Files
msd-core/tests/trae-imperative-reference.test.cjs
Tom Boucher 83a26ed1dc fix(#2939): honor the declared depth budget in shouldFlattenDispatch (#3063)
* test(#2939): prove shouldFlattenDispatch ignores the depth budget

Failing-first regression for #2939. shouldFlattenDispatch checks only
background+backgroundDispatch, never nested/subagentToolkit/maxDepth, so a
maxDepth:1 descriptor (no room for a bg orchestrator plus a leaf) is told it
may background. Row 1 (codex-like, maxDepth:1) asserts true (flatten) and
fails today; rows 2/3 guard the unchanged depth-sufficient cases.

* fix(#2939): honor the declared depth budget in shouldFlattenDispatch

shouldFlattenDispatch checked only background+backgroundDispatch, never
nested/subagentToolkit/maxDepth, so a maxDepth:1 descriptor (no room for a
backgrounded orchestrator plus a delegated leaf) was told it may background —
producing a depth-2 tree (Codex MultiAgent V2) the declared contract cannot
support.

canBackground now ALSO requires nested:true + subagentToolkit:"full" + a depth
budget > 1 (or unbounded -1), reusing the exact predicate shape from
bin/install.js _normalizeDispatchCallSpan and matching degradationFor's
treatment of maxDepth===1 as flat. Non-finite/missing maxDepth fails closed to
flatten. Correct the two existing pins that asserted the buggy output (bare
{bg,bgDispatch} now fail-closes on missing depth; the codex-like maxDepth:1 pin
flips to flatten) and add a maxDepth:2 negative-space row.

* fix(#2939): propagate depth-aware flatten to all pinned descriptors + tests

The isolated adversarial review found the depth-aware predicate reclassifies
codex/kimi/kimi-code (previously background-eligible under the two-field rule)
to flatten — the correct behavior, since each lacks what a backgrounded nesting
orchestrator needs:

  - codex: maxDepth:1 (no room for a depth-2 leaf)
  - kimi: nested:false (cannot host a nesting orchestrator)
  - kimi-code: subagentToolkit:'built-in-only' (cannot delegate to full subagents)

Only cursor (maxDepth:2) remains background-eligible. Update the three test
files that pinned the old contract (host-integration-descriptors EXPECTED_FLATTEN,
kimi-upgrades UPGRADE 2, trae-imperative-reference), and align the unbounded
convention to maxDepth < 0 (matching degradationFor/negotiateHostCapabilities)
with an accurate docstring noting the deliberate nested-check addition over
_normalizeDispatchCallSpan.

* fix(#2939): update dispatch-should-flatten CLI query pins for codex

The depth-aware rule (a0ad0f680) reclassifies codex (maxDepth:1) to flatten, but
command-routing-hub.test.cjs exercises the contract through the CLI query route
(runGsdTools query dispatch-should-flatten), not a direct shouldFlattenDispatch
call — so neither the reviewer's caller-search nor a grep for the symbol found
it; only the full gsd-test matrix did. Update the codex query assertions to
shouldFlatten=true (maxDepth:1 insufficient), preserving cursor (maxDepth:2 →
false) and the backgroundDispatch:true descriptor field.

* chore(#2939): add changeset fragment

pr:0 placeholder backfilled with the real PR number once the PR exists.

* fix(#2939): rephrase changeset for product-name-purity + opencode flatten pin

Two failures from the full gsd-test matrix on the prior sha:

1. product-name-purity: changeset fragments must not include parenthetical product
   descriptions (they render verbatim into CHANGELOG.md). 'Codex (and kimi/kimi-code)'
   tripped it — rephrase to lead with the behavior, naming runtimes inline without
   the parenthetical. lint:ci changeset-lint does not catch this; only the test does.

2. opencode-imperative-reference: the #2087-retraction pin flipped only the two
   background booleans and asserted shouldFlatten:false. Under #2939 that is no
   longer sufficient (opencode lacks nested + full toolkit + depth budget), so the
   retracted axes now correctly flatten — update the pin to true with rationale.

