Skip to main content

coverage

Generate or inspect source coverage for a selected Firedancer unit test.

ソース情報

リポジトリ
firedancer-io/firedancer
ソースの最終更新活動
2026年9月18日 21:25
検出された SKILL.md の言語
英語
スター
1,529
フォーク
535

インストール方法

デフォルトでは、最初にソースを確認する Prompt が選択されています。直接コマンドに切り替えるか、ローカルコピーをダウンロードすることもできます。

ソースファイルを確認

インストールを決める前に、SKILL.md と SkillsMP に表示されている付属ファイルをお読みください。

SKILL.md を表示中

SKILL.md
ソースの指示 · 読み取り専用プレビュー
name
coverage
description
Generate or inspect source coverage for a selected Firedancer unit test.
argument-hint
<test-name>
allowed-tools
["Bash(llvm-cov *)","Bash(CLANKER=1 ./contrib/test/single_test_cov.sh *)"]
Use the requested test, file, or function scope. ## Inspect existing coverage Read a supplied coverage report directly. For profile-based inspection, use the matching instrumented binary and source revision with the export commands below, substituting the supplied paths. State any compatibility or provenance gap; regenerate when fresh execution evidence is needed for the request. ## Generate or refresh coverage For a new report, build and run unit test `$0`. The helper writes `default.profraw` and `default.profdata` in the working directory; do not run concurrent coverage jobs there. For test MEMLOCK/huge-page prerequisites, use [testing guidance](../../../doc/testing.md); raise MEMLOCK in the same shell that runs the test. 1. Build: `make -j CC=clang EXTRAS=llvm-cov BUILDDIR=clang-cov $0` 2. Run: `CLANKER=1 ./contrib/test/single_test_cov.sh build/clang-cov/unit-test/$0` - produces `default.profdata` ## Export the requested view Use the existing artifact paths for inspection, or these defaults after a new run: - Per file: `llvm-cov show --instr-profile default.profdata build/clang-cov/unit-test/$0 -sources <src/foo/bar.c> -show-instantiations=false` - Per function: `llvm-cov show --instr-profile default.profdata build/clang-cov/unit-test/$0 --name=<function_name>` - All files overview: `llvm-cov report --instr-profile default.profdata build/clang-cov/unit-test/$0` For a new run, record the helper's exit status: it generates coverage when a profile is available, then returns the test's status (or fails if profile processing fails). Report a failed test even when coverage was generated; inspect that profile when useful for the requested investigation. For a coverage investigation, compare the requested behavior with the test and relevant report lines. A report-only request is complete once that report is produced; it does not require adding tests or a repository-wide coverage audit.
GitHubで見る