1 /* Bison Action Scanner -*- C -*-
3 Copyright (C) 2006-2012 Free Software Foundation, Inc.
5 This file is part of Bison, the GNU Compiler Compiler.
7 This program is free software: you can redistribute it and/or modify
8 it under the terms of the GNU General Public License as published by
9 the Free Software Foundation, either version 3 of the License, or
10 (at your option) any later version.
12 This program is distributed in the hope that it will be useful,
13 but WITHOUT ANY WARRANTY; without even the implied warranty of
14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 GNU General Public License for more details.
17 You should have received a copy of the GNU General Public License
18 along with this program. If not, see <http://www.gnu.org/licenses/>. */
20 %option debug nodefault noinput nounput noyywrap never-interactive
21 %option prefix="code_" outfile="lex.yy.c"
24 /* Work around a bug in flex 2.5.31. See Debian bug 333231
25 <http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=333231>. */
29 #define FLEX_PREFIX(Id) code_ ## Id
30 #include <src/flex-scanner.h>
32 #include <src/complain.h>
33 #include <src/reader.h>
34 #include <src/getargs.h>
35 #include <src/muscle-tab.h>
36 #include <src/scan-code.h>
37 #include <src/symlist.h>
40 #include <get-errno.h>
43 /* The current calling start condition: SC_RULE_ACTION or
45 # define YY_DECL static char *code_lex (code_props *self, int sc_context)
48 #define YY_USER_ACTION location_compute (loc, &loc->end, yytext, yyleng);
50 static char *fetch_type_name (char *cp, char const **type_name,
53 static void handle_action_dollar (symbol_list *rule, char *cp,
55 static void handle_action_at (symbol_list *rule, char *cp, location at_loc);
57 /* A string to be pushed to obstack after dollar/at has been handled. */
58 static char *ref_tail_fields;
60 static location the_location;
61 static location *loc = &the_location;
63 /* A string representing the most recent translation. */
64 static char *last_string;
66 /* True if an untyped $$ or $n was seen. */
67 static bool untyped_var_seen;
70 /* C and C++ comments in code. */
71 %x SC_COMMENT SC_LINE_COMMENT
72 /* Strings and characters in code. */
73 %x SC_STRING SC_CHARACTER
74 /* Whether in a rule or symbol action. Specifies the translation
76 %x SC_RULE_ACTION SC_SYMBOL_ACTION
79 /* POSIX says that a tag must be both an id and a C union member, but
80 historically almost any character is allowed in a tag. We disallow
81 NUL and newline, as this simplifies our implementation. */
84 /* Zero or more instances of backslash-newline. Following GCC, allow
85 white space between the backslash and the newline. */
86 splice (\\[ \f\t\v]*\n)*
88 /* C style identifier. Must start with letter. Will be used for
89 named symbol references. Shall be kept synchronized with
90 scan-gram.l "letter" and "id". */
91 letter [.abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ_]
92 id {letter}({letter}|[-0-9])*
93 ref -?[0-9]+|{id}|"["{id}"]"|"$"
98 /* Nesting level of the current code in braces. */
101 /* Whether a semicolon is probably needed.
103 The heuristic is that a semicolon is not needed after '{', '}',
104 ';', or a C preprocessor directive, and that whitespaces and
105 comments do not affect this flag. Note that '{' does not need a
106 semicolon because of '{}'. A semicolon may be needed before a
107 cpp directive, but don't bother.
109 While it is maintained in several start-conditions (factoring
110 opportunities), it is meaningful only for SC_RULE_ACTION. */
111 bool need_semicolon = false;
113 /* Whether in a C preprocessor directive. Don't use a start condition
114 for this because, at the end of strings and comments, we still need
115 to know whether we're in a directive. */
118 /* This scanner is special: it is invoked only once, henceforth
119 is expected to return only once. This initialization is
120 therefore done once per action to translate. */
121 aver (sc_context == SC_SYMBOL_ACTION
122 || sc_context == SC_RULE_ACTION
123 || sc_context == INITIAL);
127 /*------------------------------------------------------------.
