/* YACC parser for C expressions, for GDB.
- Copyright (C) 1986-2018 Free Software Foundation, Inc.
+ Copyright (C) 1986-2019 Free Software Foundation, Inc.
This file is part of GDB.
#include "parser-defs.h"
#include "language.h"
#include "c-lang.h"
+#include "c-support.h"
#include "bfd.h" /* Required by objfiles.h. */
#include "symfile.h" /* Required by objfiles.h. */
#include "objfiles.h" /* For have_full_symbols and have_partial_symbols */
static struct parser_state *pstate = NULL;
+/* Data that must be held for the duration of a parse. */
+
+struct c_parse_state
+{
+ /* These are used to hold type lists and type stacks that are
+ allocated during the parse. */
+ std::vector<std::unique_ptr<std::vector<struct type *>>> type_lists;
+ std::vector<std::unique_ptr<struct type_stack>> type_stacks;
+
+ /* Storage for some strings allocated during the parse. */
+ std::vector<gdb::unique_xmalloc_ptr<char>> strings;
+
+ /* When we find that lexptr (the global var defined in parse.c) is
+ pointing at a macro invocation, we expand the invocation, and call
+ scan_macro_expansion to save the old lexptr here and point lexptr
+ into the expanded text. When we reach the end of that, we call
+ end_macro_expansion to pop back to the value we saved here. The
+ macro expansion code promises to return only fully-expanded text,
+ so we don't need to "push" more than one level.
+
+ This is disgusting, of course. It would be cleaner to do all macro
+ expansion beforehand, and then hand that to lexptr. But we don't
+ really know where the expression ends. Remember, in a command like
+
+ (gdb) break *ADDRESS if CONDITION
+
+ we evaluate ADDRESS in the scope of the current frame, but we
+ evaluate CONDITION in the scope of the breakpoint's location. So
+ it's simply wrong to try to macro-expand the whole thing at once. */
+ const char *macro_original_text = nullptr;
+
+ /* We save all intermediate macro expansions on this obstack for the
+ duration of a single parse. The expansion text may sometimes have
+ to live past the end of the expansion, due to yacc lookahead.
+ Rather than try to be clever about saving the data for a single
+ token, we simply keep it all and delete it after parsing has
+ completed. */
+ auto_obstack expansion_obstack;
+};
+
+/* This is set and cleared in c_parse. */
+
+static struct c_parse_state *cpstate;
+
int yyparse (void);
static int yylex (void);
-void yyerror (const char *);
+static void yyerror (const char *);
static int type_aggregate_p (struct type *);
enum exp_opcode opcode;
struct stoken_vector svec;
- VEC (type_ptr) *tvec;
+ std::vector<struct type *> *tvec;
struct type_stack *type_stack;
static int parse_number (struct parser_state *par_state,
const char *, int, int, YYSTYPE *);
static struct stoken operator_stoken (const char *);
-static void check_parameter_typelist (VEC (type_ptr) *);
+static struct stoken typename_stoken (const char *);
+static void check_parameter_typelist (std::vector<struct type *> *);
static void write_destructor_name (struct parser_state *par_state,
struct stoken);
%token <voidval> COMPLETE
%token <tsym> TYPENAME
