X-Git-Url: https://git.saurik.com/bison.git/blobdiff_plain/1a323c2f7143659fa0dcbddeb3d2518275973c05..4d24ffb75ebfc7c4fae027e23132c6d3630998b5:/src/scan-code.l diff --git a/src/scan-code.l b/src/scan-code.l index fc35513a..31b44298 100644 --- a/src/scan-code.l +++ b/src/scan-code.l @@ -1,6 +1,6 @@ /* Bison Action Scanner -*- C -*- - Copyright (C) 2006, 2007, 2008-2009 Free Software Foundation, Inc. + Copyright (C) 2006-2012 Free Software Foundation, Inc. This file is part of Bison, the GNU Compiler Compiler. @@ -46,6 +46,9 @@ YY_DECL; #define YY_USER_ACTION location_compute (loc, &loc->end, yytext, yyleng); +static char *fetch_type_name (char *cp, char const **type_name, + location dollar_loc); + static void handle_action_dollar (symbol_list *rule, char *cp, location dollar_loc); static void handle_action_at (symbol_list *rule, char *cp, location at_loc); @@ -84,8 +87,8 @@ splice (\\[ \f\t\v]*\n)* /* C style identifier. Must start with letter. Will be used for named symbol references. Shall be kept synchronized with scan-gram.l "letter" and "id". */ -letter [-.abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ_] -id {letter}({letter}|[0-9])* +letter [.abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ_] +id {letter}({letter}|[-0-9])* ref -?[0-9]+|{id}|"["{id}"]"|"$" %% @@ -95,11 +98,12 @@ ref -?[0-9]+|{id}|"["{id}"]"|"$" int braces_level = 0; /* Whether a semicolon is probably needed. - The heuristic is that a semicolon is not needed after `{', `}', `;', - or a C preprocessor directive, and that whitespaces and comments - do not affect this flag. - Note that `{' does not need a semicolon because of `{}'. - A semicolon may be needed before a cpp direcive, but don't bother. */ + + The heuristic is that a semicolon is not needed after '{', '}', + ';', or a C preprocessor directive, and that whitespaces and + comments do not affect this flag. Note that '{' does not need a + semicolon because of '{}'. A semicolon may be needed before a + cpp directive, but don't bother. */ bool need_semicolon = false; /* Whether in a C preprocessor directive. Don't use a start condition @@ -117,7 +121,7 @@ ref -?[0-9]+|{id}|"["{id}"]"|"$" %} /*------------------------------------------------------------. - | Scanning a C comment. The initial `/ *' is already eaten. | + | Scanning a C comment. The initial '/ *' is already eaten. | `------------------------------------------------------------*/ @@ -127,7 +131,7 @@ ref -?[0-9]+|{id}|"["{id}"]"|"$" /*--------------------------------------------------------------. - | Scanning a line comment. The initial `//' is already eaten. | + | Scanning a line comment. The initial '//' is already eaten. | `--------------------------------------------------------------*/ @@ -157,7 +161,8 @@ ref -?[0-9]+|{id}|"["{id}"]"|"$" } -{ + +{ "'" { STRING_GROW; BEGIN SC_CHARACTER; @@ -181,28 +186,26 @@ ref -?[0-9]+|{id}|"["{id}"]"|"$" { "$"("<"{tag}">")?{ref} { - ref_tail_fields = 0; + ref_tail_fields = NULL; handle_action_dollar (self->rule, yytext, *loc); - if (ref_tail_fields) { + if (ref_tail_fields) obstack_sgrow (&obstack_for_string, ref_tail_fields); - } need_semicolon = true; } "@"{ref} { - ref_tail_fields = 0; + ref_tail_fields = NULL; handle_action_at (self->rule, yytext, *loc); - if (ref_tail_fields) { + if (ref_tail_fields) obstack_sgrow (&obstack_for_string, ref_tail_fields); - } need_semicolon = true; } "$" { - warn_at (*loc, _("stray `$'")); + warn_at (*loc, _("stray '$'")); obstack_sgrow (&obstack_for_string, "$]["); need_semicolon = true; } "@" { - warn_at (*loc, _("stray `@'")); + warn_at (*loc, _("stray '@'")); obstack_sgrow (&obstack_for_string, "@@"); need_semicolon = true; } @@ -220,16 +223,16 @@ ref -?