X-Git-Url: https://git.saurik.com/bison.git/blobdiff_plain/ff4a34be273490d08c7b34a51e309a76c5b2c127..e7b8bef1b1fa6eae346c4a3ee00e897a4d6f9058:/src/reader.c diff --git a/src/reader.c b/src/reader.c index 7c368621..bebbc153 100644 --- a/src/reader.c +++ b/src/reader.c @@ -1,5 +1,5 @@ /* Input parser for bison - Copyright (C) 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. @@ -21,28 +21,35 @@ #include "system.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 "quote.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; @@ -61,11 +68,26 @@ static int typed; /* Incremented for each %left, %right or %nonassoc seen */ static int lastprec; -/* Incremented for each generated symbol */ -static int gensym_count; +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; +} -static bucket *errtoken; -static bucket *undeftoken; /*===================\ @@ -117,17 +139,53 @@ read_signed_integer (FILE *stream) return sign * n; } -/*-------------------------------------------------------------------. -| Dump the string from FINPUT to FOUTPUT. MATCH is the delimiter of | -| the string (either ' or "). | -`-------------------------------------------------------------------*/ +/*--------------------------------------------------------------. +| Get the data type (alternative in the union) of the value for | +| symbol N in rule RULE. | +`--------------------------------------------------------------*/ + +static char * +get_type_name (int n, symbol_list *rule) +{ + int i; + symbol_list *rp; + + if (n < 0) + { + complain (_("invalid $ value")); + return NULL; + } + + rp = rule; + i = 0; + + while (i < n) + { + rp = rp->next; + if (rp == NULL || rp->sym == NULL) + { + complain (_("invalid $ value")); + return NULL; + } + i++; + } + + return rp->sym->type_name; +} + +/*------------------------------------------------------------. +| Dump the string from FIN to OOUT if non null. MATCH is the | +| delimiter of the string (either ' or "). | +`------------------------------------------------------------*/ static inline void -copy_string (FILE *fin, FILE *fout, int match) +copy_string2 (FILE *fin, struct obstack *oout, int match, int store) { int c; - putc (match, fout); + if (store) + obstack_1grow (oout, match); + c = getc (fin); while (c != match) @@ -142,14 +200,15 @@ copy_string (FILE *fin, FILE *fout, int match) continue; } - putc (c, fout); + obstack_1grow (oout, c); if (c == '\\') { c = getc (fin); if (c == EOF) fatal (_("unterminated string at end of file")); - putc (c, fout); + obstack_1grow (oout, c); + if (c == '\n') lineno++; } @@ -157,27 +216,63 @@ copy_string (FILE *fin, FILE *fout, int match) c = getc (fin); } - putc (c, fout); + if (store) + obstack_1grow (oout, c); } +/* FIXME. */ -/*---------------------------------------------------------------. -| Dump the comment from IN to OUT1 and OUT2. C is either `*' or | -| `/', depending upon the type of comments used. OUT2 might be | -| NULL. | -`---------------------------------------------------------------*/ +static inline void +copy_string (FILE *fin, struct obstack *oout, int match) +{ + copy_string2 (fin, oout, match, 1); +} + +/* FIXME. */ + +static inline void +copy_identifier (FILE *fin, struct obstack *oout) +{ + int c; + + while (isalnum (c = getc (fin)) || c == '_') + obstack_1grow (oout, c); + + ungetc (c, fin); +} + + +/*------------------------------------------------------------------. +| 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 *in, FILE *out1, FILE *out2, int c) +copy_comment (FILE *fin, struct obstack *oout) { int cplus_comment; int ended; + int c; + + /* We read a `/', output it. */ + obstack_1grow (oout, '/'); + + switch ((c = getc (fin))) + { + case '/': + cplus_comment = 1; + break; + case '*': + cplus_comment = 0; + break; + default: + ungetc (c, fin); + return; + } - cplus_comment = (c == '/'); - putc (c, out1); - if (out2) - putc (c, out2); - c = getc (in); + obstack_1grow (oout, c); + c = getc (fin); ended = 0; while (!ended) @@ -186,72 +281,52 @@ copy_comment2 (FILE *in, FILE *out1, FILE *out2, int c) { while (c == '*') { - putc (c, out1); - if (out2) - putc (c, out2); - c = getc (in); + obstack_1grow (oout, c); + c = getc (fin); } if (c == '/') { - putc (c, out1); - if (out2) - putc (c, out2); + obstack_1grow (oout, c); ended = 1; } } else if (c == '\n') { lineno++; - putc (c, out1); - if (out2) - putc (c, out2); + obstack_1grow (oout, c); if (cplus_comment) ended = 1; else - c = getc (in); + c = getc (fin); } else if (c == EOF) fatal (_("unterminated comment")); else { - putc (c, out1); - if (out2) - putc (c, out2); - c = getc (in); + obstack_1grow (oout, c); + c = getc (fin); } } } -/*------------------------------------------------------------. -| Dump the comment from FIN to FOUT. C is either `*' or `/', | -| depending upon the type of comments used. | -`------------------------------------------------------------*/ - -static inline void -copy_comment (FILE *fin, FILE *fout, int c) -{ - copy_comment2 (fin, fout, NULL, c); -} - - /*-----------------------------------------------------------------. -| FIN is pointing to a location (i.e., a `@'). Output to FOUT a | +| 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 `$0' with | | respect to the top of the stack. | `-----------------------------------------------------------------*/ static inline void -copy_at (FILE *fin, FILE *fout, int stack_offset) +copy_at (FILE *fin, struct obstack *oout, int stack_offset) { int c; c = getc (fin); if (c == '$') { - fprintf (fout, "yyloc"); + obstack_sgrow (oout, "yyloc"); locations_flag = 1; } else if (isdigit (c) || c == '-') @@ -260,9 +335,16 @@ copy_at (FILE *fin, FILE *fout, int stack_offset) ungetc (c, fin); n = read_signed_integer (fin); - - fprintf (fout, "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 { @@ -271,6 +353,78 @@ copy_at (FILE *fin, FILE *fout, int stack_offset) complain (_("%s is invalid"), quote (buf)); } } + + +/*-------------------------------------------------------------------. +| 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 `$0' with respect to the top of the stack. | +`-------------------------------------------------------------------*/ + +static inline void +copy_dollar (FILE *fin, struct obstack *oout, + symbol_list *rule, int stack_offset) +{ + int c = getc (fin); + const char *type_name = NULL; + + /* Get the type name if explicit. */ + if (c == '<') + { + read_type_name (fin); + type_name = token_buffer; + value_components_used = 1; + c = getc (fin); + } + + if (c == '$') + { + 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); + } + else if (isdigit (c) || c == '-') + { + int n; + ungetc (c, fin); + n = read_signed_integer (fin); + + 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 + { + char buf[] = "$c"; + buf[1] = c; + complain (_("%s is invalid"), quote (buf)); + } +} /*-------------------------------------------------------------------. | Copy the contents of a `%{ ... %}' into the definitions file. The | @@ -285,7 +439,11 @@ copy_definition (void) int after_percent; if (!no_lines_flag) - fprintf (fattrs, "#line %d \"%s\"\n", lineno, infile); + { + obstack_fgrow2 (&attrs_obstack, muscle_find ("linef"), + lineno, quotearg_style (c_quoting_style, + muscle_find ("filename"))); + } after_percent = 0; @@ -296,7 +454,7 @@ copy_definition (void) switch (c) { case '\n': - putc (c, fattrs); + obstack_1grow (&attrs_obstack, c); lineno++; break; @@ -306,22 +464,18 @@ copy_definition (void) case '\'': case '"': - copy_string (finput, fattrs, c); + copy_string (finput, &attrs_obstack, c); break; case '/': - putc (c, fattrs); - c = getc (finput); - if (c != '*' && c != '/') - continue; - copy_comment (finput, fattrs, c); + copy_comment (finput, &attrs_obstack); break; case EOF: fatal ("%s", _("unterminated `%{' definition")); default: - putc (c, fattrs); + obstack_1grow (&attrs_obstack, c); } c = getc (finput); @@ -330,12 +484,10 @@ copy_definition (void) { if (c == '}') return; - putc ('%', fattrs); + obstack_1grow (&attrs_obstack, '%'); } after_percent = 0; - } - } @@ -348,32 +500,38 @@ copy_definition (void) static void parse_token_decl (symbol_class what_is, symbol_class what_is_not) { - int token = 0; - char *typename = 0; - struct bucket *symbol = NULL; /* pts to symbol being defined */ + token_t token = tok_undef; + char *typename = NULL; + + /* The symbol being defined. */ + struct bucket *symbol = NULL; + /* After `%token' and `%nterm', any number of symbols maybe be + defined. */ for (;;) { int tmp_char = ungetc (skip_white_space (), finput); + /* `%' (for instance from `%token', or from `%%' etc.) is the + only valid means to end this declaration. */ if (tmp_char == '%') return; if (tmp_char == EOF) fatal (_("Premature EOF after %s"), token_buffer); token = lex (); - if (token == COMMA) + if (token == tok_comma) { symbol = NULL; continue; } - if (token == TYPENAME) + if (token == tok_typename) { typename = xstrdup (token_buffer); value_components_used = 1; symbol = NULL; } - else if (token == IDENTIFIER && *symval->tag == '\"' && symbol) + else if (token == tok_identifier && *symval->tag == '\"' && symbol) { if (symval->alias) warn (_("symbol `%s' used more than once as a literal string"), @@ -392,10 +550,9 @@ 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 == IDENTIFIER) + else if (token == tok_identifier) { int oldclass = symval->class; symbol = symval; @@ -414,15 +571,18 @@ parse_token_decl (symbol_class what_is, symbol_class what_is_not) complain (_("type redeclaration for %s"), symbol->tag); } } - else if (symbol && token == NUMBER) + 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 ('%'); } } @@ -439,7 +599,7 @@ parse_start_decl (void) { if (start_flag) complain (_("multiple %s declarations"), "%start"); - if (lex () != IDENTIFIER) + if (lex () != tok_identifier) complain (_("invalid %s declaration"), "%start"); else { @@ -448,43 +608,6 @@ parse_start_decl (void) } } - - -/*--------------------------------------------------------------. -| Get the data type (alternative in the union) of the value for | -| symbol n in rule rule. | -`--------------------------------------------------------------*/ - -static char * -get_type_name (int n, symbol_list * rule) -{ - int i; - symbol_list *rp; - - if (n < 0) - { - complain (_("invalid $ value")); - return NULL; - } - - rp = rule; - i = 0; - - while (i < n) - { - rp = rp->next; - if (rp == NULL || rp->sym == NULL) - { - complain (_("invalid $ value")); - return NULL; - } - i++; - } - - return rp->sym->type_name; -} - - /*-----------------------------------------------------------. | read in a %type declaration and record its information for | | get_type_name to access | @@ -495,7 +618,7 @@ parse_type_decl (void) { char *name; - if (lex () != TYPENAME) + if (lex () != tok_typename) { complain ("%s", _("%type declaration has no ")); skip_to_char ('%'); @@ -506,7 +629,7 @@ parse_type_decl (void) for (;;) { - int t; + token_t t; int tmp_char = ungetc (skip_white_space (), finput); if (tmp_char == '%') @@ -519,11 +642,11 @@ parse_type_decl (void) switch (t) { - case COMMA: - case SEMICOLON: + case tok_comma: + case tok_semicolon: break; - case IDENTIFIER: + case tok_identifier: if (symval->type_name == NULL) symval->type_name = name; else if (strcmp (name, symval->type_name) != 0) @@ -556,7 +679,7 @@ parse_assoc_decl (associativity assoc) for (;;) { - int t; + token_t t; int tmp_char = ungetc (skip_white_space (), finput); if (tmp_char == '%') @@ -568,14 +691,14 @@ parse_assoc_decl (associativity assoc) switch (t) { - case TYPENAME: + case tok_typename: name = xstrdup (token_buffer); break; - case COMMA: + case tok_comma: break; - case IDENTIFIER: + case tok_identifier: if (symval->prec != 0) complain (_("redefining precedence of %s"), symval->tag); symval->prec = lastprec; @@ -592,11 +715,10 @@ parse_assoc_decl (associativity assoc) } break; - case NUMBER: - if (prev == IDENTIFIER) + case tok_number: + if (prev == tok_identifier) { symval->user_token_number = numval; - translations = 1; } else { @@ -607,7 +729,7 @@ token_buffer); } break; - case SEMICOLON: + case tok_semicolon: return; default: @@ -616,45 +738,39 @@ token_buffer); } prev = t; - } } -/*-------------------------------------------------------------------. -| Copy the union declaration into fattrs (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 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) - fprintf (fattrs, "\n#line %d \"%s\"\n", lineno, infile); - else - fprintf (fattrs, "\n"); - - fprintf (fattrs, "typedef union"); - if (fdefines) - fprintf (fdefines, "typedef union"); + obstack_init (&union_obstack); + obstack_sgrow (&union_obstack, "union"); - c = getc (finput); - - while (c != EOF) + while (!done) { - putc (c, fattrs); - if (fdefines) - putc (c, fdefines); + c = xgetc (finput); + + /* If C contains '/', it is output by copy_comment (). */ + if (c != '/') + obstack_1grow (&union_obstack, c); switch (c) { @@ -663,36 +779,30 @@ parse_union_decl (void) break; case '/': - c = getc (finput); - if (c != '*' && c != '/') - continue; - copy_comment2 (finput, fattrs, fdefines, c); + copy_comment (finput, &union_obstack); break; - case '{': count++; break; case '}': + /* FIXME: Errr. How could this happen???. --akim */ if (count == 0) complain (_("unmatched %s"), "`}'"); count--; - if (count <= 0) - { - fprintf (fattrs, " YYSTYPE;\n"); - if (fdefines) - fprintf (fdefines, " 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)); } @@ -704,28 +814,13 @@ parse_union_decl (void) static void parse_expect_decl (void) { - int c; - size_t count; - char buffer[20]; - - c = getc (finput); - while (c == ' ' || c == '\t') - c = getc (finput); - - count = 0; - while (c >= '0' && c <= '9') - { - if (count < sizeof(buffer) - 1) - buffer[count++] = c; - c = getc (finput); - } - buffer[count] = 0; - + int c = skip_white_space (); ungetc (c, finput); - if (count <= 0 || count > 10) - complain ("%s", _("argument of %expect is not an integer")); - expected_conflicts = atoi (buffer); + if (!isdigit (c)) + complain (_("argument of %%expect is not an integer")); + else + expected_conflicts = read_signed_integer (finput); } @@ -752,14 +847,13 @@ parse_expect_decl (void) static void parse_thong_decl (void) { - int token; + 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 == TYPENAME) + if (token == tok_typename) { typename = xstrdup (token_buffer); value_components_used = 1; @@ -768,7 +862,7 @@ parse_thong_decl (void) /* process first token */ - if (token != IDENTIFIER) + if (token != tok_identifier) { complain (_("unrecognized item %s, expected an identifier"), token_buffer); @@ -782,17 +876,15 @@ parse_thong_decl (void) token = lex (); /* get number or literal string */ - if (token == NUMBER) + if (token == tok_number) { usrtoknum = numval; token = lex (); /* okay, did number, now get literal */ } - else - usrtoknum = 0; /* process literal string token */ - if (token != IDENTIFIER || *symval->tag != '\"') + if (token != tok_identifier || *symval->tag != '\"') { complain (_("expected string constant instead of %s"), token_buffer); skip_to_char ('%'); @@ -805,93 +897,187 @@ parse_thong_decl (void) symval->alias = symbol; symbol->alias = symval; - nsyms--; /* symbol and symval combined are only one symbol */ + /* symbol and symval combined are only one symbol. */ + nsyms--; +} + + +static void +parse_muscle_decl (void) +{ + int ch = ungetc (skip_white_space (), finput); + char *muscle_key; + char *muscle_value; + + /* Read key. */ + if (!isalpha (ch) && ch != '_') + { + complain (_("invalid %s declaration"), "%define"); + skip_to_char ('%'); + return; + } + copy_identifier (finput, &muscle_obstack); + obstack_1grow (&muscle_obstack, 0); + muscle_key = obstack_finish (&muscle_obstack); + + /* Read value. */ + ch = skip_white_space (); + if (ch != '"') + { + ungetc (ch, finput); + if (ch != EOF) + { + complain (_("invalid %s declaration"), "%define"); + skip_to_char ('%'); + return; + } + else + fatal (_("Premature EOF after %s"), "\""); + } + 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. */ + muscle_insert (muscle_key, muscle_value); +} + + + +/*---------------------------------. +| Parse a double quoted parameter. | +`---------------------------------*/ + +static const char * +parse_dquoted_param (const char *from) +{ + struct obstack param_obstack; + const char *param = NULL; + int c; + + obstack_init (¶m_obstack); + c = skip_white_space (); + + if (c != '"') + { + complain (_("invalid %s declaration"), from); + ungetc (c, finput); + skip_to_char ('%'); + return NULL; + } + + while ((c = literalchar ()) != '"') + obstack_1grow (¶m_obstack, c); + + obstack_1grow (¶m_obstack, '\0'); + param = obstack_finish (¶m_obstack); + + if (c != '"' || strlen (param) == 0) + { + complain (_("invalid %s declaration"), from); + if (c != '"') + ungetc (c, finput); + skip_to_char ('%'); + return NULL; + } + + return param; } +/*----------------------------------. +| Parse what comes after %skeleton. | +`----------------------------------*/ + +static void +parse_skel_decl (void) +{ + skeleton = parse_dquoted_param ("%skeleton"); +} /*----------------------------------------------------------------. | Read from finput until `%%' is seen. Discard the `%%'. Handle | | any `%' declarations, and copy the contents of any `%{ ... %}' | -| groups to fattrs. | +| groups to ATTRS_OBSTACK. | `----------------------------------------------------------------*/ 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) { - case TWO_PERCENTS: + case tok_two_percents: return; - case PERCENT_LEFT_CURLY: + case tok_percent_left_curly: copy_definition (); break; - case TOKEN: + case tok_token: parse_token_decl (token_sym, nterm_sym); break; - case NTERM: + case tok_nterm: parse_token_decl (nterm_sym, token_sym); break; - case TYPE: + case tok_type: parse_type_decl (); break; - case START: + case tok_start: parse_start_decl (); break; - case UNION: + case tok_union: parse_union_decl (); break; - case EXPECT: + case tok_expect: parse_expect_decl (); break; - case THONG: + + case tok_thong: parse_thong_decl (); break; - case LEFT: + case tok_left: parse_assoc_decl (left_assoc); break; - case RIGHT: + case tok_right: parse_assoc_decl (right_assoc); break; - case NONASSOC: + case tok_nonassoc: parse_assoc_decl (non_assoc); break; - case