mirror of
https://github.com/garrytan/gstack.git
synced 2026-10-05 02:47:17 +02:00
v1.91.8.0 test: smaller suite, derived paid selection, retired never-green evals (#2994)
* test: delete test-infrastructure dead code (G) - exit-propagation drives the runner's real strict verdict (BunTestOutputClassifier + strictTestExitCode); delete the unused shardRunLooksTruncated predicate. - delete skill-coverage-matrix registry + its gate (nothing reads it; the floor already iterates skillCensus()). - delete touchfiles-facade export-parity tests (Bun fails missing imports at link time) and the duplicated E2E_TIERS tier-value test. - delete brain-cache-spec TRANSPORT_DEFAULT_POLICY, SKILL_RUN_RETENTION_DAYS and the now-unused BrainTrustPolicy type with their literal tests. AUTOPLAN_PREFLIGHT_BUDGET_BYTES stays: skill-preflight-budget enforces it against real resolver output. - delete audit-compliance's JSDoc-comment grep. * test: replace product tests that fake the product with real-boundary tests (F) - design: serve.test.ts drove an inline mirror server; now two tests run the real serve() on an ephemeral port (reload confinement, submit exit 0). - setup-gbrain: rollback + voyage tests execute the template-extracted init blocks (3 sites) instead of drifted local bash copies. - terminal-agent: internalHandler source greps replaced by a behavioral /internal/grant + /internal/revoke auth matrix (no/wrong/valid token). - /health: server-security-surface and the server-auth / security-audit-r2 / sidebar-tabs source greps fold into one liveness-only check on the real body; the L4 sidecar wiring gets a behavioral /pty-inject-scan test. - delete tautologies (browser-manager onDisconnect, memory-command #12), ios swiftui tap fixture self-check, memory-ingest put_page grep, detach source greps, sidebar-agent absence pins, dead-CSS pins + the dead CSS, security-audit-r2 Task 1 + the test-only meta-commands re-export, duplicate generated-SKILL.md checks. - make-pdf coverage-gaps cases move into their owner test files. * test: delete tests of dead eval code (A) - A1: the retired Eng lexical oracle (evaluateEngSeedCoverage, isEngSeedDecisionAUQ), the completion-handoff detector and the retained corpus had no paid caller since v1.87.6; delete their 26 replay files, ~2.6k helper LOC and fixtures, and the dead blocks in 8 mixed files (live hasNativePlanTerminal / batching assertions stay). - A2: dead viewport approvers in autoplan-artifact-permission and their 11 replay files + fixtures; recorder/launcher cases stay. - A3: never-wired oracles and seeders (autoplan-phase-order, eng-finding-fixture, ceo-paired-fixture, design-ui-scope, plan-skill-completion, pty-current-screen, required-reads, transcript-section-logger); plan-seed-submission now decodes through the production createPtyScreen; section manifests name their actual guard. - A4: zero-reference helper exports, plus execGit and invokeAndObserve found by the reachability pass. - 52 fixtures orphaned by the deletions; touchfile and selection-table entries for every deleted path. * test: clean up the paid eval lane (B1-B4, B6, B7) - B1: delete paid files that assert nothing or cannot pass meaningfully: skill-llm-eval-spec and skill-e2e-spec-execute (test.todo), gemini-e2e (+ gemini-session-runner; no gemini CLI in CI), ship-idempotency (red since v1.63), the two opus-4-7 *-sonnet overlay wrappers, conductor-prose (+ its source-evaluation replay), codex-e2e-plan-format; drop their keys, scripts and census rows. - B2: skill-llm-eval grades browse/sections/command-list.md with one union judge that also carries the baseline score pin; regression-vs-baseline deleted (paid run: pass, c4/c4/a4). - B3: memory-pipeline, ios-qa, ios-qa-swift-build and plan-tune-cathedral make no model calls; renamed out of the paid glob so they run on every PR. Swift builds need GSTACK_TEST_SWIFT=1; device stub deleted. - B4: codex-e2e*, outside-voice, aside and ios-device cannot run in the CI image; excluded from the weekly lane with a tracked re-entry condition. - B6: fold opus-47's negative routing controls into skill-routing-e2e journey-negatives (paid run: 3/3 unrouted) and delete the file. - B7: delete the never-green brain-privacy-gate eval; a free gstack-skill-start test now proves consent precedes artifacts egress. * test: retire the finding-count cluster and trim its helpers (C) - C0/C1: the five never-green evals (skill-e2e-autoplan-chain and skill-e2e-plan-{ceo,eng,design,devex}-finding-count) failed on harness and budget, never on skill behavior; delete them, their touchfile/tier ids, AUTOPLAN_CHAIN_BUDGET and the dedicated eighth periodic slice (--slices 7). - C2: delete the helper groups whose only paid consumers were those files (11 modules), trim claude-pty-runner and eng-seeded-coverage to the paid closure, and delete the free replay tests whose assertions exercised only that dead code (89 files, 135 orphaned fixtures). Blocks that used dead code only as input for a live subject keep their assertions: the multiSelect default moved to plan-review-decisions, runner PTY tests use inline caller policies, and the timer-safe budget checks moved to eng-finding-retry-budget. - The eight production-touching files stay except ceo-current-decision-record (its template read only feeds the retired counter). - CARVE_GUARDS.autoplan is behavioral 'none'; TODOS records the lost chain and per-finding cadence coverage with their re-entry tests. * test: fold per-incident replay series into their detector owners (D) Twelve detector families move into one owner test each: 73 incident files become describe blocks in ceo-section-loading-fixture (stale-fill