%token <theclass> CLASSNAME /* ObjC Class name */
-%type <sval> name
+%type <sval> name field_name
%type <svec> string_exp
%type <ssym> name_not_typename
%type <tsym> type_name
%token <ssym> NAME_OR_INT
%token OPERATOR
-%token STRUCT CLASS UNION ENUM SIZEOF UNSIGNED COLONCOLON
+%token STRUCT CLASS UNION ENUM SIZEOF ALIGNOF UNSIGNED COLONCOLON
%token TEMPLATE
%token ERROR
%token NEW DELETE
legal basetypes. */
%token SIGNED_KEYWORD LONG SHORT INT_KEYWORD CONST_KEYWORD VOLATILE_KEYWORD DOUBLE_KEYWORD
-%token <sval> VARIABLE
+%token <sval> DOLLAR_VARIABLE
%token <opcode> ASSIGN_MODIFY
{ write_exp_elt_opcode (pstate, UNOP_SIZEOF); }
;
-exp : exp ARROW name
+exp : ALIGNOF '(' type_exp ')' %prec UNARY
+ { write_exp_elt_opcode (pstate, UNOP_ALIGNOF); }
+ ;
+
+exp : exp ARROW field_name
{ write_exp_elt_opcode (pstate, STRUCTOP_PTR);
write_exp_string (pstate, $3);
write_exp_elt_opcode (pstate, STRUCTOP_PTR); }
;
-exp : exp ARROW name COMPLETE
+exp : exp ARROW field_name COMPLETE
{ mark_struct_expression (pstate);
write_exp_elt_opcode (pstate, STRUCTOP_PTR);
write_exp_string (pstate, $3);
{ write_exp_elt_opcode (pstate, STRUCTOP_MPTR); }
;
-exp : exp '.' name
+exp : exp '.' field_name
{ write_exp_elt_opcode (pstate, STRUCTOP_STRUCT);
write_exp_string (pstate, $3);
write_exp_elt_opcode (pstate, STRUCTOP_STRUCT); }
;
-exp : exp '.' name COMPLETE
+exp : exp '.' field_name COMPLETE
{ mark_struct_expression (pstate);
write_exp_elt_opcode (pstate, STRUCTOP_STRUCT);
write_exp_string (pstate, $3);
;
function_method: exp '(' parameter_typelist ')' const_or_volatile
- { int i;
- VEC (type_ptr) *type_list = $3;
- struct type *type_elt;
- LONGEST len = VEC_length (type_ptr, type_list);
+ {
+ std::vector<struct type *> *type_list = $3;
+ LONGEST len = type_list->size ();
write_exp_elt_opcode (pstate, TYPE_INSTANCE);
/* Save the const/volatile qualifiers as
write_exp_elt_longcst (pstate,
follow_type_instance_flags ());
write_exp_elt_longcst (pstate, len);
- for (i = 0;
- VEC_iterate (type_ptr, type_list, i, type_elt);
- ++i)
+ for (type *type_elt : *type_list)
write_exp_elt_type (pstate, type_elt);
write_exp_elt_longcst(pstate, len);
write_exp_elt_opcode (pstate, TYPE_INSTANCE);
- VEC_free (type_ptr, type_list);
}
;
exp : variable
;
-exp : VARIABLE
+exp : DOLLAR_VARIABLE
{
write_dollar_variable (pstate, $1);
}
is important for example for "p
*__errno_location()". */
symbol *alias_target
- = (msymbol.minsym->type != mst_text_gnu_ifunc
+ = ((msymbol.minsym->type != mst_text_gnu_ifunc
+ && msymbol.minsym->type != mst_data_gnu_ifunc)
? find_function_alias_target (msymbol)
: NULL);
if (alias_target != NULL)
ptr_operator_ts: ptr_operator
{
$$ = get_type_stack ();
- /* This cleanup is eventually run by
- c_parse. */
- make_cleanup (type_stack_cleanup, $$);
+ cpstate->type_stacks.emplace_back ($$);
}
;
;
func_mod: '(' ')'
- { $$ = NULL; }
+ {
+ $$ = new std::vector<struct type *>;
+ cpstate->type_lists.emplace_back ($$);
+ }
| '(' parameter_typelist ')'
{ $$ = $2; }
;
type : ptype
;
-typebase /* Implements (approximately): (type-qualifier)* type-specifier */
+/* Implements (approximately): (type-qualifier)* type-specifier.
