1 /* Bison Action Scanner -*- C -*-
3 Copyright (C) 2006-2013 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 "flex-scanner.h"
35 #include "scan-code.h"
39 #include <get-errno.h>
42 /* The current calling start condition: SC_RULE_ACTION or
44 # define YY_DECL static char *code_lex (code_props *self, int sc_context)
47 #define YY_USER_ACTION location_compute (loc, &loc->end, yytext, yyleng);
49 static char *fetch_type_name (char *cp, char const **type_name,
52 static void handle_action_dollar (symbol_list *rule, char *cp,
54 static void handle_action_at (symbol_list *rule, char *cp, location at_loc);
56 /* A string to be pushed to obstack after dollar/at has been handled. */
57 static char *ref_tail_fields;
59 static location the_location;
60 static location *loc = &the_location;
62 /* A string representing the most recent translation. */
63 static char *last_string;
65 /* True if an untyped $$ or $n was seen. */
66 static bool untyped_var_seen;
69 /* C and C++ comments in code. */
70 %x SC_COMMENT SC_LINE_COMMENT
71 /* Strings and characters in code. */
72 %x SC_STRING SC_CHARACTER
73 /* Whether in a rule or symbol action. Specifies the translation
75 %x SC_RULE_ACTION SC_SYMBOL_ACTION
78 /* POSIX says that a tag must be both an id and a C union member, but
79 historically almost any character is allowed in a tag. We disallow
80 NUL and newline, as this simplifies our implementation. */
83 /* Zero or more instances of backslash-newline. Following GCC, allow
84 white space between the backslash and the newline. */
85 splice (\\[ \f\t\v]*\n)*
87 /* C style identifier. Must start with letter. Will be used for
88 named symbol references. Shall be kept synchronized with
89 scan-gram.l "letter" and "id". */
90 letter [.abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ_]
91 id {letter}({letter}|[-0-9])*
92 ref -?[0-9]+|{id}|"["{id}"]"|"$"
97 /* Nesting level of the current code in braces. */
100 /* Whether a semicolon is probably needed.
102 The heuristic is that a semicolon is not needed after '{', '}',
103 ';', or a C preprocessor directive, and that whitespaces and
104 comments do not affect this flag. Note that '{' does not need a
105 semicolon because of '{}'. A semicolon may be needed before a
106 cpp directive, but don't bother.
108 While it is maintained in several start-conditions (factoring
109 opportunities), it is meaningful only for SC_RULE_ACTION. */
110 bool need_semicolon = false;
112 /* Whether in a C preprocessor directive. Don't use a start condition
113 for this because, at the end of strings and comments, we still need
114 to know whether we're in a directive. */
117 /* This scanner is special: it is invoked only once, henceforth
118 is expected to return only once. This initialization is
119 therefore done once per action to translate. */
120 aver (sc_context == SC_SYMBOL_ACTION
121 || sc_context == SC_RULE_ACTION
122 || sc_context == INITIAL);
126 /*------------------------------------------------------------.
127 | Scanning a C comment. The initial '/ *' is already eaten. |
128 `------------------------------------------------------------*/
132 "*"{splice}"/" STRING_GROW; BEGIN sc_context;
136 /*--------------------------------------------------------------.
137 | Scanning a line comment. The initial '//' is already eaten. |
138 `--------------------------------------------------------------*/
142 "\n" STRING_GROW; BEGIN sc_context;
143 {splice} STRING_GROW;
147 /*--------------------------------------------.