race), model-overlays, coverage-audit-evidence, autoplan-phase-observer, native-auto-decide, outside-voice-evidence, eng-first-review, plan-count-completion, plan-count-file-permission, ceo-mode-option, plan-scope-selection and plan-count-prerequisite. Each block keeps its original code and fixture, so every case still runs; only tests asserting the incident file's own touchfile registration are dropped (41). Touchfile lists that named an incident now name its owner. * test: start the plan-count history PTY on its readiness marker (H) The fake CLI prints a startup marker and the runner waits for it instead of the fixed 8 s startup sleep (8.6 s -> 0.9 s locally). eng-semantic-terminal's sleeping registration cases went with C; plan-count-timeout keeps the fixed wait because it asserts deadline behavior. * test: derive paid touchfiles from each eval's static closure (E) touchfiles.test.ts now checks, per key, that the paid file's static test/helpers and test/fixtures closure (plus fixture paths it names in string literals) is covered, and names the file, path, chain and key to fix when it is not. Free *.test.ts files are no longer touchfiles, so editing a free replay test stops selecting paid evals: 950 entries removed, 653 real closure paths added. The hand-copied inventories go: periodic-fixture-selection, fake-impeccable-touchfiles and 45 per-file selection examples. Selection for the sample edits (plan-eng-review template, claude-pty-runner, plan-count-fixture, gstack-config) loses no case under either profile. CONTRIBUTING documents the rule and its lower bound. * test: skip hollow tier shards and census judges in the paid planner (B5) A paid file is now skipped for a tier lane only when every E2E id it registers is known statically and none has that tier; ids come from the touchfile registrations and literal testName/*IfSelected arguments, so a comment or skill path that quotes another id cannot unschedule it, and computed names keep today's scheduling. --list and the manifest show each skip as "skipped: no E2E_TIERS id has tier <tier>". The weekly gate census drops the LLM judges (--skip-judges); they still run in the periodic census and PR gate lanes. Gate lane 52 -> 42 files, census 41; periodic 77 -> 69. * test: run seven paid evals on the current default capture model (B8) skill-e2e-{auq-matrix,plan-format,qa-bugs,retro,workflow} pinned claude-opus-4-7 and skill-e2e-office-hours plus -brain-writeback pinned claude-sonnet-4-6; none tests a historical model, so they now capture with resolveEvalModel('capture'), and the free harness tests that execute these registrations receive the same resolver. The paid re-pin run passed all of them. skill-e2e-{design,office-hours-phase4,plan-prosons,plan} keep claude-opus-4-7: six of their cases failed on the default model (three timeouts, a missing report file, a format miss and a posture score of 3), so per the plan's fallback they keep their pins with a TODOS entry. The pre-spend estimate and drop threshold are in docs/test-audit-2026-09.md. * test: guard the reduced suite against new test-of-test files - test/test-of-test-ratchet.test.ts records the 228 free tests that import only test/ code and fails on a new one, naming the owner test to extend instead; a stale baseline entry fails with the remove instruction. - test/helpers/resolve-repo-path.ts is the one specifier/literal resolver for the ratchet and the touchfile closure invariant, with its own unit tests. - CONTRIBUTING "Test tiers" describes the paid-failure workflow (fix, then one row in the detector's owner test) and the ratchet; TEST_PORTFOLIO gains the detector -> owner-test table and no longer claims an Autoplan chain eval. - TODOS: automatic exclusion policy for chronically red periodic files (P3), the deferred native-completion table collapse, the unused CEO payment seeder; the PTY readiness item is narrowed to the paid runner. - docs/test-audit-2026-09.md collects the triage, security mapping, inventories, selection proof, behavior-commit decisions and retained false positives. * v1.91.8.0 test: smaller suite, derived paid selection, retired never-green evals Release metadata for the test-reduction branch: VERSION 1.91.8.0 (1.91.7.0 is claimed by #2983), CHANGELOG with the measured before/after table and a contributor section, durations re-recorded on Ubicloud standard-16 (857 files, 0 failures), the agents digest, CONTRIBUTING's after-measurement row, the B8 fallback TODOS entry, and the after metrics, kept-vs-plan notes, B8 run and census estimate in docs/test-audit-2026-09.md. * fix(ubicloud): skip retrieval globs that match nothing instead of reporting a failed pull * test: count issue-numbered Design findings in the UI-scope gate eval
This commit is contained in:
Garry Tan authored and GitHub committed
1 parent
dcaea52800
commit
943105f109
668 files changed
+11998
-102381
No files matched your search
@@ -5,8 +5,6 @@ import * as path from 'node:path';
|
||||
import { runGeneration } from '../scripts/gen-skill-docs';
|
||||
import { ALL_HOST_NAMES, getHostConfig } from '../hosts';
|
||||
|
||||
const ROOT = path.resolve(import.meta.dir, '..');
|
||||
|
||||
test('every host exposes the DX per-call rule before the pre-review audit and Step 0', async () => {
|
||||
const outputRoot = fs.mkdtempSync(path.join(os.tmpdir(), 'devex-rule-free-'));
|
||||
try {
|
||||
@@ -128,387 +126,3 @@ test('every host exposes the DX per-call rule before the pre-review audit and St
|
||||
}
|
||||
} finally { fs.rmSync(outputRoot, { recursive: true, force: true }); }
|
||||
}, 20_000);
|
||||
|
||||
// Import the actual paid registration in a child with only its process boundary
|
||||
// mocked. Coverage and report validation remain the production predicates.
