void c_typecheck_baset::typecheck_symbol_type(typet &type)
{
{
// add prefix
symbol_typet &symbol_type=to_symbol_type(type);
symbol_type.set_identifier(add_language_prefix(symbol_type.get_identifier()));
}
// adjust identifier, if needed
replace_symbol(type);
const irep_idt &identifier=
to_symbol_type(type).get_identifier();
symbol_tablet::symbolst::const_iterator s_it=
symbol_table.symbols.find(identifier);
if(s_it==symbol_table.symbols.end())
{
err_location(type);
str << "type symbol `" << identifier << "' not found";
throw 0;
}
const symbolt &symbol=s_it->second;
if(!symbol.is_type)
{
err_location(type);
throw "expected type symbol";
}
if(symbol.is_macro)
{
// overwrite, but preserve (add) any qualifiers and other flags
c_qualifierst c_qualifiers(type);
bool is_packed=type.get_bool(ID_C_packed);
irept alignment=type.find(ID_C_alignment);
c_qualifiers+=c_qualifierst(symbol.type);
type=symbol.type;
c_qualifiers.write(type);
if(is_packed) type.set(ID_C_packed, true);
if(alignment.is_not_nil()) type.set(ID_C_alignment, alignment);
}
// CPROVER extensions
if(symbol.base_name=="__CPROVER_rational")
{
type=rational_typet();
}
else if(symbol.base_name=="__CPROVER_integer")
{
type=integer_typet();
}
}
void c_typecastt::implicit_typecast_arithmetic(
exprt &expr,
c_typet c_type)
{
typet new_type;
const typet &expr_type=ns.follow(expr.type());
switch(c_type)
{
case PTR:
if(expr_type.id()==ID_array)
{
new_type.id(ID_pointer);
new_type.subtype()=expr_type.subtype();
break;
}
return;
case BOOL: new_type=bool_typet(); break;
case CHAR: assert(false); // should always be promoted to int
case UCHAR: assert(false); // should always be promoted to int
case SHORT: assert(false); // should always be promoted to int
case USHORT: assert(false); // should always be promoted to int
case INT: new_type=int_type(); break;
case UINT: new_type=uint_type(); break;
case LONG: new_type=long_int_type(); break;
case ULONG: new_type=long_uint_type(); break;
case LONGLONG: new_type=long_long_int_type(); break;
case ULONGLONG: new_type=long_long_uint_type(); break;
case SINGLE: new_type=float_type(); break;
case DOUBLE: new_type=double_type(); break;
case LONGDOUBLE: new_type=long_double_type(); break;
case RATIONAL: new_type=rational_typet(); break;
case REAL: new_type=real_typet(); break;
case INTEGER: new_type=integer_typet(); break;
case COMPLEX: return; // do nothing
default: return;
}
if(new_type!=expr_type)
{
if(new_type.id()==ID_pointer &&
expr_type.id()==ID_array)
{
exprt index_expr(ID_index, expr_type.subtype());
index_expr.reserve_operands(2);
index_expr.move_to_operands(expr);
index_expr.copy_to_operands(gen_zero(index_type()));
expr=exprt(ID_address_of, new_type);
expr.move_to_operands(index_expr);
}
else
do_typecast(expr, new_type);
}
}
exprt pointer_logict::pointer_expr(
const pointert &pointer,
const typet &type) const
{
if(pointer.object==null_object) // NULL?
{
if(pointer.offset==0)
{
constant_exprt result(type);
result.set_value(ID_NULL);
return result;
}
else
{
constant_exprt null(type);
null.set_value(ID_NULL);
return plus_exprt(null,
from_integer(pointer.offset, integer_typet()));
}
}
else if(pointer.object==invalid_object) // INVALID?
{
constant_exprt result(type);
result.set_value("INVALID");
return result;
}
if(pointer.object>=objects.size())
{
constant_exprt result(type);
result.set_value("INVALID-"+i2string(pointer.object));
return result;
}
const exprt &object_expr=objects[pointer.object];
exprt deep_object=object_rec(pointer.offset, type, object_expr);
exprt result;
if(type.id()==ID_pointer)
result=exprt(ID_address_of, type);
else if(type.id()==ID_reference)
result=exprt("reference_to", type);
else
assert(0);
result.copy_to_operands(deep_object);
return result;
}
void c_typecastt::implicit_typecast_arithmetic(
exprt &expr,
c_typet c_type)
{
typet new_type;
const typet &expr_type=ns.follow(expr.type());
switch(c_type)
{
case PTR:
if(expr_type.id()==ID_array)
{
new_type.id(ID_pointer);
new_type.subtype()=expr_type.subtype();
break;
}
return;
case BOOL: assert(false); // should always be promoted to int
case CHAR: assert(false); // should always be promoted to int
case UCHAR: assert(false); // should always be promoted to int
case SHORT: assert(false); // should always be promoted to int
case USHORT: assert(false); // should always be promoted to int
case INT: new_type=signed_int_type(); break;
case UINT: new_type=unsigned_int_type(); break;
case LONG: new_type=signed_long_int_type(); break;
case ULONG: new_type=unsigned_long_int_type(); break;
case LONGLONG: new_type=signed_long_long_int_type(); break;
case ULONGLONG: new_type=unsigned_long_long_int_type(); break;
case SINGLE: new_type=float_type(); break;
case DOUBLE: new_type=double_type(); break;
case LONGDOUBLE: new_type=long_double_type(); break;
case FLOAT128: new_type=ieee_float_spect::quadruple_precision().to_type(); break;
case RATIONAL: new_type=rational_typet(); break;
case REAL: new_type=real_typet(); break;
case INTEGER: new_type=integer_typet(); break;
case COMPLEX: return; // do nothing
default: return;
}
if(new_type!=expr_type)
do_typecast(expr, new_type);
}
codet java_bytecode_convertt::convert_instructions(
const instructionst &instructions,
const code_typet &method_type)
{
// Run a worklist algorithm, assuming that the bytecode has not
// been tampered with. See "Leroy, X. (2003). Java bytecode
// verification: algorithms and formalizations. Journal of Automated
// Reasoning, 30(3-4), 235-269." for a more complete treatment.