148 | Scanning user-code characters and strings. |
149 `--------------------------------------------*/
151 <SC_CHARACTER,SC_STRING>
153 {splice}|\\{splice}. STRING_GROW;
158 "'" STRING_GROW; BEGIN sc_context;
163 "\"" STRING_GROW; BEGIN sc_context;
167 <SC_RULE_ACTION,SC_SYMBOL_ACTION>
172 need_semicolon = true;
177 need_semicolon = true;
185 BEGIN SC_LINE_COMMENT;
188 warn_at (*loc, _("stray '%s'"), yytext);
189 obstack_escape (&obstack_for_string, yytext);
190 need_semicolon = true;
193 obstack_escape (&obstack_for_string, yytext);
194 need_semicolon = true;
200 "$"("<"{tag}">")?{ref} {
201 ref_tail_fields = NULL;
202 handle_action_dollar (self->rule, yytext, *loc);
204 obstack_sgrow (&obstack_for_string, ref_tail_fields);
205 need_semicolon = true;
208 ref_tail_fields = NULL;
209 handle_action_at (self->rule, yytext, *loc);
211 obstack_sgrow (&obstack_for_string, ref_tail_fields);
212 need_semicolon = true;
215 ";" STRING_GROW; need_semicolon = false;
216 "{" STRING_GROW; ++braces_level; need_semicolon = false;
218 bool outer_brace = --braces_level == 0;
220 /* As an undocumented Bison extension, append ';' before the last
221 brace in braced code, so that the user code can omit trailing
222 ';'. But do not append ';' if emulating Yacc, since Yacc does
223 not append one. This is deprecated since release 2.4.1. */
224 if (outer_brace && !yacc_flag && language_prio == default_prio
225 && skeleton_prio == default_prio && need_semicolon && ! in_cpp)
227 unsigned int indent = 0;
228 warn_at_indent (*loc, &indent,
229 _("a ';' might be needed at the end of action code"));
230 indent += SUB_INDENT;
231 warn_at_indent (*loc, &indent,
232 _("future versions of Bison will not add the ';'"));
233 obstack_1grow (&obstack_for_string, ';');
237 need_semicolon = false;
240 /* Preprocessing directives should only be recognized at the beginning
241 of lines, allowing whitespace including comments, but in C/C++,
242 '#' can only be the start of preprocessor directives or within
243 '#define' directives anyway, so don't bother with begin of line. */
244 "#" STRING_GROW; in_cpp = true;
246 {splice} STRING_GROW;
247 [\n\r] STRING_GROW; if (in_cpp) in_cpp = need_semicolon = false;
250 /* YYFAIL is undocumented and was formally deprecated in Bison
253 STRING_GROW; need_semicolon = true;
254 warn_at (*loc, _("use of YYFAIL, which is deprecated and will be"
258 /* The sole purpose of this is to make sure identifiers that merely
259 contain YYFAIL don't produce the above warning. */
260 [A-Za-z_][0-9A-Za-z_]* STRING_GROW; need_semicolon = true;
262 . STRING_GROW; need_semicolon = true;
267 "$"("<"{tag}">")?"$" {
268 const char *type_name = NULL;
269 fetch_type_name (yytext + 1, &type_name, *loc)[-1] = 0;
270 obstack_sgrow (&obstack_for_string, "]b4_dollar_dollar(");
271 obstack_quote (&obstack_for_string, type_name);
272 obstack_sgrow (&obstack_for_string, ")[");
273 self->is_value_used = true;
276 obstack_sgrow (&obstack_for_string, "]b4_at_dollar[");
277 locations_flag = true;
284 /* Escape M4 quoting characters in C code. */
285 [$@\[\]] obstack_escape (&obstack_for_string, yytext);
287 /* By default, grow the string obstack with the input. */
290 /* End of processing. */
291 <<EOF>> STRING_FINISH; return last_string;
297 is_dot_or_dash (char ch)
299 return ch == '.' || ch == '-';
303 contains_dot_or_dash (const char* p)
306 if (is_dot_or_dash (*p))
311 /* Defines a variant of a symbolic name resolution. */
314 /* Index in symbol list. */
315 unsigned symbol_index;
317 /* Matched symbol id and loc. */
321 /* Hiding named reference. */
322 named_ref* hidden_by;
324 /* Error flags. May contain zero (no errors) or
325 a combination of VARIANT_* values. */
329 /* Set when the variant refers to a symbol hidden
330 by an explicit symbol reference. */
331 #define VARIANT_HIDDEN (1 << 0)
333 /* Set when the variant refers to a symbol containing
334 dots or dashes. Will require explicit bracketing. */
335 #define VARIANT_BAD_BRACKETING (1 << 1)
337 /* Set when the variant refers to a symbol which is
338 not visible from current midrule. */
339 #define VARIANT_NOT_VISIBLE_FROM_MIDRULE (1 << 2)
341 static variant *variant_table = NULL;
342 static unsigned variant_table_size = 0;
343 static unsigned variant_count = 0;
346 variant_table_grow (void)
349 if (variant_count > variant_table_size)
351 while (variant_count > variant_table_size)
352 variant_table_size = 2 * variant_table_size + 3;
353 variant_table = xnrealloc (variant_table, variant_table_size,
354 sizeof *variant_table);
356 return &variant_table[variant_count - 1];
360 variant_table_free (void)
362 free (variant_table);
