On Fri, 20 Aug 2021 at 07:26:51 +0200, Helmut Grohne wrote: > The options you present to work around this sound plausible to me, but > they do complicate the transition plan somewhat. In essence, we'll get > two transition points. One for non-buildds and another for them.
Maybe that, but what I was suggesting most recently was one transition point for everything except reproducible-builds.org (and maybe other QA infra), and one for reproducible-builds.org (and maybe other QA infra). If we have reproducible-builds.org reporting merged-/usr vs. split-/usr variation bugs like #949270, and we have fixed all known instances of that variation, then it becomes a lot more viable to let buildd chroots migrate to merged-/usr alongside end-user systems, because at that point it no longer matters whether buildd chroots are merged-/usr or not (and reproducible-builds.org would tell us about regressions that make it matter again). smcv