On Thu, Apr 13, 2023 at 2:31 PM Do, Eling <l...@analogic.com.invalid> wrote:
> Hi Jiri, thanks for responding to my question...much appreciated. That > would be great if changing the order of things during the install would > allow the library install to complete even if the tests cannot run. Done now, it is failing merrily on setup.py now. > Any idea as to when a build/snapshot with this change might be available? > I won't promise anything. The setup.py processes the spec xml and generates the .pcl file, which means it actually runs more of the client code than what I originally expected. I already have all of it fixed on my machine, passing all the tests, but I am still reviewing it and trying to make as little of a mess as possible. Specifically, I am confused regarding where bytes (b"") or strings (u"") should be accepted, for example as message bodies, various message fields, and so on; I _can_ of course make the library convert inputs to bytes, but the question is if it is desirable. My current feeling is that I should not do this and instead have the tests pass b"" where necessary...) That makes the changes Python 2-neutral, because in python2, adding b in front of "string" does nothing. I want to carry on with this, hopefully ending up running an Apache release vote when the work is done ;P, but it is hardly my top priority. I don't want to diminish the significance of what I am trying to do here, but, to tell the truth, if I were a Qpid user wanting a broker on RHEL9, I would probably run qpid-cpp in a RHEL8-based docker, having compiled it against Python 2). Or I'd try to migrate to some other, more actively developed, broker. Are there some particular features of qpidd that you are especially attached to, that are hard to get from the alternatives? -- Mit freundlichen Grüßen / Kind regards Jiri Daněk