// first pass: get targets and map addresses to instructions
struct converted_instructiont
{
converted_instructiont(
const instructionst::const_iterator &it,
const codet &_code):source(it), code(_code), done(false)
{
}
instructionst::const_iterator source;
std::list<unsigned> successors;
std::set<unsigned> predecessors;
codet code;
stackt stack;
bool done;
};
typedef std::map<unsigned, converted_instructiont> address_mapt;
address_mapt address_map;
std::set<unsigned> targets;
for(instructionst::const_iterator
i_it=instructions.begin();
i_it!=instructions.end();
i_it++)
{
std::pair<address_mapt::iterator, bool> a_entry=
address_map.insert(std::make_pair(
i_it->address,
converted_instructiont(i_it, code_skipt())));
assert(a_entry.second);
// addresses are strictly increasing, hence we must have inserted
// a new maximal key
assert(a_entry.first==--address_map.end());
if(i_it->statement!="goto" &&
i_it->statement!="return" &&
!(i_it->statement==patternt("?return")) &&
i_it->statement!="athrow")
{
instructionst::const_iterator next=i_it;
if(++next!=instructions.end())
a_entry.first->second.successors.push_back(next->address);
}
if(i_it->statement=="goto" ||
i_it->statement==patternt("if_?cmp??") ||
i_it->statement==patternt("if??") ||
i_it->statement=="ifnonnull" ||
i_it->statement=="ifnull")
{
assert(!i_it->args.empty());
const unsigned target=safe_string2unsigned(
id2string(to_constant_expr(i_it->args[0]).get_value()));
targets.insert(target);
a_entry.first->second.successors.push_back(target);
}
else if(i_it->statement=="tableswitch" ||
i_it->statement=="lookupswitch")
{
bool is_label=true;
for(instructiont::argst::const_iterator
a_it=i_it->args.begin();
a_it!=i_it->args.end();
a_it++, is_label=!is_label)
{
if(is_label)
{
const unsigned target=safe_string2unsigned(
id2string(to_constant_expr(*a_it).get_value()));
targets.insert(target);
a_entry.first->second.successors.push_back(target);
}
}
}
}
for(address_mapt::iterator
it=address_map.begin();
it!=address_map.end();
++it)
{
for(unsigned s : it->second.successors)
{
address_mapt::iterator a_it=address_map.find(s);
assert(a_it!=address_map.end());
a_it->second.predecessors.insert(it->first);
}
}
std::set<unsigned> working_set;
if(!instructions.empty())
working_set.insert(instructions.front().address);
while(!working_set.empty())
{
std::set<unsigned>::iterator cur=working_set.begin();
address_mapt::iterator a_it=address_map.find(*cur);
assert(a_it!=address_map.end());
working_set.erase(cur);
if(a_it->second.done) continue;
working_set.insert(a_it->second.successors.begin(),
a_it->second.successors.end());
instructionst::const_iterator i_it=a_it->second.source;
stack.swap(a_it->second.stack);
a_it->second.stack.clear();
codet &c=a_it->second.code;
assert(stack.empty() ||
a_it->second.predecessors.size()<=1 ||
has_prefix(stack.front().get_string(ID_C_base_name),
"$stack"));
irep_idt statement=i_it->statement;
exprt arg0=i_it->args.size()>=1?i_it->args[0]:nil_exprt();
exprt arg1=i_it->args.size()>=2?i_it->args[1]:nil_exprt();
const bytecode_infot &bytecode_info=get_bytecode_info(statement);
// deal with _idx suffixes
if(statement.size()>=2 &&
statement[statement.size()-2]=='_' &&
isdigit(statement[statement.size()-1]))
{
arg0=constant_exprt(
std::string(id2string(statement), statement.size()-1, 1),
integer_typet());
statement=std::string(id2string(statement), 0, statement.size()-2);
}
exprt::operandst op=pop(bytecode_info.pop);
exprt::operandst results;
results.resize(bytecode_info.push, nil_exprt());
if(statement=="aconst_null")
{
assert(results.size()==1);
results[0]=gen_zero(java_reference_type(void_typet()));
}
else if(statement=="athrow")
{
assert(op.size()==1 && results.size()==1);
side_effect_expr_throwt throw_expr;
throw_expr.add_source_location()=i_it->source_location;
throw_expr.copy_to_operands(op[0]);
c=code_expressiont(throw_expr);
results[0]=op[0];
}
else if(statement=="checkcast")
{
// checkcast throws an exception in case a cast of object
// on stack to given type fails.