* chore(#2939): backfill changeset PR number 3063

---------

Co-authored-by: sim <sim@local>
2026-08-04 20:33:50 -04:00

158 lines
7.9 KiB
JavaScript

// allow-test-rule: structural-regression-guard — AC2 requires asserting no `runtime === 'trae'` string-equality branch remains in bin/install.js, src/install-engine.cts, and src/runtime-artifact-conversion.cts — the descriptor-migration contract is a property of the source text, so a source-grep is the only faithful check (#2094)
'use strict';
/**
* trae imperative reference host — ADR-1239 Phase D / #2094 (EoS/trae).
*
* Proves Trae IDE is driven through the PUBLIC Host-Integration Interface
* (the imperative adapter), that its negotiated axes classify + negotiate
* correctly, that negotiation fails CLOSED on a corrupted descriptor, and
* that the migration retired the hardcoded `runtime === 'trae'` string-
* equality branches across the install engine and artifact conversion
* modules (folded into descriptor-driven `runtime.hostBehaviors`).
*
* Trae has NO hook surface at all (`hooksSurface: "none"`,
* `extendedHookEvents: []`, `installSurface: "profile-marker-only"`) — its
* `hookBus` axis is `'engine'` (VSCode-fork extension-host lifecycle, not a
* GSD-managed hook dialect), so unlike qwen/kilo this file does not assert an
* extendedHookEvents surface. Four of Trae's six dispatch sub-axes
* (`nested`, `maxDepth`, `subagentToolkit`, `backgroundDispatch`) are
* `'undocumented'` — no authoritative Trae doc states them — which is what
* drives the fail-closed `shouldFlattenDispatch` assertion below. The real
* upgrade (SOLO stage/trigger metadata on emitted skills) is covered in
* tests/trae-upgrades.test.cjs.
*/
const { test } = require('node:test');
const assert = require('node:assert/strict');
const fs = require('node:fs');
const path = require('node:path');
const { createImperativeAdapter } = require('../gsd-core/bin/lib/adapter-imperative.cjs');
const {
profileOf,
negotiateHostCapabilities,
shouldFlattenDispatch,
PROFILE_BASELINES,
UNDOCUMENTED,
} = require('../gsd-core/bin/lib/host-integration.cjs');
const TRAE_CAP = JSON.parse(
fs.readFileSync(path.join(__dirname, '..', 'capabilities', 'trae', 'capability.json'), 'utf8'),
);
const TRAE_AXES = TRAE_CAP.runtime.hostIntegration;
// -- AC2: driven through the public interface (imperative adapter) -----------
test('createImperativeAdapter classifies trae as imperative + composes the registry', () => {
const adapter = createImperativeAdapter({ runtime: 'trae' });
assert.equal(adapter.kind, 'imperative');
assert.equal(adapter.runtime, 'trae');
assert.ok(adapter.registry && typeof adapter.registry === 'object');
assert.equal(typeof adapter.install, 'function');
assert.equal(typeof adapter.uninstall, 'function');
});
test('trae axes classify as the programmatic-cli reference profile', () => {
// Confirmed via `node -e` against the real descriptor before asserting:
// profileOf(TRAE_AXES) === 'programmatic-cli' (embeddingMode: 'imperative').
assert.equal(profileOf(TRAE_AXES), 'programmatic-cli');
});
// -- AC3: all axes populated + validated -------------------------------------
test('trae descriptor declares all 8 axes + 6 dispatch sub-axes with exact values', () => {
assert.equal(TRAE_AXES.embeddingMode, 'imperative');
assert.equal(TRAE_AXES.commandSurface, 'slash-file');
assert.equal(TRAE_AXES.modelMode, 'passive');
assert.equal(TRAE_AXES.hookBus, 'engine');
assert.equal(TRAE_AXES.stateIO, 'filesystem');
assert.equal(TRAE_AXES.transport, 'mcp');
assert.equal(TRAE_AXES.runtime, 'node');
const d = TRAE_AXES.dispatch;
assert.equal(d.namedDispatch, true);
assert.equal(d.nested, 'undocumented');
assert.equal(d.maxDepth, 'undocumented');
assert.equal(d.background, true);
assert.equal(d.subagentToolkit, 'undocumented');
assert.equal(d.backgroundDispatch, 'undocumented');
});
// -- AC5: negotiation fails CLOSED on a corrupted descriptor ------------------
test('negotiateHostCapabilities never throws for trae, even fully corrupted', () => {
