def __init__(self):
self.input_spec = [] # [(formatter, var),(formatter, var)]
self.output_spec = []
self.tests = [] # string inputs
self.test_number = int(config.get("test_number"))
self.output_folder = config.get("output_folder")
self.analyzer = config.get("analyzer")
def get_context(doc, segment, linear_value):
node = segment[0]
context = []
functions = {}
if not domutil.beyond_parent(node, linear_value):
context = domutil.get_previous_AST_elements(node, linear_value)
context.extend(segment)
elif not domutil.beyond_method(node, linear_value):
linear_value -= domutil.get_previous_AST_element_number(node) + 1
node = domutil.get_parent_AST_element(node)
context, functions = get_context(doc, [node], linear_value)
else:
function_node = domutil.get_function_element(node)
name_node = domutil.get_first_child_by_tagname(function_node, 'name')
name = domutil.get_text_content(name_node)
for call_node in doc.getElementsByTagName('call'):
call_name_node = domutil.get_first_child_by_tagname(call_node, 'name')
call_name = domutil.get_text_content(call_name_node)
if call_name == name:
if config.get('show_context_search_debug'):
blue('beyound method '+name + ' and found call node: ')
linear_value -= domutil.get_previous_AST_element_number_until_function(node) + 1
context, functions = get_context(doc, [call_node], linear_value)
functions[call_name] = function_node
return context, functions
if config.get('show_context_search_debug') == 'true':
yellow('beyound method '+name+' but did not found call node: ')
return context, functions
def instrument_output(self):
logger.info('instrumenting output')
if config.get('instrument_loop') == 'true':
if config.get('instrument_loop_parameter'):
self.do_instrument('@Post', 'variable')
self.do_instrument('@Inner')
if config.get('instrument_post') == 'true':
self.do_instrument('@Post')
if config.get('instrument_pre') == 'true':
self.do_instrument('@Pre')
def __init__(self):
self.output_folder = config.get('output_folder')
self.doc = domutil.get_doc_from_c_file(self.output_folder+'/generate.c')
self.support_doc = domutil.get_doc_from_c_file(self.output_folder+'/support.h')
# the fields used in the generated program.
# this is a big heuristic
self.needed_fields = set()
# dict to hold {alias: name}
self.alias_to_name = {}
self.code = ''
self.output_folder = config.get('output_folder')
self.struct_limit = 0
self.helium_size_defined = False
def generate_input(self, type_name, var_name, recursive_count=1):
result = ''
type_component = typeutil.parse_type(type_name)
base = type_component['base']
pointer = type_component['pointer']
array = type_component['array']
if array:
if config.get('handle_array') == 'true':
return self.generate_input_for_array(type_name, var_name)
else:
return ''
# alloc
if pointer:
# char array should be treated separately
if config.get('handle_array') == 'true' and base == 'char':
result += self.generate_input_for_array('char[]', var_name+'_array')
result += var_name +' = '+var_name + '_array;\n'
return result;
result += generate_alloc_string(type_name, var_name)
# init
if typeutil.is_primitive_type(base):
result += generate_primitive_input(type_name, var_name)
elif local.check_struct_or_typedef(base.replace('struct','').strip()):
if config.get('handle_struct') != 'true': return ''
code = local.resolve_single_simple(base.replace('struct','').strip(), t='st')
if '{' in code.strip() and code.strip().startswith('typedef struct'):
code = code.replace('typedef','').strip()
code = code[:code.rfind('}')+1] + ';'
result += self.generate_input_for_struct(type_name, var_name, code, recursive_count)
elif code.strip().startswith('struct'):
result += self.generate_input_for_struct(type_name, var_name, code, recursive_count)
elif code.startswith('typedef'):
if code.startswith('typedef struct'):
code = local.resolve_single_simple(base.replace('struct', '').strip(), t='s')
if code:
result += self.generate_input_for_struct(type_name, var_name, code, recursive_count)
else:
_,original = syntaxutil.parse_typedef_code(code)
result += self.generate_input(original+pointer+array, var_name, recursive_count)
else:
logger.warning('Not recognized code for generate input')
else:
sys_type = systype.resolve(base)
if sys_type:
result += self.generate_input(sys_type+pointer+array, var_name)
else:
logger.warning('the type '+type_name+' is not solved for init')
return result
def analyze(self):
analyzer = config.get("analyzer")
