two network/storage admin tools that only shadow needs, grouped behind one
attrset so the host list stays a list of intents rather than of packages.
winbox is a one-liner: pkgs.winbox resolves to winbox4, which nixpkgs
already builds through autoPatchelfHook against the qt/xcb stack it needs.
it does not go anywhere near nix-ld, so nothing has to be added to the
nix-ld library list for it - a comment says so, since the opposite is an
easy assumption to make about a vendored mikrotik binary. the installed
binary is WinBox, capital w, which is worth knowing at a shell prompt.
unraid-usb-creator is not in nixpkgs and upstream ships an appimage only,
so it is a repack. the appimage is a qt6/qml app - a fork of rpi-imager -
built with linuxdeploy, which means it already vendors the whole qt6 stack
plus xcb, icu, gnutls and krb5 under usr/lib. that bundle is kept verbatim
and autopatchelfed against only the handful of libraries linuxdeploy
deliberately leaves to the host: libc, libstdc++, GL, x11, fontconfig and
friends. no FHS wrapper, so no ~400mb buildFHSEnv closure. the usr/ layout
has to survive as a unit because qt.conf sitting next to the binary
declares Prefix=../ with Plugins=plugins and Qml2Imports=qml.
the non-obvious part is appendRunpaths rather than runtimeDependencies for
openssl. qt's tls backend dlopens libssl/libcrypto, so autopatchelf never
sees the dependency, and without it the app cannot fetch the release json
or stream the image. it has to be appendRunpaths because in auto-patchelf
`rpath += runtime_deps` sits inside the set-interpreter branch and so only
ever reaches executables, while `rpath.extend(append_rpaths)` runs for
every patched file. the dlopen happens from plugins/tls/libqopensslbackend
.so and glibc resolves dlopen against the calling object's DT_RUNPATH - the
wrapper's runpath is not consulted - so the path must land on the plugin
itself.
QT_QPA_PLATFORM is pinned to xcb because linuxdeploy bundled only the xcb
platform plugin; there is no libqwayland*.so anywhere in the appimage, so
this always runs through xwayland under hyprland regardless. --set-default
rather than --set, so it can still be overridden if upstream ever ships the
wayland plugin.
udisks2 is enabled from this module even though hyprland/default.nix
already enables it for udiskie. the app enumerates and unmounts removable
drives over the udisks2 d-bus api rather than poking /dev, which is how it
avoids needing root, and repeating the option keeps the module standalone
on a host with no hyprland. bool options merge on equal values, so this
costs nothing.
v1.1.0 is pinned by hash. note the tag is not returned by the github
releases api - which only lists the v2.0.x line - but the asset url that
unraid.net/download redirects to is live and matches.
verified: shadow evaluates clean, both packages build, the .desktop Exec
is rewritten from the bare name to the absolute store path, ldd on the
main binary reports nothing unresolved, and the openssl lib dir is present
on libqopensslbackend.so's runpath. not yet exercised at runtime against
an actual usb stick.
importing /etc/nixos/hardware-configuration.nix by absolute path from the
shared core module had two problems: it made the flake impure (every eval
needed --impure, since the file lives outside the store) and both hosts
resolved to whichever machine's file was on disk, so evaluating t480 on shadow
silently baked in shadow's disk uuids.
shadow now carries its own generated scan at modules/hosts/shadow/_hardware.nix.
the leading underscore keeps import-tree from picking it up as a flake-parts
module - its filter drops any path containing "/_" - so it is imported
explicitly from the host's default.nix.
it is generated with --no-filesystems and all of shadow's filesystems are
declared by hand in disks.nix instead. that split is what makes regeneration
safe: nixos-generate-config emits fileSystems for whatever is mounted at the
time, which would otherwise conflict with the entries below every time the data
drives happened to be mounted. `nix run .#sync-hardware` refreshes the file.
DUMP/GAMES/BAK are mounted by label at /mnt/{dump,games,bak} with
noauto + x-systemd.automount, so the two 2.7T spinners stay parked until
something touches the path and spin back down after 10 minutes idle. nofail
keeps a missing disk from dropping the boot into emergency mode.
/ and /boot are unchanged - still the same uuids, verified identical to the
previous evaluation. t480 keeps the absolute import for now with a TODO; its
scan can only be generated on the machine itself.
ahmza