/* Input parser for bison
- Copyright 1984, 1986, 1989, 1992, 1998, 2000
+ Copyright 1984, 1986, 1989, 1992, 1998, 2000, 2001
Free Software Foundation, Inc.
This file is part of Bison, the GNU Compiler Compiler.
#include "quote.h"
#include "getargs.h"
#include "files.h"
-#include "xalloc.h"
#include "symtab.h"
#include "options.h"
#include "lex.h"
#include "conflicts.h"
#include "muscle_tab.h"
-/* Number of slots allocated (but not necessarily used yet) in `rline' */
-static int rline_allocated;
-
typedef struct symbol_list
{
struct symbol_list *next;
bucket *sym;
+ int line;
bucket *ruleprec;
}
symbol_list;
static bucket *errtoken;
static bucket *undeftoken;
+
+
+static symbol_list *
+symbol_list_new (bucket *sym)
+{
+ symbol_list *res = XMALLOC (symbol_list, 1);
+ res->next = NULL;
+ res->sym = sym;
+ res->line = lineno;
+ res->ruleprec = NULL;
+ return res;
+}
+
\f
/*===================\
`--------------------------------------------------------------*/
static char *
-get_type_name (int n, symbol_list * rule)
+get_type_name (int n, symbol_list *rule)
{
int i;
symbol_list *rp;
/* -1 while reading a character if prev char was %. */
int after_percent;
-#if 0
if (!no_lines_flag)
{
obstack_fgrow2 (&attrs_obstack, muscle_find ("linef"),
- lineno, quotearg_style (c_quoting_style,
+ lineno, quotearg_style (c_quoting_style,
muscle_find("filename")));
}
-#endif
after_percent = 0;
static void
parse_token_decl (symbol_class what_is, symbol_class what_is_not)
{
- token_t token = 0;
- char *typename = 0;
+ token_t token = tok_undef;
+ char *typename = NULL;
/* The symbol being defined. */
struct bucket *symbol = NULL;
/* symbol and symval combined are only one symbol */
nsyms--;
}
- translations = 1;
symbol = NULL;
}
else if (token == tok_identifier)
else if (symbol && token == tok_number)
{
symbol->user_token_number = numval;
- translations = 1;
}
else
{
complain (_("`%s' is invalid in %s"),
- token_buffer, (what_is == token_sym) ? "%token" : "%nterm");
+ token_buffer,
+ (what_is == token_sym) ? "%token" : "%nterm");
skip_to_char ('%');
}
}
if (prev == tok_identifier)
{
symval->user_token_number = numval;
- translations = 1;
}
else
{
}
prev = t;
-
}
}
/*--------------------------------------------------------------.
-| Copy the union declaration into ATTRS_OBSTACK (and fdefines), |
-| where it is made into the definition of YYSTYPE, the type of |
-| elements of the parser value stack. |
+| Copy the union declaration into the stype muscle |
+| (and fdefines), where it is made into the definition of |
+| YYSTYPE, the type of elements of the parser value stack. |
`--------------------------------------------------------------*/
static void
{
int c;
int count = 0;
+ struct obstack union_obstack;
+ const char *prologue = "\
+#ifndef YYSTYPE\n\
+typedef union";
+ const char *epilogue = "\
+ yystype;\n\
+# define YYSTYPE yystype\n\
+#endif\n";
if (typed)
complain (_("multiple %s declarations"), "%union");
typed = 1;
-
- if (!no_lines_flag)
- {
- obstack_fgrow2 (&attrs_obstack, muscle_find ("linef"),
- lineno, quotearg_style (c_quoting_style,
- muscle_find("filename")));
- }
- else
+ /* FIXME: I'm worried: are you sure attrs_obstack is properly
+ filled? */
+ /* I don't see any reasons to keep this line, because we should
+ create a special skeleton for this option. */
+ if (no_lines_flag)
