looks fine, thanks

Note though, that some 32bit architectures newly won't own the lib64, e.g. ppc, sparc, s390..

Hum, so I'm not sure if that's right then or not. Is that line supposed to be there for multilib arches only?

Sorry, but I somehow do not understand the transition from scenario "for selected platforms, we own %{_lib}, which stands for lib or lib64" to scenario "own lib64 if that is available". Shouldn't we always own lib and in addition own lib64 on 64 bit platforms? Or was the previous code wrong? Could you please elaborate?

@jorton wrote:

Hum, so I'm not sure if that's right then or not. Is that line supposed to be there for multilib arches only?

I actually think that this is a good change. Nothing on earth should ever install stuff into /lib64 on 32bit arch. Since the scl meta-packages are arch specific - it is completely OK if 32bit meta packages don't own corresponding lib64 (even though they previously owned that).

@vondruch wrote:

Shouldn't we always own lib and in addition own lib64 on 64 bit platforms?

Yes. That's what this change is about, sorry I made it confusing by previous statement - scratch that (I thought that dropping lib64 ownership on those 32bit arches could cause issues).

Or was the previous code wrong? Could you please elaborate?

Not really wrong for the current set of architectures supported by Fedora, because we added aarch64 some time ago.. But at some point we might want to add a new architecture; and the %_lib == lib64 condition should work forever.

forever

s/forever/till there are no 128bits/ :-)