X-Git-Url: https://git.saurik.com/bison.git/blobdiff_plain/1ff442ca0bca362a8a06f91b4f8075fa1cd204a3..e91ff6ce3d2653b41265591dafd4bdb32ae114c2:/src/reader.c diff --git a/src/reader.c b/src/reader.c index a24a9028..6a99cfff 100644 --- a/src/reader.c +++ b/src/reader.c @@ -1,1778 +1,1767 @@ /* Input parser for bison - Copyright (C) 1984, 1986, 1989 Free Software Foundation, Inc. + Copyright (C) 1984, 1986, 1989, 1992, 1998, 2000, 2001, 2002 + Free Software Foundation, Inc. -This file is part of Bison, the GNU Compiler Compiler. + This file is part of Bison, the GNU Compiler Compiler. -Bison is free software; you can redistribute it and/or modify -it under the terms of the GNU General Public License as published by -the Free Software Foundation; either version 2, or (at your option) -any later version. + Bison is free software; you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation; either version 2, or (at your option) + any later version. -Bison is distributed in the hope that it will be useful, -but WITHOUT ANY WARRANTY; without even the implied warranty of -MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the -GNU General Public License for more details. + Bison is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. -You should have received a copy of the GNU General Public License -along with Bison; see the file COPYING. If not, write to -the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. */ + You should have received a copy of the GNU General Public License + along with Bison; see the file COPYING. If not, write to + the Free Software Foundation, Inc., 59 Temple Place - Suite 330, + Boston, MA 02111-1307, USA. */ -/* 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. - -The entry point is reader(). */ - -#include -#include #include "system.h" +#include "quotearg.h" +#include "quote.h" +#include "getargs.h" #include "files.h" -#include "new.h" #include "symtab.h" +#include "options.h" #include "lex.h" #include "gram.h" -#include "machine.h" - -#define LTYPESTR "\n#ifndef YYLTYPE\ntypedef\n 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 }\n\ - yyltype;\n\n#define YYLTYPE yyltype\n#endif\n\n" - -/* Number of slots allocated (but not necessarily used yet) in `rline' */ -int rline_allocated; - -extern char *program_name; -extern int definesflag; -extern int nolinesflag; -extern bucket *symval; -extern int numval; -extern int failure; -extern int expected_conflicts; -extern char *token_buffer; - -extern int atoi (); - -extern void init_lex(); -extern void tabinit(); -extern void output_headers(); -extern void output_trailers(); -extern void free_symtab(); -extern void open_extra_files(); -extern void fatal(); -extern void fatals(); -extern void unlex(); -extern void done(); - -extern int skip_white_space(); -extern int parse_percent_token(); -extern int lex(); - -void read_declarations(); -void copy_definition(); -void parse_token_decl(); -void parse_start_decl(); -void parse_type_decl(); -void parse_assoc_decl(); -void parse_union_decl(); -void parse_expect_decl(); -void copy_action(); -void readgram(); -void record_rule_line(); -void packsymbols(); -void output_token_defines(); -void packgram(); -int read_signed_integer(); -int get_type(); - -typedef - struct symbol_list - { - struct symbol_list *next; - bucket *sym; - bucket *ruleprec; - } - symbol_list; +#include "complain.h" +#include "output.h" +#include "reader.h" +#include "conflicts.h" +#include "muscle_tab.h" +typedef struct symbol_list +{ + struct symbol_list *next; + symbol_t *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; + symbol_t *ruleprec; +} symbol_list; int lineno; -symbol_list *grammar; -int start_flag; -bucket *startval; -char **tags; +static symbol_list *grammar = NULL; +static int start_flag = 0; +static symbol_t *startval = NULL; /* Nonzero if components of semantic values are used, implying they must be unions. */ static int value_components_used; -static int typed; /* nonzero if %union has been seen. */ +/* Nonzero if %union has been seen. */ +static int typed = 0; -static int lastprec; /* incremented for each %left, %right or %nonassoc seen */ +/* Incremented for each %left, %right or %nonassoc seen */ +static int lastprec = 0; -static int gensym_count; /* incremented for each generated symbol */ +symbol_t *errtoken = NULL; +symbol_t *undeftoken = NULL; +symbol_t *eoftoken = NULL; +symbol_t *axiom = NULL; + +static symbol_list * +symbol_list_new (symbol_t *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; +/*------------------------. +| Operations on symbols. | +`------------------------*/ -/* Nonzero if any action or guard uses the @n construct. */ -static int yylsp_needed; -extern char *version_string; +/*-----------------------------------------------------------. +| If THIS is not defined, report an error, and consider it a | +| nonterminal. | +`-----------------------------------------------------------*/ -void -reader() +static bool +symbol_check_defined (symbol_t *this) { - start_flag = 0; - startval = NULL; /* start symbol not specified yet. */ + if (this->class == unknown_sym) + { + complain + (_("symbol %s is used, but is not defined as a token and has no rules"), + this->tag); + this->class = nterm_sym; + this->number = nvars++; + } + + return TRUE; +} + + +/*-------------------------------------------------------------------. +| Assign a symbol number, and write the definition of the token name | +| into FDEFINES. Put in SYMBOLS. | +`-------------------------------------------------------------------*/ -#if 0 - translations = 0; /* initially assume token number translation not needed. */ -#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; +static bool +symbol_make_alias (symbol_t *symbol, char *typename) +{ + if (symval->alias) + warn (_("symbol `%s' used more than once as a literal string"), + symval->tag); + else if (symbol->alias) + warn (_("symbol `%s' given more than one literal string"), + symbol->tag); + else + { + symval->class = token_sym; + symval->type_name = typename; + symval->user_token_number = symbol->user_token_number; + symbol->user_token_number = SALIAS; + symval->alias = symbol; + symbol->alias = symval; + /* symbol and symval combined are only one symbol */ + nsyms--; + ntokens--; + assert (ntokens == symbol->number || ntokens == symval->number); + symbol->number = symval->number = + (symval->number < symbol->number) ? symval->number : symbol->number; + } - nsyms = 1; - nvars = 0; - nrules = 0; - nitems = 0; - rline_allocated = 10; - rline = NEW2(rline_allocated, short); + return TRUE; +} - typed = 0; - lastprec = 0; +/*---------------------------------------------------------. +| Check that THIS, and its alias, have same precedence and | +| associativity. | +`---------------------------------------------------------*/ - gensym_count = 0; +static bool +symbol_check_alias_consistence (symbol_t *this) +{ + /* Check only those who _are_ the aliases. */ + if (this->alias && this->user_token_number == SALIAS) + { + if (this->prec != this->alias->prec) + { + if (this->prec != 0 && this->alias->prec != 0) + complain (_("conflicting precedences for %s and %s"), + this->tag, this->alias->tag); + if (this->prec != 0) + this->alias->prec = this->prec; + else + this->prec = this->alias->prec; + } - semantic_parser = 0; - pure_parser = 0; - yylsp_needed = 0; + if (this->assoc != this->alias->assoc) + { + if (this->assoc != 0 && this->alias->assoc != 0) + complain (_("conflicting assoc values for %s and %s"), + this->tag, this->alias->tag); + if (this->assoc != 0) + this->alias->assoc = this->assoc; + else + this->assoc = this->alias->assoc; + } + } + return TRUE; +} - grammar = NULL; - init_lex(); - lineno = 1; +/*-------------------------------------------------------------------. +| Assign a symbol number, and write the definition of the token name | +| into FDEFINES. Put in SYMBOLS. | +`-------------------------------------------------------------------*/ - /* initialize the symbol table. */ - tabinit(); - /* construct the error token */ - errtoken = getsym("error"); - errtoken->class = STOKEN; - errtoken->user_token_number = 256; /* Value specified by posix. */ - /* construct a token that represents all undefined literal tokens. */ - /* it is always token number 2. */ - getsym("$illegal.")