// The stack isn't modified.
assert(op.size()==1 && results.size()==1);
results[0]=op[0];
}
else if(statement=="invokedynamic")
{
// not used in Java
code_typet &code_type=to_code_type(arg0.type());
const code_typet::parameterst ¶meters(code_type.parameters());
pop(parameters.size());
const typet &return_type=code_type.return_type();
if(return_type.id()!=ID_empty)
{
results.resize(1);
results[0]=nil_exprt();
}
}
else if(statement=="invokeinterface" ||
statement=="invokespecial" ||
statement=="invokevirtual" ||
statement=="invokestatic")
{
const bool use_this(statement != "invokestatic");
const bool is_virtual(
statement == "invokevirtual" || statement == "invokeinterface");
code_typet &code_type=to_code_type(arg0.type());
code_typet::parameterst ¶meters(code_type.parameters());
if(use_this)
{
if(parameters.empty() || !parameters[0].get_this())
{
const empty_typet empty;
pointer_typet object_ref_type(empty);
code_typet::parametert this_p(object_ref_type);
this_p.set_this();
this_p.set_base_name("this");
parameters.insert(parameters.begin(), this_p);
}
}
code_function_callt call;
call.add_source_location()=i_it->source_location;
call.arguments() = pop(parameters.size());
// double-check a bit
if(use_this)
{
const exprt &this_arg=call.arguments().front();
assert(this_arg.type().id()==ID_pointer);
}
// do some type adjustment for the arguments,
// as Java promotes arguments
for(unsigned i=0; i<parameters.size(); i++)
{
const typet &type=parameters[i].type();
if(type==java_boolean_type() ||
type==java_char_type() ||
type==java_byte_type() ||
type==java_short_type())
{
assert(i<call.arguments().size());
call.arguments()[i].make_typecast(type);
}
}
// do some type adjustment for return values
const typet &return_type=code_type.return_type();
if(return_type.id()!=ID_empty)
{
// return types are promoted in Java
call.lhs()=tmp_variable("return", return_type);
exprt promoted=java_bytecode_promotion(call.lhs());
results.resize(1);
results[0]=promoted;
}
assert(arg0.id()==ID_virtual_function);
// does the function symbol exist?
irep_idt id=arg0.get(ID_identifier);
if(symbol_table.symbols.find(id)==symbol_table.symbols.end())
{
// no, create stub
symbolt symbol;
symbol.name=id;
symbol.base_name=arg0.get(ID_C_base_name);
symbol.type=arg0.type();
symbol.value.make_nil();
symbol.mode=ID_java;
symbol_table.add(symbol);
}
if(is_virtual)
{
// dynamic binding
assert(use_this);
assert(!call.arguments().empty());
call.function()=arg0;
}
else
{
// static binding
/*if(id == "java::java.lang.String.charAt:(I)C")
call.function()=symbol_exprt("java::__CPROVER_uninterpreted_char_at", arg0.type());
else*/
call.function()=symbol_exprt(arg0.get(ID_identifier), arg0.type());
}
call.function().add_source_location()=i_it->source_location;
c = call;
}
else if(statement=="return")
{
assert(op.empty() && results.empty());
c=code_returnt();
}
else if(statement==patternt("?return"))
{
// Return types are promoted in java, so this might need
// conversion.
assert(op.size()==1 && results.empty());
exprt r=op[0];
if(r.type()!=method_return_type) r=typecast_exprt(r, method_return_type);
c=code_returnt(r);
}
else if(statement==patternt("?astore"))
{
assert(op.size()==3 && results.empty());
char type_char=statement[0];
exprt pointer=
typecast_exprt(op[0], java_array_type(type_char));
const dereference_exprt deref(pointer, pointer.type().subtype());
const member_exprt data_ptr(
deref, "data", pointer_typet(java_type_from_char(type_char)));
plus_exprt data_plus_offset(data_ptr, op[1], data_ptr.type());
typet element_type=data_ptr.type().subtype();
const dereference_exprt element(data_plus_offset, element_type);
c=code_assignt(element, op[2]);
}
else if(statement==patternt("?store"))
{
// store value into some local variable
assert(op.size()==1 && results.empty());
exprt var=variable(arg0, statement[0]);
const bool is_array('a' == statement[0]);
if(is_array)
var.type()=op[0].type();
c=code_assignt(var, op[0]);
}
else if(statement==patternt("?aload"))
{
assert(op.size() == 2 && results.size() == 1);
char type_char=statement[0];
exprt pointer=
typecast_exprt(op[0], java_array_type(type_char));
const dereference_exprt deref(pointer, pointer.type().subtype());
const member_exprt data_ptr(
deref, "data", pointer_typet(java_type_from_char(type_char)));
plus_exprt data_plus_offset(data_ptr, op[1], data_ptr.type());
typet element_type=data_ptr.type().subtype();
dereference_exprt element(data_plus_offset, element_type);
results[0]=java_bytecode_promotion(element);
}
else if(statement==patternt("?load"))
{
// load a value from a local variable
results[0]=variable(arg0, statement[0]);
}
else if(statement=="ldc" || statement=="ldc_w" ||
statement=="ldc2" || statement=="ldc2_w")
{
assert(op.empty() && results.size()==1);
// 1) Pushing a String causes a reference to a java.lang.String object
// to be constructed and pushed onto the operand stack.