assert.doesNotThrow(() => negotiateHostCapabilities({}));
assert.doesNotThrow(() => negotiateHostCapabilities({ ...TRAE_AXES, embeddingMode: UNDOCUMENTED }));
assert.doesNotThrow(() => negotiateHostCapabilities({ ...TRAE_AXES, embeddingMode: 'future-unknown' }));
});
test('AC-SPECIFIC: trae real dispatch axes fail CLOSED to inline (shouldFlattenDispatch === true)', () => {
// Confirmed via `node -e` against the real descriptor before asserting:
// shouldFlattenDispatch(TRAE_AXES.dispatch) === true.
//
// shouldFlattenDispatch only permits backgrounding when BOTH `background`
// AND `backgroundDispatch` are explicitly `true` (src/host-integration.cts
// shouldFlattenDispatch, `canBackground = background === true &&
// backgroundDispatch === true`). Trae's `background` is `true` but
// `backgroundDispatch` is `'undocumented'` (no authoritative Trae doc states
// whether a spawned agent can itself be backgrounded) — so canBackground is
// false and the orchestrator must run inline. Pinned here so a future
// doc-sourcing pass that fills in `backgroundDispatch` can't silently flip
// this fail-closed default without a deliberate test update.
assert.equal(shouldFlattenDispatch(TRAE_AXES.dispatch), true);
// #2939: flipping ONLY backgroundDispatch to true is NO LONGER sufficient to
// background. trae's axes declare nested:'undocumented' (not true) and
// subagentToolkit:'undocumented' (not 'full'), so even with backgroundDispatch
// true the depth/nesting gate (canNest) fails closed to flatten. A future
// doc-sourcing pass must also establish nested:true + subagentToolkit:'full' +
// a sufficient maxDepth before trae may background.
assert.equal(shouldFlattenDispatch({ ...TRAE_AXES.dispatch, backgroundDispatch: true }), true);
});
test('a partial/empty trae descriptor degrades to the safe floor, not the programmatic-cli baseline', () => {
const result = negotiateHostCapabilities({});
assert.equal(result.effective.embeddingMode, 'declarative', 'omitted embeddingMode degrades closed');
assert.equal(result.effective.hookBus, 'none');
assert.notDeepEqual(result.effective, PROFILE_BASELINES['programmatic-cli']);
assert.ok(result.warnings.length > 0);
});
// -- AC2: the folded-in behaviors ---------------------------------------------
test('trae descriptor declares runtime.hostBehaviors (the folded-in behaviors)', () => {
const hb = TRAE_CAP.runtime.hostBehaviors;
assert.ok(hb && typeof hb === 'object');
assert.equal(hb.skipSharedHooksInstall, true);
assert.equal(hb.soloStageMetadata, 'workflow');
});
// -- AC2: the hardcoded branches are retired across all folded modules -------
test('no `runtime === "trae"` string-equality branch remains in the descriptor-migrated modules (AC2)', () => {
// NOTE: this deliberately does NOT grep for `isTrae` — `isTrae` legitimately
// remains in bin/install.js as a destructured `runtimeFlags(runtime)`
// binding used ONLY by the agents-converter dispatch chain (`else if
// (isTrae) { content = convertClaudeAgentToTraeAgent(content); }`). That
// cross-runtime agents-converter dispatch is out of scope for #2094 (a
// separate migration tracked elsewhere) — trae stays in RUNTIME_FLAG_IDS by
// design until that follow-up lands. Only a `runtime === 'trae'` /
// `runtime !== 'trae'` STRING-EQUALITY comparison is a regression here.
const strip = (src) => src
.replace(/\/\*[\s\S]*?\*\//g, '')
.replace(/\/\/[^\r\n]*/g, '')
.replace(/`[^`]*`/g, '');
const repoRoot = path.join(__dirname, '..');
const files = [
path.join(repoRoot, 'bin', 'install.js'),
path.join(repoRoot, 'src', 'install-engine.cts'),
path.join(repoRoot, 'src', 'runtime-artifact-conversion.cts'),
];
for (const file of files) {
const src = fs.readFileSync(file, 'utf8');
const offenders = strip(src).match(/runtime\s*[!=]==\s*'trae'/g) || [];
assert.deepEqual(offenders, [],
`AC2: no hardcoded runtime==='trae' branch may remain in ${path.relative(repoRoot, file)}; found: ${offenders.join(', ')}`);
}
});