X-Git-Url: https://git.saurik.com/bison.git/blobdiff_plain/1509d42fcb8fb8a78d46f0c26b863cafd347bee5..ccdc1577ef7fb32a5e30cb655337e7cd42b98987:/src/output.c?ds=sidebyside diff --git a/src/output.c b/src/output.c index 9626bdeb..d6560970 100644 --- a/src/output.c +++ b/src/output.c @@ -1,237 +1,49 @@ -/* Output the generated parsing program for bison, - Copyright (C) 1984, 1986, 1989, 1992, 2000, 2001, 2002 - Free Software Foundation, Inc. +/* Output the generated parsing program for Bison. + + Copyright (C) 1984, 1986, 1989, 1992, 2000-2012 Free Software + Foundation, Inc. This file is part of Bison, the GNU Compiler Compiler. - Bison is free software; you can redistribute it and/or modify it - under the terms of the GNU General Public License as published by - the Free Software Foundation; either version 2, or (at your option) - any later version. + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. - Bison is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. You should have received a copy of the GNU General Public License - along with Bison; see the file COPYING. If not, write to the Free - Software Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA - 02111-1307, USA. */ - - -/* The parser tables consist of these tables. - - YYTRANSLATE = vector mapping yylex's token numbers into bison's - token numbers. - - YYTNAME = vector of string-names indexed by bison token number. - - YYTOKNUM = vector of yylex token numbers corresponding to entries - in YYTNAME. - - YYRLINE = vector of line-numbers of all rules. For yydebug - printouts. - - YYRHS = vector of items of all rules. This is exactly what RITEMS - contains. For yydebug and for semantic parser. - - YYPRHS[R] = index in YYRHS of first item for rule R. - - YYR1[R] = symbol number of symbol that rule R derives. - - YYR2[R] = number of symbols composing right hand side of rule R. - - YYSTOS[S] = the symbol number of the symbol that leads to state S. - - YYDEFACT[S] = default rule to reduce with in state s, when YYTABLE - doesn't specify something else to do. Zero means the default is an - error. - - YYDEFGOTO[I] = default state to go to after a reduction of a rule - that generates variable NTOKENS + I, except when YYTABLE specifies - something else to do. - - YYPACT[S] = index in YYTABLE of the portion describing state S. - The lookahead token's type is used to index that portion to find - out what to do. - - If the value in YYTABLE is positive, we shift the token and go to - that state. - - If the value is negative, it is minus a rule number to reduce by. - - If the value is zero, the default action from YYDEFACT[S] is used. - - YYPGOTO[I] = the index in YYTABLE of the portion describing what to - do after reducing a rule that derives variable I + NTOKENS. This - portion is indexed by the parser state number, S, as of before the - text for this nonterminal was read. The value from YYTABLE is the - state to go to if the corresponding value in YYCHECK is S. - - YYTABLE = a vector filled with portions for different uses, found - via YYPACT and YYPGOTO. + along with this program. If not, see . */ - YYCHECK = a vector indexed in parallel with YYTABLE. It indicates, - in a roundabout way, the bounds of the portion you are trying to - examine. - - Suppose that the portion of YYTABLE starts at index P and the index - to be examined within the portion is I. Then if YYCHECK[P+I] != I, - I is outside the bounds of what is actually allocated, and the - default (from YYDEFACT or YYDEFGOTO) should be used. Otherwise, - YYTABLE[P+I] should be used. +#include +#include "system.h" - YYFINAL = the state number of the termination state. YYFLAG = most - negative short int. Used to flag ?? */ +#include +#include +#include +#include +#include +#include +#include +#include -#include "system.h" -#include "bitsetv.h" -#include "quotearg.h" -#include "error.h" -#include "getargs.h" +#include "complain.h" #include "files.h" +#include "getargs.h" #include "gram.h" -#include "LR0.h" -#include "complain.h" +#include "muscle-tab.h" #include "output.h" -#include "lalr.h" #include "reader.h" +#include "scan-code.h" /* max_left_semantic_context */ +#include "scan-skel.h" #include "symtab.h" -#include "conflicts.h" -#include "muscle_tab.h" - -/* From src/scan-skel.l. */ -void m4_invoke PARAMS ((const char *definitions)); - - -/* Several tables will be indexed both by state and nonterminal - numbers. We call `vector' such a thing (= either a state or a - symbol number. - - Of course vector_number_t ought to be wide enough to contain - state_number_t and symbol_number_t. */ -typedef short vector_number_t; -#define VECTOR_NUMBER_MAX ((vector_number_t) SHRT_MAX) -#define VECTOR_NUMBER_MIN ((vector_number_t) SHRT_MIN) -#define state_number_to_vector_number(State) \ - ((vector_number_t) State) -#define symbol_number_to_vector_number(Symbol) \ - ((vector_number_t) (state_number_as_int (nstates) + Symbol - ntokens)) - -static int nvectors; - - -/* FROMS and TOS are indexed by vector_number_t. - - If VECTOR is a nonterminal, (FROMS[VECTOR], TOS[VECTOR]) form an - array of state numbers of the non defaulted GOTO on VECTOR. - - If VECTOR is a state, TOS[VECTOR] is the array of actions to do on - the (array of) symbols FROMS[VECTOR]. - - In both cases, TALLY[VECTOR] is the size of the arrays - FROMS[VECTOR], TOS[VECTOR]; and WIDTH[VECTOR] = - (FROMS[VECTOR][SIZE] - FROMS[VECTOR][0] + 1) where SIZE = - TALLY[VECTOR]. - - FROMS therefore contains symbol_number_t and action_number_t, - TOS state_number_t and action_number_t, - TALLY sizes, - WIDTH differences of FROMS. - - Let base_t be the type of FROMS, TOS, and WIDTH. */ -typedef int base_t; -#define BASE_MAX ((base_t) INT_MAX) -#define BASE_MIN ((base_t) INT_MIN) - -static base_t **froms = NULL; -static base_t **tos = NULL; -static unsigned int **conflict_tos = NULL; -static short *tally = NULL; -static base_t *width = NULL; - - -/* For a given state, N = ACTROW[SYMBOL]: - - If N = 0, stands for `run the default action'. - If N = MIN, stands for `raise a parse error'. - If N > 0, stands for `shift SYMBOL and go to n'. - If N < 0, stands for `reduce -N'. */ -typedef short action_t; -#define ACTION_MAX ((action_t) SHRT_MAX) -#define ACTION_MIN ((action_t) SHRT_MIN) - -static action_t *actrow = NULL; - -/* FROMS and TOS are reordered to be compressed. ORDER[VECTOR] is the - new vector number of VECTOR. We skip `empty' vectors (i.e., - TALLY[VECTOR] = 0), and call these `entries'. */ -static vector_number_t *order = NULL; -static int nentries; - -static base_t *base = NULL; -/* A distinguished value of BASE, negative infinite. During the - computation equals to BASE_MIN, later mapped to BASE_NINF to - keep parser tables small. */ -base_t base_ninf = 0; -static base_t *pos = NULL; - -static unsigned int *conflrow = 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 more or less the original hard-coded value (which was - SHRT_MAX). */ -static size_t table_size = 32768; -static base_t *table = NULL; -static base_t *check = NULL; -/* The value used in TABLE to denote explicit parse errors - (%nonassoc), a negative infinite. First defaults to ACTION_MIN, - but in order to keep small tables, renumbered as TABLE_ERROR, which - is the smallest (non error) value minus 1. */ -base_t table_ninf = 0; -static int lowzero; -static int high; +#include "tables.h" static struct obstack format_obstack; -int error_verbose = 0; - - -/*----------------------------------------------------------------. -| 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. | -`----------------------------------------------------------------*/ - -static void -table_grow (size_t desired) -{ - size_t old_size = table_size; - - while (table_size <= desired) - table_size *= 2; - - if (trace_flag & trace_resource) - fprintf (stderr, "growing table and check from: %d to %d\n", - old_size, table_size); - - table = XREALLOC (table, base_t, table_size); - check = XREALLOC (check, base_t, table_size); - if (glr_parser) - conflict_table = XREALLOC (conflict_table, unsigned int, table_size); - - for (/* Nothing. */; old_size < table_size; ++old_size) - { - table[old_size] = 0; - check[old_size] = -1; - } -} - /*-------------------------------------------------------------------. | Create a function NAME which associates to the muscle NAME the | @@ -241,101 +53,134 @@ table_grow (size_t desired) `-------------------------------------------------------------------*/ -#define GENERATE_MUSCLE_INSERT_TABLE(Name, Type) \ - \ -static void \ -Name (const char *name, \ - Type *table_data, \ - Type first, \ - int begin, \ - int end) \ -{ \ - Type min = first; \ - 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] < min) \ - min = table_data[i]; \ - if (max < table_data[i]) \ - max = table_data[i]; \ - } \ - obstack_1grow (&format_obstack, 0); \ - muscle_insert (name, obstack_finish (&format_obstack)); \ - \ - /* Build `NAME_min' and `NAME_max' in the obstack. */ \ - obstack_fgrow1 (&format_obstack, "%s_min", name); \ - obstack_1grow (&format_obstack, 0); \ - MUSCLE_INSERT_LONG_INT (obstack_finish (&format_obstack), \ - (long int) min); \ - obstack_fgrow1 (&format_obstack, "%s_max", name); \ - obstack_1grow (&format_obstack, 0); \ - MUSCLE_INSERT_LONG_INT (obstack_finish (&format_obstack), \ - (long int) max); \ +#define GENERATE_MUSCLE_INSERT_TABLE(Name, Type) \ + \ +static void \ +Name (char const *name, \ + Type *table_data, \ + Type first, \ + int begin, \ + int end) \ +{ \ + Type min = first; \ + Type max = first; \ + long int lmin; \ + long int lmax; \ + 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] < min) \ + min = table_data[i]; \ + if (max < table_data[i]) \ + max = table_data[i]; \ + } \ + obstack_1grow (&format_obstack, 0); \ + muscle_insert (name, obstack_finish (&format_obstack)); \ + \ + lmin = min; \ + lmax = max; \ + /* Build `NAME_min' and `NAME_max' in the obstack. */ \ + obstack_fgrow1 (&format_obstack, "%s_min", name); \ + obstack_1grow (&format_obstack, 0); \ + MUSCLE_INSERT_LONG_INT (obstack_finish (&format_obstack), lmin); \ + obstack_fgrow1 (&format_obstack, "%s_max", name); \ + obstack_1grow (&format_obstack, 0); \ + MUSCLE_INSERT_LONG_INT (obstack_finish (&format_obstack), lmax); \ } GENERATE_MUSCLE_INSERT_TABLE(muscle_insert_unsigned_int_table, unsigned int) GENERATE_MUSCLE_INSERT_TABLE(muscle_insert_int_table, int) -GENERATE_MUSCLE_INSERT_TABLE(muscle_insert_short_table, short) -GENERATE_MUSCLE_INSERT_TABLE(muscle_insert_base_table, base_t) -GENERATE_MUSCLE_INSERT_TABLE(muscle_insert_rule_number_table, rule_number_t) -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) +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_state_number_table, state_number) + + +/*--------------------------------------------------------------------. +| Print to OUT a representation of STRING escaped both