// 2) Pushing an int or a float causes a primitive value to be pushed
// onto the stack.
// 3) Pushing a Class constant causes a reference to a java.lang.Class
// to be pushed onto the operand stack
if(arg0.id()==ID_java_string_literal)
{
// these need to be references to java.lang.String
results[0]=arg0;
symbol_typet string_type("java::java.lang.String");
results[0].type()=pointer_typet(string_type);
}
else if(arg0.id()==ID_type)
{
irep_idt class_id=arg0.type().get(ID_identifier);
symbol_typet java_lang_Class("java::java.lang.Class");
symbol_exprt symbol_expr(id2string(class_id)+"@class_model", java_lang_Class);
address_of_exprt address_of_expr(symbol_expr);
results[0]=address_of_expr;
}
else if(arg0.id()==ID_constant)
{
results[0]=arg0;
}
else
{
error() << "unexpected ldc argument" << eom;
throw 0;
}
}
else if(statement=="goto" || statement=="goto_w")
{
assert(op.empty() && results.empty());
irep_idt number=to_constant_expr(arg0).get_value();
code_gotot code_goto(label(number));
c=code_goto;
}
else if(statement=="iconst_m1")
{
assert(results.size()==1);
results[0]=from_integer(-1, java_int_type());
}
else if(statement==patternt("?const"))
{
assert(results.size() == 1);
const char type_char=statement[0];
const bool is_double('d' == type_char);
const bool is_float('f' == type_char);
if(is_double || is_float)
{
const ieee_float_spect spec(
is_float ?
ieee_float_spect::single_precision() :
ieee_float_spect::double_precision());
ieee_floatt value(spec);
const typet &arg_type(arg0.type());
if(ID_integer == arg_type.id())
value.from_integer(arg0.get_int(ID_value));
else
value.from_expr(to_constant_expr(arg0));
results[0] = value.to_expr();
}
else
{
const unsigned int value(arg0.get_unsigned_int(ID_value));
const typet type=java_type_from_char(statement[0]);
results[0] = as_number(value, type);
}
}
else if(statement==patternt("?ipush"))
{
assert(results.size()==1);
results[0]=typecast_exprt(arg0, java_int_type());
}
else if(statement==patternt("if_?cmp??"))
{
irep_idt number=to_constant_expr(arg0).get_value();
assert(op.size()==2 && results.empty());
code_ifthenelset code_branch;
const irep_idt cmp_op=get_if_cmp_operator(statement);
binary_relation_exprt condition(op[0], cmp_op, op[1]);
cast_if_necessary(condition);
code_branch.cond()=condition;
code_branch.then_case()=code_gotot(label(number));
code_branch.then_case().add_source_location()=i_it->source_location;
code_branch.add_source_location()=i_it->source_location;
c=code_branch;
}
else if(statement==patternt("if??"))
{
const irep_idt id=
statement=="ifeq"?ID_equal:
statement=="ifne"?ID_notequal:
statement=="iflt"?ID_lt:
statement=="ifge"?ID_ge:
statement=="ifgt"?ID_gt:
statement=="ifle"?ID_le:
(assert(false), "");
irep_idt number=to_constant_expr(arg0).get_value();
assert(op.size()==1 && results.empty());
code_ifthenelset code_branch;
code_branch.cond()=binary_relation_exprt(op[0], id, gen_zero(op[0].type()));
code_branch.cond().add_source_location()=i_it->source_location;
code_branch.then_case()=code_gotot(label(number));
code_branch.then_case().add_source_location()=i_it->source_location;
code_branch.add_source_location()=i_it->source_location;
c=code_branch;
}
else if(statement==patternt("ifnonnull"))
{
irep_idt number=to_constant_expr(arg0).get_value();
assert(op.size()==1 && results.empty());
code_ifthenelset code_branch;
const typecast_exprt lhs(op[0], pointer_typet());
const exprt rhs(gen_zero(lhs.type()));
code_branch.cond()=binary_relation_exprt(lhs, ID_notequal, rhs);
code_branch.then_case()=code_gotot(label(number));
code_branch.then_case().add_source_location()=i_it->source_location;
code_branch.add_source_location()=i_it->source_location;
c=code_branch;
}
else if(statement==patternt("ifnull"))
{
assert(op.size()==1 && results.empty());
irep_idt number=to_constant_expr(arg0).get_value();
code_ifthenelset code_branch;
const typecast_exprt lhs(op[0], pointer_typet(empty_typet()));
const exprt rhs(gen_zero(lhs.type()));
code_branch.cond()=binary_relation_exprt(lhs, ID_equal, rhs);
code_branch.then_case()=code_gotot(label(number));
code_branch.then_case().add_source_location()=i_it->source_location;
code_branch.add_source_location()=i_it->source_location;
c=code_branch;
}
else if(statement=="iinc")
{
code_assignt code_assign;
code_assign.lhs()=variable(arg0, 'i');
code_assign.rhs()=plus_exprt(
variable(arg0, 'i'),
typecast_exprt(arg1, java_int_type()));
c=code_assign;
}
else if(statement==patternt("?xor"))
{
assert(op.size()==2 && results.size()==1);
results[0]=bitxor_exprt(op[0], op[1]);
}
else if(statement==patternt("?or"))
{
assert(op.size()==2 && results.size()==1);
results[0]=bitor_exprt(op[0], op[1]);
}
else if(statement==patternt("?and"))
{
assert(op.size()==2 && results.size()==1);
results[0]=bitand_exprt(op[0], op[1]);
}
else if(statement==patternt("?shl"))
{
assert(op.size()==2 && results.size()==1);
results[0]=shl_exprt(op[0], op[1]);
}
else if(statement==patternt("?shr"))
{
assert(op.size()==2 && results.size()==1);
results[0]=ashr_exprt(op[0], op[1]);
}
else if(statement==patternt("?ushr"))
