X-Git-Url: https://git.saurik.com/bison.git/blobdiff_plain/d815ec4a6290e18fac9220438622f6dd27b3227f..0d43e6057b31caf96daa30d2dd3c07932b2c461a:/src/print.c diff --git a/src/print.c b/src/print.c index b117e752..e3795a94 100644 --- a/src/print.c +++ b/src/print.c @@ -1,6 +1,6 @@ /* Print information on generated parser, for bison, - Copyright (C) 1984, 1986, 1989, 2000-2005, 2007, 2009-2011 Free + Copyright (C) 1984, 1986, 1989, 2000-2005, 2007, 2009-2012 Free Software Foundation, Inc. This file is part of Bison, the GNU Compiler Compiler. @@ -22,7 +22,6 @@ #include "system.h" #include -#include #include "LR0.h" #include "closure.h" @@ -97,7 +96,7 @@ print_core (FILE *out, state *s) sp1 = sp = ritem + sitems[i]; while (*sp >= 0) - sp++; + sp++; r = item_number_as_rule_number (*sp); @@ -105,15 +104,15 @@ print_core (FILE *out, state *s) previous_lhs = rules[r].lhs; for (sp = rules[r].rhs; sp < sp1; sp++) - fprintf (out, " %s", symbols[*sp]->tag); + fprintf (out, " %s", symbols[*sp]->tag); fputs (" .", out); for (/* Nothing */; *sp >= 0; ++sp) - fprintf (out, " %s", symbols[*sp]->tag); + fprintf (out, " %s", symbols[*sp]->tag); /* Display the lookahead tokens? */ if (report_flag & report_lookahead_tokens && item_number_is_rule_number (*sp1)) - state_rule_lookahead_tokens_print (s, &rules[r], out); + state_rule_lookahead_tokens_print (s, &rules[r], out); fputc ('\n', out); } @@ -135,10 +134,10 @@ print_transitions (state *s, FILE *out, bool display_transitions_p) /* Compute the width of the lookahead token column. */ for (i = 0; i < trans->num; i++) if (!TRANSITION_IS_DISABLED (trans, i) - && TRANSITION_IS_SHIFT (trans, i) == display_transitions_p) + && TRANSITION_IS_SHIFT (trans, i) == display_transitions_p) { - symbol *sym = symbols[TRANSITION_SYMBOL (trans, i)]; - max_length (&width, sym->tag); + symbol *sym = symbols[TRANSITION_SYMBOL (trans, i)]; + max_length (&width, sym->tag); } /* Nothing to report. */ @@ -151,20 +150,20 @@ print_transitions (state *s, FILE *out, bool display_transitions_p) /* Report lookahead tokens and shifts. */ for (i = 0; i < trans->num; i++) if (!TRANSITION_IS_DISABLED (trans, i) - && TRANSITION_IS_SHIFT (trans, i) == display_transitions_p) + && TRANSITION_IS_SHIFT (trans, i) == display_transitions_p) { - symbol *sym = symbols[TRANSITION_SYMBOL (trans, i)]; - const char *tag = sym->tag; - state *s1 = trans->states[i]; - int j; - - fprintf (out, " %s", tag); - for (j = width - strlen (tag); j > 0; --j) - fputc (' ', out); - if (display_transitions_p) - fprintf (out, _("shift, and go to state %d\n"), s1->number); - else - fprintf (out, _("go to state %d\n"), s1->number); + symbol *sym = symbols[TRANSITION_SYMBOL (trans, i)]; + const char *tag = sym->tag; + state *s1 = trans->states[i]; + int j; + + fprintf (out, " %s", tag); + for (j = width - strlen (tag); j > 0; --j) + fputc (' ', out); + if (display_transitions_p) + fprintf (out, _("shift, and go to state %d\n"), s1->number); + else + fprintf (out, _("go to state %d\n"), s1->number); } } @@ -196,12 +195,12 @@ print_errs (FILE *out, state *s) for (i = 0; i < errp->num; ++i) if (errp->symbols[i]) { - const char *tag = errp->symbols[i]->tag; - int j; - fprintf (out, " %s", tag); - for (j = width - strlen (tag); j > 0; --j) - fputc (' ', out); - fputs (_("error (nonassociative)\n"), out); + const char *tag = errp->symbols[i]->tag; + int j; + fprintf (out, " %s", tag); + for (j = width - strlen (tag); j > 0; --j) + fputc (' ', out); + fputs (_("error (nonassociative)\n"), out); } } @@ -214,8 +213,8 @@ print_errs (FILE *out, state *s) static void print_reduction (FILE *out, size_t width, - const char *lookahead_token, - rule *r, bool enabled) + const char *lookahead_token, + rule *r, bool enabled) { int j; fprintf (out, " %s", lookahead_token); @@ -267,22 +266,22 @@ print_reductions (FILE *out, state *s) if (reds->lookahead_tokens) for (i = 0; i < ntokens; i++) { - bool count = bitset_test (no_reduce_set, i); - - for (j = 0; j < reds->num; ++j) - if (bitset_test (reds->lookahead_tokens[j], i)) - { - if (! count) - { - if (reds->rules[j] != default_reduction) - max_length (&width, symbols[i]->tag); - count = true; - } - else - { - max_length (&width, symbols[i]->tag); - } - } + bool count = bitset_test (no_reduce_set, i); + + for (j = 0; j < reds->num; ++j) + if (bitset_test (reds->lookahead_tokens[j], i)) + { + if (! count) + { + if (reds->rules[j] != default_reduction) + max_length (&width, symbols[i]->tag); + count = true; + } + else + { + max_length (&width, symbols[i]->tag); + } + } } /* Nothing to report. */ @@ -296,40 +295,40 @@ print_reductions (FILE *out, state *s) if (reds->lookahead_tokens) for (i = 0; i < ntokens; i++) { - bool defaulted = false; - bool count = bitset_test (no_reduce_set, i); + bool defaulted = false; + bool count = bitset_test (no_reduce_set, i); if (count) default_reduction_only = false; - for (j = 0; j < reds->num; ++j) - if (bitset_test (reds->lookahead_tokens[j], i)) - { - if (! count) - { - if (reds->rules[j] != default_reduction) + for (j = 0; j < reds->num; ++j) + if (bitset_test (reds->lookahead_tokens[j], i)) + { + if (! count) + { + if (reds->rules[j] != default_reduction) { default_reduction_only = false; print_reduction (out, width, symbols[i]->tag, reds->rules[j], true); } - else - defaulted = true; - count = true; - } - else - { + else + defaulted = true; + count = true; + } + else + { default_reduction_only = false; - if (defaulted) - print_reduction (out, width, - symbols[i]->tag, - default_reduction, true); - defaulted = false; - print_reduction (out, width, - symbols[i]->tag, - reds->rules[j], false); - } - } + if (defaulted) + print_reduction (out, width, + symbols[i]->tag, + default_reduction, true); + defaulted = false; + print_reduction (out, width, + symbols[i]->tag, + reds->rules[j], false); + } + } } if (default_reduction) @@ -337,8 +336,8 @@ print_reductions (FILE *out, state *s) char *default_reductions = muscle_percent_define_get ("lr.default-reductions"); print_reduction (out, width, _("$default"), default_reduction, true); - aver (0 == strcmp (default_reductions, "full") - || (0 == strcmp (default_reductions, "consistent") + aver (STREQ (default_reductions, "most") + || (STREQ (default_reductions, "consistent") && default_reduction_only) || (reds->num == 1 && reds->rules[0]->number == 0)); free (default_reductions); @@ -386,15 +385,15 @@ print_state (FILE *out, state *s) | Print information on the whole grammar. | `-----------------------------------------*/ -#define END_TEST(End) \ -do { \ - if (column + strlen(buffer) > (End)) \ - { \ - fprintf (out, "%s\n ", buffer); \ - column = 3; \ - buffer[0] = 0; \ - } \ -} while (0) +#define END_TEST(End) \ + do { \ + if (column + strlen (buffer) > (End)) \ + { \ + fprintf (out, "%s\n ", buffer); \ + column = 3; \ + buffer[0] = 0; \ + } \ + } while (0) static void @@ -411,25 +410,25 @@ print_grammar (FILE *out) for (i = 0; i < max_user_token_number + 1; i++) if (token_translations[i] != undeftoken->number) { - const char *tag = symbols[token_translations[i]]->tag; - rule_number r; - item_number *rhsp; - - buffer[0] = 0; - column = strlen (tag); - fputs (tag, out); - END_TEST (65); - sprintf (buffer, " (%d)", i); - - for (r = 0; r < nrules; r++) - for (rhsp = rules[r].rhs; *rhsp >= 0; rhsp++) - if (item_number_as_symbol_number (*rhsp) == token_translations[i]) - { - END_TEST (65); - sprintf (buffer + strlen (buffer), " %d", r); - break; - } - fprintf (out, "%s\n", buffer); + const char *tag = symbols[token_translations[i]]->tag; + rule_number r; + item_number *rhsp; + + buffer[0] = 0; + column = strlen (tag); + fputs (tag, out); + END_TEST (65); + sprintf (buffer, " (%d)", i); + + for (r = 0; r < nrules; r++) + for (rhsp = rules[r].rhs; *rhsp >= 0; rhsp++) + if (item_number_as_symbol_number (*rhsp) == token_translations[i]) + { + END_TEST (65); + sprintf (buffer + strlen (buffer), " %d", r); + break; + } + fprintf (out, "%s\n", buffer); } fputs ("\n\n", out); @@ -442,17 +441,17 @@ print_grammar (FILE *out) const char *tag = symbols[i]->tag; for (r = 0; r < nrules; r++) - { - item_number *rhsp; - if (rules[r].lhs->number == i) - left_count++; - for (rhsp = rules[r].rhs; *rhsp >= 0; rhsp++) - if (item_number_as_symbol_number (*rhsp) == i) - { - right_count++; - break; - } - } + { + item_number *rhsp; + if (rules[r].lhs->number == i) + left_count++; + for (rhsp = rules[r].rhs; *rhsp >= 0; rhsp++) + if (item_number_as_symbol_number (*rhsp) == i) + { + right_count++; + break; + } + } buffer[0] = 0; fputs (tag, out); @@ -461,38 +460,38 @@ print_grammar (FILE *out) END_TEST (0); if (left_count > 0) - { - END_TEST (65); - sprintf (buffer + strlen (buffer), _(" on left:")); - - for (r = 0; r < nrules; r++) - { - if (rules[r].lhs->number == i) - { - END_TEST (65); - sprintf (buffer + strlen (buffer), " %d", r); - } - } - } + { + END_TEST (65); + sprintf (buffer + strlen (buffer), _(" on left:")); + + for (r = 0; r < nrules; r++) + { + if (rules[r].lhs->number == i) + { + END_TEST (65); + sprintf (buffer + strlen (buffer), " %d", r); + } + } + } if (right_count > 0) - { - if (left_count > 0) - sprintf (buffer + strlen (buffer), ","); - END_TEST (65); - sprintf (buffer + strlen (buffer), _(" on right:")); - for (r = 0; r < nrules; r++) - { - item_number *rhsp; - for (rhsp = rules[r].rhs; *rhsp >= 0; rhsp++) - if (item_number_as_symbol_number (*rhsp) == i) - { - END_TEST (65); - sprintf (buffer + strlen (buffer), " %d", r); - break; - } - } - } + { + if (left_count > 0) + sprintf (buffer + strlen (buffer), ","); + END_TEST (65); + sprintf (buffer + strlen (buffer), _(" on right:")); + for (r = 0; r < nrules; r++) + { + item_number *rhsp; + for (rhsp = rules[r].rhs; *rhsp >= 0; rhsp++) + if (item_number_as_symbol_number (*rhsp) == i) + { + END_TEST (65); + sprintf (buffer + strlen (buffer), " %d", r); + break; + } + } + } fprintf (out, "%s\n", buffer); } } @@ -508,7 +507,7 @@ print_results (void) reduce_output (out); grammar_rules_partial_print (out, - _("Rules useless in parser due to conflicts"), + _("Rules useless in parser due to conflicts"), rule_useless_in_parser_p); conflicts_output (out);