Github user gemmellr commented on the issue:
https://github.com/apache/activemq-artemis/pull/2418
> So openwire always returns a value on asking groupseq. It never gives an
ability to determine if it was set by client or not.
It does seem to return a value if you ask for one, which it shouldn't, but
it also does not populate the property name all the time, so its not clear why
the value would be asked for. This is showing up in the AMQP message because it
is in the property names being iterated, which it shouldn't be.
> We have to be able to set it somehow during conversion. This is being set
on message conversion. The issues here are on protocol conversion. Were not
changing amqp message
From what I see, the method is not used during conversion to an AMQP
message, with that being set directly on the underlying Properties section
within the converter. Setting it any other time is then stepping toward a
protocol violation.
I am not asking for just a sticking plaster. I did suggest that we fix
small "one line" bugs separately from comparatively massive general code
refactorings (did not say they shouldnt be done). I suggest that we avoid
actively violating the protocol as this likely will likely result (and as
mentioned, the setter was noted as being deliberately left out of the AMQP bits
when that method was added). I do not think we should emit group-sequence
values that were never set on the message.
I'm done giving feedback on Artemis issues for a while.
---