{
assert(op.size()==2 && results.size()==1);
const typet type(java_type_from_char(statement[0]));
const unsigned int width(type.get_unsigned_int(ID_width));
typet target=unsigned_long_int_type();
target.set(ID_width, width);
const typecast_exprt lhs(op[0], target);
const typecast_exprt rhs(op[1], target);
results[0]=lshr_exprt(lhs, rhs);
}
else if(statement==patternt("?add"))
{
assert(op.size()==2 && results.size()==1);
results[0]=plus_exprt(op[0], op[1]);
}
else if(statement==patternt("?sub"))
{
assert(op.size()==2 && results.size()==1);
results[0]=minus_exprt(op[0], op[1]);
}
else if(statement==patternt("?div"))
{
assert(op.size()==2 && results.size()==1);
results[0]=div_exprt(op[0], op[1]);
}
else if(statement==patternt("?mul"))
{
assert(op.size()==2 && results.size()==1);
results[0]=mult_exprt(op[0], op[1]);
}
else if(statement==patternt("?neg"))
{
assert(op.size()==1 && results.size()==1);
results[0]=unary_minus_exprt(op[0], op[0].type());
}
else if(statement==patternt("?rem"))
{
assert(op.size()==2 && results.size()==1);
if(statement=="frem" || statement=="drem")
results[0]=rem_exprt(op[0], op[1]);
else
results[0]=mod_exprt(op[0], op[1]);
}
else if(statement==patternt("?cmp"))
{
assert(op.size() == 2 && results.size() == 1);
// The integer result on the stack is:
// 0 if op[0] equals op[1]
// -1 if op[0] is less than op[1]
// 1 if op[0] is greater than op[1]
const typet t=java_int_type();
results[0]=
if_exprt(binary_relation_exprt(op[0], ID_equal, op[1]), gen_zero(t),
if_exprt(binary_relation_exprt(op[0], ID_gt, op[1]), from_integer(1, t),
from_integer(-1, t)));
}
else if(statement==patternt("?cmp?"))
{
assert(op.size()==2 && results.size()==1);
const floatbv_typet type(to_floatbv_type(java_type_from_char(statement[0])));
const ieee_float_spect spec(type);
const ieee_floatt nan(ieee_floatt::NaN(spec));
const constant_exprt nan_expr(nan.to_expr());
const int nan_value(statement[4] == 'l' ? -1 : 1);
const typet result_type(java_int_type());
const exprt nan_result(from_integer(nan_value, result_type));
// (value1 == NaN || value2 == NaN) ? nan_value : value1 < value2 ? -1 : value2 < value1 1 ? 1 : 0;
// (value1 == NaN || value2 == NaN) ? nan_value : value1 == value2 ? 0 : value1 < value2 -1 ? 1 : 0;
results[0]=
if_exprt(or_exprt(ieee_float_equal_exprt(nan_expr, op[0]), ieee_float_equal_exprt(nan_expr, op[1])), nan_result,
if_exprt(ieee_float_equal_exprt(op[0], op[1]), gen_zero(result_type),
if_exprt(binary_relation_exprt(op[0], ID_lt, op[1]), from_integer(-1, result_type), from_integer(1, result_type))));
}
else if(statement==patternt("?cmpl"))
{
assert(op.size()==2 && results.size()==1);
results[0]=binary_relation_exprt(op[0], ID_lt, op[1]);
}
else if(statement=="dup")
{
assert(op.size()==1 && results.size()==2);
results[0]=results[1]=op[0];
}
else if(statement=="dup_x1")
{
assert(op.size()==2 && results.size()==3);
results[0]=op[1];
results[1]=op[0];
results[2]=op[1];
}
else if(statement=="dup_x2")
{
assert(op.size()==3 && results.size()==4);
results[0]=op[2];
results[1]=op[0];
results[2]=op[1];
results[3]=op[2];
}
// dup2* behaviour depends on the size of the operands on the
// stack
else if(statement=="dup2")
{
assert(!stack.empty() && results.empty());
if(stack.back().type().get_unsigned_int(ID_width)==32)
op=pop(2);
else
op=pop(1);
results.insert(results.end(), op.begin(), op.end());
results.insert(results.end(), op.begin(), op.end());
}
else if(statement=="dup2_x1")
{
assert(!stack.empty() && results.empty());
if(stack.back().type().get_unsigned_int(ID_width)==32)
op=pop(3);
else
op=pop(2);
results.insert(results.end(), op.begin()+1, op.end());
results.insert(results.end(), op.begin(), op.end());
}
else if(statement=="dup2_x2")
{
assert(!stack.empty() && results.empty());
if(stack.back().type().get_unsigned_int(ID_width)==32)
op=pop(2);
else
op=pop(1);
assert(!stack.empty());
exprt::operandst op2;
if(stack.back().type().get_unsigned_int(ID_width)==32)
op2=pop(2);
else
op2=pop(1);
results.insert(results.end(), op.begin(), op.end());
results.insert(results.end(), op2.begin(), op2.end());
results.insert(results.end(), op.begin(), op.end());
}
else if(statement=="dconst")
{
assert(op.empty() && results.size()==1);
}
else if(statement=="fconst")
{
assert(op.empty() && results.size()==1);
}
else if(statement=="getfield")
{
assert(op.size()==1 && results.size()==1);
results[0]=to_member(op[0], arg0);
}
else if(statement=="getstatic")
{
assert(op.empty() && results.size()==1);
symbol_exprt symbol_expr(arg0.type());
symbol_expr.set_identifier(arg0.get_string(ID_class)+"."+arg0.get_string(ID_component_name));
results[0]=symbol_expr;
}
else if(statement=="putfield")
{
assert(op.size()==2 && results.size()==0);
c = code_assignt(to_member(op[0], arg0), op[1]);
}
else if(statement=="putstatic")
{
assert(op.size()==1 && results.empty());
symbol_exprt symbol_expr(arg0.type());
symbol_expr.set_identifier(arg0.get_string(ID_class)+"."+arg0.get_string(ID_component_name));
c=code_assignt(symbol_expr, op[0]);
}
else if(statement==patternt("?2?")) // i2c etc.
