How can I generate coverage.json without using an embedded project?

[Deep-co-de]

I’m trying to create a coverage.json compatible with Pacemaker, but my project isn’t an embedded project. The Pacemaker README.MD suggest using:

embedded-runner collect coverage.json

However, that doesn’t apply to my project. I’ve tried the following workflow instead:

cargo install mantra
cargo clean

cargo llvm-cov test
cargo llvm-cov report --json > coverage-report.json
mantra collect mantra.toml

My mantra.toml has this section:

[coverage]
files = ["coverage-report.json"]

When I run mantra collect mantra.toml, I get:

'79' requirements kept.
'2' traces kept.
Failed to update coverage data. Cause: missing field `test_runs` at line 1 column 32168419

It seems that mantra expects a coverage.json with a specific format (including a test_runs field), which cargo llvm-cov report --json does not produce.

So the question is: How can I generate a compatible coverage.json for mantra without using an embedded project?

[mhatzl]

Hi,
thanks for checking out mantra.

Getting test and coverage data in a format mantra understands is unfortunately a bit tricky at the moment. You need to follow the JSON schema defined by the struct here https://github.com/mhatzl/mantra/blob/main/schema/src/coverage.rs

It is not just taking code coverage, but also test output. That's why you see "test-run" in your error.

I am working on a new mantra version that will be able to take in JUnit and Cobertura formats. This will allow you to use cargo-nextest and llvm-cov/grcov. Hopefully I have this ready in a few months. Until then, you can try to convert your data to fit the current schema.

[Deep-co-de]

Hi,

thank you for the quick and thorough explanation! I really appreciate you clarifying why the current schema requires both test and coverage data, and for pointing me to the schema definition. It’s also exciting to hear about the upcoming version supporting JUnit and Cobertura formats—looking forward to that!

I’ll take a closer look at converting my data to fit the current schema in the meantime. Thanks again for your guidance.

[mhatzl]

just as a heads up, I haven't teste the covered-lines field that extensively. so there might be some bugs. the covered-traces one should be fine.