X-Git-Url: https://git.saurik.com/bison.git/blobdiff_plain/11d82f03b4baf017d825166bd60b21f69131547c..3735969c7de7c12abfbe0a78e108399196e271e0:/src/reader.c diff --git a/src/reader.c b/src/reader.c index 3f3a3a34..cb013b20 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. @@ -26,8 +26,8 @@ #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" @@ -36,13 +36,11 @@ #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; bucket *ruleprec; } symbol_list; @@ -66,6 +64,19 @@ 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 +134,7 @@ 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; @@ -428,8 +439,11 @@ copy_definition (void) #if 0 if (!no_lines_flag) - obstack_fgrow2 (&attrs_obstack, "#line %d %s\n", - lineno, quotearg_style (c_quoting_style, infile)); + { + obstack_fgrow2 (&attrs_obstack, muscle_find ("linef"), + lineno, quotearg_style (c_quoting_style, + muscle_find("filename"))); + } #endif after_percent = 0; @@ -487,8 +501,8 @@ 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 +551,6 @@ 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 +575,12 @@ 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; } 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 @@ -741,30 +752,43 @@ parse_union_decl (void) { int c; int count = 0; + 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; - if (!no_lines_flag) - obstack_fgrow2 (&attrs_obstack, "\n#line %d %s\n", - lineno, quotearg_style (c_quoting_style, - muscle_find("filename"))); - else + /* 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_sgrow (&attrs_obstack, "typedef union"); + obstack_init (&union_obstack); + obstack_sgrow (&union_obstack, "union"); if (defines_flag) - obstack_sgrow (&defines_obstack, "typedef union"); + obstack_sgrow (&defines_obstack, prologue); c = getc (finput); while (c != EOF) { - obstack_1grow (&attrs_obstack, c); - if (defines_flag) - obstack_1grow (&defines_obstack, c); + /* If C contains '/', it is output by copy_comment (). */ + if (c != '/') + { + obstack_1grow (&union_obstack, c); + if (defines_flag) + obstack_1grow (&defines_obstack, c); + } switch (c) { @@ -773,7 +797,7 @@ parse_union_decl (void) break; case '/': - copy_comment2 (finput, &defines_obstack, &attrs_obstack); + copy_comment2 (finput, &defines_obstack, &union_obstack); break; case '{': @@ -786,19 +810,21 @@ parse_union_decl (void) count--; if (count <= 0) { - obstack_sgrow (&attrs_obstack, " YYSTYPE;\n"); if (defines_flag) - obstack_sgrow (&defines_obstack, " YYSTYPE;\n"); + 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; } } c = getc (finput); } + } @@ -846,9 +872,8 @@ 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) { @@ -878,8 +903,6 @@ parse_thong_decl (void) usrtoknum = numval; token = lex (); /* okay, did number, now get literal */ } - else - usrtoknum = 0; /* process literal string token */ @@ -900,8 +923,6 @@ parse_thong_decl (void) nsyms--; } -/* FIXME. */ - static void parse_muscle_decl (void) { @@ -919,7 +940,7 @@ parse_muscle_decl (void) copy_identifier (finput, &muscle_obstack); obstack_1grow (&muscle_obstack, 0); muscle_key = obstack_finish (&muscle_obstack); - + /* Read value. */ ch = skip_white_space (); if (ch != '"') @@ -943,44 +964,60 @@ parse_muscle_decl (void) } -/*----------------------------------. -| Parse what comes after %skeleton. | -`----------------------------------*/ -void -parse_skel_decl (void) +/*---------------------------------. +| Parse a double quoted parameter. | +`---------------------------------*/ + +static const char * +parse_dquoted_param (const char *from) { - /* Complete with parse_dquoted_param () on the CVS branch 1.29. */ -} + struct obstack param_obstack; + const char *param = NULL; + int c; -/*------------------------------------------. -| Parse what comes after %header_extension. | -`------------------------------------------*/ + obstack_init (¶m_obstack); + c = skip_white_space (); -static void -parse_header_extension_decl (void) -{ - char buff[32]; + if (c != '"') + { + complain (_("invalid %s declaration"), from); + ungetc (c, finput); + skip_to_char ('%'); + return NULL; + } + + for (;;) + { + if (literalchar (NULL, &c, '\"')) + obstack_1grow (¶m_obstack, c); + else + break; + } + + 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; + } - if (header_extension) - complain (_("multiple %%header_extension declarations")); - fscanf (finput, "%s", buff); - header_extension = xstrdup (buff); + 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"); } /*----------------------------------------------------------------. @@ -992,16 +1029,13 @@ 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) { @@ -1052,18 +1086,10 @@ read_declarations (void) parse_assoc_decl (non_assoc); break; - case tok_hdrext: - parse_header_extension_decl (); - break; - - case tok_srcext: - parse_source_extension_decl (); - break; - case tok_define: parse_muscle_decl (); break; - + case tok_skel: parse_skel_decl (); break; @@ -1071,6 +1097,13 @@ read_declarations (void) case tok_noop: break; + case tok_stropt: + case tok_intopt: + case tok_obsolete: + abort (); + break; + + case tok_illegal: default: complain (_("unrecognized: %s"), token_buffer); skip_to_char ('%'); @@ -1100,21 +1133,18 @@ 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; - sprintf (buf, "\ncase %d:\n", nrules); - obstack_grow (&action_obstack, buf, strlen (buf)); + obstack_fgrow1 (&action_obstack, "\ncase %d:\n", nrules); if (!no_lines_flag) { - sprintf (buf, "#line %d %s\n", - lineno, quotearg_style (c_quoting_style, - muscle_find ("filename"))); - obstack_grow (&action_obstack, buf, strlen (buf)); + obstack_fgrow2 (&action_obstack, muscle_find ("linef"), + lineno, quotearg_style (c_quoting_style, + muscle_find ("filename"))); } obstack_1grow (&action_obstack, '{'); @@ -1200,7 +1230,7 @@ copy_guard (symbol_list *rule, int stack_offset) obstack_fgrow1 (&guard_obstack, "\ncase %d:\n", nrules); if (!no_lines_flag) - obstack_fgrow2 (&guard_obstack, "#line %d %s\n", + obstack_fgrow2 (&guard_obstack, muscle_find ("linef"), lineno, quotearg_style (c_quoting_style, muscle_find ("filename"))); obstack_1grow (&guard_obstack, '{'); @@ -1278,20 +1308,6 @@ 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. | @@ -1434,10 +1450,7 @@ readgram (void) nrules++; nitems++; - record_rule_line (); - - p = XCALLOC (symbol_list, 1); - p->sym = lhs; + p = symbol_list_new (lhs); crule1 = p1; if (p1) @@ -1499,36 +1512,35 @@ readgram (void) 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 (); + 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. */ + /* 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); + 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; - p->sym = sdummy; - crule1 = XCALLOC (symbol_list, 1); - p->next = crule1; + /* 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 = XCALLOC (symbol_list, 1); - p->sym = sdummy; + p = symbol_list_new (sdummy); p1->next = p; p1 = p; @@ -1538,8 +1550,7 @@ readgram (void) if (t == tok_identifier) { nitems++; - p = XCALLOC (symbol_list, 1); - p->sym = symval; + p = symbol_list_new (symval); p1->next = p; p1 = p; } @@ -1553,7 +1564,7 @@ readgram (void) } /* 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); + p = symbol_list_new (NULL); p1->next = p; p1 = p; @@ -1676,12 +1687,12 @@ read_additionnal_code (void) { char c; struct obstack el_obstack; - + obstack_init (&el_obstack); while ((c = getc (finput)) != EOF) obstack_1grow (&el_obstack, c); - + obstack_1grow (&el_obstack, 0); muscle_insert ("epilogue", obstack_finish (&el_obstack)); } @@ -1727,14 +1738,55 @@ output_token_defines (struct obstack *oout) continue; obstack_fgrow2 (oout, "# define\t%s\t%d\n", - symbol, - (translations ? bp->user_token_number : bp->value)); + 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. | +`------------------------------------------------------------------*/ + +static void +token_translations_init (void) +{ + 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 $undefined., which represents all invalid + inputs. */ + for (i = 0; i <= max_user_token_number; i++) + token_translations[i] = 2; + + for (bp = firstsymbol; bp; bp = bp->next) + { + /* Non-terminal? */ + if (bp->value >= ntokens) + continue; + /* A token string alias? */ + if (bp->user_token_number == SALIAS) + continue; + + assert (bp->user_token_number != SUNDEF); + + /* A token which translation has already been set? */ + if (token_translations[bp->user_token_number] != 2) + complain (_("tokens %s and %s both assigned number %d"), + tags[token_translations[bp->user_token_number]], + bp->tag, bp->user_token_number); + token_translations[bp->user_token_number] = bp->value; + } +} + + /*------------------------------------------------------------------. | Assign symbol numbers, and write definition of token names into | | FDEFINES. Set up vectors TAGS and SPREC of names and precedences | @@ -1744,20 +1796,21 @@ output_token_defines (struct obstack *oout) static void packsymbols (void) { - bucket *bp; + bucket *bp = NULL; int tokno = 1; - int i; int last_user_token_number; static char DOLLAR[] = "$"; 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); + /* The EOF token. */ + tags[0] = DOLLAR; + user_toknums[0] = 0; + max_user_token_number = 256; last_user_token_number = 256; @@ -1810,7 +1863,7 @@ packsymbols (void) 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; @@ -1820,53 +1873,42 @@ packsymbols (void) user_toknums[bp->value] = bp->user_token_number; sprec[bp->value] = bp->prec; sassoc[bp->value] = bp->assoc; - } - if (translations) - { - int j; + token_translations_init (); - token_translations = XCALLOC (short, max_user_token_number + 1); + error_token_number = errtoken->value; - /* 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; + 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); - 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; - } - } + start_symbol = startval->value; +} - error_token_number = errtoken->value; - output_token_defines (&output_obstack); - obstack_1grow (&output_obstack, 0); - muscle_insert ("tokendef", obstack_finish (&output_obstack)); +/*-----------------------------------. +| Output definition of token names. | +`-----------------------------------*/ + +static void +symbols_output (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 (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) { output_token_defines (&defines_obstack); @@ -1882,19 +1924,23 @@ packsymbols (void) } 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); - } + { + int i; + + 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; + /* `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 + } } } @@ -1911,14 +1957,8 @@ 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; @@ -1926,9 +1966,11 @@ 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; p = p->next; while (p && p->sym) @@ -1938,8 +1980,8 @@ 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; @@ -1949,9 +1991,9 @@ 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; @@ -1962,6 +2004,9 @@ packgram (void) } ritem[itemno] = 0; + + if (trace_flag) + ritem_print (stderr); } /*-------------------------------------------------------------------. @@ -1978,21 +2023,10 @@ 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; nvars = 0; nrules = 0; nitems = 0; - rline_allocated = 10; - rline = XCALLOC (short, rline_allocated); typed = 0; lastprec = 0; @@ -2002,7 +2036,7 @@ reader (void) grammar = NULL; - init_lex (); + lex_init (); lineno = 1; /* Initialize the muscle obstack. */ @@ -2031,18 +2065,16 @@ reader (void) 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 its type into the .tab.h file. + This is no longer need with header skeleton. */ + /* Assign the symbols their symbol numbers. Write #defines for the token symbols into FDEFINES if requested. */ packsymbols (); /* 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 (); + /* Output the headers. */ + symbols_output (); }