travis-miniconda installer fix

[rasbt]

Okay, let me fix the travis-ci from pullrequest #15
From now on, it will be a bit easier to debug it before it's merged since we will be able to see the travis reports.

• this also adds the tracking of the "build" status to the Readme file that you were asking about

I really don't understand this bug regarding the miniconda installation.
In the previous version, I selected the "latest" Miniconda Linux built from http://repo.continuum.io/miniconda/ (Miniconda-latest-Linux-x86_64.sh), which apparently doesn't work for some reason. However, selecting, e.g., Miniconda-3.9.1-Linux-x86_64.sh does work.
Anyway, let's see

[rasbt]

Ha, there we go! I really love green checkmarks ;)

[rasbt]

Btw. are you really planning to support Python 2.6? Saw it listed in the trove classifier section. Just wanted to ask, because I find Python 2.6 really annoying (six module and stuff)

[rhiever]

I'm not opposed to dropping 2.6 support. Almost tempted to drop 2.7 support.

Is this ready to merge?

[rasbt]

I'd have nothing against that as a 99% Python 3.x user, but I would keep 2.7 for now maybe. Yap, it's ready for merge. If you like, you could add more of the older Py 3.x version to the .travis.yaml file (in the top directory), currently it's only

- PYTHON_VERSION="2.7" LATEST="true" "DEAP_VERSION=1.0.1"
- PYTHON_VERSION="3.4" LATEST="true" "DEAP_VERSION=1.0.1"
- PYTHON_VERSION="3.5" LATEST="true" "DEAP_VERSION=1.0.1

thought it suffices.

[rasbt]

We could also add coverage results via coveralls later. Was a bit hesitant with this for now since I have bad experience with coveralls for "expensive" code (often aborts if evaluations take to long)

[rhiever]

Perfect. Thanks again @rasbt!

So far, the tests don't take too long because we don't have unit tests on the fit() function. Do you have a sense of how long is "too long"?

[rasbt]

Do you have a sense of how long is "too long"?

Sry, not really, it's more like coveralls is pretty flaky, sometimes the same stuff works, sometimes it doesn't (I think it has something to do with their hardware utilization on their side). But your tests seem pretty gentle so far, so let's try (I will submit another pull request for that later)

[rhiever]

Hooray!