for C and M4. | +`--------------------------------------------------------------------*/ + +static void +escaped_output (FILE *out, char const *string) +{ + char const *p; + fprintf (out, "[["); + + for (p = quotearg_style (c_quoting_style, string); *p; p++) + switch (*p) + { + case '$': fputs ("$][", out); break; + case '@': fputs ("@@", out); break; + case '[': fputs ("@{", out); break; + case ']': fputs ("@}", out); break; + default: fputc (*p, out); break; + } + fprintf (out, "]]"); +} -/*-----------------------------------------------------------------. -| Prepare the muscles related to the tokens: translate, tname, and | -| toknum. | -`-----------------------------------------------------------------*/ + +/*------------------------------------------------------------------. +| Prepare the muscles related to the symbols: translate, tname, and | +| toknum. | +`------------------------------------------------------------------*/ static void -prepare_tokens (void) +prepare_symbols (void) { + 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); + muscle_insert_symbol_number_table ("translate", - token_translations, - token_translations[0], - 1, max_user_token_number + 1); + token_translations, + token_translations[0], + 1, max_user_token_number + 1); + /* tname -- token names. */ { int i; - int j = 0; + /* We assume that the table will be output starting at column 2. */ + int j = 2; + struct quoting_options *qo = clone_quoting_options (0); + set_quoting_style (qo, c_quoting_style); + set_quoting_flags (qo, QA_SPLIT_TRIGRAPHS); 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, - 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; + char *cp = quotearg_alloc (symbols[i]->tag, -1, qo); + /* Width of the next token, including the two quotes, the + comma and the space. */ + int width = strlen (cp) + 2; + + if (j + width > 75) + { + obstack_sgrow (&format_obstack, "\n "); + j = 1; + } + + if (i) + obstack_1grow (&format_obstack, ' '); + MUSCLE_OBSTACK_SGROW (&format_obstack, cp); + free (cp); + obstack_1grow (&format_obstack, ','); + j += width; } - /* Add a NULL entry to list of tokens (well, 0, as NULL might not be - defined). */ - obstack_sgrow (&format_obstack, "0"); + free (qo); + obstack_sgrow (&format_obstack, " ]b4_null["); /* Finish table and store. */ obstack_1grow (&format_obstack, 0); @@ -345,72 +190,64 @@ prepare_tokens (void) /* Output YYTOKNUM. */ { int i; - int *values = XCALLOC (int, ntokens); + int *values = xnmalloc (ntokens, sizeof *values); for (i = 0; i < ntokens; ++i) values[i] = symbols[i]->user_token_number; muscle_insert_int_table ("toknum", values, - values[0], 1, ntokens); + values[0], 1, ntokens); free (values); } } -/*-------------------------------------------------------------. -| 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_t r; - unsigned int i = 0; - item_number_t *rhs = XMALLOC (item_number_t, nritems); - unsigned int *prhs = XMALLOC (unsigned int, nrules); - unsigned int *rline = XMALLOC (unsigned int, nrules); - symbol_number_t *r1 = XMALLOC (symbol_number_t, nrules); - unsigned int *r2 = XMALLOC (unsigned int, nrules); - short *dprec = XMALLOC (short, nrules); - short *merger = XMALLOC (short, nrules); - + 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_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; + r2[r] = rule_rhs_length(&rules[r]); /* Line where rule was defined. */ - rline[r] = rules[r].location.first_line; - /* Dynamic precedence (GLR) */ + rline[r] = rules[r].location.start.line; + /* Dynamic precedence (GLR). */ dprec[r] = rules[r].dprec; - /* Merger-function index (GLR) */ + /* Merger-function index (GLR). */ merger[r] = rules[r].merger; + /* Immediate reduction flags (GLR). */ + immediate[r] = rules[r].is_predicate; } - assert (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_short_table ("dprec", dprec, 0, 0, nrules); - muscle_insert_short_table ("merger", merger, 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); free (dprec); free (merger); + free (immediate); } /*--------------------------------------------. @@ -420,346 +257,113 @@ prepare_rules (void) static void prepare_states (void) { - state_number_t i; - symbol_number_t *values = - (symbol_number_t *) alloca (sizeof (symbol_number_t) * nstates); + state_number i; + symbol_number *values = xnmalloc (nstates, sizeof *values); for (i = 0; i < nstates; ++i) values[i] = states[i]->accessing_symbol; muscle_insert_symbol_number_table ("stos", values, - 0, 1, nstates); + 0, 1, nstates); + free (values); + + MUSCLE_INSERT_INT ("last", high); + MUSCLE_INSERT_INT ("final_state_number", final_state->number); + MUSCLE_INSERT_INT ("states_number", 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. | -`-------------------------------------------------------------------*/ +/*-------------------------------------------------------. +| Compare two symbols by type-name, and then by number. | +`-------------------------------------------------------*/ -static void -conflict_row (state_t *state) +static int +symbol_type_name_cmp (const symbol **lhs, const symbol **rhs) { - int i, j; + int res = UNIQSTR_CMP((*lhs)->type_name, (*rhs)->type_name); + if (res) + return res; + return (*lhs)->number - (*rhs)->number; +} - if (! glr_parser) - return; - for (j = 0; j < ntokens; j += 1) - if (conflrow[j]) - { - conflrow[j] = conflict_list_cnt; - - /* 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] - != rule_number_as_item_number (state->lookaheads_rule[i]->number))) - { - assert (conflict_list_free > 0); - conflict_list[conflict_list_cnt] - = state->lookaheads_rule[i]->number + 1; - conflict_list_cnt += 1; - conflict_list_free -= 1; - } - - /* Leave a 0 at the end. */ - assert (conflict_list_free > 0); - conflict_list_cnt += 1; - conflict_list_free -= 1; - } +/*----------------------------------------------------------------. +| 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; } /*------------------------------------------------------------------. -| 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 | -| ACTION_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. | +| Define b4_type_names, which is a list of (lists of the numbers of | +| symbols with same type-name). | `------------------------------------------------------------------*/ -static rule_t * -action_row (state_t *state) +static void +type_names_output (FILE *out) { int i; - rule_t *default_rule = NULL; - reductions_t *redp = state->reductions; - transitions_t *transitions = state->transitions; - errs_t *errp = state->errs; - /* Set to nonzero to inhibit having any default reduction. */ - int nodefault = 0; - int conflicted = 0; - - for (i = 0; i < ntokens; i++) - actrow[i] = conflrow[i] = 0; - - if (redp->num >= 1) + symbol **syms = symbols_by_type_name (); + fputs ("m4_define([b4_type_names],\n[", out); + for (i = 0; i < nsyms; /* nothing */) { - int j; - bitset_iterator biter; - /* loop over all the rules available here which require - lookahead */ - for (i = state->nlookaheads - 1; i >= 0; --i) - /* and find each token which the rule finds acceptable - to come next */ - BITSET_FOR_EACH (biter, state->lookaheads[i], j, 0) - { - /* and record this rule as the rule to use if that - token follows. */ - if (actrow[j] != 0) - conflicted = conflrow[j] = 1; - actrow[j] = rule_number_as_item_number (state->lookaheads_rule[i]->number); - } + // 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); } - - /* 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_EACH_SHIFT (transitions, i) - { - symbol_number_t symbol = TRANSITION_SYMBOL (transitions, i); - state_t *shift_state = transitions->states[i]; - - if (actrow[symbol] != 0) - conflicted = conflrow[symbol] = 1; - actrow[symbol] = state_number_as_int (shift_state->number); - - /* 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 ACTION_MIN as the action. */ - for (i = 0; i < errp->num; i++) - { - symbol_t *symbol = errp->symbols[i]; - actrow[symbol->number] = ACTION_MIN; - } - - /* Now find the most common reduction and make it the default action - for this state. */ - - if (redp->num >= 1 && !nodefault) - { - if (state->consistent) - default_rule = redp->rules[0]; - else - { - int max = 0; - for (i = 0; i < state->nlookaheads; i++) - { - int count = 0; - rule_t *rule = state->lookaheads_rule[i]; - symbol_number_t j; - - for (j = 0; j < ntokens; j++) - if (actrow[j] == rule_number_as_item_number (rule->number)) - count++; - - if (count > max) - { - max = count; - default_rule = rule; - } - } - - /* 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] == rule_number_as_item_number (default_rule->number) - && ! (glr_parser && conflrow[j])) - actrow[j] = 0; - } - } - } - - /* Find the rules which are reduced. */ - if (!glr_parser) - { - for (i = 0; i < ntokens; i++) - if (actrow[i] < 0 && actrow[i] != ACTION_MIN) - rules[item_number_as_rule_number (actrow[i])].useful = TRUE; - if (default_rule) - default_rule->useful = TRUE; - } - - /* If have no default rule, the default is an error. - So replace any action which says "error" with "use default". */ - - if (!default_rule) - for (i = 0; i < ntokens; i++) - if (actrow[i] == ACTION_MIN) - actrow[i] = 0; - - if (conflicted) - conflict_row (state); - - return default_rule; + fputs ("])\n\n", out); + free (syms); } -/*--------------------------------------------. -| Set FROMS, TOS, TALLY and WIDTH for STATE. | -`--------------------------------------------*/ +/*-------------------------------------. +| The list of all the symbol numbers. | +`-------------------------------------*/ static void -save_row (state_number_t state) +symbol_numbers_output (FILE *out) { - symbol_number_t i; - int count; - base_t *sp = NULL; - base_t *sp1 = NULL; - base_t *sp2 = NULL; - unsigned int *sp3 = NULL; - - /* Number of non default actions in STATE. */ - count = 0; - for (i = 0; i < ntokens; i++) - if (actrow[i] != 0) - count++; - - if (count == 0) - return; - - /* Allocate non defaulted actions. */ - froms[state] = sp1 = sp = XCALLOC (base_t, count); - tos[state] = sp2 = XCALLOC (base_t, count); - if (glr_parser) - conflict_tos[state] = sp3 = XCALLOC (unsigned int, count); - else - conflict_tos[state] = NULL; - - /* Store non defaulted actions. */ - for (i = 0; i < ntokens; i++) - if (actrow[i] != 0) - { - *sp1++ = i; - *sp2++ = actrow[i]; - if (glr_parser) - *sp3++ = conflrow[i]; - } - - tally[state] = count; - width[state] = sp1[-1] - sp[0] + 1; + 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); } -/*------------------------------------------------------------------. -| Figure out the actions for the specified state, indexed by | -| lookahead token type. | -| | -| The YYDEFACT table is output now. The detailed info is saved for | -| putting into YYTABLE later. | -`------------------------------------------------------------------*/ +/*---------------------------------. +| Output the user actions to OUT. | +`---------------------------------*/ static void -token_actions (void) -{ - state_number_t i; - rule_number_t r; - int nconflict = conflicts_total_count (); - - rule_number_t *yydefact = XCALLOC (rule_number_t, nstates); - - actrow = XCALLOC (action_t, ntokens); - conflrow = XCALLOC (unsigned int, ntokens); - - /* Now that the parser was computed, we can find which rules are - really reduced, and which are not because of SR or RR conflicts. - */ - if (!glr_parser) - for (r = 0; r < nrules; ++r) - rules[r].useful = FALSE; - - 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) - { - rule_t *default_rule = action_row (states[i]); - yydefact[i] = default_rule ? default_rule->number + 1 : 0; - save_row (i); - } - - muscle_insert_rule_number_table ("defact", yydefact, - yydefact[0], 1, nstates); - - if (!glr_parser) - for (r = 0; r < nrules ; ++r) - if (!rules[r].useful) - { - LOCATION_PRINT (stderr, rules[r].location); - fprintf (stderr, ": %s: %s: ", - _("warning"), _("rule never reduced because of conflicts")); - rule_print (&rules[r], stderr); - } - - XFREE (actrow); - XFREE (conflrow); - XFREE (yydefact); -} - - -/*-----------------------------. -| Output the actions to OOUT. | -`-----------------------------*/ - -void -actions_output (FILE *out) +user_actions_output (FILE *out) { - rule_number_t r; + rule_number r; - fputs ("m4_define([b4_actions], \n[[", out); + fputs ("m4_define([b4_actions], \n[", out); for (r = 0; r < nrules; ++r) if (rules[r].action) { - fprintf (out, " case %d:\n", r + 1); - - if (!no_lines_flag) - fprintf (out, muscle_find ("linef"), - rules[r].action_location.first_line, - quotearg_style (c_quoting_style, - muscle_find ("filename"))); - fprintf (out, " %s\n break;\n\n", - rules[r].action); + 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); + fputs ("])\n\n", out); } -/*--------------------------------------. -| Output the merge functions to OUT. | -`--------------------------------------*/ +/*------------------------------------. +| Output the merge functions to OUT. | +`------------------------------------*/ static void merger_output (FILE *out) @@ -771,571 +375,198 @@ merger_output (FILE *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); + fprintf (out, " case %d: *yy0 = %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); + fprintf (out, " case %d: yy0->%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) -{ - int i; - int first = 1; - - fputs ("m4_define([b4_tokens], \n[", out); - for (i = 0; i < ntokens; ++i) - { - symbol_t *symbol = symbols[i]; - int number = symbol->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. */ - assert (number != USER_NUMBER_ALIAS); - - /* Skip error token. */ - 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; - - /* 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, "%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, - symbol->number, - symbol->destructor, - symbol->type_name); - - first = 0; - } - fputs ("])\n\n", out); -} - -/*-------------------------------------. -| Output the symbol printers to OOUT. | -`-------------------------------------*/ +/*---------------------------------------------. +| Prepare the muscles for symbol definitions. | +`---------------------------------------------*/ static void -symbol_printers_output (FILE *out) +prepare_symbol_definitions (void) { 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, - symbol->number, - symbol->printer, - symbol->type_name); - - first = 0; - } - fputs ("])\n\n", out); -} + { + symbol *sym = symbols[i]; + 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); -/*------------------------------------------------------------------. -| Compute FROMS[VECTOR], TOS[VECTOR], TALLY[VECTOR], WIDTH[VECTOR], | -| i.e., the information related to non defaulted GOTO on the nterm | -| SYMBOL. | -| | -| DEFAULT_STATE is the principal destination on SYMBOL, i.e., the | -| default GOTO destination on SYMBOL. | -`------------------------------------------------------------------*/ +#define SET_KEY2(Entry, Suffix) \ + obstack_fgrow3 (&format_obstack, "symbol(%d, %s_%s)", \ + i, Entry, Suffix); \ + obstack_1grow (&format_obstack, 0); \ + key = obstack_finish (&format_obstack); -static void -save_column (symbol_number_t symbol, state_number_t default_state) -{ - int i; - base_t *sp; - base_t *sp1; - base_t *sp2; - int count; - vector_number_t symno = symbol_number_to_vector_number (symbol); - - goto_number_t begin = goto_map[symbol]; - goto_number_t end = goto_map[symbol + 1]; - - /* Number of non default GOTO. */ - count = 0; - for (i = begin; i < end; i++) - if (to_state[i] != default_state) - count++; - - if (count == 0) - return; - - /* Allocate room for non defaulted gotos. */ - froms[symno] = sp1 = sp = XCALLOC (base_t, count); - tos[symno] = sp2 = XCALLOC (base_t, count); - - /* Store the state numbers of the non defaulted gotos. */ - for (i = begin; i < end; i++) - if (to_state[i] != default_state) - { - *sp1++ = from_state[i]; - *sp2++ = to_state[i]; - } + // Whether the symbol has an identifier. + value = symbol_id_get (sym); + SET_KEY("has_id"); + MUSCLE_INSERT_INT (key, !!value); - tally[symno] = count; - width[symno] = sp1[-1] - sp[0] + 1; -} + // 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); -/*----------------------------------------------------------------. -| Return `the' most common destination