[0-9]+|{id}|"["{id}"]"|"$" "}" { bool outer_brace = --braces_level == 0; - /* As an undocumented Bison extension, append `;' before the last + /* As an undocumented Bison extension, append ';' before the last brace in braced code, so that the user code can omit trailing - `;'. But do not append `;' if emulating Yacc, since Yacc does + ';'. But do not append ';' if emulating Yacc, since Yacc does not append one. */ if (outer_brace && !yacc_flag && language_prio == default_prio && skeleton_prio == default_prio && need_semicolon && ! in_cpp) { - warn_at (*loc, _("a `;' might be needed at the end of action code")); - warn_at (*loc, _("future versions of Bison will not add the `;'")); - obstack_1grow (&obstack_for_string, ';'); + warn_at (*loc, _("a ';' might be needed at the end of action code")); + warn_at (*loc, _("future versions of Bison will not add the ';'")); + obstack_1grow (&obstack_for_string, ';'); } STRING_GROW; @@ -238,20 +241,37 @@ ref -?[0-9]+|{id}|"["{id}"]"|"$" /* Preprocessing directives should only be recognized at the beginning of lines, allowing whitespace including comments, but in C/C++, - `#' can only be the start of preprocessor directives or within - `#define' directives anyway, so don't bother with begin of line. */ + '#' can only be the start of preprocessor directives or within + '#define' directives anyway, so don't bother with begin of line. */ "#" STRING_GROW; in_cpp = true; {splice} STRING_GROW; - [\n\r] STRING_GROW; if (in_cpp) in_cpp = need_semicolon = false; + [\n\r] STRING_GROW; if (in_cpp) in_cpp = need_semicolon = false; [ \t\f] STRING_GROW; - . STRING_GROW; need_semicolon = true; + + /* YYFAIL is undocumented and was formally deprecated in Bison + 2.4.2. */ + YYFAIL { + STRING_GROW; need_semicolon = true; + warn_at (*loc, _("use of YYFAIL, which is deprecated and will be" + " removed")); + } + + /* The sole purpose of this is to make sure identifiers that merely + contain YYFAIL don't produce the above warning. */ + [A-Za-z_][0-9A-Za-z_]* STRING_GROW; need_semicolon = true; + + . STRING_GROW; need_semicolon = true; } { - "$$" { - obstack_sgrow (&obstack_for_string, "]b4_dollar_dollar["); + "$"("<"{tag}">")?"$" { + const char *type_name = NULL; + fetch_type_name (yytext + 1, &type_name, *loc)[-1] = 0; + obstack_sgrow (&obstack_for_string, "]b4_dollar_dollar("); + obstack_quote (&obstack_for_string, type_name); + obstack_sgrow (&obstack_for_string, ")["); self->is_value_used = true; } "@$" { @@ -332,7 +352,7 @@ typedef struct not visible from current midrule. */ #define VARIANT_NOT_VISIBLE_FROM_MIDRULE (1 << 2) -static variant *variant_table = 0; +static variant *variant_table = NULL; static unsigned variant_table_size = 0; static unsigned variant_count = 0; @@ -354,7 +374,7 @@ static void variant_table_free (void) { free (variant_table); - variant_table = 0; + variant_table = NULL; variant_table_size = variant_count = 0; } @@ -375,14 +395,14 @@ find_prefix_end (const char *prefix, char *begin, char *end) static variant * variant_add (uniqstr id, location id_loc, unsigned symbol_index, - char *cp, char *cp_end, bool exact_mode) + char *cp, char *cp_end, bool explicit_bracketing) { char *prefix_end; prefix_end = find_prefix_end (id, cp, cp_end); if (prefix_end && (prefix_end == cp_end || - (!exact_mode && is_dot_or_dash (*prefix_end)))) + (!explicit_bracketing && is_dot_or_dash (*prefix_end)))) { variant *r = variant_table_grow (); r->symbol_index = symbol_index; @@ -397,7 +417,7 @@ variant_add (uniqstr id, location id_loc, unsigned symbol_index, } static const char * -get_at_spec(unsigned symbol_index) +get_at_spec (unsigned symbol_index) { static char at_buf[20]; if (symbol_index == 0) @@ -408,8 +428,9 @@ get_at_spec(unsigned symbol_index) } static void -show_sub_messages (const char* cp, bool exact_mode, int midrule_rhs_index, - char dollar_or_at, bool is_warning) +show_sub_messages (const char* cp, bool explicit_bracketing, + int midrule_rhs_index, char dollar_or_at, + bool is_warning, unsigned indent) { unsigned i; @@ -421,16 +442,16 @@ show_sub_messages (const char* cp, bool exact_mode, int midrule_rhs_index, if (var->err == 0) { if (is_warning) - warn_at (var->loc, _(" refers to: %c%s at %s"), - dollar_or_at, var->id, at_spec); + warn_at_indent (var->loc, &indent, _("refers to: %c%s at %s"), + dollar_or_at, var->id, at_spec); else - complain_at (var->loc, _(" refers to: %c%s at %s"), - dollar_or_at, var->id, at_spec); + complain_at_indent (var->loc, &indent, _("refers to: %c%s at %s"), + dollar_or_at, var->id, at_spec); } else { static struct obstack msg_buf; - const char *tail = exact_mode ? "" : + const char *tail = explicit_bracketing ? "" : cp + strlen (var->id); const char *id = var->hidden_by ? var->hidden_by->id : var->id; @@ -440,7 +461,7 @@ show_sub_messages (const char* cp, bool exact_mode, int midrule_rhs_index, /* Create the explanation message. */ obstack_init (&msg_buf); - obstack_fgrow1 (&msg_buf, " possibly meant: %c", dollar_or_at); + obstack_fgrow1 (&msg_buf, _("possibly meant: %c"), dollar_or_at); if (contains_dot_or_dash (id)) obstack_fgrow1 (&msg_buf, "[%s]", id); else @@ -449,7 +470,7 @@ show_sub_messages (const char* cp, bool exact_mode, int midrule_rhs_index, if (var->err & VARIANT_HIDDEN) { - obstack_fgrow1 (&msg_buf, ", hiding %c", dollar_or_at); + obstack_fgrow1 (&msg_buf, _(", hiding %c"), dollar_or_at); if (contains_dot_or_dash (var->id)) obstack_fgrow1 (&msg_buf, "[%s]", var->id); else @@ -457,17 +478,22 @@ show_sub_messages (const char* cp, bool exact_mode, int midrule_rhs_index, obstack_sgrow (&msg_buf, tail); } - obstack_fgrow1 (&msg_buf, " at %s", at_spec); + obstack_fgrow1 (&msg_buf, _(" at %s"), at_spec); if (var->err & VARIANT_NOT_VISIBLE_FROM_MIDRULE) - obstack_fgrow1 (&msg_buf, ", cannot be accessed from " - "mid-rule action at $%d", midrule_rhs_index); + { + const char *format = + _(", cannot be accessed from mid-rule action at $%d"); + obstack_fgrow1 (&msg_buf, format, midrule_rhs_index); + } obstack_1grow (&msg_buf, '\0'); if (is_warning) - warn_at (id_loc, _("%s"), (char *) obstack_finish (&msg_buf)); + warn_at_indent (id_loc, &indent, "%s", + (char *) obstack_finish (&msg_buf)); else - complain_at (id_loc, _("%s"), (char *) obstack_finish (&msg_buf)); + complain_at_indent (id_loc, &indent, "%s", + (char *) obstack_finish (&msg_buf)); obstack_free (&msg_buf, 0); } } @@ -481,6 +507,9 @@ show_sub_messages (const char* cp, bool exact_mode, int midrule_rhs_index, points to LHS ($$) of the current rule or midrule. */ #define LHS_REF (INT_MIN + 1) +/* Sub-messages indent. */ +#define SUB_INDENT (4) + /* Parse named or positional reference. In case of positional references, can return negative values for $-n "deep" stack accesses. */ @@ -491,7 +520,7 @@ parse_ref (char *cp, symbol_list *rule, int rule_length, { symbol_list *l; char *cp_end; - bool exact_mode; + bool explicit_bracketing; unsigned i; unsigned valid_variants = 0; unsigned valid_variant_index = 0; @@ -520,7 +549,7 @@ parse_ref (char *cp, symbol_list *rule, int rule_length, continue; cp_end = p; - exact_mode = true; + explicit_bracketing = true; } else { @@ -536,7 +565,7 @@ parse_ref (char *cp, symbol_list *rule, int rule_length, continue; cp_end = p; - exact_mode = false; + explicit_bracketing = false; } /* Add all relevant variants. */ @@ -551,13 +580,13 @@ parse_ref (char *cp, symbol_list *rule, int rule_length, continue; var = variant_add (l->content.sym->tag, l->sym_loc, - symbol_index, cp, cp_end, exact_mode); + symbol_index, cp, cp_end, explicit_bracketing); if (var && l->named_ref) var->hidden_by = l->named_ref; if (l->named_ref) variant_add (l->named_ref->id, l->named_ref->loc, - symbol_index, cp, cp_end, exact_mode); + symbol_index, cp, cp_end, explicit_bracketing); } } @@ -573,7 +602,7 @@ parse_ref (char *cp, symbol_list *rule, int rule_length, var->err |= VARIANT_NOT_VISIBLE_FROM_MIDRULE; /* Check correct bracketing. */ - if (!exact_mode && contains_dot_or_dash (var->id)) + if (!explicit_bracketing && contains_dot_or_dash (var->id)) var->err |= VARIANT_BAD_BRACKETING; /* Check using of hidden symbols. */ @@ -590,25 +619,53 @@ parse_ref (char *cp, symbol_list *rule, int rule_length, switch (valid_variants) { case 0: - if (variant_count == 0) - complain_at (text_loc, _("invalid reference: %s, symbol not found"), - quote (text)); - else - { - complain_at (text_loc, _("invalid reference: %s"), - quote (text)); - show_sub_messages (cp, exact_mode, midrule_rhs_index, - dollar_or_at, false); - } - return INVALID_REF; + { + unsigned len = (explicit_bracketing || !ref_tail_fields) ? + cp_end - cp : ref_tail_fields - cp; + unsigned indent = 0; + + complain_at_indent (text_loc, &indent, _("invalid reference: %s"), + quote (text)); + indent += SUB_INDENT; + if (len == 0) + { + location sym_loc = text_loc; + sym_loc.start.column += 1; + sym_loc.end = sym_loc.start; + const char *format = + _("syntax error after '%c', expecting integer, letter," + " '_', '[', or '$'"); + complain_at_indent (sym_loc, &indent, format, dollar_or_at); + } + else if (midrule_rhs_index) + { + const char *format = + _("symbol not found in production before $%d: %.*s"); + complain_at_indent (rule->location, &indent, format, + midrule_rhs_index, len, cp); + } + else + { + const char *format = + _("symbol not found in production: %.*s"); + complain_at_indent (rule->location, &indent, format, + len, cp); + } + + if (variant_count > 0) + show_sub_messages (cp, explicit_bracketing, midrule_rhs_index, + dollar_or_at, false, indent); + return INVALID_REF; + } case 1: { + unsigned indent = 0; if (variant_count > 1) { - warn_at (text_loc, _("misleading reference: %s"), - quote (text)); - show_sub_messages (cp, exact_mode, midrule_rhs_index, - dollar_or_at, true); + warn_at_indent (text_loc, &indent, _("misleading reference: %s"), + quote (text)); + show_sub_messages (cp, explicit_bracketing, midrule_rhs_index, + dollar_or_at, true, indent + SUB_INDENT); } { unsigned symbol_index = @@ -618,11 +675,14 @@ parse_ref (char *cp, symbol_list *rule, int rule_length, } case 2: default: - complain_at (text_loc, _("ambiguous reference: %s"), - quote (text)); - show_sub_messages (cp, exact_mode, midrule_rhs_index, - dollar_or_at, false); - return INVALID_REF; + { + unsigned indent = 0; + complain_at_indent (text_loc, &indent, _("ambiguous reference: %s"), + quote (text)); + show_sub_messages (cp, explicit_bracketing, midrule_rhs_index, + dollar_or_at, false, indent + SUB_INDENT); + return INVALID_REF; + } } /* Not reachable. */ @@ -635,8 +695,33 @@ parse_ref (char *cp, symbol_list *rule, int rule_length, int max_left_semantic_context = 0; +/* If CP points to a typename (i.e., <.*?>), set TYPE_NAME to its + beginning (i.e., after the opening "<", and return the pointer + immediately after it. */ + +static +char * +fetch_type_name (char *cp, char const **type_name, + location dollar_loc) +{ + if (*cp == '<') + { + *type_name = ++cp; + while (*cp != '>') + ++cp; + + /* The '>' symbol will be later replaced by '\0'. Original + 'text' is needed for error messages. */ + ++cp; + if (untyped_var_seen) + complain_at (dollar_loc, _("explicit type given in untyped grammar")); + tag_seen = true; + } + return cp; +} + /*------------------------------------------------------------------. -| TEXT is pointing to a wannabee semantic value (i.e., a `$'). | +| TEXT is pointing to a wannabee semantic value (i.e., a '$'). | | | | Possible inputs: $[]($|integer) | | | @@ -648,7 +733,6 @@ handle_action_dollar (symbol_list *rule, char *text, location dollar_loc) { char const *type_name = NULL; char *cp = text + 1; - char *gt_ptr = 0; symbol_list *effective_rule; int effective_rule_length; int n; @@ -665,26 +749,14 @@ handle_action_dollar (symbol_list *rule, char *text, location dollar_loc) } /* Get the type name if explicit. */ - if (*cp == '<') - { - type_name = ++cp; - while (*cp != '>') - ++cp; - - /* The '>' symbol will be later replaced by '\0'. Original - 'text' is needed for error messages. */ - gt_ptr = cp; - ++cp; - if (untyped_var_seen) - complain_at (dollar_loc, _("explicit type given in untyped grammar")); - tag_seen = true; - } + cp = fetch_type_name (cp, &type_name, dollar_loc); n = parse_ref (cp, effective_rule, effective_rule_length, rule->midrule_parent_rhs_index, text, dollar_loc, '$'); - if (gt_ptr) - *gt_ptr = '\0'; + /* End type_name. */ + if (type_name) + cp[-1] = '\0'; switch (n) { @@ -696,26 +768,26 @@ handle_action_dollar (symbol_list *rule, char *text, location dollar_loc) type_name = symbol_list_n_type_name_get (rule, dollar_loc, 0); if (!type_name) - { - if (union_seen | tag_seen) - { - if (rule->midrule_parent_rule) - complain_at (dollar_loc, - _("$$ for the midrule at $%d of `%s'" - " has no declared type"), - rule->midrule_parent_rhs_index, - effective_rule->content.sym->tag); - else - complain_at (dollar_loc, _("$$ of `%s' has no declared type"), - rule->content.sym->tag); - } - else - untyped_var_seen = true; - type_name = ""; - } + { + if (union_seen | tag_seen) + { + if (rule->midrule_parent_rule) + complain_at (dollar_loc, + _("$$ for the midrule at $%d of %s" + " has no declared type"), + rule->midrule_parent_rhs_index, + quote (effective_rule->content.sym->tag)); + else + complain_at (dollar_loc, _("$$ of %s has no declared type"), + quote (rule->content.sym->tag)); + } + else + untyped_var_seen = true; + } - obstack_fgrow1 (&obstack_for_string, - "]b4_lhs_value([%s])[", type_name); + obstack_sgrow (&obstack_for_string, "]b4_lhs_value("); + obstack_quote (&obstack_for_string, type_name); + obstack_sgrow (&obstack_for_string, ")["); rule->action_props.is_value_used = true; break; @@ -726,18 +798,18 @@ handle_action_dollar (symbol_list *rule, char *text, location dollar_loc) type_name = symbol_list_n_type_name_get (effective_rule, dollar_loc, n); if (!type_name) - { - if (union_seen | tag_seen) - complain_at (dollar_loc, _("$%s of `%s' has no declared type"), - cp, effective_rule->content.sym->tag); - else - untyped_var_seen = true; - type_name = ""; - } + { + if (union_seen | tag_seen) + complain_at (dollar_loc, _("$%s of %s has no declared type"), + cp, quote (effective_rule->content.sym->tag)); + else + untyped_var_seen = true; + } - obstack_fgrow3 (&obstack_for_string, - "]b4_rhs_value(%d, %d, [%s])[", - effective_rule_length, n, type_name); + obstack_fgrow2 (&obstack_for_string, + "]b4_rhs_value(%d, %d, ", effective_rule_length, n); + obstack_quote (&obstack_for_string, type_name); + obstack_sgrow (&obstack_for_string, ")["); if (n > 0) symbol_list_n_get (effective_rule, n)->action_props.is_value_used = true; @@ -747,7 +819,7 @@ handle_action_dollar (symbol_list *rule, char *text, location dollar_loc) /*------------------------------------------------------. -| TEXT is a location token (i.e., a `@...'). Output to | +| TEXT is a location token (i.e., a '@...'). Output to | | OBSTACK_FOR_STRING a reference to this location. | `------------------------------------------------------*/ @@ -773,7 +845,7 @@ handle_action_at (symbol_list *rule, char *text, location at_loc) locations_flag = true; n = parse_ref (cp, effective_rule, effective_rule_length, - rule->midrule_parent_rhs_index, text, at_loc, '@'); + rule->midrule_parent_rhs_index, text, at_loc, '@'); switch (n) { case INVALID_REF: