jared/cc-os

Watch

Fork 0

Audit takedown eval for duplicates of os-sdlc lib, move onto owners #1011

New issue

Open

opened 2026-10-15 17:00:50 +00:00 by jared · 4 comments

jared commented

2026-10-15 17:00:50 +00:00

Owner

Problem

The takedown eval (plugins/os-sdlc/eval/ticket-takedown) re-implements behavior that plugins/os-sdlc/lib/os_sdlc already owns. The eval will merge into the os-sdlc lib later. Each duplicate adds code and tests that the merge must remove.

Jared's ruling, 2026-10-15: "The eval project should reuse as much of the os-sdlc code as possible. It will be integrated into os-sdlc's lib once we have the kinks ironed-out anyway. So if the code it needs already exists, use it! The eval code should really only be the difference between the existing os-sdlc code and the customizations needed for the eval portion. So, it should be using the entire Project class - not necessarily just Project::Target and Project::Config. That should reduce the number of tests needed and new code required when we merge in the future."

An ADR records this rule. It is created in the same session as this ticket. See the ADR on takedown eval reuse of os-sdlc lib.

Observed

Verified on HEAD:

eval/ticket-takedown/lib/takedown/adapter/project_suites.rb defines PATH = File.join(".sdlc", "project.yaml"), reads it with YAML.safe_load_file, and reads raw section hashes. It wraps a single OsSdlc::Project::Target through from_section. Owners: OsSdlc::Project.from_data (lib/os_sdlc/model/project.rb) and OsSdlc::Project::Config (lib/os_sdlc/adapter/project/config.rb, which owns path, read, document).
The literal File.join(@repo, ".sdlc", "project.yaml") repeats in eval tests: test/command/green_suite_test.rb:20, test/command/polish_test.rb:91, test/command/final_suite_test.rb:101 and :208, test/controller/run_cli_test.rb:26. See docs/research/lint-lessons/sdlc-dry-repeated-literal-argument.md.
Eval tests may re-test os-sdlc behavior (path ownership, target selection). Not yet enumerated.

Reproduce

n/a: audit and refactor ticket; nothing to reproduce.

Approach

Audit the whole eval, not only the instances above. For each duplicate, name the os-sdlc owner and move the eval onto it. Where the eval must differ, record the reason. The eval keeps only the difference from os-sdlc. This work needs at least one dedicated session.

Required work:

Load the project through OsSdlc::Project in the eval adapter — check: absent YAML.safe_load_file in plugins/os-sdlc/eval/ticket-takedown/lib/**/*.rb
Remove the repeated project.yaml path literal from eval tests — check: absent ".sdlc", "project.yaml" in plugins/os-sdlc/eval/ticket-takedown/test/**/*.rb

Acceptance criteria

Given the audit is done, when the ticket closes, then every duplicate found is listed with its os-sdlc owner.
Given a listed duplicate, when the ticket closes, then the eval uses the owner, or the ticket records why the eval must differ.
Given an eval test that re-tests os-sdlc behavior, when the ticket closes, then the test is deleted.
Given the change, when the eval suite and the os-sdlc suite run, then both pass.

Owner sketch

class Takedown::ProjectSuites
  def targets
    # OsSdlc::Project built from the owner's loader; eval adds only its command/suite customization
  end
end

Origin

Trigger: Jared's ruling in session 270dfee2, 2026-10-15
Improvised this session: none
Chain: eval duplicates os-sdlc project loading ← eval built standalone before the merge plan ← DESIGN (ADR on takedown eval reuse of os-sdlc lib)
Root candidate: none
Where: plugins/os-sdlc/eval/ticket-takedown/lib/takedown/adapter/project_suites.rb
Session: 270dfee2-6a92-4e87-828c-9ac8c32c1cd7
Transcript: /home/jared/.claude/projects/-home-jared-dev-cc-os--claude-worktrees-prism-architect/270dfee2-6a92-4e87-828c-9ac8c32c1cd7.jsonl

VERDICT: CREATE (ticket-skeptic: explicit user request floor applies)