{
assert(op.size()==1 && results.size()==1);
results[0]=typecast_exprt(op[0], java_type_from_char(statement[2]));
}
else if(statement=="new")
{
// use temporary since the stack symbol might get duplicated
assert(op.empty() && results.size()==1);
const pointer_typet ref_type(arg0.type());
exprt java_new_expr=side_effect_exprt(ID_java_new, ref_type);
if(!i_it->source_location.get_line().empty())
java_new_expr.add_source_location()=i_it->source_location;
const exprt tmp=tmp_variable("new", ref_type);
c=code_assignt(tmp, java_new_expr);
results[0]=tmp;
}
else if(statement=="newarray" ||
statement=="anewarray")
{
// the op is the array size
assert(op.size()==1 && results.size()==1);
char element_type;
if(statement=="newarray")
{
irep_idt id=arg0.type().id();
if(id==ID_bool)
element_type='z';
else if(id==ID_char)
element_type='c';
else if(id==ID_float)
element_type='f';
else if(id==ID_double)
element_type='d';
else if(id==ID_byte)
element_type='b';
else if(id==ID_short)
element_type='s';
else if(id==ID_int)
element_type='i';
else if(id==ID_long)
element_type='j';
else
element_type='?';
}
else
element_type='a';
const pointer_typet ref_type=java_array_type(element_type);
side_effect_exprt java_new_array(ID_java_new_array, ref_type);
java_new_array.copy_to_operands(op[0]);
if(!i_it->source_location.get_line().empty())
java_new_array.add_source_location()=i_it->source_location;
const exprt tmp=tmp_variable("newarray", ref_type);
c=code_assignt(tmp, java_new_array);
results[0]=tmp;
}
else if(statement=="multianewarray")
{
// The first argument is the type, the second argument is the dimension.
// The size of each dimension is on the stack.
irep_idt number=to_constant_expr(arg1).get_value();
unsigned dimension=safe_c_str2unsigned(number.c_str());
op=pop(dimension);
assert(results.size()==1);
// arg0.type()
const pointer_typet ref_type=java_array_type('a');
side_effect_exprt java_new_array(ID_java_new_array, ref_type);
java_new_array.operands()=op;
if(!i_it->source_location.get_line().empty())
java_new_array.add_source_location()=i_it->source_location;
const exprt tmp=tmp_variable("newarray", ref_type);
c=code_assignt(tmp, java_new_array);
results[0]=tmp;
}
else if(statement=="arraylength")
{
assert(op.size()==1 && results.size()==1);
exprt pointer=
typecast_exprt(op[0], java_array_type(statement[0]));
const dereference_exprt array(pointer, pointer.type().subtype());
assert(pointer.type().subtype().id()==ID_symbol);
const member_exprt length(array, "length", java_int_type());
results[0]=length;
}
else if(statement=="tableswitch" ||
statement=="lookupswitch")
{
assert(op.size()==1 && results.size()==0);
// we turn into switch-case
code_switcht code_switch;
code_switch.add_source_location()=i_it->source_location;
code_switch.value()=op[0];
code_blockt code_block;
code_block.add_source_location()=i_it->source_location;
bool is_label=true;
for(instructiont::argst::const_iterator
a_it=i_it->args.begin();
a_it!=i_it->args.end();
a_it++, is_label=!is_label)
{
if(is_label)
{
code_switch_caset code_case;
code_case.add_source_location()=i_it->source_location;
irep_idt number=to_constant_expr(*a_it).get_value();
code_case.code()=code_gotot(label(number));
code_case.code().add_source_location()=i_it->source_location;
if(a_it==i_it->args.begin())
code_case.set_default();
else
{
instructiont::argst::const_iterator prev=a_it;
prev--;
code_case.case_op()=typecast_exprt(*prev, op[0].type());
code_case.case_op().add_source_location()=i_it->source_location;
}
code_block.add(code_case);
}
}
code_switch.body()=code_block;
c=code_switch;
}
else if(statement=="pop" || statement=="pop2")
{
// these are skips
c=code_skipt();
// pop2 removes two single-word items from the stack (e.g. two
// integers, or an integer and an object reference) or one
// two-word item (i.e. a double or a long).
