Complete Test Coverage for Final 3 Untested Scripts

[github-actions]

Strategic Improvement

Write test suites for the remaining 3 scripts without coverage (sw-tmux-role-color.sh, sw-tmux-status.sh, sw-tracker-github.sh) to achieve 100% test coverage. This closes the gap for full platform testability and enables confident autonomous refactoring.

Acceptance Criteria

• sw-tmux-role-color-test.sh validates color assignment for all agent roles
• sw-tmux-status-test.sh validates status bar widget formatting and content
• sw-tracker-github-test.sh validates GitHub issue/PR fetch and update operations
• All 3 test suites follow existing patterns (mock binaries, PASS/FAIL counters, tmp dirs)
• npm test shows 103/103 scripts tested (100% coverage)

Context

• Priority: P6
• Complexity: fast
• Generated by: Strategic Intelligence Agent
• Strategy alignment: P6: Platform Self-Improvement (AGI-Level Readiness)

[github-actions]

⚓ Shipwright Pipeline Starting

Field	Value
Template	standard
Complexity	standard
Max Iterations	20
Triage	Write 3 focused test suites following existing patterns to achieve 100% test coverage and close platform testability gap.
Runner	ubuntu-latest
Risk Level	unknown
Workflow	25588771288

The autonomous pipeline is processing this issue. Updates will be posted as stages complete.

[github-actions]

[github-actions]

[github-actions]

📋 Implementation Plan

Click to expand full plan (34 lines)

The plan is ready. Here's a summary of what I'll implement:

Plan: Complete Test Coverage for Final 3 Untested Scripts

What I found

The three untested scripts have distinct testing challenges:

• sw-tmux-role-color.sh — A 82-line script that reads the tmux pane title and maps role keywords to hex colors. Testing requires a mock tmux binary that returns a controlled MOCK_PANE_TITLE and logs the pane-active-border-style set call so assertions can check the correct color was chosen.

• sw-tmux-status.sh — A 152-line script with 4 testable functions (stage_color, stage_icon, pipeline_widget, agent_widget). Functions can be sourced for pure-function testing, and dispatch modes tested by running the script as a subprocess with a controlled .claude/pipeline-state.md and fake heartbeat files.

• sw-tracker-github.sh — A 213-line sourced provider file (no set -euo pipefail) with 8 provider_* functions. Tested by sourcing in a subshell with a mock gh CLI that logs calls and returns realistic JSON.

Files to create/modify

File	Action
scripts/sw-tmux-role-color-test.sh	CREATE — 12 tests
scripts/sw-tmux-status-test.sh	CREATE — 16 tests
scripts/sw-tracker-github-test.sh	CREATE — 17 tests
package.json	MODIFY — add 3 entries to npm test

Key decisions

• All tests are unit-level (100%) — no E2E needed since all external calls are mocked
• Subshell isolation for sourcing sw-tracker-github.sh to prevent function pollution
• Mock tmux log via env var for reliable color capture
• tr not ${var,,} throughout to maintain Bash 3.2 compatibility

Failure modes addressed

1. Bash 3.2 compatibility — no ${var,,}, always use tr
2. Linux/macOS stat differences for heartbeat mtime — test only well-defined paths
3. Sourced provider file — always call inside ( subshell )
4. Mock tmux log path — assert non-empty before checking content

---

Generated by shipwright pipeline at 2026-05-09T02:19:00Z

[github-actions]

[github-actions]

📐 Architecture Decision Record

Click to expand ADR (42 lines)

Design: Complete Test Coverage for Final 3 Untested Scripts

Context

Three bash scripts in scripts/ lack test coverage: sw-tmux-role-color.sh (82 lines, tmux pane border colorizer), sw-tmux-status.sh (152 lines, tmux status widgets), and sw-tracker-github.sh (213 lines, sourced GitHub tracker provider). The repo's test harness is bash-based (despite the project metadata claiming vitest); existing tests follow the scripts/sw-*-test.sh convention with PASS/FAIL counters, ERR traps, and mock binaries placed on PATH via a temp directory. Constraints: Bash 3.2 compatibility (no ${var,,}, no declare -A, no readarray), set -euo pipefail in tests, atomic file writes, and npm test parallel execution across ~150 suites.

Decision

Three new sibling test scripts, each self-contained with its own TMPDIR (via mktemp -d) for full isolation under parallel npm test execution.

