def setupTypes(ast, entryFunc, inputs, nondets, entry, typedefs):
for fun in ast.ext:
if isinstance(fun, c_ast.Decl) and isinstance(fun.type,
c_ast.FuncDecl):
if fun.name.startswith('__VERIFIER_nondet_'):
info = {}
info['type'] = ext_c_generator.GnuCGenerator().visit(fun.type)
info['line'] = fun.coord.line
nondets[fun.name] = info
if isinstance(fun, c_ast.Decl) and isinstance(
fun.type, ext_c_parser.FuncDeclExt):
if fun.name.startswith('__VERIFIER_nondet_'):
info = {}
info['type'] = ext_c_generator.GnuCGenerator().visit(fun)
info['type'] = re.sub(r'^extern ', '', info['type'])
info['line'] = fun.coord.line
nondets[fun.name] = info
elif isinstance(fun, c_ast.FuncDef):
inputs[fun.decl.name] = {}
if fun.body.block_items:
for d in fun.body.block_items:
if isinstance(d, c_ast.Decl):
info = {}
typestr = ext_c_generator.GnuCGenerator().visit(d)
typestr = re.sub(r'\b%s\b' % d.name, '', typestr)
if typedefs.get(typestr):
typestr = typedefs[typestr]
info['type'] = typestr
info['line'] = d.coord.line
if d.init is None:
inputs[fun.decl.name][d.name] = info
if fun.decl.name == entryFunc:
entry['type'] = ext_c_generator.GnuCGenerator().visit(
fun.decl.type)
entry['line'] = fun.coord.line
elif isinstance(fun, c_ast.Typedef):
name = fun.name
if isinstance(fun.type, ext_c_parser.FuncDeclExt):
typestr = ext_c_generator.GnuCGenerator().visit(
fun.type.type) + ('()' + '(' +
ext_c_generator.GnuCGenerator().visit(
fun.type.args) + ')')
name = fun.name + ' (*)'
elif isinstance(fun.type, c_ast.PtrDecl) and isinstance(
fun.type.type, ext_c_parser.FuncDeclExt):
typestr = ext_c_generator.GnuCGenerator().visit(
fun.type.type.type) + (
'(*)' + '(' + ext_c_generator.GnuCGenerator().visit(
fun.type.type.args) + ')')
else:
typestr = ext_c_generator.GnuCGenerator().visit(fun.type)
while typedefs.get(typestr):
typestr = typedefs.get(typestr)
typedefs[name] = typestr
def preprocess(path):
text = preprocess_file(path)
cparser = GnuCParser()
ast = cparser.parse(text, path)
generator = ext_c_generator.GnuCGenerator()
with open(path, "w") as o:
o.write(generator.visit(ast))
return ast
def start_prediction(id, pred_dir, req_file):
P, ast = run_clf_predict(pred_dir, req_file)
F = to_func(ast)
generator = ext_c_generator.GnuCGenerator()
for name, ast in F.items():
path = os.path.join(pred_dir, name+".c")
with open(path, "w") as o:
out = generator.visit(ast)
o.write(out)
def test_double_pointer():
src = """
typedef struct Error {
int dummy;
} Error;
void func_with_p2pp(const char *, Error **);
"""
import pycparserext.ext_c_parser as ext_c_parser
import pycparserext.ext_c_generator as ext_c_generator
parser = ext_c_parser.GnuCParser()
ast = parser.parse(src)
gen = ext_c_generator.GnuCGenerator()
ast.show()
assert gen.visit(ast).find("func_with_p2pp(const char *, Error **)") != -1
def test_no_added_attr():
src = """
char* foo() {
return "";
}
int main() {
return 0;
}
"""
import pycparserext.ext_c_parser as ext_c_parser
import pycparserext.ext_c_generator as ext_c_generator
parser = ext_c_parser.GnuCParser()
ast = parser.parse(src)
gen = ext_c_generator.GnuCGenerator()
print(gen.visit(ast))
assert "attr" not in gen.visit(ast)
def test_pointer_reproduction():
src = """
struct foo {
const char *bar;
const char *baz;
};
int main () {
return 0;
}
"""
import pycparserext.ext_c_parser as ext_c_parser
import pycparserext.ext_c_generator as ext_c_generator
parser = ext_c_parser.GnuCParser()
ast = parser.parse(src)
gen = ext_c_generator.GnuCGenerator()
print(gen.visit(ast))
def processWitness(witness, benchmark, bitwidth):
try:
root = ElementTree.parse(witness).getroot()
except:
eprint("INVALID WITNESS FILE: failed to parse XML")
sys.exit(1)
# print(ElementTree.tostring(root))
ns = {'graphml': 'http://graphml.graphdrawing.org/xmlns'}
graph = root.find('./graphml:graph', ns)
if graph is None:
eprint("INVALID WITNESS FILE: failed to parse XML - no graph node")
sys.exit(1)
entryFun = validateConfig(graph, ns, witness, benchmark, bitwidth)
benchmarkString = ''
with tempfile.NamedTemporaryFile() as fp:
# preprocess and remove __attribute__
subprocess.check_call([
'gcc', '-D__attribute__(x)=', '-x', 'c', '-E', benchmark, '-o',
