codeql-action/init should explicitly delete an already present database (for self-hosted runners)

codeql-action/init のエントリポイントから開始し、報告された database,init 呼び出しで示されているデータベースパスを --overwrite オプションがどのように処理するかを追跡します。可能であれば self-hosted runner で失敗を再現し、それを引き起こす既存ディレクトリの条件を特定します。完了条件は、init が外部のクリーンアップ手順を必要とせずに報告された workspace を処理し、説明されたシナリオで動作が検証されていることです。

We are running CodeQL as an action on self-hosted runners and have enabled the 'over-write' option to clean the CodeQL database before each run.

However, some environments fail with the following message:

Failure invoking /__w/_tool/CodeQL/2.13.3-20230524/x64/codeql/codeql with arguments database,init,--db-cluster,/__w/_temp/codeql_databases,--source-root=/__w/a1_aggregation_mtf/a1_aggregation_mtf,--language=java,--begin-tracing,--trace-process-name=Runner.Worker.exe,--codescanning-config=/__w/_temp/user-config.yaml,--overwrite.
 
       Exit code 2 and error was:

       A fatal error occurred: Ran with database overwriting enabled, but the directory does not appear to be a CodeQL database or database cluster. Please check you do indeed wish to delete it, and do so manually.

As suggested in the error message, we can manually delete the directory, allowing for a successful run.

We do not understand what is causing CodeQL to believe the directory is not a CodeQL database. We are not explicitly altering the contents or structure. As the directories involved are created by CodeQL within the workflow, the hope is to have CodeQL able to clean up it's workspace from within the workflow as well.

Ideally, we want to avoid having to implement an external cleanup step for the process.

We are curious to see what logic is being used to determine the 'CodeQL database' status of the targeted directory.

Regards,
Richard Seeton