Home / Projects / Orchestrata Ultimate / Decisions / ADR index / ADR-0127

ADR-0127: The test runner stays a repo dev tool outside the integration release until codetools has an affected-set oracle

← ADR-0126 · ADR-0128 →

The record

Status: Accepted (Mohamed, 2026-10-22: "ok agree") · Date: 2026-10-23

Context

The integration release plan (wt-COMPONENTS packages/integration/increments/README.md, read 2026-10-22) has 25 increments, INT-01..INT-25, lifting 7,136 LOC of originals. INT-12..INT-22 move scripts/run_tests.py, run_tests_streaming.py, run_tests_threadpool.py and start_test_services.py into the integration package as orchestrata.run_tests and siblings.

Measured in the cut-over review (2026-10-22, read-only):

• INT-12..22 test tooling is 3,741 of the 7,136 LOC, which is 52% of the release.
• The frozen source commit e57782ea4 is stale for run_tests.py: it has changed +566/−69 lines in 11 commits since then. A lift from the frozen tree would ship a runner that is already behind.
• ha_controller_cli (185 LOC, inside INT-22) is runtime code, not test tooling.
• A separate thread proposed moving the tooling tests to codetools. That conflicts with INT-12..22 moving the same files into the integration package. The two had to be reconciled before either moved.

Decision (Mohamed, verbatim: "ok agree")

1. INT-12..INT-21, and the start_test_services half of INT-22, are dropped from the integration release. ha_controller_cli stays in the release.
2. scripts/run_tests.py stays in the repo as a developer tool. It moves to codetools only when codetools has an affected-set oracle that can replace --affected.
3. Two work items join the cut-over:
   • AST roots fix A: ast_pipeline.source_trees reads the declared roots. This landed as eddfc6668 on fix/ast-pipeline-declared-roots.
   • Affected-set cross-package dependents: a contracts change must select the kernel and volks tests.

Alternatives rejected

• Lift the runner into orchestrata.run_tests as planned. It doubles the release with test tooling, ships a stale copy of the runner, and puts dev tooling in the product.
• Move the runner to codetools now. codetools has no affected-set oracle yet, so the push gate would lose --affected (RUL-040 named verdicts).

Consequences

• The integration release drops to about 3,400 LOC of runtime code.
• The push gate keeps running scripts/run_tests.py --affected from the repo until the codetools oracle exists. A later ADR records that move.
• INT increments INT-12..21 are closed as "not in release". Whoever executes the increments plan must record this in it.

Provenance

• Mohamed, 2026-10-22, during the cut-over review, on the question of keeping the test runner out of the release: "ok agree". He then said "not an adr", and followed with "when we have finished understanding, researchimg and agreeing adr becomes a publish step for you" (see ADR-0130).
• He asked for publication on 2026-10-23: "review the decisions in the scratch pad, update them and publish them if they havent already been covered". A new ADR rather than an addendum, because the README forbids editing an accepted decision, and ADR-0113 records the L1 build, not the increments' content.
• Evidence: the session scratch notes (cut-over review findings) and the integration increments README in wt-COMPONENTS.

Related ADRs

• Addenda: ADR-0130
• Also mentioned: ADR-0113

Rulings cited

• RUL-040: RUL-040 — verification scope is measured at symbol granularity; a full-suite escalation is a bug

Specs cited

None (or no spec page exists for the ones named).