feat(coverage-gate): add target-specific policies

[martin-kolinek]

🤖

Summary

• add Cargo-style target policy selectors to cargo-coverage-gate
• support target-specific threshold overrides and no-coverable-lines assertions
• keep every selected test binary instrumented so cross-package coverage is preserved
• accept empty lcov input according to each package's effective policy
• bump cargo-coverage-gate to 0.4.0

Motivation

Addresses the coverage policy gaps reported in:

• ADO 7790459
• ADO 7790460

expect-no-coverable-lines remains appropriate for facades and re-export crates: those packages stay instrumented, their tests can cover dependencies, and the gate asserts that the package itself owns no coverable lines. Platform-gated packages can apply that assertion only on unsupported targets while retaining a positive threshold on supported targets.

Thresholds control verdicts, not instrumentation. A target-specific zero threshold disables gating without removing the package's tests from the instrumented run. If cargo-llvm-cov reports that an instrumented run produced no coverage data, an orchestrator can pass an empty lcov tracefile and let the effective policies determine whether the result is valid.

Cargo Anvil adoption is intentionally deferred until 0.4.0 has been published to crates.io; otherwise its generated setup would pin an unavailable version.

Configuration

[package.metadata.coverage-gate]
min-lines-percent = 100

[package.metadata.coverage-gate.target.'cfg(not(windows))']
expect-no-coverable-lines = true

Exact target triples take precedence over matching cfg(...) selectors; ambiguous matching cfg policies fail closed.

[Codecov Comments Bot (codecov-commenter)]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 97.7%. Comparing base (95089c5) to head (eaf6995).

❌ Your project status has failed because the head coverage (97.7%) is below the target coverage (100.0%). You can increase the head coverage or adjust the target coverage.

Additional details and impacted files

@@          Coverage Diff           @@
##            main    #101    +/-   ##
======================================
  Coverage   97.7%   97.7%            
======================================
  Files        286     287     +1     
  Lines      62174   62363   +189     
======================================
+ Hits       60751   60942   +191     
+ Misses      1423    1421     -2     

Flag	Coverage Δ
linux	97.6% <100.0%> (?)
linux-arm	97.6% <100.0%> (?)
windows	97.9% <100.0%> (?)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[martin-kolinek]

Why introduce a new enabled flag when we already have min-lines-percent and expect-no-coverable-lines?

[martin-kolinek]

🤖 Agreed — enabled was redundant with the existing policy keys. I removed it in 0070062.

Target opt-outs now use min-lines-percent = 0; a positive target-specific threshold restores gating for an exact-triple override, and expect-no-coverable-lines = true remains the self-validating assertion. I updated the design, generated README, parser, tests, and metadata-only package query accordingly.

[Copilot]

Pull request overview

This PR extends cargo-coverage-gate with Cargo-style target selectors so packages can vary coverage policy by target triple / cfg(...), and adds CLI + library surfaces to evaluate policy for an explicit target and to query “test-only” (non-instrumented) packages.

Changes:

• Add target-policy parsing and application during workspace metadata load (exact triple precedence; ambiguous matching cfg policies fail closed).
• Add --target and a metadata-only --print-test-only-packages mode (plus a new test_only_packages library API).
• Introduce target discovery/matching via cargo-platform and a new TargetContext module, and bump cargo-coverage-gate to 0.4.0.

