X-Git-Url: https://git.saurik.com/bison.git/blobdiff_plain/7612000cb004ae5b8516e3a34a79ab32050d688e..8b752b00fdaaa3802e4ab05a468f6cda4a9fa500:/src/output.c diff --git a/src/output.c b/src/output.c index 6d96d50f..4a7a7265 100644 --- a/src/output.c +++ b/src/output.c @@ -1,849 +1,530 @@ /* Output the generated parsing program for bison, - Copyright (C) 1984, 1986, 1989, 1992 Free Software Foundation, Inc. + Copyright (C) 1984, 1986, 1989, 1992, 2000, 2001, 2002 + Free Software Foundation, Inc. -This file is part of Bison, the GNU Compiler Compiler. + 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. + 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. -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. + 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. -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, 675 Mass Ave, Cambridge, MA 02139, USA. */ + 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. */ -/* functions to output parsing data to various files. Entries are: +/* The parser tables consist of these tables. Marked ones needed only + for the semantic parser. Double marked are output only if switches + are set. - output_headers () + YYTRANSLATE = vector mapping yylex's token numbers into bison's + token numbers. -Output constant strings to the beginning of certain files. + ++ YYTNAME = vector of string-names indexed by bison token number. - output_trailers() + ++ YYTOKNUM = vector of yylex token numbers corresponding to + entries in YYTNAME. -Output constant strings to the ends of certain files. + YYRLINE = vector of line-numbers of all rules. For yydebug + printouts. - output () + YYRHS = vector of items of all rules. This is exactly what RITEMS + contains. For yydebug and for semantic parser. -Output the parsing tables and the parser code to ftable. + YYPRHS[R] = index in YYRHS of first item for rule R. -The parser tables consist of these tables. -Starred ones needed only for the semantic parser. -Double starred are output only if switches are set. + YYR1[R] = symbol number of symbol that rule R derives. -yytranslate = vector mapping yylex's token numbers into bison's token numbers. + YYR2[R] = number of symbols composing right hand side of rule R. -** yytname = vector of string-names indexed by bison token number + + YYSTOS[S] = the symbol number of the symbol that leads to state + S. -** yytoknum = vector of yylex token numbers corresponding to entries in yytname + 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. -yyrline = vector of line-numbers of all rules. For yydebug printouts. + 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. -yyrhs = vector of items of all rules. - This is exactly what ritems contains. For yydebug and for semantic - parser. + 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. -yyprhs[r] = index in yyrhs of first item for rule r. + If the value in YYTABLE is positive, we shift the token and go to + that state. -yyr1[r] = symbol number of symbol that rule r derives. + If the value is negative, it is minus a rule number to reduce by. -yyr2[r] = number of symbols composing right hand side of rule r. + If the value is zero, the default action from YYDEFACT[S] is used. -* yystos[s] = the symbol number of the symbol that leads to state s. + 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. -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. + YYTABLE = a vector filled with portions for different uses, found + via YYPACT and YYPGOTO. -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. + YYCHECK = a vector indexed in parallel with YYTABLE. It indicates, + in a roundabout way, the bounds of the portion you are trying to + examine. -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. + 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. - If the value in yytable is positive, - we shift the token and go to that state. + YYFINAL = the state number of the termination state. YYFLAG = most + negative short int. Used to flag ?? */ - 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. - -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. - -YYFINAL = the state number of the termination state. -YYFLAG = most negative short int. Used to flag ?? -YYNTBASE = ntokens. - -*/ - -#include #include "system.h" -#include "machine.h" -#include "alloc.h" +#include "bitsetv.h" +#include "quotearg.h" +#include "error.h" +#include "getargs.h" #include "files.h" #include "gram.h" -#include "state.h" - - -extern int debugflag; -extern int nolinesflag; -extern int noparserflag; -extern int toknumflag; - -extern char **tags; -extern int *user_toknums; -extern int tokensetsize; -extern int final_state; -extern core **state_table; -extern shifts **shift_table; -extern errs **err_table; -extern reductions **reduction_table; -extern short *accessing_symbol; -extern unsigned *LA; -extern short *LAruleno; -extern short *lookaheads; -extern char *consistent; -extern short *goto_map; -extern short *from_state; -extern short *to_state; - -void output_token_translations(); -void output_gram(); -void output_stos(); -void output_rule_data(); -void output_defines(); -void output_actions(); -void token_actions(); -void save_row(); -void goto_actions(); -void save_column(); -void sort_actions(); -void pack_table(); -void output_base(); -void output_table(); -void output_check(); -void output_parser(); -void output_program(); -void free_itemset(); -void free_shifts(); -void free_reductions(); -void free_itemsets(); -int action_row(); -int default_goto(); -int matching_state(); -int pack_vector(); - -extern void berror(); -extern void fatals(); -extern char *int_to_string(); -extern void reader_output_yylsp(); +#include "LR0.h" +#include "complain.h" +#include "output.h" +#include "lalr.h" +#include "reader.h" +#include "symtab.h" +#include "conflicts.h" +#include "muscle_tab.