/* Input parser for bison
- Copyright (C) 1984, 86, 89, 92, 98, 2000 Free Software Foundation, Inc.
+ Copyright 1984, 1986, 1989, 1992, 1998, 2000, 2001
+ Free Software Foundation, Inc.
This file is part of Bison, the GNU Compiler Compiler.
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 <stdio.h>
#include "system.h"
+#include "obstack.h"
+#include "quotearg.h"
+#include "quote.h"
+#include "getargs.h"
#include "files.h"
-#include "alloc.h"
#include "symtab.h"
+#include "options.h"
#include "lex.h"
#include "gram.h"
-#include "machine.h"
-
-#define LTYPESTR "\
-\n\
-#ifndef YYLTYPE\n\
-typedef\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"
+#include "complain.h"
+#include "output.h"
+#include "reader.h"
+#include "conflicts.h"
+#include "muscle_tab.h"
/* 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 int noparserflag;
-extern int rawtoknumflag;
-extern bucket *symval;
-extern int numval;
-extern int expected_conflicts;
-extern char *token_buffer;
-extern int maxtoken;
-
-extern void init_lex PARAMS((void));
-extern char *grow_token_buffer PARAMS((char *));
-extern void tabinit PARAMS((void));
-extern void output_headers PARAMS((void));
-extern void output_trailers PARAMS((void));
-extern void free_symtab PARAMS((void));
-extern void open_extra_files PARAMS((void));
-extern char *int_to_string PARAMS((int));
-extern char *printable_version PARAMS((int));
-extern void fatal PARAMS((char *));
-extern void fatals PARAMS((char *, char *));
-extern void warn PARAMS((char *));
-extern void warni PARAMS((char *, int));
-extern void warns PARAMS((char *, char *));
-extern void warnss PARAMS((char *, char *, char *));
-extern void warnsss PARAMS((char *, char *, char *, char *));
-extern void unlex PARAMS((int));
-extern void done PARAMS((int));
-
-extern int skip_white_space PARAMS((void));
-extern int parse_percent_token PARAMS((void));
-extern int lex PARAMS((void));
-
-typedef
- struct symbol_list
- {
- struct symbol_list *next;
- bucket *sym;
- bucket *ruleprec;
- }
- symbol_list;
-
-
-void reader PARAMS((void));
-void reader_output_yylsp PARAMS((FILE *));
-void read_declarations PARAMS((void));
-void copy_definition PARAMS((void));
-void parse_token_decl PARAMS((int, int));
-void parse_start_decl PARAMS((void));
-void parse_type_decl PARAMS((void));
-void parse_assoc_decl PARAMS((int));
-void parse_union_decl PARAMS((void));
-void parse_expect_decl PARAMS((void));
-char *get_type_name PARAMS((int, symbol_list *));
-void copy_guard PARAMS((symbol_list *, int));
-void parse_thong_decl PARAMS((void));
-void copy_action PARAMS((symbol_list *, int));
-bucket *gensym PARAMS((void));
-void readgram PARAMS((void));
-void record_rule_line PARAMS((void));
-void packsymbols PARAMS((void));
-void output_token_defines PARAMS((FILE *));
-void packgram PARAMS((void));
-int read_signed_integer PARAMS((FILE *));
+static int rline_allocated;
-#if 0
-static int get_type PARAMS((void));
-#endif
+typedef struct symbol_list
+{
+ struct symbol_list *next;
+ bucket *sym;
+ bucket *ruleprec;
+}
+symbol_list;
int lineno;
-symbol_list *grammar;
-int start_flag;
-bucket *startval;
char **tags;
-int *user_toknums;
+short *user_toknums;
+static symbol_list *grammar;
+static int start_flag;
+static bucket *startval;
/* 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;
-static int lastprec; /* incremented for each %left, %right or %nonassoc seen */
-
-static int gensym_count; /* incremented for each generated symbol */
+/* Incremented for each %left, %right or %nonassoc seen */
+static int lastprec;
static bucket *errtoken;
static bucket *undeftoken;
+\f
-/* Nonzero if any action or guard uses the @n construct. */
-static int yylsp_needed;
-
+/*===================\
+| Low level lexing. |
+\===================*/
static void
skip_to_char (int target)
{
int c;
if (target == '\n')
- warn(_(" Skipping to next \\n"));
+ complain (_(" Skipping to next \\n"));
else
- warni(_(" Skipping to next %c"), target);
+ complain (_(" Skipping to next %c"), target);
do
- c = skip_white_space();
+ c = skip_white_space ();
while (c != target && c != EOF);
if (c != EOF)
- ungetc(c, finput);
+ ungetc (c, finput);
}
-/* Dump the string from FINPUT to FOUTPUT. MATCH is the delimiter of
- the string (either ' or "). */
+/*---------------------------------------------------------.
+| 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))
+ {
+ n = 10 * n + (c - '0');
+ c = getc (stream);
+ }
+
+ ungetc (c, stream);
+
+ return sign * n;
+}
+\f
+/*--------------------------------------------------------------.
+| 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;
+}
+\f
+/*------------------------------------------------------------.
+| 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 *finput, FILE *foutput, int match)
+copy_string2 (FILE *fin, struct obstack *oout, int match, int store)
{
int c;
- putc (match, foutput);
- c = getc (finput);
+ if (store)
+ obstack_1grow (oout, match);
+
+ c = getc (fin);
while (c != match)
{
fatal (_("unterminated string at end of file"));
if (c == '\n')
{
- warn (_("unterminated string"));
- ungetc (c, finput);
+ complain (_("unterminated string"));
+ ungetc (c, fin);
c = match; /* invent terminator */
continue;
}
- putc(c, foutput);
+ obstack_1grow (oout, c);
if (c == '\\')
{
- c = getc (finput);
+ c = getc (fin);
if (c == EOF)
fatal (_("unterminated string at end of file"));
- putc (c, foutput);
+ obstack_1grow (oout, c);
+
if (c == '\n')
lineno++;
}
- c = getc(finput);
+ c = getc (fin);
}
- putc(c, foutput);
+ 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 comment from FINPUT to FOUTPUT. C is either `*' or `/',
- depending upon the type of comments used. */
+/*-----------------------------------------------------------------.
+| Dump the wannabee comment from IN to OUT1 and OUT2 (which can be |
+| NULL). In fact we just saw a `/', which might or might not be a |
+| comment. In any case, copy what we saw. |
+| |
+| OUT2 might be NULL. |
+`-----------------------------------------------------------------*/
static inline void
-copy_comment (FILE *finput, FILE *foutput, int c)
+copy_comment2 (FILE *fin, struct obstack *oout1, struct obstack *oout2)
{
int cplus_comment;
- register int match;
- register int ended;
+ int ended;
+ int c;
- cplus_comment = (c == '/');
- putc (c, foutput);
- c = getc (finput);
+ /* We read a `/', output it. */
+ obstack_1grow (oout1, '/');
+ if (oout2)
+ obstack_1grow (oout2, '/');
+
+ switch ((c = getc (fin)))
+ {
+ case '/':
+ cplus_comment = 1;
+ break;
+ case '*':
+ cplus_comment = 0;
+ break;
+ default:
+ ungetc (c, fin);
+ return;
+ }
+
+ obstack_1grow (oout1, c);
+ if (oout2)
+ obstack_1grow (oout2, c);
+ c = getc (fin);
ended = 0;
while (!ended)
{
while (c == '*')
{
- putc(c, foutput);
- c = getc(finput);
+ obstack_1grow (oout1, c);
+ if (oout2)
+ obstack_1grow (oout2, c);
+ c = getc (fin);
}
if (c == '/')
{
- putc(c, foutput);
+ obstack_1grow (oout1, c);
+ if (oout2)
+ obstack_1grow (oout2, c);
ended = 1;
}
}
else if (c == '\n')
{
lineno++;
- putc (c, foutput);
+ obstack_1grow (oout1, c);
+ if (oout2)
+ obstack_1grow (oout2, c);
if (cplus_comment)
ended = 1;
else
- c = getc(finput);
+ c = getc (fin);
}
else if (c == EOF)
fatal (_("unterminated comment"));
else
{
- putc (c, foutput);
- c = getc (finput);
+ obstack_1grow (oout1, c);
+ if (oout2)
+ obstack_1grow (oout2, c);
+ c = getc (fin);
}
}
}
-void
-reader (void)
-{
- start_flag = 0;
- startval = NULL; /* start symbol not specified yet. */
-
-#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;
+/*-------------------------------------------------------------------.
