From f6ef5dc41d91b3bec3636c75a1f55761eb5bd87a Mon Sep 17 00:00:00 2001 From: Jon Langevin Date: Tue, 19 May 2026 16:25:44 -0400 Subject: [PATCH] ci: add codecov.yml with informational patch coverage MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Workflow main has many operator-facing CLI surfaces that get exercised by manual operators or downstream integration tests, not by unit tests. Codecov's default "auto" patch target blocks PRs whose helper functions can't be cheaply unit-tested without a full integration harness — admin-override is forced 3 times in the past week (workflow#715, workflow#720, workflow#725). Switch patch coverage to informational mode (report posted but status check passes) and set project threshold to 1% (protects against gradual rot without blocking noise). Reviewers continue to call out genuinely under-tested paths in review. Co-Authored-By: Claude Opus 4.7 (1M context) --- codecov.yml | 35 +++++++++++++++++++++++++++++++++++ 1 file changed, 35 insertions(+) create mode 100644 codecov.yml diff --git a/codecov.yml b/codecov.yml new file mode 100644 index 00000000..627a9b8f --- /dev/null +++ b/codecov.yml @@ -0,0 +1,35 @@ +# Codecov configuration for GoCodeAlone/workflow. +# +# Workflow is a large Go codebase with many code paths that exist to support +# operator-facing CLI surfaces — they get exercised by manual operators or by +# downstream integration tests, not by unit tests. Codecov's default "auto" +# patch threshold therefore blocks PRs whose helper functions can't be +# cheaply unit-tested without a full integration harness. +# +# Policy: +# - patch: informational only — codecov posts the report but does NOT fail +# the status check. Reviewers still see the patch coverage in the PR +# comment so they can call out under-tested paths in review. +# - project: 1% threshold — repo-wide coverage can dip by 1 percentage +# point before the project status fails. Protects against gradual rot +# without blocking on noise. +# +# When a PR adds significant under-tested code, reviewers should flag it +# directly rather than relying on a hard codecov threshold. This matches +# the manual-review tier in CONTRIBUTING.md. + +coverage: + status: + project: + default: + target: auto + threshold: 1% + informational: false + patch: + default: + informational: true + +comment: + layout: "reach, diff, flags, files" + behavior: default + require_changes: false