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; -static short **tos; -static short *tally; -static short *width; -static short *actrow; -static short *state_count; -static short *order; -static short *base; -static short *pos; -static short *table; -static short *check; +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; +static struct obstack format_obstack; +int error_verbose = 0; -#define GUARDSTR "\n#include \"%s\"\nextern int yyerror;\n\ -extern int yycost;\nextern char * yymsg;\nextern YYSTYPE yyval;\n\n\ -yyguard(n, yyvsp, yylsp)\nregister int n;\nregister YYSTYPE *yyvsp;\n\ -register YYLTYPE *yylsp;\n\ -{\n yyerror = 0;\nyycost = 0;\n yymsg = 0;\nswitch (n)\n {" - -#define ACTSTR "\n#include \"%s\"\nextern YYSTYPE yyval;\ -\nextern int yychar;\ -yyaction(n, yyvsp, yylsp)\nregister int n;\nregister YYSTYPE *yyvsp;\n\ -register YYLTYPE *yylsp;\n{\n switch (n)\n{" -#define ACTSTR_SIMPLE "\n switch (yyn) {\n" +/*----------------------------------------------------------------. +| 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. | +`----------------------------------------------------------------*/ - -void -output_headers() +static void +table_grow (size_t desired) { - if (semantic_parser) - fprintf(fguard, GUARDSTR, attrsfile); + size_t old_size = table_size; - if (noparserflag) - return; + while (table_size <= desired) + table_size *= 2; - fprintf(faction, (semantic_parser ? ACTSTR : ACTSTR_SIMPLE), attrsfile); -/* if (semantic_parser) JF moved this below - fprintf(ftable, "#include \"%s\"\n", attrsfile); - fprintf(ftable, "#include \n\n"); -*/ + if (trace_flag) + fprintf (stderr, "growing table and check from: %d to %d\n", + old_size, table_size); - /* Rename certain symbols if -p was specified. */ - if (spec_name_prefix) - { - fprintf(ftable, "#define yyparse %sparse\n", spec_name_prefix); - fprintf(ftable, "#define yylex %slex\n", spec_name_prefix); - fprintf(ftable, "#define yyerror %serror\n", spec_name_prefix); - fprintf(ftable, "#define yylval %slval\n", spec_name_prefix); - fprintf(ftable, "#define yychar %schar\n", spec_name_prefix); - fprintf(ftable, "#define yydebug %sdebug\n", spec_name_prefix); - fprintf(ftable, "#define yynerrs %snerrs\n", spec_name_prefix); - } -} + table = XREALLOC (table, short, table_size); + check = XREALLOC (check, short, table_size); + if (glr_parser) + conflict_table = XREALLOC (conflict_table, unsigned int, table_size); - -void -output_trailers() -{ - if (semantic_parser) - fprintf(fguard, "\n }\n}\n"); - - fprintf(faction, "\n"); - - if (noparserflag) - return; - - if (semantic_parser) - fprintf(faction, " }\n"); - fprintf(faction, "}\n"); -} - - -void -output() -{ - int c; - - /* output_token_defines(ftable); / * JF put out token defines FIRST */ - if (!semantic_parser) /* JF Put out other stuff */ + for (/* Nothing. */; old_size < table_size; ++old_size) { - rewind(fattrs); - while ((c=getc(fattrs))!=EOF) - putc(c,ftable); + table[old_size] = 0; + check[old_size] = -1; } - reader_output_yylsp(ftable); - if (debugflag) - fprintf(ftable, "#ifndef YYDEBUG\n#define YYDEBUG %d\n#endif\n\n", - !!debugflag); - - if (semantic_parser) - fprintf(ftable, "#include \"%s\"\n", attrsfile); - - if (! noparserflag) - fprintf(ftable, "#include \n\n"); - - /* Make "const" do nothing if not in ANSI C. */ - fprintf (ftable, "#ifndef __cplusplus\n#ifndef __STDC__\n#define const\n#endif\n#endif\n\n"); - - free_itemsets(); - output_defines(); - output_token_translations(); -/* if (semantic_parser) */ - /* This is now unconditional because debugging printouts can use it. */ - output_gram(); - FREE(ritem); - if (semantic_parser) - output_stos(); - output_rule_data(); - output_actions(); - if (! noparserflag) - output_parser(); - output_program(); } -void -output_token_translations() -{ - register int i, j; -/* register short *sp; JF unused */ - - if (translations) - { - fprintf(ftable, - "\n#define YYTRANSLATE(x) ((unsigned)(x) <= %d ? yytranslate[x] : %d)\n", - max_user_token_number, nsyms); - - if (ntokens < 127) /* play it very safe; check maximum element value. */ - fprintf(ftable, "\nstatic const char yytranslate[] = { 0"); - else - fprintf(ftable, "\nstatic const short yytranslate[] = { 0"); - - j = 10; - for (i = 1; i <= max_user_token_number; i++) - { - putc(',', ftable); - - if (j >= 10) - { - putc('\n', ftable); - j = 