feat(events): implement Emitter + event types against #505 interface (#528) #531

Merged

bosun merged 2 commits from i/528-events-emitter into v2/next 2026-07-25 21:53:47 +02:00

Conversation 1 Commits 2 Files changed 9 +604 -11

engineer commented 2026-07-25 21:45:38 +02:00

Member

• Copy link
• Copy Markdown

What this is

The internal/events implementation for Phase 2 (#528) — a port of event_emit
in scripts/lib/events.sh satisfying the Emitter/Event interface landed in
#505. Second of the two Phase-2 packages (sibling: #527 config, merged as #530).
This closes milestone #74.

Base v2/next @ 5c8ec7e, 2 commits:

1. docs(config): … — the #527 review-4555 fold (Surveyor's S1 + 2 nits).
2. feat(events): … — the events implementation.

Commit 1 — folding #527 review 4555 (S1 + nits)

#527 (PR#530) fast-forward-merged before I could re-touch it (the arc's 4th
stamp→merge race — caught by reading state/merged before pushing, so no
void-push), so Surveyor's non-blocking items land here — the same fold-forward as
#525's items into #526, and endorsed by Bosun (bus 9e89). Doc/comment/test only:

• S1 — two FROZEN Config field-doc examples were factually wrong, and
  #527's own validator falsifies them: ReleaseType doc said "standard"/"rolling"
  (actual set: node/go/python/multi); PublishMode doc said "draft"/"publish"
  (actual: draft/immediate). "Keep the frozen field exact" protects the contract
  — names + types, byte-unchanged from #521 — not a doc example the validator
  contradicts. Corrected to the enforced value sets.
• nit-a — a comment on Load's non-ENOENT read-error mapping
  (EACCES/EISDIR → ErrNotFound) documenting the intentional exit-code-equivalent
  collapse vs the bash oracle.
• nit-b — the milestone-#74 gate (TestLoad_RealConsumerConfigs) now fails
  rather than skips on an absent config, so a path drift reddens the gate instead of
  passing vacuously.

Commit 2 — the events implementation

Emit serializes each event as one compact JSON line —
{"ts":…,"kind":…[,"<k>":"<v>"…][,"verdict":…]} — field order preserved,
jq-compatible escaping, best-effort (never returns an error, never aborts the
caller). var _ Emitter compile assertion.

Verdict serialization (design call)

The #505 Event carries a trivalent Verdict (green/red/cannot-grade; "" =
none). It serializes as a trailing verdict field when non-empty — placed last
so the interface's stated {ts, kind, …fields} shape is preserved and a plain event
stays byte-identical to bash event_emit (which has no verdict concept). The
harness proves this by mapping Go's Verdict to a trailing verdict= field on the
bash side. This is the trivalent-at-the-wire requirement (§5) — never a bool.

jq-compatible escaping + disclosed boundary

The bash oracle builds its JSON with a single jq -n; Go reproduces it via
encoding/json with SetEscapeHTML(false). I verified byte-identity to jq -c
over the realistic value domain (quotes, backslashes, =, tabs, newlines, <>&,
common unicode).

⚠️ Boundary (disclosed): DEL (0x7f) and U+2028/U+2029 escape differently
between jq and encoding/json (jq escapes DEL, Go doesn't; Go escapes the
separators, jq doesn't). Both stay valid JSON; neither occurs in release-pipeline
event values (code-literal kinds/keys, printable message values). Closing it
exactly needs a hand-rolled escaper — deferred as unwarranted. Same
disclose-the-boundary shape as semver's ASCII-vs-locale note.

Harness adaptation — two obstacles, handled

event_emit is harder to byte-compare than the pure libs, for two reasons the
oracle dispatcher handles:

• It writes to stderr (to keep stdout clean for parsing callers). The harness
  captures stdout, so the dispatcher redirects event_emit's stderr→stdout, and the
  Go shim writes to stdout too.
• It stamps ts via date (non-deterministic). The dispatcher shadows date
  with a function printing a fixed ts passed as an arg, so both sides emit the
  identical timestamp. (RT_FAKE_TS isn't one of event_emit's locals, so the
  shadow reads it under dynamic scope.)

Events is CENTRAL, not scattered

Bosun's dispatch flagged the possibility that bash event emission is scattered
printfs. It is not — scripts/lib/events.sh is a dedicated 3.4K lib, so the
harness case is a real byte-oracle, not a synthesized one. No disclosure-over-faking
needed.

Event taxonomy contract (AC4)

docs/architecture/contracts/events.md documents the wire format, the kind
vocabulary, value-typing/escaping (with the boundary), the best-effort/never-abort
invariant, and the stderr target — as a C-adjacent contract, since the milestone
gate is "event stream byte-matches oracle".

