test: 100% line coverage + CI coverage gate

[AkashBrowserStack]

What & why

This SDK had no coverage enforcement in CI and sat at 86% locally. This PR raises it to 100% line coverage with meaningful tests and wires a hard CI gate so coverage can't silently regress.

Every line is executed by a real unit test — no # pragma skips, no coverage-gaming. Even the defensive optional-import fallbacks are exercised via reload-based tests rather than excluded.

Tests added

Area	Coverage added
snapshot.py (new test_snapshot_units.py)	log() CLI-ship failure; _wait_for_ready script-timeout adjust/restore failures; CORS-iframe origin-error skip; get_responsive_widths non-list + request failure; CDP resize → set_window_size fallback; _responsive_sleep invalid/unset; capture_responsive_dom invalid minHeight
percy/__init__.py (new test_init.py)	percy_screenshot dispatch (unsupported driver, RemoteConnection → automate, AppiumConnection delegation when appium installed + helpful error when not, unknown → None); and reload-based tests that execute the robot-library and percy_snapshot optional-import fallbacks
robot_library.py	parse-helper edge cases, full _parse_padding matrix, _get_driver missing-SeleniumLibrary error, Percy Screenshot keyword (+ region elements), reload-based test of the no-robotframework stub
driver_metadata.py	newer-Selenium client_config.remote_server_addr fallback

Instrumentation / gate

• .coveragerc — source = percy, fail_under = 100, show_missing.
• coverage + pytest-cov added to development.txt.
• New make coverage runs every module (snapshot under percy exec --testing), combines parallel data, and enforces the threshold.
• test.yml runs make coverage across the 3.8–3.10 matrix; new modules also added to make test.

Verification

make coverage -> TOTAL 514 stmts 0 miss 100%  (no pragma exclusions)  exit 0
pylint: 10.00/10

🤖 Generated with Claude Code

[AkashBrowserStack]

Claude Code PR Review

PR: #238 • Head: 083ef0e • Reviewers: stack:code-reviewer

Summary

Adds focused unit tests for the package entrypoint, snapshot.py helper/fallback paths, robot_library.py parsers and keywords, and a driver_metadata Selenium-version fallback; wires a coverage Make target plus .coveragerc with a 100% fail_under line-coverage gate and switches CI from make test to make coverage.

Review Table

Priority	Category	Check	Status	Notes
High	Security	No hardcoded secrets or credentials	Pass	Tests use only synthetic fixture values (fallback-hub, https://page); no real secrets.
High	Security	Authentication/authorization checks present	N/A	No auth surface; PR is test + CI tooling only.
High	Security	Input validation and sanitization	N/A	No new user-input boundaries introduced.
High	Security	No IDOR — resource ownership validated	N/A	No resource access in scope.
High	Security	No SQL injection (parameterized queries)	N/A	No database/SQL in scope.
High	Correctness	Logic is correct, handles edge cases	Pass	Tests faithfully target real branches (Selenium _url→remote_server_addr fallback, resize fallback, min-height parse failure, responsive widths upgrade hint, swallowed log/timeout failures). All 53 new/modified tests pass locally.
High	Correctness	Error handling is explicit, no swallowed exceptions	Pass	Test code raises/asserts explicitly; the swallow paths it exercises are pre-existing production behavior (e.g. best-effort log POST), not introduced here.
High	Correctness	No race conditions or concurrency issues	Pass	importlib.reload global-state mutations in test_init/test_robot_library are guarded by try/finally: importlib.reload(...) restore; combined run is order-independent and green.
Medium	Testing	New code has corresponding tests	Pass	This PR is the tests; the coverage/.coveragerc tooling is exercised by the CI make coverage switch.
Medium	Testing	Error paths and edge cases tested	Pass	Strong negative-path focus: unsupported types, import failures, non-numeric min-height/sleep, network failures, un-parseable iframe origin.
Medium	Testing	Existing tests still pass (no regressions)	Pass	Modified test_driver_metadata.py and test_robot_library.py add cases without altering existing ones; suites pass.
Medium	Performance	No N+1 queries or unbounded data fetching	N/A	Test code; no production data paths added.
Medium	Performance	Long-running tasks use background jobs	N/A	Not applicable to tests.
Medium	Quality	Follows existing codebase patterns	Pass	unittest + unittest.mock style consistent with existing test_cache.py/test_driver_metadata.py; pylint disables scoped narrowly.
Medium	Quality	Changes are focused (single concern)	Pass	Single concern: raise coverage + enforce it. .gitignore/development.txt additions support that concern.
Low	Quality	Meaningful names, no dead code	Pass	Descriptive test names; explanatory comments justify the reload/dispatch tricks. No dead code.
Low	Quality	Comments explain why, not what	Pass	Comments explain intent (why a class is named WebDriver, why imports are forced to fail).
Low	Quality	No unnecessary dependencies added	Pass	Only coverage==7.* added to development.txt (dev-only, standard tool).

Findings

No Critical or High findings. Minor (Low) observations, non-blocking:

• File: .coveragerc:6 and .github/workflows/test.yml:64

• Severity: Low

• Reviewer: stack:code-reviewer

• Issue: fail_under = 100 (line coverage) is strict. It is met today, but any future production line not reachable by a fast unit test will fail CI even if e2e-covered, since only line coverage is gated (no branch = True). This is the intended design of the PR, noted for awareness, not a defect.

• Suggestion: Acceptable as-is. If branch completeness is later desired, add branch = True under [run]; otherwise leave the 100% line gate.

• File: .gitignore:35

• Severity: Low

• Reviewer: stack:code-reviewer

• Issue: package-lock.json is newly ignored although a package-lock.json is already committed at repo root. Ignoring a tracked file has no effect on the tracked copy and can be mildly confusing.

• Suggestion: Either git rm --cached package-lock.json if it should not be tracked, or drop that ignore line. Cosmetic only.

• File: tests/test_snapshot_units.py:452-468

• Severity: Low

• Reviewer: stack:code-reviewer

• Issue: TestCaptureResponsiveDomMinHeight writes output_file.json to the CWD (production capture_responsive_dom opens it relative to CWD); cleanup is in tearDown, but a failure between write and teardown could leave the artifact. It is now in .gitignore, so impact is contained.

• Suggestion: Fine as-is given the .gitignore entry; optionally chdir into a tmp dir within the test for full isolation.

---

Verdict: PASS — tests are faithful to the real code paths, all 53 pass locally, and the coverage gate tooling is internally consistent.