test(cli): provision compiled host-CLI fixtures

[ScriptedAlchemy]

Summary

• Provision the deterministic Kiro/Codex fake as a dedicated Cargo example built only by test workflows; ordinary product --bin / --bins builds no longer compile it in build.rs.
• Resolve the fixture from the executing test's remapped target profile and include the Windows .exe explicitly in the nextest archive contract.
• Record each invocation as a JSON argv array, preserving arguments that contain spaces, while retaining separate missing-CLI, malformed-output, and ownership-conflict outcomes.
• Keep real vendor conformance outside the deterministic fixture protocol suite.

Fixes #936.

Implementation

• tracedecay-host-cli-fixture is a [[example]]; Linux/macOS and Windows test lanes build it explicitly with Cargo, so target linker configuration remains Cargo-owned.
• .config/nextest.toml archives target/debug/examples/tracedecay-host-cli-fixture.exe for clean Windows shards.
• Test provisioning derives target/<profile>/examples/... from current_exe() instead of embedding a build-machine OUT_DIR.
• Unix tests symlink the prebuilt executable onto isolated PATHs, avoiding the ETXTBSY fork/write-descriptor race; Windows copies the native .exe.

Evidence

• Review-critical Kiro/Codex fixture and lifecycle selection passes, including missing executable, malformed output, ownership conflict, Codex automation, rollback, and retry.
• Agent-host package: 637/637 passed.
• Nextest archive -> fresh extraction -> separate checkout remap: 5/5 fixture protocol tests passed; nextest reported one archived non-test binary.
• cargo clippy -p tracedecay-cli -p tracedecay-agent-hosts --all-targets --locked -- -D warnings: clean on the latest merged feat: TraceDecay V2 delivery (PR8–PR13) + daemon performance, memory, and reliability overhaul #707 tip.
• The std-only fixture cross-builds to a PE32+ x86-64 Windows executable locally; native Windows archive/shard checks run on the pushed exact head.

The complete local CLI package run executed 570 tests: 569 passed and the unchanged branch_list_reads_profile_sharded_branch_meta subprocess hit its existing 90-second local deadline. Focused host-CLI acceptance passed again after each latest-#707 merge; per-finding and exact-head receipts are posted below.

[changeset-bot]

⚠️ No Changeset found

Latest commit: 78cf578

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

[ScriptedAlchemy]

Review — REVISE fixture provisioning; keep the native fake CLI

Reviewed 9d51efd120480029b939cd5559c2eb2f548129ae; source review only. GitHub reports conflicts and exact-head CI pending. A compiled deterministic fake is correct for these tests; requiring ambient Kiro/Codex or renaming a shell script to .exe is not.

P1 build regression: tracedecay-cli/build.rs::main unconditionally compiles this test-only executable for every product build. The direct rustc call adds --target but does not inherit Cargo's target linker/configuration, and uses the build-host EXE_SUFFIX. That makes normal release/cross-target builds depend on linking a fixture they never use. Provision it through an explicitly enabled test-support target/feature or a dedicated workflow helper build; don't use cfg(test) in build.rs as a substitute for explicit feature selection. Keep non-test builds free of this work.

P1 distribution acceptance: prove the fixture survives nextest archive -> clean Windows shard extraction/remap and is executable at the path the tests resolve. An OUT_DIR absolute path baked into a binary is not, by itself, evidence that the helper is shipped to the executing machine. Include the helper in the selected test-artifact contract and test without access to the build machine's tree.

Preserve missing-executable, malformed-output and ownership-conflict as different outcomes. Record argv as structured entries rather than a joined string if the assertion is meant to distinguish arguments containing spaces. Keep vendor conformance separate from the fake's protocol tests. Resolve conflicts and run native Windows host-fixture plus Kiro/Codex acceptance before integration into #707.

[ScriptedAlchemy]

System-level re-review at 9d51efd12 — keep test provisioning out of the product build contract

Re-read build.rs, compile_host_cli_fixture.rs and the shared provisioner. The earlier build-boundary objection is still present: build.rs invokes the fixture compiler unconditionally, including non-test builds/checks. The native executable itself is a sound replacement for a shell script named .exe; the ownership of its build/distribution is the problem.

