def log_contain(self, keywords):
M4.exec(['sync', './vm.log'])
with open('./vm.log', 'r') as fd:
for line in fd:
for k in keywords:
if re.search(k, line.strip()):
# print(k, line)
return True
return False
def main(options):
'''
There some implementation optimizations based on the algorithm presented in the paper:
Algotirhm:
while not terminate():
s, w <- prio_pick(Q)
fill papameter to w and execute w
if length(w) > MAX:
drop s, w
s', w' <- generate_new_from(s)
First, we set a pre-defined max value as
the total number of workloads (i.e. testcases),
and we group workloads into test packages.
For each package, we create a new disk to manipulate.
Second, the prio_pick procedure favours the longest workload,
the queue Q is not necessary.
Third, since checking each workload w incurs high overhead,
we checking workloads in a batch.
I.e., for workload w is <call1, call2> and its following workload w'
is <call1, call2, call3>,
there is not need to check them separately;
when we execute w', w is also been executed.
Fourth, kmeans incurs overhead,
we use kmeans generation and random generation alternately.
'''
fs = options.fs
for cnt in range(config.get('NR_TEST_PACKAGE')):
package = TestPackage()
if options.test:
result = M4.exec(['./ctrl', 'disk', fs, f'{fs}-{cnt}'])
M4.print_result(result)
for _ in range(config.get('NR_TESTCASE_PER_PACKAGE')):
test_case = generate_test_case(package, options)
# test_case = generate_concat_test_case(package, options)
if options.test:
M4.process_exec(
['./ctrl', 'run-case', fs, f'{fs}-{cnt}', test_case.path_],
timeout=20)
# M4.print_result(result)
M4.process_exec(['./ctrl', 'kill'])
if options.test:
result = M4.exec(['rm', '-rf', f'{fs}-{cnt}'])