obstack_1grow (&attrs_obstack, '\n');
-
- obstack_sgrow (&attrs_obstack, "typedef union");
+
+ obstack_init (&union_obstack);
+ obstack_sgrow (&union_obstack, "union");
if (defines_flag)
- obstack_sgrow (&defines_obstack, "typedef union");
+ obstack_sgrow (&defines_obstack, prologue);
c = getc (finput);
while (c != EOF)
{
- obstack_1grow (&attrs_obstack, c);
- if (defines_flag)
- obstack_1grow (&defines_obstack, c);
+ /* If C contains '/', it is output by copy_comment (). */
+ if (c != '/')
+ {
+ obstack_1grow (&union_obstack, c);
+ if (defines_flag)
+ obstack_1grow (&defines_obstack, c);
+ }
switch (c)
{
break;
case '/':
- copy_comment2 (finput, &defines_obstack, &attrs_obstack);
+ copy_comment2 (finput, &defines_obstack, &union_obstack);
break;
case '{':
count--;
if (count <= 0)
{
- obstack_sgrow (&attrs_obstack, " YYSTYPE;\n");
if (defines_flag)
- obstack_sgrow (&defines_obstack, " YYSTYPE;\n");
+ obstack_sgrow (&defines_obstack, epilogue);
/* JF don't choke on trailing semi */
c = skip_white_space ();
if (c != ';')
ungetc (c, finput);
+ obstack_1grow (&union_obstack, 0);
+ muscle_insert ("stype", obstack_finish (&union_obstack));
return;
}
}
c = getc (finput);
}
+
}
token_t token;
struct bucket *symbol;
char *typename = 0;
- int usrtoknum;
+ int usrtoknum = SUNDEF;
- translations = 1;
token = lex (); /* fetch typename or first token */
if (token == tok_typename)
{
usrtoknum = numval;
token = lex (); /* okay, did number, now get literal */
}
- else
- usrtoknum = 0;
/* process literal string token */
nsyms--;
}
-/* FIXME. */
-
static void
parse_muscle_decl (void)
{
copy_identifier (finput, &muscle_obstack);
obstack_1grow (&muscle_obstack, 0);
muscle_key = obstack_finish (&muscle_obstack);
-
+
/* Read value. */
ch = skip_white_space ();
if (ch != '"')
}
-/*----------------------------------.
-| Parse what comes after %skeleton. |
-`----------------------------------*/
-void
-parse_skel_decl (void)
+/*---------------------------------.
+| Parse a double quoted parameter. |
+`---------------------------------*/
+
+static const char *
+parse_dquoted_param (const char *from)
{
- /* Complete with parse_dquoted_param () on the CVS branch 1.29. */
-}
+ struct obstack param_obstack;
+ const char *param = NULL;
+ int c;
-/*------------------------------------------.
-| Parse what comes after %header_extension. |
-`------------------------------------------*/
+ obstack_init (¶m_obstack);
+ c = skip_white_space ();
-static void
-parse_header_extension_decl (void)
-{
- char buff[32];
+ if (c != '"')
+ {
+ complain (_("invalid %s declaration"), from);
+ ungetc (c, finput);
+ skip_to_char ('%');
+ return NULL;
+ }
+
+ for (;;)
+ {
+ if (literalchar (NULL, &c, '\"'))
+ obstack_1grow (¶m_obstack, c);
+ else
+ break;
+ }
- if (header_extension)
- complain (_("multiple %%header_extension declarations"));
- fscanf (finput, "%s", buff);
- header_extension = xstrdup (buff);
+ obstack_1grow (¶m_obstack, '\0');
+ param = obstack_finish (¶m_obstack);
+
+ if (c != '"' || strlen (param) == 0)
+ {
+ complain (_("invalid %s declaration"), from);
+ if (c != '"')
+ ungetc (c, finput);
+ skip_to_char ('%');
+ return NULL;
+ }
+
+ return param;
}
-/*------------------------------------------.
-| Parse what comes after %source_extension. |
-`------------------------------------------*/
+/*----------------------------------.
