pullrequests: reword the grumpy "Rejected" to the more neutral and factual "Not approved"

"Rejecting" could be considered rude and some reviewers were thus reluctant to
use it. Rewording it might give reviewers a voice to better express their
feelings.

docs: tweak email documentation - add gmail example

smtp_auth is too obscure to have in the .ini template.

pull requests: refactor update_reviewers

This avoids a redundant database lookup for every removed reviewer,
and shows usernames (not just IDs) in associated log messages.

pull requests: prevent adding DEFAULT user as reviewer

Add a helper method to resolve reviewers, with an added check to prevent
the adding of the DEFAULT user as reviewer.

The __add_reviewers method, although internal, had a troubling interface
where the method was responsible for resolving reviewers, but the caller
was responsible for resolving mention_recipients. The method is changed
to always take two sets of database User objects, leaving the resolution
to the caller in both cases.

utils: improve extract_mentioned_users usefulness

Previously, extract_mentioned_users performed a regex search, converted
to a set to remove duplicates, converted that set back into a list, and
sorted the list on username, ignoring case. Every single caller of the
function then promptly took the carefully pruned and sorted result and
put it back into a set, rendering the entire exercise pointless.

In addition, every caller also resolved the usernames to database User
objects.

In this changeset, extract_mentioned_users is changed to return a set
of database User objects directly. A new extract_mentioned_usernames
function, which does only username extraction, is kept for the sole
purpose of providing a more testable interface (no database needed).

Bonus feature: The new extract_mentioned_users will prune non-existent
users, as well as the DEFAULT user. This means it is no longer possible
to @mention (and send notifications to) the DEFAULT user.

auth: prevent misuse of PermFunction in bool context

Evaluating a PermFunction as a boolean, rather than calling it, is
almost certainly an error. If not, "pf is not None" can be used.

notifications: untangle notification access check

This removes a broken permission check when viewing notifications (the
HasRepoPermissionAny object was created, but never actually called with
a repo_name argument as required). It would be non-trivial to actually
implement the check, as notifications don't track their repository
relationship explicitly, and even then, it's unclear why it would
make sense to allow a repository admin to see notifications to
other users.

It was never a vulnerability, due to a subsequent (and much stricter)
ownership check, which remains but has been untangled for readability.
In short, this changeset is a pure refactoring, except that specifying
a non-existent notification ID will now produce error 404, not 403.

setup.py: enable pytest-sugar plugin

pytest-sugar enables a nice visual representation of test progress and
results.
Minimum version is set to 0.7.0 as it fixes a problem where skipped tests
would not be shown in the test summary.
The upper version 1.0.0 is arbitrarily chosen and could be increased as seen
fit.

helpers: don't truncate words when when truncating texts for urlification

This would especially cause confusion when URLs were truncated.

Added tag 0.3.2 for changeset a84d40e9481f
Added signature for changeset a84d40e9481f

auth: further sanitize requests to prevent GET CSRF (CVE-2016-3691)

Routes allows GET requests to override the HTTP method, which breaks
the Kallithea CSRF protection (which only applies to POST requests).

This commit blocks such GET request, preventing CSRF attacks.

auth: properly invoke PermFunctions (CVE-2016-3114)

This fixes a vulnerability that allowed logged-in users to edit or
delete open pull requests associated with any repository to which
they had read access, plus a related vulnerability allowing logged-in
users to delete any comment from any repository, provided they could
determine the comment ID and had read access to just one repository.

changelog: fix range selection from changelog in git repos (Issue #190)

The Mercurial backend can handle short revision ids so the problem does not
occur with hg repositories. Git needs the full hash.

vcs: fix repo size calculation

tip.walk already 'dives into' the dirs - don't do that twice.

tests: finally block also needs to be executed if api_call itself fails

The cleanup code (finally block) should be executed as soon as something what
changes the db fails.

I don't remember which one but one of these tests failed and caused all the
others to fail as well because the database wasn't cleaned up after the first
failure. That made finding the root cause more difficult then it should have
been.

pytest migration: model: convert all tests to TestControllerPytest

The model tests were based on BaseTestCase which does not exist currently
for pytest-style tests. Nevertheless, there seems to be no advantage of
directly subclassing BaseTestCase over subclassing TestControllerPytest.
Thus, keep things simple and use TestControllerPytest.

pytest migration: add temporary assertRaises wrapper

Similar to the other temporary unittest-assert wrappers, add the missing
assertRaises and forward to pytest.raises.

pytest migration: simplify hg+git test class hierarchies

No need to use multiple inheritance when single inheritance can do it just
as well. Let the git/hg test cases just derive from _BaseTestCase which in
turn derives from TestControllerPytest.

Also remove empty setup_class/teardown_class methods.

pytest migration: convert functional tests with setup/teardown methods

pytest also supports setup/teardown methods like unittest (even though
pytest fixtures are more powerful and should be the end goal).
Only difference is the naming and signature of setUp (setup_method) and
tearDown (teardown_method).

test_admin_notifications: fix index test dependency

The index test only worked because another test had prepared some database
state it relied on.

More specifically, the index test creates a notification
as a newly created user, but that user had not been committed to the
database yet. When running the index test standalone, this causes the error:

IntegrityError: (IntegrityError) NOT NULL constraint failed:
notifications.created_by u'INSERT INTO notifications (subject, body,
created_by, created_on, type) VALUES (?, ?, ?, ?, ?)'
(u'test_notification_1', u'notification_1', None, '2016-03-11
21:15:19.066402', u'message')

By committing the database after the user is created, the problem is solved.

i18n: updated translation for Chinese (China)

Currently translated at 39.8% (452 of 1135 strings)

i18n: updated translation for German

Currently translated at 58.6% (666 of 1135 strings)

i18n: updated translation for French

Currently translated at 100.0% (1135 of 1135 strings)