SEMANTIC_PARSER: - if (semantic_parser == 0) - { - semantic_parser = 1; - open_extra_files (); - } + case tok_define: + parse_muscle_decl (); + break; + + case tok_skel: + parse_skel_decl (); break; - case PURE_PARSER: - pure_parser = 1; + case tok_noop: break; - case NOOP: + case tok_stropt: + case tok_intopt: + case tok_obsolete: + abort (); break; + case tok_illegal: default: complain (_("unrecognized: %s"), token_buffer); skip_to_char ('%'); @@ -914,311 +1100,93 @@ 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 n; int count; - char *type_name; - - /* offset is always 0 if parser has already popped the stack pointer */ - if (semantic_parser) - stack_offset = 0; - - fprintf (faction, "\ncase %d:\n", nrules); - if (!no_lines_flag) - fprintf (faction, "#line %d \"%s\"\n", lineno, infile); - putc ('{', faction); count = 1; - c = getc (finput); - while (count > 0) { - while (c != '}') - { - switch (c) - { - case '\n': - putc (c, faction); - lineno++; - break; - - case '{': - putc (c, faction); - count++; - break; - - case '\'': - case '"': - copy_string (finput, faction, c); - break; - - case '/': - putc (c, faction); - c = getc (finput); - if (c != '*' && c != '/') - continue; - copy_comment (finput, faction, c); - break; - - case '$': - c = getc (finput); - type_name = NULL; - - if (c == '<') - { - char *cp = token_buffer; + while ((c = getc (finput)) != '}') + switch (c) + { + case '\n': + obstack_1grow (&action_obstack, c); + lineno++; + break; - while ((c = getc (finput)) != '>' && c > 0) - { - if (cp == token_buffer + maxtoken) - cp = grow_token_buffer (cp); + case '{': + obstack_1grow (&action_obstack, c); + count++; + break; - *cp++ = c; - } - *cp = 0; - type_name = token_buffer; - value_components_used = 1; + case '\'': + case '"': + copy_string (finput, &action_obstack, c); + break; - c = getc (finput); - } - if (c == '$') - { - fprintf (faction, "yyval"); - if (!type_name) - type_name = get_type_name (0, rule); - if (type_name) - fprintf (faction, ".%s", type_name); - if (!type_name && typed) - complain (_("$$ of `%s' has no declared type"), - rule->sym->tag); - } - else if (isdigit (c) || c == '-') - { - ungetc (c, finput); - n = read_signed_integer (finput); - c = getc (finput); - - if (!type_name && n > 0) - type_name = get_type_name (n, rule); - - fprintf (faction, "yyvsp[%d]", n - stack_offset); - if (type_name) - fprintf (faction, ".%s", type_name); - if (!type_name && typed) - complain (_("$%d of `%s' has no declared type"), - n, rule->sym->tag); - continue; - } - else - { - char buf[] = "$c"; - buf[1] = c; - complain (_("%s is invalid"), quote (buf)); - } + case '/': + copy_comment (finput, &action_obstack); + break; - break; + case '$': + copy_dollar (finput, &action_obstack, + rule, stack_offset); + break; - case '@': - copy_at (finput, faction, stack_offset); - break; + case '@': + copy_at (finput, &action_obstack, + stack_offset); + break; - case EOF: - fatal (_("unmatched %s"), "`{'"); + case EOF: + fatal (_("unmatched %s"), "`{'"); - default: - putc (c, faction); - } - - c = getc (finput); - } - - /* above loop exits when c is '}' */ + default: + obstack_1grow (&action_obstack, c); + } + /* Above loop exits when C is '}'. */ if (--count) { - putc (c, faction); + obstack_1grow (&action_obstack, c); c = getc (finput); } } - fprintf (faction, ";\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 `$0' with | -| respect to the top of the stack, for the simple parser in which | -| the stack is not popped until after the guard is run. | -`-------------------------------------------------------------------*/ + static void -copy_guard (symbol_list * rule, int stack_offset) +parse_action (symbol_list *rule, int stack_offset) { - int c; - int n; - int count; - char *type_name; - int brace_flag = 0; - - /* offset is always 0 if parser has already popped the stack pointer */ - if (semantic_parser) - stack_offset = 0; - - fprintf (fguard, "\ncase %d:\n", nrules); - if (!no_lines_flag) - fprintf (fguard, "#line %d \"%s\"\n", lineno, infile); - putc ('{', fguard); - - count = 0; - c = getc (finput); - - while (brace_flag ? (count > 0) : (c != ';')) - { - switch (c) - { - case '\n': - putc (c, fguard); - lineno++; - break; - - case '{': - putc (c, fguard); - brace_flag = 1; - count++; - break; - - case '}': - putc (c, fguard); - if (count > 0) - count--; - else - { - complain (_("unmatched %s"), "`}'"); - c = getc (finput); /* skip it */ - } - break; - - case '\'': - case '"': - copy_string (finput, fguard, c); - break; - - case '/': - putc (c, fguard); - c = getc (finput); - if (c != '*' && c != '/') - continue; - copy_comment (finput, fguard, c); - break; - - case '$': - c = getc (finput); - type_name = NULL; - - if (c == '<') - { - char *cp = token_buffer; - - while ((c = getc (finput)) != '>' && c > 0) - { - if (cp == token_buffer + maxtoken) - cp = grow_token_buffer (cp); - - *cp++ = c; - } - *cp = 0; - type_name = token_buffer; - - c = getc (finput); - } - - if (c == '$') - { - fprintf (fguard, "yyval"); - if (!type_name) - type_name = rule->sym->type_name; - if (type_name) - fprintf (fguard, ".