|
||||
function runDxRegistration(scenario: string) {
|
||||
const directory = fs.realpathSync(fs.mkdtempSync(path.join(os.tmpdir(), 'devex-registration-free-')));
|
||||
const script = path.join(directory, 'registration.test.ts');
|
||||
const facts = path.join(directory, 'facts.json');
|
||||
fs.writeFileSync(script, `
|
||||
import { describe, expect, mock } from 'bun:test';
|
||||
import * as fs from 'node:fs';
|
||||
import * as path from 'node:path';
|
||||
import { DEVEX_COUNT_FILES, planDevexCountFixture, isDevexReviewIssue, devexReviewModePick }
|
||||
from ${JSON.stringify(path.join(ROOT, 'test/helpers/devex-count-fixture.ts'))};
|
||||
import captured from ${JSON.stringify(path.join(ROOT, 'test/fixtures/devex-seed-coverage-ad-v3.json'))};
|
||||
const { assertReviewReportAtBottom: actualReport, devexStep0Boundary } =
|
||||
await import(${JSON.stringify(path.join(ROOT, 'test/helpers/claude-pty-runner.ts'))});
|
||||
const scenario = ${JSON.stringify(scenario)};
|
||||
const factsPath = ${JSON.stringify(facts)};
|
||||
const finalPlan = '# Reviewed DX plan\\n\\nThe five seeded gaps each have a recorded decision.\\n\\n## GSTACK REVIEW REPORT\\n\\nDX review complete.\\n';
|
||||
const facts = { checked: false, runnerCalls: 0, reportCalls: 0, judgeCalls: 0, planPath: '', finalPlan: '' };
|
||||
const save = () => fs.writeFileSync(factsPath, JSON.stringify(facts));
|
||||
mock.module(${JSON.stringify(path.join(ROOT, 'test/helpers/e2e-gate.ts'))}, () => ({
|
||||
describeE2ETier: tier => { expect(tier).toBe('periodic'); return describe; },
|
||||
}));
|
||||
mock.module(${JSON.stringify(path.join(ROOT, 'test/helpers/claude-pty-runner.ts'))}, () => ({
|
||||
devexStep0Boundary,
|
||||
assertReviewReportAtBottom: content => {
|
||||
facts.reportCalls++;
|
||||
facts.finalPlan = content;
|
||||
save();
|
||||
expect(content).toBe(fs.readFileSync(facts.planPath, 'utf8'));
|
||||
return actualReport(content);
|
||||
},
|
||||
runPlanSkillCounting: async opts => {
|
||||
facts.runnerCalls++;
|
||||
facts.planPath = opts.expectedPlanPath;
|
||||
save();
|
||||
expect(path.dirname(path.dirname(opts.expectedPlanPath))).toBe(${JSON.stringify(directory)});
|
||||
expect(fs.existsSync(path.dirname(opts.expectedPlanPath))).toBe(true);
|
||||
// The runner owns the seeded Git project. The caller owns only its report.
|
||||
expect(opts.cwd).toBeUndefined();
|
||||
expect(opts.skillName).toBe('plan-devex-review');
|
||||
expect(opts.slashCommand).toBe('/plan-devex-review');
|
||||
expect(opts.timeoutMs).toBe(1500000);
|
||||
expect(opts.reviewCountCeiling).toBe(Infinity);
|
||||
expect(opts.pickAUQ).toBe(devexReviewModePick);
|
||||
expect(opts.isReviewAUQ).toBe(isDevexReviewIssue);
|
||||
expect(opts.isLastStep0AUQ).toBe(devexStep0Boundary);
|
||||
expect(opts.env).toEqual({ QUESTION_TUNING: 'false', EXPLAIN_LEVEL: 'default' });
|
||||
expect(opts.followUpPrompt).toBe(planDevexCountFixture(opts.expectedPlanPath) +
|
||||
'\\nFinish this DX review; I will handle subsequent reviews manually.');
|
||||
expect(opts.fixtureFiles).toEqual(DEVEX_COUNT_FILES);
|
||||
expect(opts.followUpPrompt).toContain('Use DX POLISH');
|
||||
// These are the five unresolved contracts in the current native fixture.
