Interview with Brett Cannon, Python project
Python - высокоуровневый язык программирования общего назначения, ориентированный на повышение производительности разработчика и читаемости кода.
Разработчик Python согласился ответить на мои вопросы о разработке и тестировании Python.
Please present yourself.
Brett Cannon, core developer since 2002.
How many developers involved in the development on regular basis?
According to Open HUB we had as high as 36 contributors in a single month in the past year (https://www.openhub.net/p/python/contributors/summary).
How development cycle of Python looks like?
The Python devguide covers this in detail: https://docs.python.org/devguide/.
I have found continuous integration tool based on BuildBot. What else tools, tests and testing frameworks do you use?
We use unittest primarily for our tests and our own test runner.
Do developers create tests on regular basis for new functionality and bugfixes?
Yes; it is expected all bugfixes and new features have tests.
What process do you use for testing?
We don’t have an overarching one other than tests should be written to make sure semantics don’t regress.
What kind of testing do you use (functional, compatibility, stability, unit testing etc)?
The buildbots handle compatibility. Otherwise it’s mostly unit and functional testing.
Python has performance test suite How often do you run it?
Whenever someone chooses to run it (no continuous run of it).
Do you compare results only with previous versions?
Rarely. It’s typically only used for verifying that an experimental performance improvement actually improves performance.
What are the criterias for good performance results?
Typically at least a 5% increase in performance across the board for anything that may increase code complexity.
Do you use negative testing techniques (like fuzz testing, fault injection etc)?
Victor Stinner has in the past (e.g. this is why the faulthandler module exists in the stdlib).
Do you measure code coverage?
On occasion, yes. It is a common starter project for new contributors.
How much of code covered by auto tests?
I don’t have the number handy for a recent coverage run.
What tool do you use for measuring code coverage?
Do you involve volunteers to testing? I mean things like ‘crowd sourcing’. (For example Perl community uses volunteers resources to test CPAN modules. See slides from OSCON 2010 (PDF) and http://www.cpantesters.org/)
PyPI is run independently from python-dev so that doesn’t apply here. We do include the test suite with Python so users can run it and report any issues they have.
It seems you keep all bugs in bug tracker on https://bugs.python.org/. Who is responsible for tracking of open bugs?
Everyone and no one. =) As a group we try to track bugs, but no one individual owns it.
Do you have triage process?
Yes, and it’s covered in the devguide.
Do you make regular security audit of code?
We have a Coverity project set up which regularly scans the code (maintained by Christian Heimes). They did a press release about our very low defect rate: https://www.coverity.com/press-releases/coverity-finds-python-sets-new-level-of-quality-for-open-source-software/.
(Interviewer’s note: see also Coverity Scan report (PDF))
Who is responsible for releasing of new version? What are the release criterias?
We have a release manager. Release criteria is technically no release blockers, but otherwise it is up to the RM to make that call. There is always a PEP written to outline what the targeted release dates are.
Thanks for your answers.