+| Dump the comment (actually the current string starting with a `/') |
+| from FIN to OOUT. |
+`-------------------------------------------------------------------*/
- nsyms = 1;
- nvars = 0;
- nrules = 0;
- nitems = 0;
- rline_allocated = 10;
- rline = NEW2(rline_allocated, short);
+static inline void
+copy_comment (FILE *fin, struct obstack *oout)
+{
+ copy_comment2 (fin, oout, NULL);
+}
- typed = 0;
- lastprec = 0;
- gensym_count = 0;
+/*-----------------------------------------------------------------.
+| 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 `$0' with |
+| respect to the top of the stack. |
+`-----------------------------------------------------------------*/
- semantic_parser = 0;
- pure_parser = 0;
- yylsp_needed = 0;
+static inline void
+copy_at (FILE *fin, struct obstack *oout, int stack_offset)
+{
+ int c;
- grammar = NULL;
+ c = getc (fin);
+ if (c == '$')
+ {
+ obstack_sgrow (oout, "yyloc");
+ locations_flag = 1;
+ }
+ else if (isdigit (c) || c == '-')
+ {
+ int n;
- init_lex();
- lineno = 1;
+ ungetc (c, fin);
+ n = read_signed_integer (fin);
- /* 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. */
- undeftoken = getsym("$undefined.");
- undeftoken->class = STOKEN;
- undeftoken->user_token_number = 2;
- /* Read the declaration section. Copy %{ ... %} groups to ftable and fdefines file.
- Also notice any %token, %left, etc. found there. */
- if (noparserflag)
- fprintf(ftable, "\n/* Bison-generated parse tables, made from %s\n",
- infile);
+ obstack_fgrow1 (oout, "yylsp[%d]", n - stack_offset);
+ locations_flag = 1;
+ }
else
- fprintf(ftable, "\n/* A Bison parser, made from %s\n", infile);
- fprintf(ftable, " by %s */\n\n", VERSION_STRING);
- fprintf(ftable, "#define YYBISON 1 /* Identify Bison output. */\n\n");
- read_declarations();
- /* start writing the guard and action files, if they are needed. */
- output_headers();
- /* read in the grammar, build grammar in list form. write out guards and actions. */
- readgram();
- /* Now we know whether we need the line-number stack.
- If we do, write its type into the .tab.h file. */
- if (fdefines)
- reader_output_yylsp(fdefines);
- /* write closing delimiters for actions and guards. */
- output_trailers();
- if (yylsp_needed)
- fprintf(ftable, "#define YYLSP_NEEDED\n\n");
- /* assign the symbols their symbol numbers.
- Write #defines for the token symbols into fdefines if requested. */
- packsymbols();
- /* convert the grammar into the format described in gram.h. */
- packgram();
- /* free the symbol table data structure
- since symbols are now all referred to by symbol number. */
- free_symtab();
+ {
+ char buf[] = "@c";
+ buf[1] = c;
+ complain (_("%s is invalid"), quote (buf));
+ }
}
-void
-reader_output_yylsp (FILE *f)
-{
- if (yylsp_needed)
- fprintf(f, LTYPESTR);
-}
-/* Read from finput until `%%' is seen. Discard the `%%'. Handle any
- `%' declarations, and copy the contents of any `%{ ... %}' groups
- to fattrs. */
+/*-------------------------------------------------------------------.
+| FIN is pointing to a wannabee semantic value (i.e., a `$'). |
+| |
+| Possible inputs: $[<TYPENAME>]($|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 `$0' with respect to the top of the stack. |
+`-------------------------------------------------------------------*/
-void
-read_declarations (void)
+static inline void
+copy_dollar (FILE *fin, struct obstack *oout,
+ symbol_list *rule, int stack_offset)
{
- register int c;
- register int tok;
+ int c = getc (fin);
+ const char *type_name = NULL;
- for (;;)
+ /* Get the type name if explicit. */
+ if (c == '<')
{
- c = skip_white_space();
-
- if (c == '%')
- {
- tok = parse_percent_token();
-
- switch (tok)
- {
- case TWO_PERCENTS:
- return;
-
- case PERCENT_LEFT_CURLY:
- copy_definition();
- break;
-
- case TOKEN:
- parse_token_decl (STOKEN, SNTERM);
- break;
-
- case NTERM:
- parse_token_decl (SNTERM, STOKEN);
- break;
-
- case TYPE:
- parse_type_decl();
- break;
-
- case START:
- parse_start_decl();
- break;
-
- case UNION:
- parse_union_decl();
- break;
-
- case EXPECT:
- parse_expect_decl();
- break;
- case THONG:
- parse_thong_decl();
- break;
- case LEFT:
- parse_assoc_decl(LEFT_ASSOC);
- break;
-
- case RIGHT:
- parse_assoc_decl(RIGHT_ASSOC);
- break;
-
- case NONASSOC:
- parse_assoc_decl(NON_ASSOC);
- break;
+ read_type_name (fin);
+ type_name = token_buffer;
+ value_components_used = 1;
+ c = getc (fin);
+ }
- case SEMANTIC_PARSER:
- if (semantic_parser == 0)
- {
- semantic_parser = 1;
- open_extra_files();
- }
- break;
+ if (c == '$')
+ {
+ 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);
- case PURE_PARSER:
- pure_parser = 1;
- break;
+ if (!type_name && n > 0)
+ type_name = get_type_name (n, rule);
- case NOOP:
- break;
+ obstack_fgrow1 (oout, "yyvsp[%d]", n - stack_offset);
- default:
- warns(_("unrecognized: %s"), token_buffer);
- skip_to_char('%');
- }
- }
- else if (c == EOF)
- fatal(_("no input grammar"));
- else
- {
- warns (_("unknown character: %s"), printable_version(c));
- skip_to_char('%');
- }
+ 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));
}
}
+\f
+/*-------------------------------------------------------------------.