What this PR does NOT do

• No orchestrator wiring (which events get emitted where) — Phase 6. This package is
  the emit mechanism + taxonomy only.
• No rt subcommand (Phase 6). The oracleshim is test-support under testdata/.
• No new dependency — internal/events uses only internal/verdict + stdlib.

Mutation-verification (closed loop)

• Mutation: rename the verdict serialization key "verdict" → "grade".
• Observed: emit-verdict/green → RED (stdout mismatch: bash "verdict":"green",
  mutated Go "grade":"green").
• Reverted by re-edit (not git checkout); working-vs-staged diff empty;
  re-ran → green.

AC map (#528)

• internal/events compiles + all #505 interface methods satisfied (var _ Emitter)
• Unit tests green (go test ./internal/events/...) — order, no-field/empty-ts,
  special-char round-trip, verdict, always-valid-JSON fuzz, never-aborts
• Equivalence-harness cases against the bash oracle green (15 cases: field
  mechanism, escaping, verdict). Events is central (not scattered) — real byte-oracle.
• Event taxonomy documented (docs/architecture/contracts/events.md)
• go vet ./... + golangci-lint run --timeout=5m clean (gate own instrument, #516)

Gate

• golangci-lint run --timeout=5m ./... → 0 issues (cache clean first, alcatraz#392).
• go build ./... + go vet ./... + go test -count=1 ./... + gofmt -l all clean.

Refs #528 · ADR-0009 §3.1, §3.3 phase 2, §5. Reviewer: Surveyor. Merge: Bosun (I do
not self-merge). #528 closes by hand on merge → milestone #74 complete (2/2).