## Problem The takedown eval (`plugins/os-sdlc/eval/ticket-takedown`) re-implements behavior that `plugins/os-sdlc/lib/os_sdlc` already owns. The eval will merge into the os-sdlc lib later. Each duplicate adds code and tests that the merge must remove. Jared's ruling, 2026-10-15: "The eval project should reuse as much of the os-sdlc code as possible. It will be integrated into os-sdlc's lib once we have the kinks ironed-out anyway. So if the code it needs already exists, use it! The eval code should really only be the difference between the existing os-sdlc code and the customizations needed for the eval portion. So, it should be using the entire Project class - not necessarily just Project::Target and Project::Config. That should reduce the number of tests needed and new code required when we merge in the future." An ADR records this rule. It is created in the same session as this ticket. See the ADR on takedown eval reuse of os-sdlc lib. ## Observed Verified on HEAD: 1. `eval/ticket-takedown/lib/takedown/adapter/project_suites.rb` defines `PATH = File.join(".sdlc", "project.yaml")`, reads it with `YAML.safe_load_file`, and reads raw section hashes. It wraps a single `OsSdlc::Project::Target` through `from_section`. Owners: `OsSdlc::Project.from_data` (`lib/os_sdlc/model/project.rb`) and `OsSdlc::Project::Config` (`lib/os_sdlc/adapter/project/config.rb`, which owns `path`, `read`, `document`). 2. The literal `File.join(@repo, ".sdlc", "project.yaml")` repeats in eval tests: `test/command/green_suite_test.rb:20`, `test/command/polish_test.rb:91`, `test/command/final_suite_test.rb:101` and `:208`, `test/controller/run_cli_test.rb:26`. See `docs/research/lint-lessons/sdlc-dry-repeated-literal-argument.md`. 3. Eval tests may re-test os-sdlc behavior (path ownership, target selection). Not yet enumerated. ## Reproduce n/a: audit and refactor ticket; nothing to reproduce. ## Approach Audit the whole eval, not only the instances above. For each duplicate, name the os-sdlc owner and move the eval onto it. Where the eval must differ, record the reason. The eval keeps only the difference from os-sdlc. This work needs at least one dedicated session. Required work: 1. Load the project through `OsSdlc::Project` in the eval adapter — check: absent `YAML.safe_load_file` in `plugins/os-sdlc/eval/ticket-takedown/lib/**/*.rb` 2. Remove the repeated project.yaml path literal from eval tests — check: absent `".sdlc", "project.yaml"` in `plugins/os-sdlc/eval/ticket-takedown/test/**/*.rb` ## Acceptance criteria 1. Given the audit is done, when the ticket closes, then every duplicate found is listed with its os-sdlc owner. 2. Given a listed duplicate, when the ticket closes, then the eval uses the owner, or the ticket records why the eval must differ. 3. Given an eval test that re-tests os-sdlc behavior, when the ticket closes, then the test is deleted. 4. Given the change, when the eval suite and the os-sdlc suite run, then both pass. ## Owner sketch ```ruby class Takedown::ProjectSuites def targets # OsSdlc::Project built from the owner's loader; eval adds only its command/suite customization end end ``` ## Origin - Trigger: Jared's ruling in session 270dfee2, 2026-10-15 - Improvised this session: none - Chain: eval duplicates os-sdlc project loading ← eval built standalone before the merge plan ← DESIGN (ADR on takedown eval reuse of os-sdlc lib) - Root candidate: none - Where: plugins/os-sdlc/eval/ticket-takedown/lib/takedown/adapter/project_suites.rb - Session: 270dfee2-6a92-4e87-828c-9ac8c32c1cd7 - Transcript: /home/jared/.claude/projects/-home-jared-dev-cc-os--claude-worktrees-prism-architect/270dfee2-6a92-4e87-828c-9ac8c32c1cd7.jsonl VERDICT: CREATE (ticket-skeptic: explicit user request floor applies)

jared added the

plugin/os-sdlc

filed-by/user

category/code-change

labels

2026-10-15 17:00:50 +00:00

jared commented

2026-10-15 17:01:34 +00:00

Author

Owner

The rule this ticket applies is recorded in ADR-0296 (docs/adr/0296-the-takedown-eval-holds-only-its-difference-from-os-sdlc-lib-and-reuses-os-sdlc-owners-amends-adr-0293.md).

jared commented

2026-10-15 17:01:55 +00:00

Author

Owner

Correction: the ADR is ADR-0297 (docs/adr/0297-the-takedown-eval-holds-only-its-difference-from-os-sdlc-lib-and-reuses-os-sdlc-owners-amends-adr-0293.md). 0296 is taken on branch adr-gate.