+| Copy the contents of a `%{ ... %}' into the definitions file. The |
+| `%{' has already been read. Return after reading the `%}'. |
+`-------------------------------------------------------------------*/
-
-/* Copy the contents of a `%{ ... %}' into the definitions file. The
- `%{' has already been read. Return after reading the `%}'. */
-
-void
+static void
copy_definition (void)
{
- register int c;
+ int c;
/* -1 while reading a character if prev char was %. */
- register int after_percent;
+ int after_percent;
- if (!nolinesflag)
- fprintf(fattrs, "#line %d \"%s\"\n", lineno, infile);
+#if 0
+ if (!no_lines_flag)
+ {
+ obstack_fgrow2 (&attrs_obstack, muscle_find ("linef"),
+ lineno, quotearg_style (c_quoting_style,
+ muscle_find("filename")));
+ }
+#endif
after_percent = 0;
switch (c)
{
case '\n':
- putc(c, fattrs);
+ obstack_1grow (&attrs_obstack, c);
lineno++;
break;
case '%':
- after_percent = -1;
+ after_percent = -1;
break;
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(_("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 (int what_is, int what_is_not)
+static void
+parse_token_decl (symbol_class what_is, symbol_class what_is_not)
{
- register int token = 0;
- register char *typename = 0;
- register struct bucket *symbol = NULL; /* pts to symbol being defined */
- int k;
+ 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)
- fatals ("Premature EOF after %s", token_buffer);
+ fatal (_("Premature EOF after %s"), token_buffer);
- token = lex();
- if (token == COMMA)
+ token = lex ();
+ if (token == tok_comma)
{
symbol = NULL;
continue;
}
- if (token == TYPENAME)
+ 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 == IDENTIFIER && *symval->tag == '\"'
- && symbol)
+ else if (token == tok_identifier && *symval->tag == '\"' && symbol)
{
- translations = 1;
- symval->class = STOKEN;
- symval->type_name = typename;
- symval->user_token_number = symbol->user_token_number;
- symbol->user_token_number = SALIAS;
-
- symval->alias = symbol;
- symbol->alias = symval;
- symbol = NULL;
-
- nsyms--; /* symbol and symval combined are only one symbol */
+ 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--;
+ }
+ symbol = NULL;
}
- else if (token == IDENTIFIER)
+ else if (token == tok_identifier)
{
int oldclass = symval->class;
symbol = symval;
if (symbol->class == what_is_not)
- warns(_("symbol %s redefined"), symbol->tag);
+ complain (_("symbol %s redefined"), symbol->tag);
symbol->class = what_is;
- if (what_is == SNTERM && oldclass != SNTERM)
+ if (what_is == nterm_sym && oldclass != nterm_sym)
symbol->value = nvars++;
if (typename)
{
if (symbol->type_name == NULL)
symbol->type_name = typename;
- else if (strcmp(typename, symbol->type_name) != 0)
- warns(_("type redeclaration for %s"), symbol->tag);
+ else if (strcmp (typename, symbol->type_name) != 0)
+ 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;
- }
+ }
else
{
- warnss(_("`%s' is invalid in %s"),
- token_buffer,
- (what_is == STOKEN) ? "%token" : "%nterm");
- skip_to_char('%');
+ complain (_("`%s' is invalid in %s"),
+ token_buffer, (what_is == token_sym) ? "%token" : "%nterm");
+ skip_to_char ('%');
}
}
}
-/* parse what comes after %thong
- the full syntax is
- %thong <type> token number literal
- the <type> 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 <type>, 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
-*/
-
-void
-parse_thong_decl (void)
-{
- register int token;
- register struct bucket *symbol;
- register char *typename = 0;
- int k, usrtoknum;
-
- translations = 1;
- token = lex(); /* fetch typename or first token */
- if (token == TYPENAME) {
- k = strlen(token_buffer);
- typename = NEW2(k + 1, char);
- strcpy(typename, token_buffer);
- value_components_used = 1;
- token = lex(); /* fetch first token */
- }
-
- /* process first token */
-
- if (token != IDENTIFIER)
- {
- warns(_("unrecognized item %s, expected an identifier"),
- token_buffer);
- skip_to_char('%');
- return;
- }
- symval->class = STOKEN;
- symval->type_name = typename;
- symval->user_token_number = SALIAS;
- symbol = symval;
-
- token = lex(); /* get number or literal string */
-
- if (token == NUMBER) {
- usrtoknum = numval;
- token = lex(); /* okay, did number, now get literal */
- }
- else usrtoknum = 0;
-
- /* process literal string token */
-
- if (token != IDENTIFIER || *symval->tag != '\"')
- {
- warns(_("expected string constant instead of %s"),
- token_buffer);
- skip_to_char('%');
- return;
- }
- symval->class = STOKEN;
- symval->type_name = typename;
- symval->user_token_number = usrtoknum;
-
- symval->alias = symbol;
- symbol->alias = symval;
-
- nsyms--; /* symbol and symval combined are only one symbol */
-}
-
-/* parse what comes after %start */
+/*------------------------------.
+| Parse what comes after %start |
+`------------------------------*/
-void
+static void
parse_start_decl (void)
{
if (start_flag)
- warn(_("multiple %start declarations"));
- if (lex() != IDENTIFIER)
- warn(_("invalid %start declaration"));
+ complain (_("multiple %s declarations"), "%start");
+ if (lex () != tok_identifier)
+ complain (_("invalid %s declaration"), "%start");
else
{
start_flag = 1;
}
}
+/*-----------------------------------------------------------.
+| 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
+static void
parse_type_decl (void)
{
- register int k;
- register char *name;
+ char *name;
- if (lex() != TYPENAME)
+ if (lex () != tok_typename)
{
- warn(_("%type declaration has no <typename>"));
- skip_to_char('%');
+ complain ("%s", _("%type declaration has no <typename>"));
+ skip_to_char ('%');
return;
}
- k = strlen(token_buffer);
- name = NEW2(k + 1, char);
- strcpy(name, token_buffer);
+ name = xstrdup (token_buffer);
for (;;)
{
- register int t;
+ token_t t;
int tmp_char = ungetc (skip_white_space (), finput);
if (tmp_char == '%')
return;
if (tmp_char == EOF)
- fatals ("Premature EOF after %s", token_buffer);
+ fatal (_("Premature EOF after %s"), token_buffer);
- 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 if (strcmp(name, symval->type_name) != 0)
- warns(_("type redeclaration for %s"), symval->tag);
+ else if (strcmp (name, symval->type_name) != 0)
+ complain (_("type redeclaration for %s"), symval->tag);
break;
default:
- warns(_("invalid %%type declaration due to item: `%s'"), token_buffer);
- skip_to_char('%');
+ 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 (int assoc)
+static void
+parse_assoc_decl (associativity assoc)
{
- register int k;
- register char *name = NULL;
- register int prev = 0;
+ char *name = NULL;
+ int prev = 0;
- lastprec++; /* Assign a new precedence level, never 0. */
+ lastprec++; /* Assign a new precedence level, never 0. */
for (;;)
{
- register int t;
+ token_t t;
int tmp_char = ungetc (skip_white_space (), finput);
if (tmp_char == '%')
return;
if (tmp_char == EOF)
- fatals ("Premature EOF after %s", token_buffer);
+ fatal (_("Premature EOF after %s"), token_buffer);
- 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)
- warns(_("redefining precedence of %s"), symval->tag);
+ complain (_("redefining precedence of %s"), symval->tag);
symval->prec = lastprec;
symval->assoc = assoc;
- if (symval->class == SNTERM)
- warns(_("symbol %s redefined"), symval->tag);
- symval->class = STOKEN;
+ if (symval->class == nterm_sym)
+ complain (_("symbol %s redefined"), symval->tag);
+ 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 if (strcmp(name, symval->type_name) != 0)
- warns(_("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
- {
- warns(_("invalid text (%s) - number should be after identifier"),
- token_buffer);
- skip_to_char('%');
- }
+ }
+ else
+ {
+ complain (_
+ ("invalid text (%s) - number should be after identifier"),
+token_buffer);
+ skip_to_char ('%');
+ }
break;
- case SEMICOLON:
+ case tok_semicolon:
return;
default:
- warns(_("unexpected item: %s"), token_buffer);
- skip_to_char('%');
+ 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
+static void
parse_union_decl (void)
{
- register int c;
- register int count;
- register int in_comment;
- int cplus_comment;
+ int c;
+ int count = 0;
+ struct obstack union_obstack;
if (typed)
- warn(_("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");
+ /* FIXME: I'm worried: are you sure attrs_obstack is properly
+ filled? */
+ if (no_lines_flag)
+ obstack_1grow (&attrs_obstack, '\n');
- count = 0;
- in_comment = 0;
+ obstack_init (&union_obstack);
+ obstack_sgrow (&union_obstack, "union");
+ if (defines_flag)
+ obstack_sgrow (&defines_obstack, "typedef union");
- c = getc(finput);
+ c = getc (finput);
while (c != EOF)
{
- putc(c, fattrs);
- if (fdefines)
- putc(c, fdefines);
+ /* If C contains '/', it is output by copy_comment (). */
+ if (c != '/')
+ {
+ obstack_1grow (&union_obstack, c);
+ if (defines_flag)
+ obstack_1grow (&defines_obstack, c);
+ }
switch (c)
{
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 at end of file"));
-
- if (!cplus_comment && c == '*')
- {
- c = getc(finput);
- if (c == '/')
- {
- putc('/', fattrs);
- if (fdefines)
- putc('/', fdefines);
- in_comment = 0;
- }
- }
- else
- c = getc(finput);
- }
- }
+ copy_comment2 (finput, &defines_obstack, &union_obstack);
break;
-
case '{':
count++;
break;
case '}':
if (count == 0)
- warn (_("unmatched close-brace (`}')"));
+ complain (_("unmatched %s"), "`}'");
count--;
if (count <= 0)
{
- fprintf(fattrs, " YYSTYPE;\n");
- if (fdefines)
- fprintf(fdefines, " YYSTYPE;\n");
+ if (defines_flag)
+ obstack_sgrow (&defines_obstack, " YYSTYPE;\n");
/* JF don't choke on trailing semi */
- c=skip_white_space();
- if(c!=';') ungetc(c,finput);
+ c = skip_white_space ();
+ if (c != ';')
+ ungetc (c, finput);
+ obstack_1grow (&union_obstack, 0);
+ muscle_insert ("stype", obstack_finish (&union_obstack));
return;
}
}
- c = getc(finput);
+ c = getc (finput);
}
+
}
-/* parse the declaration %expect N which says to expect N
- shift-reduce conflicts. */
-void
+/*-------------------------------------------------------.
