hyperhive/hive-c0re
müde 6f1b664c85 lifecycle: stream nixos-container stdout/stderr line-by-line
run() previously buffered the child's output via .output() and
only logged at exit — a multi-minute 'nixos-container update'
(typical on a fresh hyperhive bump) showed nothing in journald
until the very end. operator watching 'journalctl -u hive-c0re
-f' couldn't tell 'slow nix build' from 'wedged daemon'.

new shape: spawn with piped stdio, pump each line into tracing
as it arrives (stdout → INFO, stderr → WARN), keep a tail of
the last 32 stderr lines for the bail message so the eventual
'failed (status 2)' still carries the actual nix eval error.
target field 'nixos-container', argv-equivalent attached via
the 'cmdline' field so filtering by subcommand works.
2026-05-16 02:57:16 +02:00
..
assets dashboard: terminal compose with @-mention sticky recipient 2026-05-16 01:55:00 +02:00
src lifecycle: stream nixos-container stdout/stderr line-by-line 2026-05-16 02:57:16 +02:00
Cargo.toml dashboard: diff against applied/proposal/<id>, prefer fetched_sha 2026-05-15 23:18:17 +02:00