Remove DEBUG output from functional tests, fix test requirements, add VNC server

[zenmonkeykstop]

Status

Ready for review

Description of Changes

Fixes #3698 , #3718

Changes proposed in this pull request:
This PR includes @msheiny's change to selenium's output level from DEBUG to WARNING, cherrypicked from commit a1f0134

It also includes an update to the test requirements file, to remove a duplicate entry for werkzeug that probably slipped in in a merge and was preventing the tests from running.

It also includes support for viewing functional tests in progress, using VNC on OSX and Debian workstations.

Testing

To verify that functional tests run and do not include DEBUG-level output, from the securedrop directory, run:

make build-debs
cd securedrop 
./bin/dev-shell ./bin/run-test -v tests/functional/

Docker build should complete without error, and tests should run without debug-level output (specifically not the base64-encoded firefox profile data)

To verify that VNC is set up and available, kick off the functional tests as above, and once the Docker container has started and the tests are in progress, run:

make func-vnc

The system's default VNC viewer should start, displaying the functional tests' browsers.

Checklist

If you made non-trivial code changes:

• I have written a test plan and validated it for this PR

[redshiftzero]

Now that torproject.org is back up, the next step is to update the version of Tor Browser on tbb-0.9.0 (let's add to this branch @zenmonkeykstop so we can get CI fixed).

[redshiftzero]

Pushed two commits to get the docker builds working again, note I also updated the pax flags for Tor Browser 8 based on our private infrastructure repo

However we're hitting a new issue here which looks to be mozilla/geckodriver#913 likely due to the old version of geckodriver we are using. Indeed problems with Tor Browser 8 were reported over in webfp/tor-browser-selenium#105. Going to bump up versions of geckodriver (and Selenium if necessary) and see if that resolves...

[redshiftzero]

ok my attempts to resolve this CI issue so far are in commit 8da014e in tb-8-ci (not pushing here in case we take another path). I'm at the point where I've updated geckodriver to 0.21.0, firefox-esr to latest (>=60 is required by geckodriver 0.21.0), but I'm hitting the following on the first functional test:

self = <httplib.HTTPResponse instance at 0x7fbdc8448cf8>

    def _read_status(self):
        # Initialize with Simple-Response defaults
        line = self.fp.readline(_MAXLINE + 1)
        if len(line) > _MAXLINE:
            raise LineTooLong("header line")
        if self.debuglevel > 0:
            print "reply:", repr(line)
        if not line:
            # Presumably, the server closed the connection before
            # sending a valid response.
>           raise BadStatusLine(line)
E           BadStatusLine: ''

/usr/lib/python2.7/httplib.py:408: BadStatusLine

[kushaldas]

We should have the same version of firefox-esr which is being used by the Tor Browser.

[redshiftzero]

yep, in 8da014e I am using Firefox ESR 60 which is what Tor Browser 8 is based on

[zenmonkeykstop]

I tried updating to 8.0 last night as well; the teardown() function didn't seem to run between tests, so docker container ended up with a few dozen sets of firefox-esr,xvfb, and firefox.real processes, before running out of memory and giving up the ghost.

[zenmonkeykstop]

As per gitter, upgrading to geckodriver 0.22.0 seems to resolve this issue. The BadStatusLine messages are happening with 0.21.0 because it sets connection keep-alives to 5sec - this has been bumped to 90sec in 0.22.0.

[zenmonkeykstop]

@msheiny stuck in a commit to skip pw prompt for remote-viewer. Works for me on Ubuntu but give it a lash on Debian?

[msheiny]

🎉 not only are you a great canadian.. but you have fixed this PR from certain doom.

[msheiny]

errr wait .. failing tests.. take it thats a new problem?

[msheiny]

im also getting the same failures locally that are happening in CI

such as this gem
SessionNotCreatedException: Message: Unable to find a matching set of capabilities

i see a few github tickets that possibly reference a pathing issue .. checking

[redshiftzero]

There is some wonky CI logic in the staging-test-with-rebase job:

1. In the tbb-0.9.0 branch, we removed the application test run here.
2. Meanwhile, in the develop branch, we removed the testinfra test run (in this commit).

During the rebase on latest develop, we didn't reconcile these two changes. This is why we see in CI output:

    TASK [Dump raw test output] ****************************************************
    fatal: [app-staging]: FAILED! => {"msg": "'raw_test_results' is undefined"}
    	to retry, use: --limit @/root/sd/molecule/aws/side_effect.retry

[eloquence]

I'm assuming this fully resolves #3718 as well? If so, please update PR description accordingly; thanks! :)

[codecov-io]

Codecov Report

❗ No coverage uploaded for pull request base (tbb-0.9.0@4dd2687). Click here to learn what that means.
The diff coverage is n/a.

@@             Coverage Diff              @@
##             tbb-0.9.0    #3817   +/-   ##
============================================
  Coverage             ?   84.52%           
============================================
  Files                ?       43           
  Lines                ?     2720           
  Branches             ?      295           
============================================
  Hits                 ?     2299           
  Misses               ?      355           
  Partials             ?       66

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 4dd2687...d87d58d. Read the comment docs.

[redshiftzero]

all tests passing here now... so merging