Expand description
Running a subprocess that might not come back.
Command::output() waits forever. Every external program this
project talks to — hyprctl, gsettings, fc-list — is one that
can stop answering: a wedged compositor, a hung dconf, an NFS home
directory. Waiting forever on any of them means an app that has to
be killed, and for a login screen it would mean nobody can log in.
That is the same shape of mistake as waiting forever for a compositor to grant a session lock, which cost five visible seconds before anyone noticed. Unbounded waits on another process are worth treating as a category.
Deliberately not in hyprforge-paths, which has no dependencies and
is about where files live: this is about talking to another process,
not about the filesystem. It used to live in hyprforge-core, on the
theory that every caller already depended on that crate — but
hyprforge-network, hyprforge-bluetooth and hyprforge-power are
clients of other daemons (NetworkManager, BlueZ, systemd-logind) with
no Hyprland config to read or write, and pulling in 4,000+ lines of
hlconfig/hyprlang/lua machinery for a bounded-subprocess helper
made the layering a lie. This crate has no workspace dependencies at
all, same as hyprforge-paths, so a non-Hyprland app can use it
without dragging in Hyprland at all. hyprforge-core re-exports this
module as hyprforge_core::command so existing call sites keep
working; new code should depend on hyprforge-process directly.
Constants§
- TIMEOUT
- Long enough that a busy machine is not cut off, short enough that a person does not conclude the app is broken.
Functions§
- output
- Like
Command::output, but gives up.