%s", type_name); - if (!type_name && typed) - complain (_("$$ of `%s' has no declared type"), - rule->sym->tag); - } - else if (isdigit (c) || c == '-') - { - ungetc (c, finput); - n = read_signed_integer (finput); - c = getc (finput); - - if (!type_name && n > 0) - type_name = get_type_name (n, rule); - - fprintf (fguard, "yyvsp[%d]", n - stack_offset); - if (type_name) - fprintf (fguard, ".%s", type_name); - if (!type_name && typed) - complain (_("$%d of `%s' has no declared type"), - n, rule->sym->tag); - continue; - } - else - { - char buf[] = "$c"; - buf[1] = c; - complain (_("%s is invalid"), quote (buf)); - } - break; - - case '@': - copy_at (finput, fguard, stack_offset); - break; - - case EOF: - fatal ("%s", _("unterminated %guard clause")); - - default: - putc (c, fguard); - } - - if (c != '}' || count != 0) - c = getc (finput); - } - - c = skip_white_space (); - - fprintf (fguard, ";\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 | @@ -1228,365 +1196,266 @@ record_rule_line (void) static bucket * gensym (void) { + /* Incremented for each generated symbol */ + static int gensym_count = 0; + static char buf[256]; + bucket *sym; - sprintf (token_buffer, "@%d", ++gensym_count); + sprintf (buf, "@%d", ++gensym_count); + token_buffer = buf; sym = getsym (token_buffer); sym->class = nterm_sym; 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; - int t; - char *name; - - t = lex (); - - if (t != TYPENAME) - { - complain (_("invalid %s declaration"), "%type"); - return t; - } - - name = xstrdup (token_buffer); - - for (;;) - { - t = lex (); - - switch (t) - { - case SEMICOLON: - return lex (); - - case COMMA: - break; - - case 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 t; - } - } -} - -#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) { - int t; + 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 != TWO_PERCENTS && t != ENDFILE) - { - if (t == IDENTIFIER || t == 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 == IDENTIFIER) - { - lhs = symval; - - if (!start_flag) - { - startval = lhs; - start_flag = 1; - } - - t = lex (); - if (t != COLON) - { - complain (_("ill-formed rule: initial symbol not followed by colon")); - unlex (t); - } - } + 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 (nrules == 0 && t == BAR) - { - complain (_("grammar starts with vertical bar")); - lhs = symval; /* BOGUS: use a random symval */ - } - /* start a new rule and record its lhs. */ + if (t == tok_identifier) + { + lhs = symval; - nrules++; - nitems++; + if (!start_flag) + { + startval = lhs; + start_flag = 1; + } - record_rule_line (); + t = lex (); + if (t != tok_colon) + { + complain (_("ill-formed rule: initial symbol not followed by colon")); + unlex (t); + } + } - p = XCALLOC (symbol_list, 1); - p->sym = lhs; + 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. */ - crule1 = p1; - if (p1) - p1->next = p; - else - grammar = p; + nrules++; + nitems++; - p1 = p; - crule = p; + p = symbol_list_new (lhs); - /* mark the rule's lhs as a nonterminal if not already so. */ + crule1 = p1; + if (p1) + p1->next = p; + else + grammar = p; - 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); + p1 = p; + crule = p; - /* read the rhs of the rule. */ + /* mark the rule's lhs as a nonterminal if not already so. */ - for (;;) - { - t = lex (); - if (t == PREC) - { - t = lex (); - crule->ruleprec = symval; - t = lex (); - } - - if (!