1a3e68de5d
Establish an inert PostgreSQL setup signal and repeatable evidence tooling without changing harness behavior. - add explicit setup participation semantics and inertness coverage - survey Vitest setup boundaries with isolated report-only fixtures - bound interleaved campaigns by process group and campaign deadline - reject missing backend samples and enable candidate diagnostics - document the rejected boundary result and successor protocol Files changed: .../test-failures/postgres-ddl-admission-bound.md | 15 ++ docs/testing.md | 10 ++ packages/core/package.json | 2 +- .../src/__test-utils__/pg-setup-participation.ts | 22 +++ .../src/__tests__/pg-setup-participation.test.ts | 21 +++ .../__tests__/vitest-setup-pg-inertness.test.ts | 10 ++ packages/core/vitest.pg.config.ts | 10 ++ .../__tests__/pg-preadmission-campaign.test.mjs | 63 +++++++ scripts/__tests__/pg-setup-boundary-probe.test.mjs | 92 ++++++++++ scripts/pg-preadmission-campaign.mjs | 194 +++++++++++++++++++++ scripts/pg-setup-boundary-probe.mjs | 182 +++++++++++++++++++ 11 files changed, 620 insertions(+), 1 deletion(-) Fusion-Task-Id: FN-9139 Fusion-Task-Lineage: 2cf8ccbf-37f9-4fdc-8e8f-326df823e1cd Co-authored-by: Fusion (runfusion.ai) <noreply@runfusion.ai>
64 lines
3.3 KiB
JavaScript
64 lines
3.3 KiB
JavaScript
import test from "node:test";
|
|
import assert from "node:assert/strict";
|
|
import {
|
|
DEFAULT_CAMPAIGN_TIMEOUT_MS,
|
|
DEFAULT_RUN_TIMEOUT_MS,
|
|
aggregate,
|
|
campaignRunEnvironment,
|
|
classifyLog,
|
|
countProbeDegradations,
|
|
decideVerdict,
|
|
parseArgs,
|
|
remainingRunTimeoutMs,
|
|
terminateProcessGroup,
|
|
} from "../pg-preadmission-campaign.mjs";
|
|
|
|
test("parses campaign arguments", () => {
|
|
assert.deepEqual(parseArgs(["--runs", "5", "--control", "a", "--candidate", "b"]), { runs: 5, scratchDir: undefined, control: "a", candidate: "b" });
|
|
});
|
|
|
|
test("classifies failed files and project identity timeout", () => {
|
|
assert.deepEqual(classifyLog("Test Files 2 failed | 9 passed\nproject-identity.test.ts hook timed out"), { failingFiles: 2, projectIdentityTimedOut: true });
|
|
});
|
|
|
|
test("counts only degraded preadmission rows", () => {
|
|
assert.equal(countProbeDegradations('{"event":"pg-preadmission-probe","outcome":"acquired-and-released"}\n{"event":"pg-preadmission-probe","outcome":"connect-failed"}\ninvalid'), 1);
|
|
});
|
|
|
|
test("enables diagnostics and addresses the run process group", () => {
|
|
assert.deepEqual(campaignRunEnvironment("sample.jsonl"), {
|
|
FUSION_PG_TEST_TEARDOWN_DIAGNOSTICS: "1",
|
|
FUSION_PG_TEST_TEARDOWN_DIAGNOSTICS_LOG: "sample.jsonl",
|
|
});
|
|
const kills = [];
|
|
terminateProcessGroup({ pid: 42, kill: assert.fail }, "SIGKILL", (...args) => kills.push(args));
|
|
assert.deepEqual(kills, [[-42, "SIGKILL"]]);
|
|
});
|
|
|
|
test("caps each run by the remaining campaign deadline", () => {
|
|
assert.equal(remainingRunTimeoutMs(1_000, 1_000), DEFAULT_RUN_TIMEOUT_MS);
|
|
assert.equal(remainingRunTimeoutMs(1_000, 1_000 + DEFAULT_CAMPAIGN_TIMEOUT_MS - 25), 25);
|
|
assert.equal(remainingRunTimeoutMs(1_000, 1_000 + DEFAULT_CAMPAIGN_TIMEOUT_MS), 0);
|
|
});
|
|
|
|
const control = Array.from({ length: 5 }, (_, i) => ({ arm: "control", valid: true, wallTimeMs: 100 + i, failingFiles: 2, peakBackends: 10, measurementAvailable: true, projectIdentityTimedOut: false, probeRecords: 0, probeDegradations: null }));
|
|
const candidate = Array.from({ length: 5 }, (_, i) => ({ arm: "candidate", valid: true, wallTimeMs: 100 + i, failingFiles: 2, peakBackends: 11, measurementAvailable: true, projectIdentityTimedOut: false, probeRecords: 1, probeDegradations: 0 }));
|
|
test("accepts affordable synthetic evidence", () => assert.equal(decideVerdict([...control, ...candidate]), "boundary-affordable"));
|
|
test("rejects taxing synthetic evidence", () => assert.equal(decideVerdict([...control, ...candidate.map((run) => ({ ...run, wallTimeMs: 999 }))]), "boundary-taxing"));
|
|
|
|
test("retains sampled peak backends and rejects a two-backend increase", () => {
|
|
const elevatedCandidate = candidate.map((run) => ({ ...run, peakBackends: 12 }));
|
|
const report = aggregate([...control, ...elevatedCandidate], false);
|
|
assert.equal(report.candidate.peakBackends, 12);
|
|
assert.equal(report.verdict, "boundary-taxing");
|
|
});
|
|
|
|
test("rejects incomplete evidence", () => assert.equal(decideVerdict([...control, ...candidate.slice(0, 4)]), "insufficient-data"));
|
|
|
|
test("rejects evidence when backend measurement is unavailable", () => {
|
|
const unavailable = candidate.map((run, index) => index === 0
|
|
? { ...run, measurementAvailable: false, peakBackends: null }
|
|
: run);
|
|
assert.equal(decideVerdict([...control, ...unavailable]), "insufficient-data");
|
|
});
|