ChanServ changed the topic of #nixos-systemd to: NixOS <3 systemd | | Next meeting 08.12.2020 14:00 UTC (every two weeks)
elvishjerricco has quit [*.net *.split]
mcint has quit [*.net *.split]
elvishjerricco has joined #nixos-systemd
mcint has joined #nixos-systemd
<arianvp_> grml
<arianvp_> im still wondering whether nix should have a story for optional runtime dependencies
<arianvp_> maybe we should lazy load them with FUSE :P
<lukegb> Does that, err, really count as optional ;)
<Mic92> Yeah. This would horrible break if your network is down
<Mic92> Randoming haning processes that wait for fuse to return
<Mic92> *hanging
<arianvp_> :P
<arianvp_> hmm
<arianvp_> what is the point of these kind of things?
<arianvp_> ++ lib.optional withLibseccomp libseccomp
<arianvp_> why not just pass `override { libseccomp = null } if you want to disable it?
<arianvp_> why do we need this bespoke logic in the package.
<arianvp_> we do it all over the place in nixpkgs and I never got why
<gchristensen> some things should be disabled by default probably?
<Mic92> sure we need a better logic for this. There is the { libseccomp ? null } hint but it also just shadows bugs in case libseccomp gets renamed in nixpkgs
<Mic92> So the module system for package is the better choise
<Mic92> *choise
<Mic92> *choice (damn brain)
<arianvp_> hmm yeh module system would be nice; but I heard performance of module system is too bad fopr nixpkgs scale
<gchristensen> we're at like 150k function calls to evaluate pkgs.hello already
<arianvp_> must go deeper
<arianvp_> Another cleanup thing I wanted to propose
<andi-> arianvp_: you want to distinguish between a dependency not being available (bootstrap) and someone not wanting that feature.
<andi-> 150k function calls sounds cheap if the execution would be faster ;-)
ajs124 has quit [Quit: Bridge terminating on SIGTERM]
ajs124 has joined #nixos-systemd
ajs124 has quit [Remote host closed the connection]
ajs124 has joined #nixos-systemd