Thanks Preston Timmons for the report.

Only compute the CSRF_COOKIE when it is actually used. This is a
significant speedup for clients not using cookies.

Changed result of the “test_token_node_no_csrf_cookie” test:  It gets
a valid CSRF token now which seems like the correct behavior.

Changed auth_tests.test_views.LoginTest.test_login_csrf_rotate to
use get_token() to trigger CSRF cookie inclusion instead of changing
request.META["CSRF_COOKIE_USED"] directly.

When "Save as new" is chosen and errors occur, only show the "Save as new"
button and not the other save buttons.

Thanks to Tino de Bruijn for doing the real work on this fix.

This avoids calling date.tzinfo.utcoffset(date) twice. It's also
robust to cases where that function returns None -- which never
happens in practice :-)

Thanks Aymeric Augustin for the review.

Updates for languages: Indonesian, Belarusian, Persian, and Dutch.
Forward port of cb370f85 from stable/1.8.x

This is a regression caused by introducing rendered migration states in
1aa3e09c and the _meta refactoring in fb48eb05.

Thanks to Danilo Bargen for reporting the issue and Marten Kenbeek and
Tim Graham for triaging the bug and providing the initial test case.

The test failed on RHEL6 systems running python 2.7 from a RedHat
Software Collection (SCL) because this test runs an external python
script with a stripped system environment. RedHat SCLs work by setting
a number of system environment variables when these are stripped out by
this test the python 2.7 interpreter is no longer able to function
properly because it can not find the system libraries needed.

Now we use use mock to modify the system environment directly.

Thanks to Andrew Ingram for the report.

Thanks to Reto Aebersold for reporting the issue and Tim Graham and
Claude Paroz for the review.

Changed squashmigrations to not instantiate a MigrationExecutor,
but to directly use a MigrationLoader instance instead.

I changed "appears" to "appear" and emphasized the word "not" to match
the rest of the document.