->class = STOKEN; - /* Read the declaration section. Copy %{ ... %} groups to ftable and fdefines file. - Also notice any %token, %left, etc. found there. */ - fprintf(ftable, "\n/* A Bison parser, made from %s", infile); - fprintf(ftable, " with Bison version %s */\n\n", version_string); - fprintf(ftable, "#define YYBISON 1 /* Identify Bison output. */\n\n"); - read_declarations(); - /* output the definition of YYLTYPE into the fattrs and fdefines files. */ - /* fattrs winds up in the .tab.c file, before bison.simple. */ - fprintf(fattrs, LTYPESTR); - /* start writing the guard and action files, if they are needed. */ - output_headers(); - /* 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 (yylsp_needed) +static bool +symbol_pack (symbol_t *this) +{ + if (this->class == nterm_sym) + { + this->number += ntokens; + } + else if (this->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 (this->number == NUMBER_UNDEFINED) + { + if (this == eoftoken || this->alias == eoftoken) + this->number = this->alias->number = 0; + else + { + assert (this->alias->number != NUMBER_UNDEFINED); + this->number = this->alias->number; + } + } + /* Do not do processing below for SALIASs. */ + if (this->user_token_number == SALIAS) + return TRUE; + } + else /* this->class == token_sym */ { - if (fdefines) - fprintf(fdefines, LTYPESTR); + assert (this->number != NUMBER_UNDEFINED); } - /* write closing delimiters for actions and guards. */ - output_trailers(); - if (yylsp_needed) - fprintf(ftable, "#define YYLSP_NEEDED\n\n"); - /* 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(); + + symbols[this->number] = this; + return TRUE; } -/* read from finput until %% is seen. Discard the %%. -Handle any % declarations, -and copy the contents of any %{ ... %} groups to fattrs. */ -void -read_declarations () +/*--------------------------------------------------. +| Put THIS in TOKEN_TRANSLATIONS if it is a token. | +`--------------------------------------------------*/ + +static bool +symbol_translation (symbol_t *this) { - register int c; - register int tok; + /* Non-terminal? */ + if (this->class == token_sym + && this->user_token_number != SALIAS) + { + /* A token which translation has already been set? */ + if (token_translations[this->user_token_number] != undeftoken->number) + complain (_("tokens %s and %s both assigned number %d"), + symbols[token_translations[this->user_token_number]]->tag, + this->tag, this->user_token_number); - for (;;) + token_translations[this->user_token_number] = this->number; + } + + return TRUE; +} + + +/*===================\ +| Low level lexing. | +\===================*/ + +static void +skip_to_char (int target) +{ + int c; + if (target == '\n') + complain (_(" Skipping to next \\n")); + else + complain (_(" Skipping to next %c"), target); + + do + c = skip_white_space (); + while (c != target && c != EOF); + if (c != EOF) + ungetc (c, finput); +} + + +/*---------------------------------------------------------. +| Read a signed integer from STREAM and return its value. | +`---------------------------------------------------------*/ + +static inline int +read_signed_integer (FILE *stream) +{ + int c = getc (stream); + int sign = 1; + int n = 0; + + if (c == '-') + { + c = getc (stream); + sign = -1; + } + + while (isdigit (c)) { - c = skip_white_space(); + n = 10 * n + (c - '0'); + c = getc (stream); + } - if (c == '%') + ungetc (c, stream); + + return sign * n; +} + +/*--------------------------------------------------------------. +| 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) { - tok = parse_percent_token(); + complain (_("invalid $ value")); + return NULL; + } + ++i; + } - switch (tok) - { - case TWO_PERCENTS: - return; + return rp->sym->type_name; +} + +/*------------------------------------------------------------. +| Dump the string from FIN to OOUT if non null. MATCH is the | +| delimiter of the string (either ' or "). | +`------------------------------------------------------------*/ - case PERCENT_LEFT_CURLY: - copy_definition(); - break; +static inline void +copy_string2 (FILE *fin, struct obstack *oout, int match, int store) +{ + int c; - case TOKEN: - parse_token_decl (STOKEN, SNTERM); - break; - - case NTERM: - parse_token_decl (SNTERM, STOKEN); - break; - - case TYPE: - parse_type_decl(); - break; - - case START: - parse_start_decl(); - break; - - case UNION: - parse_union_decl(); - break; - - case EXPECT: - parse_expect_decl(); - break; - - case LEFT: - parse_assoc_decl(LEFT_ASSOC); - break; + if (store) + obstack_1grow (oout, match); - case RIGHT: - parse_assoc_decl(RIGHT_ASSOC); - break; + c = getc (fin); - case NONASSOC: - parse_assoc_decl(NON_ASSOC); - break; + while (c != match) + { + if (c == EOF) + fatal (_("unterminated string at end of file")); + if (c == '\n') + { + complain (_("unterminated string")); + ungetc (c, fin); + c = match; /* invent terminator */ + continue; + } - case SEMANTIC_PARSER: - if (semantic_parser == 0) - { - semantic_parser = 1; - open_extra_files(); - } - break; + obstack_1grow (oout, c); - case PURE_PARSER: - pure_parser = 1; - break; + if (c == '\\') + { + c = getc (fin); + if (c == EOF) + fatal (_("unterminated string at end of file")); + obstack_1grow (oout, c); - default: - fatal("junk after `%%' in definition section"); + if (c == '\n') + ++lineno; + } + + c = getc (fin); + } + + if (store) + obstack_1grow (oout, c); +} + +/* FIXME. */ + +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_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; + } + + obstack_1grow (oout, c); + c = getc (fin); + + ended = 0; + while (!ended) + { + if (!cplus_comment && c == '*') + { + while (c == '*') + { + obstack_1grow (oout, c); + c = getc (fin); + } + + if (c == '/') + { + obstack_1grow (oout, c); + ended = 1; } } + else if (c == '\n') + { + ++lineno; + obstack_1grow (oout, c); + if (cplus_comment) + ended = 1; + else + c = getc (fin); + } else if (c == EOF) - fatal("no input grammar"); - else if (c >= 040 && c <= 0177) - fatals ("unknown character `%c' in declaration section", c); + fatal (_("unterminated comment")); else - fatals ("unknown character with code 0x%x in declaration section", c); + { + obstack_1grow (oout, c); + c = getc (fin); + } } } -/* copy the contents of a %{ ... %} into the definitions file. -The %{ has already been read. Return after reading the %}. */ +/*-----------------------------------------------------------------. +| 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 `$0' with | +| respect to the top of the stack. | +`-----------------------------------------------------------------*/ -void -copy_definition () +static inline void +copy_at (FILE *fin, struct obstack *oout, int stack_offset) { - register int c; - register int match; - register int ended; - register int after_percent; /* -1 while reading a character if prev char was % */ - int cplus_comment; + int c; + + c = getc (fin); + if (c == '$') + { + obstack_sgrow (oout, "yyloc"); + locations_flag = 1; + } + 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 + { + /* 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 + { + char buf[] = "@c"; + buf[1] = c; + 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 | +| `%{' has already been read. Return after reading the `%}'. | +`-------------------------------------------------------------------*/ + +static void +copy_definition (void) +{ + int c; + /* -1 while reading a character if prev char was %. */ + int after_percent; - if (!nolinesflag) - fprintf(fattrs, "#line %d \"%s\"\n", lineno, infile); + if (!no_lines_flag) + { + obstack_fgrow2 (&attrs_obstack, muscle_find ("linef"), + lineno, quotearg_style (c_quoting_style, + muscle_find ("filename"))); + } after_percent = 0; - c = getc(finput); + c = getc (finput); for (;;) { switch (c) { case '\n': - putc(c, fattrs); - lineno++; + obstack_1grow (&attrs_obstack, c); + ++lineno; break; case '%': - after_percent = -1; + after_percent = -1; break; - + case '\'': case '"': - match = c; - putc(c, fattrs); - c = getc(finput); - - while (c != match) - { - if (c == EOF || c == '\n') - fatal("unterminated string"); - - putc(c, fattrs); - - if (c == '\\') - { - c = getc(finput); - if (c == EOF) - fatal("unterminated string"); - putc(c, fattrs); - if (c == '\n') - lineno++; - } - - c = getc(finput); - } - - putc(c, fattrs); + copy_string (finput, &attrs_obstack, c); break; case '/': - putc(c, fattrs); - c = getc(finput); - if (c != '*' && c != '/') - continue; - - cplus_comment = (c == '/'); - putc(c, fattrs); - c = getc(finput); - - ended = 0; - while (!ended) - { - if (!cplus_comment && c == '*') - { - while (c == '*') - { - putc(c, fattrs); - c = getc(finput); - } - - if (c == '/') - { - putc(c, fattrs); - ended = 1; - } - } - else if (c == '\n') - { - lineno++; - putc(c, fattrs); - if (cplus_comment) - ended = 1; - else - c = getc(finput); - } - else if (c == EOF) - fatal("unterminated comment in `%{' definition"); - else - { - putc(c, fattrs); - c = getc(finput); - } - } - + copy_comment (finput, &attrs_obstack); break; case EOF: - fatal("unterminated `%{' definition"); + fatal ("%s", _("unterminated `%{' definition")); default: - putc(c, fattrs); + obstack_1grow (&attrs_obstack, c); } - c = getc(finput); + c = getc (finput); if (after_percent) { if (c == '}') return; - putc('%', fattrs); + obstack_1grow (&attrs_obstack, '%'); } after_percent = 0; - } - } +/*-------------------------------------------------------------------. +| Parse what comes after %token or %nterm. For %token, WHAT_IS is | +| token_sym and WHAT_IS_NOT is nterm_sym. For %nterm, the arguments | +| are reversed. | +`-------------------------------------------------------------------*/ -/* parse what comes after %token or %nterm. -For %token, what_is is STOKEN and what_is_not is SNTERM. -For %nterm, the arguments are reversed. */ - -void -parse_token_decl (what_is, what_is_not) - int what_is, what_is_not; +static void +parse_token_decl (symbol_class what_is, symbol_class what_is_not) { -/* register int start_lineno; JF */ - register int token = 0; - register int prev; - register char *typename = 0; - int k; + token_t token = tok_undef; + char *typename = NULL; -/* start_lineno = lineno; JF */ + /* The symbol being defined. */ + symbol_t *symbol = NULL; + /* After `%token' and `%nterm', any number of symbols maybe be + defined. */ for (;;) { - if(ungetc(skip_white_space(), finput) == '%') - return; + int tmp_char = ungetc (skip_white_space (), finput); -/* if (lineno != start_lineno) - return; JF */ - - /* we have not passed a newline, so the token now starting is in this declaration */ - prev = token; + /* `%' (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) - continue; - if (token == TYPENAME) + token = lex (); + if (token == tok_comma) + { + symbol = NULL; + continue; + } + if (token == tok_typename) { - k = strlen(token_buffer); - typename = NEW2(k + 1, char); - strcpy(typename, token_buffer); + typename = xstrdup (token_buffer); value_components_used = 1; + symbol = NULL; + } + else if (token == tok_identifier && *symval->tag == '\"' && symbol) + { + symbol_make_alias (symbol, typename); + symbol = NULL; } - else if (token == IDENTIFIER) + else if (token == tok_identifier) { int oldclass = symval->class; + symbol = symval; - if (symval->class == what_is_not) - fatals("symbol %s redefined", symval->tag); - symval->class = what_is; - if (what_is == SNTERM && oldclass != SNTERM) - symval->value = nvars++; + if (symbol->class == what_is_not) + complain (_("symbol %s redefined"), symbol->tag); + symbol->class = what_is; + if (what_is == nterm_sym && oldclass != nterm_sym) + symbol->number = nvars++; + if (what_is == token_sym && symbol->number == NUMBER_UNDEFINED) + symbol->number = ntokens++; if (typename) { - if (symval->type_name == NULL) - symval->type_name = typename; - else - fatals("type redeclaration for %s", symval->tag); + if (symbol->type_name == NULL) + symbol->type_name = typename; + else if (strcmp (typename, symbol->type_name) != 0) + complain (_("type redeclaration for %s"), symbol->tag); + } + } + else if (symbol && token == tok_number) + { + symbol->user_token_number = numval; + /* User defined EOF token? */ + if (numval == 0) + { + eoftoken = symbol; + eoftoken->number = 0; + /* It is always mapped to 0, so it was already counted in + NTOKENS. */ + --ntokens; } } - else if (prev == IDENTIFIER && token == NUMBER) - { - symval->user_token_number = numval; - translations = 1; - } else - fatal("invalid text in %token or %nterm declaration"); + { + complain (_("`%s' is invalid in %s"), + token_buffer, + (what_is == token_sym) ? "%token" : "%nterm"); + skip_to_char ('%'); + } } } +/*------------------------------. +| Parse what comes after %start | +`------------------------------*/ -/* parse what comes after %start */ - -void -parse_start_decl () +static void +parse_start_decl (void) { if (start_flag) - fatal("multiple %start declarations"); - start_flag = 1; - if (lex() != IDENTIFIER) - fatal("invalid %start declaration"); - startval = symval; + complain (_("multiple %s declarations"), "%start"); + if (lex () != tok_identifier) + complain (_("invalid %s declaration"), "%start"); + else + { + start_flag = 1; + startval = symval; + } } +/*-----------------------------------------------------------. +| read in a %type declaration and record its information for | +| get_type_name to access | +`-----------------------------------------------------------*/ - -/* read in a %type declaration and record its information for get_type_name to access */ - -void -parse_type_decl () +static void +parse_type_decl (void) { - register int k; - register char *name; -/* register int start_lineno; JF */ + char *name; - if (lex() != TYPENAME) - fatal("ill-formed %type declaration"); - - k = strlen(token_buffer); - name = NEW2(k + 1, char); - strcpy(name, token_buffer); + if (lex () != tok_typename) + { + complain ("%s", _("%type declaration has no ")); + skip_to_char ('%'); + return; + } -/* start_lineno = lineno; */ + name = xstrdup (token_buffer); for (;;) { - register int t; + token_t t; + int tmp_char = ungetc (skip_white_space (), finput); - if(ungetc(skip_white_space(), finput) == '%') + if (tmp_char == '%') return; + if (tmp_char == EOF) + fatal (_("Premature EOF after %s"), token_buffer); -/* if (lineno != start_lineno) - return; JF */ - - /* we have not passed a newline, so the token now starting is in this declaration */ - - t = lex(); + t = lex (); 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 - fatals("type