+| Parse the declaration %expect N which says to expect N |
+| shift-reduce conflicts. |
+`-------------------------------------------------------*/
+
+static void
parse_expect_decl (void)
{
- register int c;
- register int count;
- char buffer[20];
+ int c = skip_white_space ();
+ ungetc (c, finput);
- c = getc(finput);
- while (c == ' ' || c == '\t')
- c = getc(finput);
+ if (!isdigit (c))
+ complain (_("argument of %%expect is not an integer"));
+ else
+ expected_conflicts = read_signed_integer (finput);
+}
- count = 0;
- while (c >= '0' && c <= '9')
+
+/*-------------------------------------------------------------------.
+| Parse what comes after %thong. the full syntax is |
+| |
+| %thong <type> token number literal |
+| |
+| the <type> 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 <type>, 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)
+{
+ token_t token;
+ struct bucket *symbol;
+ char *typename = 0;
+ int usrtoknum = SUNDEF;
+
+ token = lex (); /* fetch typename or first token */
+ if (token == tok_typename)
{
- if (count < 20)
- buffer[count++] = c;
- c = getc(finput);
+ typename = xstrdup (token_buffer);
+ value_components_used = 1;
+ token = lex (); /* fetch first token */
}
- buffer[count] = 0;
- ungetc (c, finput);
+ /* process first token */
- if (count <= 0 || count > 10)
- warn(_("argument of %expect is not an integer"));
- expected_conflicts = atoi (buffer);
-}
+ 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;
-/* that's all of parsing the declaration section */
-\f
-/* Get the data type (alternative in the union) of the value for symbol n in rule rule. */
+ token = lex (); /* get number or literal string */
-char *
-get_type_name (int n, symbol_list *rule)
-{
- static char *msg = N_("invalid $ value");
+ if (token == tok_number)
+ {
+ usrtoknum = numval;
+ token = lex (); /* okay, did number, now get literal */
+ }
- register int i;
- register symbol_list *rp;
+ /* process literal string token */
- if (n < 0)
+ if (token != tok_identifier || *symval->tag != '\"')
{
- warn(_(msg));
- return NULL;
+ 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;
- rp = rule;
- i = 0;
+ symval->alias = symbol;
+ symbol->alias = symval;
- while (i < n)
+ /* 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 != '_')
{
- rp = rp->next;
- if (rp == NULL || rp->sym == NULL)
+ 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)
{
- warn(_(msg));
- return NULL;
+ complain (_("invalid %s declaration"), "%define");
+ skip_to_char ('%');
+ return;
}
- i++;
+ else
+ fatal (_("Premature EOF after %s"), "\"");
}
+ copy_string2 (finput, &muscle_obstack, '"', 0);
+ obstack_1grow (&muscle_obstack, 0);
+ muscle_value = obstack_finish (&muscle_obstack);
- return (rp->sym->type_name);
+ /* Store the (key, value) pair in the environment. */
+ muscle_insert (muscle_key, muscle_value);
}
-
-/* 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. */
+/*----------------------------------.
+| Parse what comes after %skeleton. |
+`----------------------------------*/
void
+parse_skel_decl (void)
+{
+ /* Complete with parse_dquoted_param () on the CVS branch 1.29. */
+}
+
+/*----------------------------------------------------------------.
+| Read from finput until `%%' is seen. Discard the `%%'. Handle |
+| any `%' declarations, and copy the contents of any `%{ ... %}' |
+| groups to ATTRS_OBSTACK. |
+`----------------------------------------------------------------*/
+
+static void
+read_declarations (void)
+{
+ for (;;)
+ {
+ int c = skip_white_space ();
+
+ if (c == '%')
+ {
+ token_t tok = parse_percent_token ();
+
+ switch (tok)
+ {
+ case tok_two_percents:
+ return;
+
+ case tok_percent_left_curly:
+ copy_definition ();
+ break;
+
+ case tok_token:
+ parse_token_decl (token_sym, nterm_sym);
+ break;
+
+ case tok_nterm:
+ parse_token_decl (nterm_sym, token_sym);
+ break;
+
+ case tok_type:
+ parse_type_decl ();
+ break;
+
+ case tok_start:
+ parse_start_decl ();
+ break;
+
+ case tok_union:
+ parse_union_decl ();
+ break;
+
+ case tok_expect:
+ parse_expect_decl ();
+ break;
+
+ case tok_thong:
+ parse_thong_decl ();
+ break;
+
+ case tok_left:
+ parse_assoc_decl (left_assoc);
+ break;
+
+ case tok_right:
+ parse_assoc_decl (right_assoc);
+ break;
+
+ case tok_nonassoc:
+ parse_assoc_decl (non_assoc);
+ break;
+
+ case tok_define:
+ parse_muscle_decl ();
+ break;
+
+ case tok_skel:
+ parse_skel_decl ();
+ break;
+
+ case tok_noop:
+ break;
+
+ case tok_stropt:
+ case tok_intopt:
+ case tok_obsolete:
+ case tok_illegal:
+ abort ();
+ break;
+
+ default:
+ 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 ('%');
+ }
+ }
+}
+\f
+/*-------------------------------------------------------------------.
+| 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. |
+`-------------------------------------------------------------------*/
+
+static void
+copy_action (symbol_list *rule, int stack_offset)
+{
+ int c;
+ int count;
+ char buf[4096];
+
+ /* offset is always 0 if parser has already popped the stack pointer */
+ if (semantic_parser)
+ stack_offset = 0;
+
+ obstack_fgrow1 (&action_obstack, "\ncase %d:\n", nrules);
+
+ if (!no_lines_flag)
+ {
+ obstack_fgrow2 (&action_obstack, muscle_find ("linef"),
+ lineno, quotearg_style (c_quoting_style,
+ muscle_find ("filename")));
+ }
+ obstack_1grow (&action_obstack, '{');
+
+ count = 1;
+ c = getc (finput);
+
+ while (count > 0)
+ {
+ while (c != '}')
+ {
+ switch (c)
+ {
+ case '\n':
+ obstack_1grow (&action_obstack, c);
+ lineno++;
+ break;
+
+ case '{':
+ obstack_1grow (&action_obstack, c);
+ count++;
+ break;
+
+ case '\'':
+ case '"':
+ copy_string (finput, &action_obstack, c);
+ break;
+
+ 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);
+ }
+
+ c = getc (finput);
+ }
+
+ /* above loop exits when c is '}' */
+
+ if (--count)
+ {
+ obstack_1grow (&action_obstack, c);
+ c = getc (finput);
+ }
+ }
+
+ obstack_sgrow (&action_obstack, ";\n break;}");
+}
+\f
+/*-------------------------------------------------------------------.
