/* Output the generated parsing program for Bison.
Copyright (C) 1984, 1986, 1989, 1992, 2000, 2001, 2002, 2003, 2004,
- 2005, 2006, 2007, 2008 Free Software Foundation, Inc.
+ 2005, 2006, 2007, 2008, 2009, 2010 Free Software Foundation, Inc.
This file is part of Bison, the GNU Compiler Compiler.
#include <configmake.h>
#include <error.h>
#include <get-errno.h>
+#include <pipe.h>
#include <quotearg.h>
-#include <subpipe.h>
#include <timevar.h>
+#include <wait-process.h>
#include "complain.h"
#include "files.h"
#include "getargs.h"
#include "gram.h"
-#include "muscle_tab.h"
+#include "muscle-tab.h"
#include "output.h"
#include "reader.h"
#include "scan-code.h" /* max_left_semantic_context */
#include "symtab.h"
#include "tables.h"
+# define ARRAY_CARDINALITY(Array) (sizeof (Array) / sizeof *(Array))
static struct obstack format_obstack;
GENERATE_MUSCLE_INSERT_TABLE(muscle_insert_base_table, base_number)
GENERATE_MUSCLE_INSERT_TABLE(muscle_insert_rule_number_table, rule_number)
GENERATE_MUSCLE_INSERT_TABLE(muscle_insert_symbol_number_table, symbol_number)
-GENERATE_MUSCLE_INSERT_TABLE(muscle_insert_item_number_table, item_number)
GENERATE_MUSCLE_INSERT_TABLE(muscle_insert_state_number_table, state_number)
MUSCLE_INSERT_BOOL ("token_table", token_table_flag);
MUSCLE_INSERT_INT ("tokens_number", ntokens);
MUSCLE_INSERT_INT ("nterms_number", nvars);
+ MUSCLE_INSERT_INT ("symbols_number", nsyms);
MUSCLE_INSERT_INT ("undef_token_number", undeftoken->number);
MUSCLE_INSERT_INT ("user_token_number_max", max_user_token_number);
}
-/*-------------------------------------------------------------.
-| Prepare the muscles related to the rules: rhs, prhs, r1, r2, |
-| rline, dprec, merger. |
-`-------------------------------------------------------------*/
+/*----------------------------------------------------------------.
+| Prepare the muscles related to the rules: r1, r2, rline, dprec, |
+| merger, immediate. |
+`----------------------------------------------------------------*/
static void
prepare_rules (void)
{
- rule_number r;
- unsigned int i = 0;
- item_number *rhs = xnmalloc (nritems, sizeof *rhs);
- unsigned int *prhs = xnmalloc (nrules, sizeof *prhs);
unsigned int *rline = xnmalloc (nrules, sizeof *rline);
symbol_number *r1 = xnmalloc (nrules, sizeof *r1);
unsigned int *r2 = xnmalloc (nrules, sizeof *r2);
int *dprec = xnmalloc (nrules, sizeof *dprec);
int *merger = xnmalloc (nrules, sizeof *merger);
+ int *immediate = xnmalloc (nrules, sizeof *immediate);
+ rule_number r;
for (r = 0; r < nrules; ++r)
{
- item_number *rhsp = NULL;
- /* Index of rule R in RHS. */
- prhs[r] = i;
- /* RHS of the rule R. */
- for (rhsp = rules[r].rhs; *rhsp >= 0; ++rhsp)
- rhs[i++] = *rhsp;
/* LHS of the rule R. */
r1[r] = rules[r].lhs->number;
/* Length of rule R's RHS. */
- r2[r] = i - prhs[r];
- /* Separator in RHS. */
- rhs[i++] = -1;
+ r2[r] = rule_rhs_length(&rules[r]);
/* Line where rule was defined. */
rline[r] = rules[r].location.start.line;
/* Dynamic precedence (GLR). */
dprec[r] = rules[r].dprec;
/* Merger-function index (GLR). */
merger[r] = rules[r].merger;
+ /* Immediate reduction flags (GLR). */
+ immediate[r] = rules[r].is_predicate;
}
- aver (i == nritems);
- muscle_insert_item_number_table ("rhs", rhs, ritem[0], 1, nritems);
- muscle_insert_unsigned_int_table ("prhs", prhs, 0, 0, nrules);
muscle_insert_unsigned_int_table ("rline", rline, 0, 0, nrules);
muscle_insert_symbol_number_table ("r1", r1, 0, 0, nrules);
muscle_insert_unsigned_int_table ("r2", r2, 0, 0, nrules);
muscle_insert_int_table ("dprec", dprec, 0, 0, nrules);
muscle_insert_int_table ("merger", merger, 0, 0, nrules);
+ muscle_insert_int_table ("immediate", immediate, 0, 0, nrules);
MUSCLE_INSERT_INT ("rules_number", nrules);
MUSCLE_INSERT_INT ("max_left_semantic_context", max_left_semantic_context);
- free (rhs);
- free (prhs);
free (rline);
free (r1);
free (r2);
}
+/*-------------------------------------------------------.