redeclaration for %s", symval->tag); + else if (strcmp (name, symval->type_name) != 0) + complain (_("type redeclaration for %s"), symval->tag); break; default: - fatal("invalid %type declaration"); + complain (_("invalid %%type declaration due to item: %s"), + token_buffer); + skip_to_char ('%'); } } } -/* read in a %left, %right or %nonassoc declaration and record its information. */ -/* assoc is either LEFT_ASSOC, RIGHT_ASSOC or NON_ASSOC. */ +/*----------------------------------------------------------------. +| Read in a %left, %right or %nonassoc declaration and record its | +| information. | +`----------------------------------------------------------------*/ -void -parse_assoc_decl (assoc) -int assoc; +static void +parse_assoc_decl (associativity assoc) { - register int k; - register char *name = NULL; -/* register int start_lineno; JF */ - register int prev = 0; /* JF added = 0 to keep lint happy */ - - lastprec++; /* Assign a new precedence level, never 0. */ + char *name = NULL; + int prev = 0; -/* start_lineno = lineno; */ + /* Assign a new precedence level, never 0. */ + ++lastprec; for (;;) { - register int t; + token_t t; + int tmp_char = ungetc (skip_white_space (), finput); - if(ungetc(skip_white_space(), finput) == '%') + if (tmp_char == '%') return; + if (tmp_char == EOF) + fatal (_("Premature EOF after %s"), token_buffer); - /* if (lineno != start_lineno) - return; JF */ - - /* we have not passed a newline, so the token now starting is in this declaration */ - - t = lex(); + t = lex (); switch (t) { - - case TYPENAME: - k = strlen(token_buffer); - name = NEW2(k + 1, char); - strcpy(name, token_buffer); + case tok_typename: + name = xstrdup (token_buffer); break; - case COMMA: + case tok_comma: break; - case IDENTIFIER: + case tok_identifier: if (symval->prec != 0) - fatals("redefining precedence of %s", symval->tag); + complain (_("redefining precedence of %s"), symval->tag); symval->prec = lastprec; symval->assoc = assoc; - if (symval->class == SNTERM) - fatals("symbol %s redefined", symval->tag); - symval->class = STOKEN; + if (symval->class == nterm_sym) + complain (_("symbol %s redefined"), symval->tag); + if (symval->number == NUMBER_UNDEFINED) + { + symval->number = ntokens++; + symval->class = token_sym; + } if (name) - { /* record the type, if one is specified */ + { /* record the type, if one is specified */ if (symval->type_name == NULL) symval->type_name = name; - else - fatals("type redeclaration for %s", symval->tag); + else if (strcmp (name, symval->type_name) != 0) + complain (_("type redeclaration for %s"), symval->tag); } break; - case NUMBER: - if (prev == IDENTIFIER) - { + case tok_number: + if (prev == tok_identifier) + { symval->user_token_number = numval; - translations = 1; - } - else - fatal("invalid text in association declaration"); + } + else + { + complain + (_("invalid text (%s) - number should be after identifier"), + token_buffer); + skip_to_char ('%'); + } break; - case SEMICOLON: + case tok_semicolon: return; default: - fatal("malformatted association declaration"); + complain (_("unexpected item: %s"), token_buffer); + skip_to_char ('%'); } 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. | +`--------------------------------------------------------------*/ -void -parse_union_decl() +static void +parse_union_decl (void) { - register int c; - register int count; - register int in_comment; - int cplus_comment; - + int c; + int count = 0; + bool done = FALSE; + struct obstack union_obstack; if (typed) - fatal("multiple %union declarations"); + complain (_("multiple %s declarations"), "%union"); typed = 1; - if (!nolinesflag) - fprintf(fattrs, "\n#line %d \"%s\"\n", lineno, infile); - else - fprintf(fattrs, "\n"); - - fprintf(fattrs, "typedef union"); - if (fdefines) - fprintf(fdefines, "typedef union"); - - count = 0; - in_comment = 0; - - c = getc(finput); + MUSCLE_INSERT_INT ("stype_line", lineno); + obstack_init (&union_obstack); + obstack_sgrow (&union_obstack, "union"); - 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) { case '\n': - lineno++; + ++lineno; break; case '/': - c = getc(finput); - if (c != '*' && c != '/') - ungetc(c, finput); - else - { - putc(c, fattrs); - if (fdefines) - putc(c, fdefines); - cplus_comment = (c == '/'); - in_comment = 1; - c = getc(finput); - while (in_comment) - { - putc(c, fattrs); - if (fdefines) - putc(c, fdefines); - - if (c == '\n') - { - lineno++; - if (cplus_comment) - { - in_comment = 0; - break; - } - } - if (c == EOF) - fatal("unterminated comment"); - - if (!cplus_comment && c == '*') - { - c = getc(finput); - if (c == '/') - { - putc('/', fattrs); - if (fdefines) - putc('/', fdefines); - in_comment = 0; - } - } - else - c = getc(finput); - } - } + copy_comment (finput, &union_obstack); break; - case '{': - count++; + ++count; break; case '}': + /* FIXME: Errr. How could this happen???. --akim */ if (count == 0) - fatal ("unmatched close-brace (`}')"); + 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); } -} -/* parse the declaration %expect N which says to expect N - shift-reduce conflicts. */ - -void -parse_expect_decl() -{ - register int c; - register int count; - char buffer[20]; + /* 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)); +} - c = getc(finput); - while (c == ' ' || c == '\t') - c = getc(finput); - count = 0; - while (c >= '0' && c <= '9') - { - if (count < 20) - buffer[count++] = c; - c = getc(finput); - } - buffer[count] = 0; +/*-------------------------------------------------------. +| Parse the declaration %expect N which says to expect N | +| shift-reduce conflicts. | +`-------------------------------------------------------*/ +static void +parse_expect_decl (void) +{ + int c = skip_white_space (); ungetc (c, finput); - expected_conflicts = atoi (buffer); + if (!isdigit (c)) + complain (_("argument of %%expect is not an integer")); + else + expected_conflicts = read_signed_integer (finput); } -/* that's all of parsing the declaration section */ - -/* Get the data type (alternative in the union) of the value for symbol n in rule rule. */ -char * -get_type_name(n, rule) -int n; -symbol_list *rule; +/*-------------------------------------------------------------------. +| Parse what comes after %thong. the full syntax is | +| | +| %thong token number literal | +| | +| the or number may be omitted. The number specifies the | +| user_token_number. | +| | +| Two symbols are entered in the table, one for the token symbol and | +| one for the literal. Both are given the , if any, from the | +| declaration. The ->user_token_number of the first is SALIAS and | +| the ->user_token_number of the second is set to the number, if | +| any, from the declaration. The two symbols are linked via | +| pointers in their ->alias fields. | +| | +| During OUTPUT_DEFINES_TABLE, the symbol is reported thereafter, | +| only the literal string is retained it is the literal string that | +| is output to yytname | +`-------------------------------------------------------------------*/ + +static void +parse_thong_decl (void) { - static char *msg = "invalid $ value"; - - register int i; - register symbol_list *rp; - - if (n < 0) - fatal(msg); + token_t token; + symbol_t *symbol; + char *typename = 0; + int usrtoknum = SUNDEF; - rp = rule; - i = 0; - - while (i < n) + token = lex (); /* fetch typename or first token */ + if (token == tok_typename) { - rp = rp->next; - if (rp == NULL || rp->sym == NULL) - fatal(msg); - i++; + typename = xstrdup (token_buffer); + value_components_used = 1; + token = lex (); /* fetch first token */ } - return (rp->sym->type_name); -} - - - -/* 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. */ + /* process first token */ -void -copy_guard(rule, stack_offset) -symbol_list *rule; -int stack_offset; -{ - register int c; - register int n; - register int count; - register int match; - register int ended; - register char *type_name; - int brace_flag = 0; - int cplus_comment; - - /* 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 (!nolinesflag) - fprintf(fguard, "#line %d \"%s\"\n", lineno, infile); - putc('{', fguard); + if (token != tok_identifier) + { + complain (_("unrecognized item %s, expected an identifier"), + token_buffer); + skip_to_char ('%'); + return; + } + symval->class = token_sym; + symval->type_name = typename; + symval->user_token_number = SALIAS; + symbol = symval; - count = 0; - c = getc(finput); + token = lex (); /* get number or literal string */ - while (brace_flag ? (count > 0) : (c != ';')) + if (token == tok_number) { - switch (c) - { - case '\n': - putc(c, fguard); - lineno++; - break; - - case '{': - putc(c, fguard); - brace_flag = 1; - count++; - break; + usrtoknum = numval; + token = lex (); /* okay, did number, now get literal */ + } - case '}': - putc(c, fguard); - if (count > 0) - count--; - else - fatal("unmatched right brace ('}')"); - break; + /* process literal string token */ - case '\'': - case '"': - match = c; - putc(c, fguard); - c = getc(finput); + if (token != tok_identifier || *symval->tag != '\"') + { + complain (_("expected string constant instead of %s"), token_buffer); + skip_to_char ('%'); + return; + } + symval->class = token_sym; + symval->type_name = typename; + symval->user_token_number = usrtoknum; - while (c != match) - { - if (c == EOF || c == '\n') - fatal("unterminated string"); - - putc(c, fguard); - - if (c == '\\') - { - c = getc(finput); - if (c == EOF) - fatal("unterminated string"); - putc(c, fguard); - if (c == '\n') - lineno++; - } - - c = getc(finput); - } + symval->alias = symbol; + symbol->alias = symval; - putc(c, fguard); - break; + /* symbol and symval combined are only one symbol. */ + nsyms--; +} - case '/': - putc(c, fguard); - c = getc(finput); - if (c != '*' && c != '/') - continue; - cplus_comment = (c == '/'); - putc(c, fguard); - c = getc(finput); +static void +parse_muscle_decl (void) +{ + int ch = ungetc (skip_white_space (), finput); + char *muscle_key; + char *muscle_value; - ended = 0; - while (!ended) - { - if (!cplus_comment && c == '*') - { - while (c == '*') - { - putc(c, fguard); - c = getc(finput); - } - - if (c == '/') - { - putc(c, fguard); - ended = 1; - } - } - else if (c == '\n') - { - lineno++; - putc(c, fguard); - if (cplus_comment) - ended = 1; - else - c = getc(finput); - } - else if (c == EOF) - fatal("unterminated comment"); - else - { - putc(c, fguard); - c = getc(finput); - } - } + /* 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); - break; + /* 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); - case '$': - c = getc(finput); - type_name = NULL; + /* Store the (key, value) pair in the environment. */ + muscle_insert (muscle_key, muscle_value); +} - if (c == '<') - { - register char *cp = token_buffer; - while ((c = getc(finput)) != '>' && c > 0) - *cp++ = c; - *cp = 0; - type_name = token_buffer; - c = getc(finput); - } +/*---------------------------------. +| Parse a double quoted parameter. | +`---------------------------------*/ - 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) /* JF */ - fprintf(stderr,"%s:%d: warning: $$ of '%s' has no declared type.\n",infile,lineno,rule->sym->tag); - } +static const char * +parse_dquoted_param (const char *from) +{ + struct obstack param_obstack; + const char *param = NULL; + int c; - 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) /* JF */ - fprintf(stderr,"%s:%d: warning: $%d of '%s' has no declared type.\n",infile,lineno,n,rule->sym->tag); - continue; - } - else - fatals("$%c is invalid",c); /* JF changed style */ + obstack_init (¶m_obstack); + c = skip_white_space (); - break; + if (c != '"') + { + complain (_("invalid %s declaration"), from); + ungetc (c, finput); + skip_to_char ('%'); + return NULL; + } - case '@': - c = getc(finput); - if (isdigit(c) || c == '-') - { - ungetc (c, finput); - n = read_signed_integer(finput); - c = getc(finput); - } - else - fatals("@%c is invalid",c); /* JF changed style */ + while ((c = literalchar ()) != '"') + obstack_1grow (¶m_obstack, c); - fprintf(fguard, "yylsp[%d]", n - stack_offset); - yylsp_needed = 1; + obstack_1grow (¶m_obstack, '\0'); + param = obstack_finish (¶m_obstack); - continue; + if (c != '"' || strlen (param) == 0) + { + complain (_("invalid %s declaration"), from); + if (c != '"') + ungetc (c, finput); + skip_to_char ('%'); + return NULL; + } - case EOF: - fatal("unterminated %guard clause"); + return param; +} - default: - putc(c, fguard); - } +/*----------------------------------. +| Parse what comes after %skeleton. | +`----------------------------------*/ - if (c != '}' || count != 0) - c = getc(finput); - } +static void +parse_skel_decl (void) +{ + skeleton = parse_dquoted_param ("%skeleton"); +} - c = skip_white_space(); +/*----------------------------------------------------------------. +| Read from finput until `%%' is seen. Discard the `%%'. Handle | +| any `%' declarations, and copy the contents of any `%{ ... %}' | +| groups to ATTRS_OBSTACK. | +`----------------------------------------------------------------*/ - fprintf(fguard, ";\n break;}"); - if (c == '{') - copy_action(rule, stack_offset); - else if (c == '=') +static void +read_declarations (void) +{ + for (;;) { - c = getc(finput); - if (c == '{') - copy_action(rule, stack_offset); - } - else - ungetc(c, finput); -} - + int c = skip_white_space (); + if (c == '%') + { + token_t tok = parse_percent_token (); -/* Assuming that a { has just been seen, copy everything up to the matching } -into the actions file. -stack_offset is the number of values in the current rule so far, -which says where to find $0 with respect to the top of the stack. */ + switch (tok) + { + case tok_two_percents: + return; -void -copy_action(rule, stack_offset) -symbol_list *rule; -int stack_offset; -{ - register int c; - register int n; - register int count; - register int match; - register int ended; - register char *type_name; - int cplus_comment; + case tok_percent_left_curly: + copy_definition (); + break; - /* offset is always 0 if parser has already popped the stack pointer */ - if (semantic_parser) stack_offset = 0; + case tok_token: + parse_token_decl (token_sym, nterm_sym); + break; - fprintf(faction, "\ncase %d:\n", nrules); - if (!nolinesflag) - fprintf(faction, "#line %d \"%s\"\n", lineno, infile); - putc('{', faction); + case tok_nterm: + parse_token_decl (nterm_sym, token_sym); + break; - count = 1; - c = getc(finput); + case tok_type: + parse_type_decl (); + break; - while (count > 0) - { - while (c != '}') - { - switch (c) - { - case '\n': - putc(c, faction); - lineno++; + case tok_start: + parse_start_decl (); break; - case '{': - putc(c, faction); - count++; + case tok_union: + parse_union_decl (); break; - case '\'': - case '"': - match = c; - putc(c, faction); - c = getc(finput); - - while (c != match) - { - if (c == EOF || c == '\n') - fatal("unterminated string"); - - putc(c, faction); - - if (c == '\\') - { - c = getc(finput); - if (c == EOF) - fatal("unterminated string"); - putc(c, faction); - if (c == '\n') - lineno++; - } - - c = getc(finput); - } - - putc(c, faction); + case tok_expect: + parse_expect_decl (); break; - case '/': - putc(c, faction); - c = getc(finput); - if (c != '*' && c != '/') - continue; - - cplus_comment = (c == '/'); - putc(c, faction); - c = getc(finput); - - ended = 0; - while (!ended) - { - if (!cplus_comment && c == '*') - { - while (c == '*') - { - putc(c, faction); - c = getc(finput); - } - - if (c == '/') - { - putc(c, faction); - ended = 1; - } - } - else if (c == '\n') - { - lineno++; - putc(c, faction); - if (cplus_comment) - ended = 1; - else - c = getc(finput); - } - else if (c == EOF) - fatal("unterminated comment"); - else - { - putc(c, faction); - c = getc(finput); - } - } + case tok_thong: + parse_thong_decl (); + break; + case tok_left: + parse_assoc_decl (left_assoc); break; - case '$': - c = getc(finput); - type_name = NULL; - - if (c == '<') - { - register char *cp = token_buffer; - - while ((c = getc(finput)) != '>' && c > 0) - *cp++ = c; - *cp = 0; - type_name = token_buffer; - value_components_used = 1; - - 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) /* JF */ - fprintf(stderr,"%s:%d: warning: $$ of '%s' has no declared type.