Skip to main content

shep_core/config/
app.rs

1//! Per-app configuration schema — one sheep's Flockfile entry
2
3use core::fmt;
4
5use std::collections::BTreeMap;
6
7// use schemars::generate
8use serde::{Deserialize, Serialize};
9
10use crate::values::{MemSize, UpDuration};
11
12/// How a health probe checks a sheep
13// wire format: changing these strings is a breaking change
14#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
15#[cfg_attr(feature = "schema", derive(schemars::JsonSchema))]
16#[serde(rename_all = "snake_case")]
17pub enum ProbeKind {
18    /// HTTP GET must return 2xx
19    Http,
20    /// TCP connect must succeed
21    Tcp,
22    /// Command must exit 0
23    Exec,
24}
25
26/// Readiness/liveness probe configuration (spec §7)
27// wire format: changing field names/defaults is a breaking change
28#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
29#[cfg_attr(feature = "schema", derive(schemars::JsonSchema))]
30#[serde(deny_unknown_fields)]
31pub struct ProbeConfig {
32    /// Probe mechanism
33    pub kind: ProbeKind,
34    /// URL (http), `host:port` (tcp), or command line (exec)
35    pub target: String,
36    /// Time between probes (default 10s)
37    #[serde(default = "default_probe_interval")]
38    pub interval: UpDuration,
39    /// Per-probe timeout (default 5s)
40    #[serde(default = "default_probe_timeout")]
41    pub timeout: UpDuration,
42    /// Consecutive failures before the probe reports unhealthy (default 3)
43    #[serde(default = "default_failure_threshold")]
44    pub failure_threshold: u32,
45}
46
47fn default_probe_interval() -> UpDuration {
48    UpDuration::from_millis(10_000)
49}
50fn default_probe_timeout() -> UpDuration {
51    UpDuration::from_millis(5_000)
52}
53fn default_failure_threshold() -> u32 {
54    3
55}
56
57/// Per-app configuration — one sheep's entry in a Flockfile
58///
59/// Field names are the Flockfile contract (sheep-native; pm2 spellings are
60/// rejected — the importer translates them). Unknown fields are errors so
61/// typos fail loudly at parse time.
62///
63/// # Example
64/// ```
65/// use shep_core::config::AppConfig;
66///
67/// let app: AppConfig = toml::from_str("name = \"web\"\nscript = \"./srv\"").unwrap();
68/// assert!(app.autorestart); // spec default
69/// ```
70// wire format: changing field names/defaults is a breaking change
71#[derive(Clone, PartialEq, Eq, Serialize, Deserialize)]
72#[cfg_attr(feature = "schema", derive(schemars::JsonSchema))]
73#[serde(deny_unknown_fields, default)]
74pub struct AppConfig {
75    /// Unique sheep name (required)
76    #[cfg_attr(feature = "schema", schemars(extend("init" = {
77        "example": "my-first-sheep",
78        "group": "process",
79        "blurb": "A convenient and unique name for shep to display"
80    })))]
81    pub name: String,
82    /// Executable or script path (required)
83    #[cfg_attr(feature = "schema", schemars(extend("init" = {
84        "example": "./index.js",
85        "group": "process",
86        "blurb": "The script that shep should use to launch your app"
87    })))]
88    pub script: String,
89    /// Arguments passed to the script
90    #[cfg_attr(feature = "schema", schemars(extend("init" = {
91        "group": "inputs",
92        "blurb": "Arguments passed to the script, as a list"
93    })))]
94    pub args: Vec<String>,
95    /// Working directory (default: daemon's cwd at spawn registration)
96    #[cfg_attr(feature = "schema", schemars(extend("init" = {
97        "example": "/srv/app",
98        "group": "process",
99        "blurb": "Where the process runs. Without it, the daemon's own directory"
100    })))]
101    pub cwd: Option<String>,
102    /// Interpreter override (`"none"` = run script directly)
103    #[cfg_attr(feature = "schema", schemars(extend("init" = {
104        "example": "none",
105        "group": "process",