128 | Scanning a C comment. The initial '/ *' is already eaten. |
129 `------------------------------------------------------------*/
133 "*"{splice}"/" STRING_GROW; BEGIN sc_context;
137 /*--------------------------------------------------------------.
138 | Scanning a line comment. The initial '//' is already eaten. |
139 `--------------------------------------------------------------*/
143 "\n" STRING_GROW; BEGIN sc_context;
144 {splice} STRING_GROW;
148 /*--------------------------------------------.
149 | Scanning user-code characters and strings. |
150 `--------------------------------------------*/
152 <SC_CHARACTER,SC_STRING>
154 {splice}|\\{splice}. STRING_GROW;
159 "'" STRING_GROW; BEGIN sc_context;
164 "\"" STRING_GROW; BEGIN sc_context;
168 <SC_RULE_ACTION,SC_SYMBOL_ACTION>
173 need_semicolon = true;
178 need_semicolon = true;
186 BEGIN SC_LINE_COMMENT;
189 complain (loc, Wother, _("stray '%s'"), yytext);
190 obstack_escape (&obstack_for_string, yytext);
191 need_semicolon = true;
194 obstack_escape (&obstack_for_string, yytext);
195 need_semicolon = true;
201 "$"("<"{tag}">")?{ref} {
202 ref_tail_fields = NULL;
203 handle_action_dollar (self->rule, yytext, *loc);
205 obstack_sgrow (&obstack_for_string, ref_tail_fields);
206 need_semicolon = true;
209 ref_tail_fields = NULL;
210 handle_action_at (self->rule, yytext, *loc);
212 obstack_sgrow (&obstack_for_string, ref_tail_fields);
213 need_semicolon = true;
216 ";" STRING_GROW; need_semicolon = false;
217 "{" STRING_GROW; ++braces_level; need_semicolon = false;
219 bool outer_brace = --braces_level == 0;
221 /* As an undocumented Bison extension, append ';' before the last
222 brace in braced code, so that the user code can omit trailing
223 ';'. But do not append ';' if emulating Yacc, since Yacc does
225 if (outer_brace && !yacc_flag && language_prio == default_prio
226 && skeleton_prio == default_prio && need_semicolon && ! in_cpp)
228 complain (loc, Wdeprecated, _("a ';' might be needed at the end of action code"));
229 complain (loc, Wdeprecated, _("future versions of Bison will not add the ';'"));
230 obstack_1grow (&obstack_for_string, ';');
234 need_semicolon = false;
237 /* Preprocessing directives should only be recognized at the beginning
238 of lines, allowing whitespace including comments, but in C/C++,
239 '#' can only be the start of preprocessor directives or within
240 '#define' directives anyway, so don't bother with begin of line. */
241 "#" STRING_GROW; in_cpp = true;
243 {splice} STRING_GROW;
244 [\n\r] STRING_GROW; if (in_cpp) in_cpp = need_semicolon = false;
246 . STRING_GROW; need_semicolon = true;
251 "$"("<"{tag}">")?"$" {
252 const char *type_name = NULL;
253 fetch_type_name (yytext + 1, &type_name, *loc)[-1] = 0;
254 obstack_sgrow (&obstack_for_string, "]b4_dollar_dollar(");
255 obstack_quote (&obstack_for_string, type_name);
256 obstack_sgrow (&obstack_for_string, ")[");
257 self->is_value_used = true;
260 obstack_sgrow (&obstack_for_string, "]b4_at_dollar[");
261 muscle_percent_define_ensure("locations", the_location, true);
268 /* Escape M4 quoting characters in C code. */
