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 "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. */
107 bool need_semicolon = false;
109 /* Whether in a C preprocessor directive. Don't use a start condition
110 for this because, at the end of strings and comments, we still need
111 to know whether we're in a directive. */
114 /* This scanner is special: it is invoked only once, henceforth
115 is expected to return only once. This initialization is
116 therefore done once per action to translate. */
117 aver (sc_context == SC_SYMBOL_ACTION
118 || sc_context == SC_RULE_ACTION
119 || sc_context == INITIAL);
123 /*------------------------------------------------------------.
124 | Scanning a C comment. The initial '/ *' is already eaten. |
125 `------------------------------------------------------------*/
129 "*"{splice}"/" STRING_GROW; BEGIN sc_context;
133 /*--------------------------------------------------------------.
134 | Scanning a line comment. The initial '//' is already eaten. |
135 `--------------------------------------------------------------*/
139 "\n" STRING_GROW; BEGIN sc_context;
140 {splice} STRING_GROW;
144 /*--------------------------------------------.
145 | Scanning user-code characters and strings. |
146 `--------------------------------------------*/
148 <SC_CHARACTER,SC_STRING>
150 {splice}|\\{splice}. STRING_GROW;
155 "'" STRING_GROW; BEGIN sc_context;
160 "\"" STRING_GROW; BEGIN sc_context;
164 <SC_RULE_ACTION,SC_SYMBOL_ACTION>
169 need_semicolon = true;
174 need_semicolon = true;
182 BEGIN SC_LINE_COMMENT;
188 "$"("<"{tag}">")?{ref} {
189 ref_tail_fields = NULL;
190 handle_action_dollar (self->rule, yytext, *loc);
192 obstack_sgrow (&obstack_for_string, ref_tail_fields);
193 need_semicolon = true;
196 ref_tail_fields = NULL;
197 handle_action_at (self->rule, yytext, *loc);
199 obstack_sgrow (&obstack_for_string, ref_tail_fields);
200 need_semicolon = true;
203 warn_at (*loc, _("stray '$'"));
204 obstack_sgrow (&obstack_for_string, "$][");
205 need_semicolon = true;
208 warn_at (*loc, _("stray '@'"));
209 obstack_sgrow (&obstack_for_string, "@@");
210 need_semicolon = true;
213 obstack_sgrow (&obstack_for_string, "@{");
214 need_semicolon = true;
217 obstack_sgrow (&obstack_for_string, "@}");
218 need_semicolon = true;
221 ";" STRING_GROW; need_semicolon = false;
222 "{" STRING_GROW; ++braces_level; need_semicolon = false;
224 bool outer_brace = --braces_level == 0;
226 /* As an undocumented Bison extension, append ';' before the last
227 brace in braced code, so that the user code can omit trailing
228 ';'. But do not append ';' if emulating Yacc, since Yacc does
230 if (outer_brace && !yacc_flag && language_prio == default_prio
231 && skeleton_prio == default_prio && need_semicolon && ! in_cpp)
233 warn_at (*loc, _("a ';' might be needed at the end of action code"));
234 warn_at (*loc, _("future versions of Bison will not add the ';'"));
235 obstack_1grow (&obstack_for_string, ';');
239 need_semicolon = false;
242 /* Preprocessing directives should only be recognized at the beginning
243 of lines, allowing whitespace including comments, but in C/C++,
244 '#' can only be the start of preprocessor directives or within
245 '#define' directives anyway, so don't bother with begin of line. */
246 "#" STRING_GROW; in_cpp = true;
248 {splice} STRING_GROW;
249 [\n\r] STRING_GROW; if (in_cpp) in_cpp = need_semicolon = false;
252 /* YYFAIL is undocumented and was formally deprecated in Bison
255 STRING_GROW; need_semicolon = true;
256 warn_at (*loc, _("use of YYFAIL, which is deprecated and will be"
260 /* The sole purpose of this is to make sure identifiers that merely
261 contain YYFAIL don't produce the above warning. */
262 [A-Za-z_][0-9A-Za-z_]* STRING_GROW; need_semicolon = true;
264 . STRING_GROW; need_semicolon = true;
269 "$"("<"{tag}">")?"$" {
270 const char *type_name = NULL;
271 fetch_type_name (yytext + 1, &type_name, *loc)[-1] = 0;
272 obstack_sgrow (&obstack_for_string, "]b4_dollar_dollar(");
273 obstack_quote (&obstack_for_string, type_name);
274 obstack_sgrow (&obstack_for_string, ")[");
275 self->is_value_used = true;
278 obstack_sgrow (&obstack_for_string, "]b4_at_dollar[");
279 locations_flag = true;
284 /*-----------------------------------------.
