[connection] mark Connection as "not a request" The ORM use this information. Having a wrong value will lead to crash.

[server/session] ensure appobject obtained from the session are linked to the session

[connection] have rql_rewriter accession on Connection Code expecting a session search the RQLRewriter on this attribute.

[connection] move local_perm_cache management on Connection It belongs here anyway.

[session] use a rich object for timestamp This allows the Transaction to keep a reference to it

[migraction] do not access session.data directly use set_shared_data and get_shared_data instead

[connection] move relation management method

[connection] give access to shared data method A reference to session data are kept

[connection] move relation cache method to Connection needed for execute (and need execute themself...)

[connection] give access to session execute from connection This allows multiple function that both requires an execute method and are required by the execute method.

[connection] move last part of undo logic in connection Now that Connection have a reference to the repo it can handle all of it itself

[connection] initialize connection.user and connection.lang A RequestSessionBase need a user and lang. For now we use the session ones. Later the Connection will have it's own.

[Connection] inherit from RequestSessionBase This contains a lot method. We need them to move more code from Session to connection.

[multi-sources] drop multi-sources related test. another branch is removing the multi-source itself. We do not want to bother fixing those test.

Use new repoapi for the web stack The publisher now link repoapi.ClientConnection to request. and explicitly control there scope. Web side, appobject._cw.cnx is now a repoapi.ClientConnection. This actually kill webonly possibility until the repoapi is able to use some RPC. The change in the authentication stack is very hasty and need cleanup

PARTIAL: Using the repoapi in test Test now use the repoapi.ClientConnection to access the repo. This is very big change. The current method to access the repo are kept for backward compatibility. The new methods will be introduced much later. The TestCase keep a ClientConnection To an admin session for the whole test. This changeset does not makes all tests pass without the next one that change set htt without the next one that change the http stack

[web/test] drop proff of concept Facebook login in test The authentication stack will change. The change in API will requires rework of user cube anyway.

[test/dbapi] do not rely on the Testcase provided cnx We plan to use the repoapi in the test. So we need the DBAPI test to explicitly create a DBAPI connection.

[dbapi] explicitly use the DBAPI version of CubicwebRequestBase As we plan to use a ConnectionCubicwebRequestBase by default we better have the dbapi explicitly asking for a DBAPI implementation.

[web/application] drop unused import unused since 3c85e734ce00

[web/test] properly reset the request connection related attribute A new request does not have those attribute to None. There have special value that we now properly reinstall.

[web/test] drop two undocumented suspicious regression test Those two test are not documented and use the API in the wrong way. They do not raise exception by chance. the current request signature is:: def request(self, rollbackfirst=False, url=None, headers={}, **kwargs): So the RQL have nothing to do here.

[webrequest] introduce an alternative implementation using the repoapi DBAPI specific parts of ``CubicWebRequestBase`` are extracted in a ``DBAPICubicWebRequestBase`` decicated class and a new ``ConnectionCubicWebRequestBase`` is introduced that provide the same services but using a repoapi.ClientConnection as data source. For now the ``DBAPICubicWebRequestBase`` is still used by default.

[repoapi] add an anonymous_cnx function This function return an anonymous ClientConnection. It aims to replace ``dbapi.anonymous_session``.