269 [$@\[\]] obstack_escape (&obstack_for_string, yytext);
271 /* By default, grow the string obstack with the input. */
274 /* End of processing. */
275 <<EOF>> STRING_FINISH; return last_string;
281 is_dot_or_dash (char ch)
283 return ch == '.' || ch == '-';
287 contains_dot_or_dash (const char* p)
290 if (is_dot_or_dash (*p))
295 /* Defines a variant of a symbolic name resolution. */
298 /* Index in symbol list. */
299 unsigned symbol_index;
301 /* Matched symbol id and loc. */
305 /* Hiding named reference. */
306 named_ref* hidden_by;
308 /* Error flags. May contain zero (no errors) or
309 a combination of VARIANT_* values. */
313 /* Set when the variant refers to a symbol hidden
314 by an explicit symbol reference. */
315 #define VARIANT_HIDDEN (1 << 0)
317 /* Set when the variant refers to a symbol containing
318 dots or dashes. Will require explicit bracketing. */
319 #define VARIANT_BAD_BRACKETING (1 << 1)
321 /* Set when the variant refers to a symbol which is
322 not visible from current midrule. */
323 #define VARIANT_NOT_VISIBLE_FROM_MIDRULE (1 << 2)
325 static variant *variant_table = NULL;
326 static unsigned variant_table_size = 0;
327 static unsigned variant_count = 0;
330 variant_table_grow (void)
333 if (variant_count > variant_table_size)
335 while (variant_count > variant_table_size)
336 variant_table_size = 2 * variant_table_size + 3;
337 variant_table = xnrealloc (variant_table, variant_table_size,
338 sizeof *variant_table);
340 return &variant_table[variant_count - 1];
344 variant_table_free (void)
346 free (variant_table);
347 variant_table = NULL;
348 variant_table_size = variant_count = 0;
352 find_prefix_end (const char *prefix, char *begin, char *end)
356 for (; *prefix && ptr != end; ++prefix, ++ptr)
367 variant_add (uniqstr id, location id_loc, unsigned symbol_index,
368 char *cp, char *cp_end, bool explicit_bracketing)
372 prefix_end = find_prefix_end (id, cp, cp_end);
374 (prefix_end == cp_end ||
375 (!explicit_bracketing && is_dot_or_dash (*prefix_end))))
377 variant *r = variant_table_grow ();
378 r->symbol_index = symbol_index;
390 get_at_spec(unsigned symbol_index)
392 static char at_buf[20];
393 if (symbol_index == 0)
394 strcpy (at_buf, "$$");
396 snprintf (at_buf, sizeof at_buf, "$%u", symbol_index);
401 show_sub_message (const char* cp, bool explicit_bracketing,
402 int midrule_rhs_index, char dollar_or_at,
403 unsigned indent, const variant *var)
405 const char *at_spec = get_at_spec (var->symbol_index);
408 complain_indent (&var->loc, silent, &indent,
409 _("refers to: %c%s at %s"), dollar_or_at,
413 static struct obstack msg_buf;
414 const char *tail = explicit_bracketing ? "" : cp + strlen (var->id);
415 const char *id = var->hidden_by ? var->hidden_by->id : var->id;
416 location id_loc = var->hidden_by ? var->hidden_by->loc : var->loc;
418 /* Create the explanation message. */
419 obstack_init (&msg_buf);
421 obstack_printf (&msg_buf, _("possibly meant: %c"), dollar_or_at);
422 if (contains_dot_or_dash (id))
423 obstack_printf (&msg_buf, "[%s]", id);
425 obstack_sgrow (&msg_buf, id);
426 obstack_sgrow (&msg_buf, tail);