285 | Escape M4 quoting characters in C code. |
286 `-----------------------------------------*/
290 \$ obstack_sgrow (&obstack_for_string, "$][");
291 \@ obstack_sgrow (&obstack_for_string, "@@");
292 \[ obstack_sgrow (&obstack_for_string, "@{");
293 \] obstack_sgrow (&obstack_for_string, "@}");
296 /*-----------------------------------------------------.
297 | By default, grow the string obstack with the input. |
298 `-----------------------------------------------------*/
302 /* End of processing. */
311 is_dot_or_dash (char ch)
313 return ch == '.' || ch == '-';
317 contains_dot_or_dash (const char* p)
320 if (is_dot_or_dash (*p))
325 /* Defines a variant of a symbolic name resolution. */
328 /* Index in symbol list. */
329 unsigned symbol_index;
331 /* Matched symbol id and loc. */
335 /* Hiding named reference. */
336 named_ref* hidden_by;
338 /* Error flags. May contain zero (no errors) or
339 a combination of VARIANT_* values. */
343 /* Set when the variant refers to a symbol hidden
344 by an explicit symbol reference. */
345 #define VARIANT_HIDDEN (1 << 0)
347 /* Set when the variant refers to a symbol containing
348 dots or dashes. Will require explicit bracketing. */
349 #define VARIANT_BAD_BRACKETING (1 << 1)
351 /* Set when the variant refers to a symbol which is
352 not visible from current midrule. */
353 #define VARIANT_NOT_VISIBLE_FROM_MIDRULE (1 << 2)
355 static variant *variant_table = NULL;
356 static unsigned variant_table_size = 0;
357 static unsigned variant_count = 0;
360 variant_table_grow (void)
363 if (variant_count > variant_table_size)
365 while (variant_count > variant_table_size)
366 variant_table_size = 2 * variant_table_size + 3;
367 variant_table = xnrealloc (variant_table, variant_table_size,
368 sizeof *variant_table);
370 return &variant_table[variant_count - 1];
374 variant_table_free (void)
376 free (variant_table);
377 variant_table = NULL;
378 variant_table_size = variant_count = 0;
382 find_prefix_end (const char *prefix, char *begin, char *end)
386 for (; *prefix && ptr != end; ++prefix, ++ptr)
397 variant_add (uniqstr id, location id_loc, unsigned symbol_index,
398 char *cp, char *cp_end, bool explicit_bracketing)
402 prefix_end = find_prefix_end (id, cp, cp_end);
404 (prefix_end == cp_end ||
405 (!explicit_bracketing && is_dot_or_dash (*prefix_end))))
407 variant *r = variant_table_grow ();
408 r->symbol_index = symbol_index;
420 get_at_spec (unsigned symbol_index)
422 static char at_buf[20];
423 if (symbol_index == 0)
424 strcpy (at_buf, "$$");
426 snprintf (at_buf, sizeof at_buf, "$%u", symbol_index);
431 show_sub_messages (const char* cp, bool explicit_bracketing,
432 int midrule_rhs_index, char dollar_or_at,
433 bool is_warning, unsigned indent)
437 for (i = 0; i < variant_count; ++i)
439 const variant *var = &variant_table[i];
440 const char *at_spec = get_at_spec (var->symbol_index);
445 warn_at_indent (var->loc, &indent, _("refers to: %c%s at %s"),
446 dollar_or_at, var->id, at_spec);
448 complain_at_indent (var->loc, &indent, _("refers to: %c%s at %s"),
449 dollar_or_at, var->id, at_spec);
453 static struct obstack msg_buf;
454 const char *tail = explicit_bracketing ? "" :
455 cp + strlen (var->id);
456 const char *id = var->hidden_by ? var->hidden_by->id :