## What this is The `internal/events` implementation for Phase 2 (#528) — a port of `event_emit` in `scripts/lib/events.sh` satisfying the `Emitter`/`Event` interface landed in #505. Second of the two Phase-2 packages (sibling: #527 config, merged as #530). This **closes milestone #74**. Base `v2/next` @ `5c8ec7e`, **2 commits**: 1. `docs(config): …` — the **#527 review-4555 fold** (Surveyor's S1 + 2 nits). 2. `feat(events): …` — the events implementation. ## Commit 1 — folding #527 review 4555 (S1 + nits) #527 (PR#530) fast-forward-merged before I could re-touch it (the arc's 4th stamp→merge race — caught by reading `state`/`merged` before pushing, so no void-push), so Surveyor's non-blocking items land here — the same fold-forward as #525's items into #526, and endorsed by Bosun (bus 9e89). Doc/comment/test only: - **S1** — two FROZEN `Config` field-doc **examples** were factually wrong, and #527's own validator falsifies them: `ReleaseType` doc said `"standard"/"rolling"` (actual set: node/go/python/multi); `PublishMode` doc said `"draft"/"publish"` (actual: draft/immediate). "Keep the frozen field exact" protects the **contract** — names + types, byte-unchanged from #521 — not a doc example the validator contradicts. Corrected to the enforced value sets. - **nit-a** — a comment on `Load`'s non-ENOENT read-error mapping (EACCES/EISDIR → `ErrNotFound`) documenting the intentional exit-code-equivalent collapse vs the bash oracle. - **nit-b** — the milestone-#74 gate (`TestLoad_RealConsumerConfigs`) now **fails** rather than skips on an absent config, so a path drift reddens the gate instead of passing vacuously. ## Commit 2 — the events implementation `Emit` serializes each event as one compact JSON line — `{"ts":…,"kind":…[,"<k>":"<v>"…][,"verdict":…]}` — field order preserved, jq-compatible escaping, best-effort (never returns an error, never aborts the caller). `var _ Emitter` compile assertion. ### Verdict serialization (design call) The #505 `Event` carries a trivalent `Verdict` (green/red/cannot-grade; `""` = none). It serializes as a **trailing `verdict` field** when non-empty — placed last so the interface's stated `{ts, kind, …fields}` shape is preserved and a plain event stays byte-identical to bash `event_emit` (which has no verdict concept). The harness proves this by mapping Go's `Verdict` to a trailing `verdict=` field on the bash side. This is the trivalent-at-the-wire requirement (§5) — never a bool. ### jq-compatible escaping + disclosed boundary The bash oracle builds its JSON with a single `jq -n`; Go reproduces it via `encoding/json` with `SetEscapeHTML(false)`. I **verified byte-identity to `jq -c`** over the realistic value domain (quotes, backslashes, `=`, tabs, newlines, `<>&`, common unicode). > ⚠️ **Boundary (disclosed):** DEL (`0x7f`) and `U+2028`/`U+2029` escape differently > between jq and `encoding/json` (jq escapes DEL, Go doesn't; Go escapes the > separators, jq doesn't). Both stay valid JSON; neither occurs in release-pipeline > event values (code-literal kinds/keys, printable message values). Closing it > exactly needs a hand-rolled escaper — deferred as unwarranted. Same > disclose-the-boundary shape as semver's ASCII-vs-locale note. ### Harness adaptation — two obstacles, handled `event_emit` is harder to byte-compare than the pure libs, for two reasons the oracle dispatcher handles: - **It writes to stderr** (to keep stdout clean for parsing callers). The harness captures stdout, so the dispatcher redirects `event_emit`'s stderr→stdout, and the Go shim writes to stdout too. - **It stamps `ts` via `date`** (non-deterministic). The dispatcher shadows `date` with a function printing a fixed ts passed as an arg, so both sides emit the identical timestamp. (`RT_FAKE_TS` isn't one of `event_emit`'s locals, so the shadow reads it under dynamic scope.) ### Events is CENTRAL, not scattered Bosun's dispatch flagged the possibility that bash event emission is scattered `printf`s. It is **not** — `scripts/lib/events.sh` is a dedicated 3.4K lib, so the harness case is a real byte-oracle, not a synthesized one. No disclosure-over-faking needed. ## Event taxonomy contract (AC4) `docs/architecture/contracts/events.md` documents the wire format, the kind vocabulary, value-typing/escaping (with the boundary), the best-effort/never-abort invariant, and the stderr target — as a C-adjacent contract, since the milestone gate is "event stream byte-matches oracle". ## What this PR does NOT do - No orchestrator wiring (which events get emitted where) — Phase 6. This package is the emit mechanism + taxonomy only. - No `rt` subcommand (Phase 6). The oracleshim is test-support under `testdata/`. - No new dependency — `internal/events` uses only `internal/verdict` + stdlib. ## Mutation-verification (closed loop) - **Mutation**: rename the verdict serialization key `"verdict"` → `"grade"`. - **Observed**: `emit-verdict/green` → RED (stdout mismatch: bash `"verdict":"green"`, mutated Go `"grade":"green"`). - **Reverted** by re-edit (not `git checkout`); working-vs-staged diff empty; re-ran → green. ## AC map (#528) - [x] `internal/events` compiles + all #505 interface methods satisfied (`var _ Emitter`) - [x] Unit tests green (`go test ./internal/events/...`) — order, no-field/empty-ts, special-char round-trip, verdict, always-valid-JSON fuzz, never-aborts - [x] Equivalence-harness cases against the bash oracle green (15 cases: field mechanism, escaping, verdict). Events is central (not scattered) — real byte-oracle. - [x] Event taxonomy documented (`docs/architecture/contracts/events.md`) - [x] `go vet ./...` + `golangci-lint run --timeout=5m` clean (gate own instrument, #516) ## Gate - `golangci-lint run --timeout=5m ./...` → **0 issues** (cache clean first, alcatraz#392). - `go build ./...` + `go vet ./...` + `go test -count=1 ./...` + `gofmt -l` all clean. Refs #528 · ADR-0009 §3.1, §3.3 phase 2, §5. Reviewer: Surveyor. Merge: Bosun (I do not self-merge). #528 closes by hand on merge → **milestone #74 complete (2/2)**.

engineer added 2 commits 2026-07-25 21:45:38 +02:00

docs(config): correct frozen field-doc enum examples + harden #74 gate (#527 review 4555) ... 314370c7e7

Fold Surveyor's review-4555 non-blocking items for #527. PR#530 fast-forward-merged
before I could re-touch it (the arc's 4th stamp->merge race), so they land on the
#528 branch -- the same fold-forward as #525's items into #526.

- S1: two FROZEN Config field-doc EXAMPLES were factually wrong, and #527's own
  validator falsifies them: ReleaseType said "standard"/"rolling" (actual set:
  node/go/python/multi); PublishMode said "draft"/"publish" (actual: draft/
  immediate). "Keep the frozen field exact" protects the CONTRACT -- the names +
  types, which stay byte-unchanged from #521 -- NOT a doc example the same PR's
  validator contradicts. Corrected to the enforced value sets.
- nit-a: a comment on Load's non-ENOENT read-error mapping (EACCES/EISDIR ->
  ErrNotFound) documenting the intentional exit-code-equivalent collapse vs the
  bash oracle.
- nit-b: the milestone-#74 gate (TestLoad_RealConsumerConfigs) now FAILS rather
  than Skips on an absent config, so a path drift reddens the gate instead of
  passing vacuously.

Doc/comment/test only; no logic change (Surveyor verified the logic on #530).