428 if (var->err & VARIANT_HIDDEN)
430 obstack_printf (&msg_buf, _(", hiding %c"), dollar_or_at);
431 if (contains_dot_or_dash (var->id))
432 obstack_printf (&msg_buf, "[%s]", var->id);
434 obstack_sgrow (&msg_buf, var->id);
435 obstack_sgrow (&msg_buf, tail);
438 obstack_printf (&msg_buf, _(" at %s"), at_spec);
440 if (var->err & VARIANT_NOT_VISIBLE_FROM_MIDRULE)
441 obstack_printf (&msg_buf,
442 _(", cannot be accessed from mid-rule action at $%d"),
445 complain_indent (&id_loc, silent, &indent, "%s",
446 obstack_finish0 (&msg_buf));
447 obstack_free (&msg_buf, 0);
452 show_sub_messages (const char* cp, bool explicit_bracketing,
453 int midrule_rhs_index, char dollar_or_at,
458 for (i = 0; i < variant_count; ++i)
459 show_sub_message (cp, explicit_bracketing,
460 midrule_rhs_index, dollar_or_at,
461 indent, &variant_table[i]);
464 /* Returned from "parse_ref" when the reference
466 #define INVALID_REF (INT_MIN)
468 /* Returned from "parse_ref" when the reference
469 points to LHS ($$) of the current rule or midrule. */
470 #define LHS_REF (INT_MIN + 1)
472 /* Sub-messages indent. */
473 #define SUB_INDENT (4)
475 /* Parse named or positional reference. In case of positional
476 references, can return negative values for $-n "deep" stack
479 parse_ref (char *cp, symbol_list *rule, int rule_length,
480 int midrule_rhs_index, char *text, location text_loc,
485 bool explicit_bracketing;
487 unsigned valid_variants = 0;
488 unsigned valid_variant_index = 0;
493 if (c_isdigit (*cp) || (*cp == '-' && c_isdigit (* (cp + 1))))
495 long int num = strtol (cp, &cp, 10);
496 if (1 - INT_MAX + rule_length <= num && num <= rule_length)
500 complain (&text_loc, complaint, _("integer out of range: %s"),
508 /* Ignore the brackets. */
510 for (p = ++cp; *p != ']'; ++p)
514 explicit_bracketing = true;
518 /* Take all characters of the name. */
520 for (p = cp; *p; ++p)
521 if (is_dot_or_dash (*p))
526 for (p = cp; *p; ++p)
530 explicit_bracketing = false;
533 /* Add all relevant variants. */
535 unsigned symbol_index;
537 for (symbol_index = 0, l = rule; !symbol_list_null (l);
538 ++symbol_index, l = l->next)
541 if (l->content_type != SYMLIST_SYMBOL)
544 var = variant_add (l->content.sym->tag, l->sym_loc,
545 symbol_index, cp, cp_end, explicit_bracketing);
546 if (var && l->named_ref)
547 var->hidden_by = l->named_ref;
550 variant_add (l->named_ref->id, l->named_ref->loc,
551 symbol_index, cp, cp_end, explicit_bracketing);
556 for (i = 0; i < variant_count; ++i)
558 variant *var = &variant_table[i];
559 unsigned symbol_index = var->symbol_index;
561 /* Check visibility from mid-rule actions. */
562 if (midrule_rhs_index != 0
563 && (symbol_index == 0 || midrule_rhs_index < symbol_index))
564 var->err |= VARIANT_NOT_VISIBLE_FROM_MIDRULE;
566 /* Check correct bracketing. */
567 if (!explicit_bracketing && contains_dot_or_dash (var->id))
568 var->err |= VARIANT_BAD_BRACKETING;
570 /* Check using of hidden symbols. */
572 var->err |= VARIANT_HIDDEN;
576 valid_variant_index = i;
581 switch (valid_variants)
585 unsigned len = (explicit_bracketing || !ref_tail_fields) ?