P1 — the raw compiler invocation bypasses Cargo's target toolchain configuration. It forwards --target but not the target linker or applicable Cargo rustflags. A cross-target cargo check that previously needed only target metadata now attempts to link an unrelated test executable, and a product cross-build using a configured linker may fail because this subprocess does not inherit that configuration. The destination suffix is also taken from the build-script host rather than the target. Move this work to an explicit test-fixture provisioning step/artifact, or a test-only build contract that correctly follows the target toolchain. Do not gate it by debug/release profile: release tests and debug product builds are both valid.

P2 — verify the artifact, not only the installed test-local copy. provision_host_cli_fixture.rs obtains the source through compile-time TRACEDECAY_HOST_CLI_FIXTURE, then hard-links/copies it. It has no fallback if the build job's OUT_DIR file is absent. The Windows archive-only lane must explicitly transport that helper or reproduce it in test provisioning before tests run. I have not inspected a produced archive, so I am not claiming the helper is definitely missing; the current source establishes that it is a required external test artifact. Exercise a clean extraction without the original target directory and assert the helper runs, not merely that its bytes have an executable header.

Keep install/list/conflict/malformed/missing-program cases separate. A deterministic fake proves TraceDecay's invocation and parsing contract, not compatibility with the real vendor CLI; do not label it a vendor end-to-end acceptance lane. Preserve exact argument recording and isolated HOME/PATH.

No native compilation or archive execution performed here. Returned CI is pending, coverage in progress, profile succeeded and plugin validation queued. #707 remains draft.

[ScriptedAlchemy]

Addressed the latest review on exact head b542c6ee7.

P1 — production/cross-target build regression

• Removed the fixture compiler and all TRACEDECAY_HOST_CLI_FIXTURE / OUT_DIR plumbing from build.rs.
• Declared the fake as a Cargo [[example]], so ordinary product --bin / --bins builds exclude it.
• Added explicit Linux/macOS and native Windows test-workflow builds in ci.yml. Cargo now owns linker/target configuration.
• Local Windows-source build evidence: rustc --target x86_64-pc-windows-gnu ... host_cli_fixture.rs produced PE32+ executable (console) x86-64, for MS Windows. The repository's full Windows MSVC dependency graph cannot be cross-checked on this Linux host because ring requires Windows SDK headers; the pushed workflow performs the native MSVC build.

P1 — nextest archive distribution

• Added the exact Windows helper to profile.ci.archive.include.
• compiled_host_cli_fixture derives the executable from the running test's target/<profile>/examples directory; no build-machine absolute path is embedded.
• Archive simulation built a nextest archive, extracted it to /tmp/tracedecay-host-cli-extract-954-post-merge, remapped source metadata to a separate clean checkout, and ran the fixture target. Nextest reported Extracting 2 binaries (including 1 non-test binary) and all 5/5 tests passed from the extracted tree, including the assertion that the resolved helper is beside that test profile.

Preserved contracts

• 15/15 focused Kiro/Codex lifecycle tests passed after the latest feat: TraceDecay V2 delivery (PR8–PR13) + daemon performance, memory, and reliability overhaul #707 merge: missing executable, malformed output, ownership conflict, native format, Codex automation, rollback, and retry all remain distinct.
• Invocation logs are JSON arrays; the protocol test passes /usr/local/bin/Trace Decay/tracedecay and verifies it remains one argv entry.
• Deterministic fake protocol tests remain in host_cli_fixture; no real-vendor conformance was folded into them.
• 637/637 tracedecay-agent-hosts tests passed.
• Strict clippy passed with 0 warnings: cargo clippy -p tracedecay-cli -p tracedecay-agent-hosts --all-targets --locked -- -D warnings.

The complete local CLI run executed 570 tests: 569 passed; the unchanged branch_list_reads_profile_sharded_branch_meta subprocess hit its existing 90-second local deadline. The 15 host-CLI acceptance tests passed again on the merged tip.

[github-actions]

Performance Comparison codex/tracedecay-total-redesign-plan-reopened → cursor/host-cli-fixtures-42d7

Total Elapsed Time: 4.54s → 4.98s (+9.8%)
CPU Baseline: 77.80µs → 77.28µs (-0.7%)
Benchmark ID: index-bench-timing

timing - Execution duration of functions.