fp.name
])
with open(fp.name, 'r') as b:
needStructBody = False
skipAsm = False
for line in b:
# rewrite some GCC extensions
"""
line = re.sub(r'__extension__\s*\(\{\s*if\s*\(0\)\s*;\s*else\s+(__assert_fail\s*\("0",\s*".*",\s*\d+,\s*__extension__\s+__PRETTY_FUNCTION__\s*\));\s*\}\)', r'\1', line)
"""
line = re.sub(r'__extension__', '', line)
"""
line = re.sub(r'__restrict', 'restrict', line)
line = re.sub(r'__inline__', 'inline', line)
line = re.sub(r'__inline', 'inline', line)
line = re.sub(r'__const', 'const', line)
"""
line = re.sub(r'__signed__', 'signed', line)
"""
line = re.sub(r'__builtin_va_list', 'int', line)
"""
# a hack for some C-standards violating code in LDV benchmarks
if needStructBody and re.match(r'^\s*}\s*;\s*$', line):
line = 'int __dummy; ' + line
needStructBody = False
elif needStructBody:
needStructBody = re.match(r'^\s*$', line) is not None
elif re.match(r'^\s*struct\s+[a-zA-Z0-9_]+\s*{\s*$', line):
needStructBody = True
# remove inline asm
if re.match(
r'^\s*__asm__(\s+volatile)?\s*\("([^"]|\\")*"[^;]*$',
line):
skipAsm = True
elif skipAsm and re.search(r'\)\s*;\s*$', line):
line = '\n'
skipAsm = False
line = '\n'
if (skipAsm or re.match(
r'^\s*__asm__(\s+volatile)?\s*\("([^"]|\\")*"[^;]*\)\s*;\s*$',
line)):
line = '\n'
# remove asm renaming
line = re.sub(r'__asm__\s*\(""\s+"[a-zA-Z0-9_]+"\)', '', line)
benchmarkString += line
parser = ext_c_parser.GnuCParser()
ast = parser.parse(benchmarkString, filename=benchmark)
# ast.show(showcoord=True, buf=sys.stderr)
inputs = {}
nondets = {}
entry = {}
typedefs = {}
setupTypes(ast, entryFun, inputs, nondets, entry, typedefs)
assert entry
watch = {}
setupWatch(ast, watch)
trace = {}
entryNode = buildTrace(graph, ns, trace)
values = []
n = entryNode
missing_nondets = set(nondets)
while trace[n].get('target') is not None:
if trace[n].get('assumption') is not None:
# assumptions may use = or ==
a = re.sub(r'==', '=', trace[n]['assumption'])
a = re.sub(r'\\result', '__SV_COMP_result', a)
# we may be missing typedefs used in type casts
a_copy = a
if re.search(r'\(\s*[a-zA-Z_][a-zA-Z0-9_]*.*\)', a):
# do two rounds - strictly speaking, we'd need a fixed point here
for t in typedefs:
if t.endswith(' (*)'):
a = re.sub(r'%s' % re.escape(t), typedefs[t], a)
else:
a = re.sub(r'\b%s\b' % re.escape(t), typedefs[t], a)
for t in typedefs:
if t.endswith(' (*)'):
a = re.sub(r'%s' % re.escape(t), typedefs[t], a)
else:
a = re.sub(r'\b%s\b' % re.escape(t), typedefs[t], a)
wrapped = 'void foo() { ' + a + ';}'
try:
block_items = parser.parse(wrapped).ext[0].body.block_items
except:
eprint('Failed to parse ' + wrapped + '(expanded from ' +
a_copy + ')')
raise
for a_ast in block_items:
if isinstance(a_ast, c_ast.Assignment):
f = trace[n].get('assumption.scope')
v = ext_c_generator.GnuCGenerator().visit(a_ast.rvalue)
if (trace[n].get('startline') is not None and watch.get(
int(trace[n]['startline'])) is not None):
w = watch[int(trace[n]['startline'])]
values.append([w, v])
if w in missing_nondets:
missing_nondets.remove(w)
elif (f is not None and isinstance(a_ast.lvalue, c_ast.ID)
and inputs.get(f) is not None
and inputs[f].get(a_ast.lvalue.name) is not None):
values.append([f, a_ast.lvalue.name, v])
# else:
# print(trace[n]['startline'])
# a_ast.show()
# else:
# print(trace[n]['startline'])
# a_ast.show()
n = trace[n]['target']
if watch and not values:
eprint('inputs: ')
eprint(inputs)
eprint('nondets: ')
eprint(nondets)
eprint('watch: ')
eprint(watch)
eprint(
"WARNING: no input values found in witness file, behaviour of harness may be undefined"
)
print('IN:')
print(' ENTRY {n}()@[file {f} line {l}]'.format(n=entryFun,
f=benchmark,
l=entry['line']))
for v in values:
if len(v) == 3:
info = inputs[v[0]][v[1]]
print(
' {t} {n}@[file {f} line {l} function {fun}]={value}'.format(
t=info['type'],
n=v[1],
f=benchmark,
l=info['line'],
fun=v[0],
value=v[2]))
else:
info = nondets[v[0]]
print(' {t}@[file {f} line {l}]={value}'.format(t=info['type'],
f=benchmark,
l=info['line'],
value=v[1]))
for n in missing_nondets:
info = nondets[n]
print(' {t}@[file {f} line {l}]=0'.format(t=info['type'],
f=benchmark,
l=info['line']))