assertRedirects should assume http://testserver if it isn't provided

[Russell Keith-Magee]

[6164] fixed #987, a problem with redirects allowing relative URIs. The fix is correct, but it makes the testing API slightly unwieldy; all calls to assertRedirects() now require that '​http://testserver' be part of the asserted URL.

The vast majority of redirection tests will be internal to the site, so requiring the full dummy host prefix only serves to muddy the tests. This is made worse by the fact that urlresolvers.reverse() doesn't include the testserver component, so assertRedirects(reverse('some_name')) doesn't work.

assertRedirects should be modified to inspect the test URL; if no ':' doesn't occur in the test url, '​http://testserver' should be prepended for test purposes. i.e.,

assertRedirects('/some/url') is equivalent to assertRedirects('​http://testserver/some/url')

assertRedirects('​http://otherserver/some/url') would be untouched

[Chris Beaven]

Uh, I don't think I changed the requirement on assertRedirect tests, since the test server doesn't set HTTP_HOST or SERVER_NAME

see ​http://code.djangoproject.com/changeset/6164#file4

[Russell Keith-Magee]

Ah - Looks like I mispointed the finger. It wasn't [6164] - it was [6166], fixing #4986. The test were updated in [6169]. I have a whole lot of tests that are similarly affected.

[pat.m.boyd@…]

patch

[patrickb]

I attached a solution to only match parts of the url that were defined in the expected_url. It fixes our tests but I'm not sure if its good/clear/readable/up to django standards.

[Malcolm Tredinnick]

Faced with having to modify a couple of dozen tests in [6662], I fixed this in [6661]. Handles both relative and absolute URLs, as well as making absolute URLs from a passed in host parameter in case that's easier sometimes (one case in the tests uses django.testserver as the host, for example).