+| 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)
{
- register int c;
- register int n;
- register int count;
- register char *type_name;
+ int c;
+ int count;
int brace_flag = 0;
/* offset is always 0 if parser has already popped the stack pointer */
- if (semantic_parser) stack_offset = 0;
+ if (semantic_parser)
+ stack_offset = 0;
- fprintf(fguard, "\ncase %d:\n", nrules);
- if (!nolinesflag)
- fprintf (fguard, "#line %d \"%s\"\n", lineno, infile);
- putc('{', fguard);
+ obstack_fgrow1 (&guard_obstack, "\ncase %d:\n", nrules);
+ if (!no_lines_flag)
+ obstack_fgrow2 (&guard_obstack, muscle_find ("linef"),
+ lineno, quotearg_style (c_quoting_style,
+ muscle_find ("filename")));
+ obstack_1grow (&guard_obstack, '{');
count = 0;
- c = getc(finput);
+ c = getc (finput);
while (brace_flag ? (count > 0) : (c != ';'))
{
switch (c)
{
case '\n':
- putc(c, fguard);
+ obstack_1grow (&guard_obstack, c);
lineno++;
break;
case '{':
- putc(c, fguard);
+ obstack_1grow (&guard_obstack, c);
brace_flag = 1;
count++;
break;
case '}':
- putc(c, fguard);
+ obstack_1grow (&guard_obstack, c);
if (count > 0)
count--;
else
{
- warn(_("unmatched right brace (`}')"));
- c = getc(finput); /* skip it */
+ complain (_("unmatched %s"), "`}'");
+ c = getc (finput); /* skip it */
}
- break;
+ break;
case '\'':
case '"':
- copy_string (finput, fguard, c);
+ copy_string (finput, &guard_obstack, c);
break;
case '/':
- putc (c, fguard);
- c = getc (finput);
- if (c != '*' && c != '/')
- continue;
- copy_comment (finput, fguard, c);
+ copy_comment (finput, &guard_obstack);
break;
case '$':
- c = getc(finput);
- type_name = NULL;
-
- if (c == '<')
- {
- register 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)
- warns(_("$$ 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)
- warnss (_("$%s of `%s' has no declared type"),
- int_to_string(n), rule->sym->tag);
- continue;
- }
- else
- warns(_("$%s is invalid"), printable_version(c));
+ copy_dollar (finput, &guard_obstack, rule, stack_offset);
break;
case '@':
- c = getc (finput);
- if (c == '$')
- {
- fprintf (fguard, "yyloc");
- yylsp_needed = 1;
- }
- else if (isdigit(c) || c == '-')
- {
- ungetc (c, finput);
- n = read_signed_integer (finput);
- c = getc (finput);
- fprintf (fguard, "yylsp[%d]", n - stack_offset);
- yylsp_needed = 1;
- continue;
- }
- else
- {
- warns (_("@%s is invalid"), printable_version (c));
- n = 1;
- }
+ copy_at (finput, &guard_obstack, stack_offset);
break;
case EOF:
- fatal (_("unterminated %%guard clause"));
+ fatal ("%s", _("unterminated %guard clause"));
default:
- putc (c, fguard);
+ obstack_1grow (&guard_obstack, c);
}
if (c != '}' || count != 0)
- c = getc(finput);
+ c = getc (finput);
}
- c = skip_white_space();
+ c = skip_white_space ();
- fprintf(fguard, ";\n break;}");
+ obstack_sgrow (&guard_obstack, ";\n break;}");
if (c == '{')
- copy_action(rule, stack_offset);
+ copy_action (rule, stack_offset);
else if (c == '=')
{
- c = getc(finput); /* why not skip_white_space -wjh */
+ c = getc (finput); /* why not skip_white_space -wjh */
if (c == '{')
- copy_action(rule, stack_offset);
+ copy_action (rule, stack_offset);
}
else
- ungetc(c, finput);
+ ungetc (c, finput);
}
+\f
-
-
-/* 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 `$0'
- with respect to the top of the stack. */
-
-void
-copy_action (symbol_list *rule, int stack_offset)
+static void
+record_rule_line (void)
{
- register int c;
- register int n;
- register int count;
- register char *type_name;
-
- /* offset is always 0 if parser has already popped the stack pointer */
- if (semantic_parser)
- stack_offset = 0;
-
- fprintf (faction, "\ncase %d:\n", nrules);
- if (!nolinesflag)
- fprintf (faction, "#line %d \"%s\"\n", lineno, infile);
- putc ('{', faction);
-
- count = 1;
- c = getc(finput);
+ /* Record each rule's source line number in rline table. */
- while (count > 0)
+ if (nrules >= rline_allocated)
{
- while (c != '}')
- {
- switch (c)
- {
- case '\n':
- putc(c, faction);
- lineno++;
- break;
+ rline_allocated = nrules * 2;
+ rline = XREALLOC (rline, short, rline_allocated);
+ }
+ rline[nrules] = lineno;
+}
- case '{':
- putc(c, faction);
- count++;
- break;
- case '\'':
- case '"':
- copy_string (finput, faction, c);
- break;
+/*-------------------------------------------------------------------.
+| Generate a dummy symbol, a nonterminal, whose name cannot conflict |
+| with the user's names. |
+`-------------------------------------------------------------------*/
- case '/':
- putc(c, faction);
- c = getc(finput);
- if (c != '*' && c != '/')
- continue;
- copy_comment (finput, faction, c);
- break;
+static bucket *
+gensym (void)
+{
+ /* Incremented for each generated symbol */
+ static int gensym_count = 0;
+ static char buf[256];
- case '$':
- c = getc(finput);
- type_name = NULL;
+ bucket *sym;
- if (c == '<')
- {
- register char *cp = token_buffer;
+ sprintf (buf, "@%d", ++gensym_count);
+ token_buffer = buf;
+ sym = getsym (token_buffer);
+ sym->class = nterm_sym;
+ sym->value = nvars++;
+ return sym;
+}
- while ((c = getc(finput)) != '>' && c > 0)
- {
- if (cp == token_buffer + maxtoken)
- cp = grow_token_buffer(cp);
+#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 |
+`------------------------------------------------------------------*/
- *cp++ = c;
- }
- *cp = 0;
- type_name = token_buffer;
- value_components_used = 1;
+static int
+get_type (void)
+{
+ int k;
+ token_t token;
+ char *name;
- 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)
- warns(_("$$ 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)
- warnss(_("$%s of `%s' has no declared type"),
- int_to_string(n), rule->sym->tag);
- continue;
- }
- else
- warns(_("$%s is invalid"), printable_version(c));
+ token = lex ();
- break;
+ if (token != tok_typename)
+ {
+ complain (_("invalid %s declaration"), "%type");
+ return t;
+ }
- case '@':
- c = getc (finput);
- if (c == '$')
- {
- fprintf (faction, "yyloc");
- yylsp_needed = 1;
- }
- else if (isdigit(c) || c == '-')
- {
- ungetc (c, finput);
- n = read_signed_integer (finput);
- c = getc (finput);
- fprintf (faction, "yylsp[%d]", n - stack_offset);
- yylsp_needed = 1;
- continue;
- }
- else
- {
- warns (_("@%s is invalid"), printable_version (c));
- n = 1;
- }
- break;
+ name = xstrdup (token_buffer);
- case EOF:
- fatal(_("unmatched `{'"));
+ for (;;)
+ {
+ token = lex ();
- default:
- putc(c, faction);
- }
+ switch (token)
+ {
+ case tok_semicolon:
+ return lex ();
- c = getc(finput);
- }
+ case tok_comma:
+ break;
- /* above loop exits when c is '}' */
+ case tok_identifier:
+ if (symval->type_name == NULL)
+ symval->type_name = name;
+ else if (strcmp (name, symval->type_name) != 0)
+ complain (_("type redeclaration for %s"), symval->tag);
- if (--count)
- {
- putc(c, faction);
- c = getc(finput);
+ break;
+
+ default:
+ return token;
}
}
-
- fprintf(faction, ";\n break;}");
-}
-
-
-
-/* generate a dummy symbol, a nonterminal,
-whose name cannot conflict with the user's names. */
-
-bucket *
-gensym (void)
-{
- register bucket *sym;
-
- sprintf (token_buffer, "@%d", ++gensym_count);
- sym = getsym(token_buffer);
- sym->class = SNTERM;
- sym->value = 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. */
+#endif
+\f
+/*------------------------------------------------------------------.
