X-Git-Url: https://git.saurik.com/bison.git/blobdiff_plain/b85810aec0e5172c813b1cdd511bfce8a0cd4b4f..87675353bfb6d024d99d0b0298e65c9897600706:/src/output.c diff --git a/src/output.c b/src/output.c index d5eba988..6b082f57 100644 --- a/src/output.c +++ b/src/output.c @@ -1,5 +1,5 @@ /* Output the generated parsing program for bison, - Copyright 1984, 1986, 1989, 1992, 2000, 2001 + Copyright (C) 1984, 1986, 1989, 1992, 2000, 2001, 2002 Free Software Foundation, Inc. This file is part of Bison, the GNU Compiler Compiler. @@ -89,7 +89,9 @@ negative short int. Used to flag ?? */ #include "system.h" +#include "bitsetv.h" #include "quotearg.h" +#include "error.h" #include "getargs.h" #include "files.h" #include "gram.h" @@ -101,241 +103,341 @@ #include "symtab.h" #include "conflicts.h" #include "muscle_tab.h" -#include "skeleton.h" + +/* From lib/readpipe.h. */ +FILE *readpipe PARAMS ((const char *, ...)); + +/* From src/scan-skel.l. */ +int skel_lex PARAMS ((void)); +extern FILE *skel_in; static int nvectors; static int nentries; static short **froms = NULL; static short **tos = NULL; +static unsigned int **conflict_tos = NULL; static short *tally = NULL; static short *width = NULL; static short *actrow = NULL; +static short *conflrow = NULL; static short *state_count = NULL; static short *order = NULL; static short *base = NULL; static short *pos = NULL; + +static unsigned int *conflict_table = NULL; +static unsigned int *conflict_list = NULL; +static int conflict_list_cnt; +static int conflict_list_free; + +/* TABLE_SIZE is the allocated size of both TABLE and CHECK. + We start with the original hard-coded value: SHRT_MAX + (yes, not USHRT_MAX). */ +static size_t table_size = SHRT_MAX; static short *table = NULL; static short *check = NULL; static int lowzero; static int high; -struct obstack muscle_obstack; static struct obstack format_obstack; int error_verbose = 0; -/* Returns the number of lines of S. */ -size_t -get_lines_number (const char *s) -{ - size_t lines = 0; - size_t i; - for (i = 0; s[i]; ++i) - if (s[i] == '\n') - ++lines; +/*----------------------------------------------------------------. +| If TABLE (and CHECK) appear to be small to be addressed at | +| DESIRED, grow them. Note that TABLE[DESIRED] is to be used, so | +| the desired size is at least DESIRED + 1. | +`----------------------------------------------------------------*/ - return lines; -} +static void +table_grow (size_t desired) +{ + size_t old_size = table_size; + while (table_size <= desired) + table_size *= 2; -/* FIXME. */ + if (trace_flag) + fprintf (stderr, "growing table and check from: %d to %d\n", + old_size, table_size); -static inline void -output_table_data (struct obstack *oout, - short *table_data, - short first, - int begin, - int end) -{ - int i; - int j = 1; + table = XREALLOC (table, short, table_size); + check = XREALLOC (check, short, table_size); + if (glr_parser) + conflict_table = XREALLOC (conflict_table, unsigned int, table_size); - obstack_fgrow1 (oout, "%6d", first); - for (i = begin; i < end; ++i) + for (/* Nothing. */; old_size < table_size; ++old_size) { - obstack_1grow (oout, ','); - if (j >= 10) - { - obstack_sgrow (oout, "\n "); - j = 1; - } - else - ++j; - obstack_fgrow1 (oout, "%6d", table_data[i]); + table[old_size] = 0; + check[old_size] = -1; } - obstack_1grow (oout, 0); } -static void -output_token_translations (void) -{ - output_table_data (&format_obstack, token_translations, - 0, 1, max_user_token_number + 1); - muscle_insert ("translate", obstack_finish (&format_obstack)); - XFREE (token_translations); +/*-------------------------------------------------------------------. +| Create a function NAME which associates to the muscle NAME the | +| result of formatting the FIRST and then TABLE_DATA[BEGIN..END[ (of | +| TYPE), and to the muscle NAME_max, the max value of the | +| TABLE_DATA. | +`-------------------------------------------------------------------*/ + + +#define GENERATE_MUSCLE_INSERT_TABLE(Name, Type) \ + \ +static void \ +Name (const char *name, \ + Type *table_data, \ + Type first, \ + int begin, \ + int end) \ +{ \ + Type max = first; \ + int i; \ + int j = 1; \ + \ + obstack_fgrow1 (&format_obstack, "%6d", first); \ + for (i = begin; i < end; ++i) \ + { \ + obstack_1grow (&format_obstack, ','); \ + if (j >= 10) \ + { \ + obstack_sgrow (&format_obstack, "\n "); \ + j = 1; \ + } \ + else \ + ++j; \ + obstack_fgrow1 (&format_obstack, "%6d", table_data[i]); \ + if (table_data[i] > max) \ + max = table_data[i]; \ + } \ + obstack_1grow (&format_obstack, 0); \ + muscle_insert (name, obstack_finish (&format_obstack)); \ + \ + /* Build `NAME_max' in the obstack. */ \ + obstack_fgrow1 (&format_obstack, "%s_max", name); \ + obstack_1grow (&format_obstack, 0); \ + MUSCLE_INSERT_LONG_INT (obstack_finish (&format_obstack), \ + (long int) max); \ } +GENERATE_MUSCLE_INSERT_TABLE(muscle_insert_unsigned_int_table, unsigned int) +GENERATE_MUSCLE_INSERT_TABLE(muscle_insert_short_table, short) +GENERATE_MUSCLE_INSERT_TABLE(muscle_insert_symbol_number_table, symbol_number_t) +GENERATE_MUSCLE_INSERT_TABLE(muscle_insert_item_number_table, item_number_t) +GENERATE_MUSCLE_INSERT_TABLE(muscle_insert_state_number_table, state_number_t) + + +/*-----------------------------------------------------------------. +| Prepare the muscles related to the tokens: translate, tname, and | +| toknum. | +`-----------------------------------------------------------------*/ static void -output_gram (void) +prepare_tokens (void) { + muscle_insert_symbol_number_table ("translate", + token_translations, + 0, 1, max_user_token_number + 1); + { int i; - short *values = XCALLOC (short, nrules + 1); - for (i = 0; i < nrules + 1; ++i) - values[i] = rules[i].rhs; - output_table_data (&format_obstack, values, - 0, 1, nrules + 1); - XFREE (values); - } + int j = 0; + for (i = 0; i < nsyms; i++) + { + /* Be sure not to use twice the same QUOTEARG slot: + SYMBOL_TAG_GET uses slot 0. */ + const char *cp = + quotearg_n_style (1, c_quoting_style, + symbol_tag_get (symbols[i])); + /* Width of the next token, including the two quotes, the coma + and the space. */ + int strsize = strlen (cp) + 2; + + if (j + strsize > 75) + { + obstack_sgrow (&format_obstack, "\n "); + j = 2; + } + + obstack_sgrow (&format_obstack, cp); + obstack_sgrow (&format_obstack, ", "); + j += strsize; + } + /* Add a NULL entry to list of tokens (well, 0, as NULL might not be + defined). */ + obstack_sgrow (&format_obstack, "0"); - muscle_insert ("prhs", obstack_finish (&format_obstack)); + /* Finish table and store. */ + obstack_1grow (&format_obstack, 0); + muscle_insert ("tname", obstack_finish (&format_obstack)); + } + /* Output YYTOKNUM. */ { - short *yyrhs; int i; + short *values = XCALLOC (short, ntokens + 1); + for (i = 0; i < ntokens + 1; ++i) + values[i] = symbols[i]->user_token_number; + muscle_insert_short_table ("toknum", values, + 0, 1, ntokens + 1); + free (values); + } +} - yyrhs = XMALLOC (short, nritems); - - for (i = 1; i < nritems; ++i) - yyrhs[i] = ritem[i] >= 0 ? ritem[i] : -1; - - output_table_data (&format_obstack, yyrhs, - ritem[0], 1, nritems); - muscle_insert ("rhs", obstack_finish (&format_obstack)); - XFREE (yyrhs); - } +/*-------------------------------------------------------------. +| Prepare the muscles related to the rules: rhs, prhs, r1, r2, | +| rline, dprec, merger | +`-------------------------------------------------------------*/ -#if 0 - if (!semantic_parser) - obstack_sgrow (&table_obstack, "\n#endif\n"); -#endif +static void +prepare_rules (void) +{ + rule_number_t r; + unsigned int i = 0; + item_number_t *rhs = XMALLOC (item_number_t, nritems); + unsigned int *prhs = XMALLOC (unsigned int, nrules + 1); + unsigned int *rline = XMALLOC (unsigned int, nrules + 1); + symbol_number_t *r1 = XMALLOC (symbol_number_t, nrules + 1); + unsigned int *r2 = XMALLOC (unsigned int, nrules + 1); + short *dprec = XMALLOC (short, nrules + 1); + short *merger = XMALLOC (short, nrules + 1); + + for (r = 1; r < nrules + 1; ++r) + { + item_number_t *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; + /* Line where rule was defined. */ + rline[r] = rules[r].location.first_line; + /* Dynamic precedence (GLR) */ + dprec[r] = rules[r].dprec; + /* Merger-function index (GLR) */ + merger[r] = rules[r].merger; + } + assert (i == nritems); + + muscle_insert_item_number_table ("rhs", rhs, ritem[0], 1, nritems); + muscle_insert_unsigned_int_table ("prhs", prhs, 0, 1, nrules + 1); + muscle_insert_unsigned_int_table ("rline", rline, 0, 1, nrules + 1); + muscle_insert_symbol_number_table ("r1", r1, 0, 1, nrules + 1); + muscle_insert_unsigned_int_table ("r2", r2, 0, 1, nrules + 1); + muscle_insert_short_table ("dprec", dprec, 0, 1, nrules + 1); + muscle_insert_short_table ("merger", merger, 0, 1, nrules + 1); + + free (rhs); + free (prhs); + free (rline); + free (r1); + free (r2); + free (dprec); + free (merger); } +/*--------------------------------------------. +| Prepare the muscles related to the states. | +`--------------------------------------------*/ static void -output_stos (void) +prepare_states (void) { - int i; - short *values = (short *) alloca (sizeof (short) * nstates); + state_number_t i; + symbol_number_t *values = + (symbol_number_t *) alloca (sizeof (symbol_number_t) * nstates); for (i = 0; i < nstates; ++i) values[i] = states[i]->accessing_symbol; - output_table_data (&format_obstack, values, - 0, 1, nstates); - muscle_insert ("stos", obstack_finish (&format_obstack)); + muscle_insert_symbol_number_table ("stos", values, + 0, 1, nstates); } +/*-------------------------------------------------------------------. +| For GLR parsers, for each conflicted token in STATE, as indicated | +| by non-zero entries in conflrow, create a list of possible | +| reductions that are alternatives to the shift or reduction | +| currently recorded for that token in STATE. Store the alternative | +| reductions followed by a 0 in conflict_list, updating | +| conflict_list_cnt, and storing an index to the start of the list | +| back into conflrow. | +`-------------------------------------------------------------------*/ + static void -output_rule_data (void) +conflict_row (state_t *state) { - int i; - int j; - short *short_tab = NULL; - - { - short *values = XCALLOC (short, nrules + 1); - for (i = 0; i < nrules + 1; ++i) - values[i] = rules[i].line; - output_table_data (&format_obstack, values, - 0, 1, nrules + 1); - muscle_insert ("rline", obstack_finish (&format_obstack)); - XFREE (values); - } - + int i, j; - j = 0; - for (i = 0; i < nsyms; i++) - { - /* Be sure not to use twice the same quotearg slot. */ - const char *cp = - quotearg_n_style (1, c_quoting_style, - quotearg_style (escape_quoting_style, symbols[i]->tag)); - /* Width of the next token, including the two quotes, the coma - and the space. */ - int strsize = strlen (cp) + 2; - - if (j + strsize > 75) - { - obstack_sgrow (&format_obstack, "\n "); - j = 2; - } - - obstack_sgrow (&format_obstack, cp); - obstack_sgrow (&format_obstack, ", "); - j += strsize; - } - /* add a NULL entry to list of tokens */ - obstack_sgrow (&format_obstack, "NULL"); - - /* Finish table and store. */ - obstack_1grow (&format_obstack, 0); - muscle_insert ("tname", obstack_finish (&format_obstack)); - - /* Output YYTOKNUM. */ - { - short *values = XCALLOC (short, ntokens + 1); - for (i = 0; i < ntokens + 1; ++i) - values[i] = symbols[i]->user_token_number; - output_table_data (&format_obstack, values, - 0, 1, ntokens + 1); - muscle_insert ("toknum", obstack_finish (&format_obstack)); - XFREE (values); - } + if (! glr_parser) + return; + for (j = 0; j < ntokens; j += 1) + if (conflrow[j]) + { + conflrow[j] = conflict_list_cnt; - /* Output YYR1. */ - { - short *values = XCALLOC (short, nrules + 1); - for (i = 0; i < nrules + 1; ++i) - values[i] = rules[i].lhs; - output_table_data (&format_obstack, values, - 0, 1, nrules + 1); - muscle_insert ("r1", obstack_finish (&format_obstack)); - XFREE (values); - } + /* find all reductions for token j, and record all that do + * not match actrow[j] */ + for (i = 0; i < state->nlookaheads; i += 1) + if (bitset_test (state->lookaheads[i], j) + && actrow[j] != -state->lookaheads_rule[i]->number) + { + assert (conflict_list_free > 0); + conflict_list[conflict_list_cnt] + = state->lookaheads_rule[i]->number; + conflict_list_cnt += 1; + conflict_list_free -= 1; + } - /* Output YYR2. */ - short_tab = XMALLOC (short, nrules + 1); - for (i = 1; i < nrules; i++) - short_tab[i] = rules[i + 1].rhs - rules[i].rhs - 1; - short_tab[nrules] = nritems - rules[nrules].rhs - 1; - output_table_data (&format_obstack, short_tab, - 0, 1, nrules + 1); - muscle_insert ("r2", obstack_finish (&format_obstack)); - XFREE (short_tab); + /* Leave a 0 at the end */ + assert (conflict_list_free > 0); + conflict_list_cnt += 1; + conflict_list_free -= 1; + } } + /*------------------------------------------------------------------. | Decide what to do for each type of token if seen as the lookahead | | token in specified state. The value returned is used as the | | default action (yydefact) for the state. In addition, actrow is | | filled with what to do for each kind of token, index by symbol | | number, with zero meaning do the default action. The value | -| MINSHORT, a very negative number, means this situation is an | +| SHRT_MIN, a very negative number, means this situation is an | | error. The parser recognizes this value specially. | | | | This is where conflicts are resolved. The loop over lookahead | | rules considered lower-numbered rules last, and the last rule | | considered that likes a token gets to handle it. | +| | +| For GLR parsers, also sets conflrow[SYM] to an index into | +| conflict_list iff there is an unresolved conflict (s/r or r/r) | +| with symbol SYM. The default reduction is not used for a symbol | +| that has any such conflicts. | `------------------------------------------------------------------*/ static int action_row (state_t *state) { int i; - int default_rule = 0; - reductions *redp = state->reductions; - shifts *shiftp = state->shifts; - errs *errp = state->errs; + rule_number_t default_rule = 0; + reductions_t *redp = state->reductions; + shifts_t *shiftp = state->shifts; + errs_t *errp = state->errs; /* set nonzero to inhibit having any default reduction */ int nodefault = 0; + int conflicted = 0; for (i = 0; i < ntokens; i++) - actrow[i] = 0; + actrow[i] = conflrow[i] = 0; if (redp->nreds >= 1) { @@ -345,42 +447,41 @@ action_row (state_t *state) for (i = state->nlookaheads - 1; i >= 0; --i) /* and