mirror of
https://github.com/diegosouzapw/OmniRoute.git
synced 2026-09-14 10:52:17 +03:00
65e81158ab
Landed with the design call resolved per the owner's pick — **option 1**: the synced store is now endpoint-agnostic (persistDiscoveredModels and managedModelImport no longer drop non-chat models at write time), and chat selectability moved to read time (auto-pool expansion in autoStrategy applies filterChatSelectableModels; the models-route projection already had its chatOnly filter). Your discovery test now passes end-to-end (3/3): /api/show capabilities persist per connection and image/embedding requests route through the advertising host. Reconciliation notes: conflicted areas merged onto the current tip (adobe discovery import, requestedModel preflight signature, resolvedProvider fast-path coexists with the synced-route override — explicit resolution wins); carried base-red drains (#10055 memoization, #11071 test variants) dropped as already-landed; the managed-model-import exclusion test was propagated to the new contract (image/video models persist; the read filter still hides them from chat pickers — pinned by a new assertion). Full battery: 205/206 focused (the one red is a confirmed periodic-timer timing flake on the loaded devbox — 20/20 isolated), autoCombo vitest 30/30, combo suites 46/46, gates + typecheck clean. Thank you @yourspraveen — the capability probe + routing design was right; it just needed the store contract opened up. Fixes #11087.
66 lines
3.2 KiB
TypeScript
66 lines
3.2 KiB
TypeScript
/**
|
|
* Guard for the test-masking gate's release-scale escape hatch (gap 6).
|
|
*
|
|
* The gate diffs `base...HEAD` — three dots, so already merge-base relative, which is right
|
|
* for a normal PR. It is not the base choice that hurts. Releases SQUASH-merge into `main`, so
|
|
* a release PR's merge-base is the PREVIOUS cycle's fork point and the diff legitimately spans
|
|
* the entire cycle. In the v3.8.49 run that was ~1277 changed test files, each costing a
|
|
* `git show base:file` process plus a full regex pass: the check ran twice without finishing,
|
|
* >30 min pegged on a single core, with the release waiting on it.
|
|
*
|
|
* Measured while writing this, and worth recording because it refutes the obvious diagnosis:
|
|
*
|
|
* tracked test files ................. 3977
|
|
* absolute tautology scan ............ ~1 s (runs unconditionally, always)
|
|
* diff vs release branch ............. 0 changed test files, 0 s
|
|
* diff vs main (today) ............... 3 changed test files, 0 s
|
|
*
|
|
* It cannot be reproduced today at all, because `main` has since received the v3.8.49 squash
|
|
* and the merge-base is recent. The pathology only appears DURING a release, in the window
|
|
* before `main` receives it — the same squash-merge topology behind gap 4.
|
|
*
|
|
* So the threshold is what these tests pin. An off-by-one here either blocks a release or
|
|
* silently disables the check on a large-but-legitimate PR, and the second failure mode is the
|
|
* dangerous one.
|
|
*/
|
|
import test from "node:test";
|
|
import assert from "node:assert/strict";
|
|
|
|
// @ts-expect-error — plain .mjs gate script, no type declarations by design
|
|
import { shouldSkipDiffSubchecks } from "../../scripts/check/check-test-masking.mjs";
|
|
|
|
test("a normal PR is always analyzed", () => {
|
|
assert.equal(shouldSkipDiffSubchecks(0, 300), false);
|
|
assert.equal(shouldSkipDiffSubchecks(1, 300), false);
|
|
assert.equal(shouldSkipDiffSubchecks(42, 300), false);
|
|
});
|
|
|
|
test("the boundary is exclusive — exactly at the cap is still analyzed", () => {
|
|
assert.equal(shouldSkipDiffSubchecks(300, 300), false, "at the cap: analyze");
|
|
assert.equal(shouldSkipDiffSubchecks(301, 300), true, "one past the cap: skip");
|
|
});
|
|
|
|
test("release scale skips (the measured v3.8.49 case)", () => {
|
|
assert.equal(shouldSkipDiffSubchecks(1277, 300), true);
|
|
});
|
|
|
|
test("a cap of zero or below disables the skip — always analyze", () => {
|
|
// The documented escape hatch for anyone who wants the full pass regardless of size.
|
|
assert.equal(shouldSkipDiffSubchecks(999999, 0), false);
|
|
assert.equal(shouldSkipDiffSubchecks(999999, -1), false);
|
|
});
|
|
|
|
test("a huge explicit cap forces full analysis", () => {
|
|
// TEST_MASKING_MAX_CHANGED_TESTS=999999, the override the skip message advertises.
|
|
assert.equal(shouldSkipDiffSubchecks(1277, 999999), false);
|
|
});
|
|
|
|
test("garbage inputs never skip — an unparseable count must not disable the gate", () => {
|
|
// The safe direction: if we cannot tell how big the diff is, do the work.
|
|
assert.equal(shouldSkipDiffSubchecks(NaN, 300), false);
|
|
assert.equal(shouldSkipDiffSubchecks(-5, 300), false);
|
|
assert.equal(shouldSkipDiffSubchecks(500, NaN), false);
|
|
assert.equal(shouldSkipDiffSubchecks(500, undefined), false);
|
|
assert.equal(shouldSkipDiffSubchecks(undefined, 300), false);
|
|
});
|