When an unhandled Python error is raised in i9n tests, cleanup code becomes non-functional

[Timur Sufiev]

What's worse, the call which should take a screenshot raises another exception (since Selenium is no longer responding after that unhandled error has been raised) which completely overwrites the original exception and traceback. And without original traceback it's very difficult to understand what caused integration test to fail.

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/277755
Committed: https://git.openstack.org/cgit/openstack/horizon/commit/?id=51f35099bc0be32e26012ace18b1eb4502690349
Submitter: Jenkins
Branch: master

commit 51f35099bc0be32e26012ace18b1eb4502690349
Author: Timur Sufiev <email address hidden>
Date: Tue Feb 9 12:38:19 2016 +0300

    Don't overwrite original traceback in certain cases of i9n failures

    Achieve this by muffling exceptions (raised due to Selenium became not
    resposive) while taking integration test failure screenshot.

    Also extract the pattern of muffling and capturing exceptions common
    for dump_browser_log, dump_html_page and save_screenshot in a common
    @exceptions_captured context manager.

    Closes-Bug: #1543541
    Change-Id: I37fa18a302c553b43529df056d2beacff70f6189

[Doug Hellmann (doug-hellmann)]

This issue was fixed in the openstack/horizon 9.0.0.0b3 development milestone.