X-Git-Url: https://git.saurik.com/bison.git/blobdiff_plain/ff48177d2514d8ba381852e4d1f50b98ef3cd763..29e8831621af57c5056380b68a96337e7647b850:/src/reader.c diff --git a/src/reader.c b/src/reader.c index 8fa3dfe8..d76bd6fe 100644 --- a/src/reader.c +++ b/src/reader.c @@ -1,5 +1,5 @@ /* Input parser for bison - Copyright 1984, 1986, 1989, 1992, 1998, 2000 + Copyright 1984, 1986, 1989, 1992, 1998, 2000, 2001 Free Software Foundation, Inc. This file is part of Bison, the GNU Compiler Compiler. @@ -21,35 +21,37 @@ #include "system.h" -#include "obstack.h" #include "quotearg.h" #include "quote.h" #include "getargs.h" #include "files.h" -#include "xalloc.h" #include "symtab.h" +#include "options.h" #include "lex.h" #include "gram.h" #include "complain.h" #include "output.h" #include "reader.h" #include "conflicts.h" -#include "macrotab.h" - -/* Number of slots allocated (but not necessarily used yet) in `rline' */ -static int rline_allocated; +#include "muscle_tab.h" 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; + + /* The guard is attached to the LHS of a rule. */ + const char *guard; + int guard_line; bucket *ruleprec; -} -symbol_list; +} symbol_list; int lineno; -char **tags; -short *user_toknums; static symbol_list *grammar; static int start_flag; static bucket *startval; @@ -64,8 +66,26 @@ static int typed; /* Incremented for each %left, %right or %nonassoc seen */ static int lastprec; -static bucket *errtoken; -static bucket *undeftoken; +static bucket *errtoken = NULL; +static bucket *undeftoken = NULL; +static bucket *eoftoken = NULL; +static bucket *axiom = NULL; + +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->action = NULL; + res->action_line = 0; + res->guard = NULL; + res->guard_line = 0; + res->ruleprec = NULL; + return res; +} + /*===================\ @@ -123,7 +143,7 @@ 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 +239,22 @@ 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 +269,7 @@ 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 +279,20 @@ 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 +302,13 @@ 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 | @@ -338,9 +333,16 @@ copy_at (FILE *fin, struct obstack *oout, int stack_offset) ungetc (c, fin); n = read_signed_integer (fin); - - obstack_fgrow1 (oout, "yylsp[%d]", n - stack_offset); - locations_flag = 1; + if (n > stack_offset) + complain (_("invalid value: %s%d"), "@", n); + else + { + /* Offset is always 0 if parser has already popped the stack + pointer. */ + obstack_fgrow1 (oout, "yylsp[%d]", + n - (semantic_parser ? 0 : stack_offset)); + locations_flag = 1; + } } else { @@ -395,16 +397,24 @@ copy_dollar (FILE *fin, struct obstack *oout, ungetc (c, fin); n = read_signed_integer (fin); - if (!type_name && n > 0) - type_name = get_type_name (n, rule); - - obstack_fgrow1 (oout, "yyvsp[%d]", n - stack_offset); - - if (type_name) - obstack_fgrow1 (oout, ".%s", type_name); - if (!type_name && typed) - complain (_("$%d of `%s' has no declared type"), - n, rule->sym->tag); + if (n > stack_offset) + complain (_("invalid value: %s%d"), "$", n); + else + { + if (!type_name && n > 0) + type_name = get_type_name (n, rule); + + /* Offset is always 0 if parser has already popped the stack + pointer. */ + obstack_fgrow1 (oout, "yyvsp[%d]", + n - (semantic_parser ? 0 : stack_offset)); + + if (type_name) + obstack_fgrow1 (oout, ".