GOTO on SYMBOL (a nterm). | -`----------------------------------------------------------------*/ + SET_KEY("user_number"); + MUSCLE_INSERT_INT (key, sym->user_token_number); -static state_number_t -default_goto (symbol_number_t symbol, short state_count[]) -{ - state_number_t s; - int i; - goto_number_t m = goto_map[symbol]; - goto_number_t n = goto_map[symbol + 1]; - state_number_t default_state = (state_number_t) -1; - int max = 0; + SET_KEY("is_token"); + MUSCLE_INSERT_INT (key, + i < ntokens && sym != errtoken && sym != undeftoken); - if (m == n) - return (state_number_t) -1; + SET_KEY("number"); + MUSCLE_INSERT_INT (key, sym->number); - for (s = 0; s < nstates; s++) - state_count[s] = 0; + SET_KEY("has_type"); + MUSCLE_INSERT_INT (key, !!sym->type_name); - for (i = m; i < n; i++) - state_count[to_state[i]]++; + SET_KEY("type"); + MUSCLE_INSERT_STRING (key, sym->type_name ? sym->type_name : ""); - for (s = 0; s < nstates; s++) - if (state_count[s] > max) { - max = state_count[s]; - default_state = s; + int j; + for (j = 0; j < CODE_PROPS_SIZE; ++j) + { + /* "printer", not "%printer". */ + char const *pname = code_props_type_string (j) + 1; + code_props const *p = symbol_code_props_get (sym, j); + SET_KEY2("has", pname); + MUSCLE_INSERT_INT (key, !!p->code); + + if (p->code) + { + SET_KEY2(pname, "file"); + MUSCLE_INSERT_STRING (key, p->location.start.file); + + SET_KEY2(pname, "line"); + MUSCLE_INSERT_INT (key, p->location.start.line); + + SET_KEY(pname); + MUSCLE_INSERT_STRING_RAW (key, p->code); + } + } } - - return default_state; -} - - -/*-------------------------------------------------------------------. -| Figure out what to do after reducing with each rule, depending on | -| the saved state from before the beginning of parsing the data that | -| matched this rule. | -| | -| The YYDEFGOTO table is output now. The detailed info is saved for | -| putting into YYTABLE later. | -`-------------------------------------------------------------------*/ - -static void -goto_actions (void) -{ - symbol_number_t i; - state_number_t *yydefgoto = XMALLOC (state_number_t, nvars); - - /* For a given nterm I, STATE_COUNT[S] is the number of times there - is a GOTO to S on I. */ - short *state_count = XCALLOC (short, nstates); - for (i = ntokens; i < nsyms; ++i) - { - state_number_t default_state = default_goto (i, state_count); - save_column (i, default_state); - yydefgoto[i - ntokens] = default_state; +#undef SET_KEY2 +#undef SET_KEY } - - muscle_insert_state_number_table ("defgoto", yydefgoto, - yydefgoto[0], 1, nsyms - ntokens); - XFREE (state_count); - XFREE (yydefgoto); } -/*------------------------------------------------------------------. -| Compute ORDER, a reordering of vectors, in order to decide how to | -| pack the actions and gotos information into yytable. | -`------------------------------------------------------------------*/ +/*--------------------------------------. +| Output the tokens definition to OUT. | +`--------------------------------------*/ static void -sort_actions (void) +token_definitions_output (FILE *out) { int i; + char const *sep = ""; - nentries = 0; - - for (i = 0; i < nvectors; i++) - if (tally[i] > 0) - { - int k; - int t = tally[i]; - int w = width[i]; - int j = nentries - 1; - - while (j >= 0 && (width[order[j]] < w)) - j--; - - while (j >= 0 && (width[order[j]] == w) && (tally[order[j]] < t)) - j--; - - for (k = nentries - 1; k > j; k--) - order[k + 1] = order[k]; - - order[j + 1] = i; - nentries++; - } -} - - -/* If VECTOR is a state which actions (reflected by FROMS, TOS, TALLY - and WIDTH of VECTOR) are common to a previous state, return this - state number. - - In any other case, return -1. */ - -static state_number_t -matching_state (vector_number_t vector) -{ - vector_number_t i = order[vector]; - int t; - int w; - int prev; - - /* If VECTOR is a nterm, return -1. */ - if (i >= (int) nstates) - return -1; - - t = tally[i]; - w = width[i]; - - for (prev = vector - 1; prev >= 0; prev--) - { - vector_number_t j = order[prev]; - int k; - int match = 1; - - /* Given how ORDER was computed, if the WIDTH or TALLY is - different, there cannot be a matching state. */ - if (width[j] != w || tally[j] != t) - return -1; - - for (k = 0; match && k < t; k++) - if (tos[j][k] != tos[i][k] || froms[j][k] != froms[i][k]) - match = 0; - - if (match) - return j; - } - - return -1; -} - - -static base_t -pack_vector (vector_number_t vector) -{ - vector_number_t i = order[vector]; - int j; - int t = tally[i]; - int loc = 0; - base_t *from = froms[i]; - base_t *to = tos[i]; - unsigned int *conflict_to = conflict_tos[i]; - - assert (t); - - for (j = lowzero - from[0]; j < (int) table_size; j++) + fputs ("m4_define([b4_tokens], \n[", out); + for (i = 0; i < ntokens; ++i) { - int k; - int ok = 1; - - for (k = 0; ok && k < t; k++) - { - loc = j + state_number_as_int (from[k]); - if (loc > (int) table_size) - table_grow (loc); - - if (table[loc] != 0) - ok = 0; - } - - for (k = 0; ok && k < vector; k++) - if (pos[k] == j) - ok = 0; - - if (ok) - { - for (k = 0; k < t; k++) - { - loc = j + from[k]; - table[loc] = to[k]; - if (glr_parser && conflict_to != NULL) - conflict_table[loc] = conflict_to[k]; - check[loc] = from[k]; - } - - while (table[lowzero] != 0) - lowzero++; - - if (loc > high) - high = loc; - - if (j < BASE_MIN || BASE_MAX < j) - fatal ("base_t too small to hold %d\n", j); - return