On Fri, 2026-05-22 at 13:28 +0200, Nicolas Chauvet via rpmfusion- developers wrote: > Le ven. 22 mai 2026 à 12:37, Simone Caronni <[email protected]> a > écrit : > > > > On Fri, May 22, 2026 at 11:08 AM Nicolas Chauvet > > <[email protected]> wrote: > > > > > > That can be the way forward (should work with koji/mash), > > > especially > > > when the aarch64 arch will be added. > > > > > > That said noarch means the package will be the same with all > > > architectures, so you aren't supposed to emit any arched > > > dependency > > > from noarch package. > > > It currently works in x86 context because x86_64 can satisfy the > > > (x86-32) %{_isa} dependency, but it won't work when aarch64 will > > > be > > > introduced. > > > > > > > > > > The only issue I know related to arch package migration is that > > > sub-packages for noarch packages cannot be arched, > > > while arched packages can have noarch sub-packages. > > > > > > Staying on noarch should allow us to have the single package > The current situation is broken, as I said, a noarch package cannot > have arched dependencies. > Having a noarch package doesn't fix any issue. > Please fix. >
I ended up not having much time to properly investigate this, but the move to EL10 being x86_64-only seems to have brought a change where the 64-bit glibc package now also provides glibc32 support. In other words, it is in the 64-bit buildroot that now builds the glibc32 support needed for 32-bit compatibility. Because of that, Steam might perhaps be able to follow a similar approach by becoming an x86_64 package while still requiring some 32- bit packages. It is also worth mentioning that VirtualBox, which previously required glibc(x86-32), also had to switch in EL10 to somehow using glibc32 instead, and no longer requiring glibc(x86-32). Disclaimer: as mentioned, I only looked briefly into the subject, so these are just my initial impressions and they may very well be completely wrong. -- Sérgio M. B. _______________________________________________ rpmfusion-developers mailing list -- [email protected] To unsubscribe send an email to [email protected]