|
||||
expect(opts.fixtureFiles['docs/current-contracts.md']).toContain('There is no skip flag or offline first-run path.');
|
||||
expect(opts.fixtureFiles['docs/package-contents.txt']).toContain('that file is absent');
|
||||
expect(opts.fixtureFiles['docs/api.md']).toContain('run_eval(dataset, evaluator)');
|
||||
expect(opts.fixtureFiles['docs/api.md']).toContain('run_batch(evaluator, dataset)');
|
||||
expect(opts.fixtureFiles['docs/api.md']).toContain('AuthError("request failed")');
|
||||
expect(opts.fixtureFiles['docs/api.md']).toContain('removes the old name immediately');
|
||||
facts.checked = true;
|
||||
save();
|
||||
if (scenario === 'throw') throw new Error('controlled DX runner failure');
|
||||
// Replay public question/reply evidence only. Its historical run did not
|
||||
// complete; the terminal/report below are controlled caller-boundary inputs.
|
||||
const transcript = { status: 'ready', calls: structuredClone(captured.attempts[0].calls), assistantMessages: [] };
|
||||
if (scenario.startsWith('missing-seed-')) transcript.calls.splice(Number(scenario.slice(-1)), 1);
|
||||
if (scenario === 'missing-native') transcript.status = 'missing';
|
||||
if (scenario === 'repeated-seed') transcript.calls = Array.from({length: 5}, (_, i) =>
|
||||
({ ...structuredClone(transcript.calls[0]), toolUseId: 'repeated-' + i }));
|
||||
if (scenario === 'batched') {
|
||||
const call = structuredClone(transcript.calls[0]);
|
||||
call.questions = transcript.calls.flatMap(item => item.questions);
|
||||
call.answers = Object.fromEntries(transcript.calls.flatMap(item => Object.entries(item.answers)));
|
||||
transcript.calls = [call];
|
||||
}
|
||||
if (scenario === 'pending') transcript.calls[0].answered = false;
|
||||
if (scenario !== 'missing-report') fs.writeFileSync(opts.expectedPlanPath,
|
||||
finalPlan + (scenario === 'trailing-report' ? '\\n## Unexpected follow-up\\n' : ''));
|
||||
return { outcome: scenario === 'timeout' ? 'timeout' : scenario === 'summary' ? 'completion_summary' : 'plan_ready',
|
||||
transcript, fingerprints: [], step0Count: 2, reviewCount: scenario.startsWith('missing-seed-') ? 100 : 5,
|
||||
elapsedMs: 100, evidence: 'controlled DX observation' };
|
||||
},
|
||||
}));
|
||||
mock.module(${JSON.stringify(path.join(ROOT, 'test/helpers/plan-review-decisions.ts'))}, () => ({
|
||||
evaluatePlanReviewDecisions: () => { facts.judgeCalls++; save(); throw new Error('unexpected paid judge'); },
|
||||
}));
|
||||
await import(${JSON.stringify(path.join(ROOT, 'test/skill-e2e-plan-devex-finding-count.test.ts'))});
|
||||
`);
|
||||
try {
|
||||
const child = Bun.spawnSync([process.execPath, 'test', script], {
|
||||
cwd: ROOT, timeout: 10_000,
|
||||
env: { PATH: process.env.PATH ?? '', HOME: directory, TMPDIR: directory, TMP: directory, TEMP: directory,
|
||||
GIT_CONFIG_NOSYSTEM: '1', ...(process.env.SystemRoot ? { SystemRoot: process.env.SystemRoot } : {}) },
|
||||
});
|
||||
const output = child.stdout.toString() + child.stderr.toString();
|
||||
expect(child.signalCode ?? null, output).toBeNull();
|
||||
expect(fs.existsSync(facts), output).toBe(true);
|
||||
const observed = JSON.parse(fs.readFileSync(facts, 'utf8'));
|
||||
expect(observed.checked, output).toBe(true);
|
||||
expect(observed.runnerCalls, output).toBe(1);
|
||||
expect(observed.judgeCalls, output).toBe(0);
|
||||
expect(fs.existsSync(path.dirname(observed.planPath)), 'actual paid finally must remove its owned report directory').toBe(false);
|
||||
return { output, exitCode: child.exitCode, observed };
|
||||
} finally { fs.rmSync(directory, { recursive: true, force: true }); }
|
||||
}
|
||||
|
||||
test('the actual DX registration supplies its complete native fixture and preserves runner failure', () => {
|
||||
const result = runDxRegistration('throw');
|
||||
expect(result.exitCode, result.output).toBe(1);
|
||||
expect(result.output).toContain('controlled DX runner failure');
|
||||
expect(result.observed.reportCalls).toBe(0);
|
||||
});
|
||||
|
||||
for (const outcome of ['success', 'summary']) test(`DX registration accepts completed seed decisions and the owned final report: ${outcome}`, () => {
|
||||
const result = runDxRegistration(outcome);