106        "blurb": "What runs the script. Set it to none to exec the file directly"
107    })))]
108    pub interpreter: Option<String>,
109    /// Environment for the sheep (merged over the daemon's filtered env)
110    #[cfg_attr(feature = "schema", schemars(extend("init" = {
111        "example": "{ NODE_ENV = 'production' }",
112        "group": "inputs",
113        "blurb": "Environment variables for this app, layered over the daemon's own"
114    })))]
115    pub env: BTreeMap<String, String>,
116    /// Instance count ("cluster" = N fork instances; spec §4)
117    #[cfg_attr(feature = "schema", schemars(extend("init" = {
118        "group": "process",
119        "blurb": "How many copies of this app to run"
120    })))]
121    pub instances: u32,
122    /// Restart on unexpected exit
123    #[cfg_attr(feature = "schema", schemars(extend("init" = {
124        "group": "control",
125        "blurb": "Restarts the process automatically when it exits unexpectedly"
126    })))]
127    pub autorestart: bool,
128    /// Start when the daemon starts / on `shep muster`
129    #[cfg_attr(feature = "schema", schemars(extend("init" = {
130        "group": "control",
131        "blurb": "Start this app when the daemon starts, and on shep muster"
132    })))]
133    pub autostart: bool,
134    /// Exit codes treated as clean stop (no restart)
135    #[cfg_attr(feature = "schema", schemars(extend("init" = {
136        "group": "control",
137        "blurb": "Exit codes that mean a clean stop, so shep will not restart"
138    })))]
139    pub stop_exit_codes: Vec<i32>,
140    /// Uptime below this marks an exit as unstable
141    #[cfg_attr(feature = "schema", schemars(extend("init" = {
142        "group": "control",
143        "blurb": "An exit sooner than this counts as unstable"
144    })))]
145    pub min_uptime: UpDuration,
146    /// Consecutive unstable exits before `errored`
147    #[cfg_attr(feature = "schema", schemars(extend("init" = {
148        "group": "control",
149        "blurb": "How many unstable exits in a row before shep gives up"
150    })))]
151    pub max_restarts: u32,
152    /// Fixed delay before every restart (alternative to backoff)
153    #[cfg_attr(feature = "schema", schemars(extend("init" = {
154        "example": "3s",
155        "group": "control",
156        "blurb": "A fixed wait before every restart, instead of growing backoff"
157    })))]
158    pub restart_delay: Option<UpDuration>,
159    /// Initial backoff delay; grows ×1.5 capped at 15s (spec §4)
160    ///
161    /// Defaults to 100ms, not unset. An unstable exit (sooner than
162    /// `min_uptime`) with neither this nor `restart_delay` configured used
163    /// to restart with no delay at all, so an app that could never start
164    /// (a missing dependency, a bad config) burned its whole `max_restarts`
165    /// budget inside a second, logging the same failure dozens of times.
166    /// `min_uptime` already existed to name that case as unstable; only the
167    /// default that should have throttled it was missing.
168    ///
169    /// All of the above assumes `restart_delay` is unset. A fixed
170    /// `restart_delay` takes precedence over this field on every exit,
171    /// stable or not, so a stable exit restarts immediately only while
172    /// `restart_delay` stays unset, and setting this field to `"0"`
173    /// disables the backoff without producing an immediate restart if a
174    /// nonzero `restart_delay` is also configured.
175    #[cfg_attr(feature = "schema", schemars(extend("init" = {
176        "example": "5s",
177        "group": "control",
178        "blurb": "Starting delay between restarts, growing each time it fails again"
179    })))]
180    pub exp_backoff_restart_delay: Option<UpDuration>,
181    /// Stop signal, one of `SIGTERM`/`SIGINT`/`SIGQUIT`/`SIGUSR2` (the `SIG`
182    /// prefix and the case are both optional). Unset means `SIGTERM`.