+| 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. |
+`------------------------------------------------------------------*/
-void
+static void
readgram (void)
{
- register int t;
- register bucket *lhs = NULL;
- register symbol_list *p;
- register symbol_list *p1;
- register bucket *bp;
+ token_t t;
+ bucket *lhs = NULL;
+ symbol_list *p;
+ symbol_list *p1;
+ 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. */
+ /* Points to first symbol_list of current rule. its symbol is the
+ lhs of the rule. */
+ symbol_list *crule;
+ /* Points to the symbol_list preceding crule. */
+ symbol_list *crule1;
p1 = NULL;
- t = lex();
+ t = lex ();
- while (t != TWO_PERCENTS && t != ENDFILE)
+ while (t != tok_two_percents && t != tok_eof)
{
- if (t == IDENTIFIER || t == BAR)
+ if (t == tok_identifier || t == tok_bar)
{
- register int actionflag = 0;
- int rulelength = 0; /* number of symbols in rhs of this rule so far */
+ 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)
+ if (t == tok_identifier)
{
lhs = symval;
start_flag = 1;
}
- t = lex();
- if (t != COLON)
+ t = lex ();
+ if (t != tok_colon)
{
- warn(_("ill-formed rule: initial symbol not followed by colon"));
- unlex(t);
+ complain (_("ill-formed rule: initial symbol not followed by colon"));
+ unlex (t);
}
}
- if (nrules == 0 && t == BAR)
+ if (nrules == 0 && t == tok_bar)
{
- warn(_("grammar starts with vertical bar"));
+ complain (_("grammar starts with vertical bar"));
lhs = symval; /* BOGUS: use a random symval */
}
/* start a new rule and record its lhs. */
record_rule_line ();
- p = NEW(symbol_list);
+ p = XCALLOC (symbol_list, 1);
p->sym = lhs;
crule1 = p1;
/* mark the rule's lhs as a nonterminal if not already so. */
- if (lhs->class == SUNKNOWN)
+ if (lhs->class == unknown_sym)
{
- lhs->class = SNTERM;
+ lhs->class = nterm_sym;
lhs->value = nvars;
nvars++;
}
- else if (lhs->class == STOKEN)
- warns(_("rule given for %s, which is a token"), lhs->tag);
+ else if (lhs->class == token_sym)
+ complain (_("rule given for %s, which is a token"), lhs->tag);
/* read the rhs of the rule. */
for (;;)
{
- t = lex();
- if (t == PREC)
+ t = lex ();
+ if (t == tok_prec)
{
- t = lex();
+ t = lex ();
crule->ruleprec = symval;
- t = lex();
+ t = lex ();
}
- if (! (t == IDENTIFIER || t == LEFT_CURLY)) break;
+ if (!(t == tok_identifier || t == tok_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)
+ If one does, exit this rule now. */
+ if (t == tok_identifier)
{
- register bucket *ssave;
- register int t1;
+ bucket *ssave;
+ token_t t1;
ssave = symval;
- t1 = lex();
- unlex(t1);
+ t1 = lex ();
+ unlex (t1);
symval = ssave;
- if (t1 == COLON) break;
+ if (t1 == tok_colon)
+ break;
- if(!first_rhs) /* JF */
+ 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)
+ of a rule, so make a dummy rule to reduce it to a
+ non-terminal. */
+ if (action_flag)
{
- register bucket *sdummy;
+ 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. */
+ /* 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();
+ sdummy = gensym ();
/* Make a new rule, whose body is empty,
before the current one, so that the action
nrules++;
nitems++;
record_rule_line ();
- p = NEW(symbol_list);
+ p = XCALLOC (symbol_list, 1);
if (crule1)
crule1->next = p;
- else grammar = p;
+ else
+ grammar = p;
p->sym = sdummy;
- crule1 = NEW(symbol_list);
+ crule1 = XCALLOC (symbol_list, 1);
p->next = crule1;
crule1->next = crule;
- /* insert the dummy generated by that rule into this rule. */
+ /* Insert the dummy generated by that rule into this
+ rule. */
nitems++;
- p = NEW(symbol_list);
+ p = XCALLOC (symbol_list, 1);
p->sym = sdummy;
p1->next = p;
p1 = p;
- actionflag = 0;
+ action_flag = 0;
}
- if (t == IDENTIFIER)
+ if (t == tok_identifier)
{
nitems++;
- p = NEW(symbol_list);
+ p = XCALLOC (symbol_list, 1);
p->sym = symval;
p1->next = p;
p1 = p;
}
- else /* handle an action. */
+ else /* handle an action. */
{
- copy_action(crule, rulelength);
- actionflag = 1;
+ copy_action (crule, rulelength);
+ action_flag = 1;
xactions++; /* JF */
}
rulelength++;
- } /* end of read rhs of rule */
+ } /* end of read rhs of rule */
/* Put an empty link in the list to mark the end of this rule */
- p = NEW(symbol_list);
+ p = XCALLOC (symbol_list, 1);
p1->next = p;
p1 = p;
- if (t == PREC)
+ if (t == tok_prec)
{
- warn(_("two @prec's in a row"));
- t = lex();
+ complain (_("two @prec's in a row"));
+ t = lex ();
crule->ruleprec = symval;
- t = lex();
+ t = lex ();
}
- if (t == GUARD)
+ if (t == tok_guard)
{
- if (! semantic_parser)
- warn(_("%%guard present but %%semantic_parser not specified"));
+ if (!semantic_parser)
+ complain (_("%%guard present but %%semantic_parser not specified"));
- copy_guard(crule, rulelength);
- t = lex();
+ copy_guard (crule, rulelength);
+ t = lex ();
}
- else if (t == LEFT_CURLY)
+ else if (t == tok_left_curly)
{
- /* This case never occurs -wjh */
- if (actionflag)
- warn(_("two actions at end of one rule"));
- copy_action(crule, rulelength);
- actionflag = 1;
+ /* 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();
+ t = lex ();
}
- /* If $$ is being set in default way, warn if any type
+ /* If $$ is being set in default way, report if any type
mismatch. */
else if (!xactions
- && first_rhs
- && lhs->type_name != first_rhs->type_name)
+ && first_rhs && lhs->type_name != first_rhs->type_name)
{
if (lhs->type_name == 0
|| first_rhs->type_name == 0
- || strcmp(lhs->type_name,first_rhs->type_name))
- warnss(_("type clash (`%s' `%s') on default action"),
- lhs->type_name ? lhs->type_name : "",
- first_rhs->type_name ? first_rhs->type_name : "");
+ || strcmp (lhs->type_name, first_rhs->type_name))
+ complain (_("type clash (`%s' `%s') on default action"),
+ lhs->type_name ? lhs->type_name : "",
+ first_rhs->type_name ? first_rhs->type_name : "");
}
