René's URL Explorer Experiment

Title: ci: add strict-no-cover to detect unnecessary coverage pragmas by maxisbey · Pull Request #1897 · modelcontextprotocol/python-sdk · GitHub

Open Graph Title: ci: add strict-no-cover to detect unnecessary coverage pragmas by maxisbey · Pull Request #1897 · modelcontextprotocol/python-sdk

X Title: ci: add strict-no-cover to detect unnecessary coverage pragmas by maxisbey · Pull Request #1897 · modelcontextprotocol/python-sdk

Description: Add strict-no-cover from pydantic to the CI pipeline. This tool identifies # pragma: no cover comments on lines that are actually covered by tests, helping keep coverage pragmas accurate and preventing them from accumulating as dead annotations. Motivation Over time, # pragma: no cover annotations accumulate on lines that are actually exercised by tests — the original reason for the pragma may no longer apply (e.g., code was refactored, tests were added). These stale pragmas hide real coverage data and make it harder to identify genuinely uncovered code. Changes CI integration (e66af17) Add strict-no-cover as a dev dependency (installed from git) Add CI step to run strict-no-cover after pytest coverage (Linux only — the tool has a Windows path-handling bug) Pin coverage[toml] to >=7.10.7,<=7.13 because 7.13.1 introduced a regression in how source paths are reported in .coverage databases, breaking strict-no-cover Remove relative_files = true from [tool.coverage.run] as it interferes with strict-no-cover path matching Add pragma: lax no cover and pragma: no branch to exclude_lines in coverage config Pragma cleanup (f0c0b6e) Remove # pragma: no cover from ~150 lines that are actually covered by tests Add # pragma: lax no cover for platform-specific code (Windows skipif markers, sys.platform branches, OS-specific exception handlers) Add # pragma: no branch for branches where the exit path is never taken (async for loops, anyio.fail_after blocks, finally clauses) Pragma guide Pragma Meaning Use when # pragma: no cover Line must never be covered Strict — CI fails if covered # pragma: lax no cover Line may or may not be covered Platform/version-specific code, flaky coverage # pragma: no branch Branch exit not taken async for, with fail_after(), loops that always break How it works After coverage run -m pytest and coverage combine, the CI runs strict-no-cover which: Reads the .coverage database to find which lines were actually executed Scans source files for # pragma: no cover annotations Reports any lines that have the pragma but ARE covered (meaning the pragma is unnecessary) Exits non-zero if any are found, failing the CI check # pragma: lax no cover lines are excluded from this check, making them safe for code that is covered in some CI configurations but not others.

Open Graph Description: Add strict-no-cover from pydantic to the CI pipeline. This tool identifies # pragma: no cover comments on lines that are actually covered by tests, helping keep coverage pragmas accurate and preven...

X Description: Add strict-no-cover from pydantic to the CI pipeline. This tool identifies # pragma: no cover comments on lines that are actually covered by tests, helping keep coverage pragmas accurate and preven...

Opengraph URL: https://github.com/modelcontextprotocol/python-sdk/pull/1897

X: @github

direct link

Domain: github.com

Links:

Viewport: width=device-width