183    ///
184    /// A `String` rather than a [`KillSignal`](crate::config::KillSignal) so
185    /// the Flockfile schema and this struct's wire form stay plain text;
186    /// `normalize` is what refuses a name outside that set, the same split
187    /// `cron_restart` and the watch globs already use.
188    #[cfg_attr(feature = "schema", schemars(extend("init" = {
189        "example": "SIGTERM",
190        "group": "process",
191        "blurb": "Which signal shep sends first when stopping this app"
192    })))]
193    pub kill_signal: Option<String>,
194    /// Grace period between stop signal and SIGKILL
195    #[cfg_attr(feature = "schema", schemars(extend("init" = {
196        "group": "control",
197        "blurb": "How long shep waits after the stop signal before SIGKILL"
198    })))]
199    pub kill_timeout: UpDuration,
200    /// Send `{"kind":"shutdown"}` on the shepherd channel instead of a signal
201    #[cfg_attr(feature = "schema", schemars(extend("init" = {
202        "group": "control",
203        "blurb": "Ask the app to stop over the channel instead of signalling it"
204    })))]
205    pub shutdown_with_message: bool,
206    /// Readiness fallback window when no ready signal/probe configured
207    #[cfg_attr(feature = "schema", schemars(extend("init" = {
208        "group": "control",
209        "blurb": "How long to wait for readiness when nothing else reports it"
210    })))]
211    pub listen_timeout: UpDuration,
212    /// Drain window for the old instance during reload
213    #[cfg_attr(feature = "schema", schemars(extend("init" = {
214        "group": "control",
215        "blurb": "How long the old instance gets to drain during a reload"
216    })))]
217    pub graceful_timeout: UpDuration,
218    /// How long a triggered action gets to answer on the shepherd channel
219    /// before its row becomes `ActionOutcome::TimedOut`.
220    ///
221    /// Defaults to 3s — comfortably under the 5s an RPC caller gets when it
222    /// sends no deadline of its own (`shep-client`'s `DEFAULT_DEADLINE`,
223    /// mirrored daemon-side as `rpc`'s `DEFAULT_DEADLINE_MS`). The margin
224    /// matters more than the number: push this past that budget and a caller
225    /// using the plain default gives up with `DeadlineExceeded` before the
226    /// daemon's own honest `TimedOut` row ever reaches it. A legitimately
227    /// slow action (a cache flush, say) can still ask for longer, but its
228    /// caller has to ask for a longer deadline in step —
229    /// `Client::request_with_deadline`, the way `shep logs -f` already asks
230    /// for `LOG_PLANE_DEADLINE` rather than the client's default. `normalize`
231    /// refuses a value no caller could ever satisfy, however long a deadline
232    /// it asks for; a value merely above the *default* budget is a caller's
233    /// choice to widen its own deadline, not a config error this crate can
234    /// see.