%s", type_name); + if (!type_name && typed) + complain (_("$%d of `%s' has no declared type"), + n, rule->sym->tag); + } } else { @@ -426,11 +436,12 @@ 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, "#line %d %s\n", - lineno, quotearg_style (c_quoting_style, infile)); -#endif + { + obstack_fgrow2 (&attrs_obstack, muscle_find ("linef"), + lineno, quotearg_style (c_quoting_style, + muscle_find ("filename"))); + } after_percent = 0; @@ -487,8 +498,8 @@ copy_definition (void) static void parse_token_decl (symbol_class what_is, symbol_class what_is_not) { - token_t token = 0; - char *typename = 0; + token_t token = tok_undef; + char *typename = NULL; /* The symbol being defined. */ struct bucket *symbol = NULL; @@ -537,7 +548,6 @@ parse_token_decl (symbol_class what_is, symbol_class what_is_not) /* symbol and symval combined are only one symbol */ nsyms--; } - translations = 1; symbol = NULL; } else if (token == tok_identifier) @@ -562,12 +572,15 @@ parse_token_decl (symbol_class what_is, symbol_class what_is_not) else if (symbol && token == tok_number) { symbol->user_token_number = numval; - translations = 1; + /* User defined EOF token? */ + if (numval == 0) + eoftoken = symbol; } 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 ('%'); } } @@ -704,7 +717,6 @@ parse_assoc_decl (associativity assoc) if (prev == tok_identifier) { symval->user_token_number = numval; - translations = 1; } else { @@ -724,16 +736,15 @@ token_buffer); } prev = t; - } } /*--------------------------------------------------------------. -| Copy the union declaration into ATTRS_OBSTACK (and fdefines), | -| where it is made into the definition of YYSTYPE, the type of | -| elements of the parser value stack. | +| 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. | `--------------------------------------------------------------*/ static void @@ -741,29 +752,23 @@ parse_union_decl (void) { int c; int count = 0; - + bool done = FALSE; + struct obstack union_obstack; if (typed) complain (_("multiple %s declarations"), "%union"); typed = 1; - if (!no_lines_flag) - obstack_fgrow2 (&attrs_obstack, "\n#line %d %s\n", - lineno, quotearg_style (c_quoting_style, infile)); - else - obstack_1grow (&attrs_obstack, '\n'); - - obstack_sgrow (&attrs_obstack, "typedef union"); - if (defines_flag) - obstack_sgrow (&defines_obstack, "typedef union"); + obstack_init (&union_obstack); + obstack_sgrow (&union_obstack, "union"); - c = getc (finput); - - while (c != EOF) + while (!done) { - obstack_1grow (&attrs_obstack, c); - if (defines_flag) - obstack_1grow (&defines_obstack, c); + c = xgetc (finput); + + /* If C contains '/', it is output by copy_comment (). */ + if (c != '/') + obstack_1grow (&union_obstack, c); switch (c) { @@ -772,7 +777,7 @@ parse_union_decl (void) break; case '/': - copy_comment2 (finput, &defines_obstack, &attrs_obstack); + copy_comment (finput, &union_obstack); break; case '{': @@ -780,24 +785,22 @@ parse_union_decl (void) break; case '}': + /* FIXME: Errr. How could this happen???. --akim */ if (count == 0) complain (_("unmatched %s"), "`}'"); count--; - if (count <= 0) - { - obstack_sgrow (&attrs_obstack, " YYSTYPE;\n"); - if (defines_flag) - obstack_sgrow (&defines_obstack, " YYSTYPE;\n"); - /* JF don't choke on trailing semi */ - c = skip_white_space (); - if (c != ';') - ungetc (c, finput); - 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)); } @@ -845,9 +848,8 @@ parse_thong_decl (void) token_t token; struct bucket *symbol; char *typename = 0; - int usrtoknum; + int usrtoknum = SUNDEF; - translations = 1; token = lex (); /* fetch typename or first token */ if (token == tok_typename) { @@ -877,8 +879,6 @@ parse_thong_decl (void) usrtoknum = numval; token = lex (); /* okay, did number, now get literal */ } - else - usrtoknum = 0; /* process literal string token */ @@ -899,14 +899,13 @@ parse_thong_decl (void) nsyms--; } -/* FIXME. */ static void -parse_macro_decl (void) +parse_muscle_decl (void) { int ch = ungetc (skip_white_space (), finput); - char* macro_key; - char* macro_value; + char *muscle_key; + char *muscle_value; /* Read key. */ if (!isalpha (ch) && ch != '_') @@ -915,10 +914,10 @@ parse_macro_decl (void) skip_to_char ('%'); return; } - copy_identifier (finput, ¯o_obstack); - obstack_1grow (¯o_obstack, 0); - macro_key = obstack_finish (¯o_obstack); - + copy_identifier (finput, &muscle_obstack); + obstack_1grow (&muscle_obstack, 0); + muscle_key = obstack_finish (&muscle_obstack); + /* Read value. */ ch = skip_white_space (); if (ch != '"') @@ -933,42 +932,64 @@ parse_macro_decl (void) else fatal (_("Premature EOF after %s"), "\""); } - copy_string2 (finput, ¯o_obstack, '"', 0); - obstack_1grow (¯o_obstack, 0); - macro_value = obstack_finish (¯o_obstack); + copy_string2 (finput, &muscle_obstack, '"', 0); + obstack_1grow (&muscle_obstack, 0); + muscle_value = obstack_finish (&muscle_obstack); /* Store the (key, value) pair in the environment. */ - macro_insert (macro_key, macro_value); + muscle_insert (muscle_key, muscle_value); } -/*------------------------------------------. -| Parse what comes after %header_extension. | -`------------------------------------------*/ -static void -parse_header_extension_decl (void) + +/*---------------------------------. +| Parse a double quoted parameter. | +`---------------------------------*/ + +static const char * +parse_dquoted_param (const char *from) { - char buff[32]; + struct obstack param_obstack; + const char *param = NULL; + int c; - if (header_extension) - complain (_("multiple %%header_extension declarations")); - fscanf (finput, "%s", buff); - header_extension = xstrdup (buff); + 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 %source_extension. | -`------------------------------------------*/ +/*----------------------------------. +| Parse what comes after %skeleton. | +`----------------------------------*/ static void -parse_source_extension_decl (void) +parse_skel_decl (void) { - char buff[32]; - - if (src_extension) - complain (_("multiple %%source_extension declarations")); - fscanf (finput, "%s", buff); - src_extension = xstrdup (buff); + skeleton = parse_dquoted_param ("%skeleton"); } /*----------------------------------------------------------------. @@ -980,16 +1001,13 @@ parse_source_extension_decl (void) static void read_declarations (void) { - int c; - int tok; - for (;;) { - c = skip_white_space (); + int c = skip_white_space (); if (c == '%') { - tok = parse_percent_token (); + token_t tok = parse_percent_token (); switch (tok) { @@ -1040,21 +1058,24 @@ read_declarations (void) parse_assoc_decl (non_assoc); break; - case tok_hdrext: - parse_header_extension_decl (); + case tok_define: + parse_muscle_decl (); break; - case tok_srcext: - parse_source_extension_decl (); + case tok_skel: + parse_skel_decl (); break; - case tok_define: - parse_macro_decl (); + case tok_noop: break; - case tok_noop: + case tok_stropt: + case tok_intopt: + case tok_obsolete: + abort (); break; + case tok_illegal: default: complain (_("unrecognized: %s"), token_buffer); skip_to_char ('%'); @@ -1077,80 +1098,61 @@ read_declarations (void) | matching `}' into the actions file. STACK_OFFSET is the number of | | values in the current rule so far, which says where to find `$0' | | with respect to the top of the stack. | +| | +| This routine is used both for actions and guards. Only | +| ACTION_OBSTACK is used, but this is fine, since we use only | +| pointers to relevant portions inside this obstack. | `-------------------------------------------------------------------*/ static void -copy_action (symbol_list *rule, int stack_offset) +parse_braces (symbol_list *rule, int stack_offset) { int c; int count; - char buf[4096]; - - /* offset is always 0 if parser has already popped the stack pointer */ - if (semantic_parser) - stack_offset = 0; - - sprintf (buf, "\ncase %d:\n", nrules); - obstack_grow (&action_obstack, buf, strlen (buf)); - - if (!no_lines_flag) - { - sprintf (buf, "#line %d %s\n", - lineno, quotearg_style (c_quoting_style, infile)); - obstack_grow (&action_obstack, buf, strlen (buf)); - } - obstack_1grow (&action_obstack, '{'); count = 1; - c = getc (finput); - while (count > 0) { - while (c != '}') - { - switch (c) - { - case '\n': - obstack_1grow (&action_obstack, c); - lineno++; - break; - - case '{': - obstack_1grow (&action_obstack, c); - count++; - break; - - case '\'': - case '"': - copy_string (finput, &action_obstack, c); - break; + while ((c = getc (finput)) != '}') + switch (c) + { + case '\n': + obstack_1grow (&action_obstack, c); + lineno++; + break; - case '/': - copy_comment (finput, &action_obstack); - break; + case '{': + obstack_1grow (&action_obstack, c); + count++; + break; - case '$': - copy_dollar (finput, &action_obstack, - rule, stack_offset); - break; + case '\'': + case '"': + copy_string (finput, &action_obstack, c); + break; - case '@': - copy_at (finput, &action_obstack, - stack_offset); - break; + case '/': + copy_comment (finput, &action_obstack); + break; - case EOF: - fatal (_("unmatched %s"), "`{'"); + case '$': + copy_dollar (finput, &action_obstack, + rule, stack_offset); + break; - default: - obstack_1grow (&action_obstack, c); - } + case '@': + copy_at (finput, &action_obstack, + stack_offset); + break; - c = getc (finput); - } + case EOF: + fatal (_("unmatched %s"), "`{'"); - /* above loop exits when c is '}' */ + default: + obstack_1grow (&action_obstack, c); + } + /* Above loop exits when C is '}'. */ if (--count) { obstack_1grow (&action_obstack, c); @@ -1158,121 +1160,31 @@ copy_action (symbol_list *rule, int stack_offset) } } - obstack_sgrow (&action_obstack, ";\n break;}"); + obstack_1grow (&action_obstack, '\0'); } - -/*-------------------------------------------------------------------. -| After `%guard' is seen in the input file, copy the actual guard | -| into the guards file. If the guard is followed by an action, copy | -| that into the actions file. STACK_OFFSET is the number of values | -| in the current rule so far, which says where to find `$0' with | -| respect to the top of the stack, for the simple parser in which | -| the stack is not popped until after the guard is run. | -`-------------------------------------------------------------------*/ + static void -copy_guard (symbol_list *rule, int stack_offset) +parse_action (symbol_list *rule, int stack_offset) { - int c; - int count; - int brace_flag = 0; - - /* offset is always 0 if parser has already popped the stack pointer */ - if (semantic_parser) - stack_offset = 0; - - obstack_fgrow1 (&guard_obstack, "\ncase %d:\n", nrules); - if (!no_lines_flag) - obstack_fgrow2 (&guard_obstack, "#line %d %s\n", - lineno, quotearg_style (c_quoting_style, infile)); - obstack_1grow (&guard_obstack, '{'); - - count = 0; - c = getc (finput); - - while (brace_flag ? (count > 0) : (c != ';')) - { - switch (c) - { - case '\n': - obstack_1grow (&guard_obstack, c); - lineno++; - break; - - case '{': - obstack_1grow (&guard_obstack, c); - brace_flag = 1; - count++; - break; - - case '}': - obstack_1grow (&guard_obstack, c); - if (count > 0) - count--; - else - { - complain (_("unmatched %s"), "`}'"); - c = getc (finput); /* skip it */ - } - break; - - case '\'': - case '"': - copy_string (finput, &guard_obstack, c); - break; - - case '/': - copy_comment (finput, &guard_obstack); - break; - - case '$': - copy_dollar (finput, &guard_obstack, rule, stack_offset); - break; - - case '@': - copy_at (finput, &guard_obstack, stack_offset); - break; - - case EOF: - fatal ("%s", _("unterminated %guard clause")); - - default: - obstack_1grow (&guard_obstack, c); - } - - if (c != '}' || count != 0) - c = getc (finput); - } - - c = skip_white_space (); - - obstack_sgrow (&guard_obstack, ";\n break;}"); - if (c == '{') - copy_action (rule, stack_offset); - else if (c == '=') - { - c = getc (finput); /* why not skip_white_space -wjh */ - if (c == '{') - copy_action (rule, stack_offset); - } - else - ungetc (c, finput); + rule->action_line = lineno; + parse_braces (rule, stack_offset); + rule->action = obstack_finish (&action_obstack); } - + static