(t == IDENTIFIER || t == LEFT_CURLY)) - break; - - /* If next token is an identifier, see if a colon follows it. - If one does, exit this rule now. */ - if (t == IDENTIFIER) - { - bucket *ssave; - int t1; - - ssave = symval; - t1 = lex (); - unlex (t1); - symval = ssave; - if (t1 == 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 == 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 == PREC) - { - complain (_("two @prec's in a row")); - t = lex (); - crule->ruleprec = symval; - t = lex (); - } - if (t == 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 == 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 $$ but we need one. */ - else if (!xactions && !first_rhs && lhs->type_name != 0) - complain (_("empty rule for typed nonterminal, and no action")); - if (t == SEMICOLON) + parse_guard (crule, rulelength); t = lex (); - } -#if 0 - /* these things can appear as alternatives to rules. */ -/* NO, they cannot. - a) none of the documentation allows them - b) most of them scan forward until finding a next % - thus they may swallow lots of intervening rules -*/ - else if (t == TOKEN) - { - parse_token_decl (token_sym, nterm_sym); - t = lex (); - } - else if (t == NTERM) - { - parse_token_decl (nterm_sym, token_sym); - t = lex (); - } - else if (t == TYPE) - { - t = get_type (); - } - else if (t == UNION) - { - parse_union_decl (); - t = lex (); - } - else if (t == EXPECT) - { - parse_expect_decl (); - t = lex (); - } - else if (t == START) - { - parse_start_decl (); - t = lex (); - } -#endif + } - else - { - complain (_("invalid input: %s"), token_buffer); + if (t == tok_left_curly) + { + /* This case never occurs -wjh */ + if (action_flag) + complain (_("two actions at end of one rule")); + parse_action (crule, rulelength); + action_flag = 1; + xactions++; /* -wjh */ + t = lex (); + } + /* If $$ 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")); - /* 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. */ - fprintf (fattrs, "#ifndef YYSTYPE\n#define YYSTYPE int\n#endif\n"); - if (fdefines) - fprintf (fdefines, "#ifndef YYSTYPE\n#define YYSTYPE int\n#endif\n"); - } - /* Report any undefined symbols and consider them nonterminals. */ for (bp = firstsymbol; bp; bp = bp->next) @@ -1599,56 +1468,91 @@ 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; } - -/*--------------------------------------------------------------. -| For named tokens, but not literal ones, define the name. The | -| value is the user token number. | -`--------------------------------------------------------------*/ +/* At the end of the grammar file, some C source code must + be stored. It is going to be associated to the epilogue + directive. */ static void -output_token_defines (FILE *file) +read_additionnal_code (void) { - bucket *bp; - char *cp, *symbol; char c; + struct obstack el_obstack; - for (bp = firstsymbol; bp; bp = bp->next) + obstack_init (&el_obstack); + + if (!no_lines_flag) { - symbol = bp->tag; /* get symbol */ + 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); + + obstack_1grow (&el_obstack, 0); + muscle_insert ("epilogue", obstack_finish (&el_obstack)); +} + +/*------------------------------------------------------------------. +| 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 $undefined., which represents all invalid + inputs. */ + for (i = 0; i <= max_user_token_number; i++) + token_translations[i] = 2; + + for (bp = firstsymbol; bp; bp = bp->next) + { + /* Non-terminal? */ if (bp->value >= ntokens) continue; + /* A token string alias? */ if (bp->user_token_number == SALIAS) continue; - if ('\'' == *symbol) - continue; /* skip literal character */ - if (bp == errtoken) - continue; /* skip error token */ - if ('\"' == *symbol) - { - /* use literal string only if given a symbol with an alias */ - if (bp->alias) - symbol = bp->alias->tag; - else - continue; - } - /* Don't #define nonliteral tokens whose names contain periods. */ - cp = symbol; - while ((c = *cp++) && c != '.'); - if (c != '\0') - continue; + assert (bp->user_token_number != SUNDEF); - fprintf (file, "#define\t%s\t%d\n", symbol, - ((translations && !raw_flag) - ? bp->user_token_number : bp->value)); - if (semantic_parser) - fprintf (file, "#define\tT%s\t%d\n", symbol, bp->value); + /* A token which translation has already been set? */ + if (token_translations[bp->user_token_number] != 2) + complain (_("tokens %s and %s both assigned number %d"), + tags[token_translations[bp->user_token_number]], + bp->tag, bp->user_token_number); + token_translations[bp->user_token_number] = bp->value; } - - putc ('\n', file); } @@ -1661,18 +1565,12 @@ output_token_defines (FILE *file) static void packsymbols (void) { - bucket *bp; + bucket *bp = NULL; int tokno = 1; - int i; int last_user_token_number; - static char DOLLAR[] = "$"; - - /* 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); @@ -1688,11 +1586,18 @@ 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) { @@ -1718,18 +1623,22 @@ 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; @@ -1739,74 +1648,68 @@ 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 = 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; - if (!no_parser_flag) - output_token_defines (ftable); - 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; +} + + +/*---------------------------------------------------------------. +| Save the definition of token names in the `TOKENDEFS' muscle. | +`---------------------------------------------------------------*/ + +static void +symbols_save (void) +{ + struct obstack tokendefs; + bucket *bp; + obstack_init (&tokendefs); - if (defines_flag) + for (bp = firstsymbol; bp; bp = bp->next) { - output_token_defines (fdefines); + 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) - fprintf (fdefines, "\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 - fprintf (fdefines, "\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] != '@') - fprintf (fdefines, "#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); } @@ -1822,14 +1725,12 @@ 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; + rule_table = XCALLOC (rule_t, nrules) - 1; itemno = 0; ruleno = 1; @@ -1837,9 +1738,15 @@ packgram (void) p = grammar; while (p) { - rlhs[ruleno] = p->sym->value; - rrhs[ruleno] = itemno; - ruleprec = p->ruleprec; + bucket *ruleprec = p->ruleprec; + rule_table[ruleno].lhs = p->sym->value; + rule_table[ruleno].rhs = itemno; + rule_table[ruleno].line = p->line; + rule_table[ruleno].useful = TRUE; + rule_table[ruleno].action = p->action; + rule_table[ruleno].action_line = p->action_line; + rule_table[ruleno].guard = p->guard; + rule_table[ruleno].guard_line = p->guard_line; p = p->next; while (p && p->sym) @@ -1849,8 +1756,8 @@ 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; @@ -1860,9 +1767,9 @@ 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; @@ -1873,14 +1780,19 @@ packgram (void) } ritem[itemno] = 0; + nritems = itemno; + assert (nritems == nitems); + + if (trace_flag) + ritem_print (stderr); } /*-------------------------------------------------------------------. | Read in the grammar specification and record it in the format | -| described in gram.h. All guards are copied into the FGUARD file | -| and all actions into FACTION, 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. | +| 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. | `-------------------------------------------------------------------*/ void @@ -1889,102 +1801,80 @@ 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; - gensym_count = 0; - semantic_parser = 0; pure_parser = 0; grammar = NULL; - init_lex (); + lex_init (); lineno = 1; + /* 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; errtoken->user_token_number = 256; /* Value specified by POSIX. */ + /* Construct a token that represents all undefined literal tokens. It is always token number 2. */ undeftoken = getsym ("$undefined."); undeftoken->class = token_sym; undeftoken->user_token_number = 2; - /* Read the declaration section. Copy %{ ... %} groups to FTABLE - and FDEFINES file. Also notice any %token, %left, etc. found - there. */ - putc ('\n', ftable); - fprintf (ftable, "\ -/* %s, made from %s\n\ - by GNU bison %s. */\n\ -\n", no_parser_flag ? "Bison-generated parse tables" : "A Bison parser", infile, VERSION); + /* Initialize the obstacks. */ + obstack_init (&action_obstack); + obstack_init (&attrs_obstack); + obstack_init (&output_obstack); - fputs ("#define YYBISON 1 /* Identify Bison output. */\n\n", ftable); + finput = xfopen (infile, "r"); + + /* Read the declaration section. Copy %{ ... %} groups to + TABLE_OBSTACK and FDEFINES file. Also notice any %token, %left, + etc. found there. */ read_declarations (); - /* Start writing the guard and action files, if they are needed. */ - output_headers (); + + /* If the user did not define her EOFTOKEN, do it now. */ + if (!eoftoken) + { + eoftoken = getsym ("$"); + 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 (); - /* Now we know whether we need the line-number stack. If we do, - write its type into the .tab.h file. */ - if (fdefines) - reader_output_yylsp (fdefines); - /* Write closing delimiters for actions and guards. */ - output_trailers (); - if (locations_flag) - fputs ("#define YYLSP_NEEDED\n\n", ftable); + /* Some C code is given at the end of the grammar file. */ + read_additionnal_code (); + + 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 (); -} - - -void -reader_output_yylsp (FILE *f) -{ - if (locations_flag) - fputs ("\ -\n\ -#ifndef YYLTYPE\n\ -typedef struct yyltype\n\ -{\n\ - int timestamp;\n\ - int first_line;\n\ - int first_column;\ -\n\ - int last_line;\n\ - int last_column;\n\ - char *text;\n\ -} yyltype;\n\ -\n\ -# define YYLTYPE yyltype\n\ -#endif\n\ -\n", - f); }