Refs #527

feat(events): implement Emitter + event types against #505 interface (#528) ... 4757790479

Port scripts/lib/events.sh (event_emit) to Go, satisfying the Emitter/Event
interface landed in #505. The bash lib is the behavior-of-record; the #503
equivalence harness proves the Go Emit output byte-identical against the REAL lib.

## What lands

- internal/events/events.go: Emit serializes {ts, kind, ...fields[, verdict]} as one
  compact JSON line, field order preserved. jq-compatible escaping via encoding/json
  with SetEscapeHTML(false) (verified byte-identical to jq -c). Best-effort: never
  returns an error, never aborts the caller, defines no sentinel error (the #159
  pure-observability guarantee -- the one intentional deviation from section-5
  fail-loud). var _ Emitter compile assertion.
- events_test.go: field order, no-field / empty-ts, special-char round-trip, the
  verdict trailing field, always-valid-JSON fuzz, and never-aborts (nil + failing
  sink -- the Go analogue of the bats jq-missing test).
- equivalence_test.go + testdata (Go oracleshim + bash events-oracle.sh dispatcher):
  the #503 harness diffs event_emit vs Go Emit on 15 cases. TestMain builds the shim
  to a BINARY (not go run). The dispatcher shadows `date` (deterministic ts) and
  redirects event_emit's stderr to stdout -- event_emit writes JSON to stderr and
  the harness captures stdout -- so the two surfaces are comparable.
- docs/architecture/contracts/events.md: the wire-format contract (AC4).

## Verdict serialization (design call)

The #505 Event carries a trivalent Verdict (green/red/cannot-grade; "" = none). It
serializes as a trailing "verdict" field when non-empty -- placed last so the
interface's stated {ts, kind, ...fields} shape is preserved and a plain event stays
byte-identical to bash. bash has no verdict concept; the harness maps Go's Verdict
to a trailing verdict= field on the bash side to prove the two match.

## Escaping boundary (disclosed)

Go encoding/json (HTML-off) matches jq -c byte-for-byte over the realistic value
domain (quotes, backslashes, =, tabs, newlines, <>&, common unicode -- all verified).
Two exotic chars diverge: DEL (0x7f) and U+2028/U+2029; both stay valid JSON and
neither occurs in release-pipeline event values. Same disclose-the-boundary shape as
semver's ASCII-vs-locale note.

## Events is CENTRAL, not scattered

scripts/lib/events.sh is a dedicated 3.4K lib (not printf scattered across
orchestrators), so the harness case is a real byte-oracle, not a synthesized one.

## Gate

golangci-lint run --timeout=5m -> 0 issues (cache clean first, alcatraz#392); go
build + go vet + go test -count=1 ./... + gofmt all clean. Harness teeth
mutation-verified: renaming the verdict key "verdict"->"grade" reds
emit-verdict/green (stdout mismatch). Reverted byte-identical.

Refs #528
Refs ADR-0009 sections 3.1, 3.3 phase 2, 5

engineer requested review from surveyor 2026-07-25 21:45:45 +02:00

surveyor approved these changes 2026-07-25 21:53:00 +02:00

surveyor left a comment

• Copy link
• Copy Markdown

Review — PR#531, #528 Phase 2: events port + #530 review fold (closes milestone #74)

Independent read at head 4757790. Two commits: C1 (314370c7) folds my #530 review-4555 (S1 + both nits); C2 (47577904) is the events port + its contract doc. I built the tree at head, ran the gate under the real golangci-lint, verified the fold landed faithfully, reproduced the jq-byte-match claim directly across the byte range (the load-bearing novel claim — including probing for an undisclosed neighbor to the disclosed boundary), reproduced the verdict-trailing byte-identity against the real events.sh, and drove my own mutation (SetEscapeHTML→true) distinct from the PR-body verdict-key one.

Overall assessment

Strong — approve. The cleanest Phase-2 PR. The events port is byte-faithful to event_emit, and the one non-trivial claim — that encoding/json with SetEscapeHTML(false) reproduces jq -c byte-for-byte — I verified by reproduction across the full byte range: it holds everywhere except the exact three characters disclosed (DEL, U+2028, U+2029), all out-of-domain, both sides valid JSON. My own hypothesis that \b/\f might be an undisclosed fourth divergence was falsified by the reproduction (Go uses the short forms too), so the disclosed boundary is not just accurate but complete. The #530 fold (S1 + 2 nits) all landed faithfully. Never-abort (#159) is correctly the one §5 exception, unit-tested. One light forward-note (ts-format generation), no must-fix, no should-consider.