// http://cs.au.dk/~mis/dOvs/jvmspec/ref-pop2.html
if(statement=="pop2" &&
op[0].type().get_unsigned_int(ID_width)==32)
pop(1);
}
else if(statement=="instanceof")
{
assert(op.size()==1 && results.size()==1);
results[0]=
binary_predicate_exprt(op[0], "java_instanceof", arg0);
}
else
{
c=codet(statement);
c.operands()=op;
}
if(!i_it->source_location.get_line().empty())
c.add_source_location()=i_it->source_location;
push(results);
a_it->second.done=true;
for(std::list<unsigned>::iterator
it=a_it->second.successors.begin();
it!=a_it->second.successors.end();
++it)
{
address_mapt::iterator a_it2=address_map.find(*it);
assert(a_it2!=address_map.end());
if(!stack.empty() && a_it2->second.predecessors.size()>1)
{
// copy into temporaries
code_blockt more_code;
// introduce temporaries when successor is seen for the first
// time
if(a_it2->second.stack.empty())
{
for(stackt::iterator s_it=stack.begin();
s_it!=stack.end();
++s_it)
{
symbol_exprt lhs=tmp_variable("$stack", s_it->type());
code_assignt a(lhs, *s_it);
more_code.copy_to_operands(a);
s_it->swap(lhs);
}
}
else
{
assert(a_it2->second.stack.size()==stack.size());
stackt::const_iterator os_it=a_it2->second.stack.begin();
for(stackt::iterator s_it=stack.begin();
s_it!=stack.end();
++s_it)
{
assert(has_prefix(os_it->get_string(ID_C_base_name),
"$stack"));
symbol_exprt lhs=to_symbol_expr(*os_it);
code_assignt a(lhs, *s_it);
more_code.copy_to_operands(a);
s_it->swap(lhs);
++os_it;
}
}
if(results.empty())
{
more_code.copy_to_operands(c);
c.swap(more_code);
}
else
{
c.make_block();
forall_operands(o_it, more_code)
c.copy_to_operands(*o_it);
}
}
a_it2->second.stack=stack;
}
}
// TODO: add exception handlers from exception table
// review successor computation of athrow!
code_blockt code;
// temporaries
for(const auto & var : tmp_vars)
{
code.add(code_declt(var));
}
for(const auto & it : address_map)
{
const unsigned address=it.first;
assert(it.first==it.second.source->address);
const codet &c=it.second.code;
if(targets.find(address)!=targets.end())
code.add(code_labelt(label(i2string(address)), c));
else if(c.get_statement()!=ID_skip)
code.add(c);
}
return code;
}
SCENARIO("does_type_preserve_const_correctness",
"[core][analyses][does_remove_const][does_type_preserve_const_correctness]")
{
symbol_tablet symbol_table;
namespacet ns(symbol_table);
goto_programt program;
does_remove_constt does_remove_const(program, ns);
does_remove_const_testt does_remove_const_test(does_remove_const);
GIVEN("Const and non-const primitive and pointers to primitives")
{
c_qualifierst const_qualifier;
const_qualifier.is_constant=true;
// const int
typet const_primitive_type=integer_typet();
const_qualifier.write(const_primitive_type);
// int
typet non_const_primitive_type=integer_typet();
// pointer (can be reassigned)
// to int (value can be changed)
// int *
typet pointer_to_int_type=pointer_typet(non_const_primitive_type);
// const pointer (can't be reassigned)
// to int (value can be changed)
// int * const
typet const_pointer_to_int_type=pointer_typet(non_const_primitive_type);
const_qualifier.write(const_pointer_to_int_type);
constant_exprt constant_exprt::integer_constant(unsigned v)
{
return constant_exprt(std::to_string(v), integer_typet());
}
void boolbvt::convert_member(const member_exprt &expr, bvt &bv)
{
const exprt &struct_op=expr.struct_op();
const typet &struct_op_type=ns.follow(struct_op.type());
const bvt &struct_bv=convert_bv(struct_op);
if(struct_op_type.id()==ID_union)
{
bv=convert_bv(
byte_extract_exprt(byte_extract_id(),
struct_op,
gen_zero(integer_typet()),
expr.type()));
return;
}
else if(struct_op_type.id()==ID_struct)
{
const irep_idt &component_name=expr.get_component_name();
const struct_typet::componentst &components=
to_struct_type(struct_op_type).components();
unsigned offset=0;
for(struct_typet::componentst::const_iterator
it=components.begin();
it!=components.end();
it++)
{
const typet &subtype=it->type();
unsigned sub_width=boolbv_width(subtype);
if(it->get_name()==component_name)
{
if(!base_type_eq(subtype, expr.type(), ns))
{
#if 0
std::cout << "DEBUG " << expr.pretty() << "\n";
#endif
throw "member: component type does not match: "+
subtype.to_string()+" vs. "+
expr.type().to_string();
}
bv.resize(sub_width);
assert(offset+sub_width<=struct_bv.size());
for(unsigned i=0; i<sub_width; i++)
bv[i]=struct_bv[offset+i];
return;
}
offset+=sub_width;
}
throw "component "+id2string(component_name)+" not found in structure";
}
else
throw "member takes struct or union operand";
}
void java_bytecode_parsert::rbytecode(
methodt::instructionst &instructions)
{
u4 code_length=read_u4();
u4 address;
