Building llama.cpp with CUDA on Fedora
Following up on the earlier article about building llama.cpp with CUDA support on Windows, I tried the same thing on a Fedora 43 environment. Only the toolchain changes — the overall flow is almost identical.
Following up on the earlier article about building llama.cpp with CUDA support on Windows, I tried the same thing on a Fedora 43 environment. Only the toolchain changes — the overall flow is almost identical.
Following the previous post, I set up an SSH X11 forwarding environment and xeyes displayed fine on the Windows side. But nautilus simply refused to launch. This is a record of tracking down why this one app kept crashing even though DISPLAY was clearly working.
The Fedora 43 Server had neither an X server nor a desktop environment installed. From there, GUI apps needed to be displayed on a Windows machine. This is a record of setting up SSH X11 forwarding to forward only individual apps, without installing a desktop environment.
I set up a DNS server for my home LAN to speed up name resolution while also blocking ads. Upstream is Cloudflare's 1.1.1.1. Instead of Pi-hole, I got it all done with the lightweight dnsmasq that was already installed. This is a record of the process, including the SELinux issue I ran into along the way.
I ran into a case where trying to reach a Fedora Server machine on the same LAN via myhost.local (mDNS) failed name resolution only. By capturing packets directly with tcpdump, I traced the cause to firewalld blocking inbound mDNS traffic.