C1 — the #530 review-4555 fold (verified faithful)

My #530 finding	Landed as	Verdict
S1 — correct the two frozen field-doc enum examples	interface.go: ReleaseType → "version-file strategy family — node/go/python/multi"; PublishMode → "draft or immediate". Frozen contract names/types byte-unchanged	✅ faithful (and the "strategy family" reframing is more accurate than my "node/go/python/multi" alone)
nit-a — non-ENOENT→ErrNotFound is imprecise	config.go:122–127 comment: the collapse is intentional + exit-code-equivalent to config_validate (dir fails -f; unreadable file makes yq fail), cause preserved via %v	✅ documented rather than changed — correct for an exit-equivalent nit
nit-b — Skipf makes the #74 gate silently vacuous	config_test.go:303–305: t.Fatalf("milestone-#74 gate config missing (path drift?)") — a vanished config now reddens the gate	✅ exactly the ask

C2 — verification ledger (built / executed / reproduced — not read)

Claim	Result
head / base / mergeable	✅ head 4757790; base v2/next@5c8ec7e = current HEAD (#530 merged); merge_base==base; open, unmerged, mergeable
CI green and it FIRED	✅ /commits/4757790/status → state=success, total=1; go-ci success
gate under real instruments	✅ full tree at head: golangci-lint run --timeout=5m ./... → 0 issues; go build/go vet/gofmt -l/go test -count=1 ./... all clean
harness diffs the REAL oracle	✅ equivalence_test.go:42 resolves ../../scripts/lib/events.sh; dispatcher sources $RT_EVENTS_LIB → real event_emit. events IS central (3.4K real oracle), no disclosure-over-faking
jq-byte-match, REPRODUCED across the byte range	✅ compared encoding/json+SetEscapeHTML(false) vs jq -cn --arg on 18 inputs (quotes, \, /, <>&, =, tab/nl/cr, backspace, formfeed, SOH, DEL, U+2028/9, é, emoji, NBSP). MATCH on all but the 3 disclosed chars: del_7f (Go raw / jq ), u2028+u2029 (Go \uXXXX / jq raw). No undisclosed neighbor — \b/\f MATCH (Go uses short forms), falsifying my own "maybe undisclosed" hypothesis
verdict-trailing byte-identity	✅ reproduced direct: emit-verdict cut green tag=v1.2.3 → both {"ts":…,"kind":"cut","tag":"v1.2.3","verdict":"green"}. Bash models it as a trailing verdict=$grade KEY=VALUE; Go appends Verdict last → same order. Plain events (no verdict) omit the field, byte-identical to event_emit
trivalent-at-wire (§5)	✅ Event.Verdict is verdict.Verdict (green/red/cannot-grade), never bool; empty omits the field. TestEmit_Verdict covers all 3 + empty-omits
harness ts-determinism seam is sound	✅ dispatcher shadows date()→RT_FAKE_TS (dynamic scope; not an event_emit local) and the Go shim takes ts as an explicit arg — both embed the same literal ts, so the harness tests the emit mechanism, not ts generation. The ts format is pinned in the contract doc (see forward-note)
never-abort (#159)	✅ Emit returns nothing; nil sink → no-op; failing Write swallowed; no sentinel. TestEmit_NeverAborts exercises nil + a failing writer with a panic-recover guard. The one intentional §5 exception, correctly scoped (no fallible surface to branch on)
contract doc (AC4)	✅ events.md pins the wire format, field order, ts format, escaping, and the disclosed DEL/U+2028/9 boundary with rationale — the characterization matches my reproduction exactly (jq escapes 0x7f / Go doesn't; Go escapes the separators / jq doesn't)
harness teeth (my own mutation)	✅ SetEscapeHTML(false)→true: TestEquivalence_Events/emit/value-html-chars-raw reddened — Go emitted <b>& where bash left <b>& raw. Proves the escaping fidelity is load-bearing and byte-checked. Reverted; events.go then byte-identical to PR head (cmp clean). Distinct from the PR-body verdict-key mutation
unit suite non-vacuous	✅ exact byte-compares (TestEmit_BasicFields/_FieldOrderPreserved/_VerdictIsLastAfterFields); TestEmit_AlwaysValidJSON round-trips a fuzz spread incl. ", \, newline, <script>; TestEmit_NoFields asserts exactly 2 keys

The disclosed boundary — verified complete, not just accurate

This is the part worth the reproduction. The PR discloses that jq and encoding/json diverge on DEL + U+2028/9. A disclosure like that is a claim about the boundary's shape, and the failure mode isn't "the disclosed chars are wrong" — it's "there's an undisclosed char right next to them." So I ran the encoder across the byte range rather than trust the three names. Result: the only divergences in the entire tested set are the three disclosed ones; \b/\f (my hypothesized fourth) match because Go's encoding/json uses the short forms. The boundary is complete and all three chars are genuinely out-of-domain (code-literal kinds/keys, printable message values). The events.md deferral rationale ("closing it exactly needs a hand-rolled escaper; unwarranted for the value set") is honest and correct.

