test(hotpath): slice coverage tests + CI for sessions/lcm/capture/temporal

[ScriptedAlchemy]

What this PR is

Hotpath coverage only for the sessions/LCM/capture/temporal-query slice: tests and CI that prove the slice's hotpath instrumentation is real, feature-gated, and off by default. Benches are deliberately deferred to a later PR (they will be daemon-free when they come). No production code is touched.

Crates touched (tests only, one new tests/hotpath_coverage.rs each)

• crates/tracedecay-sessions
• crates/tracedecay-session-memory
• crates/tracedecay-session-runtime
• crates/tracedecay-session-temporal-store
• crates/tracedecay-lcm
• crates/tracedecay-capture
• crates/tracedecay-temporal-query

Plus one new workflow: .github/workflows/hotpath-slice-tests.yml. The existing index-bench workflows (hotpath-profile.yml, hotpath-comment.yml) are untouched.

What each crate's hotpath_coverage.rs proves

1. Off by default (runs in both lanes): the crate manifest's default/production feature arrays never contain hotpath.
2. Feature-off no-op (cfg(not(feature = "hotpath"))): a real workload runs with HOTPATH_OUTPUT_FORMAT/HOTPATH_OUTPUT_PATH set — no report file is written, and no listener answers on 6770/6771.
3. Feature-on instrumentation is live (cfg(feature = "hotpath")): a HotpathGuardBuilder guard (JSON report, metrics server off) wraps the same workload, and the report must contain the crate's real measured-site labels — e.g. capture.parse.record, sessions.lcm.scan_base64, sessions.shared.content_storage, usecases.context.session_grant, session_temporal.hydrate.render, temporal.rank, and the async daemon.lcm.execute/daemon.lcm.status futures spans through the mounted LCM authority.

All workloads are deterministic, daemon-free, and socket-free (session-runtime uses the registered global-db test harness on a tempdir). No new fixtures or dependencies were added; no test-only spans were introduced, so no new labels either.

CI

hotpath-slice-tests.yml runs on every PR:

• cargo test -p <7 slice crates> --locked (feature off)
• cargo test -p <7 slice crates> --features hotpath --locked (feature on)

with HOTPATH_METRICS_SERVER_OFF=1 at the job boundary. It is additive and does not replace or modify the index-bench profile lane.

Run locally

# feature off
cargo test -p tracedecay-sessions -p tracedecay-session-memory \
  -p tracedecay-session-runtime -p tracedecay-session-temporal-store \
  -p tracedecay-lcm -p tracedecay-capture -p tracedecay-temporal-query --locked

# feature on
cargo test -p tracedecay-sessions -p tracedecay-session-memory \
  -p tracedecay-session-runtime -p tracedecay-session-temporal-store \
  -p tracedecay-lcm -p tracedecay-capture -p tracedecay-temporal-query \
  --features hotpath --locked

Explicit confirmations

• hotpath / hotpath-alloc / hotpath-cpu / hotpath-mcp remain out of default and every production feature set — and the new tests now enforce that permanently.
• No existing #[hotpath::measure] (or any other hotpath macro) was added, restamped, rewritten, renamed, or moved. The in-flight measure PRs (HP-sessions, HP-temporal, HP-lcm) are untouched.
• Workspace stays pinned to hotpath 0.24 (default-features = false, features = ["threads"]); no version bump, no Cargo.toml/Cargo.lock changes.
• PR feat: TraceDecay V2 delivery (PR8–PR13) + daemon performance, memory, and reliability overhaul #707 is untouched and unmerged; this branch only reads from it as its base.

  

[changeset-bot]

⚠️ No Changeset found

Latest commit: 39937e1

Merging this PR will not cause a version bump for any packages. If these changes should not result in a new version, you're good to go. If these changes should result in a version bump, you need to add a changeset.

Click here to learn what changesets are, and how to add one.

Click here if you're a maintainer who wants to add a changeset to this PR

[github-actions]

Performance Comparison codex/tracedecay-total-redesign-plan-reopened → cursor/hotpath-slice-coverage-7e30

Total Elapsed Time: 5.09s → 5.07s (-0.4%)
CPU Baseline: 87.50µs → 83.63µs (-4.4%)
Benchmark ID: index-bench-timing

timing - Execution duration of functions.