458 location id_loc = var->hidden_by ? var->hidden_by->loc :
461 /* Create the explanation message. */
462 obstack_init (&msg_buf);
464 obstack_fgrow1 (&msg_buf, _("possibly meant: %c"), dollar_or_at);
465 if (contains_dot_or_dash (id))
466 obstack_fgrow1 (&msg_buf, "[%s]", id);
468 obstack_sgrow (&msg_buf, id);
469 obstack_sgrow (&msg_buf, tail);
471 if (var->err & VARIANT_HIDDEN)
473 obstack_fgrow1 (&msg_buf, _(", hiding %c"), dollar_or_at);
474 if (contains_dot_or_dash (var->id))
475 obstack_fgrow1 (&msg_buf, "[%s]", var->id);
477 obstack_sgrow (&msg_buf, var->id);
478 obstack_sgrow (&msg_buf, tail);
481 obstack_fgrow1 (&msg_buf, _(" at %s"), at_spec);
483 if (var->err & VARIANT_NOT_VISIBLE_FROM_MIDRULE)
486 _(", cannot be accessed from mid-rule action at $%d");
487 obstack_fgrow1 (&msg_buf, format, midrule_rhs_index);
490 obstack_1grow (&msg_buf, '\0');
492 warn_at_indent (id_loc, &indent, "%s",
493 (char *) obstack_finish (&msg_buf));
495 complain_at_indent (id_loc, &indent, "%s",
496 (char *) obstack_finish (&msg_buf));
497 obstack_free (&msg_buf, 0);
502 /* Returned from "parse_ref" when the reference
504 #define INVALID_REF (INT_MIN)
506 /* Returned from "parse_ref" when the reference
507 points to LHS ($$) of the current rule or midrule. */
508 #define LHS_REF (INT_MIN + 1)
510 /* Sub-messages indent. */
511 #define SUB_INDENT (4)
513 /* Parse named or positional reference. In case of positional
514 references, can return negative values for $-n "deep" stack
517 parse_ref (char *cp, symbol_list *rule, int rule_length,
518 int midrule_rhs_index, char *text, location text_loc,
523 bool explicit_bracketing;
525 unsigned valid_variants = 0;
526 unsigned valid_variant_index = 0;
531 if (c_isdigit (*cp) || (*cp == '-' && c_isdigit (* (cp + 1))))
533 long int num = strtol (cp, &cp, 10);
534 if (1 - INT_MAX + rule_length <= num && num <= rule_length)
538 complain_at (text_loc, _("integer out of range: %s"),
546 /* Ignore the brackets. */
548 for (p = ++cp; *p != ']'; ++p)
552 explicit_bracketing = true;
556 /* Take all characters of the name. */
558 for (p = cp; *p; ++p)
559 if (is_dot_or_dash (*p))
564 for (p = cp; *p; ++p)
568 explicit_bracketing = false;
571 /* Add all relevant variants. */
573 unsigned symbol_index;
575 for (symbol_index = 0, l = rule; !symbol_list_null (l);
576 ++symbol_index, l = l->next)
579 if (l->content_type != SYMLIST_SYMBOL)
582 var = variant_add (l->content.sym->tag, l->sym_loc,
583 symbol_index, cp, cp_end, explicit_bracketing);
584 if (var && l->named_ref)
585 var->hidden_by = l->named_ref;
588 variant_add (l->named_ref->id, l->named_ref->loc,
589 symbol_index, cp, cp_end, explicit_bracketing);
594 for (i = 0; i < variant_count; ++i)
596 variant *var = &variant_table[i];
597 unsigned symbol_index = var->symbol_index;
599 /* Check visibility from mid-rule actions. */
600 if (midrule_rhs_index != 0
601 && (symbol_index == 0 || midrule_rhs_index < symbol_index))
602 var->err |= VARIANT_NOT_VISIBLE_FROM_MIDRULE;
604 /* Check correct bracketing. */
605 if (!explicit_bracketing && contains_dot_or_dash (var->id))
606 var->err |= VARIANT_BAD_BRACKETING;
608 /* Check using of hidden symbols. */
610 var->err |= VARIANT_HIDDEN;
614 valid_variant_index = i;
619 switch (valid_variants)
623 unsigned len = (explicit_bracketing || !ref_tail_fields) ?