logger.info("doing analyzing, analyzer: " + analyzer)
if analyzer == "invariant":
template_result = compare.parse(self.output_folder + "/output.csv")
regression_result = ""
try:
regression_result = check_output(
"Rscript ~/github/PyHelium/helium/analyzer/advancedInvariant.R "
+ self.output_folder
+ "/output.csv",
shell=True,
stderr=DEVNULL,
)
regression_result = regression_result.decode("utf8")
except CalledProcessError as e:
pass
result = template_result + "\n" + regression_result
elif analyzer == "recursive":
result = recursive_compare.compare("output.csv", self.output_folder + "/output.csv")
elif analyzer == "store":
# result = recursive_compare.compare('output.csv', self.output_folder+'/output.csv')
store(len(self.input_spec))
result = ""
else:
red("unsupported analyzer")
exit(1)
return result
def get_segments(doc):
results = []
code_selection = config.get('code_selection')
if code_selection == 'annotation' or code_selection == 'assertion':
comments = doc.getElementsByTagName('comment')
for comment in comments:
comment_text = domutil.get_text_content(comment)
if '@HeliumStart' in comment_text:
results.append(get_annotation_segment(comment))
elif code_selection == 'assertion':
comments = doc.getElementsByTagName('comment')
for comment in comments:
comment_text = domutil.get_text_content(comment)
if '@HeliumAssert' in comment_text:
if domutil.has_next_AST_element(comment):
results.append([domutil.get_next_AST_element(comment)])
elif code_selection == 'loop':
for while_loop_node in doc.getElementsByTagName('while'):
results.append([while_loop_node])
for for_loop_node in doc.getElementsByTagName('for'):
results.append([for_loop_node])
else:
logger.error('unsupported code selection method: ' + repr(code_selection))
exit(1)
return results
def __init__(self):
self.doc = None
self.context = None
self.functions = None
self.inputs = None
self.resolved = {}
self.unresolved = None
self.output_folder = config.get("output_folder")
def test(self):
logger.info("testing")
self.input_spec, self.output_spec = get_io_specification(self.output_folder + "/generate.c")
# if self.input_spec and self.output_spec:
self.generate_input()
self.run()
if config.get("run_analyze") == "true":
return self.analyze()
def handle_segment(self, doc, segment):
if not segment: return
segment_text = ''
for node in segment:
segment_text += domutil.get_text_content(node) + '\n'
if config.get('max_segment_size'):
if segment_text.count('\n') > int(config.get('max_segment_size')):
return
if config.get('show_segment_size') == 'true':
print('segment size: '+str(segment_text.count('\n')))
if config.get('show_parent_function_size') == 'true':
function_node = domutil.get_parent_by_tagname(segment[0], 'function')
function_size = 0
if function_node:
text = domutil.get_text_content(function_node)
function_size = text.count('\n')
print('parent function size: '+str(function_size))
if config.get('show_segment') == 'true':
print(segment_text)
instrument_segment(doc, segment)
max_linear_value = int(config.get('max_linear_value'))
stop_criteria = StopCriteria()
for i in range(max_linear_value+1):
logger.info('context search: ' + str(i))
if config.get('show_context_search_value') == 'true':
print('context search value: '+str(i))
context, functions = get_context(doc, segment, i)
if not context: break
if self.builder.build(doc, context, functions):
if config.get('run_test') == 'true':
result = self.tester.test()
if self.analyzer == 'recursive':
if result == True:
green('found equivalent loops for recursive call')
print_nodes(segment)
elif self.analyzer == 'invariant':
if config.get('show_analyze_result') == 'true':
print(result)
stop_criteria.add(result)
stable_result = stop_criteria.get_stable()
if stable_result:
green(stable_result)
if config.get('interact_after_test') == 'true':
input('Enter to continue ...')
remove_instrument(doc)
def start(self):
for root,_,files in os.walk(self.folder):
for f in files:
if f.endswith('.c') and f != 'generate.c':
if config.get('file_include'):
if f.split('/')[-1] not in file_include: continue
if config.get('file_exclude'):
if f.split('/')[-1] in file_exclude: continue
filename = os.path.join(root, f)
doc = domutil.get_doc_from_c_file(filename)
segments = get_segments(doc)
if segments:
if config.get('show_filename') == 'true':
print(filename)
if config.get('interact_before_segment') == 'true':
c = input('Enter to start the file, anything to break the file ...')