Reviewed changes

Copilot reviewed 12 out of 13 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
crates/cargo-coverage-gate/tests/cli.rs	Adds CLI integration tests for --target and --print-test-only-packages, plus workspace test helpers.
crates/cargo-coverage-gate/src/workspace.rs	Implements parsing + resolution of target-specific policies during workspace load, including ambiguity rejection.
crates/cargo-coverage-gate/src/verdict.rs	Exposes resolve_gated internally so it can be reused by the new metadata-only query API.
crates/cargo-coverage-gate/src/target.rs	New module to resolve a target triple + rustc --print cfg and match against cargo-platform selectors.
crates/cargo-coverage-gate/src/lib.rs	Updates public API/docs: adds evaluate_many_for_target and test_only_packages, wires workspace loading with target.
crates/cargo-coverage-gate/src/error.rs	Adds error types for invalid/ambiguous target policies and target resolution failures.
crates/cargo-coverage-gate/src/bin/cargo-coverage-gate/run.rs	Adds early-exit --print-test-only-packages flow and passes target into evaluation.
crates/cargo-coverage-gate/src/bin/cargo-coverage-gate/cli.rs	Defines --target and --print-test-only-packages CLI flags and help text.
crates/cargo-coverage-gate/README.md	Regenerates user-facing docs to describe target policies and new CLI/API surfaces.
crates/cargo-coverage-gate/docs/design/README.md	Updates design doc to specify target policy schema, precedence rules, and new CLI behavior.
crates/cargo-coverage-gate/Cargo.toml	Bumps crate version to 0.4.0 and adds cargo-platform dependency.
Cargo.toml	Adds cargo-platform as a workspace dependency.
Cargo.lock	Updates lockfile for the new dependency and crate version bump.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[Sander Saares (sandersaares)]

[Copilot speaking]

Published 25 findings. No finding follows up on an existing discussion thread.

See diagnostics

Diagnostic	Value
Cache	Miss

[Copilot]

Pull request overview

Copilot reviewed 12 out of 13 changed files in this pull request and generated 1 comment.

Suppressed comments (1)

crates/cargo-coverage-gate/src/workspace.rs:234

• The PR description and config example mention a target-specific enabled = false, but target policy parsing here only accepts min-lines-percent or expect-no-coverable-lines = true (and errors if neither is set). As-is, configs using enabled = false will fail to load; either implement enabled in target policy tables (and define how it interacts with threshold/expect), or update the PR description/example to match the actual supported schema.

            let policy = if expect_no_coverable_lines {
                PolicyOverride::ExpectNoCoverableLines
            } else if let Some(value) = min_lines_percent {
                PolicyOverride::Threshold(value)
            } else {

[Copilot]

Pull request overview

Copilot reviewed 12 out of 13 changed files in this pull request and generated no new comments.

Suppressed comments (1)

Previously missed (1) — in code that hasn't changed since the last review.

crates/cargo-coverage-gate/src/verdict.rs:157

• resolve_gated is only used within this module, so widening it to pub(crate) is unnecessary and increases the internal surface area for no clear benefit. Keeping it private makes refactors easier and prevents accidental cross-module coupling.

pub(crate) fn resolve_gated<'w>(workspace: &'w Workspace, packages: &[String]) -> Result<Vec<&'w Member>, CoverageGateError> {

[Copilot]

Pull request overview

Copilot reviewed 12 out of 13 changed files in this pull request and generated no new comments.

Suppressed comments (1)

Previously missed (1) — in code that hasn't changed since the last review.

crates/cargo-coverage-gate/src/verdict.rs:157

• resolve_gated was widened to pub(crate) but appears to still be used only within verdict.rs. Keeping it private reduces the API surface inside the crate and makes future refactors easier (callers can’t accidentally start depending on it).

pub(crate) fn resolve_gated<'w>(workspace: &'w Workspace, packages: &[String]) -> Result<Vec<&'w Member>, CoverageGateError> {

[Copilot]

Pull request overview

Copilot reviewed 12 out of 13 changed files in this pull request and generated 1 comment.

[Copilot]

Pull request overview

Copilot reviewed 13 out of 14 changed files in this pull request and generated no new comments.

Suppressed comments (1)

Previously missed (1) — in code that hasn't changed since the last review.

crates/cargo-coverage-gate/src/verdict.rs:157

• resolve_gated is only used within verdict.rs (the only call site is evaluate(...) above), so making it pub(crate) unnecessarily expands the crate-internal surface area and makes refactors harder. Keeping it private avoids accidental cross-module coupling.

pub(crate) fn resolve_gated<'w>(workspace: &'w Workspace, packages: &[String]) -> Result<Vec<&'w Member>, CoverageGateError> {

[Copilot]

Pull request overview

Copilot reviewed 13 out of 14 changed files in this pull request and generated 1 comment.