624 cp_end - cp : ref_tail_fields - cp;
627 complain_at_indent (text_loc, &indent, _("invalid reference: %s"),
629 indent += SUB_INDENT;
632 location sym_loc = text_loc;
633 sym_loc.start.column += 1;
634 sym_loc.end = sym_loc.start;
636 _("syntax error after '%c', expecting integer, letter,"
637 " '_', '[', or '$'");
638 complain_at_indent (sym_loc, &indent, format, dollar_or_at);
640 else if (midrule_rhs_index)
643 _("symbol not found in production before $%d: %.*s");
644 complain_at_indent (rule->location, &indent, format,
645 midrule_rhs_index, len, cp);
650 _("symbol not found in production: %.*s");
651 complain_at_indent (rule->location, &indent, format,
655 if (variant_count > 0)
656 show_sub_messages (cp, explicit_bracketing, midrule_rhs_index,
657 dollar_or_at, false, indent);
663 if (variant_count > 1)
665 warn_at_indent (text_loc, &indent, _("misleading reference: %s"),
667 show_sub_messages (cp, explicit_bracketing, midrule_rhs_index,
668 dollar_or_at, true, indent + SUB_INDENT);
671 unsigned symbol_index =
672 variant_table[valid_variant_index].symbol_index;
673 return (symbol_index == midrule_rhs_index) ? LHS_REF : symbol_index;
680 complain_at_indent (text_loc, &indent, _("ambiguous reference: %s"),
682 show_sub_messages (cp, explicit_bracketing, midrule_rhs_index,
683 dollar_or_at, false, indent + SUB_INDENT);
692 /* Keeps track of the maximum number of semantic values to the left of
693 a handle (those referenced by $0, $-1, etc.) are required by the
694 semantic actions of this grammar. */
695 int max_left_semantic_context = 0;
698 /* If CP points to a typename (i.e., <.*?>), set TYPE_NAME to its
699 beginning (i.e., after the opening "<", and return the pointer
700 immediately after it. */
704 fetch_type_name (char *cp, char const **type_name,
713 /* The '>' symbol will be later replaced by '\0'. Original
714 'text' is needed for error messages. */
716 if (untyped_var_seen)
717 complain_at (dollar_loc, _("explicit type given in untyped grammar"));
723 /*------------------------------------------------------------------.
724 | TEXT is pointing to a wannabee semantic value (i.e., a '$'). |
726 | Possible inputs: $[<TYPENAME>]($|integer) |
728 | Output to OBSTACK_FOR_STRING a reference to this semantic value. |
729 `------------------------------------------------------------------*/
732 handle_action_dollar (symbol_list *rule, char *text, location dollar_loc)
734 char const *type_name = NULL;
736 symbol_list *effective_rule;
737 int effective_rule_length;
740 if (rule->midrule_parent_rule)
742 effective_rule = rule->midrule_parent_rule;
743 effective_rule_length = rule->midrule_parent_rhs_index - 1;
747 effective_rule = rule;
748 effective_rule_length = symbol_list_length (rule->next);
751 /* Get the type name if explicit. */
752 cp = fetch_type_name (cp, &type_name, dollar_loc);
754 n = parse_ref (cp, effective_rule, effective_rule_length,
755 rule->midrule_parent_rhs_index, text, dollar_loc, '$');
768 type_name = symbol_list_n_type_name_get (rule, dollar_loc, 0);
772 if (union_seen | tag_seen)
774 if (rule->midrule_parent_rule)
775 complain_at (dollar_loc,
776 _("$$ for the midrule at $%d of %s"
777 " has no declared type"),
778 rule->midrule_parent_rhs_index,
779 quote (effective_rule->content.sym->tag));
781 complain_at (dollar_loc, _("$$ of %s has no declared type"),
782 quote (rule->content.sym->tag));
785 untyped_var_seen = true;
788 obstack_sgrow (&obstack_for_string, "]b4_lhs_value(");
789 obstack_quote (&obstack_for_string, type_name);
790 obstack_sgrow (&obstack_for_string, ")[");
791 rule->action_props.is_value_used = true;
795 if (max_left_semantic_context < 1 - n)
796 max_left_semantic_context = 1 - n;
797 if (!type_name && 0 < n)
799 symbol_list_n_type_name_get (effective_rule, dollar_loc, n);
802 if (union_seen | tag_seen)
803 complain_at (dollar_loc, _("$%s of %s has no declared type"),
804 cp, quote (effective_rule->content.sym->tag));
806 untyped_var_seen = true;
809 obstack_fgrow2 (&obstack_for_string,
810 "]b4_rhs_value(%d, %d, ", effective_rule_length, n);
811 obstack_quote (&obstack_for_string, type_name);
812 obstack_sgrow (&obstack_for_string, ")[");
814 symbol_list_n_get (effective_rule, n)->action_props.is_value_used =
821 /*------------------------------------------------------.