• sw-tmux-role-color-test.sh — Mock tmux shim on PATH that reads MOCK_PANE_TITLE env and appends set-option invocations to $MOCK_TMUX_LOG. Each test sets the title, invokes the script, asserts the logged hex color matches the expected role mapping (builder→green, reviewer→amber, tester→cyan, etc.) plus default fallback and empty-title cases.
• sw-tmux-status-test.sh — Source the script with a guard env var to skip top-level dispatch, then call stage_color/stage_icon/pipeline_widget/agent_widget directly as pure functions. Dispatch modes (pipeline, agents) tested by subprocess invocation against a fixture .claude/pipeline-state.md and synthetic ~/.shipwright/heartbeats/*.json files written into the per-test temp HOME.
• sw-tracker-github-test.sh — Source the provider inside ( subshell ) to contain function pollution. Mock gh shim logs argv to $MOCK_GH_LOG and emits canned JSON keyed by argv pattern. Cover all 8 provider_* functions plus error paths (gh failure, malformed JSON).

Data flow: each test creates TMPDIR=$(mktemp -d), writes mocks under $TMPDIR/bin, prepends to PATH, runs assertions, increments PASS/FAIL, and cleans up via trap on EXIT. Final exit $((FAIL > 0)).

Alternatives Considered

1. Single combined test file for all three scripts — Pros: less boilerplate. Cons: violates the one-test-per-script convention used by ~140 existing suites; harder to register selectively in package.json; longer feedback loop on a single failure.
2. End-to-end tests with a real tmux server — Pros: validates real tmux integration. Cons: tmux not guaranteed in CI; flaky under parallel execution (server contention); contradicts the unit-mocking approach used everywhere else in the harness.
3. Refactor scripts to extract pure logic into a sourced lib — Pros: easier to test. Cons: scope creep beyond the stated goal; risks regressions in working production scripts; no demonstrated benefit since mocks already isolate the impure parts.

Implementation Plan

• Files to create:
  • /home/runner/work/shipwright/shipwright/scripts/sw-tmux-role-color-test.sh
  • /home/runner/work/shipwright/shipwright/scripts/sw-tmux-status-test.sh
  • /home/runner/work/shipwright/shipwright/scripts/sw-tracker-github-test.sh
• Files to modify:
  • /home/runner/work/shipwright/shipwright/package.json — append three entries to the test script (e.g., && ./scripts/sw-tmux-role-color-test.sh && ./scripts/sw-tmux-status-test.sh && ./scripts/sw-tracker-github-test.sh)
• Dependencies: none (bash, jq, mktemp already required by existing harness)
• Risk areas:
  • sw-tmux-status.sh may execute side effects at source-time; needs a SW_TMUX_STATUS_NO_DISPATCH=1 style guard or wrapping the dispatch in if [ "${BASH_SOURCE[0]}" = "$0" ]. If the script lacks this guard, the test must invoke functions only via subprocess with controlled argv, not source.
  • Heartbeat mtime differs between stat -c (Linux) and stat -f (macOS); tests must use touch -t to set predictable times and assert on bucket boundaries the script itself defines, not raw mtime math.
  • sw-tracker-github.sh lacks set -euo pipefail; sourcing it must not enable pipefail in the test shell or unrelated assertions may fail. Source inside ( set +e; . script ) subshells.
  • Parallel npm test: every temp path must derive from mktemp -d — no fixed /tmp/sw-test-* paths. MOCK_*_LOG vars set per-test, not exported globally.

Validation Criteria

• All three new test scripts exit 0 in isolation: ./scripts/sw-tmux-role-color-test.sh && ./scripts/sw-tmux-status-test.sh && ./scripts/sw-tracker-github-test.sh
• npm test passes end-to-end with all three suites registered (no regressions in existing 140+ suites)
• Each suite reports ≥12/16/17 tests via the standard PASS=/FAIL= counters with green output on success
• Running the same suite twice in parallel (./script.sh & ./script.sh & wait) succeeds — proves no shared-path collisions
• bash --posix-style 3.2 features only: grep -E "\\$\\{[^}]*,,\\}|declare -A|readarray" scripts/sw-tmux-*-test.sh scripts/sw-tracker-github-test.sh returns no matches
• Each provider_* function in sw-tracker-github.sh has at least one positive assertion and one error-path assertion
• All role keywords in sw-tmux-role-color.sh (builder, reviewer, tester, optimizer, docs, security, default) covered by a test case
• Mock log files asserted non-empty before content checks (avoids false positives when mock isn't invoked)

---

Generated by shipwright pipeline design stage at 2026-05-09T02:21:01Z

[github-actions]

[github-actions]

🔨 Build started — shipwright loop with 20 max iterations, 1 agent(s), model: opus

[github-actions]

[github-actions]

[github-actions]