Fix controller_class_name for anonymous controllers.

[mschuerig]

TestCase#process generates an incorrect controller_class_name for anonymous classes. The name has a "_controller" suffix, which ought not to be there. This change simply removes that suffix.

[rafaelfranca]

cc @pixeltrix

[josevalim]

Can we have a test pls?

[mschuerig]

On Sunday 29 April 2012, you wrote:

Can we have a test pls?

Not right now, sorry, I don't have the time right now. The existing
tests still pass, of course. Please consider the change and decide if it
is too obvious.

Michael

Michael Schuerig
mailto:michael@schuerig.de
http://www.schuerig.de/michael/

[pixeltrix]

Whilst I agree that the controller suffix shouldn't be there, how is it breaking things for you? Changing it now may break some people's tests and I've been burned for that in the past

[mschuerig]

On Sunday 29 April 2012, you wrote:

Whilst I agree that the controller suffix shouldn't be there, how is
it breaking things for you? Changing it now may break some people's
tests and I've been burned for that in the past

Simply put, if people have tests that use anonymous controllers, they
must be working around the existing behavior in some way, if they are
aware of it or not. The problem with the "_controller" suffix is, that
route caching uses nested hashes with keys like

[:action, "index"] and [:controller, "anonymous"]

A controller_name of "anonymous_controller" never matches such a key and
results in a routing error.

Michael

Michael Schuerig
mailto:michael@schuerig.de
http://www.schuerig.de/michael/

[josevalim]

This needs to be fixed, even if people are relying on it, they are relying on wrong code.

[pixeltrix]

@josevalim backport?

[pixeltrix]

Fixed by 529a3ee - not backporting as it may break things