chore(vendor): split remaining generated and data churn out of PR #707 - #735
Conversation
Byte-for-byte copies from PR #707 HEAD (12be86f), inert on master: - benchmark_data/ (new): benchmark harness, result JSONs, goldens - mockups/ (new): design mockup HTML prototypes - plugin/cursor-native-extension/ (new): self-contained extension incl. the generated rslib bundle embedded/extension.js - dashboard/codegen/schemas/dashboard-contracts.schema.json (new): generated contracts schema (pure JSON, nothing on master compiles it) 130 files, +100826 / -0. No Cargo.toml/Cargo.lock changes. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01A2BarcJk3iuv77aJQwHvnx
|
![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: 99eda612df
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".
| validation_command = ( | ||
| os.fspath(REPOSITORY_ROOT / "scripts" / "require-exact-test.sh"), | ||
| os.fspath(test_binary), | ||
| *scenario_arguments, | ||
| ) |
There was a problem hiding this comment.
When incident --authority-test is used, this command unconditionally executes scripts/require-exact-test.sh, but that script does not exist anywhere in the reviewed tree. Consequently every diagnostic-authority capture exits with an ENOENT before collecting a sample, as reproduced by the added capture-dispatch tests; include the script or perform the validation inside the harness.
Useful? React with 👍 / 👎.
| "cursor": Path( | ||
| ".cursor/projects/fixture-runtime-project/" | ||
| "agent-transcripts/fixture-session.jsonl" | ||
| ), |
There was a problem hiding this comment.
The checked-in provider tree contains the Codex and Claude transcript files but not this declared Cursor path. prepare_fixture_snapshot() nevertheless returns it as provider_files["cursor"] without validation, so Cursor/session benchmark runs receive an empty provider fixture and can produce incomplete measurements; the fixture tests also fail while attempting to read this path.
Useful? React with 👍 / 👎.
| const document = vscode.workspace.textDocuments.find( | ||
| (candidate) => candidate.uri.toString() === uri.toString(), | ||
| ); | ||
| if (document === undefined) { | ||
| continue; |
There was a problem hiding this comment.
When diagnostics are cleared for a URI that is no longer present in workspace.textDocuments, this branch drops the event instead of sending an empty diagnostics payload. The bridge therefore never communicates removal of diagnostics previously sent for that URI, allowing the TraceDecay projection to retain stale errors after a document closes or a provider clears diagnostics for an unopened file.
Useful? React with 👍 / 👎.
| invoking Cargo: | ||
|
|
||
| ```text | ||
| scripts/run-runtime-performance.sh graph-capture \ |
There was a problem hiding this comment.
Following either documented graph-capture command fails immediately because scripts/run-runtime-performance.sh is absent from the repository. The added adversarial tests also explicitly require this wrapper and fail before checking argument or environment isolation, so the benchmark harness is not usable through its advertised entry point.
Useful? React with 👍 / 👎.
| ROOT = Path(__file__).resolve().parents[3] | ||
| RUNNER = ROOT / "benchmark_data" / "runtime" / "run.py" | ||
| WRAPPER = ROOT / "scripts" / "run-runtime-performance.sh" | ||
| DOCUMENTATION = ROOT / "docs" / "development" / "runtime-performance.md" |
There was a problem hiding this comment.
The new test suite unconditionally reads docs/development/runtime-performance.md, but that file is not present anywhere in the reviewed tree. A normal unittest discover therefore raises FileNotFoundError, making the committed runtime benchmark tests impossible to pass until the documentation is included or the assertion is removed.
Useful? React with 👍 / 👎.
| env=os.environ, | ||
| log_dir=root / "host-logs", | ||
| timeout=0.2, | ||
| termination_grace=0.05, | ||
| check=False, |
There was a problem hiding this comment.
On runners where starting the nested Python process takes most of this 200 ms timeout, the process group receives SIGTERM before verbose_child.py installs its ignore handler. Both processes then exit on TERM, kill_sent remains false, and this test fails consistently instead of exercising the intended forced-kill path; wait for a readiness signal from the child before starting the timeout or asserting SIGKILL.
Useful? React with 👍 / 👎.
Absorbs the merged churn-split PRs so the merge-base advances and the byte-identical files (benchmark_data/, mockups/, plugin/cursor-native-extension/, dashboard contracts schema) leave #707's visible diff. Tree is identical to the previous branch head — no content change. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01A2BarcJk3iuv77aJQwHvnx
1 participant
Second churn split for #707 (follow-up to #734)
Carries the remaining generated/data churn from #707 (
codex/tracedecay-total-redesign-plan-reopened) so its GitHub diff shrinks to the real product change. All contents are byte-for-byte copies from #707 HEAD (12be86f52d), pure additions, inert on master.dashboard/codegen/schemas/dashboard-contracts.schema.jsonplugin/cursor-native-extension/embedded/extension.js(+26,918) andpackage-lock.json; ~1.2k lines of authored TS source (src/) land here too, review them in this PRbenchmark_data/mockups/Total: 130 files, +100,826 / −0. No
Cargo.toml/Cargo.lockchanges.Standalone-safety notes
plugin/**: the strict Claude validation and manifest schema jobs already pass on feat: consolidate TraceDecay V2 PR8-PR13 delivery #707 with this exact directory content; the master-side Rust plugin tests enumerate only known subpaths (plugin/skills/exact-coverage walk is scoped to skills).dashboard/src/contracts/generated.tswas deliberately excluded: master's dashboard build runs@rsbuild/plugin-type-checkover**/*.tsand the file importszod, which is not in master's dashboard dependencies. It stays in feat: consolidate TraceDecay V2 PR8-PR13 delivery #707 with the code that wires it.vendor/libsql/deletion also stays in feat: consolidate TraceDecay V2 PR8-PR13 delivery #707 (master path-patcheslibsql = { path = "vendor/libsql" }).Deduping #707
After this merges, #707's branch gets a no-op merge commit (tree identical to its current HEAD) so the merge-base advances and these files leave its visible diff.