Resource inventories

These manifests describe task environments, data preparation and source provenance. They are for contributors maintaining the build recipes; paths are repository-relative.

File or directory	Contents
images.json	Runtime, tool and data products, image identities, dependencies and acquisition recipes
files.json	Omitted source inputs, paths, sizes, hashes, modes and distribution conditions
sources.json	Source commits, original hashes and current packaging hashes
data/	Data preparation recipes, per-file manifests and the common data-image Dockerfile
build-inputs.json	Dockerfile FROM/COPY/ADD dependencies and omitted inputs
resource-groups.json	Omitted files grouped by purpose and distribution conditions
environments/lean-base/	Shared Lean build materials and dependency pins
environments/cpsolver/	CPSolver source and build materials; see instructions

Use tcsexam build to prepare a selection from these recipes. It downloads fixed inputs, builds dependencies and assembles the complete images used by Harbor. Data images supply files at build time. The resulting image identities are stored in the local cache for doctor and run. An external artifacts.json provides optional immutable image assignments. See resource preparation for the commands and input requirements.

Keep public practice data, hidden judging data and reference answers in their respective products. Development and public self-test images must exclude hidden inputs from every layer. A complete image inherits its inputs' distribution conditions.

Store generated images, datasets, jobs and private configuration outside the source package. Contributing describes source checks and export.