363 variant_table = NULL;
364 variant_table_size = variant_count = 0;
368 find_prefix_end (const char *prefix, char *begin, char *end)
372 for (; *prefix && ptr != end; ++prefix, ++ptr)
383 variant_add (uniqstr id, location id_loc, unsigned symbol_index,
384 char *cp, char *cp_end, bool explicit_bracketing)
388 prefix_end = find_prefix_end (id, cp, cp_end);
390 (prefix_end == cp_end ||
391 (!explicit_bracketing && is_dot_or_dash (*prefix_end))))
393 variant *r = variant_table_grow ();
394 r->symbol_index = symbol_index;
406 get_at_spec (unsigned symbol_index)
408 static char at_buf[20];
409 if (symbol_index == 0)
410 strcpy (at_buf, "$$");
412 snprintf (at_buf, sizeof at_buf, "$%u", symbol_index);
417 show_sub_messages (const char* cp, bool explicit_bracketing,
418 int midrule_rhs_index, char dollar_or_at,
419 bool is_warning, unsigned indent)
423 for (i = 0; i < variant_count; ++i)
425 const variant *var = &variant_table[i];
426 const char *at_spec = get_at_spec (var->symbol_index);
431 warn_at_indent (var->loc, &indent, _("refers to: %c%s at %s"),
432 dollar_or_at, var->id, at_spec);
434 complain_at_indent (var->loc, &indent, _("refers to: %c%s at %s"),
435 dollar_or_at, var->id, at_spec);
439 static struct obstack msg_buf;
440 const char *tail = explicit_bracketing ? "" :
441 cp + strlen (var->id);
442 const char *id = var->hidden_by ? var->hidden_by->id :
444 location id_loc = var->hidden_by ? var->hidden_by->loc :
447 /* Create the explanation message. */
448 obstack_init (&msg_buf);
450 obstack_printf (&msg_buf, _("possibly meant: %c"), dollar_or_at);
451 if (contains_dot_or_dash (id))
452 obstack_printf (&msg_buf, "[%s]", id);
454 obstack_sgrow (&msg_buf, id);
455 obstack_sgrow (&msg_buf, tail);
457 if (var->err & VARIANT_HIDDEN)
459 obstack_printf (&msg_buf, _(", hiding %c"), dollar_or_at);
460 if (contains_dot_or_dash (var->id))
461 obstack_printf (&msg_buf, "[%s]", var->id);
463 obstack_sgrow (&msg_buf, var->id);
464 obstack_sgrow (&msg_buf, tail);
467 obstack_printf (&msg_buf, _(" at %s"), at_spec);
469 if (var->err & VARIANT_NOT_VISIBLE_FROM_MIDRULE)
472 _(", cannot be accessed from mid-rule action at $%d");
473 obstack_printf (&msg_buf, format, midrule_rhs_index);
476 obstack_1grow (&msg_buf, '\0');
478 warn_at_indent (id_loc, &indent, "%s",
479 (char *) obstack_finish (&msg_buf));
481 complain_at_indent (id_loc, &indent, "%s",
482 (char *) obstack_finish (&msg_buf));
483 obstack_free (&msg_buf, 0);
488 /* Returned from "parse_ref" when the reference
490 #define INVALID_REF (INT_MIN)
492 /* Returned from "parse_ref" when the reference
493 points to LHS ($$) of the current rule or midrule. */
494 #define LHS_REF (INT_MIN + 1)
496 /* Parse named or positional reference. In case of positional
497 references, can return negative values for $-n "deep" stack
500 parse_ref (char *cp, symbol_list *rule, int rule_length,
501 int midrule_rhs_index, char *text, location text_loc,
506 bool explicit_bracketing;
508 unsigned valid_variants = 0;
509 unsigned valid_variant_index = 0;
514 if (c_isdigit (*cp) || (*cp == '-' && c_isdigit (* (cp + 1))))
516 long int num = strtol (cp, &cp, 10);
517 if (1 - INT_MAX + rule_length <= num && num <= rule_length)
521 complain_at (text_loc, _("integer out of range: %s"),
529 /* Ignore the brackets. */
531 for (p = ++cp; *p != ']'; ++p)
535 explicit_bracketing = true;
539 /* Take all characters of the name. */
541 for (p = cp; *p; ++p)
542 if (is_dot_or_dash (*p))
547 for (p = cp; *p; ++p)
551 explicit_bracketing = false;
554 /* Add all relevant variants. */
556 unsigned symbol_index;
558 for (symbol_index = 0, l = rule; !symbol_list_null (l);
559 ++symbol_index, l = l->next)
562 if (l->content_type != SYMLIST_SYMBOL)
565 var = variant_add (l->content.sym->tag, l->sym_loc,
566 symbol_index, cp, cp_end, explicit_bracketing);
567 if (var && l->named_ref)
568 var->hidden_by = l->named_ref;
571 variant_add (l->named_ref->id, l->named_ref->loc,
572 symbol_index, cp, cp_end, explicit_bracketing);
577 for (i = 0; i < variant_count; ++i)
579 variant *var = &variant_table[i];
580 unsigned symbol_index = var->symbol_index;
582 /* Check visibility from mid-rule actions. */
583 if (midrule_rhs_index != 0
584 && (symbol_index == 0 || midrule_rhs_index < symbol_index))
585 var->err |= VARIANT_NOT_VISIBLE_FROM_MIDRULE;
587 /* Check correct bracketing. */
588 if (!explicit_bracketing && contains_dot_or_dash (var->id))
589 var->err |= VARIANT_BAD_BRACKETING;
591 /* Check using of hidden symbols. */
593 var->err |= VARIANT_HIDDEN;
597 valid_variant_index = i;
602 switch (valid_variants)
606 unsigned len = (explicit_bracketing || !ref_tail_fields) ?