+| Compare two symbols by type-name, and then by number. |
+`-------------------------------------------------------*/
+
+static int
+symbol_type_name_cmp (const symbol **lhs, const symbol **rhs)
+{
+ int res = UNIQSTR_CMP((*lhs)->type_name, (*rhs)->type_name);
+ if (res)
+ return res;
+ return (*lhs)->number - (*rhs)->number;
+}
+
-/*-----------------------------------------------.
-| For each symbol type, its tags and type name. |
-`-----------------------------------------------*/
+/*----------------------------------------------------------------.
+| Return a (malloc'ed) table of the symbols sorted by type-name. |
+`----------------------------------------------------------------*/
+
+static symbol **
+symbols_by_type_name (void)
+{
+ typedef int (*qcmp_type) (const void *, const void *);
+ symbol **res = xmemdup (symbols, nsyms * sizeof *res);
+ qsort (res, nsyms, sizeof *res, (qcmp_type) &symbol_type_name_cmp);
+ return res;
+}
+
+
+/*------------------------------------------------------------------.
+| Define b4_type_names, which is a list of (lists of the numbers of |
+| symbols with same type-name). |
+`------------------------------------------------------------------*/
static void
type_names_output (FILE *out)
{
int i;
-
+ symbol **syms = symbols_by_type_name ();
fputs ("m4_define([b4_type_names],\n[", out);
- for (i = 0; i < nsyms; ++i)
+ for (i = 0; i < nsyms; /* nothing */)
{
- symbol *sym = symbols[i];
- /* Symbol-name, Symbol-number, optional typename. */
- fprintf (out, "%s[", i ? ",\n" : "");
- escaped_output (out, sym->tag);
- fprintf (out, ", %d, [[%s]]]",
- sym->number,
- sym->type_name ? sym->type_name : "");
+ // The index of the first symbol of the current type-name.
+ int i0 = i;
+ fputs (i ? ",\n[" : "[", out);
+ for (; i < nsyms && syms[i]->type_name == syms[i0]->type_name; ++i)
+ fprintf (out, "%s%d", i != i0 ? ", " : "", syms[i]->number);
+ fputs ("]", out);
}
fputs ("])\n\n", out);
+ free (syms);
+}
+
+
+/*-------------------------------------.
+| The list of all the symbol numbers. |
+`-------------------------------------*/
+
+static void
+symbol_numbers_output (FILE *out)
+{
+ int i;
+ fputs ("m4_define([b4_symbol_numbers],\n[", out);
+ for (i = 0; i < nsyms; ++i)
+ fprintf (out, "%s[%d]", i ? ", " : "", i);
+ fputs ("])\n\n", out);
}
for (r = 0; r < nrules; ++r)
if (rules[r].action)
{
- fprintf (out, "b4_case(%d, [b4_syncline(%d, ", r + 1,
- rules[r].action_location.start.line);
+ fprintf (out, "b4_%scase(%d, [b4_syncline(%d, ",
+ rules[r].is_predicate ? "predicate_" : "",
+ r + 1, rules[r].action_location.start.line);
escaped_output (out, rules[r].action_location.start.file);
fprintf (out, ")\n[ %s]])\n\n", rules[r].action);
}
fputs ("])\n\n", out);
}
-/*--------------------------------------.
