mkit/codecov.yml at main · officialunofficial/mkit · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
# Codecov config. https://docs.codecov.com/docs/codecovyml-reference
#
# mkit policy:
# - Coverage trend is informational, not a merge gate. We don't want a
#   well-tested PR blocked because some path stayed at the same %.
# - Target is the project's own moving baseline (no fixed %), with a
#   small tolerance so trivial doc/refactor PRs don't trip the patch
#   threshold.
# - Generated code, fuzz harnesses, and benchmarks are excluded from
#   the denominator — they're either not real surface or already
#   exercised by their own dedicated runs.

codecov:
  notify:
    # Wait for the rust.yml CI to finish before commenting so the PR
    # gets one coverage comment, not one per matrix job. Empty list
    # means "wait for all expected coverage uploads."
    after_n_builds: 1
    wait_for_ci: true
  require_ci_to_pass: false  # we report regardless; CI gate is rust.yml

coverage:
  precision: 1
  round: nearest
  range: "60...90"   # red → green gradient on the badge
  status:
    project:
      default:
        target: auto         # follow the moving baseline on main
        threshold: 1%        # tolerate ±1% project-wide on each PR
        if_ci_failed: success
        only_pulls: false
    patch:
      default:
        target: 80%          # new code should ship at ≥80% coverage
        # Feature-flagged transport + crypto glue (enc-transport,
        # pack-shards, sparse-checkout, history-mmr, bls-threshold) is
        # hard to unit-test exhaustively — real TCP, runtime bootstrap,
        # CLI dispatch wrappers. 10% absorbs the unavoidable
        # uncovered glue without giving up the 80%-aspirational target.
        threshold: 10%       # tolerate up to 10pp below target
        if_ci_failed: success

# Files / paths excluded from the coverage denominator. None of these
# are "production code we have a regression-test obligation for."
ignore:
  - "rust/benches/**"
  - "rust/fuzz/**"
  - "rust/tests/golden/**"          # frozen test vectors, not code
  - "rust/target/**"
  - "**/build.rs"                   # pure codegen drivers
  - "**/tests/**"                   # the tests themselves
  - "contrib/signers/mkit-sign-se/**"  # Swift; not Rust-tooled
  - "rust/crates/*/src/**/__generated/**"  # buffa-emitted code

comment:
  layout: "header, diff, files, footer"
  behavior: default
  require_changes: true   # only comment if the PR moves coverage
  require_base: false
  require_head: true