Repository navigation
test(cloud-agent-next): share one e2e scenario definition across both profiles - #6476
Merged
Merged
Conversation
![kilo-code-bot[bot]](https://avatars.githubusercontent.com/in/2193792?s=60&v=4){kind=small}
Contributor
Code Review SummaryStatus: No Issues Found | Recommendation: Merge Files Reviewed (7 files)
Previous Review Summaries (3 snapshots, latest commit 89582c3)Current summary above is authoritative. Previous snapshots are kept for context only. Previous review (commit 89582c3)Status: 7 Issues Found | Recommendation: Address before merge Executive SummaryThe Overview
Issue Details (click to expand)WARNING
SUGGESTION
Files Reviewed (50 files)
Fix these issues in Kilo Cloud Previous review (commit ea5b794)Status: No Issues Found | Recommendation: Merge Executive SummaryThe incremental commit resolving the prior shared-E2E harness remarks was reviewed at Files Reviewed (7 files)
Previous review (commit 4137be7)Status: 10 Issues Found | Recommendation: Address before merge Overview
Fix these issues in Kilo Cloud Issue Details (click to expand)WARNING
SUGGESTION
Files Reviewed (50 files)
Reviewed by deepseek-v4.1-flash · Input: 0 · Output: 0 · Cached: 0 Review guidance: REVIEW.md from base branch |
…n use Add worktree-chat, worktree-multi-chat, long-conversation and leave-and-return to the shared scenario set, so one implementation runs under both the local Docker profile and the deployed HTTP profile using only public tRPC, streams, fake-LLM /test/* and the sessionSandbox capability. No Docker file/process inspection, worker logs or Postgres. - public-surface-support and scenarios-shared-runtime: shared deadline, allocation sampling, owned-session registry with late-create cleanup, cumulative sendTurn budget, extracted from the local-only modules - removal-aware collectChildMessageText - resolve shared-only scenario names in local dispatch - lint allowlist, README/SESSION-CONTINUITY/deploy docs, e2e workflow timeout 45 -> 120 minutes
… profiles Replace the local-only and shared scenario registries with a single SHARED_SCENARIOS definition that runs on the Docker and deployed public profiles. Scenarios declare the capabilities they need; unsupported ones exit as expected-unsupported instead of being duplicated. - add file:write/file:seed/file:read fake-LLM directives with a single directive parser, op-aware tags and single-use completion - add gates (local-only) and sandboxFaults (local-only, Docker-backed) capabilities and derive the expected-unsupported exit - re-scope worktree-multi-chat onto shared-checkout write/overwrite, question ownership, answer isolation and delete survival - absorb real file turns into long-conversation; add question-idle-resume, large-stream, concurrent-chats, external-kill, kill-mid-flight and the two wrapper-freeze scenarios - delete lifecycle-continuity/lifecycle-file-state, owned-gates, worktree-support and the redundant followup/waiters-clean/gate-0/ cold-resume/warm-cold-cycles/recover-same-session/feed-stale-recovery scenarios - derive the deployed workflow timeout from the shared registry budgets
The two shared-e2e commits were authored on a branch that never ran CI, so 22 harness files were committed without oxfmt reflow and the format-check job failed. Reflow only: no logic, identifiers or string literals change.
- runConcurrentChats: own the boot stream before the echo/allocation assertions, track and close the recovery stream, and fold each hold's events exactly once in `finally` (a hold stream stays open and appends while the other hold's recovery runs) - worktree-multi-chat: register the four file turns' messageIds so chat-content isolation actually covers them - createOwnedSessionRegistry: `register` upgrades an `undefined` placeholder but never clobbers an already-defined kiloSessionId, and the interface comment now states that contract; regression tests added for both the upgrade and the defined-to-undefined clobber case - docs: drop the unsupported "249 min derived from the registry" claim from both e2e READMEs and the deployed workflow comment, correct the public-surface ceilings to 10 + 25 + 12 + 30, the long-conversation turn count to twelve, the `hot=12/12` artifact, the false claim that the long scenarios are excluded from DEFAULT_MATRIX, and the "no interrupt" claim for worktree-multi-chat
eshurakov
force-pushed
the
eshurakov/grand-swan
branch
from
September 21, 2026 14:19
ea5b794 to
89582c3
Compare
…nostics Address the Kilobot review on the shared e2e harness: - release the parked `kill-mid-flight` gate in the scenario `finally` with a bounded signal, so an early failure before the container kill cannot leave the fake's gate parked for the rest of the server's lifetime - fold each stream's buffered events into the failure diagnostic exactly once through one `foldStream` owner, instead of re-appending buffers that were already folded; events of a stream opened but not yet folded are still kept - count `message.part.updated` events in `collectCorrelatedProgress` even when no assistant child id was established, so the diagnostic distinguishes "no events for this turn" from "events but no correlated child" (the pass rule stays `correlated > 0`) - name `interrupt-then-continue` in the worktree-enrollment list and state the scenario capability rule that matches the registry
pandemicsyn
approved these changes
Sep 21, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Third PR in the deployed-E2E stack. Base: #6284.
Replaces the separate local-only and shared scenario registries with a single
SHARED_SCENARIOSdefinition that runs on the Docker and deployed publicprofiles. Scenarios declare the capabilities they need; an unsupported profile
reports
unsupportedwith the missing capability names instead of running aweakened duplicate.
file:write/file:seed/file:readfake-LLM directives with onedirective parser, op-aware tags and single-use completion (409 on replay)
gatesandsandboxFaults(local-only, Docker-backed) capabilities andderive the expected-unsupported exit
worktree-multi-chatonto shared-checkout write/overwrite, questionownership, answer isolation and delete survival
long-conversation; addquestion-idle-resume,large-stream,concurrent-chats,external-kill,kill-mid-flightand thetwo wrapper-freeze scenarios
lifecycle-continuity/lifecycle-file-state,owned-gates,worktree-supportand the redundantfollowup/waiters-clean/gate-0/cold-resume/warm-cold-cycles/recover-same-session/feed-stale-recoveryscenarios
Deployed gaps are expected and explicit: the four
sandboxFaultsscenarios areunsupportedon the deployed profile;auth-rejectruns only deployed.Verification: lint 0/0; typecheck clean; unit 7236 passed | 3 skipped;
fake-llm 16 passed; integration 700/701 (the one failure is a 5s-timeout load
flakes in
test/integration/sandbox-control.test.ts, untouched by this branch,passing in isolation). Full local E2E matrix run separately.
Test-only: no
src/orwrapper/changes.