codecov.yml

# Codecov configuration.
#
# Coverage is produced by the `coverage` job in
# .github/workflows/ci.yml, which uploads codecov.json (region
# coverage) generated by cargo-llvm-cov.

coverage:
  status:
    # Informational while a baseline settles: Codecov posts the project /
    # patch delta as a commit status and a PR comment, but never fails the
    # check on a dip. Flip `informational` to false once the baseline is
    # trusted and coverage should gate merges.
    project:
      default:
        informational: true
    # A fixed floor rather than `auto` (which targets the project figure).
    # `auto` is unreachable by construction for a grammar-dispatch patch:
    # `AGENTS.md` bans `unwrap` / `expect` in non-test code, so every
    # infallible node lookup is spelled `?` — `node.child(1)?` on a
    # `parenthesized_expression`, `child_by_field_name("operand")?` on a
    # `unary_expression` — and each mints a `None` region no valid input can
    # reach. Partials dominate the miss list for that reason (12 of 19 on
    # PR #1476). Raising the floor is the fix; adding an `expect` to win the
    # region would trade an uncoverable branch for a panic in a library that
    # parses untrusted source.
    #
    # Second thing to know before reading any patch number here: Codecov
    # silently drops a file whose diff GitHub's API refuses to send (over
    # ~3,000 changed lines), so a heavily-edited file is *absent* from the
    # patch rather than counted as missing. The tell is `<ø>` in the
    # per-file table instead of a percentage. On PR #1476 that removed
    # `src/metrics/abc.rs` — the largest changed file — from the
    # denominator entirely (283/303 lines). Production changes to a file of
    # that size are invisible to patch coverage; judge them on the covered-
    # region delta instead.
    patch:
      default:
        target: 90%
        threshold: 2%
        informational: true

comment:
  layout: "reach, diff, flags, files"
  # Skip the comment entirely when a PR does not change coverage, to keep
  # noise down on docs-only / CI-only changes.
  require_changes: true

# Coverage arrives under two flags from two jobs: `rust` (cargo-llvm-cov,
# always runs) and `python` (pytest-cov, only on PRs that touch the
# bindings). carryforward keeps the python number from vanishing on
# Rust-only PRs, where the python-test job is skipped.
#
# Measurement gap (intentional): the `python` flag measures only the pure-
# Python surface via pytest-cov. The PyO3 *Rust* layer in
# big-code-analysis-py/src/ (lib.rs, vcs.rs, node.rs, ast.rs, sarif.rs) is
# exercised by those same pytest suites, but the crate has no `#[test]`s, so
# `cargo llvm-cov nextest` (which only runs cargo tests) never instruments
# it. Its standalone Rust-line coverage therefore reads far lower than it is
# actually tested — do not "fix" that number by adding redundant cargo tests
# that re-cover behaviour pytest already drives. Closing it for real means
# instrumenting the maturin/pytest run and merging that profdata, tracked
# separately.
flag_management:
  default_rules:
    carryforward: true

# The python coverage report is written from big-code-analysis-py/ with
# relative paths (python/big_code_analysis/...). Re-root them at the repo
# so Codecov attributes lines to the checked-in files. Rust paths are
# already repo-rooted and are unaffected (the prefix does not match).
fixes:
  - "python/big_code_analysis/::big-code-analysis-py/python/big_code_analysis/"

# Exclude vendored grammars, code-generation helpers, and test/build
# harnesses from the coverage denominator so the reported number reflects
# the library's own measurable surface.
ignore:
  - "tree-sitter-ccomment/**"
  - "tree-sitter-mozcpp/**"
  - "tree-sitter-mozjs/**"
  - "tree-sitter-preproc/**"
  - "tree-sitter-tcl/**"
  - "enums/**"
  # Belt and braces (#1154). `fuzz/` is workspace-excluded, so
  # `cargo llvm-cov --workspace` cannot reach it and no line of it can
  # enter either denominator in the first place — but naming it here
  # keeps that true if the crate is ever pulled into the workspace, and
  # a fuzz target has no meaningful coverage number: it is run by
  # libFuzzer, not by the test suite.
  - "fuzz/**"
  - "xtask/**"
  # The benchmark harness, for the same reason as xtask: dev tooling this
  # repo authors but never ships, and criterion drives it rather than the
  # test suite.
  - "big-code-analysis-bench/**"
  # The PyO3 crate's *Rust* layer, for the reason the measurement-gap note
  # above already gives: pytest exercises it and `cargo llvm-cov nextest`
  # cannot see that, so it measures ~52% while being far better tested than
  # that implies. Reporting the misleading number would only invite the
  # redundant cargo tests that note warns against. The pure-Python surface
  # is still measured under the `python` flag.
  - "big-code-analysis-py/src/**"
  - "**/tests/**"
  - "**/build.rs"