235    #[cfg_attr(feature = "schema", schemars(extend("init" = {
236        "group": "control",
237        "blurb": "How long a triggered action has to answer before shep gives up"
238    })))]
239    pub action_timeout: UpDuration,
240    /// Memory ceiling — polling enforcer restarts above this
241    #[cfg_attr(feature = "schema", schemars(extend("init" = {
242        "example": "512M",
243        "group": "control",
244        "blurb": "Restart the app if it climbs above this much memory"
245    })))]
246    pub max_memory: Option<MemSize>,
247    /// Watch files and restart on change
248    #[cfg_attr(feature = "schema", schemars(extend("init" = {
249        "group": "control",
250        "blurb": "Restart when a file changes"
251    })))]
252    pub watch: bool,
253    /// Watch ignore globs (defaults added daemon-side: dot-entries, node_modules)
254    #[cfg_attr(feature = "schema", schemars(extend("init" = {
255        "group": "control",
256        "blurb": "Paths watch should skip, on top of dotfiles and node_modules"
257    })))]
258    pub ignore_watch: Vec<String>,
259    /// Watch debounce window (default 500ms, applied daemon-side)
260    #[cfg_attr(feature = "schema", schemars(extend("init" = {
261        "example": "500",
262        "group": "control",
263        "blurb": "How long to wait after a change before restarting"
264    })))]
265    pub watch_delay: Option<UpDuration>,
266    /// Cron pattern for scheduled restarts (croner dialect)
267    #[cfg_attr(feature = "schema", schemars(extend("init" = {
268        "example": "* * * * *",
269        "group": "cron",
270        "blurb": "Restart on a schedule, written as a cron pattern"
271    })))]
272    pub cron_restart: Option<String>,
273    /// Fold (group) this sheep belongs to
274    #[cfg_attr(feature = "schema", schemars(extend("init" = {
275        "example": "backend",
276        "group": "process",
277        "blurb": "A fold to group this app with others, for commands that take one"
278    })))]
279    pub fold: Option<String>,
280    /// Run as this user (unix)
281    #[cfg_attr(feature = "schema", schemars(extend("init" = {
282        "example": "www-data",
283        "group": "process",
284        "blurb": "Run as this user, on unix"
285    })))]
286    pub user: Option<String>,
287    /// Run as this group (unix)
288    #[cfg_attr(feature = "schema", schemars(extend("init" = {
289        "example": "www-data",
290        "group": "process",
291        "blurb": "Run as this group, on unix"
292    })))]
293    pub group: Option<String>,
294    /// Stdout log file (default: `$SHEP_HOME/logs/<name>-<instance>-out.log`; `merge_logs` collapses to `<name>-out.log`)
295    #[cfg_attr(feature = "schema", schemars(extend("init" = {
296        "example": "/var/log/my-first-sheep/out.log",
297        "group": "process",
298        "blurb": "Where stdout goes. Defaults to a file under $SHEP_HOME/logs"
299    })))]
300    pub out_file: Option<String>,
301    /// Stderr log file (default: `$SHEP_HOME/logs/<name>-<instance>-err.log`; `merge_logs` collapses to `<name>-err.log`)
302    #[cfg_attr(feature = "schema", schemars(extend("init" = {
303        "example": "/var/log/my-first-sheep/err.log",
304        "group": "process",
305        "blurb": "Where stderr goes. Defaults to a file under $SHEP_HOME/logs"
306    })))]
307    pub err_file: Option<String>,
308    /// Merge instance logs into one file pair
309    #[cfg_attr(feature = "schema", schemars(extend("init" = {
310        "group": "process",
311        "blurb": "Put every instance's output in one pair of files"
312    })))]
313    pub merge_logs: bool,
314    /// Open the shepherd channel on fd 3 for this app on its own, without
315    /// needing `wait_ready` or `shutdown_with_message` to imply it.
316    ///
317    /// Defaults to `false`: a socketpair plus two pump tasks per sheep is
318    /// real cost weighed against spec §14.11's single-digit-MB idle-RSS
319    /// goal, so a channel is opened only when something asks for one.
320    #[cfg_attr(feature = "schema", schemars(extend("init" = {
321        "group": "inputs",
322        "blurb": "Opens fd 3 so the app can talk to shep directly"
323    })))]
324    pub channel: bool,
325    /// Open a pipe on this sheep's stdin, so `shep whisper` can write to it.
326    ///
327    /// Defaults to `false`, and the default is the decision rather than a
328    /// convenience. Without it a sheep gets `/dev/null` on fd 0, which is what
329    /// every sheep has had until now, and three things argue for keeping it
330    /// that way unless an app asks otherwise:
331    ///
332    /// - Flipping it for the whole flock is a behaviour change to processes
333    ///   nobody asked to change.
334    /// - **Programs detect stdin.** A closed or null fd 0 is how a great many
335    ///   programs decide they are non-interactive — no prompt, no pager, no
336    ///   readline, no colour. Handing them a pipe silently moves them to the
337    ///   other branch.