jared referenced this issue

2026-10-28 15:40:25 +00:00

Revisit test fixture and factory strategy across plugin test suites #1019

jared commented

2026-10-28 15:40:45 +00:00

Author

Owner

Occurrence: Use ProjectScenario in the takedown eval tests
Seen: eval/ticket-takedown/test/command/green_suite_test.rb:20, command/polish_test.rb:91, command/final_suite_test.rb:38+228 (same two-target Hash twice), controller/run_cli_test.rb:26
Run: session 2dd497e0-555e-4518-9cdd-c68461c32463 (#1019 plan, 2026-10-28)
Adds: Required work step 2 is satisfied by reusing os-sdlc's test support rather than a new eval helper. Add two lines to eval/ticket-takedown/test/test_helper.rb: require_relative ../../../tests/support/home_and_start and ../../../tests/support/project_scenario (after #1460 splits HomeAndStart out). Do not push the whole os-sdlc tests/support dir: it eager-loads ~85 files that assume the os-sdlc test helper's DB connection and Open3 stubs. Replace each Hash+write with ProjectScenario.new(@repo, name:).write_multi_target_project_yaml(targets:). ADR-0297 is retired, so the boundary no longer forbids this. Measurement from plugins/os-sdlc should give yaml=0 afterwards (command on #1019 memo comment). Depends on #1460.

Occurrence: Use ProjectScenario in the takedown eval tests Seen: eval/ticket-takedown/test/command/green_suite_test.rb:20, command/polish_test.rb:91, command/final_suite_test.rb:38+228 (same two-target Hash twice), controller/run_cli_test.rb:26 Run: session 2dd497e0-555e-4518-9cdd-c68461c32463 (#1019 plan, 2026-10-28) Adds: Required work step 2 is satisfied by reusing os-sdlc's test support rather than a new eval helper. Add two lines to eval/ticket-takedown/test/test_helper.rb: require_relative ../../../tests/support/home_and_start and ../../../tests/support/project_scenario (after #1460 splits HomeAndStart out). Do not push the whole os-sdlc tests/support dir: it eager-loads ~85 files that assume the os-sdlc test helper's DB connection and Open3 stubs. Replace each Hash+write with ProjectScenario.new(@repo, name:).write_multi_target_project_yaml(targets:). ADR-0297 is retired, so the boundary no longer forbids this. Measurement from plugins/os-sdlc should give yaml=0 afterwards (command on #1019 memo comment). Depends on #1460.

jared referenced this issue

2026-10-28 15:42:49 +00:00

Add record-fixture reference, lens link, and test-writer index line #1458

jared referenced this issue

2026-10-28 16:46:20 +00:00

Replace hand-built ticket bodies and project.yaml Hashes in os-sdlc tests with fixtures #1461

jared referenced this issue

2026-10-28 16:46:27 +00:00

Revisit test fixture and factory strategy across plugin test suites #1019

jared commented

2026-10-28 19:00:18 +00:00

Author

Owner

Occurrence done (fast-batch-22, main 5f32ca8f, commits 97581502 and 4ae19f04): the takedown eval tests write project.yaml through ProjectScenario#write_multi_target_project_yaml (7 sites; final_suite_test had 4). The eval test_helper require_relatives home_and_start, project_scenario, and git_repo_scenario only. The #1019 measurement prints yaml=0. Eval suite 608 runs, 0 failures. The scenario also writes name: widgets; Takedown::ProjectSuites reads only targets (memo docs/crc/os-sdlc/fast-batch-22/finding-l-eval-project-scenario.md).

Occurrence done (fast-batch-22, main 5f32ca8f, commits 97581502 and 4ae19f04): the takedown eval tests write project.yaml through ProjectScenario#write_multi_target_project_yaml (7 sites; final_suite_test had 4). The eval test_helper require_relatives home_and_start, project_scenario, and git_repo_scenario only. The #1019 measurement prints yaml=0. Eval suite 608 runs, 0 failures. The scenario also writes name: widgets; Takedown::ProjectSuites reads only targets (memo docs/crc/os-sdlc/fast-batch-22/finding-l-eval-project-scenario.md).

jared referenced this issue

2026-10-28 19:00:30 +00:00

Revisit test fixture and factory strategy across plugin test suites #1019

jared referenced this issue from a commit

2026-10-28 19:01:49 +00:00

refactor(os-sdlc): takedown eval tests write project.yaml through ProjectScenario (#1011, fast-batch-22 finding L)