+------------------------------------------+----------------------------+--------------------------------+--------------------------------+--------------------------------+------------------------------+
| Function                                 | Calls                      | Avg                            | P95                            | Total                          | % Total                      |
+------------------------------------------+----------------------------+--------------------------------+--------------------------------+--------------------------------+------------------------------+
| tracedecay-index-bench                   | 1 → 1 (+0.0%)              | 5.09s → 5.07s (-0.4%)          | 5.09s → 5.07s (-0.4%)          | 5.09s → 5.07s (-0.4%)          | 100.00% → 100.00% (+0.0%)    |
+------------------------------------------+----------------------------+--------------------------------+--------------------------------+--------------------------------+------------------------------+
| query.artifact.append_pages              | 13 → 13 (+0.0%)            | 178.54ms → 180.27ms (+1.0%)    | 245.24ms → 247.07ms (+0.7%)    | 2.32s → 2.34s (+0.9%)          | 45.63% → 46.26% (+1.4%)      |
+------------------------------------------+----------------------------+--------------------------------+--------------------------------+--------------------------------+------------------------------+
| query.artifact.batch.sqlite              | 13 → 13 (+0.0%)            | 122.24ms → 123.92ms (+1.4%)    | 168.56ms → 169.08ms (+0.3%)    | 1.59s → 1.61s (+1.3%)          | 31.24% → 31.80% (+1.8%)      |
+------------------------------------------+----------------------------+--------------------------------+--------------------------------+--------------------------------+------------------------------+
| query.artifact.batch.postings            | 13 → 13 (+0.0%)            | 107.25ms → 108.74ms (+1.4%)    | 149.68ms → 151.13ms (+1.0%)    | 1.39s → 1.41s (+1.4%)          | 27.41% → 27.90% (+1.8%)      |
+------------------------------------------+----------------------------+--------------------------------+--------------------------------+--------------------------------+------------------------------+
| CodeIndexProductionOwnerV1::extract_file | 276 → 276 (+0.0%)          | 5.19ms → 4.89ms (-5.8%)        | 8.03ms → 7.40ms (-7.8%)        | 1.43s → 1.35s (-5.6%)          | 28.17% → 26.62% (-5.5%)      |
+------------------------------------------+----------------------------+--------------------------------+--------------------------------+--------------------------------+------------------------------+
| code_index.workers.install               | 83 → 83 (+0.0%)            | 16.11ms → 15.87ms (-1.5%)      | 53.71ms → 53.48ms (-0.4%)      | 1.34s → 1.32s (-1.5%)          | 26.29% → 25.99% (-1.1%)      |
+------------------------------------------+----------------------------+--------------------------------+--------------------------------+--------------------------------+------------------------------+
| code_index.build.and_publish             | 2 → 2 (+0.0%)              | 533.32ms → 522.08ms (-2.1%)    | 551.55ms → 527.70ms (-4.3%)    | 1.07s → 1.04s (-2.8%)          | 20.97% → 20.61% (-1.7%)      |
+------------------------------------------+----------------------------+--------------------------------+--------------------------------+--------------------------------+------------------------------+
| domain.canonical.sha256                  | 114526 → 114526 (+0.0%)    | 9.08µs → 8.98µs (-1.1%)        | 15.10µs → 15.13µs (+0.2%)      | 1.04s → 1.03s (-1.0%)          | 20.44% → 20.31% (-0.6%)      |
+------------------------------------------+----------------------------+--------------------------------+--------------------------------+--------------------------------+------------------------------+
| query.artifact.finalization.advance_wake | 14 → 14 (+0.0%)            | 53.80ms → 53.54ms (-0.5%)      | 264.90ms → 263.06ms (-0.7%)    | 753.20ms → 749.53ms (-0.5%)    | 14.81% → 14.79% (-0.1%)      |
+------------------------------------------+----------------------------+--------------------------------+--------------------------------+--------------------------------+------------------------------+
| code_index.chunk.index_file              | 276 → 276 (+0.0%)          | 2.37ms → 2.25ms (-5.1%)        | 3.51ms → 3.28ms (-6.6%)        | 654.29ms → 621.67ms (-5.0%)    | 12.86% → 12.27% (-4.6%)      |
+------------------------------------------+----------------------------+--------------------------------+--------------------------------+--------------------------------+------------------------------+
| query.artifact.prepare_pages             | 13 → 13 (+0.0%)            | 47.45ms → 47.43ms (-0.0%)      | 68.75ms → 68.16ms (-0.9%)      | 616.88ms → 616.54ms (-0.1%)    | 12.13% → 12.17% (+0.3%)      |
+------------------------------------------+----------------------------+--------------------------------+--------------------------------+--------------------------------+------------------------------+
| code_index.extract.parser_artifact       | 276 → 276 (+0.0%)          | 2.24ms → 2.10ms (-6.2%)        | 4.04ms → 3.18ms (-21.3%) 🚀    | 619.29ms → 578.95ms (-6.5%)    | 12.17% → 11.43% (-6.1%)      |
+------------------------------------------+----------------------------+--------------------------------+--------------------------------+--------------------------------+------------------------------+
| query.artifact.batch.parallel_prepare    | 13 → 13 (+0.0%)            | 37.15ms → 37.13ms (-0.1%)      | 53.87ms → 53.48ms (-0.7%)      | 482.94ms → 482.69ms (-0.1%)    | 9.49% → 9.53% (+0.4%)        |
+------------------------------------------+----------------------------+--------------------------------+--------------------------------+--------------------------------+------------------------------+
| code_index.chunk.build                   | 276 → 276 (+0.0%)          | 1.45ms → 1.38ms (-4.8%)        | 2.11ms → 1.99ms (-5.7%)        | 401.09ms → 381.07ms (-5.0%)    | 7.88% → 7.52% (-4.6%)        |
+------------------------------------------+----------------------------+--------------------------------+--------------------------------+--------------------------------+------------------------------+
| code_index.chunk.identify                | 276 → 276 (+0.0%)          | 1.44ms → 1.37ms (-4.9%)        | 2.09ms → 1.97ms (-5.7%)        | 396.88ms → 376.78ms (-5.1%)    | 7.80% → 7.44% (-4.6%)        |
+------------------------------------------+----------------------------+--------------------------------+--------------------------------+--------------------------------+------------------------------+