|
||||
expect(result.exitCode, result.output).toBe(0);
|
||||
expect(result.observed.reportCalls).toBe(1);
|
||||
expect(result.observed.finalPlan).toContain('## GSTACK REVIEW REPORT');
|
||||
});
|
||||
|
||||
for (const scenario of [
|
||||
...Array.from({ length: 5 }, (_, index) => 'missing-seed-' + index),
|
||||
'missing-native', 'repeated-seed', 'batched', 'pending',
|
||||
]) test(`DX registration requires complete distinct native coverage: ${scenario}`, () => {
|
||||
const result = runDxRegistration(scenario);
|
||||
expect(result.exitCode, result.output).toBe(1);
|
||||
expect(result.output).toContain('SEEDED COVERAGE FAIL');
|
||||
expect(result.observed.reportCalls).toBe(0);
|
||||
});
|
||||
|
||||
for (const scenario of ['missing-report', 'trailing-report', 'timeout']) test(`DX registration rejects incomplete delivery: ${scenario}`, () => {
|
||||
const result = runDxRegistration(scenario);
|
||||
expect(result.exitCode, result.output).toBe(1);
|
||||
expect(result.output).toContain(scenario === 'timeout' ? 'outcome=timeout' : 'D19 FAIL');
|
||||
expect(result.observed.reportCalls).toBe(scenario === 'trailing-report' ? 1 : 0);
|
||||
});
|
||||
|
||||
test('materialized DX references have working local links without inventing completed launch work', () => {
|
||||
const fixture = path.join(ROOT, 'test/fixtures/devex-existing-sdk');
|
||||
const files = ['README.md', 'docs/getting-started.md', 'docs/feedback.md', 'docs/reference-v1.md'];
|
||||
for (const file of files) {
|
||||
const body = fs.readFileSync(path.join(fixture, file), 'utf8');
|
||||
for (const [, target] of body.matchAll(/\[[^\]]+\]\(([^)]+)\)/g)) {
|
||||
const [relative, anchor] = target!.split('#');
|
||||
const destination = path.resolve(path.dirname(path.join(fixture, file)), relative || path.basename(file));
|
||||
expect(destination.startsWith(fixture + path.sep)).toBe(true);
|
||||
const linked = fs.readFileSync(destination, 'utf8');
|
||||
if (anchor) {
|
||||
const headings = [...linked.matchAll(/^#+ (.+)$/gm)].map(match => match[1]!.toLowerCase()
|
||||
.replace(/[^\w\s-]/g, '').replace(/\s/g, '-'));
|
||||
expect(headings, target).toContain(anchor);
|
||||
}
|
||||
}
|
||||
}
|
||||
const readme = fs.readFileSync(path.join(fixture, 'README.md'), 'utf8');
|
||||
expect(readme).toContain('no selected primary developer persona or peer-DX study');
|
||||
expect(readme).toContain('No first-run duration has\nbeen measured');
|
||||
expect(readme).toContain('There is no skip');
|
||||
expect(readme).toContain('no interactive demo or designed aha sequence');
|
||||
expect(readme).toContain('one ordinary passing case; it has no staged regression');
|
||||
const reference = fs.readFileSync(path.join(fixture, 'docs/reference-v1.md'), 'utf8');
|
||||
const guide = fs.readFileSync(path.join(fixture, 'docs/getting-started.md'), 'utf8');
|
||||
const errorLink = /^Reference: (docs\/[^#]+)#([^\s]+)$/m.exec(guide);
|
||||
expect(errorLink).not.toBeNull();
|
||||
expect(fs.readFileSync(path.join(fixture, errorLink![1]!), 'utf8')).toBe(reference);
|
||||
const errorHeadings = [...reference.matchAll(/^### (.+)$/gm)].map(match => match[1]!.toLowerCase().replace(/\s/g, '-'));
|
||||
expect(errorHeadings).toContain(errorLink![2]!);
|
||||
|
||||
expect(reference).toContain('cannot interrupt arbitrary application code or cap requests made by a separate');
|
||||
expect(reference).toContain('those calls have not been executed against\nthe SDK here');
|
||||
expect(reference).toContain('Fixture checks execute the local application files and explicit\ncontract doubles');
|
||||
});
|
||||
|
||||
test('materialized DX error examples identify their cause, bound and reachable code reference', () => {
|
||||
const reference = fs.readFileSync(path.join(ROOT, 'test/fixtures/devex-existing-sdk/docs/reference-v1.md'), 'utf8');
|
||||
const expected = [
|
||||
{ heading: '### SDK E002', count: 1, causes: ['MetricTypeError'], values: ['cases[0]'] },
|
||||
{ heading: '### SDK E003', count: 2, causes: ['DeadlineExceeded', 'ManagedProviderCostLimit'],
|
||||
values: ['deadline_seconds=20', 'max_cost_usd=0.25'] },
|
||||
];
|
||||
for (const spec of expected) {
|
||||