size_t bytecode_index=0; // index of bytecode instruction
for(address=0; address<code_length; address++)
{
bool wide_instruction=false;
u4 start_of_instruction=address;
u1 bytecode=read_u1();
if(bytecode==0xc4) // wide
{
wide_instruction=true;
address++;
bytecode=read_u1();
}
instructions.push_back(instructiont());
instructiont &instruction=instructions.back();
instruction.statement=bytecodes[bytecode].mnemonic;
instruction.address=start_of_instruction;
instruction.source_location
.set_java_bytecode_index(std::to_string(bytecode_index));
switch(bytecodes[bytecode].format)
{
case ' ': // no further bytes
break;
case 'c': // a constant_pool index (one byte)
if(wide_instruction)
{
instruction.args.push_back(constant(read_u2()));
address+=2;
}
else
{
instruction.args.push_back(constant(read_u1()));
address+=1;
}
break;
case 'C': // a constant_pool index (two bytes)
instruction.args.push_back(constant(read_u2()));
address+=2;
break;
case 'b': // a signed byte
{
s1 c=read_u1();
instruction.args.push_back(from_integer(c, integer_typet()));
}
address+=1;
break;
case 'o': // two byte branch offset, signed
{
s2 offset=read_u2();
instruction
.args.push_back(from_integer(address+offset, integer_typet()));
}
address+=2;
break;
case 'O': // four byte branch offset, signed
{
s4 offset=read_u4();
instruction
.args.push_back(from_integer(address+offset, integer_typet()));
}
address+=4;
break;
case 'v': // local variable index (one byte)
{
u1 v=read_u1();
instruction.args.push_back(from_integer(v, integer_typet()));
}
address+=1;
break;
case 'V':
// local variable index (two bytes) plus two signed bytes
if(wide_instruction)
{
u2 v=read_u2();
instruction.args.push_back(from_integer(v, integer_typet()));
s2 c=read_u2();
instruction.args.push_back(from_integer(c, integer_typet()));
address+=4;
}
else // local variable index (one byte) plus one signed byte
{
u1 v=read_u1();
instruction.args.push_back(from_integer(v, integer_typet()));
s1 c=read_u1();
instruction.args.push_back(from_integer(c, integer_typet()));
address+=2;
}
break;
case 'I': // two byte constant_pool index plus two bytes
{
u2 c=read_u2();
instruction.args.push_back(constant(c));
u1 b1=read_u1();
instruction.args.push_back(from_integer(b1, integer_typet()));
u1 b2=read_u1();
instruction.args.push_back(from_integer(b2, integer_typet()));
}
address+=4;
break;
case 'L': // lookupswitch
{
u4 base_offset=address;
// first a pad to 32-bit align
while(((address+1)&3)!=0) { read_u1(); address++; }
// now default value
s4 default_value=read_u4();
instruction.args
.push_back(from_integer(base_offset+default_value, integer_typet()));
address+=4;
// number of pairs
u4 npairs=read_u4();
address+=4;
for(std::size_t i=0; i<npairs; i++)
{
s4 match=read_u4();
s4 offset=read_u4();
instruction.args.push_back(from_integer(match, integer_typet()));
instruction.args
.push_back(from_integer(base_offset+offset, integer_typet()));
address+=8;
}
}
break;
case 'T': // tableswitch
{
size_t base_offset=address;
// first a pad to 32-bit align
while(((address+1)&3)!=0) { read_u1(); address++; }
// now default value
s4 default_value=read_u4();
instruction.args
.push_back(from_integer(base_offset+default_value, integer_typet()));
address+=4;
// now low value
s4 low_value=read_u4();
address+=4;
// now high value
s4 high_value=read_u4();
address+=4;
// there are high-low+1 offsets, and they are signed
for(s4 i=low_value; i<=high_value; i++)
{
s4 offset=read_u4();
instruction.args.push_back(from_integer(i, integer_typet()));
instruction.args
.push_back(from_integer(base_offset+offset, integer_typet()));
address+=4;
}
}
break;
case 'm': // multianewarray: constant-pool index plus one unsigned byte
{
u2 c=read_u2(); // constant-pool index
instruction.args.push_back(constant(c));
u1 dimensions=read_u1(); // number of dimensions
instruction.args.push_back(from_integer(dimensions, integer_typet()));
address+=3;
}
break;
case 't': // array subtype, one byte
{
typet t;
switch(read_u1())
{
case T_BOOLEAN: t.id(ID_bool); break;
case T_CHAR: t.id(ID_char); break;
case T_FLOAT: t.id(ID_float); break;
case T_DOUBLE: t.id(ID_double); break;
case T_BYTE: t.id(ID_byte); break;
case T_SHORT: t.id(ID_short); break;
case T_INT: t.id(ID_int); break;
case T_LONG: t.id(ID_long); break;
default:{};
}
instruction.args.push_back(type_exprt(t));
}
address+=1;
break;
case 's': // a signed short
{
s2 s=read_u2();
instruction.args.push_back(from_integer(s, integer_typet()));
}
address+=2;
break;
default:
throw "unknown JVM bytecode instruction";
}
bytecode_index++;
}
if(address!=code_length)
{
error() << "bytecode length mismatch" << eom;
throw 0;
}
}