/* Warn if there is no default for $$ but we need one. */
else if (!xactions && !first_rhs && lhs->type_name != 0)
- warn(_("empty rule for typed nonterminal, and no action"));
- if (t == SEMICOLON)
- t = lex();
+ complain (_("empty rule for typed nonterminal, and no action"));
+ if (t == tok_semicolon)
+ t = lex ();
}
#if 0
- /* these things can appear as alternatives to rules. */
+ /* 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)
+ else if (t == tok_token)
{
- parse_token_decl(STOKEN, SNTERM);
- t = lex();
+ parse_token_decl (token_sym, nterm_sym);
+ t = lex ();
}
- else if (t == NTERM)
+ else if (t == tok_nterm)
{
- parse_token_decl(SNTERM, STOKEN);
- t = lex();
+ parse_token_decl (nterm_sym, token_sym);
+ t = lex ();
}
- else if (t == TYPE)
+ else if (t == tok_type)
{
- t = get_type();
+ t = get_type ();
}
- else if (t == UNION)
+ else if (t == tok_union)
{
- parse_union_decl();
- t = lex();
+ parse_union_decl ();
+ t = lex ();
}
- else if (t == EXPECT)
+ else if (t == tok_expect)
{
- parse_expect_decl();
- t = lex();
+ parse_expect_decl ();
+ t = lex ();
}
- else if (t == START)
+ else if (t == tok_start)
{
- parse_start_decl();
- t = lex();
+ parse_start_decl ();
+ t = lex ();
}
#endif
else
{
- warns(_("invalid input: %s"), token_buffer);
- t = lex();
+ complain (_("invalid input: %s"), quote (token_buffer));
+ t = lex ();
}
}
/* grammar has been read. Do some checking */
if (nsyms > MAXSHORT)
- fatals(_("too many symbols (tokens plus nonterminals); maximum %s"),
- int_to_string(MAXSHORT));
+ fatal (_("too many symbols (tokens plus nonterminals); maximum %d"),
+ MAXSHORT);
if (nrules == 0)
- fatal(_("no rules in the 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");
- }
+ fatal (_("no rules in the input grammar"));
/* Report any undefined symbols and consider them nonterminals. */
for (bp = firstsymbol; bp; bp = bp->next)
- if (bp->class == SUNKNOWN)
+ if (bp->class == unknown_sym)
{
- warns(_("symbol %s is used, but is not defined as a token and has no rules"),
- bp->tag);
- bp->class = SNTERM;
+ complain (_
+ ("symbol %s is used, but is not defined as a token and has no rules"),
+ bp->tag);
+ bp->class = nterm_sym;
bp->value = nvars++;
}
ntokens = nsyms - nvars;
}
-
-void
-record_rule_line (void)
+/* 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)
{
- /* Record each rule's source line number in rline table. */
+ char c;
+ struct obstack el_obstack;
- if (nrules >= rline_allocated)
- {
- rline_allocated = nrules * 2;
- rline = (short *) xrealloc ((char *) rline,
- rline_allocated * sizeof (short));
- }
- rline[nrules] = lineno;
+ obstack_init (&el_obstack);
+
+ while ((c = getc (finput)) != EOF)
+ obstack_1grow (&el_obstack, c);
+
+ obstack_1grow (&el_obstack, 0);
+ muscle_insert ("epilogue", obstack_finish (&el_obstack));
}
+\f
+/*--------------------------------------------------------------.
+| For named tokens, but not literal ones, define the name. The |
+| value is the user token number. |
+`--------------------------------------------------------------*/
-#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)
+static void
+output_token_defines (struct obstack *oout)
{
- register int k;
- register int t;
- register char *name;
-
- t = lex();
+ bucket *bp;
+ char *cp, *symbol;
+ char c;
- if (t != TYPENAME)
+ for (bp = firstsymbol; bp; bp = bp->next)
{
- warn(_("ill-formed %type declaration"));
- return t;
- }
+ symbol = bp->tag; /* get symbol */
+
+ 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)
+ {
+ /* use literal string only if given a symbol with an alias */
+ if (bp->alias)
+ symbol = bp->alias->tag;
+ else
+ continue;
+ }
- k = strlen(token_buffer);
- name = NEW2(k + 1, char);
- strcpy(name, token_buffer);
+ /* Don't #define nonliteral tokens whose names contain periods. */
+ cp = symbol;
+ while ((c = *cp++) && c != '.');
+ if (c != '\0')
+ continue;
- for (;;)
- {
- t = lex();
+ obstack_fgrow2 (oout, "# define\t%s\t%d\n",
+ symbol, bp->user_token_number);
+ if (semantic_parser)
+ /* FIXME: This is certainly dead wrong, and should be just as
+ above. --akim. */
+ obstack_fgrow2 (oout, "# define\tT%s\t%d\n", symbol, bp->value);
+ }
+}
- switch (t)
- {
- case SEMICOLON:
- return (lex());
- case COMMA:
- break;
+/*------------------------------------------------------------------.
+| Set TOKEN_TRANSLATIONS. Check that no two symbols share the same |
+| number. |
+`------------------------------------------------------------------*/
- case IDENTIFIER:
- if (symval->type_name == NULL)
- symval->type_name = name;
- else if (strcmp(name, symval->type_name) != 0)
- warns(_("type redeclaration for %s"), symval->tag);
+static void
+token_translations_init (void)
+{
+ bucket *bp = NULL;
+ int i;
- break;
+ token_translations = XCALLOC (short, max_user_token_number + 1);
- default:
- return (t);
- }
+ /* Initialize all entries for literal tokens to 2, the internal
+ token number for $undefined., which represents all invalid
+ inputs. */
+ for (i = 0; i <= max_user_token_number; i++)
+ token_translations[i] = 2;
+
+ for (bp = firstsymbol; bp; bp = bp->next)
+ {
+ /* Non-terminal? */
+ if (bp->value >= ntokens)
+ continue;
+ /* A token string alias? */
+ if (bp->user_token_number == SALIAS)
+ continue;
+
+ assert (bp->user_token_number != SUNDEF);
+
+ /* A token which translation has already been set? */
+ if (token_translations[bp->user_token_number] != 2)
+ complain (_("tokens %s and %s both assigned number %d"),
+ tags[token_translations[bp->user_token_number]],
+ bp->tag, bp->user_token_number);
+ token_translations[bp->user_token_number] = bp->value;
}
}
-#endif
-/* Assign symbol numbers, and write definition of token names into
- fdefines. Set up vectors tags and sprec of names and precedences
- of symbols. */
+/*------------------------------------------------------------------.