const start = reference.indexOf(spec.heading);
|
||||
const next = reference.indexOf('\n##', start + spec.heading.length);
|
||||
const section = reference.slice(start, next < 0 ? undefined : next);
|
||||
const blocks = [...section.matchAll(/```text\n([\s\S]*?)\n```/g)].map(match => match[1]!);
|
||||
expect(blocks, spec.heading).toHaveLength(spec.count);
|
||||
for (const [index, block] of blocks.entries()) {
|
||||
const code = spec.heading.replace('### SDK ', 'SDK_');
|
||||
expect(block.split('\n')[0]).toStartWith(code + ':');
|
||||
expect(block).toContain('Cause: ' + spec.causes[index]);
|
||||
expect(block).toContain(spec.values[index]!);
|
||||
expect(block).toMatch(/^Next: .+/m);
|
||||
const anchor = spec.heading.replace('### ', '').toLowerCase().replace(/ /g, '-');
|
||||
expect(block).toContain('Reference: docs/reference-v1.md#' + anchor);
|
||||
}
|
||||
}
|
||||
});
|
||||
|
||||
// Materialize the documented files in a temp directory. These controls test
|
||||
// application examples against an explicit contract stub, never the absent SDK.
|
||||
const dxDocs = () => Object.fromEntries(['README.md', 'docs/getting-started.md', 'docs/reference-v1.md']
|
||||
.map(file => [file, fs.readFileSync(path.join(ROOT, 'test/fixtures/devex-existing-sdk', file), 'utf8')]));
|
||||
function dxBlock(body: string, after: string, language: string): string {
|
||||
const offset = body.indexOf(after);
|
||||
expect(offset, `missing documented example: ${after}`).toBeGreaterThanOrEqual(0);
|
||||
const match = new RegExp('```' + language + '\\n([\\s\\S]*?)\\n```').exec(body.slice(offset));
|
||||
expect(match, `missing ${language} block after ${after}`).not.toBeNull();
|
||||
return match![1]!;
|
||||
}
|
||||
function runDxDocumentationControl(code: string, payload: unknown) {
|
||||
const directory = fs.mkdtempSync(path.join(os.tmpdir(), 'devex-doc-control-'));
|
||||
try {
|
||||
const script = path.join(directory, 'control.py');
|
||||
fs.writeFileSync(script, code);
|
||||
const input = path.join(directory, 'input.json');
|
||||
fs.writeFileSync(input, JSON.stringify(payload));
|
||||
// Like bin/gstack-config, support both Python command names. Windows
|
||||
// installs normally expose python.exe; avoid preferring its python3 Store alias.
|
||||
const python = (process.platform === 'win32' ? ['python', 'python3'] : ['python3', 'python'])
|
||||
.map(command => Bun.which(command)).find((command): command is string => command !== null);
|
||||
if (!python) throw new Error('Python 3 is required for the DX documentation controls');
|
||||
const child = Bun.spawnSync([python, script, input], { cwd: directory, timeout: 10_000,
|
||||
stdin: 'ignore', stdout: 'pipe', stderr: 'pipe' });
|
||||
expect(child.signalCode ?? null, child.stderr.toString()).toBeNull();
|
||||
expect(child.exitCode, child.stderr.toString()).toBe(0);
|
||||
return child.stdout.toString();
|
||||
} finally { fs.rmSync(directory, { recursive: true, force: true }); }
|
||||
}
|
||||
|
||||
test('materialized DX success blocks print the documented structured fields without assuming SDK repr', () => {
|
||||
const docs = dxDocs();
|
||||
const first = dxBlock(docs['README.md']!, '## Quick start', 'python');
|
||||
expect(first).toBe(dxBlock(docs['docs/getting-started.md']!, '## Neutral first evaluation', 'python'));
|
||||
const examples = [
|
||||
{ code: first, expected: dxBlock(docs['README.md']!, 'Shown application output', 'text') },
|
||||
{ code: dxBlock(docs['docs/getting-started.md']!, '## Neutral first evaluation', 'python'),
|
||||
expected: dxBlock(docs['docs/getting-started.md']!, 'Shown application output', 'text') },
|
||||
{ code: dxBlock(docs['docs/getting-started.md']!, '## Caller-owned metric for free text', 'python'),
|
||||
expected: dxBlock(docs['docs/getting-started.md']!, 'Shown free-text application output', 'text') },
|
||||
];
|
||||
for (const { code } of examples) { expect(code).not.toContain('print(result)'); expect(code).toContain('result.cases'); }
|
||||
const output = runDxDocumentationControl(String.raw`
|
||||
import contextlib, io, json, sys, types
|
||||
with open(sys.argv[1], encoding='utf-8') as source:
|
||||
examples = json.load(source)
