1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
//! The apply summary line and its two extra dimensions.
//!
//! Split out of `apply_output.rs` (forjar#336): adding the drift dimension put
//! that file over the repo's 500-line ceiling, which the pre-commit ratchet
//! enforces by refusing growth. The summary is the surface the whole
//! `apply-summary-distinguishability-v1` contract is about, so it is the piece
//! that earns its own module.
use super::helpers::*;
use crate::core::types;
/// Print apply summary (JSON or text).
///
/// FJ-129: `forced_noop_count` is the number of resources that --force
/// re-applied even though the lock reported them unchanged. When > 0,
/// surface a yellow note line — that's the runtime side of contract
/// `apply-summary-distinguishability-v1`, which makes claim C3
/// (idempotency) observable through --force. The contract assertion at
/// the bottom of this function rejects nonsense states (forced-noop
/// without --force, or forced-noop > converged).
/// GH-336: the clause naming how many convergences were drift REPAIRS.
///
/// Empty at zero, and that is load-bearing: the ordinary summary must stay
/// byte-identical, because other suites assert exact substrings of it
/// (`contains("0 converged")`, `contains("1 unchanged")`).
fn drift_clause(n: usize) -> String {
if n == 0 {
String::new()
} else {
format!(" ({n} repaired drift)")
}
}
/// GH-336: name each repaired resource under the summary.
///
/// A convergence driven by observed reality is a different event from one
/// driven by a config change — the difference between a deploy and an
/// intrusion, or a deploy and a unit that keeps resetting itself. The finding
/// was already printed by the pre-apply gate, on stderr, before the run; a
/// scrollback or a CI log tail loses it by the time the summary lands.
fn print_drift_repairs(repairs: &[super::apply_drift::DriftRepair]) {
for r in repairs {
println!(
"{}",
yellow(&format!(
" drift-repaired: [{}] {} — {}",
r.machine, r.resource_id, r.detail
))
);
}
}
#[allow(clippy::too_many_arguments)]
pub(super) fn print_apply_summary(
config: &types::ForjarConfig,
results: &[types::ApplyResult],
total_converged: u32,
total_unchanged: u32,
total_failed: u32,
forced_noop_count: u32,
drift_repaired: &[super::apply_drift::DriftRepair],
dur_apply: std::time::Duration,
json: bool,
) -> Result<(), String> {
// FJ-129 contract: forced_noop must not exceed converged. If --force
// wasn't used, the caller passes 0; if --force WAS used, every forced
// no-op is by definition counted in `total_converged` (the executor
// ran the resource), so forced_noop ≤ converged is an invariant.
debug_assert!(
forced_noop_count <= total_converged,
"C3-FORCE-DISTINGUISHABLE violated: forced_noop ({}) > converged ({})",
forced_noop_count,
total_converged
);
// GH-336: a repair is a convergence, so it is already inside
// `total_converged`. The same shape as the FJ-129 assert above.
debug_assert!(
drift_repaired.len() as u32 <= total_converged,
"drift_repaired ({}) > converged ({})",
drift_repaired.len(),
total_converged
);
let actual_changes = total_converged.saturating_sub(forced_noop_count);
let drift_clause = drift_clause(drift_repaired.len());
if json {
let output = serde_json::json!({
"name": config.name,
"machines": results,
"summary": {
"total_converged": total_converged,
"total_unchanged": total_unchanged,
"total_failed": total_failed,
// FJ-129: forced_noop_count is the C3-observable extension.
// Always present so JSON consumers can branch on > 0;
// 0 means either --force wasn't used or every forced
// resource genuinely needed work.
"forced_noop_count": forced_noop_count,
"actual_changes": actual_changes,
// GH-336: the `drift:` lines go to STDERR and the JSON to
// stdout, so before this a `--json` consumer got ZERO drift
// signal — a strictly worse form of the defect than the text
// mode the issue describes. Always present, so a parser can
// branch on > 0.
"drift_repaired_count": drift_repaired.len(),
"drift_repaired": drift_repaired
.iter()
.map(|d| serde_json::json!({
"machine": d.machine,
"resource": d.resource_id,
"detail": d.detail,
}))
.collect::<Vec<_>>(),
"total_duration_seconds": dur_apply.as_secs_f64(),
}
});
println!(
"{}",
serde_json::to_string_pretty(&output)
.map_err(|e| format!("JSON serialization error: {e}"))?
);
} else {
for result in results {
let failed_str = if result.resources_failed > 0 {
red(&format!("{} failed", result.resources_failed))
} else {
format!("{} failed", result.resources_failed)
};
println!(
"{}: {} converged, {} unchanged, {} ({:.1}s)",
bold(&result.machine),
green(&result.resources_converged.to_string()),
result.resources_unchanged,
failed_str,
result.total_duration.as_secs_f64()
);
}
println!();
if total_failed > 0 {
println!(
"{}",
red(&format!(
"Apply completed with errors: {total_converged} converged{drift_clause}, {total_unchanged} unchanged, {total_failed} FAILED"
))
);
} else {
println!(
"{}",
green(&format!(
"Apply complete: {total_converged} converged{drift_clause}, {total_unchanged} unchanged."
))
);
}
print_drift_repairs(drift_repaired);
// FJ-129: When --force re-ran resources the lock reported as
// unchanged, surface that explicitly so claim C3 is observable.
// `actual_changes == 0` AND `forced_noop_count > 0` is the
// C3-PASS shape; it's worth calling out unambiguously.
if forced_noop_count > 0 {
println!(
"{}",
// SAY WHAT WAS MEASURED. Both counts are LOCK-relative, per
// contract apply-summary-distinguishability-v1
// (forced_noop_count = shadow_plan(config, real_locks).unchanged).
// The old wording — "{n} actual change(s)" — read as "nothing on
// the machine changed", so `--force` restoring a file that had
// been tampered with on disk announced "0 actual change(s)".
// That is the LOCK's answer, and it is correct; the phrasing
// claimed the MACHINE's answer, which this count never measured.
//
// The count itself is deliberately NOT changed: subtracting live
// drift here was tried and reverted as a regression (GH-208,
// FJ-129 shape 4 went 2 -> 1), and `forjar drift` is the command
// that answers the live question. See the ledger note on
// force-and-rollback-report-zero-actual-changes.
yellow(&format!(
"note: --force re-ran {forced_noop_count} resource(s) the lock reported as unchanged \
({actual_changes} differed from the lock, {forced_noop_count} matched it). \
This is lock-relative: run `forjar drift` for what the machines actually hold."
))
);
}
}
Ok(())
}