Full CLI per spec v0.2 - add/open/close/edit/status/list/remove/wipe/
completions, plus a hidden `__supervise` subcommand that IS the
supervisor process.
- profile.rs: TOML-backed profiles at ~/.config/porthole/profiles/,
validated -l/-r/-d mapping + --via grammar, atomic writes.
- instance.rs: JSON runtime state at ~/.local/state/porthole/, an
flock-based lock file that's the source of truth for "is this open"
(survives a crash/kill -9 without stale-lock cleanup), pid liveness
checked against /proc rather than trusted from disk.
- supervisor.rs: the __supervise loop - spawns ssh, traps SIGTERM/SIGINT
into a flag (rather than inferring intent from ssh's exit status),
classifies failures as fatal/known-transient/unrecognized, backs off
with a stability-reset, rotates its log.
- ssh.rs: builds the ssh invocation, including splitting --via into a
-J jump chain plus the mandatory positional target.
- open.rs: the detach/re-exec dance (setsid via pre_exec) and a bounded
wait for the supervisor to reach Up/Error before open returns, so an
immediate failure surfaces as a non-zero exit instead of a false
"opened" - this took a real bug fix during smoke testing, since the
instance file's initial state (Reconnecting, meaning "attempt in
flight") was indistinguishable from "already failed once" by state
alone.
- close.rs: SIGTERM+wait, or SIGKILL the whole process group with
--force so ssh can't be left orphaned.
Smoke-tested against invalid/unreachable hosts (no real infrastructure
touched): CLI surface, validation errors, add/edit/list/status/remove,
the reconnect/backoff loop with live state transitions, close mid-retry,
edit-while-running's warning, open --all, and wipe. cargo test: 14/14.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Found both while translating the spec into code:
- `ssh -J a,b,c` alone isn't valid - ssh -J only carries jump hosts
before the final hop; it still needs a positional destination. --via's
last entry has to double as that target, which also makes --via
mandatory (there's no other field naming "the host ssh connects to"),
reversing the earlier "no" in the required column.
- A plain single-pid SIGKILL on `close --force` would leave the `ssh`
child orphaned, since a killed process can't forward anything to it.
Documents sending SIGKILL to the whole process group instead, relying
on the supervisor's own setsid() call making it the group leader.
Bumps to v0.2. Resolves the open questions and design gaps flagged during
review before implementation starts:
- Spells out the supervisor detach/re-exec mechanism (§3), mirroring how
vmic solves the same one-shot-CLI-can't-host-a-daemon problem.
- Forces BatchMode/ExitOnForwardFailure/ConnectTimeout on every ssh
invocation so headless failures surface instead of hanging on a TTY
prompt (§3.1).
- Adds backoff + stderr-based fatal/transient failure classification so
reconnect doesn't retry forever against a permanently broken profile
(§4).
- Collapses the Instance `state` enum to up/reconnecting/error, with
"no instance file" as the sole meaning of closed, removing the prior
ambiguity around close vs. crash vs. down (§2.2, §3).
- Fixes --via's conflicting comma-list-vs-repeated-flag examples by
committing to a straight ssh -J passthrough grammar (§3.1/§5.1).
- Resolves the edit --restart open question: no --restart, stays explicit.
- Simplifies exit codes to 0/1/2 (was a 6-code table with its own TODO).
- Adds a lock file for concurrent-open safety and a log rotation policy.
- Calls out reboot/logout survival as an explicit v0.1 non-goal, with
`open --all` as the only hook left for external autostart wiring (§8).
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