|
||||
# Deliberate assumed-contract double: not an implementation of eval-sdk.
|
||||
def evaluate(target, cases, metric):
|
||||
result = []
|
||||
for case in cases:
|
||||
actual = target(case['inputs'])
|
||||
result.append(types.SimpleNamespace(actual=actual, expected=case['expected'], score=metric(actual, case['expected'])))
|
||||
return types.SimpleNamespace(cases=result)
|
||||
stub = types.ModuleType('eval_sdk'); stub.evaluate = evaluate; sys.modules['eval_sdk'] = stub
|
||||
for example in examples:
|
||||
output = io.StringIO(); namespace = {}
|
||||
with contextlib.redirect_stdout(output): exec(example['code'], namespace)
|
||||
assert output.getvalue().strip() == example['expected']
|
||||
assert json.loads(output.getvalue())[0]['score'] == 1.0
|
||||
# Preserve the caller-owned metric's mismatching-prose behavior separately.
|
||||
assert namespace['text_metric']('red', 'green') == 0.0
|
||||
print('three documented outputs match the contract stub; no SDK executed')
|
||||
`, examples);
|
||||
expect(output).toContain('three documented outputs match the contract stub; no SDK executed');
|
||||
});
|
||||
|
||||
test('materialized DX application client bounds actual local process timeouts, retries and reservations', () => {
|
||||
const guide = dxDocs()['docs/getting-started.md']!;
|
||||
const client = dxBlock(guide, 'Save as `bounded_client.py`', 'python');
|
||||
const transport = dxBlock(guide, 'Save as `fixture_transport.py`', 'python');
|
||||
const usage = dxBlock(guide, 'Use the application client in the callable', 'python');
|
||||
expect(guide).toContain('verified upper bound');
|
||||
expect(guide).toContain('not refunded');
|
||||
expect(guide).toContain('does not prove that a remote provider cancelled');
|
||||
const output = runDxDocumentationControl(String.raw`
|
||||
import json, pathlib, subprocess, sys, time, types
|
||||
with open(sys.argv[1], encoding='utf-8') as source:
|
||||
payload = json.load(source)
|
||||
pathlib.Path('bounded_client.py').write_text(payload['client'])
|
||||
pathlib.Path('fixture_transport.py').write_text(payload['transport'])
|
||||
from bounded_client import BoundedClient
|
||||
# Observe the real handles; subprocess.run still owns timeout/kill/wait.
|
||||
original_popen = subprocess.Popen
|
||||
children = []
|
||||
def capture_popen(*args, **kwargs):
|
||||
child = original_popen(*args, **kwargs)
|
||||
children.append(child)
|
||||
return child
|
||||
subprocess.Popen = capture_popen
|
||||
command = [sys.executable, 'fixture_transport.py']
|
||||
client = BoundedClient(command, timeout_seconds=1, max_attempts=2, total_cents=4, attempt_cents=2)
|
||||
assert client({'enabled': True}) == {'ready': True}
|
||||
assert client.reserved_cents == 2
|
||||
assert client({'enabled': False}) == {'ready': False}
|
||||
assert client.reserved_cents == 4
|
||||
try: client({'enabled': True}); raise AssertionError('budget exceeded')
|
||||
except RuntimeError as e: assert 'spending limit' in str(e)
|
||||
assert client.reserved_cents == 4
|
||||
# Calls sharing this application client also share one reservation ceiling.
|
||||
from concurrent.futures import ThreadPoolExecutor
|
||||
client = BoundedClient(command, timeout_seconds=1, max_attempts=2, total_cents=4, attempt_cents=2)
|
||||
def concurrent_call(_):
|
||||
try: return client({'enabled': True})
|
||||
except RuntimeError as e:
|
||||
assert 'spending limit' in str(e); return None
|
||||
with ThreadPoolExecutor(max_workers=4) as pool: outputs = list(pool.map(concurrent_call, range(4)))
