codecov.yml

# Codecov configuration.
#
# The real coverage gate is `patch`: changed lines and branches in a PR must be
# >=99% covered. Patch coverage depends only on the PR's own diff, so merging
# one PR can never push another below the bar -- this is what kills the cross-PR
# churn the old global vitest threshold caused.
#
# `project` (whole-repo total) is informational only: it is reported as a trend
# but never blocks a merge. vitest keeps a loose 90% local backstop separately.
codecov:
  # Don't post a verdict until the CI run that produced the report has finished.
  require_ci_to_pass: true
  notify:
    # Hold every status and comment until CI says the uploads are done, rather than guessing from a
    # count (#9801). `after_n_builds` fires on the Nth upload to ARRIVE, and arrival order does not
    # match importance: validate-code's rees / control-plane / engine reports land in ~2 minutes while
    # validate-tests is still running the backend lcov that covers most of src/**. So the verdict was
    # computed from a partial report -- codecov/patch went pass, then fail, while validate-tests was
    # still pending, with Codecov's own comment printing `backend | BASE 1 | HEAD 0`.
    #
    # That is not cosmetic here: the gate closes a contributor PR on ANY red check, codecov/patch
    # included, so a transient partial-report red can close a PR that has no defect.
    #
    # Raising the count cannot fix it -- see the reasoning preserved on after_n_builds below -- because
    # a count cannot express "wait for the backend upload specifically". manual_trigger can: Codecov
    # sends nothing until `codecovcli send-notifications` runs, and the codecov-notify job in
    # .github/workflows/ci.yml runs that once, after every uploading job has finished.
    #
    # Failure mode if the trigger never runs: codecov/patch simply never posts, so the gate sees an
    # incomplete check set and waits. That is fail-safe -- it cannot merge unreviewed and cannot
    # false-close -- which is the whole point of the change.
    manual_trigger: true
    # Inert while manual_trigger is true, and kept deliberately so removing manual_trigger restores the
    # correct value rather than the default. The count that ALWAYS lands whenever validate-tests runs:
    # since the 2026-07-24 unsharding, the backend suite produces ONE whole-suite lcov upload
    # (flags: backend) instead of the former 3 shard uploads this threshold was sized for. It must stay
    # 1, not 2, even though review-enrichment / control-plane add their own flag uploads on PRs that
    # touch them -- those uploads only exist on such PRs, and a floor above the guaranteed minimum would
    # leave codecov/patch permanently un-posted on every other PR.
    after_n_builds: 1

coverage:
  status:
    patch:
      default:
        target: 99%
        threshold: 0%
        # Keep the patch status in error if the upstream CI run failed.
        if_ci_failed: error
        # Suppress codecov/patch on main pushes; uploads still maintain history.
        only_pulls: true
    project:
      default:
        informational: true

# review-enrichment ships its coverage under its own flag (produced + uploaded by validate-code, ci.yml).
# Scoping the flag to review-enrichment/ keeps its project trend honest, and carryforward keeps that trend
# stable on PRs that don't re-run REES (its coverage is unchanged then, so nothing re-uploads under this flag).
# Patch gating (codecov/patch above) is unaffected by carryforward -- patch is computed on the PR's own diff,
# so a review-enrichment/** change is gated against this flag's fresh upload for that PR.
flags:
  rees:
    paths:
      - review-enrichment/
    carryforward: true
  control-plane:
    paths:
      - control-plane/
    carryforward: true
  # @loopover/engine's real behavior suite is its own node:test suite (packages/loopover-engine/test/**),
  # not vitest's root test/** -- that suite was invisible to Codecov entirely until now (#9064), even
  # though packages/loopover-engine/src/** is already in vitest's coverage.include and counted by the
  # unflagged `backend` upload for whatever a root test/** import happens to exercise. This flag is
  # additive to that: both reports' hits are unioned for the same lines, so a PR touching engine source
  # now gets credit from the suite that actually behavior-tests it, not just an incidental vitest import.
  engine:
    paths:
      - packages/loopover-engine/src/
    carryforward: true

comment:
  layout: "condensed_header, diff, flags, files"
  require_changes: false

# Coverage is collected by vitest (v8) over src/**, packages/loopover-engine/src/**,
# packages/loopover-miner/{lib,bin}/**, packages/loopover-mcp/{lib,bin}/**, and
# packages/discovery-index/src/** (plus review-enrichment/src/** via the standalone `rees` flag, c8 over
# its built dist remapped through source maps); mirror vitest's exclusions here so the Codecov total
# matches the local report.
ignore:
  - "src/env.d.ts"
  # Type-ONLY declaration modules (#9810): zero runtime statements, so v8 instruments nothing and the whole
  # file reports 0% — the same artifact src/env.d.ts is ignored for above. Adding a single field to one of
  # these otherwise fails codecov/patch on lines that can never execute. Guarded, not trusted: the
  # "codecov ignore list stays honest" test in test/unit/codecov-ignore-type-only.test.ts fails if any path
  # listed here ever gains a runtime declaration, so an ignore can never quietly start hiding real code.
  - "packages/loopover-engine/src/types/manifest-deps-types.ts"
  - "packages/loopover-engine/src/types/predicted-gate-types.ts"
  - "packages/loopover-engine/src/types/reward-risk-types.ts"
  - "apps/**"
  - "test/**"
  - "scripts/**"
  # Self-host process entry + build-time stubs: exercised by the Docker build+boot smoke test
  # (.github/workflows/selfhost.yml), not unit-coverable without booting a server/subprocess.
  - "src/server.ts"
  # review-enrichment's process entrypoint (#1473) -- same "Docker build+boot, not unit-coverable" reasoning
  # as src/server.ts above; module-scope serve() binds a real port as a side effect of importing this file,
  # so it's never imported by test/**. Pre-existing gap: this entry was missing even though
  # discovery-index/src/server.ts's own header comment already claimed review-enrichment/src/server.ts got
  # "the same treatment" -- surfaced by #8290 finally touching enough of this file's lines to trip
  # codecov/patch on it.
  - "review-enrichment/src/server.ts"
  - "src/selfhost/cf-workers-shim.ts"
  - "src/selfhost/stubs/**"
  # Postgres runtime adapters: validated by the real-Postgres integration test (test/integration/selfhost-pg.ts,
  # gated on PG_TEST_URL) + the real-PG boot. The dialect translation itself IS unit-tested (pg-dialect.ts).
  - "src/selfhost/pg-adapter.ts"
  - "src/selfhost/pg-queue.ts"
  # discovery-index's process entrypoint (#7164) -- same "Docker build+boot, not unit-coverable" reasoning
  # as src/server.ts above; app.ts (everything it wires together) is what tests actually import.
  - "packages/discovery-index/src/server.ts"
  # discovery-index's Cloudflare Container/Durable Object entry point (#7167) -- pure infra glue exercised
  # only by real Cloudflare Containers infrastructure, same reasoning as server.ts above.
  - "packages/discovery-index/src/worker.ts"
  - "packages/discovery-index/src/env.d.ts"
  - "packages/discovery-index/worker-configuration.d.ts"
  # NOTE: packages/loopover-miner/bin/loopover-miner.ts and (once #7291 migrates it)
  # packages/loopover-mcp/bin/loopover-mcp.ts are plain CLI dispatchers (no exports, terminate via
  # process.exit()) exercised only via real subprocess spawn (test/unit/support/{miner,mcp}-cli-harness.ts),
  # which v8 coverage can't instrument -- the same shape as src/server.ts above. They are deliberately NOT
  # listed here: test/unit/codecov-policy.test.ts (#4864) requires packages/loopover-miner to never get a
  # blanket ignore-list exemption, precisely so a change to the dispatcher forces either real coverage or a
  # testable-export refactor (the path bin/loopover-miner-mcp.ts's createMinerMcpServer already took)
  # instead of a silent pass. Both files are still in vitest.config.ts's coverage.include (so a PR that
  # touches one now genuinely needs to satisfy codecov/patch on it, not skip grading entirely).