338    /// - It costs a descriptor and a pump task per sheep for the whole life of
339    ///   the process, against spec §14.11's single-digit-MB idle-RSS goal — the
340    ///   same budget [`Self::channel`]'s own default is protecting.
341    ///
342    /// Unlike `channel`, nothing implies this: `wait_ready` and
343    /// `shutdown_with_message` both need fd 3 and so turn `channel` on for you,
344    /// while nothing in shep needs a sheep's stdin except an operator typing
345    /// `shep whisper`. A sheep without it answers a `no_stdin` row and names
346    /// this field.
347    ///
348    /// The pipe's write end lives as long as the sheep does, so the app sees
349    /// EOF on stdin when the process is on its way out, never before.
350    #[cfg_attr(feature = "schema", schemars(extend("init" = {
351        "group": "inputs",
352        "blurb": "Keeps stdin open so shep whisper can write to the process"
353    })))]
354    pub stdin: bool,
355    /// Expect `{"kind":"ready"}` on the shepherd channel
356    #[cfg_attr(feature = "schema", schemars(extend("init" = {
357        "group": "control",
358        "blurb": "Wait for the app to say it is ready on the channel"
359    })))]
360    pub wait_ready: bool,
361    /// Asserts that the app itself sets `SO_REUSEPORT` before it binds —
362    /// shep binds nothing, so it cannot set the option on the app's behalf.
363    /// The child process owns the mechanism (Node ≥22's `reusePort`, Go's
364    /// `net.ListenConfig.Control`, nginx's `reuseport`); shep's contribution
365    /// is permission for the old and new instance to overlap during reload,
366    /// not the socket option itself.
367    ///
368    /// **This field is inert today.** shep never reads it: reload overlap
369    /// already happens unconditionally, so setting it changes nothing and
370    /// leaving it unset costs nothing. It is kept because `shep import`
371    /// writes it for a cluster-mode pm2 app and `shep flock` displays it, so
372    /// dropping it would silently discard a value out of an imported config.
373    /// It becomes load-bearing the day shep gains a reload mode that does NOT
374    /// overlap by default, which is when the permission it describes stops
375    /// being free — see `docs/specs/deferred.md`.
376    #[cfg_attr(feature = "schema", schemars(extend("init" = {
377        "group": "process",
378        "blurb": "Not built yet. Setting it is refused rather than quietly ignored"
379    })))]
380    pub reuse_port: bool,
381    /// Readiness probe — gates reload's AwaitReady (spec §7)
382    #[cfg_attr(feature = "schema", schemars(extend("init" = {
383        "example": { "kind": "http", "target": "http://127.0.0.1:8080/ready" },
384        "group": "control",
385        "blurb": "A health check shep waits on before it treats a reload as finished"
386    })))]
387    pub readiness_probe: Option<ProbeConfig>,
388    /// Liveness probe — failures feed the restart policy (spec §7)
389    #[cfg_attr(feature = "schema", schemars(extend("init" = {
390        "example": { "kind": "http", "target": "http://127.0.0.1:8080/healthz" },
391        "group": "control",
392        "blurb": "A health check that triggers a restart when it keeps failing"
393    })))]
394    pub liveness_probe: Option<ProbeConfig>,
395    /// Watch include globs (empty = watch cwd)
396    #[cfg_attr(feature = "schema", schemars(extend("init" = {
397        "group": "control",
398        "blurb": "Which paths to watch. Empty means the working directory"
399    })))]
400    pub watch_options: Vec<String>,
401    /// Timezone for `cron_restart` (IANA name)
402    #[cfg_attr(feature = "schema", schemars(extend("init" = {
403        "example": "US/Eastern",
404        "group": "cron",
405        "blurb": "Which timezone cron_restart is read in, as an IANA name"
406    })))]
407    pub cron_timezone: Option<String>,
408    /// Env var receiving the instance slot (default `SHEP_INSTANCE`)
409    #[cfg_attr(feature = "schema", schemars(extend("init" = {
410        "example": "INSTANCE_ID",
411        "group": "inputs",
412        "blurb": "The env var each instance finds its own slot number in"
413    })))]
414    pub increment_var: Option<String>,
415}
416
417/// Debug implementation does not leak env values (IR-41)
418impl fmt::Debug for AppConfig {
419    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
420        f.debug_struct("AppConfig")
421            .field("name", &self.name)
422            .field("script", &self.script)
423            .field("env", &format_args!("<{} vars>", self.env.len()))
424            .finish_non_exhaustive()
425    }
426}
427
428impl Default for AppConfig {
429    fn default() -> Self {
430        Self {
431            name: String::new(),
432            script: String::new(),
433            args: Vec::new(),
434            cwd: None,
435            interpreter: None,
436            env: BTreeMap::new(),
437            instances: 1,
438            autorestart: true,
439            autostart: true,
440            stop_exit_codes: Vec::new(),
441            min_uptime: UpDuration::from_millis(1000),
442            max_restarts: 16,
443            restart_delay: None,
444            // Not None: see the field's doc comment. An unstable exit with
445            // no restart policy configured must not restart instantly.