+| Parse what comes after %skeleton. |
+`----------------------------------*/
static void
-parse_source_extension_decl (void)
+parse_skel_decl (void)
{
- char buff[32];
-
- if (src_extension)
- complain (_("multiple %%source_extension declarations"));
- fscanf (finput, "%s", buff);
- src_extension = xstrdup (buff);
+ skeleton = parse_dquoted_param ("%skeleton");
}
/*----------------------------------------------------------------.
static void
read_declarations (void)
{
- int c;
- int tok;
-
for (;;)
{
- c = skip_white_space ();
+ int c = skip_white_space ();
if (c == '%')
{
- tok = parse_percent_token ();
+ token_t tok = parse_percent_token ();
switch (tok)
{
parse_assoc_decl (non_assoc);
break;
- case tok_hdrext:
- parse_header_extension_decl ();
- break;
-
- case tok_srcext:
- parse_source_extension_decl ();
- break;
-
case tok_define:
parse_muscle_decl ();
break;
-
+
case tok_skel:
parse_skel_decl ();
break;
case tok_noop:
break;
+ case tok_stropt:
+ case tok_intopt:
+ case tok_obsolete:
+ abort ();
+ break;
+
+ case tok_illegal:
default:
complain (_("unrecognized: %s"), token_buffer);
skip_to_char ('%');
{
int c;
int count;
- char buf[4096];
/* offset is always 0 if parser has already popped the stack pointer */
if (semantic_parser)
if (!no_lines_flag)
{
obstack_fgrow2 (&action_obstack, muscle_find ("linef"),
- lineno, quotearg_style (c_quoting_style,
+ lineno, quotearg_style (c_quoting_style,
muscle_find ("filename")));
}
obstack_1grow (&action_obstack, '{');
}
\f
-static void
-record_rule_line (void)
-{
- /* Record each rule's source line number in rline table. */
-
- if (nrules >= rline_allocated)
- {
- rline_allocated = nrules * 2;
- rline = XREALLOC (rline, short, rline_allocated);
- }
- rline[nrules] = lineno;
-}
-
-
/*-------------------------------------------------------------------.
| Generate a dummy symbol, a nonterminal, whose name cannot conflict |
| with the user's names. |
nrules++;
nitems++;
- record_rule_line ();
-
- p = XCALLOC (symbol_list, 1);
- p->sym = lhs;
+ p = symbol_list_new (lhs);
crule1 = p1;
if (p1)
non-terminal. */
if (action_flag)
{
- bucket *sdummy;
-
/* Since the action was written out with this rule's
number, we must give the new rule this number by
inserting the new rule before it. */
/* Make a dummy nonterminal, a gensym. */
- sdummy = gensym ();
+ bucket *sdummy = gensym ();
- /* Make a new rule, whose body is empty,
- before the current one, so that the action
- just read can belong to it. */
+ /* Make a new rule, whose body is empty, before the
+ current one, so that the action just read can
+ belong to it. */
nrules++;
nitems++;
- record_rule_line ();
- p = XCALLOC (symbol_list, 1);
+ p = symbol_list_new (sdummy);
+ /* Attach its lineno to that of the host rule. */
+ p->line = crule->line;
if (crule1)
crule1->next = p;
else
grammar = p;
- p->sym = sdummy;
- crule1 = XCALLOC (symbol_list, 1);
- p->next = crule1;
+ /* End of the rule. */
+ crule1 = symbol_list_new (NULL);
crule1->next = crule;
+ p->next = crule1;
+
/* Insert the dummy generated by that rule into this
rule. */
nitems++;
- p = XCALLOC (symbol_list, 1);
- p->sym = sdummy;
+ p = symbol_list_new (sdummy);
p1->next = p;
p1 = p;
if (t == tok_identifier)
{
nitems++;
- p = XCALLOC (symbol_list, 1);
- p->sym = symval;
+ p = symbol_list_new (symval);
p1->next = p;
p1 = p;
}
} /* end of read rhs of rule */
/* Put an empty link in the list to mark the end of this rule */
- p = XCALLOC (symbol_list, 1);
+ p = symbol_list_new (NULL);
p1->next = p;
p1 = p;
{
char c;
struct obstack el_obstack;
-
+
obstack_init (&el_obstack);
while ((c = getc (finput)) != EOF)
obstack_1grow (&el_obstack, c);
-
+
obstack_1grow (&el_obstack, 0);
muscle_insert ("epilogue", obstack_finish (&el_obstack));
}
continue;
obstack_fgrow2 (oout, "# define\t%s\t%d\n",
- symbol,
- (translations ? bp->user_token_number : bp->value));
+ symbol, bp->user_token_number);
if (semantic_parser)
+ /* FIXME: This is certainly dead wrong, and should be just as
+ above. --akim. */
obstack_fgrow2 (oout, "# define\tT%s\t%d\n", symbol, bp->value);
}
}
+/*------------------------------------------------------------------.
