Lennart Poettering
Lennart Poettering
your current patch can't model the x32 abi properly. a 64bit x86 CPU after all can support three abis, not two, on the same kernel, and the SECONDARY_ABI doesn't model...
and the vpick stuff should not really take abis into consideration like this. there's no problem with running x32 and x86-64 nspawn containers and portable services on an x86-64 kernel...
I am sorry, but this is really wrong. `/usr/` trees on Debian specifically, *of* *course* support multiple ABIs! it's why `/usr/lib/` exists on Debian. A kernel for arch X often...
hmm CI is unhappy?
hmm, why the revert in the first commit though? it allows "legacy" mdns queries from the local host, which should be fine because: 1. it's the only the thing #32990...
anyway, can you elaborate why precisely you want the revert that is the first commit? I don#t get it?
> It's possible to run two (or more) RFC-compliant mDNS responders on the same machine. Neither systemd-resolved nor avahi is RFC-compliant though. Direct unicast queries don't matter much in practice...
lgtm, just some nitpicking about a log message. looks good to merge once the relevant commit entered linus tree and it looks as if it might stay there.
also needs a test
So this *must* be async, we cannot wait for DM stuff from PID1, no way. See comments above. Also, I am pretty sure this should not be restricted to notify-reload.