\n",infile,lineno,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) /* JF */ - fprintf(stderr,"%s:%d: warning: $%d of '%s' has no declared type.\n",infile,lineno,n,rule->sym->tag); - continue; - } - else - fatals("$%c is invalid",c); /* JF changed format */ + case tok_right: + parse_assoc_decl (right_assoc); + break; + case tok_nonassoc: + parse_assoc_decl (non_assoc); break; - case '@': - c = getc(finput); - if (isdigit(c) || c == '-') - { - ungetc (c, finput); - n = read_signed_integer(finput); - c = getc(finput); - } - else - fatal("invalid @-construct"); + case tok_define: + parse_muscle_decl (); + break; - fprintf(faction, "yylsp[%d]", n - stack_offset); - yylsp_needed = 1; + case tok_skel: + parse_skel_decl (); + break; - continue; + case tok_noop: + break; - case EOF: - fatal("unmatched '{'"); + case tok_stropt: + case tok_intopt: + case tok_obsolete: + assert (0); + break; + case tok_illegal: default: - putc(c, faction); + complain (_("unrecognized: %s"), token_buffer); + skip_to_char ('%'); } + } + else if (c == EOF) + fatal (_("no input grammar")); + else + { + char buf[] = "c"; + buf[0] = c; + complain (_("unknown character: %s"), quote (buf)); + skip_to_char ('%'); + } + } +} + +/*-------------------------------------------------------------------. +| Assuming that a `{' has just been seen, copy everything up to the | +| matching `}' into the actions file. STACK_OFFSET is the number of | +| values in the current rule so far, which says where to find `$0' | +| with respect to the top of the stack. | +| | +| This routine is used both for actions and guards. Only | +| ACTION_OBSTACK is used, but this is fine, since we use only | +| pointers to relevant portions inside this obstack. | +`-------------------------------------------------------------------*/ + +static void +parse_braces (symbol_list *rule, int stack_offset) +{ + int c; + int count; - c = getc(finput); - } - - /* above loop exits when c is '}' */ + count = 1; + while (count > 0) + { + while ((c = getc (finput)) != '}') + 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; + + case '/': + copy_comment (finput, &action_obstack); + break; + + case '$': + copy_dollar (finput, &action_obstack, + rule, stack_offset); + break; + + case '@': + copy_at (finput, &action_obstack, + stack_offset); + break; + + case EOF: + fatal (_("unmatched %s"), "`{'"); + + default: + obstack_1grow (&action_obstack, c); + } + /* Above loop exits when C is '}'. */ if (--count) - { - putc(c, faction); - c = getc(finput); - } + obstack_1grow (&action_obstack, c); } - fprintf(faction, ";\n break;}"); + obstack_1grow (&action_obstack, '\0'); } +static void +parse_action (symbol_list *rule, int stack_offset) +{ + rule->action_line = lineno; + parse_braces (rule, stack_offset); + rule->action = obstack_finish (&action_obstack); +} -/* generate a dummy symbol, a nonterminal, -whose name cannot conflict with the user's names. */ -bucket * -gensym() +static void +parse_guard (symbol_list *rule, int stack_offset) { - register bucket *sym; - - sprintf (token_buffer, "@%d", ++gensym_count); - sym = getsym(token_buffer); - sym->class = SNTERM; - sym->value = nvars++; - return (sym); + 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); } -/* 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. */ +/*-------------------------------------------------------------------. +| Generate a dummy symbol, a nonterminal, whose name cannot conflict | +| with the user's names. | +`-------------------------------------------------------------------*/ -void -readgram() +static symbol_t * +gensym (void) { - register int t; - register bucket *lhs; - register symbol_list *p; - register symbol_list *p1; - register bucket *bp; - - symbol_list *crule; /* points to first symbol_list of current rule. */ - /* its symbol is the lhs of the rule. */ - symbol_list *crule1; /* points to the symbol_list preceding crule. */ - - p1 = NULL; + /* Incremented for each generated symbol */ + static int gensym_count = 0; + static char buf[256]; + + symbol_t *sym; + + sprintf (buf, "@%d", ++gensym_count); + token_buffer = buf; + sym = getsym (token_buffer); + sym->class = nterm_sym; + sym->number = nvars++; + return sym; +} + +/*-------------------------------------------------------------------. +| 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; + symbol_t *lhs = NULL; + symbol_list *p = NULL; + symbol_list *p1 = NULL; - t = lex(); + /* Points to first symbol_list of current rule. its symbol is the + lhs of the rule. */ + symbol_list *crule = NULL; + /* Points to the symbol_list preceding crule. */ + symbol_list *crule1 = NULL; - while (t != TWO_PERCENTS && t != ENDFILE) - { - if (t == IDENTIFIER || t == BAR) - { - register int actionflag = 0; - int rulelength = 0; /* number of symbols in rhs of this rule so far */ - int xactions = 0; /* JF for error checking */ - bucket *first_rhs = 0; + t = lex (); - if (t == IDENTIFIER) - { - lhs = symval; - - t = lex(); - if (t != COLON) - fatal("ill-formed rule"); - } + 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 */ + symbol_t *first_rhs = 0; - if (nrules == 0) - { - if (t == BAR) - fatal("grammar starts with vertical bar"); + if (t == tok_identifier) + { + lhs = symval; - if (!start_flag) + 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); + } + } - /* start a new rule and record its 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. */ - nrules++; - nitems++; + ++nrules; + ++nritems; - record_rule_line (); + p = symbol_list_new (lhs); - p = NEW(symbol_list); - p->sym = lhs; + crule1 = p1; + if (p1) + p1->next = p; + else + grammar = p; - crule1 = p1; - if (p1) - p1->next = p; - else - grammar = p; + p1 = p; + crule = p; - p1 = p; - crule = p; + /* mark the rule's lhs as a nonterminal if not already so. */ - /* mark the rule's lhs as a nonterminal if not already so. */ + if (lhs->class == unknown_sym) + { + lhs->class = nterm_sym; + lhs->number = nvars; + ++nvars; + } + else if (lhs->class == token_sym) + complain (_("rule given for %s, which is a token"), lhs->tag); - if (lhs->class == SUNKNOWN) - { - lhs->class = SNTERM; - lhs->value = nvars; - nvars++; - } - else if (lhs->class == STOKEN) - fatals("rule given for %s, which is a token", lhs->tag); + /* read the rhs of the rule. */ - /* read the rhs of the rule. */ + for (;;) + { + t = lex (); + if (t == tok_prec) + { + t = lex (); + crule->ruleprec = symval; + t = lex (); + } + + if (!