On 19/03/07, Gordon Sim <[EMAIL PROTECTED]> wrote:
Martin Ritchie wrote:
> As to the problems you saw with the python scripts. This is down to
> the fact that the python is not compliant. The java broker stopped
> allowing AMQPLAIN as an authentication method but the python is
> obviously hardwired to only use that hence the Java broker closed the
> connection when it tried to use an 'invalid' authentication mechanism.
I came to the same diagnosis (only slower!). In fact the python client
doesn't hardcode for AMQPLAIN but the test harness does.
> I have reinstated the AMQPLAIN mechanism. It shouldn't have been
> disabled in the first place.
Actually I think its not a bad idea to disable it at some point as it
was in essence a mistake in the spec that has since been corrected.
I have a simple change for moving the python test harness over to PLAIN
but I still get errors. These are related to the access control and
occur with the c++ also. AccessManagerImpl.isAuthorized() seems to go
into a recursive loop; if I hack it to just grant access irrespective of
username or object then the tests pass. Is there some configuration
required to setup access control?
It goes in to a recursive loop because the configuration file isn't
right a change was made pre commit that should have been a post commit
improvement :(
The M2 Branch is now correct and I shall set up svnmerge on trunk to
merge it over just now.
--
Martin Ritchie