+------------------------------------------+----------------------------+----------------------------------+----------------------------------+----------------------------------+------------------------------+
| Function                                 | Calls                      | Avg                              | P95                              | Total                            | % Total                      |
+------------------------------------------+----------------------------+----------------------------------+----------------------------------+----------------------------------+------------------------------+
| tracedecay-index-bench                   | 1 → 1 (+0.0%)              | 4.54s → 4.98s (+9.7%)            | 4.54s → 4.99s (+9.9%)            | 4.54s → 4.98s (+9.7%)            | 100.00% → 100.00% (+0.0%)    |
+------------------------------------------+----------------------------+----------------------------------+----------------------------------+----------------------------------+------------------------------+
| query.artifact.append_pages              | 13 → 13 (+0.0%)            | 167.31ms → 192.83ms (+15.3%)     | 230.56ms → 326.11ms (+41.4%) ⚠️  | 2.18s → 2.51s (+15.1%)           | 47.91% → 50.31% (+5.0%)      |
+------------------------------------------+----------------------------+----------------------------------+----------------------------------+----------------------------------+------------------------------+
| query.artifact.batch.sqlite              | 13 → 13 (+0.0%)            | 113.51ms → 144.61ms (+27.4%) ⚠️  | 151.78ms → 273.94ms (+80.5%) ⚠️  | 1.48s → 1.88s (+27.0%) ⚠️        | 32.50% → 37.73% (+16.1%)     |
+------------------------------------------+----------------------------+----------------------------------+----------------------------------+----------------------------------+------------------------------+
| code_index.workers.install               | 83 → 83 (+0.0%)            | 15.88ms → 12.08ms (-23.9%) 🚀    | 53.74ms → 42.17ms (-21.5%) 🚀    | 1.32s → 1.00s (-24.2%) 🚀        | 29.03% → 20.12% (-30.7%) 🚀  |
+------------------------------------------+----------------------------+----------------------------------+----------------------------------+----------------------------------+------------------------------+
| query.artifact.batch.postings            | 13 → 13 (+0.0%)            | 76.40ms → 76.40ms (+0.0%)        | 103.74ms → 104.14ms (+0.4%)      | 993.20ms → 993.20ms (+0.0%)      | 21.88% → 19.93% (-8.9%)      |
+------------------------------------------+----------------------------+----------------------------------+----------------------------------+----------------------------------+------------------------------+
| query.artifact.batch.commit              | 13 → 13 (+0.0%)            | 34.34ms → 65.46ms (+90.6%) ⚠️    | 45.09ms → 198.71ms (+340.7%) ⚠️  | 446.40ms → 851.01ms (+90.6%) ⚠️  | 9.83% → 17.08% (+73.8%) ⚠️   |
+------------------------------------------+----------------------------+----------------------------------+----------------------------------+----------------------------------+------------------------------+
| domain.canonical.sha256                  | 114526 → 114526 (+0.0%)    | 8.99µs → 7.18µs (-20.1%) 🚀      | 12.04µs → 12.09µs (+0.4%)        | 1.03s → 822.10ms (-20.2%) 🚀     | 22.67% → 16.50% (-27.2%) 🚀  |
+------------------------------------------+----------------------------+----------------------------------+----------------------------------+----------------------------------+------------------------------+
| code_index.build.and_publish             | 2 → 2 (+0.0%)              | 504.48ms → 404.40ms (-19.8%)     | 528.22ms → 411.04ms (-22.2%) 🚀  | 1.01s → 808.81ms (-19.9%)        | 22.22% → 16.23% (-27.0%) 🚀  |
+------------------------------------------+----------------------------+----------------------------------+----------------------------------+----------------------------------+------------------------------+
| query.artifact.finalization.advance_wake | 14 → 14 (+0.0%)            | 34.40ms → 46.28ms (+34.5%) ⚠️    | 215.74ms → 217.32ms (+0.7%)      | 481.61ms → 647.95ms (+34.5%) ⚠️  | 10.61% → 13.00% (+22.5%) ⚠️  |
+------------------------------------------+----------------------------+----------------------------------+----------------------------------+----------------------------------+------------------------------+
| 🆕 query.artifact.create                 | 0 → 2 (+100.0%) ⚠️         | 0.00ns → 282.22ms (+100.0%) ⚠️   | 0.00ns → 282.33ms (+100.0%) ⚠️   | 0.00ns → 564.45ms (+100.0%) ⚠️   | 0.00% → 11.33% (+100.0%) ⚠️  |
+------------------------------------------+----------------------------+----------------------------------+----------------------------------+----------------------------------+------------------------------+
| query.artifact.prepare_pages             | 13 → 13 (+0.0%)            | 43.70ms → 38.35ms (-12.2%)       | 69.99ms → 54.23ms (-22.5%) 🚀    | 568.13ms → 498.56ms (-12.2%)     | 12.51% → 10.01% (-20.0%)     |
+------------------------------------------+----------------------------+----------------------------------+----------------------------------+----------------------------------+------------------------------+
| code_index.chunk.index_file              | 276 → 276 (+0.0%)          | 2.35ms → 1.80ms (-23.4%) 🚀      | 5.26ms → 2.61ms (-50.4%) 🚀      | 648.90ms → 497.93ms (-23.3%) 🚀  | 14.29% → 9.99% (-30.1%) 🚀   |
+------------------------------------------+----------------------------+----------------------------------+----------------------------------+----------------------------------+------------------------------+
| code_index.extract.parser_artifact       | 276 → 276 (+0.0%)          | 2.27ms → 1.65ms (-27.3%) 🚀      | 5.28ms → 2.48ms (-53.0%) 🚀      | 626.83ms → 455.98ms (-27.3%) 🚀  | 13.81% → 9.15% (-33.7%) 🚀   |
+------------------------------------------+----------------------------+----------------------------------+----------------------------------+----------------------------------+------------------------------+
| query.artifact.batch.parallel_prepare    | 13 → 13 (+0.0%)            | 35.29ms → 30.47ms (-13.7%)       | 59.34ms → 42.89ms (-27.7%) 🚀    | 458.83ms → 396.08ms (-13.7%)     | 10.11% → 7.95% (-21.4%) 🚀   |
+------------------------------------------+----------------------------+----------------------------------+----------------------------------+----------------------------------+------------------------------+
| code_index.chunk.build                   | 276 → 276 (+0.0%)          | 1.40ms → 1.09ms (-22.1%) 🚀      | 4.15ms → 1.59ms (-61.7%) 🚀      | 386.19ms → 299.84ms (-22.4%) 🚀  | 8.51% → 6.02% (-29.3%) 🚀    |
+------------------------------------------+----------------------------+----------------------------------+----------------------------------+----------------------------------+------------------------------+
| 🗑️ code_index.chunk.identify             | 276 → 0 (-100.0%) 🚀       | 1.36ms → 0.00ns (-100.0%) 🚀     | 4.13ms → 0.00ns (-100.0%) 🚀     | 376.20ms → 0.00ns (-100.0%) 🚀   | 8.29% → 0.00% (-100.0%) 🚀   |
+------------------------------------------+----------------------------+----------------------------------+----------------------------------+----------------------------------+------------------------------+