+| Set TOKEN_TRANSLATIONS. Check that no two symbols share the same |
+| number. |
+`------------------------------------------------------------------*/
+
+static void
+token_translations_init (void)
+{
+ bucket *bp = NULL;
+ int i;
+
+ token_translations = XCALLOC (short, max_user_token_number + 1);
+
+ /* Initialize all entries for literal tokens to 2, the internal
+ token number for $undefined., which represents all invalid
+ inputs. */
+ for (i = 0; i <= max_user_token_number; i++)
+ token_translations[i] = 2;
+
+ for (bp = firstsymbol; bp; bp = bp->next)
+ {
+ /* Non-terminal? */
+ if (bp->value >= ntokens)
+ continue;
+ /* A token string alias? */
+ if (bp->user_token_number == SALIAS)
+ continue;
+
+ assert (bp->user_token_number != SUNDEF);
+
+ /* A token which translation has already been set? */
+ if (token_translations[bp->user_token_number] != 2)
+ complain (_("tokens %s and %s both assigned number %d"),
+ tags[token_translations[bp->user_token_number]],
+ bp->tag, bp->user_token_number);
+ token_translations[bp->user_token_number] = bp->value;
+ }
+}
+
+
/*------------------------------------------------------------------.
| Assign symbol numbers, and write definition of token names into |
| FDEFINES. Set up vectors TAGS and SPREC of names and precedences |
static void
packsymbols (void)
{
- bucket *bp;
+ bucket *bp = NULL;
int tokno = 1;
- int i;
int last_user_token_number;
static char DOLLAR[] = "$";
tags = XCALLOC (char *, nsyms + 1);
- tags[0] = DOLLAR;
user_toknums = XCALLOC (short, nsyms + 1);
- user_toknums[0] = 0;
sprec = XCALLOC (short, nsyms);
sassoc = XCALLOC (short, nsyms);
+ /* The EOF token. */
+ tags[0] = DOLLAR;
+ user_toknums[0] = 0;
+
max_user_token_number = 256;
last_user_token_number = 256;
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)
- {
- int j;
-
- token_translations = XCALLOC (short, max_user_token_number + 1);
-
- /* initialize all entries for literal tokens to 2, the internal
- token number for $undefined., which represents all invalid
- inputs. */
- for (j = 0; j <= max_user_token_number; j++)
- token_translations[j] = 2;
-
- for (bp = firstsymbol; bp; bp = bp->next)
- {
- if (bp->value >= ntokens)
- continue; /* non-terminal */
- if (bp->user_token_number == SALIAS)
- continue;
- if (token_translations[bp->user_token_number] != 2)
- complain (_("tokens %s and %s both assigned number %d"),
- tags[token_translations[bp->user_token_number]],
- bp->tag, bp->user_token_number);
- token_translations[bp->user_token_number] = bp->value;
- }
- }
+ token_translations_init ();
error_token_number = errtoken->value;
- output_token_defines (&output_obstack);
- obstack_1grow (&output_obstack, 0);
- muscle_insert ("tokendef", obstack_finish (&output_obstack));
-
-#if 0
- if (!no_parser_flag)
- output_token_defines (&table_obstack);
-#endif
-
if (startval->class == unknown_sym)
fatal (_("the start symbol %s is undefined"), startval->tag);
else if (startval->class == token_sym)
fatal (_("the start symbol %s is a token"), startval->tag);
start_symbol = startval->value;
+}
+
+
+/*-----------------------------------.
