Am Samstag, 9. Juli 2016, 00:33:15 CEST schrieb Diederik de Haas: > On vrijdag 8 juli 2016 18:01:31 CEST Hillel Lubman wrote: > > The way I see it, improvement can be achieved by providing some relation > > between packages. I.e. making sure they enter testing all at once, and if > > one is stuck, other related ones wouldn't enter. This would ensure that > > frameworks all enter at once, and users of testing would be able to > > continue rolling the rest of their system, even if frameworks are stalled. > > > > Do you think it's a feasible solution, or it doens't fit into Debian > > methodology? Or you think frameworks shouldn't be seen as one related set > > of packages? > > _I_ think it's useful feedback (fwiw). > The way I currently see/understand them is this: > It was unforeseen that the mixture of 5.22 and 5.23 of framework packages > would cause issues and that is actually not 100% certain, but it very much > looks like the solution as getting those in (version) sync seemed to have > solved the issues every time. > > Debian has a 'feature' for coordinated transitions, namely https:// > release.debian.org/transitions/ but as can be read in the linked > documentation (https://wiki.debian.org/Teams/ReleaseTeam/Transitions) the > frameworks packages actually shouldn't qualify for that feature as there is > no ABI/API change (normally) between framework packages versions. > I may not fully understand this feature myself though.
Diederik, I think this is about "I want an always stable and releaseable testing" again. And, yes, while I think some improvements in the order packages enter testing, would certainly be nice, I also think that the main purpose of testing is just that: Preparing Debian for release by testing the heck out of it. That it is kind of a rolling distro, is in my eyes just a side effect of it. And I think for a rolling distro unstable is more suitable anyway. So that is where I will focus my energy: Testing out the new packages while fully understanding that there can be transitional states of breakage that do not even matter for the final stable distribution as it won´t have KF 5.22 packages. If you want to put energy into changing the order in which packages enter testing, fine, go ahead, I am sure many users will appreciate it: I just suggest to think clearly then what steps actually help you to move towards this goal. And which steps do not and as such as mainly a waste of energy. I think a discussion here is such a waste of energy, cause, see, we had this discussion before. I took hours to write mails explaining things, while in the end its the Debian developers and the admins of the build machines are the ones who effectively can change things around testing. But they didn´t take part in this discussion. Thus I stand by it: Any discussion here *without* involving the members of the Qt/KDE team and the people who put in a lot effort to maintain the build mechanisms for Debian is not helping to improve things *at all*. Unless probably you work on coming up with a concept that could work for them, but even then, as its them doing the actual work, you better come up with something, they agree with. I am accepting Sid/Testing for what it (currently) is and am willing to deal with temporary breakage. Cause I think for stabilising KDE´s Plasma and Applications this is more important than to try to fix temporary instabilities as side-effects of the development and package building process. That written I am training to let this discussion be to itself now. If need be I will do a summer break from debian-kde mailing list. Cause I found myself to be quite touchy on this, easily offended and maybe overly engaged. Cause in the end: *My* Plasma/Qt desktop is stable since weeks, I do not have any major issues with it at all, and it is important to me to focus on where I can really help while also clearly deciding how much energy I want to spend on this work. Honestly to all testing users my current recommendation at this exact point in time would be this: Switch to unstable, dist-upgrade and then switch back to testing again – or remain with unstable, but I think its also important that some users continue test testing. Of course, if you want to wait till Plasma 5.7 in unstable is complete, then wait. Of course, how you decide to spend your energy, by all means, is your decision. So in case you want to focus on minimising temporary instabilities in testing, by all means go for it. Good luck with that. Thanks, -- Martin
