Endpoint handling broken on SQL backend by portable-identifiers changes

[Ewan Mellor]

The changes for the portable-identifiers blueprint have broken the endpoint calls endpoint_get_by_tenant_get_page, endpoint_get_by_tenant_get_page_markers, endpoint_add, endpoint_get, and endpoint_get_by_tenant.

All of these calls need to translate from uid to id or the reverse, and they are attempting to do so, but the condition under which they do the translation is wrong. All other calls are using hasattr(api.TENANT, 'uid_to_id') or similar, but the ones in endpoint_template.py are using isinstance(api.TENANT, models.Tenant). This latter test fails, the IDs aren't translated, and so the database inserts / lookups are all using a uid where they should be using an integer ID and vice versa.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/2645

[Ewan Mellor (ewanmellor)]

https://review.openstack.org/#change,2645

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/2645
Committed: http://github.com/openstack/keystone/commit/3f2b53586ca804115d8cf91e43f63ec7a300e82c
Submitter: Jenkins
Branch: master

commit 3f2b53586ca804115d8cf91e43f63ec7a300e82c
Author: Ewan Mellor <email address hidden>
Date: Tue Dec 27 22:45:36 2011 -0800

    Bug #909255: Endpoint handling broken on SQL backend by portable-identifiers changes

    When deciding whether to translate UIDs to integer IDs or vice versa, use
    hasattr(api.TENANT, 'uid_to_id') or hasattr(api.TENANT, 'id_to_uid') in place
    of isinstance(api.TENANT, models.Tenant). This matches the code elsewhere
    (e.g. user.py) and has the effect that the IDs will be translated correctly.

    Change-Id: Ia036bf310d48f8f8aa5739cb3e09d1f7364ad1a4