test: raise coverage to 100% (and fix output_to_json metric gating)

[zkoppert]

Pull Request

Proposed Changes

stale-repos ships with --cov-fail-under=80, and the working tree sat at ~90% line coverage with eleven branches uncovered across auth.py, env.py, and stale_repos.py. Raising coverage to 100% also matches the in-flight 100%-coverage push across the rest of the github-community-projects fleet.

While writing the masking-free tests I found a latent bug in output_to_json that all three review models independently flagged: output_to_json was gating the optional daysSinceLastRelease / daysSinceLastPR JSON fields on literal "release" / "pr" keys in the per-repo dict, but set_repo_data (the only production producer of those dicts) only ever writes "days_since_last_release" / "days_since_last_pr". The result: the JSON output, including the value written to $GITHUB_OUTPUT, silently dropped both fields for every real caller, even when ADDITIONAL_METRICS was configured. Markdown output was unaffected because it gates on additional_metrics directly. I fixed this tightly-coupled bug in a separate commit by threading additional_metrics into output_to_json and mirroring the markdown writer's gate.

What changed

• I added unit tests covering every previously-uncovered branch (see the per-file breakdown below).
• I added a [report] exclude_lines block to .coveragerc covering pragma: no cover (already implicit) and if __name__ == .__main__.:.
• I bumped --cov-fail-under from 80 to 100 in the Makefile.
• I fixed output_to_json to gate optional JSON fields on additional_metrics (separate commit), matching the markdown writer's semantics. The main() caller now passes additional_metrics to output_to_json.

New test coverage (per uncovered branch)

Module	Lines	New test
auth.py	47	test_auth_to_github_raises_when_connection_is_none
env.py	118-119	test_get_env_vars_loads_dotenv_when_not_test
stale_repos.py	148-149	test_exempt_repos_env_var_is_parsed_and_applied
stale_repos.py	190-193	test_returns_none_on_type_error (get_days_since_last_release)
stale_repos.py	194-195	test_returns_none_when_no_releases
stale_repos.py	210-211	test_returns_none_when_no_pull_requests
stale_repos.py	234	test_unsupported_activity_method_raises_value_error
stale_repos.py	277, 279	test_includes_release_and_pr_keys_when_additional_metrics_set + test_release_only_metric_omits_pr_field
stale_repos.py	286-287	test_writes_to_github_output_when_env_var_set
stale_repos.py	327-328	test_github_exception_on_release_is_logged
stale_repos.py	335-336	test_github_exception_on_pr_is_logged

Before / after coverage

Module	Baseline	After
auth.py	94% (16 stmts, 1 miss)	100%
env.py	96% (45 stmts, 2 miss)	100%
markdown.py	100%	100%
stale_repos.py	84% (119 stmts, 19 miss)	100%
Total	89.81%	100.00%

Readiness Checklist

Author/Contributor

• If documentation is needed for this change, has that been included in this pull request
• run make lint and fix any issues that you have introduced
• run make test and ensure you have test coverage for the lines you are introducing

Testing

• make lint - all 5 linters pass (flake8 syntax + warnings, isort, pylint 10.00/10, mypy, black). No suppression comments added.
• make test - 53 tests pass (6 sub-tests), Required test coverage of 100% reached. Total coverage: 100.00%. Verified on the Python 3.11 baseline used by the Makefile and on the Python 3.12 fallback that uv selected on this machine.
• Multi-model code review (Opus, Sonnet, GPT-5.5) run on the staged diff before opening. All three independently surfaced the output_to_json gating bug; the fix landed in this PR as a separate commit.

Rollout

• No production behavior change beyond the output_to_json bug fix. With ADDITIONAL_METRICS unset (the default), the JSON output and $GITHUB_OUTPUT value are byte-identical to today. With ADDITIONAL_METRICS=release or pr, the JSON now correctly includes daysSinceLastRelease / daysSinceLastPR, which downstream consumers (e.g., scripts that read $GITHUB_OUTPUT) may have been silently missing.
• The threshold bump (--cov-fail-under=80 → 100) means any future PR that drops coverage will fail CI rather than slipping in unnoticed. This is the desired behavior and matches the fleet convention.

Tradeoffs / Alternatives considered

• For the output_to_json fix I considered (a) gating on "days_since_last_release" in repo_data (always true post set_repo_data, would force the field into the output of callers who don't opt in - silent behavior change) and (b) threading additional_metrics through (chosen, mirrors markdown.py, no behavior change for non-opt-in callers).
• For the if __name__ == "__main__": line I considered adding # pragma: no cover inline instead of an exclude_lines entry. The .coveragerc rule is more discoverable and matches conventions in other repos in the fleet.

[Copilot]

Pull request overview

This PR raises the project’s CI coverage bar to 100% by adding targeted unit tests for previously uncovered branches, and it fixes a production bug where output_to_json silently omitted optional daysSinceLastRelease / daysSinceLastPR fields even when ADDITIONAL_METRICS was configured. The JSON gating logic is aligned with the markdown writer by explicitly threading additional_metrics into output_to_json from main().

Changes:

• Add unit tests to cover previously uncovered branches in auth.py, env.py, and stale_repos.py.
• Fix output_to_json to include optional JSON fields based on additional_metrics (matching markdown semantics).
• Enforce 100% coverage (--cov-fail-under=100) and exclude __main__ blocks via .coveragerc.

Show a summary per file

File	Description
test_stale_repos.py	Adds branch-coverage tests for exempt repos parsing, exception paths, JSON optional-metric output, $GITHUB_OUTPUT writing, and set_repo_data exception handling.
test_env.py	Adds coverage for loading .env via load_dotenv when test=False.
test_auth.py	Adds coverage for the github3.login(...) returns None failure path.
stale_repos.py	Fixes JSON optional-metric gating by passing and honoring additional_metrics in output_to_json.
Makefile	Raises coverage enforcement threshold from 80% to 100%.
.coveragerc	Adds report-time exclusions for pragma: no cover and if __name__ == "__main__":.

Copilot's findings

• Files reviewed: 6/6 changed files
• Comments generated: 0