+| Assign symbol numbers, and write definition of token names into |
+| FDEFINES. Set up vectors TAGS and SPREC of names and precedences |
+| of symbols. |
+`------------------------------------------------------------------*/
-void
+static void
packsymbols (void)
{
- register bucket *bp;
- register int tokno = 1;
- register int i;
- register int last_user_token_number;
+ bucket *bp = NULL;
+ int tokno = 1;
+ int last_user_token_number;
+ static char DOLLAR[] = "$";
- /* int lossage = 0; JF set but not used */
+ tags = XCALLOC (char *, nsyms + 1);
+ user_toknums = XCALLOC (short, nsyms + 1);
- tags = NEW2(nsyms + 1, char *);
- tags[0] = "$";
- user_toknums = NEW2(nsyms + 1, int);
- user_toknums[0] = 0;
+ sprec = XCALLOC (short, nsyms);
+ sassoc = XCALLOC (short, nsyms);
- sprec = NEW2(nsyms, short);
- sassoc = NEW2(nsyms, short);
+ /* The EOF token. */
+ tags[0] = DOLLAR;
+ user_toknums[0] = 0;
max_user_token_number = 256;
last_user_token_number = 256;
for (bp = firstsymbol; bp; bp = bp->next)
{
- if (bp->class == SNTERM)
+ if (bp->class == nterm_sym)
{
bp->value += ntokens;
}
{
if (bp->prec != 0 && bp->alias->prec != 0
&& bp->user_token_number == SALIAS)
- warnss(_("conflicting precedences for %s and %s"),
- bp->tag, bp->alias->tag);
+ complain (_("conflicting precedences for %s and %s"),
+ bp->tag, bp->alias->tag);
if (bp->prec != 0)
bp->alias->prec = bp->prec;
else
if (bp->assoc != bp->alias->assoc)
{
- if (bp->assoc != 0 && bp->alias->assoc != 0
- && bp->user_token_number == SALIAS)
- warnss(_("conflicting assoc values for %s and %s"),
- bp->tag, bp->alias->tag);
- if (bp->assoc != 0)
- bp->alias->assoc = bp->assoc;
- else
- bp->assoc = bp->alias->assoc;
- }
+ if (bp->assoc != 0 && bp->alias->assoc != 0
+ && bp->user_token_number == SALIAS)
+ complain (_("conflicting assoc values for %s and %s"),
+ bp->tag, bp->alias->tag);
+ if (bp->assoc != 0)
+ bp->alias->assoc = bp->assoc;
+ else
+ bp->assoc = bp->alias->assoc;
+ }
if (bp->user_token_number == SALIAS)
- continue; /* do not do processing below for SALIASs */
+ continue; /* do not do processing below for SALIASs */
- }
- else /* bp->class == STOKEN */
+ }
+ else /* bp->class == token_sym */
{
bp->value = tokno++;
}
- if (bp->class == STOKEN)
+ 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;
user_toknums[bp->value] = bp->user_token_number;
sprec[bp->value] = bp->prec;
sassoc[bp->value] = bp->assoc;
-
}
- if (translations)
- {
- register int i;
+ token_translations_init ();
- token_translations = NEW2(max_user_token_number+1, short);
+ error_token_number = errtoken->value;
- /* initialize all entries for literal tokens to 2, the internal
- token number for $undefined., which represents all invalid
- inputs. */
- for (i = 0; i <= max_user_token_number; i++)
- token_translations[i] = 2;
+ if (startval->class == unknown_sym)
+ fatal (_("the start symbol %s is undefined"), startval->tag);
+ else if (startval->class == token_sym)
+ fatal (_("the start symbol %s is a token"), startval->tag);
- for (bp = firstsymbol; bp; bp = bp->next)
- {
- if (bp->value >= ntokens)
- continue; /* non-terminal */
- if (bp->user_token_number == SALIAS)
- continue;
- if (token_translations[bp->user_token_number] != 2)
- warnsss(_("tokens %s and %s both assigned number %s"),
- tags[token_translations[bp->user_token_number]],
- bp->tag,
- int_to_string(bp->user_token_number));
- token_translations[bp->user_token_number] = bp->value;
- }
- }
+ start_symbol = startval->value;
+}
- error_token_number = errtoken->value;
- if (! noparserflag)
- output_token_defines(ftable);
+/*-----------------------------------.
+| Output definition of token names. |
+`-----------------------------------*/
- 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);
+static void
+symbols_output (void)
+{
+ {
+ struct obstack tokendefs;
+ obstack_init (&tokendefs);
+ output_token_defines (&tokendefs);
+ obstack_1grow (&tokendefs, 0);
+ muscle_insert ("tokendef", xstrdup (obstack_finish (&tokendefs)));
+ obstack_free (&tokendefs, NULL);
+ }
- start_symbol = startval->value;
+#if 0
+ if (!no_parser_flag)
+ output_token_defines (&table_obstack);
+#endif
- if (definesflag)
+ if (defines_flag)
{
- output_token_defines(fdefines);
+ output_token_defines (&defines_obstack);
if (!pure_parser)
{
if (spec_name_prefix)
- fprintf(fdefines, "\nextern YYSTYPE %slval;\n", spec_name_prefix);
+ obstack_fgrow1 (&defines_obstack, "\nextern YYSTYPE %slval;\n",
+ spec_name_prefix);
else
- fprintf(fdefines, "\nextern YYSTYPE yylval;\n");
+ obstack_sgrow (&defines_obstack,
+ "\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);
- }
+ {
+ int i;
+
+ for (i = ntokens; i < nsyms; i++)
+ {
+ /* don't make these for dummy nonterminals made by gensym. */
+ if (*tags[i] != '@')
+ obstack_fgrow2 (&defines_obstack,
+ "# define\tNT%s\t%d\n", tags[i], i);
+ }
#if 0
- /* `fdefines' is now a temporary file, so we need to copy its
- contents in `done', so we can't close it here. */
- fclose(fdefines);
- fdefines = NULL;
+ /* `fdefines' is now a temporary file, so we need to copy its
+ contents in `done', so we can't close it here. */
+ fclose (fdefines);
+ fdefines = NULL;
#endif
- }
-}
-
-/* For named tokens, but not literal ones, define the name. The value
- is the user token number. */
-
-void
-output_token_defines (FILE *file)
-{
- bucket *bp;
- register char *cp, *symbol;
- register char c;
-
- for (bp = firstsymbol; bp; bp = bp->next)
- {
- symbol = bp->tag; /* get symbol */
-
- 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)
- {
- /* 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;
-
- fprintf(file, "#define\t%s\t%d\n", symbol,
- ((translations && ! rawtoknumflag)
- ? bp->user_token_number
- : bp->value));
- if (semantic_parser)
- fprintf(file, "#define\tT%s\t%d\n", symbol, bp->value);
}
-
- putc('\n', file);
}
+/*---------------------------------------------------------------.
+| Convert the rules into the representation using RRHS, RLHS and |
+| RITEMS. |
+`---------------------------------------------------------------*/
-/* convert the rules into the representation using rrhs, rlhs and ritems. */
-
-void
+static void
packgram (void)
{
- register int itemno;
- register int ruleno;
- register symbol_list *p;
-/* register bucket *bp; JF unused */
+ int itemno;
+ int ruleno;
+ symbol_list *p;
bucket *ruleprec;
- 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 (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;
itemno = 0;
ruleno = 1;
ritem[itemno++] = p->sym->value;
/* A rule gets by default the precedence and associativity
of the last token in it. */
- if (p->sym->class == STOKEN)
+ if (p->sym->class == token_sym)
{
rprec[ruleno] = p->sym->prec;
rassoc[ruleno] = p->sym->assoc;
}
- if (p) p = p->next;
+ 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;
+ rprec[ruleno] = ruleprec->prec;
+ rassoc[ruleno] = ruleprec->assoc;
rprecsym[ruleno] = ruleprec->value;
}
ritem[itemno++] = -ruleno;
ruleno++;
- if (p) p = p->next;
+ if (p)
+ p = p->next;
}
ritem[itemno] = 0;
}
\f
-/* 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 (FILE *stream)
+void
+reader (void)
{
- register int c = getc(stream);
- register int sign = 1;
- register int n;
+ start_flag = 0;
+ startval = NULL; /* start symbol not specified yet. */
- if (c == '-')
- {
- c = getc(stream);
- sign = -1;
- }
- n = 0;
- while (isdigit(c))
- {
- n = 10*n + (c - '0');
- c = getc(stream);
- }
+ nsyms = 1;
+ nvars = 0;
+ nrules = 0;
+ nitems = 0;
+ rline_allocated = 10;
+ rline = XCALLOC (short, rline_allocated);
+
+ typed = 0;
+ lastprec = 0;
+
+ semantic_parser = 0;
+ pure_parser = 0;
- ungetc(c, stream);
+ grammar = NULL;
+
+ lex_init ();
+ lineno = 1;
+
+ /* Initialize the muscle obstack. */
+ obstack_init (&muscle_obstack);
+
+ /* Initialize the symbol table. */
+ tabinit ();
+
+ /* 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;
- return n * sign;
+ /* Read the declaration section. Copy %{ ... %} groups to
+ TABLE_OBSTACK and FDEFINES file. Also notice any %token, %left,
+ etc. found there. */
+ read_declarations ();
+ /* Read in the grammar, build grammar in list form. Write out
+ guards and actions. */
+ readgram ();
+ /* Some C code is given at the end of the grammar file. */
+ read_additionnal_code ();
+
+ /* Now we know whether we need the line-number stack. If we do,
+ write its type into the .tab.h file.
+ This is no longer need with header skeleton. */
+
+ /* Assign the symbols their symbol numbers. Write #defines for the
+ token symbols into FDEFINES if requested. */
+ packsymbols ();
+ symbols_output ();
+ /* Convert the grammar into the format described in gram.h. */
+ packgram ();
}
projects / bison.git / blobdiff