dispatch/codecov.yml at main · Jamyn/dispatch · 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
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
# Codecov is advisory here: none of these statuses are in the branch ruleset's
# required checks, so a coverage drop annotates the PR without blocking it.
# Promoting one to required is a ruleset change, not a change to this file.

coverage:
  status:
    # `auto` compares against the PR base rather than an absolute number, so a
    # historically under-covered file never fails a PR that does not touch it.
    # The threshold is what makes this a regression gate instead of noise.
    project:
      default:
        target: auto
        threshold: 1%
    # Coverage of the lines this PR actually changed. Held to a real number
    # because it is the only figure the author can act on directly.
    patch:
      default:
        target: 80%
        threshold: 0%
      # The frontend has ten spec files against 428 components. Holding new
      # frontend lines to the backend's 80% would fail nearly every UI PR, so
      # it tracks its own base until there is a suite worth holding to a number.
      frontend:
        flags: [frontend]
        target: auto
        threshold: 0%

# Path-based, so these cost no extra upload and no CI change -- they slice the
# report python-test already produces. Paths carry the src/dispatch prefix
# because coverage.xml names files relative to <source>src/dispatch</source>
# and Codecov joins the two; without it every component matches nothing.
#
# No component statuses on purpose. Making three more checks fail before anyone
# has agreed these numbers should move is how coverage gates get ignored.
component_management:
  individual_components:
    # The orchestration layer -- incident and case lifecycle. Its own component
    # because */models.py is 95.5% (declarative classes, covered by being
    # imported) and drags the project figure well above what the logic scores.
    - component_id: flows
      name: flows
      paths: ["src/dispatch/**/flows.py"]

    # Provider integrations. Six of roughly seventeen provider packages have a
    # test suite, so this moves for reasons the project figure hides.
    - component_id: plugins
      name: plugins
      paths: ["src/dispatch/plugins/**"]

    # Chosen for consequence rather than size: there is no tests/auth directory.
    - component_id: auth
      name: auth
      paths: ["src/dispatch/auth/**"]

# Two uploads now, from two jobs, so the split is real rather than decorative.
# Judged separately: the frontend is 10% of a codebase larger than the backend,
# and a blended figure would let either side hide behind the other.
# No `paths:` on either -- a flag covers whatever its upload contained, and the
# two reports are disjoint. Scoping backend to src/dispatch/ would also match
# src/dispatch/static/dispatch/src/, i.e. the frontend. carryforward is off
# because both jobs run on every PR, so a report is never missing.
flags:
  backend:
    carryforward: false
  frontend:
    carryforward: false

comment:
  layout: "condensed_header, diff, flags, components, files"
  # Nothing to say on a PR that changes no covered line.
  require_changes: true

# The suite runs under pytest-xdist and pytest-cov combines the workers into a
# single report before upload, so exactly one file arrives per commit.
codecov:
  require_ci_to_pass: true