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)

i18n: updated translation for French

Currently translated at 98.5% (1119 of 1135 strings)

tests: fix Git on Windows sometimes failing on ' or ` in file:/// URLs

Some Git versions will fail when cloning a repository like:

git clone file:///path/to/strange-file-'`foobar myclone

Since '` (and some other characters) are filtered anyway (in
lib.utils.repo_name_slug) I think this test doesn't need to test those
characters.

tests: fix test_nodes.py and test_files.py which fail on windows due to mimetype differences

The mimetype of files may differ on different platforms (e.g. the mimetype of
.py files on windows is not text/x-python but text/plain). Therefore, compare
the mimetype to the value the platform uses.

compare: close Dulwich internals for Git repositories to avoid leaking open files

Fix failing test_compare.py on Windows by explicitly closing internal
object_store objects.

api: avoid sending double Content-Length header (Issue #201)

WSGI will take of Content-Length - handling it explicitly was misguided.

docs: improve mod_wsgi documentation (Issue #203)

Based on feedback from Graham Dumpleton.

* Recommend setting global Apache config WSGIRestrictEmbedded On
* Don't set the default value for processes=1 - it has side effects
* Use python-home instead of python-path
* Fix confusing mentioning of root and clarify the use of specifying user and group
* Include WSGIProcessGroup in configuration excerpt

docs: add notes about IIS, Windows Authentication and Mercurial

(The original patch from Konstantin has been heavily copyedited and modified by
Mads Kiilerich but is still [based on] Konstantin's feedback and contribution.)

status: refactor internal dict for mapping status string to description

Slightly better performance ... but more importantly: do it "the right way".

docs: document Kallithea customization options

Create basic documentation for the various customization options in
Kallithea. Move the rcextensions section to this new page and reword.

settings: rescope 'Analytics HTML' option to generic HTML/JS/CSS customization

Update the setting name 'Analytics HTML block' to better cover the fact that
it can be used to make any HTML, JavaScript or CSS tweak.

vcs: fix processing of git commands that return output on stderr

The subprocessio module used for interfacing with Git is using threading. It
might not be fully deterministic.

The subprocessio module also had the "feature" that it stopped reading and
reported failure once it found output on stderr - even if the command completed
(or would complete) with success.

Sometimes (like https://drone.io/bitbucket.org/domruf/kallithea/24 ) tests on
linux could fail on totally normal informational output like:

Couldn't run git command (['git', '-c', 'core.quotepath=false', 'checkout', 'foobranch']).
Original error was: Subprocess exited due to an error:
Switched to branch 'foobranch'

On Windows it would fail even more often.

To fix that, ignore stderr while processing output. There is a risk that it in
some cases can make the process block on stderr and thus never finish ... but
there is no reports of that yet.

i18n: add translation for Spanish

Created new translation.

db: cache SCM instance short-term (tied to SQLAlchemy session lifetime)

Repeatedly checking whether SCM instances are invalidated is slow, and
we don't actually _want_ SCM instances to invalidate half-way through
a request either.

Therefore cache them in on the db.Repository object, the lifetime
of which is directly tied to the lifetime of the SQLAlchemy session,
the lifetime of which is tied directly to the individual HTTP request.
This way, we only check for invalidation the first time the SCM instance
is accessed in a request.

This will improve performance in cases where we have (by definition) badly
written code that retrieves repo objects several times.

db: get rid of vcs_full_cache - it should always be used

It might make tests less deterministic, but it makes sure that we test what we actually use.

tests: invalidate cache for new repositories - make tests more stable if garbage has been left behind in the db

It could be argued that this should be done elsewhere in the app ...

db: name the scm_instance_cached cache entries - reduce the risk of collisions

- and same with other cache regions

db: inline repo __get_instance to scm_instance_no_cache

Having two names for the same thing just made it look even more complicated
than it is.

There _should_ not be any reason for calling scm_instance_no_cache directly
anywhere ... but for now that is how it is.

helpers: cache user_or_none results in beaker long_term cache

Avoids repeated expensive 'LIKE' lookup on email addresses extracted from
changeset authors.

This information is only used for display of changelog information.

The cache will never expire - only when the server process expires. That is ok
for this very static information and the way it is used.

scm: simplify get_repos and get rid of RepoList overloads

The RepoList variants might have made sense before any repo info was stored in
the database. Now it has no value and doesn't help at all.

pullrequests: optimize iteration over reviewers - avoid fetching users one by one

.reviewers was mainly used for iteration and then dereferencing .user one by
one. That gave lots of queries and round trips to the database and was slow.

Instead, do something else. Either query directly or use a new method for
getting the list of reviewer users.

Reviewers will explicitly be shown in the order they have been added (assuming
database id's are monotonic).

setup.py: remove unused package_data

Per the setuptools documentation, package_data is ignored if
include_package_data is True, and indeed, we instead use MANIFEST.in.

Added tag 0.3.1 for changeset 9bf8eb837e78
Added signature for changeset 9bf8eb837e78

tests: don't use the TESTS_TMP_PATH string

it differs each time you run the test
and therefore doesn't work if you use a separat kallithea test instance
(KALLITHEA_NO_TMP_PATH=1)

instead get the 'real' path from the DB

FIXME: breaks kallithea/tests/functional/test_admin_repos.py

paster: remove unused BasePasterCommand.notify_msg

This method is not referenced anywhere in the Kallithea or Paste code.
Additionally, calling it with any true-ish `log` argument would trigger
a TypeError exception.

auth: support both old and new python-pam API

python-pam 1.8.* provides a new API replacing the previously
available authenticate() function. The latest unreleased
version brings it back, but until it's available, use a
custom shim instead.

vcs: the git backend requires string encoded repository names too

Shown by test.functional.test_admin_repos.test_delete_non_ascii on Windows.

tests: binary streams/files need to be opened in binary mode

Without binary, the stream may be closed on EOF character
http://stackoverflow.com/a/9069009/7029 .

windows: node paths should always use posixpath - avoid \ in repo internal paths on windows

Using \ on Windows would result in a wrong node path which could not be checked
out with Git on Windows.

localization: set language for formencode based on config file

Before, the language would be taken from the environment - such as LANG. Now it
will use exactly what has been configured as lang in .ini .

The formencode initialization could be done in many places and there is no
obvious place ... but this one seems ok.