Run ML-powered query pack ~0.3.0 on v2.9.3+ of the CLI

[TomBolton]

This PR starts running the following versions of the ML-powered query pack:

• Version ~0.3.0 of the query pack on v2.9.3+ of the CLI
• Version ~0.2.0 of the query pack on v2.8.4 <= CLI < v2.9.3
• Version ~0.1.0 of the query pack on CLI < v2.8.4

Merge / deployment checklist

• Confirm this change is backwards compatible with existing workflows.
• Confirm the readme has been updated if necessary.
• Confirm the changelog has been updated if necessary.

[TomBolton]

@henrymercer is this how we would like to start running ~0.3.0?

Specifically, I've assumed in this PR that if the CLI is below 2.9.3, but above 2.8.4, we use ~0.2.0; otherwise we use ~0.1.0. Is this correct?

[henrymercer]

This logic looks right to me, however to merge it we'll need to do two things:

1. Compile the TypeScript files to JavaScript — there are instructions for that here: https://github.com/github/codeql-action/blob/main/CONTRIBUTING.md#development-and-testing. You should end up checking in a src/util.ts and corresponding lib/util.js and lib/util.js.map files. Let me know if you have any problems!

2. Add some tests for this functionality in config-utils.test.ts to verify that the right pack gets loaded for 2.9.3. This commit shows how we did this for 0.2.0: e26813c

[aeisenberg]

This looks generally sensible. And you'll also need to do what Henry suggests.

[TomBolton]

This PR is almost ready for another look, but one of the Basic Checks and Runner tests is failing with a duplicate test title and I'm not sure how to fix it, as the tests I added appear to be unique.

[aeisenberg]

There is an undefined in the test title. It probably means that the test macro isn't sufficient for the new tests you have written. The test macro is the thing that allows you to generate variants of a single test with some slight changes. In this case, we are using mlPoweredQueriesMacro on line 1712. The title property is what generates the title. You need to make sure that all generated titles are distinct.

You will need to change how the title is generated to ensure this is how it works. You can run the tests locally to ensure the titles are unique before pushing up a change.

[TomBolton]

Thanks for the help @aeisenberg!

[aeisenberg]

Looks reasonable, assuming all the tests pass.

[TomBolton]

@henrymercer these two PR checks (here and here), involving a multi-language repo on mac, keep getting cancelled during the go extraction.

I restarted each check to see if it was a transient problem, but it happened again. Any ideas why this might be happening?

[henrymercer]

I've raised the flaky tests internally. It seems that another rerun has sorted things with one of the jobs, so I've kicked off another run of the other job too.

[TomBolton]

Thanks for looking into it @henrymercer and doing the reruns!