if c: continue
for segment in segments:
self.handle_segment(doc, segment)
def build(path):
logger.info('building tag file')
this['path'] = path
if config.get('build_ctags') == 'true':
tag_file = ctags.build_ctags(path=path, recursive=True, opts='--languages=c,c++ -n --c-kinds=+x --exclude=helium_out')
else:
tag_file = os.path.join(path, 'tags')
with codecs.open(tag_file, encoding='utf-8') as output:
content = output.readlines()
output.close()
this['cache'] = ctags.parse_tag_lines(content)
def generate(self):
os.makedirs(self.output_folder, exist_ok=True)
with open(self.output_folder + "/generate.c", "w") as f:
f.write('#include "support.h"\n')
# main
f.write("int main() {\n")
f.write("//@HeliumInput\n")
for var_name in self.inputs:
type_name = self.inputs[var_name]
type_component = typeutil.parse_type(type_name)
base = type_component["base"]
array = type_component["array"]
pointer = type_component["pointer"]
f.write(base + pointer + " " + var_name + array + ";\n")
f.write("//@HeliumInputEnd\n")
f.write("/**********Context********/\n")
context_size = 0
function_size = 0
context_text = ""
for node in self.context:
context_text += domutil.get_text_content(node)
context_text = re.sub(r"\breturn\b[^;\n]*;", "return 0;", context_text)
if config.get("show_context") == "true":
print(context_text)
context_size = context_text.count("\n") + 1
f.write(context_text)
f.write("\n}\n")
# functions
for function in self.functions.values():
text = domutil.get_text_content(function)
function_size += text.count("\n")
f.write(text)
f.write("\n")
if config.get("show_context_size") == "true":
print("context size: " + str(context_size))
if function_size != 0:
print("context function size: " + str(function_size))
f.close()
def compile(self):
logger.info("compiling")
# print out the generated and support.h loc
if config.get("show_generate_size") == "true":
loc = 0
loc += int(check_output("cat " + self.output_folder + "/support.h | wc -l", shell=True))
loc += int(check_output("cat " + self.output_folder + "/generate.c | wc -l", shell=True))
print("total line of code generated: " + str(loc))
call("make clean -C " + self.output_folder, shell=True, stdout=DEVNULL)
if config.get("show_compile_error_message") == "true":
stderr_value = None
else:
stderr_value = DEVNULL
if call("make -C " + self.output_folder, shell=True, stdout=DEVNULL, stderr=stderr_value) == 0:
# logger.critical('compile success')
if config.get("interact_compile_success") == "true":
input("Enter to continue ...")
return True
else:
# logger.error('compile error')
if config.get("interact_compile_error") == "true":
input("Enter to continue ...")
return False
def generate_input(self):
test_generation = config.get("test_generation")
if test_generation == "random":
self.generate_random_tests()
elif test_generation == "pair":
if len(self.input_spec) >= 2:
self.generate_pairwise_tests()
else:
logger.info("only one parameter, use random generation")
self.generate_random_tests()
elif test_generation == "predefined":
self.generate_predefined_tests()
else:
logger.warning("unsupported generation type")
def instrument_input(self):
logger.info('instrumenting input')
init_code = ''
input_nodes = annotation.get_input_nodes(self.doc)
if not input_nodes: return
for input_node in input_nodes:
# input_nodes should be 'decl_stmt'
type_name, var_name, _ = syntaxutil.parse_decl_stmt(input_node)
if config.get('handle_array') == 'true':
if '[' in type_name:
text = domutil.get_text_content(input_node)
self.code = self.code.replace(text, '')
# We need to do the simplify
self.struct_limit = 30
init_code += self.generate_input(type_name, var_name)
self.code = self.code.replace('//@HeliumInputEnd', init_code, 1)
def do_instrument(self, annotate, instrument_type=''):
output_code = ''
if not instrument_type:
instrument_type = config.get('instrument_type')
instrument_node = annotation.get_comment_node_by_annotation(self.doc, annotate)
if instrument_node:
if instrument_type == 'variable':
alive_vars = io.resolve_alive_vars(instrument_node, set())
for var_name in alive_vars:
type_name = alive_vars[var_name]
self.struct_limit = 30
output_code += self.generate_output(type_name, var_name)
elif instrument_type == 'count':
output_code = 'printf("%d ", 5678);\n'
# so that it can print out something(such as precondition) even if the program crash
output_code += 'printf("\\n");'
self.code = self.code.replace('//'+annotate, output_code, 1)
def instrument_segment(doc, segment):
first_node = segment[0]
