test(dashboard): verify rendering across many projects - #715
test(dashboard): verify rendering across many projects#715ScriptedAlchemy wants to merge 5 commits into
Conversation
perf(observation) landed an import of tracedecay_runtime_core::background_cpu, but neither the module file nor its declaration was added, so the pushed integration branch does not compile: every branch cut from it fails on an unresolved import. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The batching commit added ObservationApplicationError::BatchWorkerStopped but left the ingest failure classifier's match on that enum without an arm for it, so tracedecay-sessions does not compile on this branch. Classified as retryable/unavailable rather than contended or permanent: a worker that went away before completing never reached a verdict about the observation, so the same input succeeds once a worker is running again.
live-sweep.ts proves each workspace renders; it cannot prove anything is behind that render. A workspace whose read failed still draws its chrome and so passes, which is how the Code workspace passed while /api/plugins/graph/overview was answering 200 with a null payload. live-multiproject.ts asserts the payload instead: it walks /api/projects and requires every enrolled project to serve a graph overview with a non-zero node count, reporting node/edge/file/language counts per project. When a read fails it prints coverage.omission_reasons, which is where the envelope actually carries the cause. Also covers the decode step that makes those reasons reachable: an unavailable envelope arrives as domain_state unknown with a null payload, and the existing test asserted only the state, not that the reason survived.
…plan' into HEAD # Conflicts: # crates/tracedecay-sessions/src/runtime/ingest/failure.rs
…plan-reopened' into HEAD
|
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: fd38ab903b
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| } | ||
|
|
||
| #[test] | ||
| fn a_stopped_batch_worker_is_retryable_and_unavailable() { |
There was a problem hiding this comment.
When this commit is relied on to verify rendering across many projects, this test cannot detect that regression: it only constructs BatchWorkerStopped and checks its pre-existing failure classification, without enrolling projects, opening project servers, requesting dashboard data, or rendering UI. It therefore remains green if the described 12-project journey still serves zero graphs; add a production-path many-project test and retain this classifier test separately if needed.
AGENTS.md reference: AGENTS.md:L9-L12
Useful? React with 👍 / 👎.
|
Closing as selectively superseded. The multi-project dashboard sweep is already in PR 707 as 0fc6943. The remaining classifier unit test does not exercise the claimed many-project dashboard journey; the cached-server capacity finding remains a separate production issue. |
1 participant
Reopened against
codex/tracedecay-total-redesign-plan-reopened. Verification, not a dashboard change — investigation showed the dashboard is not the blocker, so patching around a broken index would have been wrong.12 projects enrolled, 0 served a graph. Enrolment past the old cap works, but a third cap (
MAX_CACHED_PROJECT_SERVERS = 8) binds and then retries ~2/sec forever with no backoff, and no code-index generation ever sealed even for one small project on a clean run — every reconcile failure wasthe read port was cancelled, accruing during a silent period.