+
+ When type-specifier is only ever a single word, like 'float' then these
+ arrive as pre-built TYPENAME tokens thanks to the classify_name
+ function. However, when a type-specifier can contain multiple words,
+ for example 'double' can appear as just 'double' or 'long double', and
+ similarly 'long' can appear as just 'long' or in 'long double', then
+ these type-specifiers are parsed into their own tokens in the function
+ lex_one_token and the ident_tokens array. These separate tokens are all
+ recognised here. */
+typebase
: TYPENAME
{ $$ = $1.type; }
| INT_KEYWORD
{ check_parameter_typelist ($1); }
| nonempty_typelist ',' DOTDOTDOT
{
- VEC_safe_push (type_ptr, $1, NULL);
+ $1->push_back (NULL);
check_parameter_typelist ($1);
$$ = $1;
}
nonempty_typelist
: type
{
- VEC (type_ptr) *typelist = NULL;
- VEC_safe_push (type_ptr, typelist, $1);
+ std::vector<struct type *> *typelist
+ = new std::vector<struct type *>;
+ cpstate->type_lists.emplace_back (typelist);
+
+ typelist->push_back ($1);
$$ = typelist;
}
| nonempty_typelist ',' type
{
- VEC_safe_push (type_ptr, $1, $3);
+ $1->push_back ($3);
$$ = $1;
}
;
}
;
-
+/* This rule exists in order to allow some tokens that would not normally
+ match the 'name' rule to appear as fields within a struct. The example
+ that initially motivated this was the RISC-V target which models the
+ floating point registers as a union with fields called 'float' and
+ 'double'. The 'float' string becomes a TYPENAME token and can appear
+ anywhere a 'name' can, however 'double' is its own token,
+ DOUBLE_KEYWORD, and doesn't match the 'name' rule.*/
+field_name
+ : name
+ | DOUBLE_KEYWORD { $$ = typename_stoken ("double"); }
+ | INT_KEYWORD { $$ = typename_stoken ("int"); }
+ | LONG { $$ = typename_stoken ("long"); }
+ | SHORT { $$ = typename_stoken ("short"); }
+ | SIGNED_KEYWORD { $$ = typename_stoken ("signed"); }
+ | UNSIGNED { $$ = typename_stoken ("unsigned"); }
+ ;
name : NAME { $$ = $1.stoken; }
| BLOCKNAME { $$ = $1.stoken; }
st.ptr = buf;
/* The toplevel (c_parse) will free the memory allocated here. */
- make_cleanup (free, buf);
+ cpstate->strings.emplace_back (buf);
+ return st;
+};
+
+/* Returns a stoken of the type named TYPE. */
+
+static struct stoken
+typename_stoken (const char *type)
+{
+ struct stoken st = { type, 0 };
+ st.length = strlen (type);
return st;
};
/* Validate a parameter typelist. */
static void
-check_parameter_typelist (VEC (type_ptr) *params)
+check_parameter_typelist (std::vector<struct type *> *params)
{
struct type *type;
int ix;
- for (ix = 0; VEC_iterate (type_ptr, params, ix, type); ++ix)
+ for (ix = 0; ix < params->size (); ++ix)
{
+ type = (*params)[ix];
if (type != NULL && TYPE_CODE (check_typedef (type)) == TYPE_CODE_VOID)
{
if (ix == 0)
{
- if (VEC_length (type_ptr, params) == 1)
+ if (params->size () == 1)
{
/* Ok. */
break;
}
- VEC_free (type_ptr, params);
error (_("parameter types following 'void'"));
}
else
- {
- VEC_free (type_ptr, params);
- error (_("'void' invalid as parameter type"));
- }
+ error (_("'void' invalid as parameter type"));
}
}
}
parse_number (struct parser_state *par_state,
const char *buf, int len, int parsed_float, YYSTYPE *putithere)
{
- /* FIXME: Shouldn't these be unsigned? We don't deal with negative values
- here, and we do kind of silly things like cast to unsigned. */
- LONGEST n = 0;
- LONGEST prevn = 0;
+ ULONGEST n = 0;
+ ULONGEST prevn = 0;
ULONGEST un;
int i = 0;
len -= 2;
}
/* Handle suffixes: 'f' for float, 'l' for long double. */
