Repository navigation
test(e2e): the persistent-questions session ends with the token that opened it - #1144
Conversation
…opened it, so the DELETE no longer 404s when a second passes
|
@coderabbitai review |
|
No actionable comments were generated in the recent review. Walkthrough
Merge Risk:
|
Action not completed
|
Cherry-pick CreatedA cherry-pick PR to Please review and merge if this change should also be in If the cherry-pick is not needed, close the PR. |
Performance Test ResultsStatus: ✅ All tests passed Summary
Total: 101 tests across 21 projects 📊 View full report in workflow run Generated at: 2026-11-05T18:43:14.111Z |
What and why
The pre-release E2E gate of Publish Release failed on
persistent-questions.e2e.spec.ts› "keeps an answer for the next session when the waiting session ends" (run 42081062660). It also failed twice on #1141, and passed elsewhere.The test ends its MCP session with a DELETE, but it signed the DELETE with a token minted then, not the one the session was opened with. A streamable-HTTP session id is bound to the signature of the token that created it (
authSig).TestTokenFactorytokens have nojtiand second-resolutioniat/exp, so the two tokens were byte-identical only when both were minted in the same second. When the test crossed a second boundary,session:verifydidn't verify the session id and the DELETE answered 404.The test now opens the session with a token it keeps, and ends it with that token. The server is unchanged: a session id is honored only for the token that opened it.
Tests
persistent-questions.e2e.spec.tspasses locally (thefrontmcp inboxcase needs the CLI build and runs in CI).Summary by CodeRabbit