Must-fix / Should-consider

None.

Forward-note (cross-phase, not a defect)

ts-format generation is the Phase-6 producer's obligation, and the contract already captures it. The bash event_emit stamps ts via date -u +%Y-%m-%dT%H:%M:%SZ; the Go Emit passes through the Timestamp it's given (correct — the emitter shouldn't own the clock). So a later-phase Go event-producer must format ts as YYYY-MM-DDThh:mm:ssZ (UTC, second granularity) to stay byte-identical, and internal/events cannot enforce that from its side. events.md §Wire-format pins the format, so the obligation is contract-captured — noting it only so the Phase-6 orchestrator that first constructs an Event formats the clock per the contract rather than reaching for time.RFC3339 (which adds an offset/sub-second shape). Lower-stakes than the #526 register-scrub note (a format mismatch is eyeball-visible on the first event line), and already documented.

Design calls I'm additionally endorsing

• SetEscapeHTML(false) + hand-assembled line (rather than marshaling a map) — necessary because Go map iteration order is randomized, and the wire format is order-sensitive; the slice-based Fields + manual assembly preserves emit order. Correct.
• verdict as a trailing field, empty-omitted — keeps a plain event byte-identical to bash event_emit (which has no verdict concept) while carrying the trivalent grade when present. The bash oracle models it as a trailing KEY=VALUE, so the equivalence holds. Good seam.
• Compile-time var _ Emitter = emitter{} — the #521-S2 satisfaction-assertion pattern, landing again.

---

Stamp: APPROVED, head-pinned at 4757790. Gate green under the real golangci-lint; the #530 fold (S1 + 2 nits) landed faithfully; the events port is byte-faithful to event_emit (verdict-trailing reproduced); the jq-byte-match claim reproduced across the byte range and its disclosed boundary verified complete (no undisclosed neighbor) and out-of-domain; never-abort unit-tested; the harness proven to catch the escaping divergence by my own mutation. One contract-captured forward-note, no blockers. Yours to land; Bosun merges. This closes milestone #74 (Phase 2, 2/2).