-| Output the merge functions to OUT. |
-`--------------------------------------*/
+/*------------------------------------.
+| Output the merge functions to OUT. |
+`------------------------------------*/
static void
merger_output (FILE *out)
fputs ("]])\n\n", out);
}
-/*--------------------------------------.
-| Output the tokens definition to OUT. |
-`--------------------------------------*/
+
+/*---------------------------------------------.
+| Prepare the muscles for symbol definitions. |
+`---------------------------------------------*/
static void
-token_definitions_output (FILE *out)
+prepare_symbol_definitions (void)
{
int i;
- char const *sep = "";
-
- fputs ("m4_define([b4_tokens], \n[", out);
- for (i = 0; i < ntokens; ++i)
+ for (i = 0; i < nsyms; ++i)
{
symbol *sym = symbols[i];
- int number = sym->user_token_number;
-
- /* At this stage, if there are literal aliases, they are part of
- SYMBOLS, so we should not find symbols which are the aliases
- here. */
- aver (number != USER_NUMBER_ALIAS);
-
- /* Skip error token. */
- if (sym == errtoken)
- continue;
-
- /* If this string has an alias, then it is necessarily the alias
- which is to be output. */
- if (sym->alias)
- sym = sym->alias;
-
- /* Don't output literal chars or strings (when defined only as a
- string). Note that must be done after the alias resolution:
- think about `%token 'f' "f"'. */
- if (sym->tag[0] == '\'' || sym->tag[0] == '\"')
- continue;
-
- /* Don't #define nonliteral tokens whose names contain periods
- or '$' (as does the default value of the EOF token). */
- if (strchr (sym->tag, '.') || strchr (sym->tag, '$'))
- continue;
-
- fprintf (out, "%s[[[%s]], %d]",
- sep, sym->tag, number);
- sep = ",\n";
+ const char *key;
+ const char *value;
+
+#define SET_KEY(Entry) \
+ obstack_fgrow2 (&format_obstack, "symbol(%d, %s)", i, Entry); \
+ obstack_1grow (&format_obstack, 0); \
+ key = obstack_finish (&format_obstack);
+
+ // Whether the symbol has an identifier.
+ value = symbol_id_get (sym);
+ SET_KEY("has_id");
+ MUSCLE_INSERT_INT (key, !!value);
+
+ // Its identifier.
+ SET_KEY("id");
+ MUSCLE_INSERT_STRING (key, value ? value : "");
+
+ // Its tag. Typically for documentation purpose.
+ SET_KEY("tag");
+ MUSCLE_INSERT_STRING (key, sym->tag);
+
+ SET_KEY("user_number");
+ MUSCLE_INSERT_INT (key, sym->user_token_number);
+
+ SET_KEY("is_token");
+ MUSCLE_INSERT_INT (key,
+ i < ntokens && sym != errtoken && sym != undeftoken);
+
+ SET_KEY("number");
+ MUSCLE_INSERT_INT (key, sym->number);
+
+ SET_KEY("has_type");
+ MUSCLE_INSERT_INT (key, !!sym->type_name);
+
+ SET_KEY("type");
+ MUSCLE_INSERT_STRING (key, sym->type_name ? sym->type_name : "");
+
+#define CODE_PROP(PropName) \
+ do { \
+ code_props const *p = symbol_ ## PropName ## _get (sym); \
+ SET_KEY("has_" #PropName); \
+ MUSCLE_INSERT_INT (key, !!p->code); \
+ \
+ if (p->code) \
+ { \
+ SET_KEY(#PropName "_file"); \
+ MUSCLE_INSERT_STRING (key, p->location.start.file); \
+ \
+ SET_KEY(#PropName "_line"); \
+ MUSCLE_INSERT_INT (key, p->location.start.line); \
+ \
+ SET_KEY(#PropName); \
+ MUSCLE_INSERT_STRING_RAW (key, p->code); \
+ } \
+ } while (0)
+
+ CODE_PROP(destructor);
+ CODE_PROP(printer);
+#undef CODE_PROP
+#undef SET_KEY
}
- fputs ("])\n\n", out);
}
-/*---------------------------------------------------.
