/* Process source files and output type information.
- Copyright (C) 2006, 2007, 2010, 2012 Free Software Foundation, Inc.
+ Copyright (C) 2006-2021 Free Software Foundation, Inc.
This file is part of GCC.
along with GCC; see the file COPYING3. If not see
<http://www.gnu.org/licenses/>. */
-#ifdef GENERATOR_FILE
-#include "bconfig.h"
-#else
+#ifdef HOST_GENERATOR_FILE
#include "config.h"
+#define GENERATOR_FILE 1
+#else
+#include "bconfig.h"
#endif
#include "system.h"
#include "gengtype.h"
"union",
"struct",
"enum",
- "VEC",
"...",
"ptr_alias",
"nested_ptr",
"a string constant",
"a character constant",
"an array declarator",
+ "a C++ keyword to ignore"
};
/* This array is indexed by token code minus FIRST_TOKEN_WITH_VALUE. */
"'\"%s\"'",
"\"'%s'\"",
"'[%s]'",
+ "'%s'",
};
/* Produce a printable representation for a token defined by CODE and
return v;
}
+/* As per require, but do not advance. */
+static const char *
+require_without_advance (int t)
+{
+ int u = token ();
+ const char *v = T.value;
+ if (u != t)
+ {
+ parse_error ("expected %s, have %s",
+ print_token (t, 0), print_token (u, v));
+ return 0;
+ }
+ return v;
+}
+
/* If the next token does not have one of the codes T1 or T2, report a
parse error; otherwise return the token's value. */
static const char *
return v;
}
+/* If the next token does not have one of the codes T1, T2, T3 or T4, report a
+ parse error; otherwise return the token's value. */
+static const char *
+require4 (int t1, int t2, int t3, int t4)
+{
+ int u = token ();
+ const char *v = advance ();
+ if (u != t1 && u != t2 && u != t3 && u != t4)
+ {
+ parse_error ("expected %s, %s, %s or %s, have %s",
+ print_token (t1, 0), print_token (t2, 0),
+ print_token (t3, 0), print_token (t4, 0),
+ print_token (u, v));
+ return 0;
+ }
+ return v;
+}
+
/* Near-terminals. */
/* C-style string constant concatenation: STRING+
/* The caller has detected a template declaration that starts
with TMPL_NAME. Parse up to the closing '>'. This recognizes
- simple template declarations of the form ID<ID1,ID2,...,IDn>.
+ simple template declarations of the form ID<ID1,ID2,...,IDn>,
+ potentially with a single level of indirection e.g.
+ ID<ID1 *, ID2, ID3 *, ..., IDn>.
It does not try to parse anything more sophisticated than that.
Returns the template declaration string "ID<ID1,ID2,...,IDn>". */
require_template_declaration (const char *tmpl_name)
{
char *str;
+ int num_indirections = 0;
/* Recognize the opening '<'. */
require ('<');
str = concat (tmpl_name, "<", (char *) 0);
/* Read the comma-separated list of identifiers. */
- while (token () != '>')
+ int depth = 1;
+ while (depth > 0)
{
- const char *id = require2 (ID, ',');
+ if (token () == ENUM)
+ {
+ advance ();
+ str = concat (str, "enum ", (char *) 0);
+ continue;
+ }
+ if (token () == NUM
+ || token () == ':'
+ || token () == '+')
+ {
+ str = concat (str, advance (), (char *) 0);
+ continue;
+ }
+ if (token () == '<')
+ {
+ advance ();
+ str = concat (str, "<", (char *) 0);
+ depth += 1;
+ continue;
+ }
+ if (token () == '>')
+ {
+ advance ();
+ str = concat (str, ">", (char *) 0);
+ depth -= 1;
+ continue;
+ }
+ const char *id = require4 (SCALAR, ID, '*', ',');
if (id == NULL)
- id = ",";
+ {
+ if (T.code == '*')
+ {
+ id = "*";
+ if (num_indirections++)
+ parse_error ("only one level of indirection is supported"
+ " in template arguments");
+ }
+ else
+ id = ",";
+ }
+ else
+ num_indirections = 0;
str = concat (str, id, (char *) 0);
}
-
- /* Recognize the closing '>'. */
- require ('>');
- str = concat (str, ">", (char *) 0);
-
return str;
}
-/* typedef_name: either an ID, or VEC(x,y), or a template type
- specification of the form ID<t1,t2,...,tn>.