446            exp_backoff_restart_delay: Some(UpDuration::from_millis(100)),
447            kill_signal: None,
448            kill_timeout: UpDuration::from_millis(1600),
449            shutdown_with_message: false,
450            listen_timeout: UpDuration::from_millis(3000),
451            graceful_timeout: UpDuration::from_millis(8000),
452            action_timeout: UpDuration::from_millis(3000),
453            max_memory: None,
454            watch: false,
455            ignore_watch: Vec::new(),
456            watch_delay: None,
457            cron_restart: None,
458            fold: None,
459            user: None,
460            group: None,
461            out_file: None,
462            err_file: None,
463            merge_logs: false,
464            channel: false,
465            stdin: false,
466            wait_ready: false,
467            reuse_port: false,
468            readiness_probe: None,
469            liveness_probe: None,
470            watch_options: Vec::new(),
471            cron_timezone: None,
472            increment_var: None,
473        }
474    }
475}
476
477impl AppConfig {
478    /// A minimal config with spec defaults — the programmatic entry point
479    #[must_use]
480    pub fn minimal(name: &str, script: &str) -> Self {
481        Self {
482            name: name.to_string(),
483            script: script.to_string(),
484            ..Self::default()
485        }
486    }
487}
488
489#[cfg(test)]
490mod tests {
491    use super::*;
492    use crate::values::{MemSize, UpDuration};
493
494    #[test]
495    fn minimal_config_gets_spec_defaults() {
496        let app = AppConfig::minimal("web", "./server");
497        assert_eq!(app.name, "web");
498        assert_eq!(app.script, "./server");
499        assert!(app.autorestart);
500        assert!(app.autostart);
501        assert_eq!(app.instances, 1);
502        assert_eq!(app.min_uptime, UpDuration::from_millis(1000));
503        assert_eq!(app.max_restarts, 16);
504        assert_eq!(app.kill_timeout, UpDuration::from_millis(1600));
505        assert_eq!(app.listen_timeout, UpDuration::from_millis(3000));
506        assert_eq!(app.graceful_timeout, UpDuration::from_millis(8000));
507        assert_eq!(app.action_timeout, UpDuration::from_millis(3000));
508        assert!(app.max_memory.is_none());
509        assert!(app.fold.is_none());
510        assert!(!app.channel);
511    }
512
513    /// fails if `exp_backoff_restart_delay` reverts to `None`. Defect 2:
514    /// with no restart policy configured, a crash-looping app used to
515    /// restart with no delay at all, burning `max_restarts` in well under a
516    /// second. `restart_delay` (the daemon-side function that reads this
517    /// field) is what actually applies the throttle; this test only pins
518    /// that the default an unconfigured app gets is "on".
