SQLAlchemy >= 0.9.5 breaks tests

[Thomas Herve]

We're getting a lot of test failures since yesterday night ending up with "TypeError: 'NoneType' object does not support item assignment" during update code. I tracked it down to the new SQLAlchemy releases, trying to understand what's going on.

[Thomas Herve (therve)]

I found the issue, it has to do with MutableDict, opened https://bitbucket.org/zzzeek/sqlalchemy/issue/3093/mutabledictsetdefault-doesnt-return-the upstream.

[Liusheng (liusheng)]

this bug just need Updating the requirements from global requirements, like: https://review.openstack.org/#/c/96819/

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/102155

[Pavlo Shchelokovskyy (pshchelo)]

Liusheng, it will not help as SQLAlchemy 0.9.6 is already on PYPI, it still has this problem, but global requirements only blacklist 0.9.5

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/102155
Committed: https://git.openstack.org/cgit/openstack/heat/commit/?id=8e7294a7c3178222f94ce1cea45d77d70d5ac95e
Submitter: Jenkins
Branch: master

commit 8e7294a7c3178222f94ce1cea45d77d70d5ac95e
Author: Thomas Herve <email address hidden>
Date: Tue Jun 24 11:31:46 2014 +0200

    Don't rely on setdefault returning a value

    Workaround a bug in sqlalchemy>=0.9.5 where setdefault doesn't return
    the proper value.

    Change-Id: Ibc93024f7f41f4ced81525fac42566a18b1c641c
    Closes-Bug: #1333556

[Matt Riedemann (mriedem)]

sqlalchemy 0.9.5 was removed from the gate and 0.9.6 was released the same day to fix the issues.

However, this is still showing up in the check queue, so is there a problem with the heat patch?

http://logs.openstack.org/47/102347/1/check/gate-heat-python27/9941609/console.html

http://logstash.openstack.org/#eyJzZWFyY2giOiJtZXNzYWdlOlwiRkFJTDogaGVhdC50ZXN0cy50ZXN0X2F1dG9zY2FsaW5nX3VwZGF0ZV9wb2xpY3kuQXV0b1NjYWxpbmdHcm91cFRlc3QudGVzdF9hdXRvc2NhbGluZ19ncm91cF91cGRhdGVfbm9fcmVwbGFjZVwiIEFORCB0YWdzOmNvbnNvbGUgQU5EIHByb2plY3Q6XCJvcGVuc3RhY2svaGVhdFwiIEFORCBidWlsZF9xdWV1ZTpcImdhdGVcIiIsImZpZWxkcyI6W10sIm9mZnNldCI6MCwidGltZWZyYW1lIjoiMTcyODAwIiwiZ3JhcGhtb2RlIjoiY291bnQiLCJ0aW1lIjp7InVzZXJfaW50ZXJ2YWwiOjB9LCJzdGFtcCI6MTQwMzcwODIzNTI5NiwibW9kZSI6IiIsImFuYWx5emVfZmllbGQiOiIifQ==

[Matt Riedemann (mriedem)]

Nevermind comment 6, looks like maybe my change got caught up before the fix was merged to heat.

[Matt Riedemann (mriedem)]

I don't think this is a duplicate of bug 1333410, that was for a different kind of error introduced with sqla 0.9.5.