On Sat, Aug 1, 2020 at 7:17 AM Kevin Kofler <kevin.kofler(a)chello.at> wrote:
Neal Gompa wrote:
> I think it does have value, however I think the Red Hat compiler team
> drastically underestimated how much breakage we're willing to tolerate
> for it.
I think you mean "overestimated" there, not "underestimated",
don't you?
Yeah, I meant overestimated here... That's what I get for replying
right after waking up. :)
> That's not true. Since Koji 1.18, it's been possible to
modify the
> build process by setting simple RPM macros and mock flags in build
> tags. And with the module builds (which operate in chain builds on
> side tags), there is a higher potential for modifications that can
> result in a different set of binaries since it'll generate macros
> packages on demand to do complex build environment changes.
But the annobin side tag would have the exact same RPM macros and mock flags
set as regular Rawhide. (Ideally, none, because Rawhide should be the
default target of the specfiles.) Modules would of course need their own
annobin side tags (one per module build tag) if you want to cover them too.
Then there'd be the problem of where we'd have the build capacity. We
barely have enough for what we do now...
--
真実はいつも一つ!/ Always, there's only one truth!