#Fedora 35 is almost out... so boring!
So I just formatted a 40GB partition on my laptop to try Rawhide (future Fedora 36) while keeping the stable root filesystem just in case:
mkfs.xfs -f -L fedora36 /dev/nvme0n1p4
mount LABEL=fedora36 /rawhide
I labeled it fedora36 to avoid confusion in 6 months, when I wipe my old fedora35 partition to make room for Fedora 37. I've already been through this a number of times 🙂
Next, I copy my current system to the new partition:
rsync --stats -HAXphax / /rawhide/
sent 13.90G bytes received 7.84M bytes 103.40M bytes/sec
total size is 15.24G speedup is 1.10
Only 103MB/s on an NVME drive? 🙄
Probably rsync is not the right tool for copying filesystems locally, but it's convenient because it preserves extended attributes, hardlinks, and even sparse files. I don't think it preserves reflinks, but I doubt Fedora has any of them in the root filesystem...
And on modern PCs, there's already the stupid EFI boot partition that has to be shared among multiple operating systems. Since Fedora puts its EFI bootloader and configs in the unversioned "fedora" directory, ironically, it is easier to make it coexist with Windows and Ubuntu than with another version of Fedora! 🤦♂️
Recently, /boot/efi/EFI/fedora/grub.cfg has become a stub that loads /boot/grub2/grub.cfg:
search --no-floppy --fs-uuid --set=dev UUID_OF_YOUR_ROOT
...but this indirection doesn't make sharing any easier: they're still hard-coding a particular Fedora instance in a shared path that can't be configured (I believe "EFI/fedora" is hard-coded in grubx64.efi).
And finally, I launch the big transaction:
dnf --disablerepo=updates --disablerepo=fedora --enablerepo=rawhide --enablerepo=rpmfusion-free-rawhide --enablerepo=rpmfusion-nonfree-rawhide update
Install 2 Packages
Upgrade 1129 Packages
Remove 2 Packages
Skip 2 Packages
Total size: 2.0 G
Is this ok [y/N]: y
So these are the changes expected in #Fedora 36:
I already see the OpenSSL 3.0 package, but several packages still depend on the OpenSSL 1.1 libs.
I'm not planning to use DNS-over-TLS because I expect that frequent reconnects will slow down DNS queries and cause reliability issues. What we really need is DNS-over-HTTP/3 (r DNS-over-QUIC, but nobody supports it).
@codewiz Because nobody was maintaining it and there was some big API changes :/
On the upside the new API is nice: https://invent.kde.org/plasma/kdeplasma-addons/-/merge_requests/91 (WIP)
The social network of the future: No ads, no corporate surveillance, ethical design, and decentralization! Own your data with Mastodon!