You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Commit 3f171a0
Browse filesBrowse the repository at this point in the historyBrowse files
test(everything): mark the tests that pin known bugs
These suites are characterization tests: they assert current behavior so
the SDK v2 migration can show it changed nothing. Some of them pin
behavior that is wrong. Each of those now carries a uniform marker
immediately above its test declaration:
// KNOWN BUG #<issue>: pins current (wrong) behavior; the fix changes this assertion.
// KNOWN BUG (no issue): <what is wrong>; the fix changes this assertion.
so `git grep "KNOWN BUG"` lists exactly the tests a bug fix is expected
to change. Regression guards for fixed bugs, correct SDK-mandated
behavior (the #3267 priming event) and pins of intended design carry no
marker. Comments only; no assertion changes.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Signed-off-by: cliffhall <[email protected]>
expect(texts[0]).toBe("⚠️ User cancelled the elicitation dialog.");
389
389
});
390
390
391
+
// KNOWN BUG (no issue): trigger-elicitation-request is listed for a URL-only elicitation client, whose call then fails; the fix changes this assertion.
391
392
it("is listed for a URL-only client, which then cannot answer its form request",async()=>{
392
393
// Characterization: the tool is gated on `elicitation` alone, so a
393
394
// client that declared only URL mode sees it and the call fails.
// KNOWN BUG (no issue): answers 400, not the 404 the spec requires, for an unknown or terminated session ID, and the error omits the request id; the fix changes this assertion.
255
258
it("answers a POST for an unknown session with a 400 that omits the request id",async()=>{
256
259
// Characterization: no body parser runs before this check, so
257
260
// `req.body` is undefined and the error carries no `id`.
// KNOWN BUG (no issue): trigger-sampling-request-async drops the status message of a task that is already finished when created; the fix changes this assertion.
425
426
it("does not poll a client task that is already finished when it is created",async()=>{
426
427
// Characterization: the status message is only read from polls, so a
427
428
// task that fails before it is returned is reported with "No message".
// KNOWN BUG (no issue): trigger-elicitation-request-async polls for longer than the 10-minute TTL it requests for its task; the fix changes this assertion.
588
590
it("fails on the last poll instead when the client honors the 10-minute TTL",async()=>{
589
591
// Characterization: 600 one-second polls outlast the 600000 ms TTL the
590
592
// tool asks for, so a client that expires the task on time answers the
0 commit comments