|
||||
assert outputs.count({'ready': True}) == 2 and outputs.count(None) == 2
|
||||
assert client.reserved_cents == 4
|
||||
# Real child failure/retry and real child timeout: no network or SDK involved.
|
||||
pathlib.Path('controlled_transport.py').write_text('''import sys, time
|
||||
mode = sys.argv[1]
|
||||
with open('attempts', 'a') as f: f.write('attempt\\n')
|
||||
if mode == 'stall': time.sleep(30)
|
||||
if mode == 'fail': sys.exit(75)
|
||||
''')
|
||||
for mode in ('fail', 'stall'):
|
||||
first_child = len(children)
|
||||
pathlib.Path('attempts').unlink(missing_ok=True)
|
||||
client = BoundedClient([sys.executable, 'controlled_transport.py', mode], timeout_seconds=0.2,
|
||||
max_attempts=2, total_cents=6, attempt_cents=2)
|
||||
started = time.monotonic()
|
||||
try: client({'enabled': True}); raise AssertionError('failed transport succeeded')
|
||||
except (RuntimeError, subprocess.TimeoutExpired): pass
|
||||
assert time.monotonic() - started < 3
|
||||
attempts = pathlib.Path('attempts').read_text().splitlines()
|
||||
owned_children = children[first_child:]
|
||||
assert len(attempts) == len(owned_children) == 2 and client.reserved_cents == 4
|
||||
for child in owned_children:
|
||||
assert child.poll() is not None, 'transport process leaked'
|
||||
assert child.wait(timeout=0) == child.returncode
|
||||
assert child.returncode != 0
|
||||
if mode == 'fail': assert child.returncode == 75
|
||||
# Insufficient reservation prevents even the retry from starting.
|
||||
pathlib.Path('attempts').unlink()
|
||||
client = BoundedClient([sys.executable, 'controlled_transport.py', 'fail'], timeout_seconds=1,
|
||||
max_attempts=2, total_cents=2, attempt_cents=2)
|
||||
try: client({'enabled': True}); raise AssertionError('budget exceeded')
|
||||
except RuntimeError as e: assert 'spending limit' in str(e)
|
||||
assert len(pathlib.Path('attempts').read_text().splitlines()) == 1
|
||||
assert client.reserved_cents == 2
|
||||
# The full shown usage sends independent limits to the SDK contract double.
|
||||
seen = []
|
||||
def evaluate(target, cases, metric, **options):
|
||||
seen.append(options)
|
||||
assert target(cases[0]['inputs']) == cases[0]['expected']
|
||||
return types.SimpleNamespace(cases=[])
|
||||
stub = types.ModuleType('eval_sdk'); stub.evaluate = evaluate; sys.modules['eval_sdk'] = stub
|
||||
exec(payload['usage'], {})
|
||||
assert seen == [{'deadline_seconds': 20, 'max_cost_usd': 0.25}]
|
||||
print('local timeout/retry/reservation bounds verified; no SDK/provider call')
|
||||
`, { client, transport, usage });
|
||||
expect(output).toContain('local timeout/retry/reservation bounds verified; no SDK/provider call');
|
||||
});
|
||||
|
||||
test('materialized DX CLI cases and import targets match the exact shown invocation in an offline contract double', () => {
|
||||
const reference = dxDocs()['docs/reference-v1.md']!;
|
||||
const cli = reference.slice(reference.indexOf('## CLI'), reference.indexOf('## Errors'));
|
||||
const payload = { app: dxBlock(cli, 'Save as `app.py`', 'python'), cases: dxBlock(cli, 'Save as `cases.json`', 'json'),
|
||||
command: dxBlock(cli, 'Run with the assumed SDK', 'bash') };
|
||||
expect(JSON.parse(payload.cases)).toEqual([{ inputs: { enabled: true }, expected: { ready: true } }]);
|
||||
const output = runDxDocumentationControl(String.raw`
|
||||
import argparse, importlib, json, pathlib, shlex, sys
|
||||
with open(sys.argv[1], encoding='utf-8') as source:
|
||||
payload = json.load(source)
|
||||
pathlib.Path('app.py').write_text(payload['app']); pathlib.Path('cases.json').write_text(payload['cases'])
|
||||
# Parse the documented command as an explicit contract double, not the absent CLI.
|
||||
args = shlex.split(payload['command']); assert args[:2] == ['eval-sdk', 'run']
|
||||
parser = argparse.ArgumentParser()
|
||||
for flag in ('target', 'cases', 'metric', 'deadline-seconds', 'max-cost-usd'): parser.add_argument('--' + flag, required=True)
|
||||
parser.add_argument('--no-input', action='store_true')
|
||||
options = parser.parse_args(args[2:])
|
||||
assert options.no_input and options.deadline_seconds == '20' and options.max_cost_usd == '0.25'
|
||||
def resolve(value):
|
||||
module, name = value.split(':'); return getattr(importlib.import_module(module), name)
|
||||
target, metric = resolve(options.target), resolve(options.metric)
|
||||
cases = json.loads(pathlib.Path(options.cases).read_text())
|
||||
assert isinstance(cases, list) and len(cases) == 1
|
||||
for case in cases:
|
||||
assert set(case) == {'inputs', 'expected'}
|
||||
assert metric(target(case['inputs']), case['expected']) == 1.0
|
||||
print('shown cases file, CLI arguments and import targets agree; no SDK executed')
|
||||
`, payload);
|
||||
expect(output).toContain('shown cases file, CLI arguments and import targets agree; no SDK executed');
|
||||
});
|
||||
Reference in new issue
Block a user