-
- FIXME cxx-conversion. VEC(x,y) is currently translated to the
- template 'vec_t<x>'. This is to support the transition to C++ and
- avoid re-writing all the 'VEC(x,y)' declarations in the code. This
- needs to be fixed when the branch is merged into trunk. */
+/* typedef_name: either an ID, or a template type
+ specification of the form ID<t1,t2,...,tn>. */
static const char *
typedef_name (void)
{
- if (token () == VEC_TOKEN)
- {
- const char *c1, *r;
- advance ();
- require ('(');
- c1 = require2 (ID, SCALAR);
- require (',');
- require (ID);
- require (')');
- r = concat ("vec_t<", c1, ">", (char *) 0);
- free (CONST_CAST (char *, c1));
- return r;
- }
-
const char *id = require (ID);
if (token () == '<')
return require_template_declaration (id);
}
/* Absorb a sequence of tokens, possibly including ()[]{}-delimited
- expressions, until we encounter a semicolon outside any such
- delimiters; absorb that too. If IMMEDIATE is true, it is an error
- if the semicolon is not the first token encountered. */
+ expressions, until we encounter an end-of-statement marker (a ';' or
+ a '}') outside any such delimiters; absorb that too. */
+
static void
-consume_until_semi (bool immediate)
+consume_until_eos (void)
{
- if (immediate && token () != ';')
- require (';');
for (;;)
switch (token ())
{
case ';':
advance ();
return;
- default:
- advance ();
- break;
+
+ case '{':
+ consume_balanced ('{', '}');
+ return;
case '(':
consume_balanced ('(', ')');
break;
+
case '[':
consume_balanced ('[', ']');
break;
- case '{':
- consume_balanced ('{', '}');
- break;
case '}':
case ']':
case ')':
parse_error ("unmatched '%c' while scanning for ';'", token ());
- return;
+ return;
case EOF_TOKEN:
parse_error ("unexpected end of file while scanning for ';'");
return;
+
+ default:
+ advance ();
+ break;
}
}
/* Absorb a sequence of tokens, possibly including ()[]{}-delimited
expressions, until we encounter a comma or semicolon outside any
- such delimiters; absorb that too. If IMMEDIATE is true, it is an
- error if the comma or semicolon is not the first token encountered.
- Returns true if the loop ended with a comma. */
+ such delimiters; absorb that too. Returns true if the loop ended
+ with a comma. */
+
static bool
-consume_until_comma_or_semi (bool immediate)
+consume_until_comma_or_eos ()
{
- if (immediate && token () != ',' && token () != ';')
- require2 (',', ';');
for (;;)
switch (token ())
{
case ',':
advance ();
return true;
+
case ';':
advance ();
return false;
- default:
- advance ();
- break;
+
+ case '{':
+ consume_balanced ('{', '}');
+ return false;
case '(':
consume_balanced ('(', ')');
break;
+
case '[':
consume_balanced ('[', ']');
break;
- case '{':
- consume_balanced ('{', '}');
- break;
case '}':
case ']':
case EOF_TOKEN:
parse_error ("unexpected end of file while scanning for ',' or ';'");
return false;
+
+ default:
+ advance ();
+ break;
}
}
\f
/* One GTY(()) option:
ID str_optvalue_opt
| PTR_ALIAS type_optvalue
- | PARAM_IS type_optvalue
| NESTED_PTR nestedptr_optvalue
*/
static options_p
advance ();
return type_optvalue (prev, "ptr_alias");
- case PARAM_IS:
- return type_optvalue (prev, advance ());
-
case NESTED_PTR:
advance ();
return nestedptr_optvalue (prev);
return 0;
return gtymarker ();
}
+
+
\f
/* Declarators. The logic here is largely lifted from c-parser.c.
Note that we do not have to process abstract declarators, which can
return ty;
}
-static type_p inner_declarator (type_p, const char **, options_p *);
+static type_p inner_declarator (type_p, const char **, options_p *, bool);
/* direct_declarator:
'(' inner_declarator ')'
+ '(' \epsilon ')' <-- C++ ctors/dtors
gtymarker_opt ID array_and_function_declarators_opt
Subroutine of declarator, mutually recursive with inner_declarator;
- do not use elsewhere. */
+ do not use elsewhere.
+
+ IN_STRUCT is true if we are called while parsing structures or classes. */
+
static type_p
-direct_declarator (type_p ty, const char **namep, options_p *optsp)
+direct_declarator (type_p ty, const char **namep, options_p *optsp,
+ bool in_struct)
{
/* The first token in a direct-declarator must be an ID, a
GTY marker, or an open parenthesis. */
case GTY_TOKEN:
*optsp = gtymarker ();
/* fall through */
+
case ID:
*namep = require (ID);
+ /* If the next token is '(', we are parsing a function declaration.