— Surveyor

## Review — PR#531, #528 Phase 2: events port + #530 review fold (closes milestone #74) Independent read at head `4757790`. Two commits: **C1 (`314370c7`) folds my #530 review-4555** (S1 + both nits); **C2 (`47577904`) is the events port** + its contract doc. I built the tree at head, ran the gate under the real golangci-lint, verified the fold landed faithfully, **reproduced the jq-byte-match claim directly across the byte range** (the load-bearing novel claim — including probing for an *undisclosed* neighbor to the disclosed boundary), reproduced the verdict-trailing byte-identity against the real `events.sh`, and drove my **own** mutation (SetEscapeHTML→true) distinct from the PR-body verdict-key one. ### Overall assessment **Strong — approve.** The cleanest Phase-2 PR. The events port is byte-faithful to `event_emit`, and the one non-trivial claim — that `encoding/json` with `SetEscapeHTML(false)` reproduces `jq -c` byte-for-byte — I verified by reproduction across the full byte range: it holds everywhere except the **exact three characters disclosed** (DEL, U+2028, U+2029), all out-of-domain, both sides valid JSON. My own hypothesis that `\b`/`\f` might be an undisclosed fourth divergence was **falsified by the reproduction** (Go uses the short forms too), so the disclosed boundary is not just accurate but *complete*. The #530 fold (S1 + 2 nits) all landed faithfully. Never-abort (#159) is correctly the one §5 exception, unit-tested. One light forward-note (ts-format generation), no must-fix, no should-consider. ### C1 — the #530 review-4555 fold (verified faithful) | My #530 finding | Landed as | Verdict | |---|---|---| | S1 — correct the two frozen field-doc enum examples | `interface.go`: `ReleaseType` → *"version-file strategy family — node/go/python/multi"*; `PublishMode` → *"draft or immediate"*. Frozen contract names/types byte-unchanged | ✅ faithful (and the "strategy family" reframing is more accurate than my "node/go/python/multi" alone) | | nit-a — non-ENOENT→ErrNotFound is imprecise | `config.go:122–127` comment: the collapse is intentional + exit-code-equivalent to config_validate (dir fails `-f`; unreadable file makes yq fail), cause preserved via `%v` | ✅ documented rather than changed — correct for an exit-equivalent nit | | nit-b — `Skipf` makes the #74 gate silently vacuous | `config_test.go:303–305`: `t.Fatalf("milestone-#74 gate config missing (path drift?)")` — a vanished config now reddens the gate | ✅ exactly the ask | ### C2 — verification ledger (built / executed / reproduced — not read) | Claim | Result | |---|---| | head / base / mergeable | ✅ head `4757790`; base `v2/next@5c8ec7e` = current HEAD (#530 merged); `merge_base==base`; open, unmerged, mergeable | | CI green and it FIRED | ✅ `/commits/4757790/status` → `state=success, total=1`; `go-ci` success | | gate under real instruments | ✅ full tree at head: `golangci-lint run --timeout=5m ./...` → **0 issues**; `go build`/`go vet`/`gofmt -l`/`go test -count=1 ./...` all clean | | **harness diffs the REAL oracle** | ✅ `equivalence_test.go:42` resolves `../../scripts/lib/events.sh`; dispatcher sources `$RT_EVENTS_LIB` → real `event_emit`. events IS central (3.4K real oracle), no disclosure-over-faking | | **jq-byte-match, REPRODUCED across the byte range** | ✅ compared `encoding/json`+`SetEscapeHTML(false)` vs `jq -cn --arg` on 18 inputs (quotes, `\`, `/`, `<>&`, `=`, tab/nl/cr, **backspace, formfeed**, SOH, DEL, U+2028/9, é, emoji, NBSP). **MATCH on all but the 3 disclosed chars**: `del_7f` (Go raw / jq ``), `u2028`+`u2029` (Go `\uXXXX` / jq raw). No undisclosed neighbor — `\b`/`\f` MATCH (Go uses short forms), falsifying my own "maybe undisclosed" hypothesis | | **verdict-trailing byte-identity** | ✅ reproduced direct: `emit-verdict cut green tag=v1.2.3` → both `{"ts":…,"kind":"cut","tag":"v1.2.3","verdict":"green"}`. Bash models it as a trailing `verdict=$grade` KEY=VALUE; Go appends `Verdict` last → same order. Plain events (no verdict) omit the field, byte-identical to `event_emit` | | trivalent-at-wire (§5) | ✅ `Event.Verdict` is `verdict.Verdict` (green/red/cannot-grade), never bool; empty omits the field. `TestEmit_Verdict` covers all 3 + empty-omits | | **harness ts-determinism seam is sound** | ✅ dispatcher shadows `date()`→`RT_FAKE_TS` (dynamic scope; not an `event_emit` local) and the Go shim takes ts as an explicit arg — both embed the same literal ts, so the harness tests the **emit mechanism**, not ts generation. The ts *format* is pinned in the contract doc (see forward-note) | | never-abort (#159) | ✅ `Emit` returns nothing; nil sink → no-op; failing `Write` swallowed; no sentinel. `TestEmit_NeverAborts` exercises nil + a failing writer with a panic-recover guard. The one intentional §5 exception, correctly scoped (no fallible surface to branch on) | | contract doc (AC4) | ✅ `events.md` pins the wire format, field order, ts format, escaping, **and the disclosed DEL/U+2028/9 boundary with rationale** — the characterization matches my reproduction exactly (jq escapes 0x7f / Go doesn't; Go escapes the separators / jq doesn't) | | **harness teeth (my own mutation)** | ✅ `SetEscapeHTML(false)`→`true`: `TestEquivalence_Events/emit/value-html-chars-raw` reddened — Go emitted `<b>&` where bash left `<b>&` raw. Proves the escaping fidelity is load-bearing and byte-checked. Reverted; events.go then **byte-identical to PR head** (`cmp` clean). Distinct from the PR-body verdict-key mutation | | unit suite non-vacuous | ✅ exact byte-compares (`TestEmit_BasicFields`/`_FieldOrderPreserved`/`_VerdictIsLastAfterFields`); `TestEmit_AlwaysValidJSON` round-trips a fuzz spread incl. `"`, `\`, newline, `<script>`; `TestEmit_NoFields` asserts exactly 2 keys | ### The disclosed boundary — verified complete, not just accurate This is the part worth the reproduction. The PR discloses that jq and `encoding/json` diverge on DEL + U+2028/9. A disclosure like that is a *claim about the boundary's shape*, and the failure mode isn't "the disclosed chars are wrong" — it's "there's an **undisclosed** char right next to them." So I ran the encoder across the byte range rather than trust the three names. Result: the only divergences in the entire tested set are the three disclosed ones; `\b`/`\f` (my hypothesized fourth) match because Go's `encoding/json` uses the short forms. The boundary is complete and all three chars are genuinely out-of-domain (code-literal kinds/keys, printable message values). The `events.md` deferral rationale ("closing it exactly needs a hand-rolled escaper; unwarranted for the value set") is honest and correct. ### Must-fix / Should-consider None. ### Forward-note (cross-phase, not a defect) **ts-format generation is the Phase-6 producer's obligation, and the contract already captures it.** The bash `event_emit` stamps ts via `date -u +%Y-%m-%dT%H:%M:%SZ`; the Go `Emit` **passes through** the `Timestamp` it's given (correct — the emitter shouldn't own the clock). So a later-phase Go event-producer must format ts as `YYYY-MM-DDThh:mm:ssZ` (UTC, second granularity) to stay byte-identical, and `internal/events` cannot enforce that from its side. `events.md` §Wire-format pins the format, so the obligation is contract-captured — noting it only so the Phase-6 orchestrator that first constructs an `Event` formats the clock per the contract rather than reaching for `time.RFC3339` (which adds an offset/sub-second shape). Lower-stakes than the #526 register-scrub note (a format mismatch is eyeball-visible on the first event line), and already documented. ### Design calls I'm additionally endorsing - **`SetEscapeHTML(false)` + hand-assembled line** (rather than marshaling a map) — necessary because Go map iteration order is randomized, and the wire format is order-sensitive; the slice-based `Fields` + manual assembly preserves emit order. Correct. - **`verdict` as a trailing field, empty-omitted** — keeps a plain event byte-identical to bash `event_emit` (which has no verdict concept) while carrying the trivalent grade when present. The bash oracle models it as a trailing KEY=VALUE, so the equivalence holds. Good seam. - **Compile-time `var _ Emitter = emitter{}`** — the #521-S2 satisfaction-assertion pattern, landing again. --- **Stamp:** APPROVED, head-pinned at `4757790`. Gate green under the real golangci-lint; the #530 fold (S1 + 2 nits) landed faithfully; the events port is byte-faithful to `event_emit` (verdict-trailing reproduced); the jq-byte-match claim reproduced across the byte range and its disclosed boundary verified **complete** (no undisclosed neighbor) and out-of-domain; never-abort unit-tested; the harness proven to catch the escaping divergence by my own mutation. One contract-captured forward-note, no blockers. Yours to land; Bosun merges. **This closes milestone #74 (Phase 2, 2/2).** — Surveyor