586 cp_end - cp : ref_tail_fields - cp;
589 complain_indent (&text_loc, complaint, &indent,
590 _("invalid reference: %s"), quote (text));
591 indent += SUB_INDENT;
594 location sym_loc = text_loc;
595 sym_loc.start.column += 1;
596 sym_loc.end = sym_loc.start;
597 complain_indent (&sym_loc, complaint, &indent,
598 _("syntax error after '%c', expecting integer, "
599 "letter, '_', '[', or '$'"),
602 else if (midrule_rhs_index)
603 complain_indent (&rule->location, complaint, &indent,
604 _("symbol not found in production before $%d: "
606 midrule_rhs_index, len, cp);
608 complain_indent (&rule->location, complaint, &indent,
609 _("symbol not found in production: %.*s"),
612 if (variant_count > 0)
613 show_sub_messages (cp, explicit_bracketing, midrule_rhs_index,
614 dollar_or_at, indent);
620 if (variant_count > 1)
622 complain_indent (&text_loc, Wother, &indent,
623 _("misleading reference: %s"), quote (text));
624 show_sub_messages (cp, explicit_bracketing, midrule_rhs_index,
625 dollar_or_at, indent + SUB_INDENT);
628 unsigned symbol_index =
629 variant_table[valid_variant_index].symbol_index;
630 return (symbol_index == midrule_rhs_index) ? LHS_REF : symbol_index;
637 complain_indent (&text_loc, complaint, &indent,
638 _("ambiguous reference: %s"), quote (text));
639 show_sub_messages (cp, explicit_bracketing, midrule_rhs_index,
640 dollar_or_at, indent + SUB_INDENT);
649 /* Keeps track of the maximum number of semantic values to the left of
650 a handle (those referenced by $0, $-1, etc.) are required by the
651 semantic actions of this grammar. */
652 int max_left_semantic_context = 0;
655 /* If CP points to a typename (i.e., <.*?>), set TYPE_NAME to its
656 beginning (i.e., after the opening "<", and return the pointer
657 immediately after it. */
661 fetch_type_name (char *cp, char const **type_name,
670 /* The '>' symbol will be later replaced by '\0'. Original
671 'text' is needed for error messages. */
673 if (untyped_var_seen)
674 complain (&dollar_loc, complaint,
675 _("explicit type given in untyped grammar"));
681 /*------------------------------------------------------------------.
682 | TEXT is pointing to a wannabee semantic value (i.e., a '$'). |
684 | Possible inputs: $[<TYPENAME>]($|integer) |
686 | Output to OBSTACK_FOR_STRING a reference to this semantic value. |
687 `------------------------------------------------------------------*/
690 handle_action_dollar (symbol_list *rule, char *text, location dollar_loc)
692 char const *type_name = NULL;
694 symbol_list *effective_rule;
695 int effective_rule_length;
698 if (rule->midrule_parent_rule)
700 effective_rule = rule->midrule_parent_rule;
701 effective_rule_length = rule->midrule_parent_rhs_index - 1;
705 effective_rule = rule;
706 effective_rule_length = symbol_list_length (rule->next);
709 /* Get the type name if explicit. */
710 cp = fetch_type_name (cp, &type_name, dollar_loc);
712 n = parse_ref (cp, effective_rule, effective_rule_length,
713 rule->midrule_parent_rhs_index, text, dollar_loc, '$');
726 type_name = symbol_list_n_type_name_get (rule, dollar_loc, 0);
730 if (union_seen | tag_seen)
732 if (rule->midrule_parent_rule)
733 complain (&dollar_loc, complaint,
734 _("$$ for the midrule at $%d of %s"
735 " has no declared type"),
736 rule->midrule_parent_rhs_index,
737 quote (effective_rule->content.sym->tag));
739 complain (&dollar_loc, complaint,
740 _("$$ of %s has no declared type"),
741 quote (rule->content.sym->tag));
744 untyped_var_seen = true;
747 obstack_sgrow (&obstack_for_string, "]b4_lhs_value(");
748 obstack_quote (&obstack_for_string, type_name);
749 obstack_sgrow (&obstack_for_string, ")[");
750 rule->action_props.is_value_used = true;
754 if (max_left_semantic_context < 1 - n)
755 max_left_semantic_context = 1 - n;
756 if (!type_name && 0 < n)
758 symbol_list_n_type_name_get (effective_rule, dollar_loc, n);
761 if (union_seen | tag_seen)
762 complain (&dollar_loc, complaint,
763 _("$%s of %s has no declared type"), cp,
764 quote (effective_rule->content.sym->tag));
766 untyped_var_seen = true;
769 obstack_printf (&obstack_for_string,
770 "]b4_rhs_value(%d, %d, ", effective_rule_length, n);
771 obstack_quote (&obstack_for_string, type_name);
772 obstack_sgrow (&obstack_for_string, ")[");
774 symbol_list_n_get (effective_rule, n)->action_props.is_value_used =
781 /*------------------------------------------------------.