-| Output the symbol destructors or printers to OUT. |
-`---------------------------------------------------*/
+/*--------------------------------------.
+| Output the tokens definition to OUT. |
+`--------------------------------------*/
static void
-symbol_code_props_output (FILE *out, char const *what,
- code_props const *(*get)(symbol const *))
+token_definitions_output (FILE *out)
{
int i;
char const *sep = "";
- fputs ("m4_define([b4_symbol_", out);
- fputs (what, out);
- fputs ("], \n[", out);
- for (i = 0; i < nsyms; ++i)
+ fputs ("m4_define([b4_tokens], \n[", out);
+ for (i = 0; i < ntokens; ++i)
{
symbol *sym = symbols[i];
- char const *code = (*get) (sym)->code;
- if (code)
+ int number = sym->user_token_number;
+ uniqstr id = symbol_id_get (sym);
+
+ /* At this stage, if there are literal string aliases, they are
+ part of SYMBOLS, so we should not find their aliased symbols
+ here. */
+ aver (number != USER_NUMBER_HAS_STRING_ALIAS);
+
+ /* Skip error token and tokens without identifier. */
+ if (sym != errtoken && id)
{
- location loc = (*get) (sym)->location;
- /* Filename, lineno,
- Symbol-name, Symbol-number,
- code, optional typename. */
- fprintf (out, "%s[", sep);
+ fprintf (out, "%s[[[%s]], %d]",
+ sep, id, number);
sep = ",\n";
- escaped_output (out, loc.start.file);
- fprintf (out, ", %d, ", loc.start.line);
- escaped_output (out, sym->tag);
- fprintf (out, ", %d, [[%s]]", sym->number, code);
- if (sym->type_name)
- fprintf (out, ", [[%s]]", sym->type_name);
- fputc (']', out);
}
}
fputs ("])\n\n", out);
muscle_insert_unsigned_int_table ("conflicting_rules", conflict_list,
0, 1, conflict_list_cnt);
}
+
+
+/*--------------------------------------------.
+| Output the definitions of all the muscles. |
+`--------------------------------------------*/
+
+static void
+muscles_output (FILE *out)
+{
+ fputs ("m4_init()\n", out);
+ merger_output (out);
+ symbol_numbers_output (out);
+ token_definitions_output (out);
+ type_names_output (out);
+ user_actions_output (out);
+ // Must be last.
+ muscles_m4_output (out);
+}
\f
/*---------------------------.
| Call the skeleton parser. |
output_skeleton (void)
{
FILE *in;
- FILE *out;
int filter_fd[2];
- char const *argv[9];
+ char const *argv[10];
pid_t pid;
/* Compute the names of the package data dir and skeleton files. */
{
int i = 0;
argv[i++] = m4;
+
+ /* When POSIXLY_CORRECT is set, GNU M4 1.6 and later disable GNU
+ extensions, which Bison's skeletons depend on. With older M4,
+ it has no effect. M4 1.4.12 added a -g/--gnu command-line
+ option to make it explicit that a program wants GNU M4
+ extensions even when POSIXLY_CORRECT is set.
+
+ See the thread starting at
+ <http://lists.gnu.org/archive/html/bug-bison/2008-07/msg00000.html>
+ for details. */
+ if (*M4_GNU_OPTION)
+ argv[i++] = M4_GNU_OPTION;
+
argv[i++] = "-I";
argv[i++] = pkgdatadir;
if (trace_flag & trace_m4)
argv[i++] = full_m4bison;
argv[i++] = full_skeleton;
argv[i++] = NULL;
+ aver (i <= ARRAY_CARDINALITY (argv));
}
- /* When POSIXLY_CORRECT is set, some future versions of GNU M4 (most likely
- 2.0) may drop some of the GNU extensions that Bison's skeletons depend
- upon. So that the next release of Bison is forward compatible with those
- future versions of GNU M4, we unset POSIXLY_CORRECT here.
