There seem to be two more occurences

sdks/python/apache_beam/io/tfrecordio_test.py
65:# ...   print data

sdks/python/apache_beam/utils/annotations.py
44:    print arg1, '*', arg2, '=',
55:    print arg1, '*', arg2, '(the old way)=',
63:    print arg1, '*', arg2, '(the experimental way)=',
69:  print multiply(5, 6)
70:  print old_multiply(5,6)
71:  print exp_multiply(5,6)

Thanks @superbobry, those instances were in comments, not in live code but I have converted them for completeness.

@charlesccychen @holdenk Your reviews please?

@pabloem Your review please?

@holdenk This is merged now but I am wondering how the testing did not catch this and how we can prevent similar regression in the future. https://github.com/apache/beam/blob/master/sdks/python/scripts/run_mini_py3lint.sh#L51 should be able to spot these as syntax errors but perhaps it is only look at a portion of the codebase instead of looking at all Python files in the repo. On Travis, it takes 20 sec to do this flake8 test on the entire repo.

utils module is included py3 lint checks.

The first three files contained print statements in live code. These files are in .test-infra/jenkins which flake8 does not currently test.