519    #[test]
520    fn unstable_restarts_are_throttled_by_default() {
521        let app = AppConfig::minimal("web", "./srv");
522        assert_eq!(
523            app.exp_backoff_restart_delay,
524            Some(UpDuration::from_millis(100))
525        );
526    }
527
528    /// fails if `stdin` defaults to anything but false. The default is the
529    /// whole decision: piping stdin for every sheep would change how a great
530    /// many programs behave (a closed stdin is how they decide they are
531    /// non-interactive), and would hold a descriptor and a task per sheep for
532    /// the life of the process.
533    #[test]
534    fn stdin_is_not_piped_unless_the_app_asks() {
535        let app = AppConfig::minimal("web", "./srv");
536        assert!(!app.stdin);
537        let parsed: AppConfig = toml::from_str("name = \"web\"\nscript = \"./srv\"").unwrap();
538        assert!(!parsed.stdin);
539    }
540
541    /// fails if the Flockfile key is spelled anything but `stdin`. It is a
542    /// contract with every config file already written against it the moment
543    /// this ships, and `deny_unknown_fields` means a rename is a hard parse
544    /// failure for the operator rather than a silently ignored key.
545    #[test]
546    fn the_flockfile_key_is_stdin() {
547        let parsed: AppConfig =
548            toml::from_str("name = \"web\"\nscript = \"./srv\"\nstdin = true").unwrap();
549        assert!(parsed.stdin);
550    }
551
552    #[test]
553    fn toml_round_trip_with_newtypes() {
554        let toml_src = r#"
555name = "worker"
556script = "python3"
557args = ["job.py", "--fast"]
558max_memory = "512M"
559min_uptime = "5s"
560fold = "backend"
561env = { RUST_LOG = "info" }
562"#;
563        let app: AppConfig = toml::from_str(toml_src).unwrap();
564        assert_eq!(app.max_memory, Some("512M".parse::<MemSize>().unwrap()));
565        assert_eq!(app.min_uptime, UpDuration::from_millis(5000));
566        assert_eq!(app.fold.as_deref(), Some("backend"));
567        assert_eq!(app.env.get("RUST_LOG").map(String::as_str), Some("info"));
568        assert_eq!(app.args, vec!["job.py", "--fast"]);
569    }
570
571    #[test]
572    fn unknown_fields_are_rejected() {
573        let err = toml::from_str::<AppConfig>(
574            "name = \"x\"\nscript = \"y\"\nmax_memory_restart = \"1G\"",
575        )
576        .unwrap_err();
577        assert!(err.to_string().contains("max_memory_restart"), "{err}");
578    }
579
580    #[test]
581    fn probe_config_parses_with_defaults() {
582        let src = r#"
583name = "api"
584script = "./api"
585
586[readiness_probe]
587kind = "http"
588target = "http://127.0.0.1:8080/healthz"
589"#;
590        let app: AppConfig = toml::from_str(src).unwrap();
591        let probe = app.readiness_probe.unwrap();
592        assert_eq!(probe.kind, ProbeKind::Http);
593        assert_eq!(probe.target, "http://127.0.0.1:8080/healthz");
594        assert_eq!(probe.interval, UpDuration::from_millis(10_000));
595        assert_eq!(probe.timeout, UpDuration::from_millis(5_000));
596        assert_eq!(probe.failure_threshold, 3);
597        assert!(app.liveness_probe.is_none());
598    }
599
600    #[test]
601    fn debug_redacts_env_values() {
602        // IR-41: env may carry secrets; Debug output lands in daemon logs.
603        // Exact string pinned so a lazy derive(Debug) refactor fails here.
604        let mut app = AppConfig::minimal("web", "./srv");
605        app.env
606            .insert("DATABASE_URL".to_string(), "postgres://secret".to_string());
607        app.env.insert("RUST_LOG".to_string(), "info".to_string());
608        assert_eq!(
609            format!("{app:?}"),
610            "AppConfig { name: \"web\", script: \"./srv\", env: <2 vars>, .. }"
611        );
612    }
613}