Tests: retry request on ConnectionError.

[mauritsvanrees]

On Jenkins we often get one ConnectionError in a seemingly random test.
Happens on both Python 2 and 3.
Retrying after a short pause helps.
Fixes #648

Note: I added a print statement so we know if it happens, and maybe detect a pattern in the urls where this goes wrong. I tried logger.warning and logger.error, but this never showed up in the test output.

In a few local parallel runs, I have seen errors on these urls:

http://localhost:64140/plone/@users
http://localhost:64117/plone/@types/Document
http://localhost:49516/plone/@users/noam

[mauritsvanrees]

Meh, bin/code-analysis on Travis does not like the print statement. Fair enough. I will remove it.

[mauritsvanrees]

I have amended the commit. This was the print statement if anyone wants to try that:

print('\n### Got ConnectionError at %s, retrying once.' % url)

To get the url out from amidst the many deprecation and other warnings:

bin/test -s plone.restapi -m test_documentation 2>&1 | grep ConnectionError

[Rotonen]

Is there something which still needs to happen after the ZServer ProcessStarting event fires before plone.rest or plone.restapi endpoints are actually available? Or is that somehow different for Python 3 / WSGI?

The root cause should probably get figured out eventually.

[coveralls]

Coverage increased (+0.2%) to 96.225% when pulling e6b601c on issue-648-retry-on-connectionerror into 8f4b1e3 on master.

[mauritsvanrees]

Agreed, the root cause would be good to know.
But for now, the Jenkins jobs are green, except for one with a spurious error in plone.i18n in Japanese. So should be safe to merge.

[gforcada]

LGTM, maybe rather than a print statement, we can add one deprecation warning, it is not perfect, but does its job, I.e. get printed on Jenkins log output 😃

[gforcada]

And we, hopefully, will finish the year with Jenkins green 🎉