mirror of
git://git.yoctoproject.org/poky.git
synced 2025-07-19 12:59:02 +02:00
resulttool/manualexecution: Refactor and simplify codebase
Simplify and removed unnecessary codes. Refactor to allow pythonic loop. (From OE-Core rev: 84c6a992e1114685194f6e8a554bce7753c090cc) Signed-off-by: Yeoh Ee Peng <ee.peng.yeoh@intel.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
This commit is contained in:
parent
f1e470d8fa
commit
996d1f4518
|
@ -24,24 +24,12 @@ def load_json_file(file):
|
|||
with open(file, "r") as f:
|
||||
return json.load(f)
|
||||
|
||||
|
||||
class ManualTestRunner(object):
|
||||
def __init__(self):
|
||||
self.jdata = ''
|
||||
self.test_module = ''
|
||||
self.test_cases_id = ''
|
||||
self.configuration = ''
|
||||
self.starttime = ''
|
||||
self.result_id = ''
|
||||
self.write_dir = ''
|
||||
|
||||
def _get_testcases(self, file):
|
||||
self.jdata = load_json_file(file)
|
||||
self.test_cases_id = []
|
||||
self.test_module = self.jdata[0]['test']['@alias'].split('.', 2)[0]
|
||||
for i in self.jdata:
|
||||
self.test_cases_id.append(i['test']['@alias'])
|
||||
|
||||
|
||||
def _get_input(self, config):
|
||||
while True:
|
||||
output = input('{} = '.format(config))
|
||||
|
@ -67,45 +55,42 @@ class ManualTestRunner(object):
|
|||
extra_config = set(store_map['manual']) - set(self.configuration)
|
||||
for config in sorted(extra_config):
|
||||
print('---------------------------------------------')
|
||||
print('This is configuration #%s. Please provide configuration value(use "None" if not applicable).'
|
||||
% config)
|
||||
print('This is configuration #%s. Please provide configuration value(use "None" if not applicable).' % config)
|
||||
print('---------------------------------------------')
|
||||
value_conf = self._get_input('Configuration Value')
|
||||
print('---------------------------------------------\n')
|
||||
self.configuration[config] = value_conf
|
||||
|
||||
def _create_result_id(self):
|
||||
self.result_id = 'manual_' + self.test_module + '_' + self.starttime
|
||||
self.result_id = 'manual_%s_%s' % (self.test_module, self.starttime)
|
||||
|
||||
def _execute_test_steps(self, test_id):
|
||||
def _execute_test_steps(self, test):
|
||||
test_result = {}
|
||||
total_steps = len(self.jdata[test_id]['test']['execution'].keys())
|
||||
print('------------------------------------------------------------------------')
|
||||
print('Executing test case:' + '' '' + self.test_cases_id[test_id])
|
||||
print('Executing test case: %s' % test['test']['@alias'])
|
||||
print('------------------------------------------------------------------------')
|
||||
print('You have total ' + str(total_steps) + ' test steps to be executed.')
|
||||
print('You have total %s test steps to be executed.' % len(test['test']['execution']))
|
||||
print('------------------------------------------------------------------------\n')
|
||||
for step, _ in sorted(self.jdata[test_id]['test']['execution'].items(), key=lambda x: int(x[0])):
|
||||
print('Step %s: ' % step + self.jdata[test_id]['test']['execution']['%s' % step]['action'])
|
||||
expected_output = self.jdata[test_id]['test']['execution']['%s' % step]['expected_results']
|
||||
for step, _ in sorted(test['test']['execution'].items(), key=lambda x: int(x[0])):
|
||||
print('Step %s: %s' % (step, test['test']['execution'][step]['action']))
|
||||
expected_output = test['test']['execution'][step]['expected_results']
|
||||
if expected_output:
|
||||
print('Expected output: ' + expected_output)
|
||||
print('Expected output: %s' % expected_output)
|
||||
while True:
|
||||
done = input('\nPlease provide test results: (P)assed/(F)ailed/(B)locked/(S)kipped? \n')
|
||||
done = done.lower()
|
||||
done = input('\nPlease provide test results: (P)assed/(F)ailed/(B)locked/(S)kipped? \n').lower()
|
||||
result_types = {'p':'PASSED',
|
||||
'f':'FAILED',
|
||||
'b':'BLOCKED',
|
||||
's':'SKIPPED'}
|
||||
'f':'FAILED',
|
||||
'b':'BLOCKED',
|
||||
's':'SKIPPED'}
|
||||
if done in result_types:
|
||||
for r in result_types:
|
||||
if done == r:
|
||||
res = result_types[r]
|
||||
if res == 'FAILED':
|
||||
log_input = input('\nPlease enter the error and the description of the log: (Ex:log:211 Error Bitbake)\n')
|
||||
test_result.update({self.test_cases_id[test_id]: {'status': '%s' % res, 'log': '%s' % log_input}})
|
||||
test_result.update({test['test']['@alias']: {'status': '%s' % res, 'log': '%s' % log_input}})
|
||||
else:
|
||||
test_result.update({self.test_cases_id[test_id]: {'status': '%s' % res}})
|
||||
test_result.update({test['test']['@alias']: {'status': '%s' % res}})
|
||||
break
|
||||
print('Invalid input!')
|
||||
return test_result
|
||||
|
@ -120,9 +105,9 @@ class ManualTestRunner(object):
|
|||
self._create_result_id()
|
||||
self._create_write_dir()
|
||||
test_results = {}
|
||||
print('\nTotal number of test cases in this test suite: ' + '%s\n' % len(self.jdata))
|
||||
for i in range(0, len(self.jdata)):
|
||||
test_result = self._execute_test_steps(i)
|
||||
print('\nTotal number of test cases in this test suite: %s\n' % len(self.jdata))
|
||||
for t in self.jdata:
|
||||
test_result = self._execute_test_steps(t)
|
||||
test_results.update(test_result)
|
||||
return self.configuration, self.result_id, self.write_dir, test_results
|
||||
|
||||
|
@ -130,8 +115,7 @@ def manualexecution(args, logger):
|
|||
testrunner = ManualTestRunner()
|
||||
get_configuration, get_result_id, get_write_dir, get_test_results = testrunner.run_test(args.file)
|
||||
resultjsonhelper = OETestResultJSONHelper()
|
||||
resultjsonhelper.dump_testresult_file(get_write_dir, get_configuration, get_result_id,
|
||||
get_test_results)
|
||||
resultjsonhelper.dump_testresult_file(get_write_dir, get_configuration, get_result_id, get_test_results)
|
||||
return 0
|
||||
|
||||
def register_commands(subparsers):
|
||||
|
|
Loading…
Reference in New Issue
Block a user