find each token which the rule finds acceptable to come next */ - for (j = 0; j < ntokens; j++) + BITSET_EXECUTE (state->lookaheads[i], 0, j, + { /* and record this rule as the rule to use if that token follows. */ - if (BITISSET (LA (state->lookaheadsp + i), j)) - actrow[j] = -LAruleno[state->lookaheadsp + i]; + if (actrow[j] != 0) + conflicted = conflrow[j] = 1; + actrow[j] = -state->lookaheads_rule[i]->number; + }); } /* Now see which tokens are allowed for shifts in this state. For them, record the shift as the thing to do. So shift is preferred to reduce. */ - for (i = 0; i < shiftp->nshifts; i++) - { - int symbol; - int shift_state = shiftp->shifts[i]; - if (!shift_state) - continue; - - symbol = states[shift_state]->accessing_symbol; - - if (ISVAR (symbol)) - break; + for (i = 0; i < shiftp->nshifts && SHIFT_IS_SHIFT (shiftp, i); i++) + if (!SHIFT_IS_DISABLED (shiftp, i)) + { + symbol_number_t symbol = SHIFT_SYMBOL (shiftp, i); + state_number_t shift_state = shiftp->shifts[i]; - actrow[symbol] = shift_state; + if (actrow[symbol] != 0) + conflicted = conflrow[symbol] = 1; + actrow[symbol] = state_number_as_int (shift_state); - /* Do not use any default reduction if there is a shift for - error */ - if (symbol == error_token_number) - nodefault = 1; - } + /* Do not use any default reduction if there is a shift for + error */ + if (symbol == errtoken->number) + nodefault = 1; + } /* See which tokens are an explicit error in this state (due to - %nonassoc). For them, record MINSHORT as the action. */ + %nonassoc). For them, record SHRT_MIN as the action. */ for (i = 0; i < errp->nerrs; i++) { int symbol = errp->errs[i]; - actrow[symbol] = MINSHORT; + actrow[symbol] = SHRT_MIN; } /* Now find the most common reduction and make it the default action @@ -396,11 +497,11 @@ action_row (state_t *state) for (i = 0; i < state->nlookaheads; i++) { int count = 0; - int rule = -LAruleno[state->lookaheadsp + i]; - int j; + rule_number_t rule = state->lookaheads_rule[i]->number; + symbol_number_t j; for (j = 0; j < ntokens; j++) - if (actrow[j] == rule) + if (actrow[j] == -rule) count++; if (count > max) @@ -410,17 +511,19 @@ action_row (state_t *state) } } - /* actions which match the default are replaced with zero, - which means "use the default" */ + /* GLR parsers need space for conflict lists, so we can't + default conflicted entries. For non-conflicted entries + or as long as we are not building a GLR parser, + actions that match the default are replaced with zero, + which means "use the default". */ if (max > 0) { int j; for (j = 0; j < ntokens; j++) - if (actrow[j] == default_rule) + if (actrow[j] == -default_rule + && ! (glr_parser && conflrow[j])) actrow[j] = 0; - - default_rule = -default_rule; } } } @@ -430,21 +533,25 @@ action_row (state_t *state) if (default_rule == 0) for (i = 0; i < ntokens; i++) - if (actrow[i] == MINSHORT) + if (actrow[i] == SHRT_MIN) actrow[i] = 0; + if (conflicted) + conflict_row (state); + return default_rule; } static void -save_row (int state) +save_row (state_number_t state) { - int i; + symbol_number_t i; int count; - short *sp; - short *sp1; - short *sp2; + short *sp = NULL; + short *sp1 = NULL; + short *sp2 = NULL; + unsigned int *sp3 = NULL; count = 0; for (i = 0; i < ntokens; i++) @@ -456,12 +563,18 @@ save_row (int state) froms[state] = sp1 = sp = XCALLOC (short, count); tos[state] = sp2 = XCALLOC (short, count); + if (glr_parser) + conflict_tos[state] = sp3 = XCALLOC (unsigned int, count); + else + conflict_tos[state] = NULL; for (i = 0; i < ntokens; i++) if (actrow[i] != 0) { *sp1++ = i; *sp2++ = actrow[i]; + if (glr_parser) + *sp3++ = conflrow[i]; } tally[state] = count; @@ -480,21 +593,33 @@ save_row (int state) static void token_actions (void) { - int i; + state_number_t i; + int nconflict = conflicts_total_count (); + short *yydefact = XCALLOC (short, nstates); actrow = XCALLOC (short, ntokens); + + conflrow = XCALLOC (short, ntokens); + if (glr_parser) + { + conflict_list = XCALLOC (unsigned int, 1 + 2 * nconflict); + conflict_list_free = 2 * nconflict; + conflict_list_cnt = 1; + } + else + conflict_list_free = conflict_list_cnt = 0; + for (i = 0; i < nstates; ++i) { yydefact[i] = action_row (states[i]); save_row (i); } - output_table_data (&format_obstack, yydefact, - yydefact[0], 1, nstates); - muscle_insert ("defact", obstack_finish (&format_obstack)); - + muscle_insert_short_table ("defact", yydefact, + yydefact[0], 1, nstates); XFREE (actrow); + XFREE (conflrow); XFREE (yydefact); } @@ -504,129 +629,178 @@ token_actions (void) `-----------------------------*/ void -actions_output (FILE *out, size_t *line) +actions_output (FILE *out) { - int rule; - for (rule = 1; rule < nrules + 1; ++rule) - if (rules[rule].action) + rule_number_t r; + + fputs ("m4_define([b4_actions], \n[[", out); + for (r = 1; r < nrules + 1; ++r) + if (rules[r].action) { - fprintf (out, " case %d:\n", rule); + fprintf (out, " case %d:\n", r); if (!no_lines_flag) fprintf (out, muscle_find ("linef"), - rules[rule].action_line, + rules[r].action_location.first_line, quotearg_style (c_quoting_style, muscle_find ("filename"))); - /* As a Bison extension, add the ending semicolon. Since some - Yacc don't do that, help people using bison as a Yacc - finding their missing semicolons. */ - fprintf (out, "{ %s%s }\n break;\n\n", - rules[rule].action, - yacc_flag ? ";" : ""); - - /* We always output 4 '\n' per action. */ - *line += 4; - /* Plus one if !no_lines_flag. */ - if (!no_lines_flag) - ++*line; - /* Get the number of lines written by the user. */ - *line += get_lines_number (rules[rule].action); + fprintf (out, " %s\n break;\n\n", + rules[r].action); } + fputs ("]])\n\n", out); } +/*--------------------------------------. +| Output the merge functions to OUT. | +`--------------------------------------*/ -/*----------------------------. -| Output the guards to OOUT. | -`----------------------------*/ - -void -guards_output (FILE *out, size_t *line) +static void +merger_output (FILE *out) { - int rule; - for (rule = 1; rule < nrules + 1; ++rule) - if (rules[rule].action) - { - fprintf (out, " case %d:\n", rule); + int n; + merger_list* p; - if (!no_lines_flag) - fprintf (out, muscle_find ("linef"), - rules[rule].guard_line, - quotearg_style (c_quoting_style, - muscle_find ("filename"))); - fprintf (out, "{ %s; }\n break;\n\n", - rules[rule].guard); - - /* We always output 4 '\n' per action. */ - *line += 4; - /* Plus one if !no_lines_flag. */ - if (!no_lines_flag) - ++*line; - /* Get the number of lines written by the user. */ - *line += get_lines_number (rules[rule].guard); - } + fputs ("m4_define([b4_mergers], \n[[", out); + for (n = 1, p = merge_functions; p != NULL; n += 1, p = p->next) + { + if (p->type[0] == '\0') + fprintf (out, " case %d: yyval = %s (*yy0, *yy1); break;\n", + n, p->name); + else + fprintf (out, " case %d: yyval.%s = %s (*yy0, *yy1); break;\n", + n, p->type, p->name); + } + fputs ("]])\n\n", out); } - /*---------------------------------------. | Output the tokens definition to OOUT. | `---------------------------------------*/ void -token_definitions_output (FILE *out, size_t *line) +token_definitions_output (FILE *out) { int i; + int first = 1; + + fputs ("m4_define([b4_tokens], \n[", out); for (i = 0; i < ntokens; ++i) { - bucket *symbol = symbols[i]; + symbol_t *symbol = symbols[i]; int number = symbol->user_token_number; - if (number == SALIAS) - continue; + /* At this stage, if there are literal aliases, they are part of + SYMBOLS, so we should not find symbols which are the aliases + here. */ + assert (number != USER_NUMBER_ALIAS); + /* Skip error token. */ - if (symbol->value == error_token_number) + if (symbol == errtoken) + continue; + + /* If this string has an alias, then it is necessarily the alias + which is to be output. */ + if (symbol->alias) + symbol = symbol->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 (symbol->tag[0] == '\'' || symbol->tag[0] == '\"') continue; - if (symbol->tag[0] == '\'') - continue; /* skip literal character */ - if (symbol->tag[0] == '\"') - { - /* use literal string only if given a symbol with an alias */ - if (symbol->alias) - symbol = symbol->alias; - else - continue; - } /* Don't #define nonliteral tokens whose names contain periods or '$' (as does the default value of the EOF token). */ if (strchr (symbol->tag, '.') || strchr (symbol->tag, '$')) continue; - fprintf (out, "# define %s\t%d\n", - symbol->tag, number); - ++*line; - if (semantic_parser) - { - /* FIXME: This is probably wrong, and should be just as - above. --akim. */ - fprintf (out, "# define T%s\t%d\n", symbol->tag, symbol->value); - ++*line; - } + fprintf (out, "%s[[[%s]], [%d]]", + first ? "" : ",\n", symbol->tag, number); + + first = 0; } + fputs ("])\n\n", out); +} + + +/*----------------------------------------. +| Output the symbol destructors to OOUT. | +`----------------------------------------*/ + +static void +symbol_destructors_output (FILE *out) +{ + int i; + int first = 1; + + fputs ("m4_define([b4_symbol_destructors], \n[", out); + for (i = 0; i < nsyms; ++i) + if (symbols[i]->destructor) + { + symbol_t *symbol = symbols[i]; + + /* Filename, lineno, + Symbol-name, Symbol-number, + destructor, typename. */ + fprintf (out, "%s[[[%s]], [[%d]], [[%s]], [[%d]], [[%s]], [[%s]]]", + first ? "" : ",\n", + infile, symbol->destructor_location.first_line, + symbol_tag_get (symbol), + symbol->number, + symbol->destructor, + symbol->type_name); + + first = 0; + } + fputs ("])\n\n", out); +} + + +/*-------------------------------------. +| Output the symbol printers to OOUT. | +`-------------------------------------*/ + +static void +symbol_printers_output (FILE *out) +{ + int i; + int first = 1; + + fputs ("m4_define([b4_symbol_printers], \n[", out); + for (i = 0; i < nsyms; ++i) + if (symbols[i]->destructor) + { + symbol_t *symbol = symbols[i]; + + /* Filename, lineno, + Symbol-name, Symbol-number, + destructor, typename. */ + fprintf (out, "%s[[[%s]], [[%d]], [[%s]], [[%d]], [[%s]], [[%s]]]", + first ? "" : ",\n", + infile, symbol->printer_location.first_line, + symbol_tag_get (symbol), + symbol->number, + symbol->printer, + symbol->type_name); + + first = 0; + } + fputs ("])\n\n", out); } static void -save_column (int symbol, int default_state) +save_column (symbol_number_t symbol, state_number_t default_state) { int i; short *sp; short *sp1; short *sp2; int count; - int symno = symbol - ntokens + nstates; + int symno = symbol - ntokens + state_number_as_int (nstates); - short begin = goto_map[symbol]; - short end = goto_map[symbol + 1]; + int begin = goto_map[symbol]; + int end = goto_map[symbol + 1]; count = 0; for (i = begin; i < end; i++) @@ -650,29 +824,31 @@ save_column (int symbol, int default_state) width[symno] = sp1[-1] - sp[0] + 1; } -static int -default_goto (int symbol) + +static state_number_t +default_goto (symbol_number_t symbol) { + state_number_t s; int i; int m = goto_map[symbol]; int n = goto_map[symbol + 1]; - int default_state = -1; + state_number_t default_state = (state_number_t) -1; int max = 0; if (m == n) - return -1; + return (state_number_t) -1; - for (i = 0; i < nstates; i++) - state_count[i] = 0; + for (s = 0; s < nstates; s++) + state_count[s] = 0; for (i = m; i < n; i++) state_count[to_state[i]]++; - for (i = 0; i < nstates; i++) - if (state_count[i] > max) + for (s = 0; s < nstates; s++) + if (state_count[s] > max) { - max = state_count[i]; - default_state = i; + max = state_count[s]; + default_state = s; } return default_state; @@ -691,21 +867,19 @@ default_goto (int symbol) static void goto_actions (void) { - int i; - short *yydefgoto = XMALLOC (short, nsyms - ntokens); + symbol_number_t i; + state_number_t *yydefgoto = XMALLOC (state_number_t, nsyms - ntokens); state_count = XCALLOC (short, nstates); for (i = ntokens; i < nsyms; ++i) { - int default_state = default_goto (i); + state_number_t default_state = default_goto (i); save_column (i, default_state); yydefgoto[i - ntokens] = default_state; } - output_table_data (&format_obstack, yydefgoto, - yydefgoto[0], 1, nsyms - ntokens); - muscle_insert ("defgoto", obstack_finish (&format_obstack)); - + muscle_insert_state_number_table ("defgoto", yydefgoto, + yydefgoto[0], 1, nsyms - ntokens); XFREE (state_count); XFREE (yydefgoto); } @@ -753,7 +927,7 @@ matching_state (int vector) int w; int prev; - if (i >= nstates) + if (i >= (int) nstates) return -1; t = tally[i]; @@ -789,19 +963,20 @@ pack_vector (int vector) int loc = 0; short *from = froms[i]; short *to = tos[i]; + unsigned int *conflict_to = conflict_tos[i]; assert (t); - for (j = lowzero - from[0]; j < MAXTABLE; j++) + for (j = lowzero - from[0]; j < (int) table_size; j++) { int k; int ok = 1; for (k = 0; ok && k < t; k++) { - loc = j + from[k]; - if (loc > MAXTABLE) - fatal (_("maximum table size (%d) exceeded"), MAXTABLE); + loc = j + state_number_as_int (from[k]); + if (loc > (int) table_size) + table_grow (loc); if (table[loc] != 0) ok = 0; @@ -815,9 +990,11 @@ pack_vector (int vector) { for (k = 0; k < t; k++) { - loc = j + from[k]; - table[loc] = to[k]; - check[loc] = from[k]; + loc = j + state_number_as_int (from[k]); + table[loc] = state_number_as_int (to[k]); + if (glr_parser && conflict_to != NULL) + conflict_table[loc] = conflict_to[k]; + check[loc] = state_number_as_int (from[k]); } while (table[lowzero] != 0) @@ -844,16 +1021,18 @@ pack_table (void) base = XCALLOC (short, nvectors); pos = XCALLOC (short, nentries); - table = XCALLOC (short, MAXTABLE); - check = XCALLOC (short, MAXTABLE); + table = XCALLOC (short, table_size); + if (glr_parser) + conflict_table = XCALLOC (unsigned int, table_size); + check = XCALLOC (short, table_size); lowzero = 0; high = 0; for (i = 0; i < nvectors; i++) - base[i] = MINSHORT; + base[i] = SHRT_MIN; - for (i = 0; i < MAXTABLE; i++) + for (i = 0; i < (int) table_size; i++) check[i] = -1; for (i = 0; i < nentries; i++) @@ -873,29 +1052,28 @@ pack_table (void) { XFREE (froms[i]); XFREE (tos[i]); + XFREE (conflict_tos[i]); } XFREE (froms); XFREE (tos); + XFREE (conflict_tos); XFREE (pos); } -/* the following functions output yytable, yycheck +/* the following functions output yytable, yycheck, yyconflp, yyconfl, and the vectors whose elements index the portion starts */ static void output_base (void) { /* Output pact. */ - output_table_data (&format_obstack, base, - base[0], 1, nstates); - muscle_insert ("pact", obstack_finish (&format_obstack)); + muscle_insert_short_table ("pact", base, + base[0], 1, nstates); /* Output pgoto. */ - output_table_data (&format_obstack, base, - base[nstates], nstates + 1, nvectors); - muscle_insert ("pgoto", obstack_finish (&format_obstack)); - + muscle_insert_short_table ("pgoto", base, + base[nstates], nstates + 1, nvectors); XFREE (base); } @@ -903,39 +1081,67 @@ output_base (void) static void output_table (void) { - output_table_data (&format_obstack, table, - table[0], 1, high + 1); - muscle_insert ("table", obstack_finish (&format_obstack)); + muscle_insert_short_table ("table", table, + table[0], 1, high + 1); XFREE (table); } +static void +output_conflicts (void) +{ + /* GLR parsing slightly modifies yytable and yycheck + (and thus yypact) so that in states with unresolved conflicts, + the default reduction is not used in the conflicted entries, so + that there is a place to put a conflict pointer. This means that + yyconflp and yyconfl are nonsense for a non-GLR parser, so we + avoid accidents by not writing them out in that case. */ + if (! glr_parser) + return; + + muscle_insert_unsigned_int_table ("conflict_list_heads", conflict_table, + conflict_table[0], 1, high+1); + muscle_insert_unsigned_int_table ("conflicting_rules", conflict_list, + conflict_list[0], 1, conflict_list_cnt); + + XFREE (conflict_table); + XFREE (conflict_list); +} + + static void output_check (void) { - output_table_data (&format_obstack, check, - check[0], 1, high + 1); - muscle_insert ("check", obstack_finish (&format_obstack)); + muscle_insert_short_table ("check", check, + check[0], 1, high + 1); XFREE (check); } -/* compute and output yydefact, yydefgoto, yypact, yypgoto, yytable - and yycheck. */ +/*-----------------------------------------------------------------. +| Compute and output yydefact, yydefgoto, yypact, yypgoto, yytable | +| and yycheck. | +`-----------------------------------------------------------------*/ static void output_actions (void) { - int i; - nvectors = nstates + nvars; + /* That's a poor way to make sure the sizes are properly corelated, + in particular the signedness is not taking into account, but it's + not useless. */ + assert (sizeof (nvectors) >= sizeof (nstates)); + assert (sizeof (nvectors) >= sizeof (nvars)); + + nvectors = state_number_as_int (nstates) + nvars; froms = XCALLOC (short *, nvectors); tos = XCALLOC (short *, nvectors); + conflict_tos = XCALLOC (unsigned int *, nvectors); tally = XCALLOC (short, nvectors); width = XCALLOC (short, nvectors); token_actions (); - XFREE (LA); - XFREE (LAruleno); + bitsetv_free (LA); + free (LArule); goto_actions (); XFREE (goto_map + ntokens); @@ -947,20 +1153,52 @@ output_actions (void) output_base (); output_table (); + output_conflicts (); output_check (); - - for (i = 0; i < nstates; ++i) - { - free (states[i]->shifts); - XFREE (states[i]->reductions); - free (states[i]->errs); - free (states[i]); - } - XFREE (states); } +/*----------------------. +| Run our backend, M4. | +`----------------------*/ + +static void +m4_invoke (const char *definitions) +{ + /* Invoke m4 on the definition of the muscles, and the skeleton. */ + const char *bison_pkgdatadir = getenv ("BISON_PKGDATADIR"); + const char *m4 = getenv ("M4"); + int pkg_data_len; + char *full_skeleton; + + if (!m4) + m4 = M4; + if (!bison_pkgdatadir) + bison_pkgdatadir = PKGDATADIR; + pkg_data_len = strlen (bison_pkgdatadir); + full_skeleton = XMALLOC (char, pkg_data_len + strlen (skeleton) + 2); + if (bison_pkgdatadir[pkg_data_len-1] == '/') + sprintf (full_skeleton, "%s%s", bison_pkgdatadir, skeleton); + else + sprintf (full_skeleton, "%s/%s", bison_pkgdatadir, skeleton); + if (trace_flag) + fprintf (stderr, + "running: %s -I %s m4sugar/m4sugar.m4 %s %s\n", + m4, bison_pkgdatadir, definitions, full_skeleton); + skel_in = readpipe (m4, + "-I", bison_pkgdatadir, + "m4sugar/m4sugar.m4", + definitions, + full_skeleton, + NULL); + XFREE (full_skeleton); + if (!skel_in) + error (EXIT_FAILURE, errno, "cannot run m4"); + skel_lex (); + +} + /*---------------------------. | Call the skeleton parser. | `---------------------------*/ @@ -968,105 +1206,96 @@ output_actions (void) static void output_skeleton (void) { - /* Find the right skeleton file. */ - if (!skeleton) - { - if (semantic_parser) - skeleton = skeleton_find ("BISON_HAIRY", BISON_HAIRY); - else - skeleton = skeleton_find ("BISON_SIMPLE", BISON_SIMPLE); - } + /* Store the definition of all the muscles. */ + const char *tempdir = getenv ("TMPDIR"); + char *tempfile = NULL; + FILE *out = NULL; + int fd; + + if (tempdir == NULL) + tempdir = DEFAULT_TMPDIR; + tempfile = xmalloc (strlen (tempdir) + 11); + sprintf (tempfile, "%s/bsnXXXXXX", tempdir); + fd = mkstemp (tempfile); + if (fd == -1) + error (EXIT_FAILURE, errno, "%s", tempfile); + + out = fdopen (fd, "w"); + if (out == NULL) + error (EXIT_FAILURE, errno, "%s", tempfile); + + /* There are no comments, especially not `#': we do want M4 expansion + after `#': think of CPP macros! */ + fputs ("m4_changecom()\n", out); + fputs ("m4_init()\n", out); + + actions_output (out); + merger_output (out); + token_definitions_output (out); + symbol_destructors_output (out); + symbol_printers_output (out); + + muscles_m4_output (out); + + fputs ("m4_wrap([m4_divert_pop(0)])\n", out); + fputs ("m4_divert_push(0)dnl\n", out); + xfclose (out); - /* Parse the skeleton file and output the needed parsers. */ - muscle_insert ("skeleton", skeleton); - process_skeleton (skeleton); + m4_invoke (tempfile); + + /* If `debugging', keep this file alive. */ + if (!trace_flag) + unlink (tempfile); + + free (tempfile); } static void prepare (void) { MUSCLE_INSERT_INT ("last", high); - MUSCLE_INSERT_INT ("flag", MINSHORT); + MUSCLE_INSERT_INT ("flag", SHRT_MIN); MUSCLE_INSERT_INT ("pure", pure_parser); MUSCLE_INSERT_INT ("nsym", nsyms); MUSCLE_INSERT_INT ("debug", debug_flag); - MUSCLE_INSERT_INT ("final", final_state); - MUSCLE_INSERT_INT ("maxtok", max_user_token_number); - MUSCLE_INSERT_INT ("error-verbose", error_verbose); + MUSCLE_INSERT_INT ("final", final_state->number); + MUSCLE_INSERT_INT ("undef_token_number", undeftoken->number); + MUSCLE_INSERT_INT ("user_token_number_max", max_user_token_number); + MUSCLE_INSERT_INT ("error_verbose", error_verbose); MUSCLE_INSERT_STRING ("prefix", spec_name_prefix ? spec_name_prefix : "yy"); /* FIXME: This is wrong: the muscles should decide whether they hold a copy or not, but the situation is too obscure currently. */ - MUSCLE_INSERT_STRING ("output-infix", output_infix ? output_infix : ""); - MUSCLE_INSERT_STRING ("output-prefix", short_base_name); + MUSCLE_INSERT_STRING ("output_infix", output_infix ? output_infix : ""); + MUSCLE_INSERT_STRING ("output_prefix", short_base_name); + MUSCLE_INSERT_STRING ("output_parser_name", parser_file_name); + MUSCLE_INSERT_STRING ("output_header_name", spec_defines_file); MUSCLE_INSERT_INT ("nnts", nvars); MUSCLE_INSERT_INT ("nrules", nrules); MUSCLE_INSERT_INT ("nstates", nstates); MUSCLE_INSERT_INT ("ntokens", ntokens); - MUSCLE_INSERT_INT ("locations-flag", locations_flag); -} - -/*-------------------------. -| Output the header file. | -`-------------------------*/ + MUSCLE_INSERT_INT ("locations_flag", locations_flag); + MUSCLE_INSERT_INT ("defines_flag", defines_flag); -static void -header_output (void) -{ - size_t dummy_line; - FILE *out = xfopen (spec_defines_file, "w"); - char *macro_name = compute_header_macro (); - - fprintf (out, "#ifndef %s\n", macro_name); - fprintf (out, "# define %s\n\n", macro_name); - - token_definitions_output (out, &dummy_line); - fprintf (out, "\ -#ifndef YYSTYPE\n\ -typedef %s -yystype;\n\ -# define YYSTYPE yystype\n\ -#endif\n", - muscle_find ("stype")); - - if (!pure_parser) - fprintf (out, "\nextern YYSTYPE %slval;\n", - spec_name_prefix ? spec_name_prefix : "yy"); - - if (locations_flag) - { - fputs ("\n\n", out); - fprintf (out, "\ -#ifndef YYLTYPE\n\ -typedef struct yyltype\n\ -{\n\ - int first_line;\n\ - int first_column;\n\ - int last_line;\n\ - int last_column;\n\ -} yyltype;\n\ -# define YYLTYPE yyltype\n\ -#endif\n"); - if (!pure_parser) - fprintf (out, "\nextern YYLTYPE %slloc;\n", - spec_name_prefix ? spec_name_prefix : "yy"); - } + /* Copy definitions in directive. */ + obstack_1grow (&pre_prologue_obstack, 0); + obstack_1grow (&post_prologue_obstack, 0); + muscle_insert ("pre_prologue", obstack_finish (&pre_prologue_obstack)); + muscle_insert ("post_prologue", obstack_finish (&post_prologue_obstack)); - if (semantic_parser) + /* Find the right skeleton file. */ + if (!skeleton) { - int i; - - for (i = ntokens; i < nsyms; i++) - /* don't make these for dummy nonterminals made by gensym. */ - if (*symbols[i]->tag != '@') - fprintf (out, "# define NT%s\t%d\n", symbols[i]->tag, i); + if (glr_parser) + skeleton = "glr.c"; + else + skeleton = "yacc.c"; } - fprintf (out, "\n#endif /* not %s */\n", macro_name); - free (macro_name); - xfclose (out); + /* Parse the skeleton file and output the needed parsers. */ + muscle_insert ("skeleton", skeleton); } @@ -1079,30 +1308,17 @@ output (void) { obstack_init (&format_obstack); - output_token_translations (); - output_gram (); - - XFREE (ritem); - if (semantic_parser) - output_stos (); - output_rule_data (); + prepare_tokens (); + prepare_rules (); + prepare_states (); output_actions (); prepare (); - /* Copy definitions in directive. */ - obstack_1grow (&attrs_obstack, 0); - muscle_insert ("prologue", obstack_finish (&attrs_obstack)); /* Process the selected skeleton file. */ output_skeleton (); - /* Output the header if needed. */ - if (defines_flag) - header_output (); - - free (rules + 1); - obstack_free (&muscle_obstack, NULL); obstack_free (&format_obstack, NULL); - obstack_free (&action_obstack, NULL); - obstack_free (&attrs_obstack, NULL); + obstack_free (&pre_prologue_obstack, NULL); + obstack_free (&post_prologue_obstack, NULL); }