607 cp_end - cp : ref_tail_fields - cp;
610 complain_at_indent (text_loc, &indent, _("invalid reference: %s"),
612 indent += SUB_INDENT;
615 location sym_loc = text_loc;
616 sym_loc.start.column += 1;
617 sym_loc.end = sym_loc.start;
619 _("syntax error after '%c', expecting integer, letter,"
620 " '_', '[', or '$'");
621 complain_at_indent (sym_loc, &indent, format, dollar_or_at);
623 else if (midrule_rhs_index)
626 _("symbol not found in production before $%d: %.*s");
627 complain_at_indent (rule->location, &indent, format,
628 midrule_rhs_index, len, cp);
633 _("symbol not found in production: %.*s");
634 complain_at_indent (rule->location, &indent, format,
638 if (variant_count > 0)
639 show_sub_messages (cp, explicit_bracketing, midrule_rhs_index,
640 dollar_or_at, false, indent);
646 if (variant_count > 1)
648 warn_at_indent (text_loc, &indent, _("misleading reference: %s"),
650 show_sub_messages (cp, explicit_bracketing, midrule_rhs_index,
651 dollar_or_at, true, indent + SUB_INDENT);
654 unsigned symbol_index =
655 variant_table[valid_variant_index].symbol_index;
656 return (symbol_index == midrule_rhs_index) ? LHS_REF : symbol_index;
663 complain_at_indent (text_loc, &indent, _("ambiguous reference: %s"),
665 show_sub_messages (cp, explicit_bracketing, midrule_rhs_index,
666 dollar_or_at, false, indent + SUB_INDENT);
675 /* Keeps track of the maximum number of semantic values to the left of
676 a handle (those referenced by $0, $-1, etc.) are required by the
677 semantic actions of this grammar. */
678 int max_left_semantic_context = 0;
681 /* If CP points to a typename (i.e., <.*?>), set TYPE_NAME to its
682 beginning (i.e., after the opening "<", and return the pointer
683 immediately after it. */
687 fetch_type_name (char *cp, char const **type_name,
696 /* The '>' symbol will be later replaced by '\0'. Original
697 'text' is needed for error messages. */
699 if (untyped_var_seen)
700 complain_at (dollar_loc, _("explicit type given in untyped grammar"));
706 /*------------------------------------------------------------------.