- else if (len >= 1 && tolower (p[len - 1]) == 'f')
+ else if (len >= 1 && TOLOWER (p[len - 1]) == 'f')
{
putithere->typed_val_float.type
= parse_type (par_state)->builtin_float;
len -= 1;
}
- else if (len >= 1 && tolower (p[len - 1]) == 'l')
+ else if (len >= 1 && TOLOWER (p[len - 1]) == 'l')
{
putithere->typed_val_float.type
= parse_type (par_state)->builtin_long_double;
on 0x123456789 when LONGEST is 32 bits. */
if (c != 'l' && c != 'u' && n != 0)
{
- if ((unsigned_p && (ULONGEST) prevn >= (ULONGEST) n))
+ if (unsigned_p && prevn >= n)
error (_("Numeric constant too large."));
}
prevn = n;
the case where it is we just always shift the value more than
once, with fewer bits each time. */
- un = (ULONGEST)n >> 2;
+ un = n >> 2;
if (long_p == 0
&& (un >> (gdbarch_int_bit (parse_gdbarch (par_state)) - 2)) == 0)
{
if (output)
obstack_grow_str (output, "\\x");
++tokptr;
- if (!isxdigit (*tokptr))
+ if (!ISXDIGIT (*tokptr))
error (_("\\x escape without a following hex digit"));
- while (isxdigit (*tokptr))
+ while (ISXDIGIT (*tokptr))
{
if (output)
obstack_1grow (output, *tokptr);
if (output)
obstack_grow_str (output, "\\");
for (i = 0;
- i < 3 && isdigit (*tokptr) && *tokptr != '8' && *tokptr != '9';
+ i < 3 && ISDIGIT (*tokptr) && *tokptr != '8' && *tokptr != '9';
++i)
{
if (output)
obstack_1grow (output, *tokptr);
}
++tokptr;
- if (!isxdigit (*tokptr))
+ if (!ISXDIGIT (*tokptr))
error (_("\\%c escape without a following hex digit"), c);
- for (i = 0; i < len && isxdigit (*tokptr); ++i)
+ for (i = 0; i < len && ISXDIGIT (*tokptr); ++i)
{
if (output)
obstack_1grow (output, *tokptr);
{".*", DOT_STAR, BINOP_END, FLAG_CXX}
};
-/* Identifier-like tokens. */
+/* Identifier-like tokens. Only type-specifiers than can appear in
+ multi-word type names (for example 'double' can appear in 'long
+ double') need to be listed here. type-specifiers that are only ever
+ single word (like 'float') are handled by the classify_name function. */
static const struct token ident_tokens[] =
{
{"unsigned", UNSIGNED, OP_NULL, 0},
{"struct", STRUCT, OP_NULL, 0},
{"signed", SIGNED_KEYWORD, OP_NULL, 0},
{"sizeof", SIZEOF, OP_NULL, 0},
+ {"_Alignof", ALIGNOF, OP_NULL, 0},
+ {"alignof", ALIGNOF, OP_NULL, FLAG_CXX},
{"double", DOUBLE_KEYWORD, OP_NULL, 0},
{"false", FALSEKEYWORD, OP_NULL, FLAG_CXX},
{"class", CLASS, OP_NULL, FLAG_CXX},
{"typeid", TYPEID, OP_TYPEID, FLAG_CXX}
};
-/* When we find that lexptr (the global var defined in parse.c) is
- pointing at a macro invocation, we expand the invocation, and call
- scan_macro_expansion to save the old lexptr here and point lexptr
- into the expanded text. When we reach the end of that, we call
- end_macro_expansion to pop back to the value we saved here. The
- macro expansion code promises to return only fully-expanded text,
- so we don't need to "push" more than one level.
-
- This is disgusting, of course. It would be cleaner to do all macro
- expansion beforehand, and then hand that to lexptr. But we don't
- really know where the expression ends. Remember, in a command like
-
- (gdb) break *ADDRESS if CONDITION
-
- we evaluate ADDRESS in the scope of the current frame, but we
- evaluate CONDITION in the scope of the breakpoint's location. So
- it's simply wrong to try to macro-expand the whole thing at once. */
-static const char *macro_original_text;
-
-/* We save all intermediate macro expansions on this obstack for the
- duration of a single parse. The expansion text may sometimes have
- to live past the end of the expansion, due to yacc lookahead.