(t == tok_identifier || t == tok_left_curly)) + break; - for (;;) - { - 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) - { - register bucket *ssave; - register 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 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 (actionflag) - { - register bucket *sdummy; - - /* Since the action was written out with this rule's */ - /* number, we must write 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 = NEW(symbol_list); - if (crule1) - crule1->next = p; - else grammar = p; - p->sym = sdummy; - crule1 = NEW(symbol_list); - p->next = crule1; - crule1->next = crule; - - /* insert the dummy generated by that rule into this rule. */ - nitems++; - p = NEW(symbol_list); - p->sym = sdummy; - p1->next = p; - p1 = p; - - actionflag = 0; - } - - if (t == IDENTIFIER) - { - nitems++; - p = NEW(symbol_list); - p->sym = symval; - p1->next = p; - p1 = p; - } - else /* handle an action. */ - { - copy_action(crule, rulelength); - actionflag = 1; - xactions++; /* JF */ - } - rulelength++; - } + /* If next token is an identifier, see if a colon follows it. + If one does, exit this rule now. */ + if (t == tok_identifier) + { + symbol_t *ssave; + token_t t1; + + ssave = symval; + t1 = lex (); + unlex (t1); + symval = ssave; + if (t1 == tok_colon) + { + warn (_("previous rule lacks an ending `;'")); + 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. */ + symbol_t *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; + ++nritems; + p = symbol_list_new (sdummy); + /* Attach its lineno to that of the host rule. */ + p->line = crule->line; + /* Move the action from the host rule to this one. */ + p->action = crule->action; + p->action_line = crule->action_line; + crule->action = NULL; + + 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. */ + ++nritems; + p = symbol_list_new (sdummy); + p1->next = p; + p1 = p; + + action_flag = 0; + } + + if (t == tok_identifier) + { + ++nritems; + 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 (); + } - /* Put an empty link in the list to mark the end of this rule */ - p = NEW(symbol_list); - p1->next = p; - p1 = p; + if (t == tok_guard) + { + if (!semantic_parser) + complain (_("%%guard present but %%semantic_parser not specified")); - if (t == PREC) - { - t = lex(); - crule->ruleprec = symval; - t = lex(); - } - if (t == GUARD) - { - if (! semantic_parser) - fatal("%guard present but %semantic_parser not specified"); + parse_guard (crule, rulelength); + t = lex (); + } - copy_guard(crule, rulelength); - t = lex(); - } - else if (t == LEFT_CURLY) - { - if (actionflag) fatal("two actions at end of one rule"); - copy_action(crule, rulelength); - t = lex(); - } - /* If $$ is being set in default way, - warn 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)) - fprintf(stderr, "%s:%d: warning: type clash ('%s' '%s') on default action\n", - infile, - lineno, + 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) - fprintf(stderr, - "%s:%d: warning: empty rule for typed nonterminal, and no action\n", - infile, - lineno); - if (t == SEMICOLON) - t = lex(); - } - /* these things can appear as alternatives to rules. */ - else if (t == TOKEN) - { - parse_token_decl(STOKEN, SNTERM); - t = lex(); - } - else if (t == NTERM) - { - parse_token_decl(SNTERM, STOKEN); - 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(); - } - else - fatal("invalid input"); - } - - if (nsyms > MAXSHORT) - fatals("too many symbols (tokens plus nonterminals); maximum %d", - MAXSHORT); - if (nrules == 0) - fatal("no input grammar"); - - if (typed == 0 /* JF put out same default YYSTYPE as YACC does */ - && !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) - if (bp->class == SUNKNOWN) + } + /* 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_two_percents || t == tok_eof) + warn (_("previous rule lacks an ending `;'")); + if (t == tok_semicolon) + t = lex (); + } + else { - fprintf(stderr, "symbol %s used, not defined as token, and no rules for it\n", - bp->tag); - failure = 1; - bp->class = SNTERM; - bp->value = nvars++; + complain (_("invalid input: %s"), quote (token_buffer)); + t = lex (); } - ntokens = nsyms - nvars; -} - + /* grammar has been read. Do some checking */ -void -record_rule_line () -{ - /* Record each rule's source line number in rline table. */ + if (nrules == 0) + fatal (_("no rules in the input grammar")); - if (nrules >= rline_allocated) - { - rline_allocated = nrules * 2; - rline = (short *) realloc (rline, - rline_allocated * sizeof (short)); - if (rline == 0) - { - fprintf (stderr, "%s: memory exhausted\n", program_name); - done (1); - } - } - rline[nrules] = lineno; + /* Report any undefined symbols and consider them nonterminals. */ + symbols_do (symbol_check_defined, NULL); + + /* 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; + nritems += 3; + grammar = p; + startval = axiom; + + if (nsyms > SHRT_MAX) + fatal (_("too many symbols (tokens plus nonterminals); maximum %d"), + SHRT_MAX); + + assert (nsyms == ntokens + nvars); } - -/* read in a %type declaration and record its information for get_type_name to access */ - -int -get_type() +/* 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 +read_additionnal_code (void) { - register int k; - register int t; - register char *name; - - t = lex(); + int c; + struct obstack el_obstack; - if (t != TYPENAME) - fatal("ill-formed %type declaration"); + obstack_init (&el_obstack); - k = strlen(token_buffer); - name = NEW2(k + 1, char); - strcpy(name, token_buffer); - - for (;;) + if (!no_lines_flag) { - t = lex(); - - switch (t) - { - case SEMICOLON: - return (lex()); - - case COMMA: - break; - - case IDENTIFIER: - if (symval->type_name == NULL) - symval->type_name = name; - else - fatals("type redeclaration for %s", symval->tag); - - break; - - default: - return (t); - } + 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)); +} -/* assign symbol numbers, and write definition of token names into fdefines. -Set up vectors tags and sprec of names and precedences of symbols. */ + +/*------------------------------------------------------------------. +| Set TOKEN_TRANSLATIONS. Check that no two symbols share the same | +| number. | +`------------------------------------------------------------------*/ -void -packsymbols() +static void +token_translations_init (void) { - register bucket *bp; - register int tokno = 1; - register int i; - register int last_user_token_number; - - /* int lossage = 0; JF set but not used */ - - tags = NEW2(nsyms + 1, char *); - tags[0] = "$"; - - sprec = NEW2(nsyms, short); - sassoc = NEW2(nsyms, short); - - max_user_token_number = 256; - last_user_token_number = 256; + int num_256_available_p = TRUE; + int i; - for (bp = firstsymbol; bp; bp = bp->next) + /* Find the highest user token number, and whether 256, the POSIX + preferred user token number for the error token, is used. */ + max_user_token_number = 0; + for (i = 0; i < ntokens; ++i) { - if (bp->class == SNTERM) + symbol_t *this = symbols[i]; + if (this->user_token_number != SUNDEF) { - bp->value += ntokens; + if (this->user_token_number > max_user_token_number) + max_user_token_number = this->user_token_number; + if (this->user_token_number == 256) + num_256_available_p = FALSE; } - else - { - if (translations && !