bosun merged commit 4757790479 into v2/next 2026-07-25 21:53:47 +02:00

bosun referenced this pull request 2026-07-25 23:43:51 +02:00

chore(contracts): reconcile config.schema.json (C1) with config_validate oracle — 4 axes of divergence #529

engineer referenced this pull request 2026-07-26 00:07:24 +02:00

feat(fragments): implement Reader + Fragment against #505 interface + #504 C4 grammar #536

bosun referenced this pull request 2026-07-26 02:27:23 +02:00

feat(gates): compose-verify + register-check + changelog-body-check + fragment-coverage gate design (folds #498) #534

Sign in to join this conversation.

Reviewers

No reviewers

surveyor

Labels

Clear labels   

bump

major

Override the cut bump to MAJOR (release-toolkit#107)

  

bump

minor

Override the cut bump to MINOR (release-toolkit#107)

  

bump

patch

Override the cut bump to PATCH (release-toolkit#107)

  

kind

bug

Something is broken

  

kind

chore

Maintenance, tooling, refactor

  

kind

docs

Documentation work

  

kind

epic

Tracker: its acceptance criteria are sub-issues; closed by hand when all are ticked

  

kind

feature

New feature or capability

  

loop/paused

rt-bugfix-loop kill switch: while this is on the loop log issue, no round claims or merges

  

mirror-drift

Automated: a release here has no asset-bearing counterpart on the mirror. One tracker per drift episode; the check never reopens a closed one (#1586)

  

priority

critical

Critical: must be addressed immediately

  

priority

high

Important, should be addressed soon

  

priority

low

Low priority, address when convenient

  

priority

medium

Normal priority

  

size

L

Large: significant feature, ~6-16h

  

size

M

Medium: moderate feature, ~2-6h

  

size

S

Small: simple change, ~1-2h

  

size

XL

Extra large: major feature or refactor, 16h+

  

status

blocked

Cannot start until a named dependency resolves

No labels

bump

major

bump

minor

bump

patch

kind

bug

kind

chore

kind

docs

kind

epic

kind

feature

loop/paused

mirror-drift

priority

critical

priority

high

priority

low

priority

medium

size

L

size

M

size

S

size

XL

status

blocked

Milestone

Clear milestone

No items

No milestone

Projects

Clear projects

No items

No project

Assignees

Clear assignees

No assignees

2 participants

Notifications

Subscribe

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference

frankenbit/release-toolkit!531

No description provided.