782 | TEXT is a location token (i.e., a '@...'). Output to |
783 | OBSTACK_FOR_STRING a reference to this location. |
784 `------------------------------------------------------*/
787 handle_action_at (symbol_list *rule, char *text, location at_loc)
790 symbol_list *effective_rule;
791 int effective_rule_length;
794 if (rule->midrule_parent_rule)
796 effective_rule = rule->midrule_parent_rule;
797 effective_rule_length = rule->midrule_parent_rhs_index - 1;
801 effective_rule = rule;
802 effective_rule_length = symbol_list_length (rule->next);
805 muscle_percent_define_ensure("locations", at_loc, true);
807 n = parse_ref (cp, effective_rule, effective_rule_length,
808 rule->midrule_parent_rhs_index, text, at_loc, '@');
815 obstack_sgrow (&obstack_for_string, "]b4_lhs_location[");
819 obstack_printf (&obstack_for_string, "]b4_rhs_location(%d, %d)[",
820 effective_rule_length, n);
826 /*-------------------------.
827 | Initialize the scanner. |
828 `-------------------------*/
830 /* Translate the dollars and ats in \a self, in the context \a sc_context
831 (SC_RULE_ACTION, SC_SYMBOL_ACTION, INITIAL). */
834 translate_action (code_props *self, int sc_context)
837 static bool initialized = false;
840 obstack_init (&obstack_for_string);
845 loc->start = loc->end = self->location.start;
846 yy_switch_to_buffer (yy_scan_string (self->code));
847 res = code_lex (self, sc_context);
848 yy_delete_buffer (YY_CURRENT_BUFFER);
853 /*------------------------------------------------------------------------.
854 | Implementation of the public interface as documented in "scan-code.h". |
855 `------------------------------------------------------------------------*/
858 code_props_none_init (code_props *self)
860 *self = code_props_none;
863 code_props code_props_none = CODE_PROPS_NONE_INIT;
866 code_props_plain_init (code_props *self, char const *code,
869 code_props_none_init (self);
870 self->kind = CODE_PROPS_PLAIN;
872 self->location = code_loc;
876 code_props_symbol_action_init (code_props *self, char const *code,
879 code_props_none_init (self);
880 self->kind = CODE_PROPS_SYMBOL_ACTION;
882 self->location = code_loc;
886 code_props_rule_action_init (code_props *self, char const *code,
887 location code_loc, symbol_list *rule,
888 named_ref *name, bool is_predicate)
890 code_props_none_init (self);
891 self->kind = CODE_PROPS_RULE_ACTION;
893 self->location = code_loc;
895 self->named_ref = name;
896 self->is_predicate = is_predicate;
900 code_props_translate_code (code_props *self)
904 case CODE_PROPS_NONE:
906 case CODE_PROPS_PLAIN:
907 self->code = translate_action (self, INITIAL);
909 case CODE_PROPS_SYMBOL_ACTION:
910 self->code = translate_action (self, SC_SYMBOL_ACTION);
912 case CODE_PROPS_RULE_ACTION:
913 self->code = translate_action (self, SC_RULE_ACTION);
919 code_scanner_last_string_free (void)
925 code_scanner_free (void)
927 obstack_free (&obstack_for_string, 0);
928 variant_table_free ();
930 /* Reclaim Flex's buffers. */