void -record_rule_line (void) +parse_guard (symbol_list *rule, int stack_offset) { - /* 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; + 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); } + /*-------------------------------------------------------------------. | Generate a dummy symbol, a nonterminal, whose name cannot conflict | @@ -1295,364 +1207,253 @@ 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; - record_rule_line (); + if (t == tok_identifier) + { + lhs = symval; - p = XCALLOC (symbol_list, 1); - p->sym = lhs; + if (!start_flag) + { + startval = lhs; + start_flag = 1; + } - crule1 = p1; - if (p1) - p1->next = p; - else - grammar = p; + t = lex (); + if (t != tok_colon) + { + complain (_("ill-formed rule: initial symbol not followed by colon")); + unlex (t); + } + } - p1 = p; - crule = p; + 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. */ - /* mark the rule's lhs as a nonterminal if not already so. */ + nrules++; + nitems++; - 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); + p = symbol_list_new (lhs); - /* read the rhs of the rule. */ + crule1 = p1; + if (p1) + p1->next = p; + else + grammar = p; - for (;;) - { - t = lex (); - if (t == tok_prec) - { - t = lex (); - crule->ruleprec = symval; - t = lex (); - } + p1 = p; + crule = p; - if (!(t == tok_identifier || t == tok_left_curly)) - break; + /* mark the rule's lhs as a nonterminal if not already so. */ - /* 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 (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 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; - } + /* read the rhs of the rule. */ - 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 */ + for (;;) + { + t = lex (); + if (t == tok_prec) + { + t = lex (); + crule->ruleprec = symval; + t = lex (); + } + + if (!(t == tok_identifier || t == tok_left_curly)) + break; - /* Put an empty link in the list to mark the end of this rule */ - p = XCALLOC (symbol_list, 1); - p1->next = p; - p1 = p; + /* 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. */ + { + 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_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 (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) + parse_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 (); - t = lex (); - } -#endif + } - else - { - complain (_("invalid input: %s"), quote (token_buffer)); + 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_semicolon) t = lex (); - } - } + } + else + { + complain (_("invalid input: %s"), quote (token_buffer)); + t = lex (); + } /* grammar has been read. Do some checking */ - if (nsyms > MAXSHORT) - fatal (_("too many symbols (tokens plus nonterminals); maximum %d"), - MAXSHORT); if (nrules == 0) fatal (_("no rules in the input grammar")); -#if 0 /* This code is in the skeleton now. */ - /* JF put out same default YYSTYPE as YACC does */ - if (typed == 0 - && !value_components_used) - { - /* We used to use `unsigned long' as YYSTYPE on MSDOS, - but it seems better to be consistent. - Most programs should declare their own type anyway. */ - obstack_sgrow (&attrs_obstack, - "#ifndef YYSTYPE\n#define YYSTYPE int\n#endif\n"); - if (defines_flag) - obstack_sgrow (&defines_obstack, "\ -# ifndef YYSTYPE\n\ -# define YYSTYPE int\n\ -# endif\n"); - } -#endif - /* Report any undefined symbols and consider them nonterminals. */ for (bp = firstsymbol; bp; bp = bp->next) @@ -1665,102 +1466,107 @@ readgram (void) bp->value = nvars++; } + /* Insert the initial rule, which line is that of the first rule + (not that of the start symbol): + + axiom: %start EOF. */ + p = symbol_list_new (axiom); + p->line = grammar->line; + p->next = symbol_list_new (startval); + p->next->next = symbol_list_new (eoftoken); + p->next->next->next = symbol_list_new (NULL); + p->next->next->next->next = grammar; + nrules += 1; + nitems += 