On Tue, Jun 5, 2018 at 9:21 PM Josh Boyer <jwboyer@xxxxxxxxxxxxxxxxx> wrote: > > On Tue, Jun 5, 2018 at 2:13 AM Zbigniew Jędrzejewski-Szmek > <zbyszek@xxxxxxxxx> wrote: > > > > On Tue, Jun 05, 2018 at 02:26:25AM +0100, Peter Robinson wrote: > > > On Tue, Jun 5, 2018 at 12:39 AM, Adam Williamson > > > <adamwill@xxxxxxxxxxxxxxxxx> wrote: > > > > Hi, folks! > > > > > > > > We currently have a Final release criterion that reads as follows: > > > > > > > > "A spin-kickstarts package which contains the exact kickstart files > > > > used to build the release must be present in the release repository. > > > > The included kickstarts must define the correct set of release > > > > repositories. > > > > > > > > Why? > > > > > > > > This is considered part of Fedora's duty to be 'self-hosting': the > > > > kickstarts used to produce the release images are a vital piece of > > > > information required to duplicate that release, so they must be > > > > preserved along with the release." > > > > > > > > Lately this requirement has been fairly annoying in practice. Updating > > > > the package prior to release does not appear to be in anyone's regular > > > > schedule, so invariably what happens is shortly before the release > > > > deadline I realize we haven't built a 'release' spin-kickstarts package > > > > and have to file a blocker bug and ping people with the necessary > > > > permissions (of which there are only a few) to build one in a tearing > > > > hurry. Then we have to approve the blocker bug and push the updated > > > > package through the freeze, all wasting time we could be spending on > > > > more important fixes. > > > > > > > > The benefit here is really fairly tiny, as well. It's arguable whether > > > > anyone cares particularly whether a Fedora release, as a frozen > > > > artifact, is 100% internally reproducible (and I'm not sure whether our > > > > releases actually *are* reproducible in any case, these days, I'm not > > > > at all sure we ship all the necessary metadata and so on for *every > > > > single deliverable* within the distribution). > > > > > > > > These days I'd suggest it should be quite acceptable to simply use git > > > > tags for this purpose. It should be quite easy for rel-eng to adjust > > > > the release scripts to create a tag in the fedora-kickstarts repo (and > > > > why not fedora-comps too, while we're at it) for each 'candidate' > > > > compose, named for the compose ID. That would make it very easy to > > > > access the correct kickstarts for any Fedora candidate compose just by > > > > a 'git checkout', with no need for the cumbersome work of getting the > > > > package into the compose. > > > > > > > > Naturally this would go along with updates to any relevant docs or wiki > > > > pages, recommending to use the git repository instead of the RPM > > > > packages, and explaining the tagging scheme. As for the package, we > > > > could either keep it but not sweat about updating it for each release, > > > > retire it entirely, or change it to contain only a text file pointing > > > > to the git repository (or to the doc / wiki page that explains the git > > > > repo location and tagging strategy). > > > > > > > > Thoughts? Thanks! > > > > > > It makes perfect sense, the package is not actually shipped as part of > > > any of the actual deliverable artifacts and they're widely available > > > in a public git repository for people to consume so it's not reducing > > > the ability to reproduce Fedora, we don't rush around and ensure all > > > the tools that might need last minute patches in the compose process > > > are all tagged stable in the release either so I don't see actually > > > shipping this package as stable makes any difference in reality, we > > > don't even use the package in the compose process, we pull the > > > kickstarts directly from git. > > > > +1 too. > > Yes, let's do what makes sense. I like the proposal. > I'm not in favor of this change. Actually, what concerns me is that there's an underlying problem that caused Adam to propose this: there are only a few people with the necessary permissions to be able to prepare the release, and that this is a manual process that is forgotten often. If we were to say: "hey guys, we're just gonna use these autogen'd git tag thingies from now on", what stops us from just autogenerating a push to generate the required package and have it built in Koji to be part of the distribution release? And perhaps this is something that doesn't get talked about much anymore, but I know that when I'm presenting Fedora to people, they *like* that it's only a couple of package installs away to get everything in place to make spins or remixes. And being able to offer in the distribution *everything* needed to build it is powerful. We've been bad about this in recent releases, sure. But why can't we look at making that process better, rather than throwing out the baby with the bathwater? -- 真実はいつも一つ！/ Always, there's only one truth! _______________________________________________ devel mailing list -- devel@xxxxxxxxxxxxxxxxxxxxxxx To unsubscribe send an email to devel-leave@xxxxxxxxxxxxxxxxxxxxxxx Fedora Code of Conduct: https://getfedora.org/code-of-conduct.html List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines List Archives: https://lists.fedoraproject.org/archives/list/devel@xxxxxxxxxxxxxxxxxxxxxxx/message/LFG5I75ALLR6D5UYC77ORUIVSH4OSSHE/