822 | TEXT is a location token (i.e., a '@...'). Output to |
823 | OBSTACK_FOR_STRING a reference to this location. |
824 `------------------------------------------------------*/
827 handle_action_at (symbol_list *rule, char *text, location at_loc)
830 symbol_list *effective_rule;
831 int effective_rule_length;
834 if (rule->midrule_parent_rule)
836 effective_rule = rule->midrule_parent_rule;
837 effective_rule_length = rule->midrule_parent_rhs_index - 1;
841 effective_rule = rule;
842 effective_rule_length = symbol_list_length (rule->next);
845 locations_flag = true;
847 n = parse_ref (cp, effective_rule, effective_rule_length,
848 rule->midrule_parent_rhs_index, text, at_loc, '@');
855 obstack_sgrow (&obstack_for_string, "]b4_lhs_location[");
859 obstack_fgrow2 (&obstack_for_string, "]b4_rhs_location(%d, %d)[",
860 effective_rule_length, n);
866 /*-------------------------.
867 | Initialize the scanner. |
868 `-------------------------*/
870 /* Translate the dollars and ats in \a self, in the context \a sc_context
871 (SC_RULE_ACTION, SC_SYMBOL_ACTION, INITIAL). */
874 translate_action (code_props *self, int sc_context)
877 static bool initialized = false;
880 obstack_init (&obstack_for_string);
885 loc->start = loc->end = self->location.start;
886 yy_switch_to_buffer (yy_scan_string (self->code));
887 res = code_lex (self, sc_context);
888 yy_delete_buffer (YY_CURRENT_BUFFER);
893 /*------------------------------------------------------------------------.
894 | Implementation of the public interface as documented in "scan-code.h". |
895 `------------------------------------------------------------------------*/
898 code_props_none_init (code_props *self)
900 *self = code_props_none;
903 code_props const code_props_none = CODE_PROPS_NONE_INIT;
906 code_props_plain_init (code_props *self, char const *code,
909 self->kind = CODE_PROPS_PLAIN;
911 self->location = code_loc;
912 self->is_value_used = false;
914 self->named_ref = NULL;
918 code_props_symbol_action_init (code_props *self, char const *code,
921 self->kind = CODE_PROPS_SYMBOL_ACTION;
923 self->location = code_loc;
924 self->is_value_used = false;
926 self->named_ref = NULL;
930 code_props_rule_action_init (code_props *self, char const *code,
931 location code_loc, symbol_list *rule,
934 self->kind = CODE_PROPS_RULE_ACTION;
936 self->location = code_loc;
937 self->is_value_used = false;
939 self->named_ref = name;
943 code_props_translate_code (code_props *self)
947 case CODE_PROPS_NONE:
949 case CODE_PROPS_PLAIN:
950 self->code = translate_action (self, INITIAL);
952 case CODE_PROPS_SYMBOL_ACTION:
953 self->code = translate_action (self, SC_SYMBOL_ACTION);
955 case CODE_PROPS_RULE_ACTION:
956 self->code = translate_action (self, SC_RULE_ACTION);
962 code_scanner_last_string_free (void)
968 code_scanner_free (void)
970 obstack_free (&obstack_for_string, 0);
971 variant_table_free ();
973 /* Reclaim Flex's buffers. */