- Rather than try to be clever about saving the data for a single
- token, we simply keep it all and delete it after parsing has
- completed. */
-static struct obstack expansion_obstack;
static void
scan_macro_expansion (char *expansion)
char *copy;
/* We'd better not be trying to push the stack twice. */
- gdb_assert (! macro_original_text);
+ gdb_assert (! cpstate->macro_original_text);
/* Copy to the obstack, and then free the intermediate
expansion. */
- copy = (char *) obstack_copy0 (&expansion_obstack, expansion,
+ copy = (char *) obstack_copy0 (&cpstate->expansion_obstack, expansion,
strlen (expansion));
xfree (expansion);
/* Save the old lexptr value, so we can return to it when we're done
parsing the expanded text. */
- macro_original_text = lexptr;
+ cpstate->macro_original_text = lexptr;
lexptr = copy;
}
static int
scanning_macro_expansion (void)
{
- return macro_original_text != 0;
+ return cpstate->macro_original_text != 0;
}
static void
finished_macro_expansion (void)
{
/* There'd better be something to pop back to. */
- gdb_assert (macro_original_text);
+ gdb_assert (cpstate->macro_original_text);
/* Pop back to the original text. */
- lexptr = macro_original_text;
- macro_original_text = 0;
-}
-
-static void
-scan_macro_cleanup (void *dummy)
-{
- if (macro_original_text)
- finished_macro_expansion ();
-
- obstack_free (&expansion_obstack, NULL);
+ lexptr = cpstate->macro_original_text;
+ cpstate->macro_original_text = 0;
}
/* Return true iff the token represents a C++ cast operator. */
last_was_structop = true;
goto symbol; /* Nope, must be a symbol. */
}
- /* FALL THRU into number case. */
+ /* FALL THRU. */
case '0':
case '1':
case '@':
{
const char *p = &tokstart[1];
- size_t len = strlen ("entry");
if (parse_language (par_state)->la_language == language_objc)
{
size_t len = strlen ("selector");
if (strncmp (p, "selector", len) == 0
- && (p[len] == '\0' || isspace (p[len])))
+ && (p[len] == '\0' || ISSPACE (p[len])))
{
lexptr = p + len;
return SELECTOR;
goto parse_string;
}
- while (isspace (*p))
+ while (ISSPACE (*p))
p++;
- if (strncmp (p, "entry", len) == 0 && !isalnum (p[len])
+ size_t len = strlen ("entry");
+ if (strncmp (p, "entry", len) == 0 && !c_ident_is_alnum (p[len])
&& p[len] != '_')
{
lexptr = &p[len];
}
}
- if (!(c == '_' || c == '$'
- || (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z')))
+ if (!(c == '_' || c == '$' || c_ident_is_alpha (c)))
/* We must have come across a bad character (e.g. ';'). */
error (_("Invalid character '%c' in expression."), c);
/* It's a name. See how long it is. */
namelen = 0;
for (c = tokstart[namelen];
- (c == '_' || c == '$' || (c >= '0' && c <= '9')
- || (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') || c == '<');)
+ (c == '_' || c == '$' || c_ident_is_alnum (c) || c == '<');)
{
/* Template parameter lists are part of the name.
FIXME: This mishandles `print $a<4&&$a>3'. */
}
if (*tokstart == '$')
- return VARIABLE;
+ return DOLLAR_VARIABLE;
if (parse_completion && *lexptr == '\0')
saw_name_at_eof = 1;
}
/* An object of this type is pushed on a FIFO by the "outer" lexer. */
-typedef struct
+struct token_and_value
{
int token;
YYSTYPE value;
-} token_and_value;
-
-DEF_VEC_O (token_and_value);
+};
/* A FIFO of tokens that have been read but not yet returned to the
parser. */
-static VEC (token_and_value) *token_fifo;
+static std::vector<token_and_value> token_fifo;
/* Non-zero if the lexer should return tokens from the FIFO. */
static int popping;
const struct block *search_block;
bool is_quoted_name, last_lex_was_structop;
- if (popping && !VEC_empty (token_and_value, token_fifo))
+ if (popping && !token_fifo.empty ())
goto do_pop;
popping = 0;
/* Read any sequence of alternating "::" and name-like tokens into
the token FIFO. */
current.value = yylval;
- VEC_safe_push (token_and_value, token_fifo, ¤t);
+ token_fifo.push_back (current);
last_was_coloncolon = current.token == COLONCOLON;
while (1)
{