707 | TEXT is pointing to a wannabee semantic value (i.e., a '$'). |
709 | Possible inputs: $[<TYPENAME>]($|integer) |
711 | Output to OBSTACK_FOR_STRING a reference to this semantic value. |
712 `------------------------------------------------------------------*/
715 handle_action_dollar (symbol_list *rule, char *text, location dollar_loc)
717 char const *type_name = NULL;
719 symbol_list *effective_rule;
720 int effective_rule_length;
723 if (rule->midrule_parent_rule)
725 effective_rule = rule->midrule_parent_rule;
726 effective_rule_length = rule->midrule_parent_rhs_index - 1;
730 effective_rule = rule;
731 effective_rule_length = symbol_list_length (rule->next);
734 /* Get the type name if explicit. */
735 cp = fetch_type_name (cp, &type_name, dollar_loc);
737 n = parse_ref (cp, effective_rule, effective_rule_length,
738 rule->midrule_parent_rhs_index, text, dollar_loc, '$');
751 type_name = symbol_list_n_type_name_get (rule, dollar_loc, 0);
755 if (union_seen | tag_seen)
757 if (rule->midrule_parent_rule)
758 complain_at (dollar_loc,
759 _("$$ for the midrule at $%d of %s"
760 " has no declared type"),
761 rule->midrule_parent_rhs_index,
762 quote (effective_rule->content.sym->tag));
764 complain_at (dollar_loc, _("$$ of %s has no declared type"),
765 quote (rule->content.sym->tag));
768 untyped_var_seen = true;
771 obstack_sgrow (&obstack_for_string, "]b4_lhs_value(");
772 obstack_quote (&obstack_for_string, type_name);
773 obstack_sgrow (&obstack_for_string, ")[");
774 rule->action_props.is_value_used = true;
778 if (max_left_semantic_context < 1 - n)
779 max_left_semantic_context = 1 - n;
780 if (!type_name && 0 < n)
782 symbol_list_n_type_name_get (effective_rule, dollar_loc, n);
785 if (union_seen | tag_seen)
786 complain_at (dollar_loc, _("$%s of %s has no declared type"),
787 cp, quote (effective_rule->content.sym->tag));
789 untyped_var_seen = true;
792 obstack_printf (&obstack_for_string,
793 "]b4_rhs_value(%d, %d, ", effective_rule_length, n);
794 obstack_quote (&obstack_for_string, type_name);
795 obstack_sgrow (&obstack_for_string, ")[");
797 symbol_list_n_get (effective_rule, n)->action_props.is_value_used =
804 /*------------------------------------------------------.
805 | TEXT is a location token (i.e., a '@...'). Output to |
806 | OBSTACK_FOR_STRING a reference to this location. |
807 `------------------------------------------------------*/
810 handle_action_at (symbol_list *rule, char *text, location at_loc)
813 symbol_list *effective_rule;
814 int effective_rule_length;
817 if (rule->midrule_parent_rule)
819 effective_rule = rule->midrule_parent_rule;
820 effective_rule_length = rule->midrule_parent_rhs_index - 1;
824 effective_rule = rule;
825 effective_rule_length = symbol_list_length (rule->next);
828 locations_flag = true;
830 n = parse_ref (cp, effective_rule, effective_rule_length,
831 rule->midrule_parent_rhs_index, text, at_loc, '@');
838 obstack_sgrow (&obstack_for_string, "]b4_lhs_location[");
842 obstack_printf (&obstack_for_string, "]b4_rhs_location(%d, %d)[",
843 effective_rule_length, n);
849 /*-------------------------.
850 | Initialize the scanner. |
851 `-------------------------*/
853 /* Translate the dollars and ats in \a self, in the context \a sc_context
854 (SC_RULE_ACTION, SC_SYMBOL_ACTION, INITIAL). */
857 translate_action (code_props *self, int sc_context)
860 static bool initialized = false;
863 obstack_init (&obstack_for_string);
868 loc->start = loc->end = self->location.start;
869 yy_switch_to_buffer (yy_scan_string (self->code));
870 res = code_lex (self, sc_context);
871 yy_delete_buffer (YY_CURRENT_BUFFER);
876 /*------------------------------------------------------------------------.
877 | Implementation of the public interface as documented in "scan-code.h". |
878 `------------------------------------------------------------------------*/
881 code_props_none_init (code_props *self)
883 *self = code_props_none;
886 code_props const code_props_none = CODE_PROPS_NONE_INIT;
889 code_props_plain_init (code_props *self, char const *code,
892 self->kind = CODE_PROPS_PLAIN;
894 self->location = code_loc;
895 self->is_value_used = false;
897 self->named_ref = NULL;
901 code_props_symbol_action_init (code_props *self, char const *code,
904 self->kind = CODE_PROPS_SYMBOL_ACTION;
906 self->location = code_loc;
907 self->is_value_used = false;
909 self->named_ref = NULL;
913 code_props_rule_action_init (code_props *self, char const *code,
914 location code_loc, symbol_list *rule,
917 self->kind = CODE_PROPS_RULE_ACTION;
919 self->location = code_loc;
920 self->is_value_used = false;
922 self->named_ref = name;
926 code_props_translate_code (code_props *self)
930 case CODE_PROPS_NONE:
932 case CODE_PROPS_PLAIN:
933 self->code = translate_action (self, INITIAL);
935 case CODE_PROPS_SYMBOL_ACTION:
936 self->code = translate_action (self, SC_SYMBOL_ACTION);
938 case CODE_PROPS_RULE_ACTION:
939 self->code = translate_action (self, SC_RULE_ACTION);
945 code_scanner_last_string_free (void)
951 code_scanner_free (void)
953 obstack_free (&obstack_for_string, 0);
954 variant_table_free ();
956 /* Reclaim Flex's buffers. */