+ Functions are ignored by gengtype, so we return NULL. */
+ if (token () == '(')
+ return NULL;
break;
case '(':
+ /* If the declarator starts with a '(', we have three options. We
+ are either parsing 'TYPE (*ID)' (i.e., a function pointer)
+ or 'TYPE(...)'.
+
+ The latter will be a constructor iff we are inside a
+ structure or class. Otherwise, it could be a typedef, but
+ since we explicitly reject typedefs inside structures, we can
+ assume that we found a ctor and return NULL. */
advance ();
- ty = inner_declarator (ty, namep, optsp);
+ if (in_struct && token () != '*')
+ {
+ /* Found a constructor. Find and consume the closing ')'. */
+ while (token () != ')')
+ advance ();
+ advance ();
+ /* Tell the caller to ignore this. */
+ return NULL;
+ }
+ ty = inner_declarator (ty, namep, optsp, in_struct);
require (')');
break;
+ case IGNORABLE_CXX_KEYWORD:
+ /* Any C++ keyword like 'operator' means that we are not looking
+ at a regular data declarator. */
+ return NULL;
+
default:
- parse_error ("expected '(', 'GTY', or an identifier, have %s",
+ parse_error ("expected '(', ')', 'GTY', or an identifier, have %s",
print_cur_token ());
/* Do _not_ advance if what we have is a close squiggle brace, as
we will get much better error recovery that way. */
direct_declarator
Mutually recursive subroutine of direct_declarator; do not use
- elsewhere. */
+ elsewhere.
+
+ IN_STRUCT is true if we are called while parsing structures or classes. */
static type_p
-inner_declarator (type_p ty, const char **namep, options_p *optsp)
+inner_declarator (type_p ty, const char **namep, options_p *optsp,
+ bool in_struct)
{
if (token () == '*')
{
type_p inner;
advance ();
- inner = inner_declarator (ty, namep, optsp);
+ inner = inner_declarator (ty, namep, optsp, in_struct);
if (inner == 0)
return 0;
else
return create_pointer (ty);
}
else
- return direct_declarator (ty, namep, optsp);
+ return direct_declarator (ty, namep, optsp, in_struct);
}
/* declarator: '*'+ direct_declarator
This is the sole public interface to this part of the grammar.
Arguments are the type known so far, a pointer to where the name
may be stored, and a pointer to where GTY options may be stored.
- Returns the final type. */
+
+ IN_STRUCT is true when we are called to parse declarators inside
+ a structure or class.
+
+ Returns the final type. */
static type_p
-declarator (type_p ty, const char **namep, options_p *optsp)
+declarator (type_p ty, const char **namep, options_p *optsp,
+ bool in_struct = false)
{
*namep = 0;
*optsp = 0;
advance ();
ty = create_pointer (ty);
}
- return direct_declarator (ty, namep, optsp);
+ return direct_declarator (ty, namep, optsp, in_struct);
}
\f
/* Types and declarations. */
(
type bitfield ';'
| type declarator bitfield? ( ',' declarator bitfield? )+ ';'
- )+
+ )*
Knows that such declarations must end with a close brace (or,
erroneously, at EOF).
const char *name;
bool another;
- do
+ while (token () != '}' && token () != EOF_TOKEN)
{
ty = type (&opts, true);
+ /* Ignore access-control keywords ("public:" etc). */
+ while (!ty && token () == IGNORABLE_CXX_KEYWORD)
+ {
+ const char *keyword = advance ();
+ if (strcmp (keyword, "public:") != 0
+ && strcmp (keyword, "private:") != 0
+ && strcmp (keyword, "protected:") != 0)
+ break;
+ ty = type (&opts, true);
+ }
+
if (!ty || token () == ':')
{
- consume_until_semi (false);
+ consume_until_eos ();
continue;
}
do
{
- dty = declarator (ty, &name, &dopts);
+ dty = declarator (ty, &name, &dopts, true);
+
/* There could be any number of weird things after the declarator,
notably bitfield declarations and __attribute__s. If this
function returns true, the last thing was a comma, so we have
more than one declarator paired with the current type. */
- another = consume_until_comma_or_semi (false);
+ another = consume_until_comma_or_eos ();
if (!dty)
continue;
}
while (another);
}
- while (token () != '}' && token () != EOF_TOKEN);
return nreverse_pairs (f);
}
/* Return true if OPTS contain the option named STR. */
-static bool
+bool
opts_have (options_p opts, const char *str)
{
for (options_p opt = opts; opt; opt = opt->next)
Returns a partial type; under some conditions (notably
"struct foo GTY((...)) thing;") it may write an options
structure to *OPTSP.