Subsequent ones do not have any special meaning. */
current.token = lex_one_token (pstate, &ignore);
current.value = yylval;
- VEC_safe_push (token_and_value, token_fifo, ¤t);
+ token_fifo.push_back (current);
if ((last_was_coloncolon && current.token != NAME)
|| (!last_was_coloncolon && current.token != COLONCOLON))
/* We always read one extra token, so compute the number of tokens
to examine accordingly. */
- last_to_examine = VEC_length (token_and_value, token_fifo) - 2;
+ last_to_examine = token_fifo.size () - 2;
next_to_examine = 0;
- current = *VEC_index (token_and_value, token_fifo, next_to_examine);
+ current = token_fifo[next_to_examine];
++next_to_examine;
name_obstack.clear ();
while (next_to_examine <= last_to_examine)
{
- token_and_value *next;
+ token_and_value next;
- next = VEC_index (token_and_value, token_fifo, next_to_examine);
+ next = token_fifo[next_to_examine];
++next_to_examine;
- if (next->token == NAME && last_was_coloncolon)
+ if (next.token == NAME && last_was_coloncolon)
{
int classification;
- yylval = next->value;
+ yylval = next.value;
classification = classify_inner_name (pstate, search_block,
context_type);
/* We keep going until we either run out of names, or until
/* We don't want to put a leading "::" into the name. */
obstack_grow_str (&name_obstack, "::");
}
- obstack_grow (&name_obstack, next->value.sval.ptr,
- next->value.sval.length);
+ obstack_grow (&name_obstack, next.value.sval.ptr,
+ next.value.sval.length);
yylval.sval.ptr = (const char *) obstack_base (&name_obstack);
yylval.sval.length = obstack_object_size (&name_obstack);
context_type = yylval.tsym.type;
}
- else if (next->token == COLONCOLON && !last_was_coloncolon)
+ else if (next.token == COLONCOLON && !last_was_coloncolon)
last_was_coloncolon = 1;
else
{
if (checkpoint > 0)
{
current.value.sval.ptr
- = (const char *) obstack_copy0 (&expansion_obstack,
+ = (const char *) obstack_copy0 (&cpstate->expansion_obstack,
current.value.sval.ptr,
current.value.sval.length);
- VEC_replace (token_and_value, token_fifo, 0, ¤t);
+ token_fifo[0] = current;
if (checkpoint > 1)
- VEC_block_remove (token_and_value, token_fifo, 1, checkpoint - 1);
+ token_fifo.erase (token_fifo.begin () + 1,
+ token_fifo.begin () + checkpoint);
}
do_pop:
- current = *VEC_index (token_and_value, token_fifo, 0);
- VEC_ordered_remove (token_and_value, token_fifo, 0);
+ current = token_fifo[0];
+ token_fifo.erase (token_fifo.begin ());
yylval = current.value;
return current.token;
}
int
c_parse (struct parser_state *par_state)
{
- int result;
- struct cleanup *back_to;
-
/* Setting up the parser state. */
scoped_restore pstate_restore = make_scoped_restore (&pstate);
gdb_assert (par_state != NULL);
pstate = par_state;
+ c_parse_state cstate;
+ scoped_restore cstate_restore = make_scoped_restore (&cpstate, &cstate);
+
gdb::unique_xmalloc_ptr<struct macro_scope> macro_scope;
if (expression_context_block)
scoped_restore restore_macro_scope
= make_scoped_restore (&expression_macro_scope, macro_scope.get ());
- /* Initialize macro expansion code. */
- obstack_init (&expansion_obstack);
- gdb_assert (! macro_original_text);
- /* Note that parsing (within yyparse) freely installs cleanups
- assuming they'll be run here (below). */
- back_to = make_cleanup (scan_macro_cleanup, 0);
-
scoped_restore restore_yydebug = make_scoped_restore (&yydebug,
parser_debug);
last_was_structop = false;
saw_name_at_eof = 0;
- VEC_free (token_and_value, token_fifo);
+ token_fifo.clear ();
popping = 0;
name_obstack.clear ();
- result = yyparse ();
- do_cleanups (back_to);
-
- return result;
+ return yyparse ();
}
#ifdef YYBISON
break;
case NSSTRING:
- case VARIABLE:
+ case DOLLAR_VARIABLE:
parser_fprintf (file, "sval<%s>", copy_name (value.sval));
break;
#endif
-void
+static void
yyerror (const char *msg)
{
if (prev_lexptr)
lexptr = prev_lexptr;
- error (_("A %s in expression, near `%s'."), (msg ? msg : "error"), lexptr);
+ error (_("A %s in expression, near `%s'."), msg, lexptr);
}