---

Generated with hotpath-rs

[ScriptedAlchemy]

Latest-base follow-up: #707 advanced after the first evidence post, so I merged its new tip abf7f47a8 and pushed exact head edb7c3b40.

Fresh evidence on edb7c3b40:

• 11/11 review-critical fixture, missing/malformed/conflict, and Codex automation tests passed with --no-tests=fail.
• Strict CLI + agent-host clippy passed with -D warnings (0 errors, 0 warnings).
• Remote comparison is 0 behind / 7 ahead of feat: TraceDecay V2 delivery (PR8–PR13) + daemon performance, memory, and reliability overhaul #707; GitHub reports MERGEABLE.

The intervening #707 delta touched search-eval/configuration files only; it did not alter the fixture, archive contract, CLI host lifecycle, or agent-host code. Exact-head GitHub checks have restarted, including the native Windows archive/shards.

[ScriptedAlchemy]

Final base convergence: pushed 78cf57842, which contains current #707 tip abf58b20e.

• Remote comparison: 0 behind / 10 ahead.
• GitHub mergeability: MERGEABLE.
• The final four base commits touched graph-db/session Git-correlation code only; none touched the fixture, CLI host lifecycle, agent-hosts, archive config, or CI workflow. The immediately preceding integrated tip passed 11/11 review-critical tests and strict CLI/agent-host clippy.
• Exact-head checks restarted and remain queued/pending. The Actions job watcher hit the shared GitHub REST rate limit before native Windows jobs began; it did not report a CI test failure.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-09-07T06:51:29.309559Z	78cf578	Draft marked ready

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" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.