Improvements

September 9, 2013 - 3 minute read -
lint pep8 performance testing gsoc

Commits

Others couldn't come here because of Github API :)

Things done in past few days:

  • Integration tests for mark_tracking
  • Sphinx Docs: A lot of modules had errors. The most common problem is ImportError because sphinx is unable to find the modules in PYTHONPATH. Thus, these modules need to be mocked. I have a directory fake-lib which has all these mocked modules. This fake-lib has to be added to the PYTHONPATH in the conf.py file. The docs are at http://gst-switch.readthedocs.org/en/latest/
  • Imgur Frame Upload: For testing in travis, I am uploading the key frames that are generated to imgur. For example in this build the output key frames are: img1(before PIP change) and img2(after PIP change). So there are two main problems with the build process -
    1. An unknown dependency
    2. Gstreamer's python binding dont produce a videotestsrc output. Maybe a problem with the typelibs.
  • Modified the controller tests: The controller tests were too heavy causing them to be unreliable at times. Now they are perfectly reliable.
  • Docs: Added docs for all modules and classes.
  • Linting: All code is pylint clean. The corrosponding pylintrc files are present. These files contain the excluded warnings and errors. Can be done by make lint
  • pep8: All code is pep8 clean. Can be done by make pep8
  • pre-commit hooks: Added pre-commit hooks which will run lint on C code and pylint and pep8 on python code. It will not commit unless these tests pass.
  • Build failures: If any of the tests (lint, pep8, unittests) fail, the build fails and the repository is flagged red.
  • Performance Tests: These are extreme torture tests. I run it on my system alone and not on travis. All of these tests are not meant to pass, some failures are expected. Currently the tests include things test cases where I call the get_compose_port method N number of times, where i vary N in [100, 200, 300, 400, 500]. The test mostly fails at the 500 mark. One test case tests changing the PIP mode after some delay. I vary this delay like [1, 0.6, 0.5, 0.2] in seconds. On my system, below 0.6, i.e for test cases 0.5 and 0.2 the test fails.

Comments Section

Feel free to comment on the post but keep it clean and on topic.

blog comments powered by Disqus