3; + grammar = p; + startval = axiom; + + if (nsyms > MAXSHORT) + fatal (_("too many symbols (tokens plus nonterminals); maximum %d"), + MAXSHORT); + ntokens = nsyms - nvars; } /* At the end of the grammar file, some C source code must - be stored. It is going to be associated to the user_code + be stored. It is going to be associated to the epilogue directive. */ static void read_additionnal_code (void) { char c; - struct obstack uc_obstack; - - obstack_init (&uc_obstack); + 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) - obstack_1grow (&uc_obstack, c); - - obstack_1grow (&uc_obstack, 0); - macro_insert ("user_code", obstack_finish (&uc_obstack)); + obstack_1grow (&el_obstack, c); + + obstack_1grow (&el_obstack, 0); + muscle_insert ("epilogue", obstack_finish (&el_obstack)); } -/*--------------------------------------------------------------. -| For named tokens, but not literal ones, define the name. The | -| value is the user token number. | -`--------------------------------------------------------------*/ +/*------------------------------------------------------------------. +| Set TOKEN_TRANSLATIONS. Check that no two symbols share the same | +| number. | +`------------------------------------------------------------------*/ static void -output_token_defines (struct obstack *oout) +token_translations_init (void) { - bucket *bp; - char *cp, *symbol; - char c; + bucket *bp = NULL; + int i; + + token_translations = XCALLOC (short, max_user_token_number + 1); + + /* Initialize all entries for literal tokens to 2, the internal + token number for $undefined., which represents all invalid + inputs. */ + for (i = 0; i <= max_user_token_number; i++) + token_translations[i] = 2; for (bp = firstsymbol; bp; bp = bp->next) { - symbol = bp->tag; /* get symbol */ - + /* Non-terminal? */ if (bp->value >= ntokens) continue; + /* A token string alias? */ 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; + assert (bp->user_token_number != SUNDEF); - obstack_fgrow2 (oout, "# define\t%s\t%d\n", - symbol, - (translations ? bp->user_token_number : bp->value)); - if (semantic_parser) - obstack_fgrow2 (oout, "# define\tT%s\t%d\n", symbol, bp->value); + /* A token which translation has already been set? */ + if (token_translations[bp->user_token_number] != 2) + complain (_("tokens %s and %s both assigned number %d"), + symbols[token_translations[bp->user_token_number]]->tag, + bp->tag, bp->user_token_number); + token_translations[bp->user_token_number] = bp->value; } - - /* obstack_1grow (oout, '\n'); */ } -/*------------------------------------------------------------------. -| Assign symbol numbers, and write definition of token names into | -| FDEFINES. Set up vectors TAGS and SPREC of names and precedences | -| of symbols. | -`------------------------------------------------------------------*/ +/*----------------------------------------------------------------. +| Assign symbol numbers, and write definition of token names into | +| FDEFINES. Set up vectors SYMBOL_TABLE, TAGS of symbols. | +`----------------------------------------------------------------*/ static void packsymbols (void) { - bucket *bp; + bucket *bp = NULL; int tokno = 1; - int i; int last_user_token_number; - static char DOLLAR[] = "$"; - - /* int lossage = 0; JF set but not used */ - tags = XCALLOC (char *, nsyms + 1); - tags[0] = DOLLAR; - user_toknums = XCALLOC (short, nsyms + 1); - user_toknums[0] = 0; - - sprec = XCALLOC (short, nsyms); - sassoc = XCALLOC (short, nsyms); + symbols = XCALLOC (bucket *, nsyms); max_user_token_number = 256; last_user_token_number = 256; @@ -1773,11 +1579,18 @@ packsymbols (void) } else if (bp->alias) { - /* this symbol and its alias are a single token defn. - allocate a tokno, and assign to both check agreement of - ->prec and ->assoc fields and make both the same */ - if (bp->value == 0) - bp->value = bp->alias->value = tokno++; + /* This symbol and its alias are a single token defn. + Allocate a tokno, and assign to both check agreement of + prec and assoc fields and make both the same */ + if (bp->value == -1) + { + if (bp == eoftoken || bp->alias == eoftoken) + bp->value = bp->alias->value = 0; + else + { + bp->value = bp->alias->value = tokno++; + } + } if (bp->prec != bp->alias->prec) { @@ -1803,101 +1616,90 @@ packsymbols (void) bp->assoc = bp->alias->assoc; } + /* Do not do processing below for SALIASs. */ if (bp->user_token_number == SALIAS) - continue; /* do not do processing below for SALIASs */ + continue; } - else /* bp->class == token_sym */ + else /* bp->class == token_sym */ { - bp->value = tokno++; + if (bp == eoftoken) + bp->value = 0; + else + bp->value = tokno++; } if (bp->class == token_sym) { - if (translations && !(bp->user_token_number)) + if (bp->user_token_number == SUNDEF) bp->user_token_number = ++last_user_token_number; if (bp->user_token_number > max_user_token_number) max_user_token_number = bp->user_token_number; } - tags[bp->value] = bp->tag; - user_toknums[bp->value] = bp->user_token_number; - sprec[bp->value] = bp->prec; - sassoc[bp->value] = bp->assoc; - + symbols[bp->value] = bp; } - if (translations) - { - int j; - - token_translations = XCALLOC (short, max_user_token_number + 1); - - /* initialize all entries for literal tokens to 2, the internal - token number for $undefined., which represents all invalid - inputs. */ - for (j = 0; j <= max_user_token_number; j++) - token_translations[j] = 2; - - for (bp = firstsymbol; bp; bp = bp->next) - { - if (bp->value >= ntokens) - continue; /* non-terminal */ - if (bp->user_token_number == SALIAS) - continue; - if (token_translations[bp->user_token_number] != 2) - complain (_("tokens %s and %s both assigned number %d"), - tags[token_translations[bp->user_token_number]], - bp->tag, bp->user_token_number); - token_translations[bp->user_token_number] = bp->value; - } - } + token_translations_init (); error_token_number = errtoken->value; - output_token_defines (&output_obstack); - obstack_1grow (&output_obstack, 0); - macro_insert ("tokendef", obstack_finish (&output_obstack)); - - /* if (!no_parser_flag) - output_token_defines (&table_obstack); */ - if (startval->class == unknown_sym) fatal (_("the start symbol %s is undefined"), startval->tag); else if (startval->class == token_sym) fatal (_("the start symbol %s is a token"), startval->tag); start_symbol = startval->value; +} + - if (defines_flag) +/*---------------------------------------------------------------. +| Save the definition of token names in the `TOKENDEFS' muscle. | +`---------------------------------------------------------------*/ + +static void +symbols_save (void) +{ + struct obstack tokendefs; + bucket *bp; + obstack_init (&tokendefs); + + for (bp = firstsymbol; bp; bp = bp->next) { - output_token_defines (&defines_obstack); + char *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; } + /* Don't #define nonliteral tokens whose names contain periods. */ + if (strchr (symbol, '.')) + continue; + + obstack_fgrow2 (&tokendefs, "# define %s\t%d\n", + symbol, bp->user_token_number); if (semantic_parser) - 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 + /* 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); } @@ -1913,14 +1715,12 @@ packgram (void) int ruleno; symbol_list *p; - bucket *ruleprec; + /* We use short to index items. */ + if (nitems >= MAXSHORT) + fatal (_("too many items (max %d)"), MAXSHORT); 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; + rules = XCALLOC (rule_t, nrules) - 1; itemno = 0; ruleno = 1; @@ -1928,9 +1728,15 @@ packgram (void) p = grammar; while (p) { - rlhs[ruleno] = p->sym->value; - rrhs[ruleno] = itemno; - ruleprec = p->ruleprec; + bucket *ruleprec = p->ruleprec; + rules[ruleno].lhs = p->sym->value; + rules[ruleno].rhs = itemno; + rules[ruleno].line = p->line; + rules[ruleno].useful = TRUE; + rules[ruleno].action = p->action; + rules[ruleno].action_line = p->action_line; + rules[ruleno].guard = p->guard; + rules[ruleno].guard_line = p->guard_line; p = p->next; while (p && p->sym) @@ -1940,8 +1746,8 @@ packgram (void) of the last token in it. */ if (p->sym->class == token_sym) { - rprec[ruleno] = p->sym->prec; - rassoc[ruleno] = p->sym->assoc; + rules[ruleno].prec = p->sym->prec; + rules[ruleno].assoc = p->sym->assoc; } if (p) p = p->next; @@ -1951,9 +1757,9 @@ packgram (void) the specified symbol's precedence replaces the default. */ if (ruleprec) { - rprec[ruleno] = ruleprec->prec; - rassoc[ruleno] = ruleprec->assoc; - rprecsym[ruleno] = ruleprec->value; + rules[ruleno].prec = ruleprec->prec; + rules[ruleno].assoc = ruleprec->assoc; + rules[ruleno].precsym = ruleprec->value; } ritem[itemno++] = -ruleno; @@ -1964,6 +1770,11 @@ packgram (void) } ritem[itemno] = 0; + nritems = itemno; + assert (nritems == nitems); + + if (trace_flag) + ritem_print (stderr); } /*-------------------------------------------------------------------. @@ -1980,21 +1791,10 @@ reader (void) start_flag = 0; startval = NULL; /* start symbol not specified yet. */ -#if 0 - /* initially assume token number translation not needed. */ - translations = 0; -#endif - /* Nowadays translations is always set to 1, since we give `error' a - user-token-number to satisfy the Posix demand for YYERRCODE==256. - */ - translations = 1; - - nsyms = 1; + nsyms = 0; nvars = 0; nrules = 0; nitems = 0; - rline_allocated = 10; - rline = XCALLOC (short, rline_allocated); typed = 0; lastprec = 0; @@ -2004,15 +1804,20 @@ reader (void) grammar = NULL; - init_lex (); + lex_init (); lineno = 1; - /* Initialize the macro obstack. */ - obstack_init (¯o_obstack); + /* Initialize the muscle obstack. */ + obstack_init (&muscle_obstack); /* Initialize the symbol table. */ tabinit (); + /* Construct the axiom symbol. */ + axiom = getsym ("$axiom"); + axiom->class = nterm_sym; + axiom->value = nvars++; + /* Construct the error token */ errtoken = getsym ("error"); errtoken->class = token_sym; @@ -2024,37 +1829,42 @@ reader (void) undeftoken->class = token_sym; undeftoken->user_token_number = 2; + /* Initialize the obstacks. */ + obstack_init (&action_obstack); + obstack_init (&attrs_obstack); + obstack_init (&output_obstack); + + finput = xfopen (infile, "r"); + /* Read the declaration section. Copy %{ ... %} groups to TABLE_OBSTACK and FDEFINES file. Also notice any %token, %left, etc. found there. */ read_declarations (); - /* Start writing the guard and action files, if they are needed. */ -#if 0 - output_headers (); -#endif + + /* If the user did not define her EOFTOKEN, do it now. */ + if (!eoftoken) + { + eoftoken = getsym ("$"); + eoftoken->class = token_sym; + /* Value specified by POSIX. */ + eoftoken->user_token_number = 0; + } + /* Read in the grammar, build grammar in list form. Write out guards and actions. */ readgram (); /* Some C code is given at the end of the grammar file. */ read_additionnal_code (); - /* Now we know whether we need the line-number stack. If we do, - write its type into the .tab.h file. */ -#if 0 - if (defines_flag) - reader_output_yylsp (&defines_obstack); -#endif - /* Write closing delimiters for actions and guards. */ -#if 0 - output_trailers (); - if (locations_flag) - obstack_sgrow (&table_obstack, "#define YYLSP_NEEDED 1\n\n"); -#endif + + lex_free (); + xfclose (finput); + /* Assign the symbols their symbol numbers. Write #defines for the token symbols into FDEFINES if requested. */ packsymbols (); + /* Save them. */ + symbols_save (); + /* Convert the grammar into the format described in gram.h. */ packgram (); - /* Free the symbol table data structure since symbols are now all - referred to by symbol number. */ - free_symtab (); }