Enable cover tox env on Travis

[theacodes]

Replace fragile app engine sdk download script with gcprepotools script -
this ensures the SDK is up-to-date and is a no-op if the SDK is installed
and up-to-date.

Resolves #617

[theacodes]

(do not merge until I confirm the test god is behaving as expected)

[dhermes]

Is readthedocs.org build system using fetch_gae_sdk.py?

[theacodes]

Is readthedocs.org build system using fetch_gae_sdk.py?

@dhermes god I hope not. Let me check.

[theacodes]

@dhermes doesn't appear so. We mock all the app engine imports for the docs build.

[theacodes]

(As an aside, this should also fix our GCE system tests on jenkins because tox is obtuse about environment variables)

[theacodes]

Travis is only slow when you need it to be fast.

[pferate]

What do you think about changing the coveralls command to:

coverage report --show-missing --fail-under=100 && coveralls

and dropping {[coverbase]commands} from coveralls?

[theacodes]

@pferate good point, there's no need for the coveralls tox env to re-run the tests.

[dhermes]

there's no need for the coveralls tox env to re-run the tests

You could just drop the coveralls tox env entirely?

[theacodes]

@dhermes it works! the cover environment is failing because of missing branches in sqlalchemy.

[theacodes]

You could just drop the coveralls tox env entirely?

That's what I did.

[theacodes]

Okay, so I'm assuming we don't want to merge this because it makes travis fail. Does anyone want to tackle the missing branches in sqlalchemy so I can rebase and merge this?

[pferate]

I could take a stab at it.

[theacodes]

@pferate that'd be swell. Otherwise I'll try to do it this afternoon.

[dhermes]

It looks like oauth2client.client has also grown a branch miss via #588:

oauth2client/client.py                                696      0    202      1    99%   1945->1947

[pferate]

Got the SQLAlchemy tests taken care of, then I found that the new PKCE code has a partial branch as well.

[pferate]

Haha @dhermes barely beat me at posting it here.

[pferate]

@jonparrott, #629 has been submitted for you!

[theacodes]

Rebased. Waiting on travis. @dhermes any concerns with merging this?

[dhermes]

Nope no concerns

[pferate]

Hmm... Once again I'm seeing strange failures in my Travis fork, but it's not failing on the main repository. Is anybody else seeing anything similar?

For the pypy tox environment:
./scripts/install.sh: line 31: /home/travis/.pyenv/bin/pyenv: No such file or directory

It looks like ${HOME}/.pyenv exists, since it passes if [ ! -d "${HOME}/.pyenv" ]; then, but it fails on the next line, as if the cloned repo isn't there.

I cleared my Travis cache and restarted the test, but it failed again.... 😦

Any ideas?

[theacodes]

@pferate not sure.

It's still reporting branch misses after rebase. What even?

[pferate]

Hmm.. tox -e cover reported no missed branches for me and you didn't change anything in oauth2client or tests.

What lines are being reported with missed branches?

[pferate]

It looks like an empty .pyenv directory is getting created somehow. I added some tracers to a test branch and got this:

ls -la "${HOME}"

total 80
drwxr-xr-x 18 travis travis  4096 Aug 12 22:14 .
drwxr-xr-x  3 root   root      33 Feb  5  2015 ..
--- OTHER DIRECTORIES OMITTED ---
drwxrwxr-x  2 travis travis     6 Aug 12 22:14 .pyenv
--- OTHER DIRECTORIES OMITTED ---

ls -la "${HOME}/.pyenv"

total 4
drwxrwxr-x  2 travis travis    6 Aug 12 22:14 .
drwxr-xr-x 18 travis travis 4096 Aug 12 22:14 ..

Maybe Travis is creating it for me, since it's called out as a cached directory?
$ git grep pyenv

.travis.yml:  - ${HOME}/.pyenv
scripts/install.sh:    if [ ! -d "${HOME}/.pyenv" ]; then
scripts/install.sh:        git clone https://github.com/yyuu/pyenv.git ${HOME}/.pyenv
scripts/install.sh:    ${HOME}/.pyenv/bin/pyenv install --skip-existing pypy-2.6.0
scripts/run.sh:# pyenv.
scripts/run.sh:    PATH="${HOME}/.pyenv/versions/pypy-2.6.0/bin:${PATH}"

I wonder if it's not failing in the main repo since it had pyenv cloned before #590 and #595, and is keeping the cached version. I'm thinking that if you delete the cache on the main repo, you will see the same failures that I am.

[theacodes]

Seems we figured out the pyenv issues separately. Merging this.