On Fri, 2020-06-19 at 22:17 +0100, Gordon Sim wrote: > On 19/06/2020 9:40 pm, Andrew Stitcher wrote: > > On Fri, 2020-06-19 at 13:59 +0100, Gordon Sim wrote: > > > ... > > > > > > Looks like it changed back for cyrus-sasl anyway: > > > https://github.com/apache/qpid-proton/commit/885d68aeaf522021a35b7b5cecb7c7c53663929b#diff-47e0b33a5461eff21e6cbbd017a969edL123 > > > > I don't exactly see where in that change any behaviour change > > happened > > - could you point it out (I know it was my change, but it was a > > long > > time ago!). > > Lines 123-125 in the diff: > > https://github.com/apache/qpid-proton/commit/885d68aeaf522021a35b7b5cecb7c7c53663929b#diff-47e0b33a5461eff21e6cbbd017a969edL123-L125 > > It removes the handling for case SASL_CB_USER.
Ah! I can't see any good reason why I reverted that earlier compatibility change. The default (non-cyrus) SASL implementation still doesn't send authzid at all, so it would make more sense for the cyrus impl to act the same way, certainly until we actually have some way/use for setting the authzid itself. Probably best to raise a new JIRA to note the regression from the earlier change - it's amazing it took so long to notice! I'd also note that raising a bug against RabbitMQ would be in order too as it shouldn't die if it receives an authzid equal to the authcid in any case. Andrew --------------------------------------------------------------------- To unsubscribe, e-mail: users-unsubscr...@qpid.apache.org For additional commands, e-mail: users-h...@qpid.apache.org
