number they apply | +| to. | | | | Bison used to allow some %directives in the rules sections, but | | this is no longer consider appropriate: (i) the documented grammar | @@ -1569,15 +1557,6 @@ readgram (void) t = lex (); } - if (t == tok_guard) - { - if (!semantic_parser) - complain (_("%%guard present but %%semantic_parser not specified")); - - parse_guard (crule, rulelength); - t = lex (); - } - if (t == tok_left_curly) { /* This case never occurs -wjh */ @@ -1663,7 +1642,7 @@ read_additionnal_code (void) } while ((c = getc (finput)) != EOF) - obstack_1grow (&el_obstack, c); + copy_character (&el_obstack, c); obstack_1grow (&el_obstack, 0); muscle_insert ("epilogue", obstack_finish (&el_obstack)); @@ -1678,15 +1657,38 @@ read_additionnal_code (void) static void token_translations_init (void) { - int last_user_token_number = 256; + int num_256_available_p = TRUE; int i; - /* Set the user numbers. */ + /* Find the highest user token number, and whether 256, the POSIX + preferred user token number for the error token, is used. */ + max_user_token_number = 0; + for (i = 0; i < ntokens; ++i) + { + symbol_t *this = symbols[i]; + if (this->user_token_number != USER_NUMBER_UNDEFINED) + { + if (this->user_token_number > max_user_token_number) + max_user_token_number = this->user_token_number; + if (this->user_token_number == 256) + num_256_available_p = FALSE; + } + } + + /* If 256 is not used, assign it to error, to follow POSIX. */ + if (num_256_available_p + && errtoken->user_token_number == USER_NUMBER_UNDEFINED) + errtoken->user_token_number = 256; + + /* Set the missing user numbers. */ + if (max_user_token_number < 256) + max_user_token_number = 256; + for (i = 0; i < ntokens; ++i) { symbol_t *this = symbols[i]; - if (this->user_token_number == SUNDEF) - this->user_token_number = ++last_user_token_number; + if (this->user_token_number == USER_NUMBER_UNDEFINED) + this->user_token_number = ++max_user_token_number; if (this->user_token_number > max_user_token_number) max_user_token_number = this->user_token_number; } @@ -1698,7 +1700,6 @@ token_translations_init (void) inputs. */ for (i = 0; i < max_user_token_number + 1; i++) token_translations[i] = undeftoken->number; - symbols_do (symbol_translation, NULL); } @@ -1757,13 +1758,13 @@ packgram (void) rules[ruleno].useful = TRUE; rules[ruleno].action = p->action; rules[ruleno].action_line = p->action_line; - rules[ruleno].guard = p->guard; - rules[ruleno].guard_line = p->guard_line; p = p->next; while (p && p->sym) { - ritem[itemno++] = p->sym->number; + /* item_number_t = token_number_t. + But the former needs to contain more: negative rule numbers. */ + ritem[itemno++] = token_number_as_item_number (p->sym->number); /* A rule gets by default the precedence and associativity of the last token in it. */ if (p->sym->class == token_sym) @@ -1793,13 +1794,12 @@ packgram (void) ritem_print (stderr); } -/*-------------------------------------------------------------------. -| Read in the grammar specification and record it in the format | -| described in gram.h. All guards are copied into the GUARD_OBSTACK | -| and all actions into ACTION_OBSTACK, in each case forming the body | -| of a C function (YYGUARD or YYACTION) which contains a switch | -| statement to decide which guard or action to execute. | -`-------------------------------------------------------------------*/ +/*------------------------------------------------------------------. +| Read in the grammar specification and record it in the format | +| described in gram.h. All actions are copied into ACTION_OBSTACK, | +| in each case forming the body of a C function (YYACTION) which | +| contains a switch statement to decide which action to execute. | +`------------------------------------------------------------------*/ void reader (void) @@ -1822,7 +1822,6 @@ reader (void) errtoken = getsym ("error"); errtoken->class = token_sym; errtoken->number = ntokens++; - errtoken->user_token_number = 256; /* Value specified by POSIX. */ /* Construct a token that represents all undefined literal tokens. It is always token number 2. */ @@ -1832,8 +1831,9 @@ reader (void) /* Initialize the obstacks. */ obstack_init (&action_obstack); - obstack_init (&attrs_obstack); obstack_init (&output_obstack); + obstack_init (&pre_prologue_obstack); + obstack_init (&post_prologue_obstack); finput = xfopen (infile, "r"); @@ -1853,7 +1853,7 @@ reader (void) } /* Read in the grammar, build grammar in list form. Write out - guards and actions. */ + actions. */ readgram (); /* Some C code is given at the end of the grammar file. */ read_additionnal_code ();