This only happened to come up because of a multilib glitch, but I'm wondering why rdma-core is on the default Workstation image. It seems unlikely that any of our users have Infiniband on their desktops.
----- Forwarded message from Jonathan Billings billings@negate.org -----
Date: Tue, 26 Jan 2021 14:58:14 -0500 From: Jonathan Billings billings@negate.org To: users@lists.fedoraproject.org Subject: Re: rdma-core-32.0-1.fc33.i686 has inferior architecture Archived-At: https://lists.fedoraproject.org/archives/list/users@lists.fedoraproject.org/message/5RYUPZZ23EKR22HKNQ7AHK4CP2QDPUDK/
On Tue, Jan 26, 2021 at 10:33:20AM -0500, Matthew Miller wrote:
On Mon, Jan 25, 2021 at 06:40:10PM -0800, ToddAndMargo via users wrote:
Hi Matthew,
I just did a # dnf remove rdma-core-32.0-1.fc33.x86_64 and now `dnf upgrade` is happy.
FWIW the root cause is here: https://bugzilla.redhat.com/show_bug.cgi?id=1919864 ... the package was accidentally built x86_64 only, which is why DNF couldn't find the matching i686 package. But this is only needed for special-purpose cluster networking hardware, and I'm pretty sure 0% of desktop users even need this package. I'm not sure what has pulled it in for so many people.
Looking at the /var/log/anaconda files, it appears that rdma-core was included in the LiveCD image I used to load the fedora workstation that has it installed. I used the default Fedora Workstation (Fedora-Workstation-Live-x86_64-33-1.2.iso). If I mount the rootfs inside the squashfs on the ISO, I can run:
# mkdir /mnt/livecd /mnt/squashfs /mnt/rootfs # mount /slow/images/Fedora-Workstation-Live-x86_64-33-1.2.iso /mnt/livecd # mount /mnt/livecd/LiveOS/squashfs.img /mnt/squashfs # mount /mnt/squashfs/LiveOS/rootfs.img /mnt/rootfs # chroot /mnt/rootfs/ rpm -q rdma-core rdma-core-31.0-1.fc33.x86_64
Might want to remove that from the build if it shouldn't be installed.
-- Jonathan Billings billings@negate.org _______________________________________________ users mailing list -- users@lists.fedoraproject.org To unsubscribe send an email to users-leave@lists.fedoraproject.org Fedora Code of Conduct: https://docs.fedoraproject.org/en-US/project/code-of-conduct/ List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines List Archives: https://lists.fedoraproject.org/archives/list/users@lists.fedoraproject.org
----- End forwarded message -----
On Tue, Jan 26, 2021 at 2:02 PM Matthew Miller mattdm@mattdm.org wrote:
This only happened to come up because of a multilib glitch, but I'm wondering why rdma-core is on the default Workstation image. It seems unlikely that any of our users have Infiniband on their desktops.
My guess: it's brought in by multipathd which is brought in by anaconda.
On Tue, Jan 26, 2021 at 2:06 PM Chris Murphy lists@colorremedies.com wrote:
On Tue, Jan 26, 2021 at 2:02 PM Matthew Miller mattdm@mattdm.org wrote:
This only happened to come up because of a multilib glitch, but I'm wondering why rdma-core is on the default Workstation image. It seems unlikely that any of our users have Infiniband on their desktops.
My guess: it's brought in by multipathd which is brought in by anaconda.
OK that is not correct...
I can remove rdma-core with impunity, nothing else gets removed. And yet it's not in comps. Weak dependency?
It appears not related to librdmacm. When I try to remove that with dnf, it tries to remove gnome-boxes, qemu, et al.
On Tue, Jan 26, 2021 at 9:07 PM Chris Murphy lists@colorremedies.com wrote:
On Tue, Jan 26, 2021 at 2:02 PM Matthew Miller mattdm@mattdm.org wrote:
This only happened to come up because of a multilib glitch, but I'm wondering why rdma-core is on the default Workstation image. It seems unlikely that any of our users have Infiniband on their desktops.
My guess: it's brought in by multipathd which is brought in by anaconda.
Your guess is wrong, it was due to libpcap being linked against libibverbs to be able to decode IB packets which in turn pulled in rdma-core, I split out the one library into libibverbs-core to pull in the bare minimum pieces that would never be used in the vast majority of use cases, but the maintainer didn't like that, rdma-core itself can now be removed, and shouldn't be pulled in for the latest rawhide images, but libibverbs still has a lot of stuff that doesn't need to be there for non IB users, but now I'm ¯_(ツ)_/¯ because it's like arguing with a wall. Details in: https://bugzilla.redhat.com/show_bug.cgi?id=1901086 https://bugzilla.redhat.com/show_bug.cgi?id=1837812
desktop@lists.fedoraproject.org