last_node = segment[-1]
parent = first_node.parentNode
pre = doc.createElement('comment')
pre.appendChild(doc.createTextNode('\n//@Pre\n'))
parent.insertBefore(pre, first_node)
post = doc.createElement('comment')
post.appendChild(doc.createTextNode('\n//@Post\n'))
parent.insertBefore(post, last_node.nextSibling)
segment.insert(0, pre)
segment.append(post)
# loop invariant
if config.get('instrument_loop') == 'true':
for node in segment:
if domutil.is_element(node) and node.tagName == 'for' or node.tagName == 'while':
block_node = domutil.get_first_child_by_tagname(node, 'block')
if not block_node: continue
inner = doc.createElement('comment')
inner.appendChild(doc.createTextNode('\n//@Inner\n'))
block_node.insertBefore(inner, block_node.lastChild)
return segment
def resolve(node, exception = set()):
"""Resolve the identifiers used in node
"""
# to_resolve = extract_to_resolve(node, exception)
to_resolve = extract_to_resolve_from_code(domutil.get_text_content_except(node, 'comment'), exception)
result = {}
resolved = set()
unresolved = set()
while to_resolve:
new_resolve = {}
local_result, local_resolved = local.resolve(to_resolve)
result.update(local_result)
sys_resolved = sys.resolve(to_resolve - local_resolved)
unresolved.update(to_resolve - local_resolved - sys_resolved)
resolved.update(local_resolved | sys_resolved)
new_code = ''
for text,_,_ in local_result.values():
new_code += text
to_resolve = extract_to_resolve_from_code(new_code, exception | resolved | unresolved)
if unresolved:
if config.get('show_unresolved') == 'true':
print('unresolved: '+repr(list(unresolved)))
return result
def build(self, doc, context, functions):
"""
:param doc: document of the context
:param context: context nodes
:param functions: functions because of the linear search up above function.
functions = {call_name: node, ...}
"""
self.doc = doc
self.context = context
self.functions = functions
self.inputs = io.resolve_undefined_vars(context, set())
self.resolved = {}
self.unresolved = set()
self.generate()
self.resolve()
self.generate_support()
Instrumenter().instrument()
compile_success = self.compile()
if config.get("show_compile_result") == "true":
if compile_success:
green("compile success")
else:
red("compile error")
return compile_success
def __init__(self, folder):
self.builder = Builder()
self.tester = Tester()
self.folder = folder
self.analyzer = config.get('analyzer')
self.start()
def run(self):
fout = open(self.output_folder + "/output.csv", "w")
fin = open(self.output_folder + "/input.csv", "w")
var_list = [var for _, var in self.output_spec]
fout.write(",".join(var_list))
output_number = len(var_list)
fout.write("\n")
input_var_list = [var for _, var in self.input_spec]
fin.write(",".join(input_var_list))
fin.write("\n")
logger.info("Running " + str(len(self.tests)) + " tests")
run_success = False
for test in self.tests:
if config.get("show_test_data") == "true":
logger.debug("input:" + repr(test))
fin.write(",".join(test.split()))
fin.write("\n")
p = Popen(self.output_folder + "/a.out", shell=True, stdin=PIPE, stdout=PIPE, stderr=PIPE)
# print(test)
if not test.strip():
continue
timeout = int(config.get("timeout"))
try:
stdout, _ = p.communicate(input=test.encode("utf8"), timeout=timeout / 1000)
stdout = stdout.decode("utf8")
except TimeoutExpired:
logger.warning("testing timeout")
p.kill()
stdout = ""
except UnicodeDecodeError:
logger.warning("UnicodeDecodeError")
stdout = ""
# stderr = stderr.decode('utf8')
# -6: 134: assertion fail
# -11: 139: segment fault
if p.returncode == 0:
run_success = True
if config.get("show_returncode") == "true":
print("return code: " + str(p.returncode))
if p.returncode != 0:
# stdout = ''
pass
if config.get("show_test_data") == "true":
logger.debug("output:" + stdout)
output = stdout.split()
if not output:
fout.write(",".join(["NA" for _ in range(output_number)]))
fout.write("\n")
else:
if config.get("instrument_type") == "count":
# fout.write(str(len(output)))
fout.write(str(stdout.count("5678")))
if config.get("instrument_loop_parameter") == "true":
output = stdout.replace("5678", "").split()
for _ in range(output_number - len(output) - 1):
output.append("NA")
fout.write(",")
fout.write(",".join(output))
fout.write("\n")
else:
for _ in range(output_number - len(output)):
output.append("NA")
fout.write(",".join(output))
fout.write("\n")
logger.info("done running testing")
if run_success and config.get("show_run_success") == "true":
green("run success")
fout.close()
fin.close()