---

Generated with hotpath-rs

[ScriptedAlchemy]

Review of record (integration lane, base 7f1f8e02c).

The feature-on guard-report tests are the valuable part and are accepted as written (subject to every asserted label still existing after 3e26c7596, which removed many spans — verified during integration). Three changes will land as a follow-up commit on the merged head:

1. Fold the workflow. Origin already carries .github/workflows/hotpath-coverage.yml (storage slice, job slice-tests). This PR adds a second workflow whose job is also named slice-tests, so the PR shows two identically named checks with different results. The seven crates run as a second job sessions-slice-tests inside hotpath-coverage.yml; hotpath-slice-tests.yml is dropped.
2. Drop hotpath_stays_out_of_default_and_production_features and the hand-rolled manifest_feature_array (duplicated 7×). Scanning Cargo.toml text is a source-shape scan used as acceptance, which the repo rules exclude; the feature-off CI lane building with default features and asserting no report is written is the behavioral check.
3. Drop the 6770/6771 TcpStream::connect probe. It asserts a property of the host (no other process listening), not of the crate, and false-fails on any machine running a hotpath-enabled daemon — including the dev box this is integrated on.

Red checks here are from a stale-base merge commit; re-judged on the merged #707 head.

[ScriptedAlchemy]

Integrated into codex/tracedecay-total-redesign-plan-reopened (#707) by fast-forward push, rebased on 77be8b23c:

• 622d066a9 test(hotpath): add slice coverage for sessions/lcm/capture/temporal — this PR's seven tests/hotpath_coverage.rs files (trailer Merges #790).
• 94eac52fe test(hotpath): fold slice coverage into the hotpath-coverage workflow — the review fixes above: the seven crates run as job sessions-slice-tests inside .github/workflows/hotpath-coverage.yml (per-package <crate>/hotpath feature syntax, HOTPATH_METRICS_SERVER_OFF=1); hotpath-slice-tests.yml removed; the Cargo.toml string-scan test and the 6770/6771 port probe dropped from all seven suites.

Evidence on the integrated head (cf8269f80 base, identical patches):

• every asserted label still has a live production measure after the #774/#775/#776/#778/#779/#784/#786 instrumentation reverts (16 labels checked, e.g. sessions.shared.content_storage → crates/tracedecay-sessions/src/runtime/shared.rs:676, daemon.lcm.execute → lcm_authority.rs:178, temporal.rank → ranking.rs:125);
• feature-off lane: 7 binaries × 1 passed; 0 failed; feature-on lane (--features tracedecay-sessions/hotpath,…): 7 × 1 passed; 0 failed, guard reports carry the labels;
• cargo fmt --check, clippy --tests -D warnings for the seven crates, workflow YAML parse: pass;
• MCP suite on the integrated head: 384 passed / 1 failed (context_relevance_eval_test::context_eval_fixture_scores_real_queries, mean_recall_at_5 0.29 < 0.3, unrelated to this slice — retrieval eval on the base; being re-run in isolation).

Closing: the content is on the target branch under the commits above.