-
- FIXME: A user might set POSIXLY_CORRECT to affect processes run from
- macros like m4_syscmd in a custom skeleton. For now, Bison makes no
- promises about the behavior of custom skeletons, so this scenario is not a
- concern. However, we eventually want to eliminate this shortcoming. The
- next release of GNU M4 (1.4.12 or 1.6) will accept the -g command-line
- option as a no-op, and later releases will accept it to indicate that
- POSIXLY_CORRECT should be ignored. Once the GNU M4 versions that accept
- -g are pervasive, Bison should use -g instead of unsetting
- POSIXLY_CORRECT.
- See the thread starting at
- <http://lists.gnu.org/archive/html/bug-bison/2008-07/msg00000.html>
- for details. */
- unsetenv ("POSIXLY_CORRECT");
- init_subpipe ();
- pid = create_subpipe (argv, filter_fd);
+ /* The ugly cast is because gnulib gets the const-ness wrong. */
+ pid = create_pipe_bidi ("m4", m4, (char **)(void*)argv, false, true,
+ true, filter_fd);
free (full_m4sugar);
free (full_m4bison);
free (full_skeleton);
- out = fdopen (filter_fd[0], "w");
- if (! out)
- error (EXIT_FAILURE, get_errno (),
- "fdopen");
-
- /* Output the definitions of all the muscles. */
- fputs ("m4_init()\n", out);
-
- type_names_output (out);
- user_actions_output (out);
- merger_output (out);
- token_definitions_output (out);
- symbol_code_props_output (out, "destructors", &symbol_destructor_get);
- symbol_code_props_output (out, "printers", &symbol_printer_get);
-
- muscles_m4_output (out);
- xfclose (out);
+ if (trace_flag & trace_muscles)
+ muscles_output (stderr);
+ {
+ FILE *out = fdopen (filter_fd[1], "w");
+ if (! out)
+ error (EXIT_FAILURE, get_errno (),
+ "fdopen");
+ muscles_output (out);
+ xfclose (out);
+ }
/* Read and process m4's output. */
timevar_push (TV_M4);
- end_of_output_subpipe (pid, filter_fd);
- in = fdopen (filter_fd[1], "r");
+ in = fdopen (filter_fd[0], "r");
if (! in)
error (EXIT_FAILURE, get_errno (),
"fdopen");
scan_skel (in);
+ /* scan_skel should have read all of M4's output. Otherwise, when we
+ close the pipe, we risk letting M4 report a broken-pipe to the
+ Bison user. */
+ aver (feof (in));
xfclose (in);
- reap_subpipe (pid, m4);
+ wait_subprocess (pid, "m4", false, false, true, true, NULL);
timevar_pop (TV_M4);
}
use_push_for_pull_flag = true;
/* Flags. */
- MUSCLE_INSERT_BOOL ("debug_flag", debug_flag);
MUSCLE_INSERT_BOOL ("defines_flag", defines_flag);
- MUSCLE_INSERT_BOOL ("error_verbose_flag", error_verbose);
MUSCLE_INSERT_BOOL ("glr_flag", glr_parser);
- MUSCLE_INSERT_BOOL ("locations_flag", locations_flag);
MUSCLE_INSERT_BOOL ("nondeterministic_flag", nondeterministic_parser);
MUSCLE_INSERT_BOOL ("synclines_flag", !no_lines_flag);
MUSCLE_INSERT_BOOL ("tag_seen_flag", tag_seen);
prepare_rules ();
prepare_states ();
prepare_actions ();
+ prepare_symbol_definitions ();
prepare ();
projects / bison.git / blobdiff