+| Output definition of token names. |
+`-----------------------------------*/
+
+static void
+symbols_output (void)
+{
+ {
+ struct obstack tokendefs;
+ obstack_init (&tokendefs);
+ output_token_defines (&tokendefs);
+ obstack_1grow (&tokendefs, 0);
+ muscle_insert ("tokendef", xstrdup (obstack_finish (&tokendefs)));
+ obstack_free (&tokendefs, NULL);
+ }
if (defines_flag)
{
output_token_defines (&defines_obstack);
if (!pure_parser)
+ obstack_fgrow1 (&defines_obstack, "\nextern YYSTYPE %slval;\n",
+ spec_name_prefix);
+ if (semantic_parser)
{
- if (spec_name_prefix)
- obstack_fgrow1 (&defines_obstack, "\nextern YYSTYPE %slval;\n",
- spec_name_prefix);
- else
- obstack_sgrow (&defines_obstack,
- "\nextern YYSTYPE yylval;\n");
- }
+ int i;
- if (semantic_parser)
- for (i = ntokens; i < nsyms; i++)
- {
- /* don't make these for dummy nonterminals made by gensym. */
- if (*tags[i] != '@')
- obstack_fgrow2 (&defines_obstack,
- "# define\tNT%s\t%d\n", tags[i], i);
- }
+ 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
+ }
}
}
int ruleno;
symbol_list *p;
- bucket *ruleprec;
-
ritem = XCALLOC (short, nitems + 1);
- rlhs = XCALLOC (short, nrules) - 1;
- rrhs = XCALLOC (short, nrules) - 1;
- rprec = XCALLOC (short, nrules) - 1;
- rprecsym = XCALLOC (short, nrules) - 1;
- rassoc = XCALLOC (short, nrules) - 1;
+ rule_table = XCALLOC (rule_t, nrules) - 1;
itemno = 0;
ruleno = 1;
p = grammar;
while (p)
{
- rlhs[ruleno] = p->sym->value;
- rrhs[ruleno] = itemno;
- ruleprec = p->ruleprec;
+ bucket *ruleprec = p->ruleprec;
+ rule_table[ruleno].lhs = p->sym->value;
+ rule_table[ruleno].rhs = itemno;
+ rule_table[ruleno].line = p->line;
+ rule_table[ruleno].useful = TRUE;
p = p->next;
while (p && p->sym)
of the last token in it. */
if (p->sym->class == token_sym)
{
- rprec[ruleno] = p->sym->prec;
- rassoc[ruleno] = p->sym->assoc;
+ rule_table[ruleno].prec = p->sym->prec;
+ rule_table[ruleno].assoc = p->sym->assoc;
}
if (p)
p = p->next;
the specified symbol's precedence replaces the default. */
if (ruleprec)
{
- rprec[ruleno] = ruleprec->prec;
- rassoc[ruleno] = ruleprec->assoc;
- rprecsym[ruleno] = ruleprec->value;
+ rule_table[ruleno].prec = ruleprec->prec;
+ rule_table[ruleno].assoc = ruleprec->assoc;
+ rule_table[ruleno].precsym = ruleprec->value;
}
ritem[itemno++] = -ruleno;
}
ritem[itemno] = 0;
+
+ if (trace_flag)
+ ritem_print (stderr);
}
\f
/*-------------------------------------------------------------------.
start_flag = 0;
startval = NULL; /* start symbol not specified yet. */
-#if 0
- /* initially assume token number translation not needed. */
- translations = 0;
-#endif
- /* Nowadays translations is always set to 1, since we give `error' a
- user-token-number to satisfy the Posix demand for YYERRCODE==256.
- */
- translations = 1;
-
nsyms = 1;
nvars = 0;
nrules = 0;
nitems = 0;
- rline_allocated = 10;
- rline = XCALLOC (short, rline_allocated);
typed = 0;
lastprec = 0;
grammar = NULL;
- init_lex ();
+ lex_init ();
lineno = 1;
/* Initialize the muscle obstack. */
packsymbols ();
/* Convert the grammar into the format described in gram.h. */
packgram ();
- /* Free the symbol table data structure since symbols are now all
- referred to by symbol number. */
- free_symtab ();
+ /* Output the headers. */
+ symbols_output ();
}
projects / bison.git / blobdiff