(bp->user_token_number)) - 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; - bp->value = tokno++; - } - - tags[bp->value] = bp->tag; - sprec[bp->value] = bp->prec; - sassoc[bp->value] = bp->assoc; - } - if (translations) - { - register int i; + /* If 256 is not used, assign it to error, to follow POSIX. */ + if (num_256_available_p && errtoken->user_token_number == SUNDEF) + errtoken->user_token_number = 256; - token_translations = NEW2(max_user_token_number+1, short); + /* Set the missing user numbers. */ + if (max_user_token_number < 256) + max_user_token_number = 256; - /* initialize all entries for literal tokens to 2, - the internal token number for $illegal., 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) + for (i = 0; i < ntokens; ++i) { - if (bp->value >= ntokens) continue; - if (translations) - { - if (token_translations[bp->user_token_number] != 2) - { - /* JF made this a call to fatals() */ - fatals( "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; - } + symbol_t *this = symbols[i]; + if (this->user_token_number == SUNDEF) + this->user_token_number = ++max_user_token_number; + if (this->user_token_number > max_user_token_number) + max_user_token_number = this->user_token_number; } - error_token_number = errtoken->value; - - output_token_defines(ftable); - - if (startval->class == SUNKNOWN) - fatals("the start symbol %s is undefined", startval->tag); - else if (startval->class == STOKEN) - fatals("the start symbol %s is a token", startval->tag); + token_translations = XCALLOC (token_number_t, max_user_token_number + 1); - start_symbol = startval->value; - - if (definesflag) - { - output_token_defines(fdefines); - - if (!pure_parser) - { - if (spec_name_prefix) - fprintf(fdefines, "\nextern YYSTYPE %slval;\n", spec_name_prefix); - else - fprintf(fdefines, "\nextern YYSTYPE yylval;\n"); - } - - 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 - } + /* 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 + 1; i++) + token_translations[i] = undeftoken->number; + symbols_do (symbol_translation, NULL); } - -void -output_token_defines(file) -FILE *file; -{ - bucket *bp; - for (bp = firstsymbol; bp; bp = bp->next) - { - if (bp->value >= ntokens) continue; +/*----------------------------------------------------------------. +| Assign symbol numbers, and write definition of token names into | +| FDEFINES. Set up vectors SYMBOL_TABLE, TAGS of symbols. | +`----------------------------------------------------------------*/ - /* For named tokens, but not literal ones, define the name. */ - /* The value is the user token number. */ +static void +packsymbols (void) +{ + symbols = XCALLOC (symbol_t *, nsyms); - if ('\'' != *tags[bp->value] && bp != errtoken) - { - register char *cp = tags[bp->value]; - register char c; + symbols_do (symbol_check_alias_consistence, NULL); + symbols_do (symbol_pack, NULL); - /* Don't #define nonliteral tokens whose names contain periods. */ + token_translations_init (); - while ((c = *cp++) && c != '.'); - if (!c) - { - fprintf(file, "#define\t%s\t%d\n", tags[bp->value], - (translations ? bp->user_token_number : bp->value)); - if (semantic_parser) - fprintf(file, "#define\tT%s\t%d\n", tags[bp->value], - bp->value); - } - } - } + 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); - putc('\n', file); + start_symbol = startval->number; } +/*---------------------------------------------------------------. +| Convert the rules into the representation using RRHS, RLHS and | +| RITEM. | +`---------------------------------------------------------------*/ -/* convert the rules into the representation using rrhs, rlhs and ritems. */ - -void -packgram() +static void +packgram (void) { - register int itemno; - register int ruleno; - register symbol_list *p; -/* register bucket *bp; JF unused */ - - bucket *ruleprec; + int itemno; + int ruleno; + symbol_list *p; - ritem = NEW2(nitems + 1, short); - rlhs = NEW2(nrules, short) - 1; - rrhs = NEW2(nrules, short) - 1; - rprec = NEW2(nrules, short) - 1; - rprecsym = NEW2(nrules, short) - 1; - rassoc = NEW2(nrules, short) - 1; + ritem = XCALLOC (item_number_t, nritems + 1); + rules = XCALLOC (rule_t, nrules) - 1; itemno = 0; ruleno = 1; @@ -1780,65 +1769,136 @@ packgram() p = grammar; while (p) { - rlhs[ruleno] = p->sym->value; - rrhs[ruleno] = itemno; - ruleprec = p->ruleprec; + symbol_t *ruleprec = p->ruleprec; + rules[ruleno].user_number = ruleno; + rules[ruleno].number = ruleno; + rules[ruleno].lhs = p->sym; + rules[ruleno].rhs = ritem + 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) { - ritem[itemno++] = p->sym->value; + /* item_number_t = token_number_t. + But the former needs to contain more: negative rule numbers. */ + ritem[itemno++] = token_number_as_item_number (p->sym->number); /* A rule gets by default the precedence and associativity of the last token in it. */ - if (p->sym->class == STOKEN) - { - rprec[ruleno] = p->sym->prec; - rassoc[ruleno] = p->sym->assoc; - } - if (p) p = p->next; + if (p->sym->class == token_sym) + rules[ruleno].prec = p->sym; + if (p) + p = p->next; } /* If this rule has a %prec, - the specified symbol's precedence replaces the default. */ + the specified symbol's precedence replaces the default. */ if (ruleprec) { - rprec[ruleno] = ruleprec->prec; - rassoc[ruleno] = ruleprec->assoc; - rprecsym[ruleno] = ruleprec->value; + rules[ruleno].precsym = ruleprec; + rules[ruleno].prec = ruleprec; } - ritem[itemno++] = -ruleno; - ruleno++; + ++ruleno; - if (p) p = p->next; + if (p) + p = p->next; } ritem[itemno] = 0; + assert (itemno == nritems); + + if (trace_flag) + ritem_print (stderr); } -/* Read a signed integer from STREAM and return its value. */ +/*-------------------------------------------------------------------. +| Read in the grammar specification and record it in the format | +| described in gram.h. All guards are copied into the GUARD_OBSTACK | +| and all actions into ACTION_OBSTACK, in each case forming the body | +| of a C function (YYGUARD or YYACTION) which contains a switch | +| statement to decide which guard or action to execute. | +`-------------------------------------------------------------------*/ -int -read_signed_integer (stream) - FILE *stream; +void +reader (void) { - register int c = getc(stream); - register int sign = 1; - register int n; + lex_init (); + lineno = 1; - if (c == '-') - { - c = getc(stream); - sign = -1; - } - n = 0; - while (isdigit(c)) + /* Initialize the muscle obstack. */ + obstack_init (&muscle_obstack); + + /* Initialize the symbol table. */ + symbols_new (); + + /* Construct the axiom symbol. */ + axiom = getsym ("$axiom"); + axiom->class = nterm_sym; + axiom->number = nvars++; + + /* Construct the error token */ + errtoken = getsym ("error"); + errtoken->class = token_sym; + errtoken->number = ntokens++; + + /* Construct a token that represents all undefined literal tokens. + It is always token number 2. */ + undeftoken = getsym ("$undefined."); + undeftoken->class = token_sym; + undeftoken->number = ntokens++; + + /* Initialize the obstacks. */ + obstack_init (&action_obstack); + obstack_init (&attrs_obstack); + obstack_init (&output_obstack); + + finput = xfopen (infile, "r"); + + /* Read the declaration section. Copy %{ ... %} groups to + TABLE_OBSTACK and FDEFINES file. Also notice any %token, %left, + etc. found there. */ + read_declarations (); + + /* If the user did not define her EOFTOKEN, do it now. */ + if (!eoftoken) { - n = 10*n + (c - '0'); - c = getc(stream); + eoftoken = getsym ("$"); + eoftoken->class = token_sym; + eoftoken->number = 0; + /* Value specified by POSIX. */ + eoftoken->user_token_number = 0; } - ungetc(c, stream); + /* 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 (); + + lex_free (); + xfclose (finput); + + /* 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 (); - return n * sign; + /* The grammar as a symbol_list is no longer needed. */ + LIST_FREE (symbol_list, grammar); +} + +void +grammar_free (void) +{ + XFREE (ritem); + free (rules + 1); + /* Free the symbol table data structure. */ + symbols_free (); }