1; - } - else - { - j++; - } - - fprintf(ftable, "%6d", token_translations[i]); - } - - fprintf(ftable, "\n};\n"); - } - else - { - fprintf(ftable, "\n#define YYTRANSLATE(x) (x)\n"); - } -} - - -void -output_gram() -{ - register int i; - register int j; - register short *sp; - - /* With the ordinary parser, - yyprhs and yyrhs are needed only for yydebug. */ - /* With the noparser option, all tables are generated */ - if (! semantic_parser && ! noparserflag) - fprintf(ftable, "\n#if YYDEBUG != 0"); - - fprintf(ftable, "\nstatic const short yyprhs[] = { 0"); - - j = 10; - for (i = 1; i <= nrules; i++) - { - putc(',', ftable); - - if (j >= 10) - { - putc('\n', ftable); - j = 1; - } - else - { - j++; - } - - fprintf(ftable, "%6d", rrhs[i]); - } - - fprintf(ftable, "\n};\n"); - - fprintf(ftable, "\nstatic const short yyrhs[] = {%6d", ritem[0]); - - j = 10; - for (sp = ritem + 1; *sp; sp++) - { - putc(',', ftable); - - if (j >= 10) - { - putc('\n', ftable); - j = 1; - } - else - { - j++; - } - - if (*sp > 0) - fprintf(ftable, "%6d", *sp); - else - fprintf(ftable, " 0"); - } - - fprintf(ftable, "\n};\n"); - - if (! semantic_parser && ! noparserflag) - fprintf(ftable, "\n#endif\n"); +/*-------------------------------------------------------------------. +| 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) -void -output_stos() -{ - register int i; - register int j; - fprintf(ftable, "\nstatic const short yystos[] = { 0"); +/*-----------------------------------------------------------------. +| Prepare the muscles related to the tokens: translate, tname, and | +| toknum. | +`-----------------------------------------------------------------*/ - j = 10; - for (i = 1; i < nstates; i++) - { - putc(',', ftable); - - if (j >= 10) - { - putc('\n', ftable); - j = 1; - } - else - { - j++; - } - - fprintf(ftable, "%6d", accessing_symbol[i]); - } - - fprintf(ftable, "\n};\n"); -} - - -void -output_rule_data() +static void +prepare_tokens (void) { - register int i; - register int j; - - fprintf(ftable, "\n#if YYDEBUG != 0\n"); - fprintf(ftable, "static const short yyrline[] = { 0"); - - j = 10; - for (i = 1; i <= nrules; i++) - { - putc(',', ftable); - - if (j >= 10) - { - putc('\n', ftable); - j = 1; - } - else - { - j++; - } - - fprintf(ftable, "%6d", rline[i]); - } - fprintf(ftable, "\n};\n#endif\n\n"); - - if (toknumflag || noparserflag) - { - fprintf(ftable, "#define YYNTOKENS %d\n", ntokens); - fprintf(ftable, "#define YYNNTS %d\n", nvars); - fprintf(ftable, "#define YYNRULES %d\n", nrules); - fprintf(ftable, "#define YYNSTATES %d\n", nstates); - fprintf(ftable, "#define YYMAXUTOK %d\n\n", max_user_token_number); - } - - if (! toknumflag && ! noparserflag) - fprintf(ftable, "\n#if YYDEBUG != 0 || defined (YYERROR_VERBOSE)\n\n"); - - /* Output the table of symbol names. */ - - fprintf(ftable, - "static const char * const yytname[] = { \"%s\"", - tags[0]); - - j = strlen (tags[0]) + 44; - for (i = 1; i < nsyms; i++) - /* this used to be i<=nsyms, but that output a final "" symbol - almost by accident */ - { - register char *p; - putc(',', ftable); - j++; - - if (j > 75) - { - putc('\n', ftable); - j = 0; - } - - putc ('\"', ftable); - j++; - - for (p = tags[i]; p && *p; p++) - { - if (*p == '"' || *p == '\\') - { - fprintf(ftable, "\\%c", *p); - j += 2; - } - else if (*p == '\n') - { - fprintf(ftable, "\\n"); - j += 2; - } - else if (*p == '\t') - { - fprintf(ftable, "\\t"); - j += 2; - } - else if (*p == '\b') - { - fprintf(ftable, "\\b"); - j += 2; - } - else if (*p < 040 || *p >= 0177) - { - fprintf(ftable, "\\%03o", *p); - j += 4; - } - else - { - putc(*p, ftable); - j++; - } - } - - putc ('\"', ftable); - j++; - } - fprintf(ftable, ", NULL\n};\n"); /* add a NULL entry to list of tokens */ - - if (! toknumflag && ! noparserflag) - fprintf(ftable, "#endif\n\n"); + muscle_insert_symbol_number_table ("translate", + token_translations, + 0, 1, max_user_token_number + 1); + + { + int i; + 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, + 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; + } - if (toknumflag) - { - fprintf(ftable, "static const short yytoknum[] = { 0"); - j = 10; - for (i = 1; i <= ntokens; i++) { - putc(',', ftable); - if (j >= 10) - { - putc('\n', ftable); - j = 1; - } - else - j++; - fprintf(ftable, "%6d", user_toknums[i]); + obstack_sgrow (&format_obstack, cp); + obstack_sgrow (&format_obstack, ", "); + j += strsize; } - fprintf(ftable, "\n};\n\n"); - } + /* Add a NULL entry to list of tokens (well, 0, as NULL might not be + defined). */ + obstack_sgrow (&format_obstack, "0"); - fprintf(ftable, "static const short yyr1[] = { 0"); - - j = 10; - for (i = 1; i <= nrules; i++) - { - putc(',', ftable); - - if (j >= 10) - { - putc('\n', ftable); - j = 1; - } - else - { - j++; - } + /* Finish table and store. */ + obstack_1grow (&format_obstack, 0); + muscle_insert ("tname", obstack_finish (&format_obstack)); + } - fprintf(ftable, "%6d", rlhs[i]); - } + /* Output YYTOKNUM. */ + { + 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); + } +} - FREE(rlhs + 1); - fprintf(ftable, "\n};\n\nstatic