Closes #44: Call correct python executable with subprocess.

[ghost]

No description provided.

[jgonggrijp]

The build currently fails because of the inner workings of cram, not because of your changes. I was able to fix this by modifying the test (see 9188784). However, this revealed that the use of python -m breaks backward compatibility with Python 2.6 (see Travis build 69). Python 2.6 could only do -m with single modules, not with packages containing a __main__ module.

At this point, I believe it is best to drop Python 2.6 support, as python -m is clearly the way forward. See the tip of the feature/subprocess-call branch, which builds upon your pull request. Since this is a breaking change, I want to include these changes with the 1.0 release. I will make an intermediate deprecation release to warn users that Python 2.6 support will be removed.

You pull request will remain open in the meanwhile, but rest assured that it is accepted and that it will be merged when the 1.0 release is near.

[jgonggrijp]

At first you opened this pull request with two commits, the first of which missed a comma and the second of which fixed that mistake. I merged your pull request indirectly, basing a new feature branch on those two commits and finally merging that into develop, today. See 6e44350, mentioned above.

In the meanwhile, you squashed your two commits into one and changed history for your patch-1 branch. This escaped my attention. The consequence is that GitHub currently doesn't understand that your PR got merged. This again escaped my attention, until now, while I have already based several commits and a new release branch on the old history of your patch-1 branch.

I will now mark this PR as closed. You can probably still convince GitHub to mark it as merged, by changing history again so that your patch-1 branch points to 0fba3da, which I based my feature branch on.