continuous integration

[GoogleCodeExporter]

It would be rather nice to use a continuous integration server such as https://travis-ci.org/ to automatically build the library and run the unit tests.

Similarly, using a service such as https://readthedocs.org/ the docs would automatically be rebuilt whenever we get a new commit.

• travis-cl.org only seems to work with github at moment; maybe someone knows if it can be used with Google Code (please comment on this Issue!) or if we just want to periodically clone to github
• readthedocs should be ridiculously straightforward

Original issue reported on code.google.com by orbeckst on 24 Apr 2013 at 3:55

• Blocked on: Move MDAnalysis to GitHub due to Google Code closure #223

[hainm]

is there any progress on building mdanalysis by conda yet? If not, I highly recommend to use it. I can just install pytraj from binstar within 30 seconds.

Hai

[tylerjereddy]

I recently made conda builds on mac and linux and it seemed to work well from binstar. I've mentioned this on the mailing list before, but we still don't have a way to easily make a windows conda build. If there's enough interest we could probably make the mac / linux conda builds more official, but my impression was that we wanted to refine other aspects of the development process first.

[orbeckst]

@richardjgowers : This is looking really good. I was surprised that Travis-CL lets you run tests for more than 1h (if I read the "run time" correctly).

Please merge #278 whenever you think it is ready. Big step forward!

[richardjgowers]

The limit is 50 mins... and I'm hitting 45 with the python 2.6 tests :D I think (luckily) 2.6 and 2.7 are run in parallel, so 1h+ is the cumulative time.

I think going forward we'll have to look at shrinking some of the tests, (iirc some of the PDB file tests take a minute?). Once we're tracking coverage then we can safely redo the tests and know we're not losing coverage.

[orbeckst]

On 19 May, 2015, at 11:57, Richard Gowers wrote:

The limit is 50 mins... and I'm hitting 45 with the python 2.6 tests :D I think (luckily) 2.6 and 2.7 are run in parallel, so 1h+ is the cumulative time.

I see.

I think going forward we'll have to look at shrinking some of the tests, (iirc some of the PDB file tests take a minute?). Once we're tracking coverage then we can safely redo the tests and know we're not losing coverage.

... or/and get faster PDB and GRO readers. #212 is already high priority

Oliver Beckstein * orbeckst@gmx.net
skype: orbeckst * orbeckst@gmail.com