const short yyr2[] = { 0"); +/*-------------------------------------------------------------. +| Prepare the muscles related to the rules: rhs, prhs, r1, r2, | +| rline, dprec, merger | +`-------------------------------------------------------------*/ - j = 10; - for (i = 1; i < nrules; i++) +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) { - putc(',', ftable); - - if (j >= 10) - { - putc('\n', ftable); - j = 1; - } - else - { - j++; - } - - fprintf(ftable, "%6d", rrhs[i + 1] - rrhs[i] - 1); + 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; } - - putc(',', ftable); - if (j >= 10) - putc('\n', ftable); - - fprintf(ftable, "%6d\n};\n", nitems - rrhs[nrules] - 1); - FREE(rrhs + 1); + 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. | +`--------------------------------------------*/ -void -output_defines() +static void +prepare_states (void) { - fprintf(ftable, "\n\n#define\tYYFINAL\t\t%d\n", final_state); - fprintf(ftable, "#define\tYYFLAG\t\t%d\n", MINSHORT); - fprintf(ftable, "#define\tYYNTBASE\t%d\n", ntokens); + 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; + 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. | +`-------------------------------------------------------------------*/ -/* compute and output yydefact, yydefgoto, yypact, yypgoto, yytable and yycheck. */ - -void -output_actions() +static void +conflict_row (state_t *state) { - nvectors = nstates + nvars; - - froms = NEW2(nvectors, short *); - tos = NEW2(nvectors, short *); - tally = NEW2(nvectors, short); - width = NEW2(nvectors, short); - - token_actions(); - free_shifts(); - free_reductions(); - FREE(lookaheads); - FREE(LA); - FREE(LAruleno); - FREE(accessing_symbol); - - goto_actions(); - FREE(goto_map + ntokens); - FREE(from_state); - FREE(to_state); - - sort_actions(); - pack_table(); - output_base(); - output_table(); - output_check(); -} + int i, j; + if (! glr_parser) + return; + for (j = 0; j < ntokens; j += 1) + if (conflrow[j]) + { + conflrow[j] = conflict_list_cnt; -/* 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. */ - -void -token_actions() -{ - register int i; - register int j; - register int k; - - actrow = NEW2(ntokens, short); - - k = action_row(0); - fprintf(ftable, "\nstatic const short yydefact[] = {%6d", k); - save_row(0); - - j = 10; - for (i = 1; i < nstates; i++) - { - putc(',', ftable); - - if (j >= 10) - { - putc('\n', ftable); - j = 1; - } - else - { - j++; - } - - k = action_row(i); - fprintf(ftable, "%6d", k); - save_row(i); - } + /* 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; + } - fprintf(ftable, "\n};\n"); - FREE(actrow); + /* 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 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. */ - -int -action_row(state) -int state; +/*------------------------------------------------------------------. +| 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 | +| 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) { - register int i; - register int j; - register int k; - register int m; - register int n; - register int count; - register int default_rule; - register int nreds; - register int max; - register int rule; - register int shift_state; - register int symbol; - register unsigned mask; - register unsigned *wordp; - register reductions *redp; - register shifts *shiftp; - register errs *errp; - int nodefault = 0; /* set nonzero to inhibit having any default reduction */ + int i; + rule_number_t default_rule = 0; + reductions_t *redp = state->reductions; + transitions_t *transitions = state->transitions; + 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; - - default_rule = 0; - nreds = 0; - redp = reduction_table[state]; + actrow[i] = conflrow[i] = 0; - if (redp) + if (redp->num >= 1) { - nreds = redp->nreds; - - if (nreds >= 1) + 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) { - /* loop over all the rules available here which require lookahead */ - m = lookaheads[state]; - n = lookaheads[state + 1]; - - for (i = n - 1; i >= m; i--) - { - rule = - LAruleno[i]; - wordp = LA + i * tokensetsize; - mask = 1; - - /* and find each token which the rule finds acceptable to come next */ - for (j = 0; j < ntokens; j++) - { - /* and record this rule as the rule to use if that token follows. */ - if (mask & *wordp) - actrow[j] = rule; - - mask <<= 1; - if (mask == 0) - { - mask = 1; - wordp++; - } - } - } + /* and record this rule as the rule to use if that + token follows. */ + if (actrow[j] != 0) + conflicted = conflrow[j] = 1; + actrow[j] = -state->lookaheads_rule[i]->number; } } - shiftp = shift_table[state]; - - /* 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. */ - - if (shiftp) - { - k = shiftp->nshifts; - - for (i = 0; i < k; i++) - { - shift_state = shiftp->shifts[i]; - if (! shift_state) continue; - - symbol = accessing_symbol[shift_state]; - - if (ISVAR(symbol)) - break; - - actrow[symbol] = shift_state; - - /* do not use any default reduction if there is a shift for error */ - - if (symbol == error_token_number) nodefault = 1; - } - } + /* 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 < transitions->num && TRANSITION_IS_SHIFT (transitions, i); i++) + if (!TRANSITION_IS_DISABLED (transitions, i)) + { + symbol_number_t symbol = TRANSITION_SYMBOL (transitions, i); + state_number_t shift_state = transitions->states[i]; - errp = err_table[state]; + if (actrow[symbol] != 0) + conflicted = conflrow[symbol] = 1; + actrow[symbol] = state_number_as_int (shift_state); - /* See which tokens are an explicit error in this state - (due to %nonassoc). For them, record MINSHORT as the action. */ + /* Do not use any default reduction if there is a shift for + error */ + if (symbol == errtoken->number) + nodefault = 1; + } - if (errp) + /* See which tokens are an explicit error in this state (due to + %nonassoc). For them, record SHRT_MIN as the action. */ + for (i = 0; i < errp->num; i++) { - k = errp->nerrs; - - for (i = 0; i < k; i++) - { - symbol = errp->errs[i]; - actrow[symbol] = MINSHORT; - } + symbol_number_t symbol = errp->symbols[i]; + actrow[symbol] = SHRT_MIN; } - /* now find the most common reduction and make it the default action for this state. */ + /* Now find the most common reduction and make it the default action + for this state. */ - if (nreds >= 1 && ! nodefault) + if (redp->num >= 1 && !nodefault) { - if (consistent[state]) + if (state->consistent) default_rule = redp->rules[0]; else { - max = 0; - for (i = m; i < n; i++) + int max = 0; + for (i = 0; i < state->nlookaheads; i++) { - count = 0; - rule = - LAruleno[i]; - + int count = 0; + rule_number_t rule = state->lookaheads_rule[i]->number; + symbol_number_t j; + for (j = 0; j < ntokens; j++) - { - if (actrow[j] == rule) - count++; - } - + if (actrow[j] == -rule) + count++; + if (count > max) { max = count; default_rule = rule; } } - - /* 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) - actrow[j] = 0; - } - - default_rule = - default_rule; + if (actrow[j] == -default_rule + && ! (glr_parser && conflrow[j])) + actrow[j] = 0; } } } @@ -852,363 +533,469 @@ int state; So replace any action which says "error" with "use default". */ if (default_rule == 0) - for (j = 0; j < ntokens; j++) - { - if (actrow[j] == MINSHORT) - actrow[j] = 0; - } + for (i = 0; i < ntokens; i++) + if (actrow[i] == SHRT_MIN) + actrow[i] = 0; + + if (conflicted) + conflict_row (state); - return (default_rule); + return default_rule; } -void -save_row(state) -int state; +static void +save_row (state_number_t state) { - register int i; - register int count; - register short *sp; - register short *sp1; - register short *sp2; + symbol_number_t i; + int count; + short *sp = NULL; + short *sp1 = NULL; + short *sp2 = NULL; + unsigned int *sp3 = NULL; count = 0; for (i = 0; i < ntokens; i++) - { - if (actrow[i] != 0) - count++; - } + if (actrow[i] != 0) + count++; if (count == 0) return; - froms[state] = sp1 = sp = NEW2(count, short); - tos[state] = sp2 = NEW2(count, short); + 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 (actrow[i] != 0) + { + *sp1++ = i; + *sp2++ = actrow[i]; + if (glr_parser) + *sp3++ = conflrow[i]; + } tally[state] = count; width[state] = sp1[-1] - sp[0] + 1; } +/*------------------------------------------------------------------. +| 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. | +`------------------------------------------------------------------*/ -/* 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. */ - -void -goto_actions() +static void +token_actions (void) { - register int i; - register int j; - register int k; + state_number_t i; + int nconflict = conflicts_total_count (); - state_count = NEW2(nstates, short); + short *yydefact = XCALLOC (short, nstates); - k = default_goto(ntokens); - fprintf(ftable, "\nstatic const short yydefgoto[] = {%6d", k); - save_column(ntokens, k); + actrow = XCALLOC (short, ntokens); - j = 10; - for (i = ntokens + 1; i < nsyms; i++) + conflrow = XCALLOC (short, ntokens); + if (glr_parser) { - putc(',', ftable); - - if (j >= 10) - { - putc('\n', ftable); - j = 1; - } - else - { - j++; - } + 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; - k = default_goto(i); - fprintf(ftable, "%6d", k); - save_column(i, k); + for (i = 0; i < nstates; ++i) + { + yydefact[i] = action_row (states[i]); + save_row (i); } - fprintf(ftable, "\n};\n"); - FREE(state_count); + muscle_insert_short_table ("defact", yydefact, + yydefact[0], 1, nstates); + XFREE (actrow); + XFREE (conflrow); + XFREE (yydefact); } +/*-----------------------------. +| Output the actions to OOUT. | +`-----------------------------*/ -int -default_goto(symbol) -int symbol; +void +actions_output (FILE *out) { - register int i; - register int m; - register int n; - register int default_state; - register int max; - - m = goto_map[symbol]; - n = goto_map[symbol + 1]; + rule_number_t r; - if (m == n) - return (-1); - - for (i = 0; i < nstates; i++) - state_count[i] = 0; + fputs ("m4_define([b4_actions], \n[[", out); + for (r = 1; r < nrules + 1; ++r) + if (rules[r].action) + { + fprintf (out, " case %d:\n", r); + + 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); + } + fputs ("]])\n\n", out); +} - for (i = m; i < n; i++) - state_count[to_state[i]]++; +/*--------------------------------------. +| Output the merge functions to OUT. | +`--------------------------------------*/ - max = 0; - default_state = -1; +static void +merger_output (FILE *out) +{ + int n; + merger_list* p; - for (i = 0; i < nstates; i++) + fputs ("m4_define([b4_mergers], \n[[", out); + for (n = 1, p = merge_functions; p != NULL; n += 1, p = p->next) { - if (state_count[i] > max) - { - max = state_count[i]; - default_state = i; - } + 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); } - - return (default_state); + fputs ("]])\n\n", out); } +/*---------------------------------------. +| Output the tokens definition to OOUT. | +`---------------------------------------*/ void -save_column(symbol, default_state) -int symbol; -int default_state; +token_definitions_output (FILE *out) { - register int i; - register int m; - register int n; - register short *sp; - register short *sp1; - register short *sp2; - register int count; - register int symno; - - m = goto_map[symbol]; - n = goto_map[symbol + 1]; + int i; + int first = 1; - count = 0; - for (i = m; i < n; i++) + fputs ("m4_define([b4_tokens], \n[", out); + for (i = 0; i < ntokens; ++i) { - if (to_state[i] != default_state) - count++; + 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); +} - if (count == 0) - return; - - symno = symbol - ntokens + nstates; - froms[symno] = sp1 = sp = NEW2(count, short); - tos[symno] = sp2 = NEW2(count, short); +/*----------------------------------------. +| Output the symbol destructors to OOUT. | +`----------------------------------------*/ - for (i = m; i < n; i++) - { - if (to_state[i] != default_state) - { - *sp1++ = from_state[i]; - *sp2++ = to_state[i]; - } - } +static void +symbol_destructors_output (FILE *out) +{ + int i; + int first = 1; - tally[symno] = count; - width[symno] = sp1[-1] - sp[0] + 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. | +`-------------------------------------*/ -/* the next few functions decide how to pack - the actions and gotos information into yytable. */ +static void +symbol_printers_output (FILE *out) +{ + int i; + int first = 1; -void -sort_actions() + 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); +} + + +static void +save_column (symbol_number_t symbol, state_number_t default_state) { - register int i; - register int j; - register int k; - register int t; - register int w; + int i; + short *sp; + short *sp1; + short *sp2; + int count; + int symno = symbol - ntokens + state_number_as_int (nstates); - order = NEW2(nvectors, short); - nentries = 0; + int begin = goto_map[symbol]; + int end = goto_map[symbol + 1]; - for (i = 0; i < nvectors; i++) - { - if (tally[i] > 0) - { - t = tally[i]; - w = width[i]; - j = nentries - 1; + count = 0; + for (i = begin; i < end; i++) + if (to_state[i] != default_state) + count++; - while (j >= 0 && (width[order[j]] < w)) - j--; + if (count == 0) + return; - while (j >= 0 && (width[order[j]] == w) && (tally[order[j]] < t)) - j--; + froms[symno] = sp1 = sp = XCALLOC (short, count); + tos[symno] = sp2 = XCALLOC (short, count); - for (k = nentries - 1; k > j; k--) - order[k + 1] = order[k]; + for (i = begin; i < end; i++) + if (to_state[i] != default_state) + { + *sp1++ = from_state[i]; + *sp2++ = to_state[i]; + } - order[j + 1] = i; - nentries++; - } - } + tally[symno] = count; + width[symno] = sp1[-1] - sp[0] + 1; } -void -pack_table() +static state_number_t +default_goto (symbol_number_t symbol) { - register int i; - register int place; - register int state; + state_number_t s; + int i; + int m = goto_map[symbol]; + int n = goto_map[symbol + 1]; + state_number_t default_state = (state_number_t) -1; + int max = 0; - base = NEW2(nvectors, short); - pos = NEW2(nentries, short); - table = NEW2(MAXTABLE, short); - check = NEW2(MAXTABLE, short); + if (m == n) + return (state_number_t) -1; - lowzero = 0; - high = 0; + for (s = 0; s < nstates; s++) + state_count[s] = 0; - for (i = 0; i < nvectors; i++) - base[i] = MINSHORT; + for (i = m; i < n; i++) + state_count[to_state[i]]++; - for (i = 0; i < MAXTABLE; i++) - check[i] = -1; + for (s = 0; s < nstates; s++) + if (state_count[s] > max) + { + max = state_count[s]; + default_state = s; + } - for (i = 0; i < nentries; i++) - { - state = matching_state(i); + return default_state; +} - if (state < 0) - place = pack_vector(i); - else - place = base[state]; - pos[i] = place; - base[order[i]] = place; - } +/*-------------------------------------------------------------------. +| 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. | +`-------------------------------------------------------------------*/ - for (i = 0; i < nvectors; i++) +static void +goto_actions (void) +{ + 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) { - if (froms[i]) - FREE(froms[i]); - if (tos[i]) - FREE(tos[i]); + state_number_t default_state = default_goto (i); + save_column (i, default_state); + yydefgoto[i - ntokens] = default_state; } - FREE(froms); - FREE(tos); - FREE(pos); + muscle_insert_state_number_table ("defgoto", yydefgoto, + yydefgoto[0], 1, nsyms - ntokens); + XFREE (state_count); + XFREE (yydefgoto); } +/* The next few functions decide how to pack the actions and gotos + information into yytable. */ -int -matching_state(vector) -int vector; +static void +sort_actions (void) { - register int i; - register int j; - register int k; - register int t; - register int w; - register int match; - register int prev; - - i = order[vector]; - if (i >= nstates) - return (-1); + int i; + + order = XCALLOC (short, nvectors); + 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++; + } +} + + +static int +matching_state (int vector) +{ + int i = order[vector]; + int t; + int w; + int prev; + + if (i >= (int) nstates) + return -1; t = tally[i]; w = width[i]; for (prev = vector - 1; prev >= 0; prev--) { - j = order[prev]; + int j = order[prev]; + int k; + int match = 1; + if (width[j] != w || tally[j] != t) - return (-1); + return -1; - match = 1; for (k = 0; match && k < t; k++) - { - if (tos[j][k] != tos[i][k] || froms[j][k] != froms[i][k]) - match = 0; - } + if (tos[j][k] != tos[i][k] || froms[j][k] != froms[i][k]) + match = 0; if (match) - return (j); + return j; } - return (-1); + return -1; } - -int -pack_vector(vector) -int vector; +static int +pack_vector (int vector) { - register int i; - register int j; - register int k; - register int t; - register int loc; - register int ok; - register short *from; - register short *to; - - i = order[vector]; - t = tally[i]; - - if (t == 0) - berror("pack_vector"); + int i = order[vector]; + int j; + int t = tally[i]; + int loc = 0; + short *from = froms[i]; + short *to = tos[i]; + unsigned int *conflict_to = conflict_tos[i]; - from = froms[i]; - to = tos[i]; + assert (t); - for (j = lowzero - from[0]; j < MAXTABLE; j++) + for (j = lowzero - from[0]; j < (int) table_size; j++) { - ok = 1; + int k; + int ok = 1; for (k = 0; ok && k < t; k++) { - loc = j + from[k]; - if (loc > MAXTABLE) - fatals(_("maximum table size (%s) exceeded"), int_to_string(MAXTABLE)); + 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 (pos[k] == j) + ok = 0; if (ok) { 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) @@ -1217,268 +1004,322 @@ int vector; if (loc > high) high = loc; - return (j); + return j; } } - - berror("pack_vector"); - return 0; /* JF keep lint happy */ +#define pack_vector_succeeded 0 + assert (pack_vector_succeeded); + return 0; } +static void +pack_table (void) +{ + int i; + int place; + int state; -/* the following functions output yytable, yycheck - and the vectors whose elements index the portion starts */ + base = XCALLOC (short, nvectors); + pos = XCALLOC (short, nentries); + table = XCALLOC (short, table_size); + if (glr_parser) + conflict_table = XCALLOC (unsigned int, table_size); + check = XCALLOC (short, table_size); -void -output_base() -{ - register int i; - register int j; + lowzero = 0; + high = 0; + + for (i = 0; i < nvectors; i++) + base[i] = SHRT_MIN; - fprintf(ftable, "\nstatic const short yypact[] = {%6d", base[0]); + for (i = 0; i < (int) table_size; i++) + check[i] = -1; - j = 10; - for (i = 1; i < nstates; i++) + for (i = 0; i < nentries; i++) { - putc(',', ftable); + state = matching_state (i); - if (j >= 10) - { - putc('\n', ftable); - j = 1; - } + if (state < 0) + place = pack_vector (i); else - { - j++; - } + place = base[state]; - fprintf(ftable, "%6d", base[i]); + pos[i] = place; + base[order[i]] = place; } - fprintf(ftable, "\n};\n\nstatic const short yypgoto[] = {%6d", base[nstates]); - - j = 10; - for (i = nstates + 1; i < nvectors; i++) + for (i = 0; i < nvectors; i++) { - putc(',', ftable); + XFREE (froms[i]); + XFREE (tos[i]); + XFREE (conflict_tos[i]); + } - if (j >= 10) - { - putc('\n', ftable); - j = 1; - } - else - { - j++; - } + XFREE (froms); + XFREE (tos); + XFREE (conflict_tos); + XFREE (pos); +} - fprintf(ftable, "%6d", base[i]); - } +/* the following functions output yytable, yycheck, yyconflp, yyconfl, + and the vectors whose elements index the portion starts */ - fprintf(ftable, "\n};\n"); - FREE(base); +static void +output_base (void) +{ + /* Output pact. */ + muscle_insert_short_table ("pact", base, + base[0], 1, nstates); + + /* Output pgoto. */ + muscle_insert_short_table ("pgoto", base, + base[nstates], nstates + 1, nvectors); + XFREE (base); } -void -output_table() +static void +output_table (void) { - register int i; - register int j; + muscle_insert_short_table ("table", table, + table[0], 1, high + 1); + XFREE (table); +} - fprintf(ftable, "\n\n#define\tYYLAST\t\t%d\n\n", high); - fprintf(ftable, "\nstatic const short yytable[] = {%6d", table[0]); - j = 10; - for (i = 1; i <= high; i++) - { - putc(',', ftable); +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; - if (j >= 10) - { - putc('\n', ftable); - j = 1; - } - else - { - j++; - } + 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); +} - fprintf(ftable, "%6d", table[i]); - } - fprintf(ftable, "\n};\n"); - FREE(table); +static void +output_check (void) +{ + muscle_insert_short_table ("check", check, + check[0], 1, high + 1); + XFREE (check); } +/*-----------------------------------------------------------------. +| Compute and output yydefact, yydefgoto, yypact, yypgoto, yytable | +| and yycheck. | +`-----------------------------------------------------------------*/ -void -output_check() +static void +output_actions (void) { - register int i; - register int j; + /* 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)); - fprintf(ftable, "\nstatic const short yycheck[] = {%6d", check[0]); + nvectors = state_number_as_int (nstates) + nvars; - j = 10; - for (i = 1; i <= high; i++) - { - putc(',', ftable); + froms = XCALLOC (short *, nvectors); + tos = XCALLOC (short *, nvectors); + conflict_tos = XCALLOC (unsigned int *, nvectors); + tally = XCALLOC (short, nvectors); + width = XCALLOC (short, nvectors); - if (j >= 10) - { - putc('\n', ftable); - j = 1; - } - else - { - j++; - } + token_actions (); + bitsetv_free (LA); + free (LArule); - fprintf(ftable, "%6d", check[i]); - } + goto_actions (); + XFREE (goto_map + ntokens); + XFREE (from_state); + XFREE (to_state); - fprintf(ftable, "\n};\n"); - FREE(check); -} + sort_actions (); + pack_table (); + output_base (); + output_table (); + output_conflicts (); + output_check (); +} -/* copy the parser code into the ftable file at the end. */ + +/*----------------------. +| Run our backend, M4. | +`----------------------*/ -void -output_parser() +static void +m4_invoke (const char *definitions) { - register int c; -#ifdef DONTDEF - FILE *fpars; -#else -#define fpars fparser -#endif - - if (pure_parser) - fprintf(ftable, "#define YYPURE 1\n\n"); - -#ifdef DONTDEF /* JF no longer needed 'cuz open_extra_files changes the - currently open parser from bison.simple to bison.hairy */ - if (semantic_parser) - fpars = fparser; - else fpars = fparser1; -#endif - - /* Loop over lines in the standard parser file. */ - - while (1) - { - int write_line = 1; + /* 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 (); +} - c = getc(fpars); - /* See if the line starts with `#line. - If so, set write_line to 0. */ - if (nolinesflag) - if (c == '#') - { - c = getc(fpars); - if (c == 'l') - { - c = getc(fpars); - if (c == 'i') - { - c = getc(fpars); - if (c == 'n') - { - c = getc(fpars); - if (c == 'e') - write_line = 0; - else - fprintf(ftable, "#lin"); - } - else - fprintf(ftable, "#li"); - } - else - fprintf(ftable, "#l"); - } - else - fprintf(ftable, "#"); - } +/*---------------------------. +| Call the skeleton parser. | +`---------------------------*/ - /* now write out the line... */ - for (; c != '\n' && c != EOF; c = getc(fpars)) - if (write_line) - if (c == '$') - { - /* `$' in the parser file indicates where to put the actions. - Copy them in at this point. */ - rewind(faction); - for(c=getc(faction);c!=EOF;c=getc(faction)) - putc(c,ftable); - } - else - putc(c, ftable); - if (c == EOF) - break; - putc(c, ftable); - } +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); + + muscles_m4_output (out); + + fputs ("m4_wrap([m4_divert_pop(0)])\n", out); + fputs ("m4_divert_push(0)dnl\n", out); + xfclose (out); + + m4_invoke (tempfile); + + /* If `debugging', keep this file alive. */ + if (!trace_flag) + unlink (tempfile); + + free (tempfile); } -void -output_program() +static void +prepare (void) { - register int c; - extern int lineno; - - if (!nolinesflag) - fprintf(ftable, "#line %d \"%s\"\n", lineno, infile); - - c = getc(finput); - while (c != EOF) + MUSCLE_INSERT_INT ("last", high); + 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->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_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); + MUSCLE_INSERT_INT ("defines_flag", defines_flag); + + /* 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)); + + /* Find the right skeleton file. */ + if (!skeleton) { - putc(c, ftable); - c = getc(finput); + if (glr_parser) + skeleton = "glr.c"; + else + skeleton = "yacc.c"; } -} - -void -free_itemsets() -{ - register core *cp,*cptmp; - - FREE(state_table); - - for (cp = first_state; cp; cp = cptmp) { - cptmp=cp->next; - FREE(cp); - } + /* Parse the skeleton file and output the needed parsers. */ + muscle_insert ("skeleton", skeleton); } +/*----------------------------------------------------------. +| Output the parsing tables and the parser code to ftable. | +`----------------------------------------------------------*/ + void -free_shifts() +output (void) { - register shifts *sp,*sptmp;/* JF derefrenced freed ptr */ + obstack_init (&format_obstack); - FREE(shift_table); + prepare_tokens (); + prepare_rules (); + prepare_states (); + output_actions (); - for (sp = first_shift; sp; sp = sptmp) { - sptmp=sp->next; - FREE(sp); - } -} - - -void -free_reductions() -{ - register reductions *rp,*rptmp;/* JF fixed freed ptr */ + prepare (); - FREE(reduction_table); + /* Process the selected skeleton file. */ + output_skeleton (); - for (rp = first_reduction; rp; rp = rptmp) { - rptmp=rp->next; - FREE(rp); - } + obstack_free (&format_obstack, NULL); + obstack_free (&pre_prologue_obstack, NULL); + obstack_free (&post_prologue_obstack, NULL); }