More specifically, issues for one distro (or maybe family of distros if applicable) and not other distros.
I’ll start: NixOS. I love how I can have my whole operating system configuration defined deterministically with configuration files, but what I believe is a serious flaw, is when you choose the “unstable” channel, the channel for receiving the latest packages rather than ones up to 6 months old: stable. Unstable package often build dependencies on device, and I’ve often faced build failures, why are new package versions given if they fail to build?!
Why can’t Nixpkgs backend ensure that new package versions build successfully before shipping them to end users?! It would take a lot of work to do so, but I can’t think of any other distros that have this problem: where installing a new package version that came out a few days or even a week ago, has a chance of failing after its made available in the distro’s official package manager.
There are a handful of workarounds for your NixOS configuration in this case, but it happens too often for me and I shouldn’t have to edit my config for to work around it.
I’ve been using the unstable channel and am rarely seeing packages being built locally. But yeah, there’s no negative caching, if it’s something that fails to build in NixOS CI, local build will be attempted and will fail again.
Package update must build successfully before it’s merged to git, and a git master must build basic system successfully before it’s pushed to unstable channel. Requiring all packages to build successfully is infeasible for unstable, that’s what stable releases are for.
Do you have this problem with some particular packages, or is it always something else?