-*/
+
+ NESTED is true when parsing a declaration already known to have a
+ GTY marker. In these cases, typedef and enum declarations are not
+ allowed because gengtype only understands types at the global
+ scope. */
+
static type_p
type (options_p *optsp, bool nested)
{
return create_scalar_type (s);
case ID:
- case VEC_TOKEN:
s = typedef_name ();
return resolve_typedef (s, &lexer_line);
+ case IGNORABLE_CXX_KEYWORD:
+ /* By returning NULL here, we indicate to the caller that they
+ should ignore everything following this keyword up to the
+ next ';' or '}'. */
+ return NULL;
+
case STRUCT:
case UNION:
{
+ type_p base_class = NULL;
options_p opts = 0;
/* GTY annotations follow attribute syntax
GTY_BEFORE_ID is for union/struct declarations
/* Top-level structures that are not explicitly tagged GTY(())
are treated as mere forward declarations. This is because
there are a lot of structures that we don't need to know
- about, and some of those have weird macro stuff in them
- that we can't handle. */
+ about, and some of those have C++ and macro constructs that
+ we cannot handle. */
if (nested || token () == GTY_TOKEN)
{
is_gty = GTY_BEFORE_ID;
opts = gtymarker_opt ();
}
+ bool is_user_gty = opts_have (opts, "user");
+
+ if (token () == ':')
+ {
+ if (is_gty && !is_user_gty)
+ {
+ /* For GTY-marked types that are not "user", parse some C++
+ inheritance specifications.
+ We require single-inheritance from a non-template type. */
+ advance ();
+ const char *basename = require (ID);
+ /* This may be either an access specifier, or the base name. */
+ if (strcmp (basename, "public") == 0
+ || strcmp (basename, "protected") == 0
+ || strcmp (basename, "private") == 0)
+ basename = require (ID);
+ base_class = find_structure (basename, TYPE_STRUCT);
+ if (!base_class)
+ parse_error ("unrecognized base class: %s", basename);
+ require_without_advance ('{');
+ }
+ else
+ {
+ /* For types lacking GTY-markings, skip over C++ inheritance
+ specification (and thus avoid having to parse e.g. template
+ types). */
+ while (token () != '{')
+ advance ();
+ }
+ }
+
if (is_gty)
{
- bool is_user_gty = opts_have (opts, "user");
if (token () == '{')
{
pair_p fields;
fields = NULL;
kind = TYPE_USER_STRUCT;
consume_balanced ('{', '}');
+ return create_user_defined_type (s, &lexer_line);
}
- return new_structure (s, kind, &lexer_line, fields, opts);
+ return new_structure (s, kind, &lexer_line, fields, opts,
+ base_class);
}
}
else if (token () == '{')
return find_structure (s, kind);
}
+ case TYPEDEF:
+ /* In C++, a typedef inside a struct/class/union defines a new
+ type for that inner scope. We cannot support this in
+ gengtype because we have no concept of scoping.
+
+ We handle typedefs in the global scope separately (see
+ parse_file), so if we find a 'typedef', we must be inside
+ a struct. */
+ gcc_assert (nested);
+ parse_error ("typedefs not supported in structures marked with "
+ "automatic GTY markers. Use GTY((user)) to mark "
+ "this structure.");
+ advance ();
+ return NULL;
+
case ENUM:
advance ();
if (token () == ID)
if (token () == '{')
consume_balanced ('{', '}');
+
+ /* If after parsing the enum we are at the end of the statement,
+ and we are currently inside a structure, then this was an
+ enum declaration inside this scope.
+
+ We cannot support this for the same reason we cannot support
+ 'typedef' inside structures (see the TYPEDEF handler above).
+ If this happens, emit an error and return NULL. */
+ if (nested && token () == ';')
+ {
+ parse_error ("enum definitions not supported in structures marked "
+ "with automatic GTY markers. Use GTY((user)) to mark "
+ "this structure.");
+ advance ();
+ return NULL;
+ }
+
return create_scalar_type (s);
default:
/* Yet another place where we could have junk (notably attributes)
after the declarator. */
- another = consume_until_comma_or_semi (false);
+ another = consume_until_comma_or_eos ();
if (dty)
do_typedef (name, dty, &lexer_line);
}
projects / thirdparty / gcc.git / blobdiff