j; - } + symbol *sym = symbols[i]; + 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) + { + fprintf (out, "%s[[[%s]], %d]", + sep, id, number); + sep = ",\n"; + } } -#define pack_vector_succeeded 0 - assert (pack_vector_succeeded); - return 0; + fputs ("])\n\n", out); } -/*-------------------------------------------------------------. -| Remap the negative infinite in TAB from NINF to the greatest | -| possible smallest value. Return it. | -| | -| In most case this allows us to use shorts instead of ints in | -| parsers. | -`-------------------------------------------------------------*/ - -static base_t -table_ninf_remap (base_t tab[], size_t size, base_t ninf) -{ - base_t res = 0; - size_t i; - - for (i = 0; i < size; i++) - if (tab[i] < res && tab[i] != ninf) - res = base[i]; - - --res; - - for (i = 0; i < size; i++) - if (tab[i] == ninf) - tab[i] = res; - - return res; -} - static void -pack_table (void) +prepare_actions (void) { - int i; - - base = XCALLOC (base_t, nvectors); - pos = XCALLOC (base_t, nentries); - table = XCALLOC (base_t, table_size); - if (glr_parser) - conflict_table = XCALLOC (unsigned int, table_size); - check = XCALLOC (base_t, table_size); - - lowzero = 0; - high = 0; - - for (i = 0; i < nvectors; i++) - base[i] = BASE_MIN; - - for (i = 0; i < (int) table_size; i++) - check[i] = -1; - - for (i = 0; i < nentries; i++) - { - state_number_t state = matching_state (i); - base_t place; - - if (state < 0) - /* A new set of state actions, or a nonterminal. */ - place = pack_vector (i); - else - /* Action of I were already coded for STATE. */ - place = base[state]; - - pos[i] = place; - base[order[i]] = place; - } - - /* Use the greatest possible negative infinites. */ - base_ninf = table_ninf_remap (base, nvectors, BASE_MIN); - table_ninf = table_ninf_remap (table, high + 1, ACTION_MIN); - - for (i = 0; i < nvectors; i++) - { - XFREE (froms[i]); - XFREE (tos[i]); - XFREE (conflict_tos[i]); - } + /* Figure out the actions for the specified state, indexed by + lookahead token type. */ - free (froms); - free (tos); - free (conflict_tos); - free (pos); -} + muscle_insert_rule_number_table ("defact", yydefact, + yydefact[0], 1, nstates); + /* Figure out what to do after reducing with each rule, depending on + the saved state from before the beginning of parsing the data + that matched this rule. */ + muscle_insert_state_number_table ("defgoto", yydefgoto, + yydefgoto[0], 1, nsyms - ntokens); -/* the following functions output yytable, yycheck, yyconflp, yyconfl, - and the vectors whose elements index the portion starts. */ -static void -output_base (void) -{ /* Output PACT. */ muscle_insert_base_table ("pact", base, - base[0], 1, nstates); + base[0], 1, nstates); MUSCLE_INSERT_INT ("pact_ninf", base_ninf); /* Output PGOTO. */ muscle_insert_base_table ("pgoto", base, - base[nstates], nstates + 1, nvectors); - XFREE (base); -} + base[nstates], nstates + 1, nvectors); - -static void -output_table (void) -{ muscle_insert_base_table ("table", table, - table[0], 1, high + 1); + table[0], 1, high + 1); MUSCLE_INSERT_INT ("table_ninf", table_ninf); - XFREE (table); -} + muscle_insert_base_table ("check", check, + check[0], 1, high + 1); -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; + /* 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 could avoid accidents by not writing them out in + that case. Nevertheless, it seems even better to be able to use + the GLR skeletons even without the non-deterministic tables. */ muscle_insert_unsigned_int_table ("conflict_list_heads", conflict_table, - conflict_table[0], 1, high+1); + 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) -{ - muscle_insert_base_table ("check", check, - check[0], 1, high + 1); - XFREE (check); + 0, 1, conflict_list_cnt); } -/*-----------------------------------------------------------------. -| Compute and output yydefact, yydefgoto, yypact, yypgoto, yytable | -| and yycheck. | -`-----------------------------------------------------------------*/ +/*--------------------------------------------. +| Output the definitions of all the muscles. | +`--------------------------------------------*/ static void -prepare_actions (void) +muscles_output (FILE *out) { - /* 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 (base_t *, nvectors); - tos = XCALLOC (base_t *, nvectors); - conflict_tos = XCALLOC (unsigned int *, nvectors); - tally = XCALLOC (short, nvectors); - width = XCALLOC (base_t, nvectors); - - token_actions (); - bitsetv_free (LA); - free (LArule); - - goto_actions (); - XFREE (goto_map + ntokens); - XFREE (from_state); - XFREE (to_state); - - order = XCALLOC (vector_number_t, nvectors); - sort_actions (); - pack_table (); - free (order); - - free (tally); - free (width); - - output_base (); - output_table (); - output_conflicts (); - - output_check (); + 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); } - /*---------------------------. | Call the skeleton parser. | @@ -1344,101 +575,144 @@ prepare_actions (void) static void output_skeleton (void) { - /* 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); + int filter_fd[2]; + pid_t pid; + + /* Compute the names of the package data dir and skeleton files. */ + char const *m4 = (m4 = getenv ("M4")) ? m4 : M4; + char const *datadir = pkgdatadir (); + char *m4sugar = xconcatenated_filename (datadir, "m4sugar/m4sugar.m4", NULL); + char *m4bison = xconcatenated_filename (datadir, "bison.m4", NULL); + char *skel = (IS_PATH_WITH_DIR (skeleton) + ? xstrdup (skeleton) + : xconcatenated_filename (datadir, skeleton, NULL)); + + /* Test whether m4sugar.m4 is readable, to check for proper + installation. A faulty installation can cause deadlock, so a + cheap sanity check is worthwhile. */ + xfclose (xfopen (m4sugar, "r")); + + /* Create an m4 subprocess connected to us via two pipes. */ + + if (trace_flag & trace_tools) + fprintf (stderr, "running: %s %s - %s %s\n", + m4, m4sugar, m4bison, skel); + + /* Some future version of GNU M4 (most likely 1.6) may treat the -dV in a + position-dependent manner. Keep it as the first argument so that all + files are traced. + + See the thread starting at + + for details. */ + { + char const *argv[10]; + 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 + + for details. */ + if (*M4_GNU_OPTION) + argv[i++] = M4_GNU_OPTION; + + argv[i++] = "-I"; + argv[i++] = datadir; + if (trace_flag & trace_m4) + argv[i++] = "-dV"; + argv[i++] = m4sugar; + argv[i++] = "-"; + argv[i++] = m4bison; + argv[i++] = skel; + argv[i++] = NULL; + aver (i <= ARRAY_CARDINALITY (argv)); + + /* 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); + } - muscles_m4_output (out); + free (m4sugar); + free (m4bison); + free (skel); - fputs ("m4_wrap([m4_divert_pop(0)])\n", out); - fputs ("m4_divert_push(0)dnl\n", out); - xfclose (out); + if (trace_flag & trace_muscles) + muscles_output (stderr); + { + FILE *out = xfdopen (filter_fd[1], "w"); + muscles_output (out); + xfclose (out); + } + /* Read and process m4's output. */ timevar_push (TV_M4); - m4_invoke (tempfile); + { + FILE *in = xfdopen (filter_fd[0], "r"); + 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); + } + wait_subprocess (pid, "m4", false, false, true, true, NULL); timevar_pop (TV_M4); - - /* If `debugging', keep this file alive. */ - if (!(trace_flag & trace_tools)) - unlink (tempfile); - - free (tempfile); } static void prepare (void) { - /* Flags. */ - MUSCLE_INSERT_INT ("locations_flag", locations_flag); - MUSCLE_INSERT_INT ("defines_flag", defines_flag); - MUSCLE_INSERT_INT ("error_verbose", error_verbose); - MUSCLE_INSERT_INT ("pure", pure_parser); - MUSCLE_INSERT_INT ("debug", debug_flag); - - /* 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 ("prefix", spec_name_prefix ? spec_name_prefix : "yy"); - 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); - - /* Symbols. */ - MUSCLE_INSERT_INT ("tokens_number", ntokens); - MUSCLE_INSERT_INT ("nterms_number", nvars); - MUSCLE_INSERT_INT ("undef_token_number", undeftoken->number); - MUSCLE_INSERT_INT ("user_token_number_max", max_user_token_number); - - /* Rules. */ - MUSCLE_INSERT_INT ("rules_number", nrules); - - /* States. */ - MUSCLE_INSERT_INT ("last", high); - MUSCLE_INSERT_INT ("final_state_number", final_state->number); - MUSCLE_INSERT_INT ("states_number", nstates); - - /* User Code. */ - 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)); + /* BISON_USE_PUSH_FOR_PULL is for the test suite and should not be + documented for the user. */ + char const *cp = getenv ("BISON_USE_PUSH_FOR_PULL"); + bool use_push_for_pull_flag = cp && *cp && strtol (cp, 0, 10); - /* Find the right skeleton file. */ - if (!skeleton) - { - if (glr_parser) - skeleton = "glr.c"; - else - skeleton = "yacc.c"; - } + /* Flags. */ + MUSCLE_INSERT_BOOL ("defines_flag", defines_flag); + MUSCLE_INSERT_BOOL ("glr_flag", glr_parser); + MUSCLE_INSERT_BOOL ("nondeterministic_flag", nondeterministic_parser); + MUSCLE_INSERT_BOOL ("synclines_flag", !no_lines_flag); + MUSCLE_INSERT_BOOL ("tag_seen_flag", tag_seen); + MUSCLE_INSERT_BOOL ("token_table_flag", token_table_flag); + MUSCLE_INSERT_BOOL ("use_push_for_pull_flag", use_push_for_pull_flag); + MUSCLE_INSERT_BOOL ("yacc_flag", yacc_flag); + + /* File names. */ + if (spec_name_prefix) + MUSCLE_INSERT_STRING ("prefix", spec_name_prefix); + + MUSCLE_INSERT_STRING ("file_name_all_but_ext", all_but_ext); + +#define DEFINE(Name) MUSCLE_INSERT_STRING (#Name, Name ? Name : "") + DEFINE (dir_prefix); + DEFINE (parser_file_name); + DEFINE (spec_defines_file); + DEFINE (spec_file_prefix); + DEFINE (spec_graph_file); + DEFINE (spec_name_prefix); + DEFINE (spec_outfile); + DEFINE (spec_verbose_file); +#undef DEFINE + + /* Find the right skeleton file, and add muscles about the skeletons. */ + if (skeleton) + MUSCLE_INSERT_C_STRING ("skeleton", skeleton); + else + skeleton = language->skeleton; - /* Parse the skeleton file and output the needed parsers. */ - muscle_insert ("skeleton", skeleton); + /* About the skeletons. */ + { + /* b4_pkgdatadir is used inside m4_include in the skeletons, so digraphs + would never be expanded. Hopefully no one has M4-special characters in + his Bison installation path. */ + MUSCLE_INSERT_STRING_RAW ("pkgdatadir", pkgdatadir ()); + } } @@ -1451,12 +725,11 @@ output (void) { obstack_init (&format_obstack); - prepare_tokens (); + prepare_symbols (); prepare_rules (); prepare_states (); - timevar_push (TV_ACTIONS); prepare_actions (); - timevar_pop (TV_ACTIONS); + prepare_symbol_definitions (); prepare (); @@ -1464,6 +737,11 @@ output (void) output_skeleton (); obstack_free (&format_obstack, NULL); - obstack_free (&pre_prologue_obstack, NULL); - obstack_free (&post_prologue_obstack, NULL); +} + +char const * +pkgdatadir (void) +{ + char const *cp = getenv ("BISON_PKGDATADIR"); + return cp ? cp : PKGDATADIR; }